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 #define BPF_LOAD_ACQ 0x100 /* load-acquire */ 55 #define BPF_STORE_REL 0x110 /* store-release */ 56 57 enum bpf_cond_pseudo_jmp { 58 BPF_MAY_GOTO = 0, 59 }; 60 61 /* Register numbers */ 62 enum { 63 BPF_REG_0 = 0, 64 BPF_REG_1, 65 BPF_REG_2, 66 BPF_REG_3, 67 BPF_REG_4, 68 BPF_REG_5, 69 BPF_REG_6, 70 BPF_REG_7, 71 BPF_REG_8, 72 BPF_REG_9, 73 BPF_REG_10, 74 __MAX_BPF_REG, 75 }; 76 77 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 78 #define MAX_BPF_REG __MAX_BPF_REG 79 80 struct bpf_insn { 81 __u8 code; /* opcode */ 82 __u8 dst_reg:4; /* dest register */ 83 __u8 src_reg:4; /* source register */ 84 __s16 off; /* signed offset */ 85 __s32 imm; /* signed immediate constant */ 86 }; 87 88 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for 89 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for 90 * the trailing flexible array member) instead. 91 */ 92 struct bpf_lpm_trie_key { 93 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 94 __u8 data[0]; /* Arbitrary size */ 95 }; 96 97 /* Header for bpf_lpm_trie_key structs */ 98 struct bpf_lpm_trie_key_hdr { 99 __u32 prefixlen; 100 }; 101 102 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */ 103 struct bpf_lpm_trie_key_u8 { 104 union { 105 struct bpf_lpm_trie_key_hdr hdr; 106 __u32 prefixlen; 107 }; 108 __u8 data[]; /* Arbitrary size */ 109 }; 110 111 struct bpf_cgroup_storage_key { 112 __u64 cgroup_inode_id; /* cgroup inode id */ 113 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 114 }; 115 116 enum bpf_cgroup_iter_order { 117 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 118 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 119 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 120 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 121 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 122 /* 123 * Walks the immediate children of the specified parent 124 * cgroup_subsys_state. Unlike BPF_CGROUP_ITER_DESCENDANTS_PRE, 125 * BPF_CGROUP_ITER_DESCENDANTS_POST, and BPF_CGROUP_ITER_ANCESTORS_UP 126 * the iterator does not include the specified parent as one of the 127 * returned iterator elements. 128 */ 129 BPF_CGROUP_ITER_CHILDREN, 130 }; 131 132 union bpf_iter_link_info { 133 struct { 134 __u32 map_fd; 135 } map; 136 struct { 137 enum bpf_cgroup_iter_order order; 138 139 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 140 * both are zero, the walk starts from the default cgroup v2 141 * root. For walking v1 hierarchy, one should always explicitly 142 * specify cgroup_fd. 143 */ 144 __u32 cgroup_fd; 145 __u64 cgroup_id; 146 } cgroup; 147 /* Parameters of task iterators. */ 148 struct { 149 __u32 tid; 150 __u32 pid; 151 __u32 pid_fd; 152 } task; 153 }; 154 155 /* BPF syscall commands, see bpf(2) man-page for more details. */ 156 /** 157 * DOC: eBPF Syscall Preamble 158 * 159 * The operation to be performed by the **bpf**\ () system call is determined 160 * by the *cmd* argument. Each operation takes an accompanying argument, 161 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 162 * below). The size argument is the size of the union pointed to by *attr*. 163 */ 164 /** 165 * DOC: eBPF Syscall Commands 166 * 167 * BPF_MAP_CREATE 168 * Description 169 * Create a map and return a file descriptor that refers to the 170 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 171 * is automatically enabled for the new file descriptor. 172 * 173 * Applying **close**\ (2) to the file descriptor returned by 174 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 175 * 176 * Return 177 * A new file descriptor (a nonnegative integer), or -1 if an 178 * error occurred (in which case, *errno* is set appropriately). 179 * 180 * BPF_MAP_LOOKUP_ELEM 181 * Description 182 * Look up an element with a given *key* in the map referred to 183 * by the file descriptor *map_fd*. 184 * 185 * The *flags* argument may be specified as one of the 186 * following: 187 * 188 * **BPF_F_LOCK** 189 * Look up the value of a spin-locked map without 190 * returning the lock. This must be specified if the 191 * elements contain a spinlock. 192 * 193 * Return 194 * Returns zero on success. On error, -1 is returned and *errno* 195 * is set appropriately. 196 * 197 * BPF_MAP_UPDATE_ELEM 198 * Description 199 * Create or update an element (key/value pair) in a specified map. 200 * 201 * The *flags* argument should be specified as one of the 202 * following: 203 * 204 * **BPF_ANY** 205 * Create a new element or update an existing element. 206 * **BPF_NOEXIST** 207 * Create a new element only if it did not exist. 208 * **BPF_EXIST** 209 * Update an existing element. 210 * **BPF_F_LOCK** 211 * Update a spin_lock-ed map element. 212 * 213 * Return 214 * Returns zero on success. On error, -1 is returned and *errno* 215 * is set appropriately. 216 * 217 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 218 * **E2BIG**, **EEXIST**, or **ENOENT**. 219 * 220 * **E2BIG** 221 * The number of elements in the map reached the 222 * *max_entries* limit specified at map creation time. 223 * **EEXIST** 224 * If *flags* specifies **BPF_NOEXIST** and the element 225 * with *key* already exists in the map. 226 * **ENOENT** 227 * If *flags* specifies **BPF_EXIST** and the element with 228 * *key* does not exist in the map. 229 * 230 * BPF_MAP_DELETE_ELEM 231 * Description 232 * Look up and delete an element by key in a specified map. 233 * 234 * Return 235 * Returns zero on success. On error, -1 is returned and *errno* 236 * is set appropriately. 237 * 238 * BPF_MAP_GET_NEXT_KEY 239 * Description 240 * Look up an element by key in a specified map and return the key 241 * of the next element. Can be used to iterate over all elements 242 * in the map. 243 * 244 * Return 245 * Returns zero on success. On error, -1 is returned and *errno* 246 * is set appropriately. 247 * 248 * The following cases can be used to iterate over all elements of 249 * the map: 250 * 251 * * If *key* is not found, the operation returns zero and sets 252 * the *next_key* pointer to the key of the first element. 253 * * If *key* is found, the operation returns zero and sets the 254 * *next_key* pointer to the key of the next element. 255 * * If *key* is the last element, returns -1 and *errno* is set 256 * to **ENOENT**. 257 * 258 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 259 * **EINVAL** on error. 260 * 261 * BPF_PROG_LOAD 262 * Description 263 * Verify and load an eBPF program, returning a new file 264 * descriptor associated with the program. 265 * 266 * Applying **close**\ (2) to the file descriptor returned by 267 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 268 * 269 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 270 * automatically enabled for the new file descriptor. 271 * 272 * Return 273 * A new file descriptor (a nonnegative integer), or -1 if an 274 * error occurred (in which case, *errno* is set appropriately). 275 * 276 * BPF_OBJ_PIN 277 * Description 278 * Pin an eBPF program or map referred by the specified *bpf_fd* 279 * to the provided *pathname* on the filesystem. 280 * 281 * The *pathname* argument must not contain a dot ("."). 282 * 283 * On success, *pathname* retains a reference to the eBPF object, 284 * preventing deallocation of the object when the original 285 * *bpf_fd* is closed. This allow the eBPF object to live beyond 286 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 287 * process. 288 * 289 * Applying **unlink**\ (2) or similar calls to the *pathname* 290 * unpins the object from the filesystem, removing the reference. 291 * If no other file descriptors or filesystem nodes refer to the 292 * same object, it will be deallocated (see NOTES). 293 * 294 * The filesystem type for the parent directory of *pathname* must 295 * be **BPF_FS_MAGIC**. 296 * 297 * Return 298 * Returns zero on success. On error, -1 is returned and *errno* 299 * is set appropriately. 300 * 301 * BPF_OBJ_GET 302 * Description 303 * Open a file descriptor for the eBPF object pinned to the 304 * specified *pathname*. 305 * 306 * Return 307 * A new file descriptor (a nonnegative integer), or -1 if an 308 * error occurred (in which case, *errno* is set appropriately). 309 * 310 * BPF_PROG_ATTACH 311 * Description 312 * Attach an eBPF program to a *target_fd* at the specified 313 * *attach_type* hook. 314 * 315 * The *attach_type* specifies the eBPF attachment point to 316 * attach the program to, and must be one of *bpf_attach_type* 317 * (see below). 318 * 319 * The *attach_bpf_fd* must be a valid file descriptor for a 320 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 321 * or sock_ops type corresponding to the specified *attach_type*. 322 * 323 * The *target_fd* must be a valid file descriptor for a kernel 324 * object which depends on the attach type of *attach_bpf_fd*: 325 * 326 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 327 * **BPF_PROG_TYPE_CGROUP_SKB**, 328 * **BPF_PROG_TYPE_CGROUP_SOCK**, 329 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 330 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 331 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 332 * **BPF_PROG_TYPE_SOCK_OPS** 333 * 334 * Control Group v2 hierarchy with the eBPF controller 335 * enabled. Requires the kernel to be compiled with 336 * **CONFIG_CGROUP_BPF**. 337 * 338 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 339 * 340 * Network namespace (eg /proc/self/ns/net). 341 * 342 * **BPF_PROG_TYPE_LIRC_MODE2** 343 * 344 * LIRC device path (eg /dev/lircN). Requires the kernel 345 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 346 * 347 * **BPF_PROG_TYPE_SK_SKB**, 348 * **BPF_PROG_TYPE_SK_MSG** 349 * 350 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 351 * 352 * Return 353 * Returns zero on success. On error, -1 is returned and *errno* 354 * is set appropriately. 355 * 356 * BPF_PROG_DETACH 357 * Description 358 * Detach the eBPF program associated with the *target_fd* at the 359 * hook specified by *attach_type*. The program must have been 360 * previously attached using **BPF_PROG_ATTACH**. 361 * 362 * Return 363 * Returns zero on success. On error, -1 is returned and *errno* 364 * is set appropriately. 365 * 366 * BPF_PROG_TEST_RUN 367 * Description 368 * Run the eBPF program associated with the *prog_fd* a *repeat* 369 * number of times against a provided program context *ctx_in* and 370 * data *data_in*, and return the modified program context 371 * *ctx_out*, *data_out* (for example, packet data), result of the 372 * execution *retval*, and *duration* of the test run. 373 * 374 * The sizes of the buffers provided as input and output 375 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 376 * be provided in the corresponding variables *ctx_size_in*, 377 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 378 * of these parameters are not provided (ie set to NULL), the 379 * corresponding size field must be zero. 380 * 381 * Some program types have particular requirements: 382 * 383 * **BPF_PROG_TYPE_SK_LOOKUP** 384 * *data_in* and *data_out* must be NULL. 385 * 386 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 387 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 388 * 389 * *ctx_out*, *data_in* and *data_out* must be NULL. 390 * *repeat* must be zero. 391 * 392 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 393 * 394 * Return 395 * Returns zero on success. On error, -1 is returned and *errno* 396 * is set appropriately. 397 * 398 * **ENOSPC** 399 * Either *data_size_out* or *ctx_size_out* is too small. 400 * **ENOTSUPP** 401 * This command is not supported by the program type of 402 * the program referred to by *prog_fd*. 403 * 404 * BPF_PROG_GET_NEXT_ID 405 * Description 406 * Fetch the next eBPF program currently loaded into the kernel. 407 * 408 * Looks for the eBPF program with an id greater than *start_id* 409 * and updates *next_id* on success. If no other eBPF programs 410 * remain with ids higher than *start_id*, returns -1 and sets 411 * *errno* to **ENOENT**. 412 * 413 * Return 414 * Returns zero on success. On error, or when no id remains, -1 415 * is returned and *errno* is set appropriately. 416 * 417 * BPF_MAP_GET_NEXT_ID 418 * Description 419 * Fetch the next eBPF map currently loaded into the kernel. 420 * 421 * Looks for the eBPF map with an id greater than *start_id* 422 * and updates *next_id* on success. If no other eBPF maps 423 * remain with ids higher than *start_id*, returns -1 and sets 424 * *errno* to **ENOENT**. 425 * 426 * Return 427 * Returns zero on success. On error, or when no id remains, -1 428 * is returned and *errno* is set appropriately. 429 * 430 * BPF_PROG_GET_FD_BY_ID 431 * Description 432 * Open a file descriptor for the eBPF program corresponding to 433 * *prog_id*. 434 * 435 * Return 436 * A new file descriptor (a nonnegative integer), or -1 if an 437 * error occurred (in which case, *errno* is set appropriately). 438 * 439 * BPF_MAP_GET_FD_BY_ID 440 * Description 441 * Open a file descriptor for the eBPF map corresponding to 442 * *map_id*. 443 * 444 * Return 445 * A new file descriptor (a nonnegative integer), or -1 if an 446 * error occurred (in which case, *errno* is set appropriately). 447 * 448 * BPF_OBJ_GET_INFO_BY_FD 449 * Description 450 * Obtain information about the eBPF object corresponding to 451 * *bpf_fd*. 452 * 453 * Populates up to *info_len* bytes of *info*, which will be in 454 * one of the following formats depending on the eBPF object type 455 * of *bpf_fd*: 456 * 457 * * **struct bpf_prog_info** 458 * * **struct bpf_map_info** 459 * * **struct bpf_btf_info** 460 * * **struct bpf_link_info** 461 * * **struct bpf_token_info** 462 * 463 * Return 464 * Returns zero on success. On error, -1 is returned and *errno* 465 * is set appropriately. 466 * 467 * BPF_PROG_QUERY 468 * Description 469 * Obtain information about eBPF programs associated with the 470 * specified *attach_type* hook. 471 * 472 * The *target_fd* must be a valid file descriptor for a kernel 473 * object which depends on the attach type of *attach_bpf_fd*: 474 * 475 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 476 * **BPF_PROG_TYPE_CGROUP_SKB**, 477 * **BPF_PROG_TYPE_CGROUP_SOCK**, 478 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 479 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 480 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 481 * **BPF_PROG_TYPE_SOCK_OPS** 482 * 483 * Control Group v2 hierarchy with the eBPF controller 484 * enabled. Requires the kernel to be compiled with 485 * **CONFIG_CGROUP_BPF**. 486 * 487 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 488 * 489 * Network namespace (eg /proc/self/ns/net). 490 * 491 * **BPF_PROG_TYPE_LIRC_MODE2** 492 * 493 * LIRC device path (eg /dev/lircN). Requires the kernel 494 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 495 * 496 * **BPF_PROG_QUERY** always fetches the number of programs 497 * attached and the *attach_flags* which were used to attach those 498 * programs. Additionally, if *prog_ids* is nonzero and the number 499 * of attached programs is less than *prog_cnt*, populates 500 * *prog_ids* with the eBPF program ids of the programs attached 501 * at *target_fd*. 502 * 503 * The following flags may alter the result: 504 * 505 * **BPF_F_QUERY_EFFECTIVE** 506 * Only return information regarding programs which are 507 * currently effective at the specified *target_fd*. 508 * 509 * Return 510 * Returns zero on success. On error, -1 is returned and *errno* 511 * is set appropriately. 512 * 513 * BPF_RAW_TRACEPOINT_OPEN 514 * Description 515 * Attach an eBPF program to a tracepoint *name* to access kernel 516 * internal arguments of the tracepoint in their raw form. 517 * 518 * The *prog_fd* must be a valid file descriptor associated with 519 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 520 * 521 * No ABI guarantees are made about the content of tracepoint 522 * arguments exposed to the corresponding eBPF program. 523 * 524 * Applying **close**\ (2) to the file descriptor returned by 525 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 526 * 527 * Return 528 * A new file descriptor (a nonnegative integer), or -1 if an 529 * error occurred (in which case, *errno* is set appropriately). 530 * 531 * BPF_BTF_LOAD 532 * Description 533 * Verify and load BPF Type Format (BTF) metadata into the kernel, 534 * returning a new file descriptor associated with the metadata. 535 * BTF is described in more detail at 536 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 537 * 538 * The *btf* parameter must point to valid memory providing 539 * *btf_size* bytes of BTF binary metadata. 540 * 541 * The returned file descriptor can be passed to other **bpf**\ () 542 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 543 * associate the BTF with those objects. 544 * 545 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 546 * parameters to specify a *btf_log_buf*, *btf_log_size* and 547 * *btf_log_level* which allow the kernel to return freeform log 548 * output regarding the BTF verification process. 549 * 550 * Return 551 * A new file descriptor (a nonnegative integer), or -1 if an 552 * error occurred (in which case, *errno* is set appropriately). 553 * 554 * BPF_BTF_GET_FD_BY_ID 555 * Description 556 * Open a file descriptor for the BPF Type Format (BTF) 557 * corresponding to *btf_id*. 558 * 559 * Return 560 * A new file descriptor (a nonnegative integer), or -1 if an 561 * error occurred (in which case, *errno* is set appropriately). 562 * 563 * BPF_TASK_FD_QUERY 564 * Description 565 * Obtain information about eBPF programs associated with the 566 * target process identified by *pid* and *fd*. 567 * 568 * If the *pid* and *fd* are associated with a tracepoint, kprobe 569 * or uprobe perf event, then the *prog_id* and *fd_type* will 570 * be populated with the eBPF program id and file descriptor type 571 * of type **bpf_task_fd_type**. If associated with a kprobe or 572 * uprobe, the *probe_offset* and *probe_addr* will also be 573 * populated. Optionally, if *buf* is provided, then up to 574 * *buf_len* bytes of *buf* will be populated with the name of 575 * the tracepoint, kprobe or uprobe. 576 * 577 * The resulting *prog_id* may be introspected in deeper detail 578 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 579 * 580 * Return 581 * Returns zero on success. On error, -1 is returned and *errno* 582 * is set appropriately. 583 * 584 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 585 * Description 586 * Look up an element with the given *key* in the map referred to 587 * by the file descriptor *fd*, and if found, delete the element. 588 * 589 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 590 * types, the *flags* argument needs to be set to 0, but for other 591 * map types, it may be specified as: 592 * 593 * **BPF_F_LOCK** 594 * Look up and delete the value of a spin-locked map 595 * without returning the lock. This must be specified if 596 * the elements contain a spinlock. 597 * 598 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 599 * implement this command as a "pop" operation, deleting the top 600 * element rather than one corresponding to *key*. 601 * The *key* and *key_len* parameters should be zeroed when 602 * issuing this operation for these map types. 603 * 604 * This command is only valid for the following map types: 605 * * **BPF_MAP_TYPE_QUEUE** 606 * * **BPF_MAP_TYPE_STACK** 607 * * **BPF_MAP_TYPE_HASH** 608 * * **BPF_MAP_TYPE_PERCPU_HASH** 609 * * **BPF_MAP_TYPE_LRU_HASH** 610 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 611 * 612 * Return 613 * Returns zero on success. On error, -1 is returned and *errno* 614 * is set appropriately. 615 * 616 * BPF_MAP_FREEZE 617 * Description 618 * Freeze the permissions of the specified map. 619 * 620 * Write permissions may be frozen by passing zero *flags*. 621 * Upon success, no future syscall invocations may alter the 622 * map state of *map_fd*. Write operations from eBPF programs 623 * are still possible for a frozen map. 624 * 625 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 626 * 627 * Return 628 * Returns zero on success. On error, -1 is returned and *errno* 629 * is set appropriately. 630 * 631 * BPF_BTF_GET_NEXT_ID 632 * Description 633 * Fetch the next BPF Type Format (BTF) object currently loaded 634 * into the kernel. 635 * 636 * Looks for the BTF object with an id greater than *start_id* 637 * and updates *next_id* on success. If no other BTF objects 638 * remain with ids higher than *start_id*, returns -1 and sets 639 * *errno* to **ENOENT**. 640 * 641 * Return 642 * Returns zero on success. On error, or when no id remains, -1 643 * is returned and *errno* is set appropriately. 644 * 645 * BPF_MAP_LOOKUP_BATCH 646 * Description 647 * Iterate and fetch multiple elements in a map. 648 * 649 * Two opaque values are used to manage batch operations, 650 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 651 * to NULL to begin the batched operation. After each subsequent 652 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 653 * *out_batch* as the *in_batch* for the next operation to 654 * continue iteration from the current point. Both *in_batch* and 655 * *out_batch* must point to memory large enough to hold a key, 656 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH, 657 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters 658 * must be at least 4 bytes wide regardless of key size. 659 * 660 * The *keys* and *values* are output parameters which must point 661 * to memory large enough to hold *count* items based on the key 662 * and value size of the map *map_fd*. The *keys* buffer must be 663 * of *key_size* * *count*. The *values* buffer must be of 664 * *value_size* * *count*. 665 * 666 * The *elem_flags* argument may be specified as one of the 667 * following: 668 * 669 * **BPF_F_LOCK** 670 * Look up the value of a spin-locked map without 671 * returning the lock. This must be specified if the 672 * elements contain a spinlock. 673 * 674 * On success, *count* elements from the map are copied into the 675 * user buffer, with the keys copied into *keys* and the values 676 * copied into the corresponding indices in *values*. 677 * 678 * If an error is returned and *errno* is not **EFAULT**, *count* 679 * is set to the number of successfully processed elements. 680 * 681 * Return 682 * Returns zero on success. On error, -1 is returned and *errno* 683 * is set appropriately. 684 * 685 * May set *errno* to **ENOSPC** to indicate that *keys* or 686 * *values* is too small to dump an entire bucket during 687 * iteration of a hash-based map type. 688 * 689 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 690 * Description 691 * Iterate and delete all elements in a map. 692 * 693 * This operation has the same behavior as 694 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 695 * 696 * * Every element that is successfully returned is also deleted 697 * from the map. This is at least *count* elements. Note that 698 * *count* is both an input and an output parameter. 699 * * Upon returning with *errno* set to **EFAULT**, up to 700 * *count* elements may be deleted without returning the keys 701 * and values of the deleted elements. 702 * 703 * Return 704 * Returns zero on success. On error, -1 is returned and *errno* 705 * is set appropriately. 706 * 707 * BPF_MAP_UPDATE_BATCH 708 * Description 709 * Update multiple elements in a map by *key*. 710 * 711 * The *keys* and *values* are input parameters which must point 712 * to memory large enough to hold *count* items based on the key 713 * and value size of the map *map_fd*. The *keys* buffer must be 714 * of *key_size* * *count*. The *values* buffer must be of 715 * *value_size* * *count*. 716 * 717 * Each element specified in *keys* is sequentially updated to the 718 * value in the corresponding index in *values*. The *in_batch* 719 * and *out_batch* parameters are ignored and should be zeroed. 720 * 721 * The *elem_flags* argument should be specified as one of the 722 * following: 723 * 724 * **BPF_ANY** 725 * Create new elements or update a existing elements. 726 * **BPF_NOEXIST** 727 * Create new elements only if they do not exist. 728 * **BPF_EXIST** 729 * Update existing elements. 730 * **BPF_F_LOCK** 731 * Update spin_lock-ed map elements. This must be 732 * specified if the map value contains a spinlock. 733 * 734 * On success, *count* elements from the map are updated. 735 * 736 * If an error is returned and *errno* is not **EFAULT**, *count* 737 * is set to the number of successfully processed elements. 738 * 739 * Return 740 * Returns zero on success. On error, -1 is returned and *errno* 741 * is set appropriately. 742 * 743 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 744 * **E2BIG**. **E2BIG** indicates that the number of elements in 745 * the map reached the *max_entries* limit specified at map 746 * creation time. 747 * 748 * May set *errno* to one of the following error codes under 749 * specific circumstances: 750 * 751 * **EEXIST** 752 * If *flags* specifies **BPF_NOEXIST** and the element 753 * with *key* already exists in the map. 754 * **ENOENT** 755 * If *flags* specifies **BPF_EXIST** and the element with 756 * *key* does not exist in the map. 757 * 758 * BPF_MAP_DELETE_BATCH 759 * Description 760 * Delete multiple elements in a map by *key*. 761 * 762 * The *keys* parameter is an input parameter which must point 763 * to memory large enough to hold *count* items based on the key 764 * size of the map *map_fd*, that is, *key_size* * *count*. 765 * 766 * Each element specified in *keys* is sequentially deleted. The 767 * *in_batch*, *out_batch*, and *values* parameters are ignored 768 * and should be zeroed. 769 * 770 * The *elem_flags* argument may be specified as one of the 771 * following: 772 * 773 * **BPF_F_LOCK** 774 * Look up the value of a spin-locked map without 775 * returning the lock. This must be specified if the 776 * elements contain a spinlock. 777 * 778 * On success, *count* elements from the map are updated. 779 * 780 * If an error is returned and *errno* is not **EFAULT**, *count* 781 * is set to the number of successfully processed elements. If 782 * *errno* is **EFAULT**, up to *count* elements may be been 783 * deleted. 784 * 785 * Return 786 * Returns zero on success. On error, -1 is returned and *errno* 787 * is set appropriately. 788 * 789 * BPF_LINK_CREATE 790 * Description 791 * Attach an eBPF program to a *target_fd* at the specified 792 * *attach_type* hook and return a file descriptor handle for 793 * managing the link. 794 * 795 * Return 796 * A new file descriptor (a nonnegative integer), or -1 if an 797 * error occurred (in which case, *errno* is set appropriately). 798 * 799 * BPF_LINK_UPDATE 800 * Description 801 * Update the eBPF program in the specified *link_fd* to 802 * *new_prog_fd*. 803 * 804 * Return 805 * Returns zero on success. On error, -1 is returned and *errno* 806 * is set appropriately. 807 * 808 * BPF_LINK_GET_FD_BY_ID 809 * Description 810 * Open a file descriptor for the eBPF Link corresponding to 811 * *link_id*. 812 * 813 * Return 814 * A new file descriptor (a nonnegative integer), or -1 if an 815 * error occurred (in which case, *errno* is set appropriately). 816 * 817 * BPF_LINK_GET_NEXT_ID 818 * Description 819 * Fetch the next eBPF link currently loaded into the kernel. 820 * 821 * Looks for the eBPF link with an id greater than *start_id* 822 * and updates *next_id* on success. If no other eBPF links 823 * remain with ids higher than *start_id*, returns -1 and sets 824 * *errno* to **ENOENT**. 825 * 826 * Return 827 * Returns zero on success. On error, or when no id remains, -1 828 * is returned and *errno* is set appropriately. 829 * 830 * BPF_ENABLE_STATS 831 * Description 832 * Enable eBPF runtime statistics gathering. 833 * 834 * Runtime statistics gathering for the eBPF runtime is disabled 835 * by default to minimize the corresponding performance overhead. 836 * This command enables statistics globally. 837 * 838 * Multiple programs may independently enable statistics. 839 * After gathering the desired statistics, eBPF runtime statistics 840 * may be disabled again by calling **close**\ (2) for the file 841 * descriptor returned by this function. Statistics will only be 842 * disabled system-wide when all outstanding file descriptors 843 * returned by prior calls for this subcommand are closed. 844 * 845 * Return 846 * A new file descriptor (a nonnegative integer), or -1 if an 847 * error occurred (in which case, *errno* is set appropriately). 848 * 849 * BPF_ITER_CREATE 850 * Description 851 * Create an iterator on top of the specified *link_fd* (as 852 * previously created using **BPF_LINK_CREATE**) and return a 853 * file descriptor that can be used to trigger the iteration. 854 * 855 * If the resulting file descriptor is pinned to the filesystem 856 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 857 * for that path will trigger the iterator to read kernel state 858 * using the eBPF program attached to *link_fd*. 859 * 860 * Return 861 * A new file descriptor (a nonnegative integer), or -1 if an 862 * error occurred (in which case, *errno* is set appropriately). 863 * 864 * BPF_LINK_DETACH 865 * Description 866 * Forcefully detach the specified *link_fd* from its 867 * corresponding attachment point. 868 * 869 * Return 870 * Returns zero on success. On error, -1 is returned and *errno* 871 * is set appropriately. 872 * 873 * BPF_PROG_BIND_MAP 874 * Description 875 * Bind a map to the lifetime of an eBPF program. 876 * 877 * The map identified by *map_fd* is bound to the program 878 * identified by *prog_fd* and only released when *prog_fd* is 879 * released. This may be used in cases where metadata should be 880 * associated with a program which otherwise does not contain any 881 * references to the map (for example, embedded in the eBPF 882 * program instructions). 883 * 884 * Return 885 * Returns zero on success. On error, -1 is returned and *errno* 886 * is set appropriately. 887 * 888 * BPF_TOKEN_CREATE 889 * Description 890 * Create BPF token with embedded information about what 891 * BPF-related functionality it allows: 892 * - a set of allowed bpf() syscall commands; 893 * - a set of allowed BPF map types to be created with 894 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 895 * - a set of allowed BPF program types and BPF program attach 896 * types to be loaded with BPF_PROG_LOAD command, if 897 * BPF_PROG_LOAD itself is allowed. 898 * 899 * BPF token is created (derived) from an instance of BPF FS, 900 * assuming it has necessary delegation mount options specified. 901 * This BPF token can be passed as an extra parameter to various 902 * bpf() syscall commands to grant BPF subsystem functionality to 903 * unprivileged processes. 904 * 905 * When created, BPF token is "associated" with the owning 906 * user namespace of BPF FS instance (super block) that it was 907 * derived from, and subsequent BPF operations performed with 908 * BPF token would be performing capabilities checks (i.e., 909 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 910 * that user namespace. Without BPF token, such capabilities 911 * have to be granted in init user namespace, making bpf() 912 * syscall incompatible with user namespace, for the most part. 913 * 914 * Return 915 * A new file descriptor (a nonnegative integer), or -1 if an 916 * error occurred (in which case, *errno* is set appropriately). 917 * 918 * BPF_PROG_STREAM_READ_BY_FD 919 * Description 920 * Read data of a program's BPF stream. The program is identified 921 * by *prog_fd*, and the stream is identified by the *stream_id*. 922 * The data is copied to a buffer pointed to by *stream_buf*, and 923 * filled less than or equal to *stream_buf_len* bytes. 924 * 925 * Return 926 * Number of bytes read from the stream on success, or -1 if an 927 * error occurred (in which case, *errno* is set appropriately). 928 * 929 * BPF_PROG_ASSOC_STRUCT_OPS 930 * Description 931 * Associate a BPF program with a struct_ops map. The struct_ops 932 * map is identified by *map_fd* and the BPF program is 933 * identified by *prog_fd*. 934 * 935 * Return 936 * 0 on success or -1 if an error occurred (in which case, 937 * *errno* is set appropriately). 938 * 939 * NOTES 940 * eBPF objects (maps and programs) can be shared between processes. 941 * 942 * * After **fork**\ (2), the child inherits file descriptors 943 * referring to the same eBPF objects. 944 * * File descriptors referring to eBPF objects can be transferred over 945 * **unix**\ (7) domain sockets. 946 * * File descriptors referring to eBPF objects can be duplicated in the 947 * usual way, using **dup**\ (2) and similar calls. 948 * * File descriptors referring to eBPF objects can be pinned to the 949 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 950 * 951 * An eBPF object is deallocated only after all file descriptors referring 952 * to the object have been closed and no references remain pinned to the 953 * filesystem or attached (for example, bound to a program or device). 954 */ 955 enum bpf_cmd { 956 BPF_MAP_CREATE, 957 BPF_MAP_LOOKUP_ELEM, 958 BPF_MAP_UPDATE_ELEM, 959 BPF_MAP_DELETE_ELEM, 960 BPF_MAP_GET_NEXT_KEY, 961 BPF_PROG_LOAD, 962 BPF_OBJ_PIN, 963 BPF_OBJ_GET, 964 BPF_PROG_ATTACH, 965 BPF_PROG_DETACH, 966 BPF_PROG_TEST_RUN, 967 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 968 BPF_PROG_GET_NEXT_ID, 969 BPF_MAP_GET_NEXT_ID, 970 BPF_PROG_GET_FD_BY_ID, 971 BPF_MAP_GET_FD_BY_ID, 972 BPF_OBJ_GET_INFO_BY_FD, 973 BPF_PROG_QUERY, 974 BPF_RAW_TRACEPOINT_OPEN, 975 BPF_BTF_LOAD, 976 BPF_BTF_GET_FD_BY_ID, 977 BPF_TASK_FD_QUERY, 978 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 979 BPF_MAP_FREEZE, 980 BPF_BTF_GET_NEXT_ID, 981 BPF_MAP_LOOKUP_BATCH, 982 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 983 BPF_MAP_UPDATE_BATCH, 984 BPF_MAP_DELETE_BATCH, 985 BPF_LINK_CREATE, 986 BPF_LINK_UPDATE, 987 BPF_LINK_GET_FD_BY_ID, 988 BPF_LINK_GET_NEXT_ID, 989 BPF_ENABLE_STATS, 990 BPF_ITER_CREATE, 991 BPF_LINK_DETACH, 992 BPF_PROG_BIND_MAP, 993 BPF_TOKEN_CREATE, 994 BPF_PROG_STREAM_READ_BY_FD, 995 BPF_PROG_ASSOC_STRUCT_OPS, 996 __MAX_BPF_CMD, 997 BPF_COMMON_ATTRS = 1 << 16, /* Indicate carrying syscall common attrs. */ 998 }; 999 1000 enum bpf_map_type { 1001 BPF_MAP_TYPE_UNSPEC, 1002 BPF_MAP_TYPE_HASH, 1003 BPF_MAP_TYPE_ARRAY, 1004 BPF_MAP_TYPE_PROG_ARRAY, 1005 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 1006 BPF_MAP_TYPE_PERCPU_HASH, 1007 BPF_MAP_TYPE_PERCPU_ARRAY, 1008 BPF_MAP_TYPE_STACK_TRACE, 1009 BPF_MAP_TYPE_CGROUP_ARRAY, 1010 BPF_MAP_TYPE_LRU_HASH, 1011 BPF_MAP_TYPE_LRU_PERCPU_HASH, 1012 BPF_MAP_TYPE_LPM_TRIE, 1013 BPF_MAP_TYPE_ARRAY_OF_MAPS, 1014 BPF_MAP_TYPE_HASH_OF_MAPS, 1015 BPF_MAP_TYPE_DEVMAP, 1016 BPF_MAP_TYPE_SOCKMAP, 1017 BPF_MAP_TYPE_CPUMAP, 1018 BPF_MAP_TYPE_XSKMAP, 1019 BPF_MAP_TYPE_SOCKHASH, 1020 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 1021 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 1022 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 1023 * both cgroup-attached and other progs and supports all functionality 1024 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 1025 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 1026 */ 1027 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 1028 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 1029 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1030 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 1031 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 1032 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 1033 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 1034 * deprecated. 1035 */ 1036 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 1037 BPF_MAP_TYPE_QUEUE, 1038 BPF_MAP_TYPE_STACK, 1039 BPF_MAP_TYPE_SK_STORAGE, 1040 BPF_MAP_TYPE_DEVMAP_HASH, 1041 BPF_MAP_TYPE_STRUCT_OPS, 1042 BPF_MAP_TYPE_RINGBUF, 1043 BPF_MAP_TYPE_INODE_STORAGE, 1044 BPF_MAP_TYPE_TASK_STORAGE, 1045 BPF_MAP_TYPE_BLOOM_FILTER, 1046 BPF_MAP_TYPE_USER_RINGBUF, 1047 BPF_MAP_TYPE_CGRP_STORAGE, 1048 BPF_MAP_TYPE_ARENA, 1049 BPF_MAP_TYPE_INSN_ARRAY, 1050 BPF_MAP_TYPE_RHASH, 1051 __MAX_BPF_MAP_TYPE 1052 }; 1053 1054 /* Note that tracing related programs such as 1055 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1056 * are not subject to a stable API since kernel internal data 1057 * structures can change from release to release and may 1058 * therefore break existing tracing BPF programs. Tracing BPF 1059 * programs correspond to /a/ specific kernel which is to be 1060 * analyzed, and not /a/ specific kernel /and/ all future ones. 1061 */ 1062 enum bpf_prog_type { 1063 BPF_PROG_TYPE_UNSPEC, 1064 BPF_PROG_TYPE_SOCKET_FILTER, 1065 BPF_PROG_TYPE_KPROBE, 1066 BPF_PROG_TYPE_SCHED_CLS, 1067 BPF_PROG_TYPE_SCHED_ACT, 1068 BPF_PROG_TYPE_TRACEPOINT, 1069 BPF_PROG_TYPE_XDP, 1070 BPF_PROG_TYPE_PERF_EVENT, 1071 BPF_PROG_TYPE_CGROUP_SKB, 1072 BPF_PROG_TYPE_CGROUP_SOCK, 1073 BPF_PROG_TYPE_LWT_IN, 1074 BPF_PROG_TYPE_LWT_OUT, 1075 BPF_PROG_TYPE_LWT_XMIT, 1076 BPF_PROG_TYPE_SOCK_OPS, 1077 BPF_PROG_TYPE_SK_SKB, 1078 BPF_PROG_TYPE_CGROUP_DEVICE, 1079 BPF_PROG_TYPE_SK_MSG, 1080 BPF_PROG_TYPE_RAW_TRACEPOINT, 1081 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1082 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1083 BPF_PROG_TYPE_LIRC_MODE2, 1084 BPF_PROG_TYPE_SK_REUSEPORT, 1085 BPF_PROG_TYPE_FLOW_DISSECTOR, 1086 BPF_PROG_TYPE_CGROUP_SYSCTL, 1087 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1088 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1089 BPF_PROG_TYPE_TRACING, 1090 BPF_PROG_TYPE_STRUCT_OPS, 1091 BPF_PROG_TYPE_EXT, 1092 BPF_PROG_TYPE_LSM, 1093 BPF_PROG_TYPE_SK_LOOKUP, 1094 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1095 BPF_PROG_TYPE_NETFILTER, 1096 __MAX_BPF_PROG_TYPE 1097 }; 1098 1099 enum bpf_attach_type { 1100 BPF_CGROUP_INET_INGRESS, 1101 BPF_CGROUP_INET_EGRESS, 1102 BPF_CGROUP_INET_SOCK_CREATE, 1103 BPF_CGROUP_SOCK_OPS, 1104 BPF_SK_SKB_STREAM_PARSER, 1105 BPF_SK_SKB_STREAM_VERDICT, 1106 BPF_CGROUP_DEVICE, 1107 BPF_SK_MSG_VERDICT, 1108 BPF_CGROUP_INET4_BIND, 1109 BPF_CGROUP_INET6_BIND, 1110 BPF_CGROUP_INET4_CONNECT, 1111 BPF_CGROUP_INET6_CONNECT, 1112 BPF_CGROUP_INET4_POST_BIND, 1113 BPF_CGROUP_INET6_POST_BIND, 1114 BPF_CGROUP_UDP4_SENDMSG, 1115 BPF_CGROUP_UDP6_SENDMSG, 1116 BPF_LIRC_MODE2, 1117 BPF_FLOW_DISSECTOR, 1118 BPF_CGROUP_SYSCTL, 1119 BPF_CGROUP_UDP4_RECVMSG, 1120 BPF_CGROUP_UDP6_RECVMSG, 1121 BPF_CGROUP_GETSOCKOPT, 1122 BPF_CGROUP_SETSOCKOPT, 1123 BPF_TRACE_RAW_TP, 1124 BPF_TRACE_FENTRY, 1125 BPF_TRACE_FEXIT, 1126 BPF_MODIFY_RETURN, 1127 BPF_LSM_MAC, 1128 BPF_TRACE_ITER, 1129 BPF_CGROUP_INET4_GETPEERNAME, 1130 BPF_CGROUP_INET6_GETPEERNAME, 1131 BPF_CGROUP_INET4_GETSOCKNAME, 1132 BPF_CGROUP_INET6_GETSOCKNAME, 1133 BPF_XDP_DEVMAP, 1134 BPF_CGROUP_INET_SOCK_RELEASE, 1135 BPF_XDP_CPUMAP, 1136 BPF_SK_LOOKUP, 1137 BPF_XDP, 1138 BPF_SK_SKB_VERDICT, 1139 BPF_SK_REUSEPORT_SELECT, 1140 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1141 BPF_PERF_EVENT, 1142 BPF_TRACE_KPROBE_MULTI, 1143 BPF_LSM_CGROUP, 1144 BPF_STRUCT_OPS, 1145 BPF_NETFILTER, 1146 BPF_TCX_INGRESS, 1147 BPF_TCX_EGRESS, 1148 BPF_TRACE_UPROBE_MULTI, 1149 BPF_CGROUP_UNIX_CONNECT, 1150 BPF_CGROUP_UNIX_SENDMSG, 1151 BPF_CGROUP_UNIX_RECVMSG, 1152 BPF_CGROUP_UNIX_GETPEERNAME, 1153 BPF_CGROUP_UNIX_GETSOCKNAME, 1154 BPF_NETKIT_PRIMARY, 1155 BPF_NETKIT_PEER, 1156 BPF_TRACE_KPROBE_SESSION, 1157 BPF_TRACE_UPROBE_SESSION, 1158 BPF_TRACE_FSESSION, 1159 BPF_TRACE_FENTRY_MULTI, 1160 BPF_TRACE_FEXIT_MULTI, 1161 BPF_TRACE_FSESSION_MULTI, 1162 __MAX_BPF_ATTACH_TYPE 1163 }; 1164 1165 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1166 1167 /* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[] 1168 * in sync with the definitions below. 1169 */ 1170 enum bpf_link_type { 1171 BPF_LINK_TYPE_UNSPEC = 0, 1172 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1173 BPF_LINK_TYPE_TRACING = 2, 1174 BPF_LINK_TYPE_CGROUP = 3, 1175 BPF_LINK_TYPE_ITER = 4, 1176 BPF_LINK_TYPE_NETNS = 5, 1177 BPF_LINK_TYPE_XDP = 6, 1178 BPF_LINK_TYPE_PERF_EVENT = 7, 1179 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1180 BPF_LINK_TYPE_STRUCT_OPS = 9, 1181 BPF_LINK_TYPE_NETFILTER = 10, 1182 BPF_LINK_TYPE_TCX = 11, 1183 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1184 BPF_LINK_TYPE_NETKIT = 13, 1185 BPF_LINK_TYPE_SOCKMAP = 14, 1186 BPF_LINK_TYPE_TRACING_MULTI = 15, 1187 __MAX_BPF_LINK_TYPE, 1188 }; 1189 1190 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1191 1192 enum bpf_perf_event_type { 1193 BPF_PERF_EVENT_UNSPEC = 0, 1194 BPF_PERF_EVENT_UPROBE = 1, 1195 BPF_PERF_EVENT_URETPROBE = 2, 1196 BPF_PERF_EVENT_KPROBE = 3, 1197 BPF_PERF_EVENT_KRETPROBE = 4, 1198 BPF_PERF_EVENT_TRACEPOINT = 5, 1199 BPF_PERF_EVENT_EVENT = 6, 1200 }; 1201 1202 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1203 * 1204 * NONE(default): No further bpf programs allowed in the subtree. 1205 * 1206 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1207 * the program in this cgroup yields to sub-cgroup program. 1208 * 1209 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1210 * that cgroup program gets run in addition to the program in this cgroup. 1211 * 1212 * Only one program is allowed to be attached to a cgroup with 1213 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1214 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1215 * release old program and attach the new one. Attach flags has to match. 1216 * 1217 * Multiple programs are allowed to be attached to a cgroup with 1218 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1219 * (those that were attached first, run first) 1220 * The programs of sub-cgroup are executed first, then programs of 1221 * this cgroup and then programs of parent cgroup. 1222 * When children program makes decision (like picking TCP CA or sock bind) 1223 * parent program has a chance to override it. 1224 * 1225 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1226 * programs for a cgroup. Though it's possible to replace an old program at 1227 * any position by also specifying BPF_F_REPLACE flag and position itself in 1228 * replace_bpf_fd attribute. Old program at this position will be released. 1229 * 1230 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1231 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1232 * Ex1: 1233 * cgrp1 (MULTI progs A, B) -> 1234 * cgrp2 (OVERRIDE prog C) -> 1235 * cgrp3 (MULTI prog D) -> 1236 * cgrp4 (OVERRIDE prog E) -> 1237 * cgrp5 (NONE prog F) 1238 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1239 * if prog F is detached, the execution is E,D,A,B 1240 * if prog F and D are detached, the execution is E,A,B 1241 * if prog F, E and D are detached, the execution is C,A,B 1242 * 1243 * All eligible programs are executed regardless of return code from 1244 * earlier programs. 1245 */ 1246 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1247 #define BPF_F_ALLOW_MULTI (1U << 1) 1248 /* Generic attachment flags. */ 1249 #define BPF_F_REPLACE (1U << 2) 1250 #define BPF_F_BEFORE (1U << 3) 1251 #define BPF_F_AFTER (1U << 4) 1252 #define BPF_F_ID (1U << 5) 1253 #define BPF_F_PREORDER (1U << 6) 1254 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1255 1256 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1257 * verifier will perform strict alignment checking as if the kernel 1258 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1259 * and NET_IP_ALIGN defined to 2. 1260 */ 1261 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1262 1263 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1264 * verifier will allow any alignment whatsoever. On platforms 1265 * with strict alignment requirements for loads ands stores (such 1266 * as sparc and mips) the verifier validates that all loads and 1267 * stores provably follow this requirement. This flag turns that 1268 * checking and enforcement off. 1269 * 1270 * It is mostly used for testing when we want to validate the 1271 * context and memory access aspects of the verifier, but because 1272 * of an unaligned access the alignment check would trigger before 1273 * the one we are interested in. 1274 */ 1275 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1276 1277 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1278 * Verifier does sub-register def/use analysis and identifies instructions whose 1279 * def only matters for low 32-bit, high 32-bit is never referenced later 1280 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1281 * that it is safe to ignore clearing high 32-bit for these instructions. This 1282 * saves some back-ends a lot of code-gen. However such optimization is not 1283 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1284 * hence hasn't used verifier's analysis result. But, we really want to have a 1285 * way to be able to verify the correctness of the described optimization on 1286 * x86_64 on which testsuites are frequently exercised. 1287 * 1288 * So, this flag is introduced. Once it is set, verifier will randomize high 1289 * 32-bit for those instructions who has been identified as safe to ignore them. 1290 * Then, if verifier is not doing correct analysis, such randomization will 1291 * regress tests to expose bugs. 1292 */ 1293 #define BPF_F_TEST_RND_HI32 (1U << 2) 1294 1295 /* The verifier internal test flag. Behavior is undefined */ 1296 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1297 1298 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1299 * restrict map and helper usage for such programs. Sleepable BPF programs can 1300 * only be attached to hooks where kernel execution context allows sleeping. 1301 * Such programs are allowed to use helpers that may sleep like 1302 * bpf_copy_from_user(). 1303 */ 1304 #define BPF_F_SLEEPABLE (1U << 4) 1305 1306 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1307 * fully support xdp frags. 1308 */ 1309 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1310 1311 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1312 * program becomes device-bound but can access XDP metadata. 1313 */ 1314 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1315 1316 /* The verifier internal test flag. Behavior is undefined */ 1317 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1318 1319 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1320 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1321 */ 1322 enum { 1323 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1324 }; 1325 1326 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1327 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1328 */ 1329 enum { 1330 /* Get return uprobe. */ 1331 BPF_F_UPROBE_MULTI_RETURN = (1U << 0), 1332 1333 /* Get path from provided path_fd. */ 1334 BPF_F_UPROBE_MULTI_PATH_FD = (1U << 1), 1335 }; 1336 1337 /* link_create.netfilter.flags used in LINK_CREATE command for 1338 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1339 */ 1340 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1341 1342 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1343 * the following extensions: 1344 * 1345 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1346 * insn[0].imm: map fd or fd_idx 1347 * insn[1].imm: 0 1348 * insn[0].off: 0 1349 * insn[1].off: 0 1350 * ldimm64 rewrite: address of map 1351 * verifier type: CONST_PTR_TO_MAP 1352 */ 1353 #define BPF_PSEUDO_MAP_FD 1 1354 #define BPF_PSEUDO_MAP_IDX 5 1355 1356 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1357 * insn[0].imm: map fd or fd_idx 1358 * insn[1].imm: offset into value 1359 * insn[0].off: 0 1360 * insn[1].off: 0 1361 * ldimm64 rewrite: address of map[0]+offset 1362 * verifier type: PTR_TO_MAP_VALUE 1363 */ 1364 #define BPF_PSEUDO_MAP_VALUE 2 1365 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1366 1367 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1368 * insn[0].imm: kernel btd id of VAR 1369 * insn[1].imm: 0 1370 * insn[0].off: 0 1371 * insn[1].off: 0 1372 * ldimm64 rewrite: address of the kernel variable 1373 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1374 * is struct/union. 1375 */ 1376 #define BPF_PSEUDO_BTF_ID 3 1377 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1378 * insn[0].imm: insn offset to the func 1379 * insn[1].imm: 0 1380 * insn[0].off: 0 1381 * insn[1].off: 0 1382 * ldimm64 rewrite: address of the function 1383 * verifier type: PTR_TO_FUNC. 1384 */ 1385 #define BPF_PSEUDO_FUNC 4 1386 1387 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1388 * offset to another bpf function 1389 */ 1390 #define BPF_PSEUDO_CALL 1 1391 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1392 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1393 */ 1394 #define BPF_PSEUDO_KFUNC_CALL 2 1395 1396 enum bpf_addr_space_cast { 1397 BPF_ADDR_SPACE_CAST = 1, 1398 }; 1399 1400 /* flags for BPF_MAP_UPDATE_ELEM command */ 1401 enum { 1402 BPF_ANY = 0, /* create new element or update existing */ 1403 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1404 BPF_EXIST = 2, /* update existing element */ 1405 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1406 BPF_F_CPU = 8, /* cpu flag for percpu maps, upper 32-bit of flags is a cpu number */ 1407 BPF_F_ALL_CPUS = 16, /* update value across all CPUs for percpu maps */ 1408 }; 1409 1410 /* flags for BPF_MAP_CREATE command */ 1411 enum { 1412 BPF_F_NO_PREALLOC = (1U << 0), 1413 /* Instead of having one common LRU list in the 1414 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1415 * which can scale and perform better. 1416 * Note, the LRU nodes (including free nodes) cannot be moved 1417 * across different LRU lists. 1418 */ 1419 BPF_F_NO_COMMON_LRU = (1U << 1), 1420 /* Specify numa node during map creation */ 1421 BPF_F_NUMA_NODE = (1U << 2), 1422 1423 /* Flags for accessing BPF object from syscall side. */ 1424 BPF_F_RDONLY = (1U << 3), 1425 BPF_F_WRONLY = (1U << 4), 1426 1427 /* Flag for stack_map, store build_id+offset instead of pointer */ 1428 BPF_F_STACK_BUILD_ID = (1U << 5), 1429 1430 /* Zero-initialize hash function seed. This should only be used for testing. */ 1431 BPF_F_ZERO_SEED = (1U << 6), 1432 1433 /* Flags for accessing BPF object from program side. */ 1434 BPF_F_RDONLY_PROG = (1U << 7), 1435 BPF_F_WRONLY_PROG = (1U << 8), 1436 1437 /* Clone map from listener for newly accepted socket */ 1438 BPF_F_CLONE = (1U << 9), 1439 1440 /* Enable memory-mapping BPF map */ 1441 BPF_F_MMAPABLE = (1U << 10), 1442 1443 /* Share perf_event among processes */ 1444 BPF_F_PRESERVE_ELEMS = (1U << 11), 1445 1446 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1447 BPF_F_INNER_MAP = (1U << 12), 1448 1449 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1450 BPF_F_LINK = (1U << 13), 1451 1452 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1453 BPF_F_PATH_FD = (1U << 14), 1454 1455 /* Flag for value_type_btf_obj_fd, the fd is available */ 1456 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1457 1458 /* BPF token FD is passed in a corresponding command's token_fd field */ 1459 BPF_F_TOKEN_FD = (1U << 16), 1460 1461 /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */ 1462 BPF_F_SEGV_ON_FAULT = (1U << 17), 1463 1464 /* Do not translate kernel bpf_arena pointers to user pointers */ 1465 BPF_F_NO_USER_CONV = (1U << 18), 1466 1467 /* Enable BPF ringbuf overwrite mode */ 1468 BPF_F_RB_OVERWRITE = (1U << 19), 1469 }; 1470 1471 /* Flags for BPF_PROG_QUERY. */ 1472 1473 /* Query effective (directly attached + inherited from ancestor cgroups) 1474 * programs that will be executed for events within a cgroup. 1475 * attach_flags with this flag are always returned 0. 1476 */ 1477 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1478 1479 /* Flags for BPF_PROG_TEST_RUN */ 1480 1481 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1482 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1483 /* If set, XDP frames will be transmitted after processing */ 1484 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1485 /* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */ 1486 #define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2) 1487 1488 /* type for BPF_ENABLE_STATS */ 1489 enum bpf_stats_type { 1490 /* enabled run_time_ns and run_cnt */ 1491 BPF_STATS_RUN_TIME = 0, 1492 }; 1493 1494 enum bpf_stack_build_id_status { 1495 /* user space need an empty entry to identify end of a trace */ 1496 BPF_STACK_BUILD_ID_EMPTY = 0, 1497 /* with valid build_id and offset */ 1498 BPF_STACK_BUILD_ID_VALID = 1, 1499 /* couldn't get build_id, fallback to ip */ 1500 BPF_STACK_BUILD_ID_IP = 2, 1501 }; 1502 1503 #define BPF_BUILD_ID_SIZE 20 1504 struct bpf_stack_build_id { 1505 __s32 status; 1506 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1507 union { 1508 __u64 offset; 1509 __u64 ip; 1510 }; 1511 }; 1512 1513 struct bpf_common_attr { 1514 __aligned_u64 log_buf; 1515 __u32 log_size; 1516 __u32 log_level; 1517 __u32 log_true_size; 1518 }; 1519 1520 #define BPF_OBJ_NAME_LEN 16U 1521 1522 enum { 1523 BPF_STREAM_STDOUT = 1, 1524 BPF_STREAM_STDERR = 2, 1525 }; 1526 1527 union bpf_attr { 1528 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1529 __u32 map_type; /* one of enum bpf_map_type */ 1530 __u32 key_size; /* size of key in bytes */ 1531 __u32 value_size; /* size of value in bytes */ 1532 __u32 max_entries; /* max number of entries in a map */ 1533 __u32 map_flags; /* BPF_MAP_CREATE related 1534 * flags defined above. 1535 */ 1536 __u32 inner_map_fd; /* fd pointing to the inner map */ 1537 __u32 numa_node; /* numa node (effective only if 1538 * BPF_F_NUMA_NODE is set). 1539 */ 1540 char map_name[BPF_OBJ_NAME_LEN]; 1541 __u32 map_ifindex; /* ifindex of netdev to create on */ 1542 __u32 btf_fd; /* fd pointing to a BTF type data */ 1543 __u32 btf_key_type_id; /* BTF type_id of the key */ 1544 __u32 btf_value_type_id; /* BTF type_id of the value */ 1545 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1546 * struct stored as the 1547 * map value 1548 */ 1549 /* Any per-map-type extra fields 1550 * 1551 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1552 * number of hash functions (if 0, the bloom filter will default 1553 * to using 5 hash functions). 1554 * 1555 * BPF_MAP_TYPE_ARENA - contains the address where user space 1556 * is going to mmap() the arena. It has to be page aligned. 1557 * 1558 * BPF_MAP_TYPE_RHASH - initial table size hint 1559 * (nelem_hint). 0 = use rhashtable default. Must be 1560 * <= min(max_entries, U16_MAX). Upper 32 bits reserved, 1561 * must be zero. 1562 */ 1563 __u64 map_extra; 1564 1565 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1566 * type data for 1567 * btf_vmlinux_value_type_id. 1568 */ 1569 /* BPF token FD to use with BPF_MAP_CREATE operation. 1570 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1571 */ 1572 __s32 map_token_fd; 1573 1574 /* Hash of the program that has exclusive access to the map. 1575 */ 1576 __aligned_u64 excl_prog_hash; 1577 /* Size of the passed excl_prog_hash. */ 1578 __u32 excl_prog_hash_size; 1579 }; 1580 1581 struct { /* anonymous struct used by BPF_MAP_*_ELEM and BPF_MAP_FREEZE commands */ 1582 __u32 map_fd; 1583 __aligned_u64 key; 1584 union { 1585 __aligned_u64 value; 1586 __aligned_u64 next_key; 1587 }; 1588 __u64 flags; 1589 }; 1590 1591 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1592 __aligned_u64 in_batch; /* start batch, 1593 * NULL to start from beginning 1594 */ 1595 __aligned_u64 out_batch; /* output: next start batch */ 1596 __aligned_u64 keys; 1597 __aligned_u64 values; 1598 __u32 count; /* input/output: 1599 * input: # of key/value 1600 * elements 1601 * output: # of filled elements 1602 */ 1603 __u32 map_fd; 1604 __u64 elem_flags; 1605 __u64 flags; 1606 } batch; 1607 1608 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1609 __u32 prog_type; /* one of enum bpf_prog_type */ 1610 __u32 insn_cnt; 1611 __aligned_u64 insns; 1612 __aligned_u64 license; 1613 __u32 log_level; /* verbosity level of verifier */ 1614 __u32 log_size; /* size of user buffer */ 1615 __aligned_u64 log_buf; /* user supplied buffer */ 1616 __u32 kern_version; /* not used */ 1617 __u32 prog_flags; 1618 char prog_name[BPF_OBJ_NAME_LEN]; 1619 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1620 /* For some prog types expected attach type must be known at 1621 * load time to verify attach type specific parts of prog 1622 * (context accesses, allowed helpers, etc). 1623 */ 1624 __u32 expected_attach_type; 1625 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1626 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1627 __aligned_u64 func_info; /* func info */ 1628 __u32 func_info_cnt; /* number of bpf_func_info records */ 1629 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1630 __aligned_u64 line_info; /* line info */ 1631 __u32 line_info_cnt; /* number of bpf_line_info records */ 1632 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1633 union { 1634 /* valid prog_fd to attach to bpf prog */ 1635 __u32 attach_prog_fd; 1636 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1637 __u32 attach_btf_obj_fd; 1638 }; 1639 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1640 __aligned_u64 fd_array; /* array of FDs */ 1641 __aligned_u64 core_relos; 1642 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1643 /* output: actual total log contents size (including termintaing zero). 1644 * It could be both larger than original log_size (if log was 1645 * truncated), or smaller (if log buffer wasn't filled completely). 1646 */ 1647 __u32 log_true_size; 1648 /* BPF token FD to use with BPF_PROG_LOAD operation. 1649 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1650 */ 1651 __s32 prog_token_fd; 1652 /* The fd_array_cnt can be used to pass the length of the 1653 * fd_array array. In this case all the [map] file descriptors 1654 * passed in this array will be bound to the program, even if 1655 * the maps are not referenced directly. The functionality is 1656 * similar to the BPF_PROG_BIND_MAP syscall, but maps can be 1657 * used by the verifier during the program load. If provided, 1658 * then the fd_array[0,...,fd_array_cnt-1] is expected to be 1659 * continuous. 1660 */ 1661 __u32 fd_array_cnt; 1662 /* Pointer to a buffer containing the signature of the BPF 1663 * program. 1664 */ 1665 __aligned_u64 signature; 1666 /* Size of the signature buffer in bytes. */ 1667 __u32 signature_size; 1668 /* ID of the kernel keyring to be used for signature 1669 * verification. 1670 */ 1671 __s32 keyring_id; 1672 }; 1673 1674 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1675 __aligned_u64 pathname; 1676 __u32 bpf_fd; 1677 __u32 file_flags; 1678 /* Same as dirfd in openat() syscall; see openat(2) 1679 * manpage for details of path FD and pathname semantics; 1680 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1681 * file_flags field, otherwise it should be set to zero; 1682 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1683 */ 1684 __s32 path_fd; 1685 }; 1686 1687 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1688 union { 1689 __u32 target_fd; /* target object to attach to or ... */ 1690 __u32 target_ifindex; /* target ifindex */ 1691 }; 1692 __u32 attach_bpf_fd; 1693 __u32 attach_type; 1694 __u32 attach_flags; 1695 __u32 replace_bpf_fd; 1696 union { 1697 __u32 relative_fd; 1698 __u32 relative_id; 1699 }; 1700 __u64 expected_revision; 1701 }; 1702 1703 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1704 __u32 prog_fd; 1705 __u32 retval; 1706 __u32 data_size_in; /* input: len of data_in */ 1707 __u32 data_size_out; /* input/output: len of data_out 1708 * returns ENOSPC if data_out 1709 * is too small. 1710 */ 1711 __aligned_u64 data_in; 1712 __aligned_u64 data_out; 1713 __u32 repeat; 1714 __u32 duration; 1715 __u32 ctx_size_in; /* input: len of ctx_in */ 1716 __u32 ctx_size_out; /* input/output: len of ctx_out 1717 * returns ENOSPC if ctx_out 1718 * is too small. 1719 */ 1720 __aligned_u64 ctx_in; 1721 __aligned_u64 ctx_out; 1722 __u32 flags; 1723 __u32 cpu; 1724 __u32 batch_size; 1725 } test; 1726 1727 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1728 union { 1729 __u32 start_id; 1730 __u32 prog_id; 1731 __u32 map_id; 1732 __u32 btf_id; 1733 __u32 link_id; 1734 }; 1735 __u32 next_id; 1736 __u32 open_flags; 1737 __s32 fd_by_id_token_fd; 1738 }; 1739 1740 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1741 __u32 bpf_fd; 1742 __u32 info_len; 1743 __aligned_u64 info; 1744 } info; 1745 1746 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1747 union { 1748 __u32 target_fd; /* target object to query or ... */ 1749 __u32 target_ifindex; /* target ifindex */ 1750 }; 1751 __u32 attach_type; 1752 __u32 query_flags; 1753 __u32 attach_flags; 1754 __aligned_u64 prog_ids; 1755 union { 1756 __u32 prog_cnt; 1757 __u32 count; 1758 }; 1759 __u32 :32; 1760 /* output: per-program attach_flags. 1761 * not allowed to be set during effective query. 1762 */ 1763 __aligned_u64 prog_attach_flags; 1764 __aligned_u64 link_ids; 1765 __aligned_u64 link_attach_flags; 1766 __u64 revision; 1767 } query; 1768 1769 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1770 __u64 name; 1771 __u32 prog_fd; 1772 __u32 :32; 1773 __aligned_u64 cookie; 1774 } raw_tracepoint; 1775 1776 struct { /* anonymous struct for BPF_BTF_LOAD */ 1777 __aligned_u64 btf; 1778 __aligned_u64 btf_log_buf; 1779 __u32 btf_size; 1780 __u32 btf_log_size; 1781 __u32 btf_log_level; 1782 /* output: actual total log contents size (including termintaing zero). 1783 * It could be both larger than original log_size (if log was 1784 * truncated), or smaller (if log buffer wasn't filled completely). 1785 */ 1786 __u32 btf_log_true_size; 1787 __u32 btf_flags; 1788 /* BPF token FD to use with BPF_BTF_LOAD operation. 1789 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1790 */ 1791 __s32 btf_token_fd; 1792 }; 1793 1794 struct { 1795 __u32 pid; /* input: pid */ 1796 __u32 fd; /* input: fd */ 1797 __u32 flags; /* input: flags */ 1798 __u32 buf_len; /* input/output: buf len */ 1799 __aligned_u64 buf; /* input/output: 1800 * tp_name for tracepoint 1801 * symbol for kprobe 1802 * filename for uprobe 1803 */ 1804 __u32 prog_id; /* output: prod_id */ 1805 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1806 __u64 probe_offset; /* output: probe_offset */ 1807 __u64 probe_addr; /* output: probe_addr */ 1808 } task_fd_query; 1809 1810 struct { /* struct used by BPF_LINK_CREATE command */ 1811 union { 1812 __u32 prog_fd; /* eBPF program to attach */ 1813 __u32 map_fd; /* struct_ops to attach */ 1814 }; 1815 union { 1816 __u32 target_fd; /* target object to attach to or ... */ 1817 __u32 target_ifindex; /* target ifindex */ 1818 }; 1819 __u32 attach_type; /* attach type */ 1820 __u32 flags; /* extra flags */ 1821 union { 1822 __u32 target_btf_id; /* btf_id of target to attach to */ 1823 struct { 1824 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1825 __u32 iter_info_len; /* iter_info length */ 1826 }; 1827 struct { 1828 /* black box user-provided value passed through 1829 * to BPF program at the execution time and 1830 * accessible through bpf_get_attach_cookie() BPF helper 1831 */ 1832 __u64 bpf_cookie; 1833 } perf_event; 1834 struct { 1835 __u32 flags; 1836 __u32 cnt; 1837 __aligned_u64 syms; 1838 __aligned_u64 addrs; 1839 __aligned_u64 cookies; 1840 } kprobe_multi; 1841 struct { 1842 /* this is overlaid with the target_btf_id above. */ 1843 __u32 target_btf_id; 1844 /* black box user-provided value passed through 1845 * to BPF program at the execution time and 1846 * accessible through bpf_get_attach_cookie() BPF helper 1847 */ 1848 __u64 cookie; 1849 } tracing; 1850 struct { 1851 __u32 pf; 1852 __u32 hooknum; 1853 __s32 priority; 1854 __u32 flags; 1855 } netfilter; 1856 struct { 1857 union { 1858 __u32 relative_fd; 1859 __u32 relative_id; 1860 }; 1861 __u64 expected_revision; 1862 } tcx; 1863 struct { 1864 __aligned_u64 path; 1865 __aligned_u64 offsets; 1866 __aligned_u64 ref_ctr_offsets; 1867 __aligned_u64 cookies; 1868 __u32 cnt; 1869 __u32 flags; 1870 __u32 pid; 1871 __u32 path_fd; 1872 } uprobe_multi; 1873 struct { 1874 union { 1875 __u32 relative_fd; 1876 __u32 relative_id; 1877 }; 1878 __u64 expected_revision; 1879 } netkit; 1880 struct { 1881 union { 1882 __u32 relative_fd; 1883 __u32 relative_id; 1884 }; 1885 __u64 expected_revision; 1886 } cgroup; 1887 struct { 1888 __aligned_u64 ids; 1889 __aligned_u64 cookies; 1890 __u32 cnt; 1891 } tracing_multi; 1892 }; 1893 } link_create; 1894 1895 struct { /* struct used by BPF_LINK_UPDATE command */ 1896 __u32 link_fd; /* link fd */ 1897 union { 1898 /* new program fd to update link with */ 1899 __u32 new_prog_fd; 1900 /* new struct_ops map fd to update link with */ 1901 __u32 new_map_fd; 1902 }; 1903 __u32 flags; /* extra flags */ 1904 union { 1905 /* expected link's program fd; is specified only if 1906 * BPF_F_REPLACE flag is set in flags. 1907 */ 1908 __u32 old_prog_fd; 1909 /* expected link's map fd; is specified only 1910 * if BPF_F_REPLACE flag is set. 1911 */ 1912 __u32 old_map_fd; 1913 }; 1914 } link_update; 1915 1916 struct { 1917 __u32 link_fd; 1918 } link_detach; 1919 1920 struct { /* struct used by BPF_ENABLE_STATS command */ 1921 __u32 type; 1922 } enable_stats; 1923 1924 struct { /* struct used by BPF_ITER_CREATE command */ 1925 __u32 link_fd; 1926 __u32 flags; 1927 } iter_create; 1928 1929 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1930 __u32 prog_fd; 1931 __u32 map_fd; 1932 __u32 flags; /* extra flags */ 1933 } prog_bind_map; 1934 1935 struct { /* struct used by BPF_TOKEN_CREATE command */ 1936 __u32 flags; 1937 __u32 bpffs_fd; 1938 } token_create; 1939 1940 struct { 1941 __aligned_u64 stream_buf; 1942 __u32 stream_buf_len; 1943 __u32 stream_id; 1944 __u32 prog_fd; 1945 } prog_stream_read; 1946 1947 struct { 1948 __u32 map_fd; 1949 __u32 prog_fd; 1950 __u32 flags; 1951 } prog_assoc_struct_ops; 1952 1953 } __attribute__((aligned(8))); 1954 1955 /* The description below is an attempt at providing documentation to eBPF 1956 * developers about the multiple available eBPF helper functions. It can be 1957 * parsed and used to produce a manual page. The workflow is the following, 1958 * and requires the rst2man utility: 1959 * 1960 * $ ./scripts/bpf_doc.py \ 1961 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1962 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1963 * $ man /tmp/bpf-helpers.7 1964 * 1965 * Note that in order to produce this external documentation, some RST 1966 * formatting is used in the descriptions to get "bold" and "italics" in 1967 * manual pages. Also note that the few trailing white spaces are 1968 * intentional, removing them would break paragraphs for rst2man. 1969 * 1970 * Start of BPF helper function descriptions: 1971 * 1972 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1973 * Description 1974 * Perform a lookup in *map* for an entry associated to *key*. 1975 * Return 1976 * Map value associated to *key*, or **NULL** if no entry was 1977 * found. 1978 * 1979 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1980 * Description 1981 * Add or update the value of the entry associated to *key* in 1982 * *map* with *value*. *flags* is one of: 1983 * 1984 * **BPF_NOEXIST** 1985 * The entry for *key* must not exist in the map. 1986 * **BPF_EXIST** 1987 * The entry for *key* must already exist in the map. 1988 * **BPF_ANY** 1989 * No condition on the existence of the entry for *key*. 1990 * 1991 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1992 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1993 * elements always exist), the helper would return an error. 1994 * Return 1995 * 0 on success, or a negative error in case of failure. 1996 * 1997 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1998 * Description 1999 * Delete entry with *key* from *map*. 2000 * Return 2001 * 0 on success, or a negative error in case of failure. 2002 * 2003 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 2004 * Description 2005 * For tracing programs, safely attempt to read *size* bytes from 2006 * kernel space address *unsafe_ptr* and store the data in *dst*. 2007 * 2008 * Generally, use **bpf_probe_read_user**\ () or 2009 * **bpf_probe_read_kernel**\ () instead. 2010 * Return 2011 * 0 on success, or a negative error in case of failure. 2012 * 2013 * u64 bpf_ktime_get_ns(void) 2014 * Description 2015 * Return the time elapsed since system boot, in nanoseconds. 2016 * Does not include time the system was suspended. 2017 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 2018 * Return 2019 * Current *ktime*. 2020 * 2021 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 2022 * Description 2023 * This helper is a "printk()-like" facility for debugging. It 2024 * prints a message defined by format *fmt* (of size *fmt_size*) 2025 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 2026 * available. It can take up to three additional **u64** 2027 * arguments (as an eBPF helpers, the total number of arguments is 2028 * limited to five). 2029 * 2030 * Each time the helper is called, it appends a line to the trace. 2031 * Lines are discarded while *\/sys/kernel/tracing/trace* is 2032 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 2033 * The format of the trace is customizable, and the exact output 2034 * one will get depends on the options set in 2035 * *\/sys/kernel/tracing/trace_options* (see also the 2036 * *README* file under the same directory). However, it usually 2037 * defaults to something like: 2038 * 2039 * :: 2040 * 2041 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 2042 * 2043 * In the above: 2044 * 2045 * * ``telnet`` is the name of the current task. 2046 * * ``470`` is the PID of the current task. 2047 * * ``001`` is the CPU number on which the task is 2048 * running. 2049 * * In ``.N..``, each character refers to a set of 2050 * options (whether irqs are enabled, scheduling 2051 * options, whether hard/softirqs are running, level of 2052 * preempt_disabled respectively). **N** means that 2053 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 2054 * are set. 2055 * * ``419421.045894`` is a timestamp. 2056 * * ``0x00000001`` is a fake value used by BPF for the 2057 * instruction pointer register. 2058 * * ``<formatted msg>`` is the message formatted with 2059 * *fmt*. 2060 * 2061 * The conversion specifiers supported by *fmt* are similar, but 2062 * more limited than for printk(). They are **%d**, **%i**, 2063 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 2064 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 2065 * of field, padding with zeroes, etc.) is available, and the 2066 * helper will return **-EINVAL** (but print nothing) if it 2067 * encounters an unknown specifier. 2068 * 2069 * Also, note that **bpf_trace_printk**\ () is slow, and should 2070 * only be used for debugging purposes. For this reason, a notice 2071 * block (spanning several lines) is printed to kernel logs and 2072 * states that the helper should not be used "for production use" 2073 * the first time this helper is used (or more precisely, when 2074 * **trace_printk**\ () buffers are allocated). For passing values 2075 * to user space, perf events should be preferred. 2076 * Return 2077 * The number of bytes written to the buffer, or a negative error 2078 * in case of failure. 2079 * 2080 * u32 bpf_get_prandom_u32(void) 2081 * Description 2082 * Get a pseudo-random number. 2083 * 2084 * From a security point of view, this helper uses its own 2085 * pseudo-random internal state, and cannot be used to infer the 2086 * seed of other random functions in the kernel. However, it is 2087 * essential to note that the generator used by the helper is not 2088 * cryptographically secure. 2089 * Return 2090 * A random 32-bit unsigned value. 2091 * 2092 * u32 bpf_get_smp_processor_id(void) 2093 * Description 2094 * Get the SMP (symmetric multiprocessing) processor id. Note that 2095 * all programs run with migration disabled, which means that the 2096 * SMP processor id is stable during all the execution of the 2097 * program. 2098 * Return 2099 * The SMP id of the processor running the program. 2100 * Attributes 2101 * __bpf_fastcall 2102 * 2103 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 2104 * Description 2105 * Store *len* bytes from address *from* into the packet 2106 * associated to *skb*, at *offset*. The *flags* are a combination 2107 * of the following values: 2108 * 2109 * **BPF_F_RECOMPUTE_CSUM** 2110 * Automatically update *skb*\ **->csum** after storing the 2111 * bytes. 2112 * **BPF_F_INVALIDATE_HASH** 2113 * Set *skb*\ **->hash**, *skb*\ **->swhash** and *skb*\ 2114 * **->l4hash** to 0. 2115 * 2116 * A call to this helper is susceptible to change the underlying 2117 * packet buffer. Therefore, at load time, all checks on pointers 2118 * previously done by the verifier are invalidated and must be 2119 * performed again, if the helper is used in combination with 2120 * direct packet access. 2121 * Return 2122 * 0 on success, or a negative error in case of failure. 2123 * 2124 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 2125 * Description 2126 * Recompute the layer 3 (e.g. IP) checksum for the packet 2127 * associated to *skb*. Computation is incremental, so the helper 2128 * must know the former value of the header field that was 2129 * modified (*from*), the new value of this field (*to*), and the 2130 * number of bytes (2 or 4) for this field, stored in *size*. 2131 * Alternatively, it is possible to store the difference between 2132 * the previous and the new values of the header field in *to*, by 2133 * setting *from* and *size* to 0. For both methods, *offset* 2134 * indicates the location of the IP checksum within the packet. 2135 * 2136 * This helper works in combination with **bpf_csum_diff**\ (), 2137 * which does not update the checksum in-place, but offers more 2138 * flexibility and can handle sizes larger than 2 or 4 for the 2139 * checksum to update. 2140 * 2141 * A call to this helper is susceptible to change the underlying 2142 * packet buffer. Therefore, at load time, all checks on pointers 2143 * previously done by the verifier are invalidated and must be 2144 * performed again, if the helper is used in combination with 2145 * direct packet access. 2146 * Return 2147 * 0 on success, or a negative error in case of failure. 2148 * 2149 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 2150 * Description 2151 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 2152 * packet associated to *skb*. Computation is incremental, so the 2153 * helper must know the former value of the header field that was 2154 * modified (*from*), the new value of this field (*to*), and the 2155 * number of bytes (2 or 4) for this field, stored on the lowest 2156 * four bits of *flags*. Alternatively, it is possible to store 2157 * the difference between the previous and the new values of the 2158 * header field in *to*, by setting *from* and the four lowest 2159 * bits of *flags* to 0. For both methods, *offset* indicates the 2160 * location of the IP checksum within the packet. In addition to 2161 * the size of the field, *flags* can be added (bitwise OR) actual 2162 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 2163 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 2164 * for updates resulting in a null checksum the value is set to 2165 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 2166 * that the modified header field is part of the pseudo-header. 2167 * Flag **BPF_F_IPV6** should be set for IPv6 packets. 2168 * 2169 * This helper works in combination with **bpf_csum_diff**\ (), 2170 * which does not update the checksum in-place, but offers more 2171 * flexibility and can handle sizes larger than 2 or 4 for the 2172 * checksum to update. 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_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2183 * Description 2184 * This special helper is used to trigger a "tail call", or in 2185 * other words, to jump into another eBPF program. The same stack 2186 * frame is used (but values on stack and in registers for the 2187 * caller are not accessible to the callee). This mechanism allows 2188 * for program chaining, either for raising the maximum number of 2189 * available eBPF instructions, or to execute given programs in 2190 * conditional blocks. For security reasons, there is an upper 2191 * limit to the number of successive tail calls that can be 2192 * performed. 2193 * 2194 * Upon call of this helper, the program attempts to jump into a 2195 * program referenced at index *index* in *prog_array_map*, a 2196 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2197 * *ctx*, a pointer to the context. 2198 * 2199 * If the call succeeds, the kernel immediately runs the first 2200 * instruction of the new program. This is not a function call, 2201 * and it never returns to the previous program. If the call 2202 * fails, then the helper has no effect, and the caller continues 2203 * to run its subsequent instructions. A call can fail if the 2204 * destination program for the jump does not exist (i.e. *index* 2205 * is superior to the number of entries in *prog_array_map*), or 2206 * if the maximum number of tail calls has been reached for this 2207 * chain of programs. This limit is defined in the kernel by the 2208 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2209 * which is currently set to 33. 2210 * Return 2211 * 0 on success, or a negative error in case of failure. 2212 * 2213 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2214 * Description 2215 * Clone and redirect the packet associated to *skb* to another 2216 * net device of index *ifindex*. Both ingress and egress 2217 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2218 * value in *flags* is used to make the distinction (ingress path 2219 * is selected if the flag is present, egress path otherwise). 2220 * This is the only flag supported for now. 2221 * 2222 * In comparison with **bpf_redirect**\ () helper, 2223 * **bpf_clone_redirect**\ () has the associated cost of 2224 * duplicating the packet buffer, but this can be executed out of 2225 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2226 * efficient, but it is handled through an action code where the 2227 * redirection happens only after the eBPF program has returned. 2228 * 2229 * A call to this helper is susceptible to change the underlying 2230 * packet buffer. Therefore, at load time, all checks on pointers 2231 * previously done by the verifier are invalidated and must be 2232 * performed again, if the helper is used in combination with 2233 * direct packet access. 2234 * Return 2235 * 0 on success, or a negative error in case of failure. Positive 2236 * error indicates a potential drop or congestion in the target 2237 * device. The particular positive error codes are not defined. 2238 * 2239 * u64 bpf_get_current_pid_tgid(void) 2240 * Description 2241 * Get the current pid and tgid. 2242 * Return 2243 * A 64-bit integer containing the current tgid and pid, and 2244 * created as such: 2245 * *current_task*\ **->tgid << 32 \|** 2246 * *current_task*\ **->pid**. 2247 * 2248 * u64 bpf_get_current_uid_gid(void) 2249 * Description 2250 * Get the current uid and gid. 2251 * Return 2252 * A 64-bit integer containing the current GID and UID, and 2253 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2254 * 2255 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2256 * Description 2257 * Copy the **comm** attribute of the current task into *buf* of 2258 * *size_of_buf*. The **comm** attribute contains the name of 2259 * the executable (excluding the path) for the current task. The 2260 * *size_of_buf* must be strictly positive. On success, the 2261 * helper makes sure that the *buf* is NUL-terminated. On failure, 2262 * it is filled with zeroes. 2263 * Return 2264 * 0 on success, or a negative error in case of failure. 2265 * 2266 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2267 * Description 2268 * Retrieve the classid for the current task, i.e. for the net_cls 2269 * cgroup to which *skb* belongs. 2270 * 2271 * This helper can be used on TC egress path, but not on ingress. 2272 * 2273 * The net_cls cgroup provides an interface to tag network packets 2274 * based on a user-provided identifier for all traffic coming from 2275 * the tasks belonging to the related cgroup. See also the related 2276 * kernel documentation, available from the Linux sources in file 2277 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2278 * 2279 * The Linux kernel has two versions for cgroups: there are 2280 * cgroups v1 and cgroups v2. Both are available to users, who can 2281 * use a mixture of them, but note that the net_cls cgroup is for 2282 * cgroup v1 only. This makes it incompatible with BPF programs 2283 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2284 * only hold data for one version of cgroups at a time). 2285 * 2286 * This helper is only available is the kernel was compiled with 2287 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2288 * "**y**" or to "**m**". 2289 * Return 2290 * The classid, or 0 for the default unconfigured classid. 2291 * 2292 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2293 * Description 2294 * Push a *vlan_tci* (VLAN tag control information) of protocol 2295 * *vlan_proto* to the packet associated to *skb*, then update 2296 * the checksum. Note that if *vlan_proto* is different from 2297 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2298 * be **ETH_P_8021Q**. 2299 * 2300 * A call to this helper is susceptible to change the underlying 2301 * packet buffer. Therefore, at load time, all checks on pointers 2302 * previously done by the verifier are invalidated and must be 2303 * performed again, if the helper is used in combination with 2304 * direct packet access. 2305 * Return 2306 * 0 on success, or a negative error in case of failure. 2307 * 2308 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2309 * Description 2310 * Pop a VLAN header from the packet associated to *skb*. 2311 * 2312 * A call to this helper is susceptible to change the underlying 2313 * packet buffer. Therefore, at load time, all checks on pointers 2314 * previously done by the verifier are invalidated and must be 2315 * performed again, if the helper is used in combination with 2316 * direct packet access. 2317 * Return 2318 * 0 on success, or a negative error in case of failure. 2319 * 2320 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2321 * Description 2322 * Get tunnel metadata. This helper takes a pointer *key* to an 2323 * empty **struct bpf_tunnel_key** of **size**, that will be 2324 * filled with tunnel metadata for the packet associated to *skb*. 2325 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2326 * indicates that the tunnel is based on IPv6 protocol instead of 2327 * IPv4. 2328 * 2329 * The **struct bpf_tunnel_key** is an object that generalizes the 2330 * principal parameters used by various tunneling protocols into a 2331 * single struct. This way, it can be used to easily make a 2332 * decision based on the contents of the encapsulation header, 2333 * "summarized" in this struct. In particular, it holds the IP 2334 * address of the remote end (IPv4 or IPv6, depending on the case) 2335 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2336 * this struct exposes the *key*\ **->tunnel_id**, which is 2337 * generally mapped to a VNI (Virtual Network Identifier), making 2338 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2339 * () helper. 2340 * 2341 * Let's imagine that the following code is part of a program 2342 * attached to the TC ingress interface, on one end of a GRE 2343 * tunnel, and is supposed to filter out all messages coming from 2344 * remote ends with IPv4 address other than 10.0.0.1: 2345 * 2346 * :: 2347 * 2348 * int ret; 2349 * struct bpf_tunnel_key key = {}; 2350 * 2351 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2352 * if (ret < 0) 2353 * return TC_ACT_SHOT; // drop packet 2354 * 2355 * if (key.remote_ipv4 != 0x0a000001) 2356 * return TC_ACT_SHOT; // drop packet 2357 * 2358 * return TC_ACT_OK; // accept packet 2359 * 2360 * This interface can also be used with all encapsulation devices 2361 * that can operate in "collect metadata" mode: instead of having 2362 * one network device per specific configuration, the "collect 2363 * metadata" mode only requires a single device where the 2364 * configuration can be extracted from this helper. 2365 * 2366 * This can be used together with various tunnels such as VXLan, 2367 * Geneve, GRE or IP in IP (IPIP). 2368 * Return 2369 * 0 on success, or a negative error in case of failure. 2370 * 2371 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2372 * Description 2373 * Populate tunnel metadata for packet associated to *skb.* The 2374 * tunnel metadata is set to the contents of *key*, of *size*. The 2375 * *flags* can be set to a combination of the following values: 2376 * 2377 * **BPF_F_TUNINFO_IPV6** 2378 * Indicate that the tunnel is based on IPv6 protocol 2379 * instead of IPv4. 2380 * **BPF_F_ZERO_CSUM_TX** 2381 * For IPv4 packets, add a flag to tunnel metadata 2382 * indicating that checksum computation should be skipped 2383 * and checksum set to zeroes. 2384 * **BPF_F_DONT_FRAGMENT** 2385 * Add a flag to tunnel metadata indicating that the 2386 * packet should not be fragmented. 2387 * **BPF_F_SEQ_NUMBER** 2388 * Add a flag to tunnel metadata indicating that a 2389 * sequence number should be added to tunnel header before 2390 * sending the packet. This flag was added for GRE 2391 * encapsulation, but might be used with other protocols 2392 * as well in the future. 2393 * **BPF_F_NO_TUNNEL_KEY** 2394 * Add a flag to tunnel metadata indicating that no tunnel 2395 * key should be set in the resulting tunnel header. 2396 * 2397 * Here is a typical usage on the transmit path: 2398 * 2399 * :: 2400 * 2401 * struct bpf_tunnel_key key; 2402 * populate key ... 2403 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2404 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2405 * 2406 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2407 * helper for additional information. 2408 * Return 2409 * 0 on success, or a negative error in case of failure. 2410 * 2411 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2412 * Description 2413 * Read the value of a perf event counter. This helper relies on a 2414 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2415 * the perf event counter is selected when *map* is updated with 2416 * perf event file descriptors. The *map* is an array whose size 2417 * is the number of available CPUs, and each cell contains a value 2418 * relative to one CPU. The value to retrieve is indicated by 2419 * *flags*, that contains the index of the CPU to look up, masked 2420 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2421 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2422 * current CPU should be retrieved. 2423 * 2424 * Note that before Linux 4.13, only hardware perf event can be 2425 * retrieved. 2426 * 2427 * Also, be aware that the newer helper 2428 * **bpf_perf_event_read_value**\ () is recommended over 2429 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2430 * quirks where error and counter value are used as a return code 2431 * (which is wrong to do since ranges may overlap). This issue is 2432 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2433 * time provides more features over the **bpf_perf_event_read**\ 2434 * () interface. Please refer to the description of 2435 * **bpf_perf_event_read_value**\ () for details. 2436 * Return 2437 * The value of the perf event counter read from the map, or a 2438 * negative error code in case of failure. 2439 * 2440 * long bpf_redirect(u32 ifindex, u64 flags) 2441 * Description 2442 * Redirect the packet to another net device of index *ifindex*. 2443 * This helper is somewhat similar to **bpf_clone_redirect**\ 2444 * (), except that the packet is not cloned, which provides 2445 * increased performance. 2446 * 2447 * Except for XDP, both ingress and egress interfaces can be used 2448 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2449 * to make the distinction (ingress path is selected if the flag 2450 * is present, egress path otherwise). Currently, XDP only 2451 * supports redirection to the egress interface, and accepts no 2452 * flag at all. 2453 * 2454 * The same effect can also be attained with the more generic 2455 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2456 * redirect target instead of providing it directly to the helper. 2457 * Return 2458 * For XDP, the helper returns **XDP_REDIRECT** on success or 2459 * **XDP_ABORTED** on error. For other program types, the values 2460 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2461 * error. 2462 * 2463 * u32 bpf_get_route_realm(struct sk_buff *skb) 2464 * Description 2465 * Retrieve the realm or the route, that is to say the 2466 * **tclassid** field of the destination for the *skb*. The 2467 * identifier retrieved is a user-provided tag, similar to the 2468 * one used with the net_cls cgroup (see description for 2469 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2470 * held by a route (a destination entry), not by a task. 2471 * 2472 * Retrieving this identifier works with the clsact TC egress hook 2473 * (see also **tc-bpf(8)**), or alternatively on conventional 2474 * classful egress qdiscs, but not on TC ingress path. In case of 2475 * clsact TC egress hook, this has the advantage that, internally, 2476 * the destination entry has not been dropped yet in the transmit 2477 * path. Therefore, the destination entry does not need to be 2478 * artificially held via **netif_keep_dst**\ () for a classful 2479 * qdisc until the *skb* is freed. 2480 * 2481 * This helper is available only if the kernel was compiled with 2482 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2483 * Return 2484 * The realm of the route for the packet associated to *skb*, or 0 2485 * if none was found. 2486 * 2487 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2488 * Description 2489 * Write raw *data* blob into a special BPF perf event held by 2490 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2491 * event must have the following attributes: **PERF_SAMPLE_RAW** 2492 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2493 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2494 * 2495 * The *flags* are used to indicate the index in *map* for which 2496 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2497 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2498 * to indicate that the index of the current CPU core should be 2499 * used. 2500 * 2501 * The value to write, of *size*, is passed through eBPF stack and 2502 * pointed by *data*. 2503 * 2504 * The context of the program *ctx* needs also be passed to the 2505 * helper. 2506 * 2507 * On user space, a program willing to read the values needs to 2508 * call **perf_event_open**\ () on the perf event (either for 2509 * one or for all CPUs) and to store the file descriptor into the 2510 * *map*. This must be done before the eBPF program can send data 2511 * into it. An example is available in file 2512 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2513 * tree (the eBPF program counterpart is in 2514 * *samples/bpf/trace_output.bpf.c*). 2515 * 2516 * **bpf_perf_event_output**\ () achieves better performance 2517 * than **bpf_trace_printk**\ () for sharing data with user 2518 * space, and is much better suitable for streaming data from eBPF 2519 * programs. 2520 * 2521 * Note that this helper is not restricted to tracing use cases 2522 * and can be used with programs attached to TC or XDP as well, 2523 * where it allows for passing data to user space listeners. Data 2524 * can be: 2525 * 2526 * * Only custom structs, 2527 * * Only the packet payload, or 2528 * * A combination of both. 2529 * Return 2530 * 0 on success, or a negative error in case of failure. 2531 * 2532 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2533 * Description 2534 * This helper was provided as an easy way to load data from a 2535 * packet. It can be used to load *len* bytes from *offset* from 2536 * the packet associated to *skb*, into the buffer pointed by 2537 * *to*. 2538 * 2539 * Since Linux 4.7, usage of this helper has mostly been replaced 2540 * by "direct packet access", enabling packet data to be 2541 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2542 * pointing respectively to the first byte of packet data and to 2543 * the byte after the last byte of packet data. However, it 2544 * remains useful if one wishes to read large quantities of data 2545 * at once from a packet into the eBPF stack. 2546 * Return 2547 * 0 on success, or a negative error in case of failure. 2548 * 2549 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2550 * Description 2551 * Walk a user or a kernel stack and return its id. To achieve 2552 * this, the helper needs *ctx*, which is a pointer to the context 2553 * on which the tracing program is executed, and a pointer to a 2554 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2555 * 2556 * The last argument, *flags*, holds the number of stack frames to 2557 * skip (from 0 to 255), masked with 2558 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2559 * a combination of the following flags: 2560 * 2561 * **BPF_F_USER_STACK** 2562 * Collect a user space stack instead of a kernel stack. 2563 * **BPF_F_FAST_STACK_CMP** 2564 * Compare stacks by hash only. 2565 * **BPF_F_REUSE_STACKID** 2566 * If two different stacks hash into the same *stackid*, 2567 * discard the old one. 2568 * 2569 * The stack id retrieved is a 32 bit long integer handle which 2570 * can be further combined with other data (including other stack 2571 * ids) and used as a key into maps. This can be useful for 2572 * generating a variety of graphs (such as flame graphs or off-cpu 2573 * graphs). 2574 * 2575 * For walking a stack, this helper is an improvement over 2576 * **bpf_probe_read**\ (), which can be used with unrolled loops 2577 * but is not efficient and consumes a lot of eBPF instructions. 2578 * Instead, **bpf_get_stackid**\ () can collect up to 2579 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2580 * this limit can be controlled with the **sysctl** program, and 2581 * that it should be manually increased in order to profile long 2582 * user stacks (such as stacks for Java programs). To do so, use: 2583 * 2584 * :: 2585 * 2586 * # sysctl kernel.perf_event_max_stack=<new value> 2587 * Return 2588 * The positive or null stack id on success, or a negative error 2589 * in case of failure. 2590 * 2591 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2592 * Description 2593 * Compute a checksum difference, from the raw buffer pointed by 2594 * *from*, of length *from_size* (that must be a multiple of 4), 2595 * towards the raw buffer pointed by *to*, of size *to_size* 2596 * (same remark). An optional *seed* can be added to the value 2597 * (this can be cascaded, the seed may come from a previous call 2598 * to the helper). 2599 * 2600 * This is flexible enough to be used in several ways: 2601 * 2602 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2603 * checksum, it can be used when pushing new data. 2604 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2605 * checksum, it can be used when removing data from a packet. 2606 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2607 * can be used to compute a diff. Note that *from_size* and 2608 * *to_size* do not need to be equal. 2609 * 2610 * This helper can be used in combination with 2611 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2612 * which one can feed in the difference computed with 2613 * **bpf_csum_diff**\ (). 2614 * Return 2615 * The checksum result, or a negative error code in case of 2616 * failure. 2617 * 2618 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2619 * Description 2620 * Retrieve tunnel options metadata for the packet associated to 2621 * *skb*, and store the raw tunnel option data to the buffer *opt* 2622 * of *size*. 2623 * 2624 * This helper can be used with encapsulation devices that can 2625 * operate in "collect metadata" mode (please refer to the related 2626 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2627 * more details). A particular example where this can be used is 2628 * in combination with the Geneve encapsulation protocol, where it 2629 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2630 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2631 * the eBPF program. This allows for full customization of these 2632 * headers. 2633 * Return 2634 * The size of the option data retrieved. 2635 * 2636 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2637 * Description 2638 * Set tunnel options metadata for the packet associated to *skb* 2639 * to the option data contained in the raw buffer *opt* of *size*. 2640 * 2641 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2642 * helper for additional information. 2643 * Return 2644 * 0 on success, or a negative error in case of failure. 2645 * 2646 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2647 * Description 2648 * Change the protocol of the *skb* to *proto*. Currently 2649 * supported are transition from IPv4 to IPv6, and from IPv6 to 2650 * IPv4. The helper takes care of the groundwork for the 2651 * transition, including resizing the socket buffer. The eBPF 2652 * program is expected to fill the new headers, if any, via 2653 * **skb_store_bytes**\ () and to recompute the checksums with 2654 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2655 * (). The main case for this helper is to perform NAT64 2656 * operations out of an eBPF program. 2657 * 2658 * Internally, the GSO type is marked as dodgy so that headers are 2659 * checked and segments are recalculated by the GSO/GRO engine. 2660 * The size for GSO target is adapted as well. 2661 * 2662 * All values for *flags* are reserved for future usage, and must 2663 * be left at zero. 2664 * 2665 * A call to this helper is susceptible to change the underlying 2666 * packet buffer. Therefore, at load time, all checks on pointers 2667 * previously done by the verifier are invalidated and must be 2668 * performed again, if the helper is used in combination with 2669 * direct packet access. 2670 * Return 2671 * 0 on success, or a negative error in case of failure. 2672 * 2673 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2674 * Description 2675 * Change the packet type for the packet associated to *skb*. This 2676 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2677 * the eBPF program does not have a write access to *skb*\ 2678 * **->pkt_type** beside this helper. Using a helper here allows 2679 * for graceful handling of errors. 2680 * 2681 * The major use case is to change incoming *skb*s to 2682 * **PACKET_HOST** in a programmatic way instead of having to 2683 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2684 * example. 2685 * 2686 * Note that *type* only allows certain values. At this time, they 2687 * are: 2688 * 2689 * **PACKET_HOST** 2690 * Packet is for us. 2691 * **PACKET_BROADCAST** 2692 * Send packet to all. 2693 * **PACKET_MULTICAST** 2694 * Send packet to group. 2695 * **PACKET_OTHERHOST** 2696 * Send packet to someone else. 2697 * Return 2698 * 0 on success, or a negative error in case of failure. 2699 * 2700 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2701 * Description 2702 * Check whether *skb* is a descendant of the cgroup2 held by 2703 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2704 * Return 2705 * The return value depends on the result of the test, and can be: 2706 * 2707 * * 0, if the *skb* failed the cgroup2 descendant test. 2708 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2709 * * A negative error code, if an error occurred. 2710 * 2711 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2712 * Description 2713 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2714 * not set, in particular if the hash was cleared due to mangling, 2715 * recompute this hash. Later accesses to the hash can be done 2716 * directly with *skb*\ **->hash**. 2717 * 2718 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2719 * prototype with **bpf_skb_change_proto**\ (), or calling 2720 * **bpf_skb_store_bytes**\ () with the 2721 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2722 * the hash and to trigger a new computation for the next call to 2723 * **bpf_get_hash_recalc**\ (). 2724 * Return 2725 * The 32-bit hash. 2726 * 2727 * u64 bpf_get_current_task(void) 2728 * Description 2729 * Get the current task. 2730 * Return 2731 * A pointer to the current task struct. 2732 * 2733 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2734 * Description 2735 * Attempt in a safe way to write *len* bytes from the buffer 2736 * *src* to *dst* in memory. It only works for threads that are in 2737 * user context, and *dst* must be a valid user space address. 2738 * 2739 * This helper should not be used to implement any kind of 2740 * security mechanism because of TOC-TOU attacks, but rather to 2741 * debug, divert, and manipulate execution of semi-cooperative 2742 * processes. 2743 * 2744 * Keep in mind that this feature is meant for experiments, and it 2745 * has a risk of crashing the system and running programs. 2746 * Therefore, when an eBPF program using this helper is attached, 2747 * a warning including PID and process name is printed to kernel 2748 * logs. 2749 * Return 2750 * 0 on success, or a negative error in case of failure. 2751 * 2752 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2753 * Description 2754 * Check whether the probe is being run is the context of a given 2755 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2756 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2757 * Return 2758 * The return value depends on the result of the test, and can be: 2759 * 2760 * * 1, if current task belongs to the cgroup2. 2761 * * 0, if current task does not belong to the cgroup2. 2762 * * A negative error code, if an error occurred. 2763 * 2764 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2765 * Description 2766 * Resize (trim or grow) the packet associated to *skb* to the 2767 * new *len*. The *flags* are reserved for future usage, and must 2768 * be left at zero. 2769 * 2770 * The basic idea is that the helper performs the needed work to 2771 * change the size of the packet, then the eBPF program rewrites 2772 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2773 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2774 * and others. This helper is a slow path utility intended for 2775 * replies with control messages. And because it is targeted for 2776 * slow path, the helper itself can afford to be slow: it 2777 * implicitly linearizes, unclones and drops offloads from the 2778 * *skb*. 2779 * 2780 * A call to this helper is susceptible to change the underlying 2781 * packet buffer. Therefore, at load time, all checks on pointers 2782 * previously done by the verifier are invalidated and must be 2783 * performed again, if the helper is used in combination with 2784 * direct packet access. 2785 * Return 2786 * 0 on success, or a negative error in case of failure. 2787 * 2788 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2789 * Description 2790 * Pull in non-linear data in case the *skb* is non-linear and not 2791 * all of *len* are part of the linear section. Make *len* bytes 2792 * from *skb* readable and writable. If a zero value is passed for 2793 * *len*, then all bytes in the linear part of *skb* will be made 2794 * readable and writable. 2795 * 2796 * This helper is only needed for reading and writing with direct 2797 * packet access. 2798 * 2799 * For direct packet access, testing that offsets to access 2800 * are within packet boundaries (test on *skb*\ **->data_end**) is 2801 * susceptible to fail if offsets are invalid, or if the requested 2802 * data is in non-linear parts of the *skb*. On failure the 2803 * program can just bail out, or in the case of a non-linear 2804 * buffer, use a helper to make the data available. The 2805 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2806 * the data. Another one consists in using **bpf_skb_pull_data** 2807 * to pull in once the non-linear parts, then retesting and 2808 * eventually access the data. 2809 * 2810 * At the same time, this also makes sure the *skb* is uncloned, 2811 * which is a necessary condition for direct write. As this needs 2812 * to be an invariant for the write part only, the verifier 2813 * detects writes and adds a prologue that is calling 2814 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2815 * the very beginning in case it is indeed cloned. 2816 * 2817 * A call to this helper is susceptible to change the underlying 2818 * packet buffer. Therefore, at load time, all checks on pointers 2819 * previously done by the verifier are invalidated and must be 2820 * performed again, if the helper is used in combination with 2821 * direct packet access. 2822 * Return 2823 * 0 on success, or a negative error in case of failure. 2824 * 2825 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2826 * Description 2827 * Add the checksum *csum* into *skb*\ **->csum** in case the 2828 * driver has supplied a checksum for the entire packet into that 2829 * field. Return an error otherwise. This helper is intended to be 2830 * used in combination with **bpf_csum_diff**\ (), in particular 2831 * when the checksum needs to be updated after data has been 2832 * written into the packet through direct packet access. 2833 * Return 2834 * The checksum on success, or a negative error code in case of 2835 * failure. 2836 * 2837 * void bpf_set_hash_invalid(struct sk_buff *skb) 2838 * Description 2839 * Invalidate the current *skb*\ **->hash**. It can be used after 2840 * mangling on headers through direct packet access, in order to 2841 * indicate that the hash is outdated and to trigger a 2842 * recalculation the next time the kernel tries to access this 2843 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2844 * Return 2845 * void. 2846 * 2847 * long bpf_get_numa_node_id(void) 2848 * Description 2849 * Return the id of the current NUMA node. The primary use case 2850 * for this helper is the selection of sockets for the local NUMA 2851 * node, when the program is attached to sockets using the 2852 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2853 * but the helper is also available to other eBPF program types, 2854 * similarly to **bpf_get_smp_processor_id**\ (). 2855 * Return 2856 * The id of current NUMA node. 2857 * 2858 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2859 * Description 2860 * Grows headroom of packet associated to *skb* and adjusts the 2861 * offset of the MAC header accordingly, adding *len* bytes of 2862 * space. It automatically extends and reallocates memory as 2863 * required. 2864 * 2865 * This helper can be used on a layer 3 *skb* to push a MAC header 2866 * for redirection into a layer 2 device. 2867 * 2868 * All values for *flags* are reserved for future usage, and must 2869 * be left at zero. 2870 * 2871 * A call to this helper is susceptible to change the underlying 2872 * packet buffer. Therefore, at load time, all checks on pointers 2873 * previously done by the verifier are invalidated and must be 2874 * performed again, if the helper is used in combination with 2875 * direct packet access. 2876 * Return 2877 * 0 on success, or a negative error in case of failure. 2878 * 2879 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2880 * Description 2881 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2882 * it is possible to use a negative value for *delta*. This helper 2883 * can be used to prepare the packet for pushing or popping 2884 * headers. 2885 * 2886 * A call to this helper is susceptible to change the underlying 2887 * packet buffer. Therefore, at load time, all checks on pointers 2888 * previously done by the verifier are invalidated and must be 2889 * performed again, if the helper is used in combination with 2890 * direct packet access. 2891 * Return 2892 * 0 on success, or a negative error in case of failure. 2893 * 2894 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2895 * Description 2896 * Copy a NUL terminated string from an unsafe kernel address 2897 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2898 * more details. 2899 * 2900 * Generally, use **bpf_probe_read_user_str**\ () or 2901 * **bpf_probe_read_kernel_str**\ () instead. 2902 * Return 2903 * On success, the strictly positive length of the string, 2904 * including the trailing NUL character. On error, a negative 2905 * value. 2906 * 2907 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2908 * Description 2909 * If the **struct sk_buff** pointed by *skb* has a known socket, 2910 * retrieve the cookie (generated by the kernel) of this socket. 2911 * If no cookie has been set yet, generate a new cookie. Once 2912 * generated, the socket cookie remains stable for the life of the 2913 * socket. This helper can be useful for monitoring per socket 2914 * networking traffic statistics as it provides a global socket 2915 * identifier that can be assumed unique. 2916 * Return 2917 * A 8-byte long unique number on success, or 0 if the socket 2918 * field is missing inside *skb*. 2919 * 2920 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2921 * Description 2922 * Equivalent to bpf_get_socket_cookie() helper that accepts 2923 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2924 * Return 2925 * A 8-byte long unique number. 2926 * 2927 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2928 * Description 2929 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2930 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2931 * Return 2932 * A 8-byte long unique number. 2933 * 2934 * u64 bpf_get_socket_cookie(struct sock *sk) 2935 * Description 2936 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2937 * *sk*, but gets socket from a BTF **struct sock**. This helper 2938 * also works for sleepable programs. 2939 * Return 2940 * A 8-byte long unique number or 0 if *sk* is NULL. 2941 * 2942 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2943 * Description 2944 * Get the owner UID of the socked associated to *skb*. 2945 * Return 2946 * The owner UID of the socket associated to *skb*. If the socket 2947 * is **NULL**, or if it is not a full socket (i.e. if it is a 2948 * time-wait or a request socket instead), **overflowuid** value 2949 * is returned (note that **overflowuid** might also be the actual 2950 * UID value for the socket). 2951 * 2952 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2953 * Description 2954 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2955 * to value *hash*. 2956 * Return 2957 * 0 2958 * 2959 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2960 * Description 2961 * Emulate a call to **setsockopt()** on the socket associated to 2962 * *bpf_socket*, which must be a full socket. The *level* at 2963 * which the option resides and the name *optname* of the option 2964 * must be specified, see **setsockopt(2)** for more information. 2965 * The option value of length *optlen* is pointed by *optval*. 2966 * 2967 * *bpf_socket* should be one of the following: 2968 * 2969 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2970 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2971 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2972 * 2973 * This helper actually implements a subset of **setsockopt()**. 2974 * It supports the following *level*\ s: 2975 * 2976 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2977 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2978 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2979 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2980 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2981 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2982 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2983 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2984 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2985 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2986 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2987 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2988 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**. 2989 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2990 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2991 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2992 * Return 2993 * 0 on success, or a negative error in case of failure. 2994 * 2995 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2996 * Description 2997 * Grow or shrink the room for data in the packet associated to 2998 * *skb* by *len_diff*, and according to the selected *mode*. 2999 * 3000 * By default, the helper will reset any offloaded checksum 3001 * indicator of the skb to CHECKSUM_NONE. This can be avoided 3002 * by the following flag: 3003 * 3004 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 3005 * checksum data of the skb to CHECKSUM_NONE. 3006 * 3007 * There are two supported modes at this time: 3008 * 3009 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 3010 * (room space is added or removed between the layer 2 and 3011 * layer 3 headers). 3012 * 3013 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 3014 * (room space is added or removed between the layer 3 and 3015 * layer 4 headers). 3016 * 3017 * The following flags are supported at this time: 3018 * 3019 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 3020 * Adjusting mss in this way is not allowed for datagrams. 3021 * 3022 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 3023 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 3024 * Any new space is reserved to hold a tunnel header. 3025 * Configure skb offsets and other fields accordingly. 3026 * 3027 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 3028 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 3029 * Use with ENCAP_L3 flags to further specify the tunnel type. 3030 * 3031 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 3032 * Use with ENCAP_L3/L4 flags to further specify the tunnel 3033 * type; *len* is the length of the inner MAC header. 3034 * 3035 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 3036 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 3037 * L2 type as Ethernet. 3038 * 3039 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 3040 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 3041 * Indicate the new IP header version after decapsulating the outer 3042 * IP header. Used when the inner and outer IP versions are different. 3043 * 3044 * A call to this helper is susceptible to change the underlying 3045 * packet buffer. Therefore, at load time, all checks on pointers 3046 * previously done by the verifier are invalidated and must be 3047 * performed again, if the helper is used in combination with 3048 * direct packet access. 3049 * Return 3050 * 0 on success, or a negative error in case of failure. 3051 * 3052 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 3053 * Description 3054 * Redirect the packet to the endpoint referenced by *map* at 3055 * index *key*. Depending on its type, this *map* can contain 3056 * references to net devices (for forwarding packets through other 3057 * ports), or to CPUs (for redirecting XDP frames to another CPU; 3058 * but this is only implemented for native XDP (with driver 3059 * support) as of this writing). 3060 * 3061 * The lower two bits of *flags* are used as the return code if 3062 * the map lookup fails. This is so that the return value can be 3063 * one of the XDP program return codes up to **XDP_TX**, as chosen 3064 * by the caller. The higher bits of *flags* can be set to 3065 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 3066 * 3067 * With BPF_F_BROADCAST the packet will be broadcasted to all the 3068 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 3069 * interface will be excluded when do broadcasting. 3070 * 3071 * See also **bpf_redirect**\ (), which only supports redirecting 3072 * to an ifindex, but doesn't require a map to do so. 3073 * Return 3074 * **XDP_REDIRECT** on success, or the value of the two lower bits 3075 * of the *flags* argument on error. 3076 * 3077 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 3078 * Description 3079 * Redirect the packet to the socket referenced by *map* (of type 3080 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3081 * egress interfaces can be used for redirection. The 3082 * **BPF_F_INGRESS** value in *flags* is used to make the 3083 * distinction (ingress path is selected if the flag is present, 3084 * egress path otherwise). This is the only flag supported for now. 3085 * Return 3086 * **SK_PASS** on success, or **SK_DROP** on error. 3087 * 3088 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3089 * Description 3090 * Add an entry to, or update a *map* referencing sockets. The 3091 * *skops* is used as a new value for the entry associated to 3092 * *key*. *flags* is one of: 3093 * 3094 * **BPF_NOEXIST** 3095 * The entry for *key* must not exist in the map. 3096 * **BPF_EXIST** 3097 * The entry for *key* must already exist in the map. 3098 * **BPF_ANY** 3099 * No condition on the existence of the entry for *key*. 3100 * 3101 * If the *map* has eBPF programs (parser and verdict), those will 3102 * be inherited by the socket being added. If the socket is 3103 * already attached to eBPF programs, this results in an error. 3104 * Return 3105 * 0 on success, or a negative error in case of failure. 3106 * 3107 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 3108 * Description 3109 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 3110 * *delta* (which can be positive or negative). Note that this 3111 * operation modifies the address stored in *xdp_md*\ **->data**, 3112 * so the latter must be loaded only after the helper has been 3113 * called. 3114 * 3115 * The use of *xdp_md*\ **->data_meta** is optional and programs 3116 * are not required to use it. The rationale is that when the 3117 * packet is processed with XDP (e.g. as DoS filter), it is 3118 * possible to push further meta data along with it before passing 3119 * to the stack, and to give the guarantee that an ingress eBPF 3120 * program attached as a TC classifier on the same device can pick 3121 * this up for further post-processing. Since TC works with socket 3122 * buffers, it remains possible to set from XDP the **mark** or 3123 * **priority** pointers, or other pointers for the socket buffer. 3124 * Having this scratch space generic and programmable allows for 3125 * more flexibility as the user is free to store whatever meta 3126 * data they need. 3127 * 3128 * A call to this helper is susceptible to change the underlying 3129 * packet buffer. Therefore, at load time, all checks on pointers 3130 * previously done by the verifier are invalidated and must be 3131 * performed again, if the helper is used in combination with 3132 * direct packet access. 3133 * Return 3134 * 0 on success, or a negative error in case of failure. 3135 * 3136 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 3137 * Description 3138 * Read the value of a perf event counter, and store it into *buf* 3139 * of size *buf_size*. This helper relies on a *map* of type 3140 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 3141 * counter is selected when *map* is updated with perf event file 3142 * descriptors. The *map* is an array whose size is the number of 3143 * available CPUs, and each cell contains a value relative to one 3144 * CPU. The value to retrieve is indicated by *flags*, that 3145 * contains the index of the CPU to look up, masked with 3146 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 3147 * **BPF_F_CURRENT_CPU** to indicate that the value for the 3148 * current CPU should be retrieved. 3149 * 3150 * This helper behaves in a way close to 3151 * **bpf_perf_event_read**\ () helper, save that instead of 3152 * just returning the value observed, it fills the *buf* 3153 * structure. This allows for additional data to be retrieved: in 3154 * particular, the enabled and running times (in *buf*\ 3155 * **->enabled** and *buf*\ **->running**, respectively) are 3156 * copied. In general, **bpf_perf_event_read_value**\ () is 3157 * recommended over **bpf_perf_event_read**\ (), which has some 3158 * ABI issues and provides fewer functionalities. 3159 * 3160 * These values are interesting, because hardware PMU (Performance 3161 * Monitoring Unit) counters are limited resources. When there are 3162 * more PMU based perf events opened than available counters, 3163 * kernel will multiplex these events so each event gets certain 3164 * percentage (but not all) of the PMU time. In case that 3165 * multiplexing happens, the number of samples or counter value 3166 * will not reflect the case compared to when no multiplexing 3167 * occurs. This makes comparison between different runs difficult. 3168 * Typically, the counter value should be normalized before 3169 * comparing to other experiments. The usual normalization is done 3170 * as follows. 3171 * 3172 * :: 3173 * 3174 * normalized_counter = counter * t_enabled / t_running 3175 * 3176 * Where t_enabled is the time enabled for event and t_running is 3177 * the time running for event since last normalization. The 3178 * enabled and running times are accumulated since the perf event 3179 * open. To achieve scaling factor between two invocations of an 3180 * eBPF program, users can use CPU id as the key (which is 3181 * typical for perf array usage model) to remember the previous 3182 * value and do the calculation inside the eBPF program. 3183 * Return 3184 * 0 on success, or a negative error in case of failure. 3185 * 3186 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3187 * Description 3188 * For an eBPF program attached to a perf event, retrieve the 3189 * value of the event counter associated to *ctx* and store it in 3190 * the structure pointed by *buf* and of size *buf_size*. Enabled 3191 * and running times are also stored in the structure (see 3192 * description of helper **bpf_perf_event_read_value**\ () for 3193 * more details). 3194 * Return 3195 * 0 on success, or a negative error in case of failure. 3196 * 3197 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3198 * Description 3199 * Emulate a call to **getsockopt()** on the socket associated to 3200 * *bpf_socket*, which must be a full socket. The *level* at 3201 * which the option resides and the name *optname* of the option 3202 * must be specified, see **getsockopt(2)** for more information. 3203 * The retrieved value is stored in the structure pointed by 3204 * *opval* and of length *optlen*. 3205 * 3206 * *bpf_socket* should be one of the following: 3207 * 3208 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3209 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3210 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3211 * 3212 * This helper actually implements a subset of **getsockopt()**. 3213 * It supports the same set of *optname*\ s that is supported by 3214 * the **bpf_setsockopt**\ () helper. The exceptions are 3215 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3216 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3217 * Return 3218 * 0 on success, or a negative error in case of failure. 3219 * 3220 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3221 * Description 3222 * Used for error injection, this helper uses kprobes to override 3223 * the return value of the probed function, and to set it to *rc*. 3224 * The first argument is the context *regs* on which the kprobe 3225 * works. 3226 * 3227 * This helper works by setting the PC (program counter) 3228 * to an override function which is run in place of the original 3229 * probed function. This means the probed function is not run at 3230 * all. The replacement function just returns with the required 3231 * value. 3232 * 3233 * This helper has security implications, and thus is subject to 3234 * restrictions. It is only available if the kernel was compiled 3235 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3236 * option, and in this case it only works on functions tagged with 3237 * **ALLOW_ERROR_INJECTION** in the kernel code. 3238 * Return 3239 * 0 3240 * 3241 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3242 * Description 3243 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3244 * for the full TCP socket associated to *bpf_sock_ops* to 3245 * *argval*. 3246 * 3247 * The primary use of this field is to determine if there should 3248 * be calls to eBPF programs of type 3249 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3250 * code. A program of the same type can change its value, per 3251 * connection and as necessary, when the connection is 3252 * established. This field is directly accessible for reading, but 3253 * this helper must be used for updates in order to return an 3254 * error if an eBPF program tries to set a callback that is not 3255 * supported in the current kernel. 3256 * 3257 * *argval* is a flag array which can combine these flags: 3258 * 3259 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3260 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3261 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3262 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3263 * 3264 * Therefore, this function can be used to clear a callback flag by 3265 * setting the appropriate bit to zero. e.g. to disable the RTO 3266 * callback: 3267 * 3268 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3269 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3270 * 3271 * Here are some examples of where one could call such eBPF 3272 * program: 3273 * 3274 * * When RTO fires. 3275 * * When a packet is retransmitted. 3276 * * When the connection terminates. 3277 * * When a packet is sent. 3278 * * When a packet is received. 3279 * Return 3280 * Code **-EINVAL** if the socket is not a full TCP socket; 3281 * otherwise, a positive number containing the bits that could not 3282 * be set is returned (which comes down to 0 if all bits were set 3283 * as required). 3284 * 3285 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3286 * Description 3287 * This helper is used in programs implementing policies at the 3288 * socket level. If the message *msg* is allowed to pass (i.e. if 3289 * the verdict eBPF program returns **SK_PASS**), redirect it to 3290 * the socket referenced by *map* (of type 3291 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3292 * egress interfaces can be used for redirection. The 3293 * **BPF_F_INGRESS** value in *flags* is used to make the 3294 * distinction (ingress path is selected if the flag is present, 3295 * egress path otherwise). This is the only flag supported for now. 3296 * Return 3297 * **SK_PASS** on success, or **SK_DROP** on error. 3298 * 3299 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3300 * Description 3301 * For socket policies, apply the verdict of the eBPF program to 3302 * the next *bytes* (number of bytes) of message *msg*. 3303 * 3304 * For example, this helper can be used in the following cases: 3305 * 3306 * * A single **sendmsg**\ () or **sendfile**\ () system call 3307 * contains multiple logical messages that the eBPF program is 3308 * supposed to read and for which it should apply a verdict. 3309 * * An eBPF program only cares to read the first *bytes* of a 3310 * *msg*. If the message has a large payload, then setting up 3311 * and calling the eBPF program repeatedly for all bytes, even 3312 * though the verdict is already known, would create unnecessary 3313 * overhead. 3314 * 3315 * When called from within an eBPF program, the helper sets a 3316 * counter internal to the BPF infrastructure, that is used to 3317 * apply the last verdict to the next *bytes*. If *bytes* is 3318 * smaller than the current data being processed from a 3319 * **sendmsg**\ () or **sendfile**\ () system call, the first 3320 * *bytes* will be sent and the eBPF program will be re-run with 3321 * the pointer for start of data pointing to byte number *bytes* 3322 * **+ 1**. If *bytes* is larger than the current data being 3323 * processed, then the eBPF verdict will be applied to multiple 3324 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3325 * consumed. 3326 * 3327 * Note that if a socket closes with the internal counter holding 3328 * a non-zero value, this is not a problem because data is not 3329 * being buffered for *bytes* and is sent as it is received. 3330 * Return 3331 * 0 3332 * 3333 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3334 * Description 3335 * For socket policies, prevent the execution of the verdict eBPF 3336 * program for message *msg* until *bytes* (byte number) have been 3337 * accumulated. 3338 * 3339 * This can be used when one needs a specific number of bytes 3340 * before a verdict can be assigned, even if the data spans 3341 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3342 * case would be a user calling **sendmsg**\ () repeatedly with 3343 * 1-byte long message segments. Obviously, this is bad for 3344 * performance, but it is still valid. If the eBPF program needs 3345 * *bytes* bytes to validate a header, this helper can be used to 3346 * prevent the eBPF program to be called again until *bytes* have 3347 * been accumulated. 3348 * Return 3349 * 0 3350 * 3351 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3352 * Description 3353 * For socket policies, pull in non-linear data from user space 3354 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3355 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3356 * respectively. 3357 * 3358 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3359 * *msg* it can only parse data that the (**data**, **data_end**) 3360 * pointers have already consumed. For **sendmsg**\ () hooks this 3361 * is likely the first scatterlist element. But for calls relying 3362 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3363 * be the range (**0**, **0**) because the data is shared with 3364 * user space and by default the objective is to avoid allowing 3365 * user space to modify data while (or after) eBPF verdict is 3366 * being decided. This helper can be used to pull in data and to 3367 * set the start and end pointer to given values. Data will be 3368 * copied if necessary (i.e. if data was not linear and if start 3369 * and end pointers do not point to the same chunk). 3370 * 3371 * A call to this helper is susceptible to change the underlying 3372 * packet buffer. Therefore, at load time, all checks on pointers 3373 * previously done by the verifier are invalidated and must be 3374 * performed again, if the helper is used in combination with 3375 * direct packet access. 3376 * 3377 * All values for *flags* are reserved for future usage, and must 3378 * be left at zero. 3379 * Return 3380 * 0 on success, or a negative error in case of failure. 3381 * 3382 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3383 * Description 3384 * Bind the socket associated to *ctx* to the address pointed by 3385 * *addr*, of length *addr_len*. This allows for making outgoing 3386 * connection from the desired IP address, which can be useful for 3387 * example when all processes inside a cgroup should use one 3388 * single IP address on a host that has multiple IP configured. 3389 * 3390 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3391 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3392 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3393 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3394 * behavior and lets the kernel efficiently pick up an unused 3395 * port as long as 4-tuple is unique. Passing non-zero port might 3396 * lead to degraded performance. 3397 * Return 3398 * 0 on success, or a negative error in case of failure. 3399 * 3400 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3401 * Description 3402 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3403 * possible to both shrink and grow the packet tail. 3404 * Shrink done via *delta* being a negative integer. 3405 * 3406 * A call to this helper is susceptible to change the underlying 3407 * packet buffer. Therefore, at load time, all checks on pointers 3408 * previously done by the verifier are invalidated and must be 3409 * performed again, if the helper is used in combination with 3410 * direct packet access. 3411 * Return 3412 * 0 on success, or a negative error in case of failure. 3413 * 3414 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3415 * Description 3416 * Retrieve the XFRM state (IP transform framework, see also 3417 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3418 * 3419 * The retrieved value is stored in the **struct bpf_xfrm_state** 3420 * pointed by *xfrm_state* and of length *size*. 3421 * 3422 * All values for *flags* are reserved for future usage, and must 3423 * be left at zero. 3424 * 3425 * This helper is available only if the kernel was compiled with 3426 * **CONFIG_XFRM** configuration option. 3427 * Return 3428 * 0 on success, or a negative error in case of failure. 3429 * 3430 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3431 * Description 3432 * Return a user or a kernel stack in bpf program provided buffer. 3433 * To achieve this, the helper needs *ctx*, which is a pointer 3434 * to the context on which the tracing program is executed. 3435 * To store the stacktrace, the bpf program provides *buf* with 3436 * a nonnegative *size*. 3437 * 3438 * The last argument, *flags*, holds the number of stack frames to 3439 * skip (from 0 to 255), masked with 3440 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3441 * the following flags: 3442 * 3443 * **BPF_F_USER_STACK** 3444 * Collect a user space stack instead of a kernel stack. 3445 * **BPF_F_USER_BUILD_ID** 3446 * Collect (build_id, file_offset) instead of ips for user 3447 * stack, only valid if **BPF_F_USER_STACK** is also 3448 * specified. 3449 * 3450 * *file_offset* is an offset relative to the beginning 3451 * of the executable or shared object file backing the vma 3452 * which the *ip* falls in. It is *not* an offset relative 3453 * to that object's base address. Accordingly, it must be 3454 * adjusted by adding (sh_addr - sh_offset), where 3455 * sh_{addr,offset} correspond to the executable section 3456 * containing *file_offset* in the object, for comparisons 3457 * to symbols' st_value to be valid. 3458 * 3459 * **bpf_get_stack**\ () can collect up to 3460 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3461 * to sufficient large buffer size. Note that 3462 * this limit can be controlled with the **sysctl** program, and 3463 * that it should be manually increased in order to profile long 3464 * user stacks (such as stacks for Java programs). To do so, use: 3465 * 3466 * :: 3467 * 3468 * # sysctl kernel.perf_event_max_stack=<new value> 3469 * Return 3470 * The non-negative copied *buf* length equal to or less than 3471 * *size* on success, or a negative error in case of failure. 3472 * 3473 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3474 * Description 3475 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3476 * it provides an easy way to load *len* bytes from *offset* 3477 * from the packet associated to *skb*, into the buffer pointed 3478 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3479 * a fifth argument *start_header* exists in order to select a 3480 * base offset to start from. *start_header* can be one of: 3481 * 3482 * **BPF_HDR_START_MAC** 3483 * Base offset to load data from is *skb*'s mac header. 3484 * **BPF_HDR_START_NET** 3485 * Base offset to load data from is *skb*'s network header. 3486 * 3487 * In general, "direct packet access" is the preferred method to 3488 * access packet data, however, this helper is in particular useful 3489 * in socket filters where *skb*\ **->data** does not always point 3490 * to the start of the mac header and where "direct packet access" 3491 * is not available. 3492 * Return 3493 * 0 on success, or a negative error in case of failure. 3494 * 3495 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3496 * Description 3497 * Do FIB lookup in kernel tables using parameters in *params*. 3498 * If lookup is successful and result shows packet is to be 3499 * forwarded, the neighbor tables are searched for the nexthop. 3500 * If successful (ie., FIB lookup shows forwarding and nexthop 3501 * is resolved), the nexthop address is returned in ipv4_dst 3502 * or ipv6_dst based on family, smac is set to mac address of 3503 * egress device, dmac is set to nexthop mac address, rt_metric 3504 * is set to metric from route (IPv4/IPv6 only), and ifindex 3505 * is set to the device index of the nexthop from the FIB lookup. 3506 * 3507 * *plen* argument is the size of the passed in struct. 3508 * *flags* argument can be a combination of one or more of the 3509 * following values: 3510 * 3511 * **BPF_FIB_LOOKUP_DIRECT** 3512 * Do a direct table lookup vs full lookup using FIB 3513 * rules. 3514 * **BPF_FIB_LOOKUP_TBID** 3515 * Used with BPF_FIB_LOOKUP_DIRECT. 3516 * Use the routing table ID present in *params*->tbid 3517 * for the fib lookup. 3518 * **BPF_FIB_LOOKUP_OUTPUT** 3519 * Perform lookup from an egress perspective (default is 3520 * ingress). 3521 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3522 * Skip the neighbour table lookup. *params*->dmac 3523 * and *params*->smac will not be set as output. A common 3524 * use case is to call **bpf_redirect_neigh**\ () after 3525 * doing **bpf_fib_lookup**\ (). 3526 * **BPF_FIB_LOOKUP_SRC** 3527 * Derive and set source IP addr in *params*->ipv{4,6}_src 3528 * for the nexthop. If the src addr cannot be derived, 3529 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3530 * case, *params*->dmac and *params*->smac are not set either. 3531 * **BPF_FIB_LOOKUP_MARK** 3532 * Use the mark present in *params*->mark for the fib lookup. 3533 * This option should not be used with BPF_FIB_LOOKUP_DIRECT, 3534 * as it only has meaning for full lookups. 3535 * 3536 * *ctx* is either **struct xdp_md** for XDP programs or 3537 * **struct sk_buff** tc cls_act programs. 3538 * Return 3539 * * < 0 if any input argument is invalid 3540 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3541 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3542 * packet is not forwarded or needs assist from full stack 3543 * 3544 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3545 * was exceeded and output params->mtu_result contains the MTU. 3546 * 3547 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3548 * Description 3549 * Add an entry to, or update a sockhash *map* referencing sockets. 3550 * The *skops* is used as a new value for the entry associated to 3551 * *key*. *flags* is one of: 3552 * 3553 * **BPF_NOEXIST** 3554 * The entry for *key* must not exist in the map. 3555 * **BPF_EXIST** 3556 * The entry for *key* must already exist in the map. 3557 * **BPF_ANY** 3558 * No condition on the existence of the entry for *key*. 3559 * 3560 * If the *map* has eBPF programs (parser and verdict), those will 3561 * be inherited by the socket being added. If the socket is 3562 * already attached to eBPF programs, this results in an error. 3563 * Return 3564 * 0 on success, or a negative error in case of failure. 3565 * 3566 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3567 * Description 3568 * This helper is used in programs implementing policies at the 3569 * socket level. If the message *msg* is allowed to pass (i.e. if 3570 * the verdict eBPF program returns **SK_PASS**), redirect it to 3571 * the socket referenced by *map* (of type 3572 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3573 * egress interfaces can be used for redirection. The 3574 * **BPF_F_INGRESS** value in *flags* is used to make the 3575 * distinction (ingress path is selected if the flag is present, 3576 * egress path otherwise). This is the only flag supported for now. 3577 * Return 3578 * **SK_PASS** on success, or **SK_DROP** on error. 3579 * 3580 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3581 * Description 3582 * This helper is used in programs implementing policies at the 3583 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3584 * if the verdict eBPF program returns **SK_PASS**), redirect it 3585 * to the socket referenced by *map* (of type 3586 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3587 * egress interfaces can be used for redirection. The 3588 * **BPF_F_INGRESS** value in *flags* is used to make the 3589 * distinction (ingress path is selected if the flag is present, 3590 * egress otherwise). This is the only flag supported for now. 3591 * Return 3592 * **SK_PASS** on success, or **SK_DROP** on error. 3593 * 3594 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3595 * Description 3596 * Encapsulate the packet associated to *skb* within a Layer 3 3597 * protocol header. This header is provided in the buffer at 3598 * address *hdr*, with *len* its size in bytes. *type* indicates 3599 * the protocol of the header and can be one of: 3600 * 3601 * **BPF_LWT_ENCAP_SEG6** 3602 * IPv6 encapsulation with Segment Routing Header 3603 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3604 * the IPv6 header is computed by the kernel. 3605 * **BPF_LWT_ENCAP_SEG6_INLINE** 3606 * Only works if *skb* contains an IPv6 packet. Insert a 3607 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3608 * the IPv6 header. 3609 * **BPF_LWT_ENCAP_IP** 3610 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3611 * must be IPv4 or IPv6, followed by zero or more 3612 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3613 * total bytes in all prepended headers. Please note that 3614 * if **skb_is_gso**\ (*skb*) is true, no more than two 3615 * headers can be prepended, and the inner header, if 3616 * present, should be either GRE or UDP/GUE. 3617 * 3618 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3619 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3620 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3621 * **BPF_PROG_TYPE_LWT_XMIT**. 3622 * 3623 * A call to this helper is susceptible to change the underlying 3624 * packet buffer. Therefore, at load time, all checks on pointers 3625 * previously done by the verifier are invalidated and must be 3626 * performed again, if the helper is used in combination with 3627 * direct packet access. 3628 * Return 3629 * 0 on success, or a negative error in case of failure. 3630 * 3631 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3632 * Description 3633 * Store *len* bytes from address *from* into the packet 3634 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3635 * inside the outermost IPv6 Segment Routing Header can be 3636 * modified through this helper. 3637 * 3638 * A call to this helper is susceptible to change the underlying 3639 * packet buffer. Therefore, at load time, all checks on pointers 3640 * previously done by the verifier are invalidated and must be 3641 * performed again, if the helper is used in combination with 3642 * direct packet access. 3643 * Return 3644 * 0 on success, or a negative error in case of failure. 3645 * 3646 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3647 * Description 3648 * Adjust the size allocated to TLVs in the outermost IPv6 3649 * Segment Routing Header contained in the packet associated to 3650 * *skb*, at position *offset* by *delta* bytes. Only offsets 3651 * after the segments are accepted. *delta* can be as well 3652 * positive (growing) as negative (shrinking). 3653 * 3654 * A call to this helper is susceptible to change the underlying 3655 * packet buffer. Therefore, at load time, all checks on pointers 3656 * previously done by the verifier are invalidated and must be 3657 * performed again, if the helper is used in combination with 3658 * direct packet access. 3659 * Return 3660 * 0 on success, or a negative error in case of failure. 3661 * 3662 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3663 * Description 3664 * Apply an IPv6 Segment Routing action of type *action* to the 3665 * packet associated to *skb*. Each action takes a parameter 3666 * contained at address *param*, and of length *param_len* bytes. 3667 * *action* can be one of: 3668 * 3669 * **SEG6_LOCAL_ACTION_END_X** 3670 * End.X action: Endpoint with Layer-3 cross-connect. 3671 * Type of *param*: **struct in6_addr**. 3672 * **SEG6_LOCAL_ACTION_END_T** 3673 * End.T action: Endpoint with specific IPv6 table lookup. 3674 * Type of *param*: **int**. 3675 * **SEG6_LOCAL_ACTION_END_B6** 3676 * End.B6 action: Endpoint bound to an SRv6 policy. 3677 * Type of *param*: **struct ipv6_sr_hdr**. 3678 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3679 * End.B6.Encap action: Endpoint bound to an SRv6 3680 * encapsulation policy. 3681 * Type of *param*: **struct ipv6_sr_hdr**. 3682 * 3683 * A call to this helper is susceptible to change the underlying 3684 * packet buffer. Therefore, at load time, all checks on pointers 3685 * previously done by the verifier are invalidated and must be 3686 * performed again, if the helper is used in combination with 3687 * direct packet access. 3688 * Return 3689 * 0 on success, or a negative error in case of failure. 3690 * 3691 * long bpf_rc_repeat(void *ctx) 3692 * Description 3693 * This helper is used in programs implementing IR decoding, to 3694 * report a successfully decoded repeat key message. This delays 3695 * the generation of a key up event for previously generated 3696 * key down event. 3697 * 3698 * Some IR protocols like NEC have a special IR message for 3699 * repeating last button, for when a button is held down. 3700 * 3701 * The *ctx* should point to the lirc sample as passed into 3702 * the program. 3703 * 3704 * This helper is only available is the kernel was compiled with 3705 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3706 * "**y**". 3707 * Return 3708 * 0 3709 * 3710 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3711 * Description 3712 * This helper is used in programs implementing IR decoding, to 3713 * report a successfully decoded key press with *scancode*, 3714 * *toggle* value in the given *protocol*. The scancode will be 3715 * translated to a keycode using the rc keymap, and reported as 3716 * an input key down event. After a period a key up event is 3717 * generated. This period can be extended by calling either 3718 * **bpf_rc_keydown**\ () again with the same values, or calling 3719 * **bpf_rc_repeat**\ (). 3720 * 3721 * Some protocols include a toggle bit, in case the button was 3722 * released and pressed again between consecutive scancodes. 3723 * 3724 * The *ctx* should point to the lirc sample as passed into 3725 * the program. 3726 * 3727 * The *protocol* is the decoded protocol number (see 3728 * **enum rc_proto** for some predefined values). 3729 * 3730 * This helper is only available is the kernel was compiled with 3731 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3732 * "**y**". 3733 * Return 3734 * 0 3735 * 3736 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3737 * Description 3738 * Return the cgroup v2 id of the socket associated with the *skb*. 3739 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3740 * helper for cgroup v1 by providing a tag resp. identifier that 3741 * can be matched on or used for map lookups e.g. to implement 3742 * policy. The cgroup v2 id of a given path in the hierarchy is 3743 * exposed in user space through the f_handle API in order to get 3744 * to the same 64-bit id. 3745 * 3746 * This helper can be used on TC egress path, but not on ingress, 3747 * and is available only if the kernel was compiled with the 3748 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3749 * Return 3750 * The id is returned or 0 in case the id could not be retrieved. 3751 * 3752 * u64 bpf_get_current_cgroup_id(void) 3753 * Description 3754 * Get the current cgroup id based on the cgroup within which 3755 * the current task is running. 3756 * Return 3757 * A 64-bit integer containing the current cgroup id based 3758 * on the cgroup within which the current task is running. 3759 * 3760 * void *bpf_get_local_storage(void *map, u64 flags) 3761 * Description 3762 * Get the pointer to the local storage area. 3763 * The type and the size of the local storage is defined 3764 * by the *map* argument. 3765 * The *flags* meaning is specific for each map type, 3766 * and has to be 0 for cgroup local storage. 3767 * 3768 * Depending on the BPF program type, a local storage area 3769 * can be shared between multiple instances of the BPF program, 3770 * running simultaneously. 3771 * 3772 * A user should care about the synchronization by himself. 3773 * For example, by using the **BPF_ATOMIC** instructions to alter 3774 * the shared data. 3775 * Return 3776 * A pointer to the local storage area. 3777 * 3778 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3779 * Description 3780 * Select a **SO_REUSEPORT** socket from a 3781 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3782 * It checks the selected socket is matching the incoming 3783 * request in the socket buffer. 3784 * Return 3785 * 0 on success, or a negative error in case of failure. 3786 * 3787 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3788 * Description 3789 * Return id of cgroup v2 that is ancestor of cgroup associated 3790 * with the *skb* at the *ancestor_level*. The root cgroup is at 3791 * *ancestor_level* zero and each step down the hierarchy 3792 * increments the level. If *ancestor_level* == level of cgroup 3793 * associated with *skb*, then return value will be same as that 3794 * of **bpf_skb_cgroup_id**\ (). 3795 * 3796 * The helper is useful to implement policies based on cgroups 3797 * that are upper in hierarchy than immediate cgroup associated 3798 * with *skb*. 3799 * 3800 * The format of returned id and helper limitations are same as in 3801 * **bpf_skb_cgroup_id**\ (). 3802 * Return 3803 * The id is returned or 0 in case the id could not be retrieved. 3804 * 3805 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3806 * Description 3807 * Look for TCP socket matching *tuple*, optionally in a child 3808 * network namespace *netns*. The return value must be checked, 3809 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3810 * 3811 * The *ctx* should point to the context of the program, such as 3812 * the skb or socket (depending on the hook in use). This is used 3813 * to determine the base network namespace for the lookup. 3814 * 3815 * *tuple_size* must be one of: 3816 * 3817 * **sizeof**\ (*tuple*\ **->ipv4**) 3818 * Look for an IPv4 socket. 3819 * **sizeof**\ (*tuple*\ **->ipv6**) 3820 * Look for an IPv6 socket. 3821 * 3822 * If the *netns* is a negative signed 32-bit integer, then the 3823 * socket lookup table in the netns associated with the *ctx* 3824 * will be used. For the TC hooks, this is the netns of the device 3825 * in the skb. For socket hooks, this is the netns of the socket. 3826 * If *netns* is any other signed 32-bit value greater than or 3827 * equal to zero then it specifies the ID of the netns relative to 3828 * the netns associated with the *ctx*. *netns* values beyond the 3829 * range of 32-bit integers are reserved for future use. 3830 * 3831 * All values for *flags* are reserved for future usage, and must 3832 * be left at zero. 3833 * 3834 * This helper is available only if the kernel was compiled with 3835 * **CONFIG_NET** configuration option. 3836 * Return 3837 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3838 * For sockets with reuseport option, the **struct bpf_sock** 3839 * result is from *reuse*\ **->socks**\ [] using the hash of the 3840 * tuple. 3841 * 3842 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3843 * Description 3844 * Look for UDP socket matching *tuple*, optionally in a child 3845 * network namespace *netns*. The return value must be checked, 3846 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3847 * 3848 * The *ctx* should point to the context of the program, such as 3849 * the skb or socket (depending on the hook in use). This is used 3850 * to determine the base network namespace for the lookup. 3851 * 3852 * *tuple_size* must be one of: 3853 * 3854 * **sizeof**\ (*tuple*\ **->ipv4**) 3855 * Look for an IPv4 socket. 3856 * **sizeof**\ (*tuple*\ **->ipv6**) 3857 * Look for an IPv6 socket. 3858 * 3859 * If the *netns* is a negative signed 32-bit integer, then the 3860 * socket lookup table in the netns associated with the *ctx* 3861 * will be used. For the TC hooks, this is the netns of the device 3862 * in the skb. For socket hooks, this is the netns of the socket. 3863 * If *netns* is any other signed 32-bit value greater than or 3864 * equal to zero then it specifies the ID of the netns relative to 3865 * the netns associated with the *ctx*. *netns* values beyond the 3866 * range of 32-bit integers are reserved for future use. 3867 * 3868 * All values for *flags* are reserved for future usage, and must 3869 * be left at zero. 3870 * 3871 * This helper is available only if the kernel was compiled with 3872 * **CONFIG_NET** configuration option. 3873 * Return 3874 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3875 * For sockets with reuseport option, the **struct bpf_sock** 3876 * result is from *reuse*\ **->socks**\ [] using the hash of the 3877 * tuple. 3878 * 3879 * long bpf_sk_release(void *sock) 3880 * Description 3881 * Release the reference held by *sock*. *sock* must be a 3882 * non-**NULL** pointer that was returned from 3883 * **bpf_sk_lookup_xxx**\ (). 3884 * Return 3885 * 0 on success, or a negative error in case of failure. 3886 * 3887 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3888 * Description 3889 * Push an element *value* in *map*. *flags* is one of: 3890 * 3891 * **BPF_EXIST** 3892 * If the queue/stack is full, the oldest element is 3893 * removed to make room for this. 3894 * Return 3895 * 0 on success, or a negative error in case of failure. 3896 * 3897 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3898 * Description 3899 * Pop an element from *map*. 3900 * Return 3901 * 0 on success, or a negative error in case of failure. 3902 * 3903 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3904 * Description 3905 * Get an element from *map* without removing it. 3906 * Return 3907 * 0 on success, or a negative error in case of failure. 3908 * 3909 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3910 * Description 3911 * For socket policies, insert *len* bytes into *msg* at offset 3912 * *start*. 3913 * 3914 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3915 * *msg* it may want to insert metadata or options into the *msg*. 3916 * This can later be read and used by any of the lower layer BPF 3917 * hooks. 3918 * 3919 * This helper may fail if under memory pressure (a malloc 3920 * fails) in these cases BPF programs will get an appropriate 3921 * error and BPF programs will need to handle them. 3922 * Return 3923 * 0 on success, or a negative error in case of failure. 3924 * 3925 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3926 * Description 3927 * Will remove *len* bytes from a *msg* starting at byte *start*. 3928 * This may result in **ENOMEM** errors under certain situations if 3929 * an allocation and copy are required due to a full ring buffer. 3930 * However, the helper will try to avoid doing the allocation 3931 * if possible. Other errors can occur if input parameters are 3932 * invalid either due to *start* byte not being valid part of *msg* 3933 * payload and/or *pop* value being to large. 3934 * Return 3935 * 0 on success, or a negative error in case of failure. 3936 * 3937 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3938 * Description 3939 * This helper is used in programs implementing IR decoding, to 3940 * report a successfully decoded pointer movement. 3941 * 3942 * The *ctx* should point to the lirc sample as passed into 3943 * the program. 3944 * 3945 * This helper is only available is the kernel was compiled with 3946 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3947 * "**y**". 3948 * Return 3949 * 0 3950 * 3951 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3952 * Description 3953 * Acquire a spinlock represented by the pointer *lock*, which is 3954 * stored as part of a value of a map. Taking the lock allows to 3955 * safely update the rest of the fields in that value. The 3956 * spinlock can (and must) later be released with a call to 3957 * **bpf_spin_unlock**\ (\ *lock*\ ). 3958 * 3959 * Spinlocks in BPF programs come with a number of restrictions 3960 * and constraints: 3961 * 3962 * * **bpf_spin_lock** objects are only allowed inside maps of 3963 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3964 * list could be extended in the future). 3965 * * BTF description of the map is mandatory. 3966 * * The BPF program can take ONE lock at a time, since taking two 3967 * or more could cause dead locks. 3968 * * Only one **struct bpf_spin_lock** is allowed per map element. 3969 * * When the lock is taken, calls (either BPF to BPF or helpers) 3970 * are not allowed. 3971 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3972 * allowed inside a spinlock-ed region. 3973 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3974 * the lock, on all execution paths, before it returns. 3975 * * The BPF program can access **struct bpf_spin_lock** only via 3976 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3977 * helpers. Loading or storing data into the **struct 3978 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3979 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3980 * of the map value must be a struct and have **struct 3981 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3982 * Nested lock inside another struct is not allowed. 3983 * * The **struct bpf_spin_lock** *lock* field in a map value must 3984 * be aligned on a multiple of 4 bytes in that value. 3985 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3986 * the **bpf_spin_lock** field to user space. 3987 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3988 * a BPF program, do not update the **bpf_spin_lock** field. 3989 * * **bpf_spin_lock** cannot be on the stack or inside a 3990 * networking packet (it can only be inside of a map values). 3991 * * **bpf_spin_lock** is available to root only. 3992 * * Tracing programs and socket filter programs cannot use 3993 * **bpf_spin_lock**\ () due to insufficient preemption checks 3994 * (but this may change in the future). 3995 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3996 * Return 3997 * 0 3998 * 3999 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 4000 * Description 4001 * Release the *lock* previously locked by a call to 4002 * **bpf_spin_lock**\ (\ *lock*\ ). 4003 * Return 4004 * 0 4005 * 4006 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 4007 * Description 4008 * This helper gets a **struct bpf_sock** pointer such 4009 * that all the fields in this **bpf_sock** can be accessed. 4010 * Return 4011 * A **struct bpf_sock** pointer on success, or **NULL** in 4012 * case of failure. 4013 * 4014 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 4015 * Description 4016 * This helper gets a **struct bpf_tcp_sock** pointer from a 4017 * **struct bpf_sock** pointer. 4018 * Return 4019 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 4020 * case of failure. 4021 * 4022 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 4023 * Description 4024 * Set ECN (Explicit Congestion Notification) field of IP header 4025 * to **CE** (Congestion Encountered) if current value is **ECT** 4026 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 4027 * and IPv4. 4028 * Return 4029 * 1 if the **CE** flag is set (either by the current helper call 4030 * or because it was already present), 0 if it is not set. 4031 * 4032 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 4033 * Description 4034 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 4035 * **bpf_sk_release**\ () is unnecessary and not allowed. 4036 * Return 4037 * A **struct bpf_sock** pointer on success, or **NULL** in 4038 * case of failure. 4039 * 4040 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 4041 * Description 4042 * Look for TCP socket matching *tuple*, optionally in a child 4043 * network namespace *netns*. The return value must be checked, 4044 * and if non-**NULL**, released via **bpf_sk_release**\ (). 4045 * 4046 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 4047 * that it also returns timewait or request sockets. Use 4048 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 4049 * full structure. 4050 * 4051 * This helper is available only if the kernel was compiled with 4052 * **CONFIG_NET** configuration option. 4053 * Return 4054 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 4055 * For sockets with reuseport option, the **struct bpf_sock** 4056 * result is from *reuse*\ **->socks**\ [] using the hash of the 4057 * tuple. 4058 * 4059 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4060 * Description 4061 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 4062 * the listening socket in *sk*. 4063 * 4064 * *iph* points to the start of the IPv4 or IPv6 header, while 4065 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4066 * **sizeof**\ (**struct ipv6hdr**). 4067 * 4068 * *th* points to the start of the TCP header, while *th_len* 4069 * contains the length of the TCP header (at least 4070 * **sizeof**\ (**struct tcphdr**)). 4071 * Return 4072 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 4073 * error otherwise. 4074 * 4075 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 4076 * Description 4077 * Get name of sysctl in /proc/sys/ and copy it into provided by 4078 * program buffer *buf* of size *buf_len*. 4079 * 4080 * The buffer is always NUL terminated, unless it's zero-sized. 4081 * 4082 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 4083 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 4084 * only (e.g. "tcp_mem"). 4085 * Return 4086 * Number of character copied (not including the trailing NUL). 4087 * 4088 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 4089 * truncated name in this case). 4090 * 4091 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 4092 * Description 4093 * Get current value of sysctl as it is presented in /proc/sys 4094 * (incl. newline, etc), and copy it as a string into provided 4095 * by program buffer *buf* of size *buf_len*. 4096 * 4097 * The whole value is copied, no matter what file position user 4098 * space issued e.g. sys_read at. 4099 * 4100 * The buffer is always NUL terminated, unless it's zero-sized. 4101 * Return 4102 * Number of character copied (not including the trailing NUL). 4103 * 4104 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 4105 * truncated name in this case). 4106 * 4107 * **-EINVAL** if current value was unavailable, e.g. because 4108 * sysctl is uninitialized and read returns -EIO for it. 4109 * 4110 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 4111 * Description 4112 * Get new value being written by user space to sysctl (before 4113 * the actual write happens) and copy it as a string into 4114 * provided by program buffer *buf* of size *buf_len*. 4115 * 4116 * User space may write new value at file position > 0. 4117 * 4118 * The buffer is always NUL terminated, unless it's zero-sized. 4119 * Return 4120 * Number of character copied (not including the trailing NUL). 4121 * 4122 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 4123 * truncated name in this case). 4124 * 4125 * **-EINVAL** if sysctl is being read. 4126 * 4127 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 4128 * Description 4129 * Override new value being written by user space to sysctl with 4130 * value provided by program in buffer *buf* of size *buf_len*. 4131 * 4132 * *buf* should contain a string in same form as provided by user 4133 * space on sysctl write. 4134 * 4135 * User space may write new value at file position > 0. To override 4136 * the whole sysctl value file position should be set to zero. 4137 * Return 4138 * 0 on success. 4139 * 4140 * **-E2BIG** if the *buf_len* is too big. 4141 * 4142 * **-EINVAL** if sysctl is being read. 4143 * 4144 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 4145 * Description 4146 * Convert the initial part of the string from buffer *buf* of 4147 * size *buf_len* to a long integer according to the given base 4148 * and save the result in *res*. 4149 * 4150 * The string may begin with an arbitrary amount of white space 4151 * (as determined by **isspace**\ (3)) followed by a single 4152 * optional '**-**' sign. 4153 * 4154 * Five least significant bits of *flags* encode base, other bits 4155 * are currently unused. 4156 * 4157 * Base must be either 8, 10, 16 or 0 to detect it automatically 4158 * similar to user space **strtol**\ (3). 4159 * Return 4160 * Number of characters consumed on success. Must be positive but 4161 * no more than *buf_len*. 4162 * 4163 * **-EINVAL** if no valid digits were found or unsupported base 4164 * was provided. 4165 * 4166 * **-ERANGE** if resulting value was out of range. 4167 * 4168 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 4169 * Description 4170 * Convert the initial part of the string from buffer *buf* of 4171 * size *buf_len* to an unsigned long integer according to the 4172 * given base and save the result in *res*. 4173 * 4174 * The string may begin with an arbitrary amount of white space 4175 * (as determined by **isspace**\ (3)). 4176 * 4177 * Five least significant bits of *flags* encode base, other bits 4178 * are currently unused. 4179 * 4180 * Base must be either 8, 10, 16 or 0 to detect it automatically 4181 * similar to user space **strtoul**\ (3). 4182 * Return 4183 * Number of characters consumed on success. Must be positive but 4184 * no more than *buf_len*. 4185 * 4186 * **-EINVAL** if no valid digits were found or unsupported base 4187 * was provided. 4188 * 4189 * **-ERANGE** if resulting value was out of range. 4190 * 4191 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4192 * Description 4193 * Get a bpf-local-storage from a *sk*. 4194 * 4195 * Logically, it could be thought of getting the value from 4196 * a *map* with *sk* as the **key**. From this 4197 * perspective, the usage is not much different from 4198 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4199 * helper enforces the key must be a full socket and the map must 4200 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4201 * 4202 * Underneath, the value is stored locally at *sk* instead of 4203 * the *map*. The *map* is used as the bpf-local-storage 4204 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4205 * searched against all bpf-local-storages residing at *sk*. 4206 * 4207 * *sk* is a kernel **struct sock** pointer for LSM program. 4208 * *sk* is a **struct bpf_sock** pointer for other program types. 4209 * 4210 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4211 * used such that a new bpf-local-storage will be 4212 * created if one does not exist. *value* can be used 4213 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4214 * the initial value of a bpf-local-storage. If *value* is 4215 * **NULL**, the new bpf-local-storage will be zero initialized. 4216 * Return 4217 * A bpf-local-storage pointer is returned on success. 4218 * 4219 * **NULL** if not found or there was an error in adding 4220 * a new bpf-local-storage. 4221 * 4222 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4223 * Description 4224 * Delete a bpf-local-storage from a *sk*. 4225 * Return 4226 * 0 on success. 4227 * 4228 * **-ENOENT** if the bpf-local-storage cannot be found. 4229 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4230 * 4231 * long bpf_send_signal(u32 sig) 4232 * Description 4233 * Send signal *sig* to the process of the current task. 4234 * The signal may be delivered to any of this process's threads. 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 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4247 * Description 4248 * Try to issue a SYN cookie for the packet with corresponding 4249 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4250 * 4251 * *iph* points to the start of the IPv4 or IPv6 header, while 4252 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4253 * **sizeof**\ (**struct ipv6hdr**). 4254 * 4255 * *th* points to the start of the TCP header, while *th_len* 4256 * contains the length of the TCP header with options (at least 4257 * **sizeof**\ (**struct tcphdr**)). 4258 * Return 4259 * On success, lower 32 bits hold the generated SYN cookie in 4260 * followed by 16 bits which hold the MSS value for that cookie, 4261 * and the top 16 bits are unused. 4262 * 4263 * On failure, the returned value is one of the following: 4264 * 4265 * **-EINVAL** SYN cookie cannot be issued due to error 4266 * 4267 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4268 * 4269 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4270 * 4271 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4272 * 4273 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4274 * Description 4275 * Write raw *data* blob into a special BPF perf event held by 4276 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4277 * event must have the following attributes: **PERF_SAMPLE_RAW** 4278 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4279 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4280 * 4281 * The *flags* are used to indicate the index in *map* for which 4282 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4283 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4284 * to indicate that the index of the current CPU core should be 4285 * used. 4286 * 4287 * The value to write, of *size*, is passed through eBPF stack and 4288 * pointed by *data*. 4289 * 4290 * *ctx* is a pointer to in-kernel struct sk_buff. 4291 * 4292 * This helper is similar to **bpf_perf_event_output**\ () but 4293 * restricted to raw_tracepoint bpf programs. 4294 * Return 4295 * 0 on success, or a negative error in case of failure. 4296 * 4297 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4298 * Description 4299 * Safely attempt to read *size* bytes from user space address 4300 * *unsafe_ptr* and store the data in *dst*. 4301 * Return 4302 * 0 on success, or a negative error in case of failure. 4303 * 4304 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4305 * Description 4306 * Safely attempt to read *size* bytes from kernel space address 4307 * *unsafe_ptr* and store the data in *dst*. 4308 * Return 4309 * 0 on success, or a negative error in case of failure. 4310 * 4311 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4312 * Description 4313 * Copy a NUL terminated string from an unsafe user address 4314 * *unsafe_ptr* to *dst*. The *size* should include the 4315 * terminating NUL byte. In case the string length is smaller than 4316 * *size*, the target is not padded with further NUL bytes. If the 4317 * string length is larger than *size*, just *size*-1 bytes are 4318 * copied and the last byte is set to NUL. 4319 * 4320 * On success, returns the number of bytes that were written, 4321 * including the terminal NUL. This makes this helper useful in 4322 * tracing programs for reading strings, and more importantly to 4323 * get its length at runtime. See the following snippet: 4324 * 4325 * :: 4326 * 4327 * SEC("kprobe/sys_open") 4328 * void bpf_sys_open(struct pt_regs *ctx) 4329 * { 4330 * char buf[PATHLEN]; // PATHLEN is defined to 256 4331 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4332 * ctx->di); 4333 * 4334 * // Consume buf, for example push it to 4335 * // userspace via bpf_perf_event_output(); we 4336 * // can use res (the string length) as event 4337 * // size, after checking its boundaries. 4338 * } 4339 * 4340 * In comparison, using **bpf_probe_read_user**\ () helper here 4341 * instead to read the string would require to estimate the length 4342 * at compile time, and would often result in copying more memory 4343 * than necessary. 4344 * 4345 * Another useful use case is when parsing individual process 4346 * arguments or individual environment variables navigating 4347 * *current*\ **->mm->arg_start** and *current*\ 4348 * **->mm->env_start**: using this helper and the return value, 4349 * one can quickly iterate at the right offset of the memory area. 4350 * Return 4351 * On success, the strictly positive length of the output string, 4352 * including the trailing NUL character. On error, a negative 4353 * value. 4354 * 4355 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4356 * Description 4357 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4358 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4359 * Return 4360 * On success, the strictly positive length of the string, including 4361 * the trailing NUL character. On error, a negative value. 4362 * 4363 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4364 * Description 4365 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4366 * *rcv_nxt* is the ack_seq to be sent out. 4367 * Return 4368 * 0 on success, or a negative error in case of failure. 4369 * 4370 * long bpf_send_signal_thread(u32 sig) 4371 * Description 4372 * Send signal *sig* to the thread corresponding to the current task. 4373 * Return 4374 * 0 on success or successfully queued. 4375 * 4376 * **-EBUSY** if work queue under nmi is full. 4377 * 4378 * **-EINVAL** if *sig* is invalid. 4379 * 4380 * **-EPERM** if no permission to send the *sig*. 4381 * 4382 * **-EAGAIN** if bpf program can try again. 4383 * 4384 * u64 bpf_jiffies64(void) 4385 * Description 4386 * Obtain the 64bit jiffies 4387 * Return 4388 * The 64 bit jiffies 4389 * 4390 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4391 * Description 4392 * For an eBPF program attached to a perf event, retrieve the 4393 * branch records (**struct perf_branch_entry**) associated to *ctx* 4394 * and store it in the buffer pointed by *buf* up to size 4395 * *size* bytes. 4396 * Return 4397 * On success, number of bytes written to *buf*. On error, a 4398 * negative value. 4399 * 4400 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4401 * instead return the number of bytes required to store all the 4402 * branch entries. If this flag is set, *buf* may be NULL. 4403 * 4404 * **-EINVAL** if arguments invalid or **size** not a multiple 4405 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4406 * 4407 * **-ENOENT** if architecture does not support branch records. 4408 * 4409 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4410 * Description 4411 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4412 * *namespace* will be returned in *nsdata*. 4413 * Return 4414 * 0 on success, or one of the following in case of failure: 4415 * 4416 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4417 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4418 * 4419 * **-ENOENT** if pidns does not exists for the current task. 4420 * 4421 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4422 * Description 4423 * Write raw *data* blob into a special BPF perf event held by 4424 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4425 * event must have the following attributes: **PERF_SAMPLE_RAW** 4426 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4427 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4428 * 4429 * The *flags* are used to indicate the index in *map* for which 4430 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4431 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4432 * to indicate that the index of the current CPU core should be 4433 * used. 4434 * 4435 * The value to write, of *size*, is passed through eBPF stack and 4436 * pointed by *data*. 4437 * 4438 * *ctx* is a pointer to in-kernel struct xdp_buff. 4439 * 4440 * This helper is similar to **bpf_perf_eventoutput**\ () but 4441 * restricted to raw_tracepoint bpf programs. 4442 * Return 4443 * 0 on success, or a negative error in case of failure. 4444 * 4445 * u64 bpf_get_netns_cookie(void *ctx) 4446 * Description 4447 * Retrieve the cookie (generated by the kernel) of the network 4448 * namespace the input *ctx* is associated with. The network 4449 * namespace cookie remains stable for its lifetime and provides 4450 * a global identifier that can be assumed unique. If *ctx* is 4451 * NULL, then the helper returns the cookie for the initial 4452 * network namespace. The cookie itself is very similar to that 4453 * of **bpf_get_socket_cookie**\ () helper, but for network 4454 * namespaces instead of sockets. 4455 * Return 4456 * A 8-byte long opaque number. 4457 * 4458 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4459 * Description 4460 * Return id of cgroup v2 that is ancestor of the cgroup associated 4461 * with the current task at the *ancestor_level*. The root cgroup 4462 * is at *ancestor_level* zero and each step down the hierarchy 4463 * increments the level. If *ancestor_level* == level of cgroup 4464 * associated with the current task, then return value will be the 4465 * same as that of **bpf_get_current_cgroup_id**\ (). 4466 * 4467 * The helper is useful to implement policies based on cgroups 4468 * that are upper in hierarchy than immediate cgroup associated 4469 * with the current task. 4470 * 4471 * The format of returned id and helper limitations are same as in 4472 * **bpf_get_current_cgroup_id**\ (). 4473 * Return 4474 * The id is returned or 0 in case the id could not be retrieved. 4475 * 4476 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4477 * Description 4478 * Helper is overloaded depending on BPF program type. This 4479 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4480 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4481 * 4482 * Assign the *sk* to the *skb*. When combined with appropriate 4483 * routing configuration to receive the packet towards the socket, 4484 * will cause *skb* to be delivered to the specified socket. 4485 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4486 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4487 * interfere with successful delivery to the socket. 4488 * 4489 * This operation is only valid from TC ingress path. 4490 * 4491 * The *flags* argument must be zero. 4492 * Return 4493 * 0 on success, or a negative error in case of failure: 4494 * 4495 * **-EINVAL** if specified *flags* are not supported. 4496 * 4497 * **-ENOENT** if the socket is unavailable for assignment. 4498 * 4499 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4500 * 4501 * **-EOPNOTSUPP** if the operation is not supported, for example 4502 * a call from outside of TC ingress. 4503 * 4504 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4505 * Description 4506 * Helper is overloaded depending on BPF program type. This 4507 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4508 * 4509 * Select the *sk* as a result of a socket lookup. 4510 * 4511 * For the operation to succeed passed socket must be compatible 4512 * with the packet description provided by the *ctx* object. 4513 * 4514 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4515 * be an exact match. While IP family (**AF_INET** or 4516 * **AF_INET6**) must be compatible, that is IPv6 sockets 4517 * that are not v6-only can be selected for IPv4 packets. 4518 * 4519 * Only TCP listeners and UDP unconnected sockets can be 4520 * selected. *sk* can also be NULL to reset any previous 4521 * selection. 4522 * 4523 * *flags* argument can combination of following values: 4524 * 4525 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4526 * socket selection, potentially done by a BPF program 4527 * that ran before us. 4528 * 4529 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4530 * load-balancing within reuseport group for the socket 4531 * being selected. 4532 * 4533 * On success *ctx->sk* will point to the selected socket. 4534 * 4535 * Return 4536 * 0 on success, or a negative errno in case of failure. 4537 * 4538 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4539 * not compatible with packet family (*ctx->family*). 4540 * 4541 * * **-EEXIST** if socket has been already selected, 4542 * potentially by another program, and 4543 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4544 * 4545 * * **-EINVAL** if unsupported flags were specified. 4546 * 4547 * * **-EPROTOTYPE** if socket L4 protocol 4548 * (*sk->protocol*) doesn't match packet protocol 4549 * (*ctx->protocol*). 4550 * 4551 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4552 * state (TCP listening or UDP unconnected). 4553 * 4554 * u64 bpf_ktime_get_boot_ns(void) 4555 * Description 4556 * Return the time elapsed since system boot, in nanoseconds. 4557 * Does include the time the system was suspended. 4558 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4559 * Return 4560 * Current *ktime*. 4561 * 4562 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4563 * Description 4564 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4565 * out the format string. 4566 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4567 * the format string itself. The *data* and *data_len* are format string 4568 * arguments. The *data* are a **u64** array and corresponding format string 4569 * values are stored in the array. For strings and pointers where pointees 4570 * are accessed, only the pointer values are stored in the *data* array. 4571 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4572 * 4573 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4574 * Reading kernel memory may fail due to either invalid address or 4575 * valid address but requiring a major memory fault. If reading kernel memory 4576 * fails, the string for **%s** will be an empty string, and the ip 4577 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4578 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4579 * Return 4580 * 0 on success, or a negative error in case of failure: 4581 * 4582 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4583 * by returning 1 from bpf program. 4584 * 4585 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4586 * 4587 * **-E2BIG** if *fmt* contains too many format specifiers. 4588 * 4589 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4590 * 4591 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4592 * Description 4593 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4594 * The *m* represents the seq_file. The *data* and *len* represent the 4595 * data to write in bytes. 4596 * Return 4597 * 0 on success, or a negative error in case of failure: 4598 * 4599 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4600 * 4601 * u64 bpf_sk_cgroup_id(void *sk) 4602 * Description 4603 * Return the cgroup v2 id of the socket *sk*. 4604 * 4605 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4606 * returned from **bpf_sk_lookup_xxx**\ (), 4607 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4608 * same as in **bpf_skb_cgroup_id**\ (). 4609 * 4610 * This helper is available only if the kernel was compiled with 4611 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4612 * Return 4613 * The id is returned or 0 in case the id could not be retrieved. 4614 * 4615 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4616 * Description 4617 * Return id of cgroup v2 that is ancestor of cgroup associated 4618 * with the *sk* at the *ancestor_level*. The root cgroup is at 4619 * *ancestor_level* zero and each step down the hierarchy 4620 * increments the level. If *ancestor_level* == level of cgroup 4621 * associated with *sk*, then return value will be same as that 4622 * of **bpf_sk_cgroup_id**\ (). 4623 * 4624 * The helper is useful to implement policies based on cgroups 4625 * that are upper in hierarchy than immediate cgroup associated 4626 * with *sk*. 4627 * 4628 * The format of returned id and helper limitations are same as in 4629 * **bpf_sk_cgroup_id**\ (). 4630 * Return 4631 * The id is returned or 0 in case the id could not be retrieved. 4632 * 4633 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4634 * Description 4635 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4636 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4637 * of new data availability is sent. 4638 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4639 * of new data availability is sent unconditionally. 4640 * If **0** is specified in *flags*, an adaptive notification 4641 * of new data availability is sent. 4642 * 4643 * An adaptive notification is a notification sent whenever the user-space 4644 * process has caught up and consumed all available payloads. In case the user-space 4645 * process is still processing a previous payload, then no notification is needed 4646 * as it will process the newly added payload automatically. 4647 * Return 4648 * 0 on success, or a negative error in case of failure. 4649 * 4650 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4651 * Description 4652 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4653 * *flags* must be 0. 4654 * Return 4655 * Valid pointer with *size* bytes of memory available; NULL, 4656 * otherwise. 4657 * 4658 * void bpf_ringbuf_submit(void *data, u64 flags) 4659 * Description 4660 * Submit reserved ring buffer sample, pointed to by *data*. 4661 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4662 * of new data availability is sent. 4663 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4664 * of new data availability is sent unconditionally. 4665 * If **0** is specified in *flags*, an adaptive notification 4666 * of new data availability is sent. 4667 * 4668 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4669 * Return 4670 * Nothing. Always succeeds. 4671 * 4672 * void bpf_ringbuf_discard(void *data, u64 flags) 4673 * Description 4674 * Discard reserved ring buffer sample, pointed to by *data*. 4675 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4676 * of new data availability is sent. Discarded records remain in 4677 * the ring buffer until consumed by user space, so a later submit 4678 * using adaptive wakeup might not wake up the consumer. 4679 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4680 * of new data availability is sent unconditionally. 4681 * If **0** is specified in *flags*, an adaptive notification 4682 * of new data availability is sent. 4683 * 4684 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4685 * Return 4686 * Nothing. Always succeeds. 4687 * 4688 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4689 * Description 4690 * Query various characteristics of provided ring buffer. What 4691 * exactly is queries is determined by *flags*: 4692 * 4693 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4694 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4695 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4696 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4697 * 4698 * Data returned is just a momentary snapshot of actual values 4699 * and could be inaccurate, so this facility should be used to 4700 * power heuristics and for reporting, not to make 100% correct 4701 * calculation. 4702 * Return 4703 * Requested value, or 0, if *flags* are not recognized. 4704 * 4705 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4706 * Description 4707 * Change the skbs checksum level by one layer up or down, or 4708 * reset it entirely to none in order to have the stack perform 4709 * checksum validation. The level is applicable to the following 4710 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4711 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4712 * through **bpf_skb_adjust_room**\ () helper with passing in 4713 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4714 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4715 * the UDP header is removed. Similarly, an encap of the latter 4716 * into the former could be accompanied by a helper call to 4717 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4718 * skb is still intended to be processed in higher layers of the 4719 * stack instead of just egressing at tc. 4720 * 4721 * There are three supported level settings at this time: 4722 * 4723 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4724 * with CHECKSUM_UNNECESSARY. 4725 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4726 * with CHECKSUM_UNNECESSARY. 4727 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4728 * sets CHECKSUM_NONE to force checksum validation by the stack. 4729 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4730 * skb->csum_level. 4731 * Return 4732 * 0 on success, or a negative error in case of failure. In the 4733 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4734 * is returned or the error code -EACCES in case the skb is not 4735 * subject to CHECKSUM_UNNECESSARY. 4736 * 4737 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4738 * Description 4739 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4740 * Return 4741 * *sk* if casting is valid, or **NULL** otherwise. 4742 * 4743 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4744 * Description 4745 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4746 * Return 4747 * *sk* if casting is valid, or **NULL** otherwise. 4748 * 4749 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4750 * Description 4751 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4752 * Return 4753 * *sk* if casting is valid, or **NULL** otherwise. 4754 * 4755 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4756 * Description 4757 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4758 * Return 4759 * *sk* if casting is valid, or **NULL** otherwise. 4760 * 4761 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4762 * Description 4763 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4764 * Return 4765 * *sk* if casting is valid, or **NULL** otherwise. 4766 * 4767 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4768 * Description 4769 * Return a user or a kernel stack in bpf program provided buffer. 4770 * Note: the user stack will only be populated if the *task* is 4771 * the current task; all other tasks will return -EOPNOTSUPP. 4772 * To achieve this, the helper needs *task*, which is a valid 4773 * pointer to **struct task_struct**. To store the stacktrace, the 4774 * bpf program provides *buf* with a nonnegative *size*. 4775 * 4776 * The last argument, *flags*, holds the number of stack frames to 4777 * skip (from 0 to 255), masked with 4778 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4779 * the following flags: 4780 * 4781 * **BPF_F_USER_STACK** 4782 * Collect a user space stack instead of a kernel stack. 4783 * The *task* must be the current task. 4784 * **BPF_F_USER_BUILD_ID** 4785 * Collect buildid+offset instead of ips for user stack, 4786 * only valid if **BPF_F_USER_STACK** is also specified. 4787 * 4788 * **bpf_get_task_stack**\ () can collect up to 4789 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4790 * to sufficient large buffer size. Note that 4791 * this limit can be controlled with the **sysctl** program, and 4792 * that it should be manually increased in order to profile long 4793 * user stacks (such as stacks for Java programs). To do so, use: 4794 * 4795 * :: 4796 * 4797 * # sysctl kernel.perf_event_max_stack=<new value> 4798 * Return 4799 * The non-negative copied *buf* length equal to or less than 4800 * *size* on success, or a negative error in case of failure. 4801 * 4802 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4803 * Description 4804 * Load header option. Support reading a particular TCP header 4805 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4806 * 4807 * If *flags* is 0, it will search the option from the 4808 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4809 * has details on what skb_data contains under different 4810 * *skops*\ **->op**. 4811 * 4812 * The first byte of the *searchby_res* specifies the 4813 * kind that it wants to search. 4814 * 4815 * If the searching kind is an experimental kind 4816 * (i.e. 253 or 254 according to RFC6994). It also 4817 * needs to specify the "magic" which is either 4818 * 2 bytes or 4 bytes. It then also needs to 4819 * specify the size of the magic by using 4820 * the 2nd byte which is "kind-length" of a TCP 4821 * header option and the "kind-length" also 4822 * includes the first 2 bytes "kind" and "kind-length" 4823 * itself as a normal TCP header option also does. 4824 * 4825 * For example, to search experimental kind 254 with 4826 * 2 byte magic 0xeB9F, the searchby_res should be 4827 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4828 * 4829 * To search for the standard window scale option (3), 4830 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4831 * Note, kind-length must be 0 for regular option. 4832 * 4833 * Searching for No-Op (0) and End-of-Option-List (1) are 4834 * not supported. 4835 * 4836 * *len* must be at least 2 bytes which is the minimal size 4837 * of a header option. 4838 * 4839 * Supported flags: 4840 * 4841 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4842 * saved_syn packet or the just-received syn packet. 4843 * 4844 * Return 4845 * > 0 when found, the header option is copied to *searchby_res*. 4846 * The return value is the total length copied. On failure, a 4847 * negative error code is returned: 4848 * 4849 * **-EINVAL** if a parameter is invalid. 4850 * 4851 * **-ENOMSG** if the option is not found. 4852 * 4853 * **-ENOENT** if no syn packet is available when 4854 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4855 * 4856 * **-ENOSPC** if there is not enough space. Only *len* number of 4857 * bytes are copied. 4858 * 4859 * **-EFAULT** on failure to parse the header options in the 4860 * packet. 4861 * 4862 * **-EPERM** if the helper cannot be used under the current 4863 * *skops*\ **->op**. 4864 * 4865 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4866 * Description 4867 * Store header option. The data will be copied 4868 * from buffer *from* with length *len* to the TCP header. 4869 * 4870 * The buffer *from* should have the whole option that 4871 * includes the kind, kind-length, and the actual 4872 * option data. The *len* must be at least kind-length 4873 * long. The kind-length does not have to be 4 byte 4874 * aligned. The kernel will take care of the padding 4875 * and setting the 4 bytes aligned value to th->doff. 4876 * 4877 * This helper will check for duplicated option 4878 * by searching the same option in the outgoing skb. 4879 * 4880 * This helper can only be called during 4881 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4882 * 4883 * Return 4884 * 0 on success, or negative error in case of failure: 4885 * 4886 * **-EINVAL** If param is invalid. 4887 * 4888 * **-ENOSPC** if there is not enough space in the header. 4889 * Nothing has been written 4890 * 4891 * **-EEXIST** if the option already exists. 4892 * 4893 * **-EFAULT** on failure to parse the existing header options. 4894 * 4895 * **-EPERM** if the helper cannot be used under the current 4896 * *skops*\ **->op**. 4897 * 4898 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4899 * Description 4900 * Reserve *len* bytes for the bpf header option. The 4901 * space will be used by **bpf_store_hdr_opt**\ () later in 4902 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4903 * 4904 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4905 * the total number of bytes will be reserved. 4906 * 4907 * This helper can only be called during 4908 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4909 * 4910 * Return 4911 * 0 on success, or negative error in case of failure: 4912 * 4913 * **-EINVAL** if a parameter is invalid. 4914 * 4915 * **-ENOSPC** if there is not enough space in the header. 4916 * 4917 * **-EPERM** if the helper cannot be used under the current 4918 * *skops*\ **->op**. 4919 * 4920 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4921 * Description 4922 * Get a bpf_local_storage from an *inode*. 4923 * 4924 * Logically, it could be thought of as getting the value from 4925 * a *map* with *inode* as the **key**. From this 4926 * perspective, the usage is not much different from 4927 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4928 * helper enforces the key must be an inode and the map must also 4929 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4930 * 4931 * Underneath, the value is stored locally at *inode* instead of 4932 * the *map*. The *map* is used as the bpf-local-storage 4933 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4934 * searched against all bpf_local_storage residing at *inode*. 4935 * 4936 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4937 * used such that a new bpf_local_storage will be 4938 * created if one does not exist. *value* can be used 4939 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4940 * the initial value of a bpf_local_storage. If *value* is 4941 * **NULL**, the new bpf_local_storage will be zero initialized. 4942 * Return 4943 * A bpf_local_storage pointer is returned on success. 4944 * 4945 * **NULL** if not found or there was an error in adding 4946 * a new bpf_local_storage. 4947 * 4948 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4949 * Description 4950 * Delete a bpf_local_storage from an *inode*. 4951 * Return 4952 * 0 on success. 4953 * 4954 * **-ENOENT** if the bpf_local_storage cannot be found. 4955 * 4956 * long bpf_d_path(const struct path *path, char *buf, u32 sz) 4957 * Description 4958 * Return full path for given **struct path** object, which 4959 * needs to be the kernel BTF *path* object. The path is 4960 * returned in the provided buffer *buf* of size *sz* and 4961 * is zero terminated. 4962 * 4963 * Return 4964 * On success, the strictly positive length of the string, 4965 * including the trailing NUL character. On error, a negative 4966 * value. 4967 * 4968 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4969 * Description 4970 * Read *size* bytes from user space address *user_ptr* and store 4971 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4972 * Return 4973 * 0 on success, or a negative error in case of failure. 4974 * 4975 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4976 * Description 4977 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4978 * using *ptr*->type_id. This value should specify the type 4979 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4980 * can be used to look up vmlinux BTF type ids. Traversing the 4981 * data structure using BTF, the type information and values are 4982 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4983 * the pointer data is carried out to avoid kernel crashes during 4984 * operation. Smaller types can use string space on the stack; 4985 * larger programs can use map data to store the string 4986 * representation. 4987 * 4988 * The string can be subsequently shared with userspace via 4989 * bpf_perf_event_output() or ring buffer interfaces. 4990 * bpf_trace_printk() is to be avoided as it places too small 4991 * a limit on string size to be useful. 4992 * 4993 * *flags* is a combination of 4994 * 4995 * **BTF_F_COMPACT** 4996 * no formatting around type information 4997 * **BTF_F_NONAME** 4998 * no struct/union member names/types 4999 * **BTF_F_PTR_RAW** 5000 * show raw (unobfuscated) pointer values; 5001 * equivalent to printk specifier %px. 5002 * **BTF_F_ZERO** 5003 * show zero-valued struct/union members; they 5004 * are not displayed by default 5005 * 5006 * Return 5007 * The number of bytes that were written (or would have been 5008 * written if output had to be truncated due to string size), 5009 * or a negative error in cases of failure. 5010 * 5011 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 5012 * Description 5013 * Use BTF to write to seq_write a string representation of 5014 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 5015 * *flags* are identical to those used for bpf_snprintf_btf. 5016 * Return 5017 * 0 on success or a negative error in case of failure. 5018 * 5019 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 5020 * Description 5021 * See **bpf_get_cgroup_classid**\ () for the main description. 5022 * This helper differs from **bpf_get_cgroup_classid**\ () in that 5023 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 5024 * associated socket instead of the current process. 5025 * Return 5026 * The id is returned or 0 in case the id could not be retrieved. 5027 * 5028 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 5029 * Description 5030 * Redirect the packet to another net device of index *ifindex* 5031 * and fill in L2 addresses from neighboring subsystem. This helper 5032 * is somewhat similar to **bpf_redirect**\ (), except that it 5033 * populates L2 addresses as well, meaning, internally, the helper 5034 * relies on the neighbor lookup for the L2 address of the nexthop. 5035 * 5036 * The helper will perform a FIB lookup based on the skb's 5037 * networking header to get the address of the next hop, unless 5038 * this is supplied by the caller in the *params* argument. The 5039 * *plen* argument indicates the len of *params* and should be set 5040 * to 0 if *params* is NULL. 5041 * 5042 * The *flags* argument is reserved and must be 0. The helper is 5043 * currently only supported for tc BPF program types, and enabled 5044 * for IPv4 and IPv6 protocols. 5045 * Return 5046 * The helper returns **TC_ACT_REDIRECT** on success or 5047 * **TC_ACT_SHOT** on error. 5048 * 5049 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 5050 * Description 5051 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 5052 * pointer to the percpu kernel variable on *cpu*. A ksym is an 5053 * extern variable decorated with '__ksym'. For ksym, there is a 5054 * global var (either static or global) defined of the same name 5055 * in the kernel. The ksym is percpu if the global var is percpu. 5056 * The returned pointer points to the global percpu var on *cpu*. 5057 * 5058 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 5059 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 5060 * happens if *cpu* is larger than nr_cpu_ids. The caller of 5061 * bpf_per_cpu_ptr() must check the returned value. 5062 * Return 5063 * A pointer pointing to the kernel percpu variable on *cpu*, or 5064 * NULL, if *cpu* is invalid. 5065 * 5066 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 5067 * Description 5068 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 5069 * pointer to the percpu kernel variable on this cpu. See the 5070 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 5071 * 5072 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 5073 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 5074 * never return NULL. 5075 * Return 5076 * A pointer pointing to the kernel percpu variable on this cpu. 5077 * 5078 * long bpf_redirect_peer(u32 ifindex, u64 flags) 5079 * Description 5080 * Redirect the packet to another net device of index *ifindex*. 5081 * This helper is somewhat similar to **bpf_redirect**\ (), except 5082 * that the redirection happens to the *ifindex*' peer device and 5083 * the netns switch takes place from ingress to ingress without 5084 * going through the CPU's backlog queue. 5085 * 5086 * *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during 5087 * the netns switch. 5088 * 5089 * The *flags* argument is reserved and must be 0. The helper is 5090 * currently only supported for tc BPF program types at the 5091 * ingress hook and for veth and netkit target device types. The 5092 * peer device must reside in a different network namespace. 5093 * Return 5094 * The helper returns **TC_ACT_REDIRECT** on success or 5095 * **TC_ACT_SHOT** on error. 5096 * 5097 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 5098 * Description 5099 * Get a bpf_local_storage from the *task*. 5100 * 5101 * Logically, it could be thought of as getting the value from 5102 * a *map* with *task* as the **key**. From this 5103 * perspective, the usage is not much different from 5104 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 5105 * helper enforces the key must be a task_struct and the map must also 5106 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 5107 * 5108 * Underneath, the value is stored locally at *task* instead of 5109 * the *map*. The *map* is used as the bpf-local-storage 5110 * "type". The bpf-local-storage "type" (i.e. the *map*) is 5111 * searched against all bpf_local_storage residing at *task*. 5112 * 5113 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5114 * used such that a new bpf_local_storage will be 5115 * created if one does not exist. *value* can be used 5116 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5117 * the initial value of a bpf_local_storage. If *value* is 5118 * **NULL**, the new bpf_local_storage will be zero initialized. 5119 * Return 5120 * A bpf_local_storage pointer is returned on success. 5121 * 5122 * **NULL** if not found or there was an error in adding 5123 * a new bpf_local_storage. 5124 * 5125 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 5126 * Description 5127 * Delete a bpf_local_storage from a *task*. 5128 * Return 5129 * 0 on success. 5130 * 5131 * **-ENOENT** if the bpf_local_storage cannot be found. 5132 * 5133 * struct task_struct *bpf_get_current_task_btf(void) 5134 * Description 5135 * Return a BTF pointer to the "current" task. 5136 * This pointer can also be used in helpers that accept an 5137 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 5138 * Return 5139 * Pointer to the current task. 5140 * 5141 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 5142 * Description 5143 * Set or clear certain options on *bprm*: 5144 * 5145 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 5146 * which sets the **AT_SECURE** auxv for glibc. The bit 5147 * is cleared if the flag is not specified. 5148 * Return 5149 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 5150 * 5151 * u64 bpf_ktime_get_coarse_ns(void) 5152 * Description 5153 * Return a coarse-grained version of the time elapsed since 5154 * system boot, in nanoseconds. Does not include time the system 5155 * was suspended. 5156 * 5157 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 5158 * Return 5159 * Current *ktime*. 5160 * 5161 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 5162 * Description 5163 * Returns the stored IMA hash of the *inode* (if it's available). 5164 * If the hash is larger than *size*, then only *size* 5165 * bytes will be copied to *dst* 5166 * Return 5167 * The **hash_algo** is returned on success, 5168 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if 5169 * invalid arguments are passed. 5170 * 5171 * struct socket *bpf_sock_from_file(struct file *file) 5172 * Description 5173 * If the given file represents a socket, returns the associated 5174 * socket. 5175 * Return 5176 * A pointer to a struct socket on success or NULL if the file is 5177 * not a socket. 5178 * 5179 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 5180 * Description 5181 * Check packet size against exceeding MTU of net device (based 5182 * on *ifindex*). This helper will likely be used in combination 5183 * with helpers that adjust/change the packet size. 5184 * 5185 * The argument *len_diff* can be used for querying with a planned 5186 * size change. This allows to check MTU prior to changing packet 5187 * ctx. Providing a *len_diff* adjustment that is larger than the 5188 * actual packet size (resulting in negative packet size) will in 5189 * principle not exceed the MTU, which is why it is not considered 5190 * a failure. Other BPF helpers are needed for performing the 5191 * planned size change; therefore the responsibility for catching 5192 * a negative packet size belongs in those helpers. 5193 * 5194 * Specifying *ifindex* zero means the MTU check is performed 5195 * against the current net device. This is practical if this isn't 5196 * used prior to redirect. 5197 * 5198 * On input *mtu_len* must be a valid pointer, else verifier will 5199 * reject BPF program. If the value *mtu_len* is initialized to 5200 * zero then the ctx packet size is use. When value *mtu_len* is 5201 * provided as input this specify the L3 length that the MTU check 5202 * is done against. Remember XDP and TC length operate at L2, but 5203 * this value is L3 as this correlate to MTU and IP-header tot_len 5204 * values which are L3 (similar behavior as bpf_fib_lookup). 5205 * 5206 * The Linux kernel route table can configure MTUs on a more 5207 * specific per route level, which is not provided by this helper. 5208 * For route level MTU checks use the **bpf_fib_lookup**\ () 5209 * helper. 5210 * 5211 * *ctx* is either **struct xdp_md** for XDP programs or 5212 * **struct sk_buff** for tc cls_act programs. 5213 * 5214 * The *flags* argument can be a combination of one or more of the 5215 * following values: 5216 * 5217 * **BPF_MTU_CHK_SEGS** 5218 * This flag will only works for *ctx* **struct sk_buff**. 5219 * If packet context contains extra packet segment buffers 5220 * (often knows as GSO skb), then MTU check is harder to 5221 * check at this point, because in transmit path it is 5222 * possible for the skb packet to get re-segmented 5223 * (depending on net device features). This could still be 5224 * a MTU violation, so this flag enables performing MTU 5225 * check against segments, with a different violation 5226 * return code to tell it apart. Check cannot use len_diff. 5227 * 5228 * On return *mtu_len* pointer contains the MTU value of the net 5229 * device. Remember the net device configured MTU is the L3 size, 5230 * which is returned here and XDP and TC length operate at L2. 5231 * Helper take this into account for you, but remember when using 5232 * MTU value in your BPF-code. 5233 * 5234 * Return 5235 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5236 * 5237 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5238 * 5239 * MTU violations return positive values, but also populate MTU 5240 * value in *mtu_len* pointer, as this can be needed for 5241 * implementing PMTU handing: 5242 * 5243 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5244 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5245 * 5246 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5247 * Description 5248 * For each element in **map**, call **callback_fn** function with 5249 * **map**, **callback_ctx** and other map-specific parameters. 5250 * The **callback_fn** should be a static function and 5251 * the **callback_ctx** should be a pointer to the stack. 5252 * The **flags** is used to control certain aspects of the helper. 5253 * Currently, the **flags** must be 0. 5254 * 5255 * The following are a list of supported map types and their 5256 * respective expected callback signatures: 5257 * 5258 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5259 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5260 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5261 * 5262 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5263 * 5264 * For per_cpu maps, the map_value is the value on the cpu where the 5265 * bpf_prog is running. 5266 * 5267 * If **callback_fn** return 0, the helper will continue to the next 5268 * element. If return value is 1, the helper will skip the rest of 5269 * elements and return. Other return values are not used now. 5270 * 5271 * Return 5272 * The number of traversed map elements for success, **-EINVAL** for 5273 * invalid **flags**. 5274 * 5275 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5276 * Description 5277 * Outputs a string into the **str** buffer of size **str_size** 5278 * based on a format string stored in a read-only map pointed by 5279 * **fmt**. 5280 * 5281 * Each format specifier in **fmt** corresponds to one u64 element 5282 * in the **data** array. For strings and pointers where pointees 5283 * are accessed, only the pointer values are stored in the *data* 5284 * array. The *data_len* is the size of *data* in bytes - must be 5285 * a multiple of 8. 5286 * 5287 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5288 * memory. Reading kernel memory may fail due to either invalid 5289 * address or valid address but requiring a major memory fault. If 5290 * reading kernel memory fails, the string for **%s** will be an 5291 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5292 * Not returning error to bpf program is consistent with what 5293 * **bpf_trace_printk**\ () does for now. 5294 * 5295 * Return 5296 * The strictly positive length of the formatted string, including 5297 * the trailing zero character. If the return value is greater than 5298 * **str_size**, **str** contains a truncated string, guaranteed to 5299 * be zero-terminated except when **str_size** is 0. 5300 * 5301 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5302 * 5303 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5304 * Description 5305 * Execute bpf syscall with given arguments. 5306 * Return 5307 * A syscall result. 5308 * 5309 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5310 * Description 5311 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5312 * Return 5313 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5314 * 5315 * long bpf_sys_close(u32 fd) 5316 * Description 5317 * Execute close syscall for given FD. 5318 * Return 5319 * A syscall result. 5320 * 5321 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5322 * Description 5323 * Initialize the timer. 5324 * First 4 bits of *flags* specify clockid. 5325 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5326 * All other bits of *flags* are reserved. 5327 * The verifier will reject the program if *timer* is not from 5328 * the same *map*. 5329 * Return 5330 * 0 on success. 5331 * **-EBUSY** if *timer* is already initialized. 5332 * **-EINVAL** if invalid *flags* are passed. 5333 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5334 * The user space should either hold a file descriptor to a map with timers 5335 * or pin such map in bpffs. When map is unpinned or file descriptor is 5336 * closed all timers in the map will be cancelled and freed. 5337 * 5338 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5339 * Description 5340 * Configure the timer to call *callback_fn* static function. 5341 * Return 5342 * 0 on success. 5343 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5344 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5345 * The user space should either hold a file descriptor to a map with timers 5346 * or pin such map in bpffs. When map is unpinned or file descriptor is 5347 * closed all timers in the map will be cancelled and freed. 5348 * 5349 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5350 * Description 5351 * Set timer expiration N nanoseconds from the current time. The 5352 * configured callback will be invoked in soft irq context on some cpu 5353 * and will not repeat unless another bpf_timer_start() is made. 5354 * In such case the next invocation can migrate to a different cpu. 5355 * Since struct bpf_timer is a field inside map element the map 5356 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5357 * of BPF program to make sure that callback_fn code stays valid. 5358 * When user space reference to a map reaches zero all timers 5359 * in a map are cancelled and corresponding program's refcnts are 5360 * decremented. This is done to make sure that Ctrl-C of a user 5361 * process doesn't leave any timers running. If map is pinned in 5362 * bpffs the callback_fn can re-arm itself indefinitely. 5363 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5364 * cancel and free the timer in the given map element. 5365 * The map can contain timers that invoke callback_fn-s from different 5366 * programs. The same callback_fn can serve different timers from 5367 * different maps if key/value layout matches across maps. 5368 * Every bpf_timer_set_callback() can have different callback_fn. 5369 * 5370 * *flags* can be one of: 5371 * 5372 * **BPF_F_TIMER_ABS** 5373 * Start the timer in absolute expire value instead of the 5374 * default relative one. 5375 * **BPF_F_TIMER_CPU_PIN** 5376 * Timer will be pinned to the CPU of the caller. 5377 * 5378 * Return 5379 * 0 on success. 5380 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5381 * or invalid *flags* are passed. 5382 * 5383 * long bpf_timer_cancel(struct bpf_timer *timer) 5384 * Description 5385 * Cancel the timer and wait for callback_fn to finish if it was running. 5386 * Return 5387 * 0 if the timer was not active. 5388 * 1 if the timer was active. 5389 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5390 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5391 * own timer which would have led to a deadlock otherwise. 5392 * 5393 * u64 bpf_get_func_ip(void *ctx) 5394 * Description 5395 * Get address of the traced function (for tracing and kprobe programs). 5396 * 5397 * When called for kprobe program attached as uprobe it returns 5398 * probe address for both entry and return uprobe. 5399 * 5400 * Return 5401 * Address of the traced function for kprobe. 5402 * 0 for kprobes placed within the function (not at the entry). 5403 * Address of the probe for uprobe and return uprobe. 5404 * 5405 * u64 bpf_get_attach_cookie(void *ctx) 5406 * Description 5407 * Get bpf_cookie value provided (optionally) during the program 5408 * attachment. It might be different for each individual 5409 * attachment, even if BPF program itself is the same. 5410 * Expects BPF program context *ctx* as a first argument. 5411 * 5412 * Supported for the following program types: 5413 * - kprobe/uprobe; 5414 * - tracepoint; 5415 * - perf_event. 5416 * Return 5417 * Value specified by user at BPF link creation/attachment time 5418 * or 0, if it was not specified. 5419 * 5420 * long bpf_task_pt_regs(struct task_struct *task) 5421 * Description 5422 * Get the struct pt_regs associated with **task**. 5423 * Return 5424 * A pointer to struct pt_regs. 5425 * 5426 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5427 * Description 5428 * Get branch trace from hardware engines like Intel LBR. The 5429 * hardware engine is stopped shortly after the helper is 5430 * called. Therefore, the user need to filter branch entries 5431 * based on the actual use case. To capture branch trace 5432 * before the trigger point of the BPF program, the helper 5433 * should be called at the beginning of the BPF program. 5434 * 5435 * The data is stored as struct perf_branch_entry into output 5436 * buffer *entries*. *size* is the size of *entries* in bytes. 5437 * *flags* is reserved for now and must be zero. 5438 * 5439 * Return 5440 * On success, number of bytes written to *buf*. On error, a 5441 * negative value. 5442 * 5443 * **-EINVAL** if *flags* is not zero. 5444 * 5445 * **-ENOENT** if architecture does not support branch records. 5446 * 5447 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5448 * Description 5449 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5450 * to format and can handle more format args as a result. 5451 * 5452 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5453 * Return 5454 * The number of bytes written to the buffer, or a negative error 5455 * in case of failure. 5456 * 5457 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5458 * Description 5459 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5460 * Return 5461 * *sk* if casting is valid, or **NULL** otherwise. 5462 * 5463 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5464 * Description 5465 * Get the address of a kernel symbol, returned in *res*. *res* is 5466 * set to 0 if the symbol is not found. 5467 * Return 5468 * On success, zero. On error, a negative value. 5469 * 5470 * **-EINVAL** if *flags* is not zero. 5471 * 5472 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5473 * 5474 * **-ENOENT** if symbol is not found. 5475 * 5476 * **-EPERM** if caller does not have permission to obtain kernel address. 5477 * 5478 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5479 * Description 5480 * Find vma of *task* that contains *addr*, call *callback_fn* 5481 * function with *task*, *vma*, and *callback_ctx*. 5482 * The *callback_fn* should be a static function and 5483 * the *callback_ctx* should be a pointer to the stack. 5484 * The *flags* is used to control certain aspects of the helper. 5485 * Currently, the *flags* must be 0. 5486 * 5487 * The expected callback signature is 5488 * 5489 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5490 * 5491 * Return 5492 * 0 on success. 5493 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5494 * **-EBUSY** if failed to try lock mmap_lock. 5495 * **-EINVAL** for invalid **flags**. 5496 * 5497 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5498 * Description 5499 * For **nr_loops**, call **callback_fn** function 5500 * with **callback_ctx** as the context parameter. 5501 * The **callback_fn** should be a static function and 5502 * the **callback_ctx** should be a pointer to the stack. 5503 * The **flags** is used to control certain aspects of the helper. 5504 * Currently, the **flags** must be 0. Currently, nr_loops is 5505 * limited to 1 << 23 (~8 million) loops. 5506 * 5507 * long (\*callback_fn)(u64 index, void \*ctx); 5508 * 5509 * where **index** is the current index in the loop. The index 5510 * is zero-indexed. 5511 * 5512 * If **callback_fn** returns 0, the helper will continue to the next 5513 * loop. If return value is 1, the helper will skip the rest of 5514 * the loops and return. Other return values are not used now, 5515 * and will be rejected by the verifier. 5516 * 5517 * Return 5518 * The number of loops performed, **-EINVAL** for invalid **flags**, 5519 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5520 * 5521 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5522 * Description 5523 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5524 * to be null-terminated and **s1_sz** is the maximum storage 5525 * size of **s1**. **s2** must be a read-only string. 5526 * Return 5527 * An integer less than, equal to, or greater than zero 5528 * if the first **s1_sz** bytes of **s1** is found to be 5529 * less than, to match, or be greater than **s2**. 5530 * 5531 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5532 * Description 5533 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5534 * returned in **value**. 5535 * 5536 * Return 5537 * 0 on success. 5538 * **-EINVAL** if n >= argument register count of traced function. 5539 * 5540 * long bpf_get_func_ret(void *ctx, u64 *value) 5541 * Description 5542 * Get return value of the traced function (for tracing programs) 5543 * in **value**. 5544 * 5545 * Return 5546 * 0 on success. 5547 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5548 * 5549 * long bpf_get_func_arg_cnt(void *ctx) 5550 * Description 5551 * Get number of registers of the traced function (for tracing programs) where 5552 * function arguments are stored in these registers. 5553 * 5554 * Return 5555 * The number of argument registers of the traced function. 5556 * 5557 * int bpf_get_retval(void) 5558 * Description 5559 * Get the BPF program's return value that will be returned to the upper layers. 5560 * 5561 * This helper is currently supported by cgroup programs and only by the hooks 5562 * where BPF program's return value is returned to the userspace via errno. 5563 * Return 5564 * The BPF program's return value. 5565 * 5566 * int bpf_set_retval(int retval) 5567 * Description 5568 * Set the BPF program's return value that will be returned to the upper layers. 5569 * 5570 * This helper is currently supported by cgroup programs and only by the hooks 5571 * where BPF program's return value is returned to the userspace via errno. 5572 * 5573 * Note that there is the following corner case where the program exports an error 5574 * via bpf_set_retval but signals success via 'return 1': 5575 * 5576 * bpf_set_retval(-EPERM); 5577 * return 1; 5578 * 5579 * In this case, the BPF program's return value will use helper's -EPERM. This 5580 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5581 * 5582 * Return 5583 * 0 on success, or a negative error in case of failure. 5584 * 5585 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5586 * Description 5587 * Get the total size of a given xdp buff (linear and paged area) 5588 * Return 5589 * The total size of a given xdp buffer. 5590 * 5591 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5592 * Description 5593 * This helper is provided as an easy way to load data from a 5594 * xdp buffer. It can be used to load *len* bytes from *offset* from 5595 * the frame associated to *xdp_md*, into the buffer pointed by 5596 * *buf*. 5597 * Return 5598 * 0 on success, or a negative error in case of failure. 5599 * 5600 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5601 * Description 5602 * Store *len* bytes from buffer *buf* into the frame 5603 * associated to *xdp_md*, at *offset*. 5604 * Return 5605 * 0 on success, or a negative error in case of failure. 5606 * 5607 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5608 * Description 5609 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5610 * address space, and stores the data in *dst*. *flags* is not 5611 * used yet and is provided for future extensibility. This helper 5612 * can only be used by sleepable programs. 5613 * Return 5614 * 0 on success, or a negative error in case of failure. On error 5615 * *dst* buffer is zeroed out. 5616 * 5617 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5618 * Description 5619 * Change the __sk_buff->tstamp_type to *tstamp_type* 5620 * and set *tstamp* to the __sk_buff->tstamp together. 5621 * 5622 * If there is no need to change the __sk_buff->tstamp_type, 5623 * the tstamp value can be directly written to __sk_buff->tstamp 5624 * instead. 5625 * 5626 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5627 * will be kept during bpf_redirect_*(). A non zero 5628 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5629 * *tstamp_type*. 5630 * 5631 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5632 * with a zero *tstamp*. 5633 * 5634 * Only IPv4 and IPv6 skb->protocol are supported. 5635 * 5636 * This function is most useful when it needs to set a 5637 * mono delivery time to __sk_buff->tstamp and then 5638 * bpf_redirect_*() to the egress of an iface. For example, 5639 * changing the (rcv) timestamp in __sk_buff->tstamp at 5640 * ingress to a mono delivery time and then bpf_redirect_*() 5641 * to sch_fq@phy-dev. 5642 * Return 5643 * 0 on success. 5644 * **-EINVAL** for invalid input 5645 * **-EOPNOTSUPP** for unsupported protocol 5646 * 5647 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5648 * Description 5649 * Returns a calculated IMA hash of the *file*. 5650 * If the hash is larger than *size*, then only *size* 5651 * bytes will be copied to *dst* 5652 * Return 5653 * The **hash_algo** is returned on success, 5654 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if 5655 * invalid arguments are passed. 5656 * 5657 * void *bpf_kptr_xchg(void *dst, void *ptr) 5658 * Description 5659 * Exchange kptr at pointer *dst* with *ptr*, and return the old value. 5660 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise 5661 * it must be a referenced pointer which will be released when this helper 5662 * is called. 5663 * Return 5664 * The old value of kptr (which can be NULL). The returned pointer 5665 * if not NULL, is a reference which must be released using its 5666 * corresponding release function, or moved into a BPF map before 5667 * program exit. 5668 * 5669 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5670 * Description 5671 * Perform a lookup in *percpu map* for an entry associated to 5672 * *key* on *cpu*. 5673 * Return 5674 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5675 * was found or *cpu* is invalid. 5676 * 5677 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5678 * Description 5679 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5680 * Return 5681 * *sk* if casting is valid, or **NULL** otherwise. 5682 * 5683 * long bpf_dynptr_from_mem(void *data, u64 size, u64 flags, struct bpf_dynptr *ptr) 5684 * Description 5685 * Get a dynptr to local memory *data*. 5686 * 5687 * *data* must be a ptr to a map value. 5688 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5689 * *flags* is currently unused. 5690 * Return 5691 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5692 * -EINVAL if flags is not 0. 5693 * 5694 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5695 * Description 5696 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5697 * through the dynptr interface. *flags* must be 0. 5698 * 5699 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5700 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5701 * reservation fails. This is enforced by the verifier. 5702 * Return 5703 * 0 on success, or a negative error in case of failure. 5704 * 5705 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5706 * Description 5707 * Submit reserved ring buffer sample, pointed to by *data*, 5708 * through the dynptr interface. This is a no-op if the dynptr is 5709 * invalid/null. 5710 * 5711 * For more information on *flags*, please see 5712 * 'bpf_ringbuf_submit'. 5713 * Return 5714 * Nothing. Always succeeds. 5715 * 5716 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5717 * Description 5718 * Discard reserved ring buffer sample through the dynptr 5719 * interface. This is a no-op if the dynptr is invalid/null. 5720 * 5721 * For more information on *flags*, please see 5722 * 'bpf_ringbuf_discard'. 5723 * Return 5724 * Nothing. Always succeeds. 5725 * 5726 * long bpf_dynptr_read(void *dst, u64 len, const struct bpf_dynptr *src, u64 offset, u64 flags) 5727 * Description 5728 * Read *len* bytes from *src* into *dst*, starting from *offset* 5729 * into *src*. 5730 * *flags* is currently unused. 5731 * Return 5732 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5733 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5734 * *flags* is not 0. 5735 * 5736 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u64 offset, void *src, u64 len, u64 flags) 5737 * Description 5738 * Write *len* bytes from *src* into *dst*, starting from *offset* 5739 * into *dst*. 5740 * 5741 * *flags* must be 0 except for skb-type dynptrs. 5742 * 5743 * For skb-type dynptrs: 5744 * * All data slices of the dynptr are automatically 5745 * invalidated after **bpf_dynptr_write**\ (). This is 5746 * because writing may pull the skb and change the 5747 * underlying packet buffer. 5748 * 5749 * * For *flags*, please see the flags accepted by 5750 * **bpf_skb_store_bytes**\ (). 5751 * Return 5752 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5753 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5754 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5755 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5756 * 5757 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u64 offset, u64 len) 5758 * Description 5759 * Get a pointer to the underlying dynptr data. 5760 * 5761 * *len* must be a statically known value. The returned data slice 5762 * is invalidated whenever the dynptr is invalidated. 5763 * 5764 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5765 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5766 * Return 5767 * Pointer to the underlying dynptr data, NULL if the dynptr is 5768 * read-only, if the dynptr is invalid, or if the offset and length 5769 * is out of bounds. 5770 * 5771 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5772 * Description 5773 * Try to issue a SYN cookie for the packet with corresponding 5774 * IPv4/TCP headers, *iph* and *th*, without depending on a 5775 * listening socket. 5776 * 5777 * *iph* points to the IPv4 header. 5778 * 5779 * *th* points to the start of the TCP header, while *th_len* 5780 * contains the length of the TCP header (at least 5781 * **sizeof**\ (**struct tcphdr**)). 5782 * Return 5783 * On success, lower 32 bits hold the generated SYN cookie in 5784 * followed by 16 bits which hold the MSS value for that cookie, 5785 * and the top 16 bits are unused. 5786 * 5787 * On failure, the returned value is one of the following: 5788 * 5789 * **-EINVAL** if *th_len* is invalid. 5790 * 5791 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5792 * Description 5793 * Try to issue a SYN cookie for the packet with corresponding 5794 * IPv6/TCP headers, *iph* and *th*, without depending on a 5795 * listening socket. 5796 * 5797 * *iph* points to the IPv6 header. 5798 * 5799 * *th* points to the start of the TCP header, while *th_len* 5800 * contains the length of the TCP header (at least 5801 * **sizeof**\ (**struct tcphdr**)). 5802 * Return 5803 * On success, lower 32 bits hold the generated SYN cookie in 5804 * followed by 16 bits which hold the MSS value for that cookie, 5805 * and the top 16 bits are unused. 5806 * 5807 * On failure, the returned value is one of the following: 5808 * 5809 * **-EINVAL** if *th_len* is invalid. 5810 * 5811 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5812 * 5813 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5814 * Description 5815 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5816 * without depending on a listening socket. 5817 * 5818 * *iph* points to the IPv4 header. 5819 * 5820 * *th* points to the TCP header. 5821 * Return 5822 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5823 * 5824 * On failure, the returned value is one of the following: 5825 * 5826 * **-EACCES** if the SYN cookie is not valid. 5827 * 5828 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5829 * Description 5830 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5831 * without depending on a listening socket. 5832 * 5833 * *iph* points to the IPv6 header. 5834 * 5835 * *th* points to the TCP header. 5836 * Return 5837 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5838 * 5839 * On failure, the returned value is one of the following: 5840 * 5841 * **-EACCES** if the SYN cookie is not valid. 5842 * 5843 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5844 * 5845 * u64 bpf_ktime_get_tai_ns(void) 5846 * Description 5847 * A nonsettable system-wide clock derived from wall-clock time but 5848 * ignoring leap seconds. This clock does not experience 5849 * discontinuities and backwards jumps caused by NTP inserting leap 5850 * seconds as CLOCK_REALTIME does. 5851 * 5852 * See: **clock_gettime**\ (**CLOCK_TAI**) 5853 * Return 5854 * Current *ktime*. 5855 * 5856 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5857 * Description 5858 * Drain samples from the specified user ring buffer, and invoke 5859 * the provided callback for each such sample: 5860 * 5861 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5862 * 5863 * If **callback_fn** returns 0, the helper will continue to try 5864 * and drain the next sample, up to a maximum of 5865 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5866 * the helper will skip the rest of the samples and return. Other 5867 * return values are not used now, and will be rejected by the 5868 * verifier. 5869 * Return 5870 * The number of drained samples if no error was encountered while 5871 * draining samples, or 0 if no samples were present in the ring 5872 * buffer. If a user-space producer was epoll-waiting on this map, 5873 * and at least one sample was drained, they will receive an event 5874 * notification notifying them of available space in the ring 5875 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5876 * function, no wakeup notification will be sent. If the 5877 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5878 * be sent even if no sample was drained. 5879 * 5880 * On failure, the returned value is one of the following: 5881 * 5882 * **-EBUSY** if the ring buffer is contended, and another calling 5883 * context was concurrently draining the ring buffer. 5884 * 5885 * **-EINVAL** if user-space is not properly tracking the ring 5886 * buffer due to the producer position not being aligned to 8 5887 * bytes, a sample not being aligned to 8 bytes, or the producer 5888 * position not matching the advertised length of a sample. 5889 * 5890 * **-E2BIG** if user-space has tried to publish a sample which is 5891 * larger than the size of the ring buffer, or which cannot fit 5892 * within a struct bpf_dynptr. 5893 * 5894 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5895 * Description 5896 * Get a bpf_local_storage from the *cgroup*. 5897 * 5898 * Logically, it could be thought of as getting the value from 5899 * a *map* with *cgroup* as the **key**. From this 5900 * perspective, the usage is not much different from 5901 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5902 * helper enforces the key must be a cgroup struct and the map must also 5903 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5904 * 5905 * In reality, the local-storage value is embedded directly inside of the 5906 * *cgroup* object itself, rather than being located in the 5907 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5908 * queried for some *map* on a *cgroup* object, the kernel will perform an 5909 * O(n) iteration over all of the live local-storage values for that 5910 * *cgroup* object until the local-storage value for the *map* is found. 5911 * 5912 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5913 * used such that a new bpf_local_storage will be 5914 * created if one does not exist. *value* can be used 5915 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5916 * the initial value of a bpf_local_storage. If *value* is 5917 * **NULL**, the new bpf_local_storage will be zero initialized. 5918 * Return 5919 * A bpf_local_storage pointer is returned on success. 5920 * 5921 * **NULL** if not found or there was an error in adding 5922 * a new bpf_local_storage. 5923 * 5924 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5925 * Description 5926 * Delete a bpf_local_storage from a *cgroup*. 5927 * Return 5928 * 0 on success. 5929 * 5930 * **-ENOENT** if the bpf_local_storage cannot be found. 5931 */ 5932 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5933 FN(unspec, 0, ##ctx) \ 5934 FN(map_lookup_elem, 1, ##ctx) \ 5935 FN(map_update_elem, 2, ##ctx) \ 5936 FN(map_delete_elem, 3, ##ctx) \ 5937 FN(probe_read, 4, ##ctx) \ 5938 FN(ktime_get_ns, 5, ##ctx) \ 5939 FN(trace_printk, 6, ##ctx) \ 5940 FN(get_prandom_u32, 7, ##ctx) \ 5941 FN(get_smp_processor_id, 8, ##ctx) \ 5942 FN(skb_store_bytes, 9, ##ctx) \ 5943 FN(l3_csum_replace, 10, ##ctx) \ 5944 FN(l4_csum_replace, 11, ##ctx) \ 5945 FN(tail_call, 12, ##ctx) \ 5946 FN(clone_redirect, 13, ##ctx) \ 5947 FN(get_current_pid_tgid, 14, ##ctx) \ 5948 FN(get_current_uid_gid, 15, ##ctx) \ 5949 FN(get_current_comm, 16, ##ctx) \ 5950 FN(get_cgroup_classid, 17, ##ctx) \ 5951 FN(skb_vlan_push, 18, ##ctx) \ 5952 FN(skb_vlan_pop, 19, ##ctx) \ 5953 FN(skb_get_tunnel_key, 20, ##ctx) \ 5954 FN(skb_set_tunnel_key, 21, ##ctx) \ 5955 FN(perf_event_read, 22, ##ctx) \ 5956 FN(redirect, 23, ##ctx) \ 5957 FN(get_route_realm, 24, ##ctx) \ 5958 FN(perf_event_output, 25, ##ctx) \ 5959 FN(skb_load_bytes, 26, ##ctx) \ 5960 FN(get_stackid, 27, ##ctx) \ 5961 FN(csum_diff, 28, ##ctx) \ 5962 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5963 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5964 FN(skb_change_proto, 31, ##ctx) \ 5965 FN(skb_change_type, 32, ##ctx) \ 5966 FN(skb_under_cgroup, 33, ##ctx) \ 5967 FN(get_hash_recalc, 34, ##ctx) \ 5968 FN(get_current_task, 35, ##ctx) \ 5969 FN(probe_write_user, 36, ##ctx) \ 5970 FN(current_task_under_cgroup, 37, ##ctx) \ 5971 FN(skb_change_tail, 38, ##ctx) \ 5972 FN(skb_pull_data, 39, ##ctx) \ 5973 FN(csum_update, 40, ##ctx) \ 5974 FN(set_hash_invalid, 41, ##ctx) \ 5975 FN(get_numa_node_id, 42, ##ctx) \ 5976 FN(skb_change_head, 43, ##ctx) \ 5977 FN(xdp_adjust_head, 44, ##ctx) \ 5978 FN(probe_read_str, 45, ##ctx) \ 5979 FN(get_socket_cookie, 46, ##ctx) \ 5980 FN(get_socket_uid, 47, ##ctx) \ 5981 FN(set_hash, 48, ##ctx) \ 5982 FN(setsockopt, 49, ##ctx) \ 5983 FN(skb_adjust_room, 50, ##ctx) \ 5984 FN(redirect_map, 51, ##ctx) \ 5985 FN(sk_redirect_map, 52, ##ctx) \ 5986 FN(sock_map_update, 53, ##ctx) \ 5987 FN(xdp_adjust_meta, 54, ##ctx) \ 5988 FN(perf_event_read_value, 55, ##ctx) \ 5989 FN(perf_prog_read_value, 56, ##ctx) \ 5990 FN(getsockopt, 57, ##ctx) \ 5991 FN(override_return, 58, ##ctx) \ 5992 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5993 FN(msg_redirect_map, 60, ##ctx) \ 5994 FN(msg_apply_bytes, 61, ##ctx) \ 5995 FN(msg_cork_bytes, 62, ##ctx) \ 5996 FN(msg_pull_data, 63, ##ctx) \ 5997 FN(bind, 64, ##ctx) \ 5998 FN(xdp_adjust_tail, 65, ##ctx) \ 5999 FN(skb_get_xfrm_state, 66, ##ctx) \ 6000 FN(get_stack, 67, ##ctx) \ 6001 FN(skb_load_bytes_relative, 68, ##ctx) \ 6002 FN(fib_lookup, 69, ##ctx) \ 6003 FN(sock_hash_update, 70, ##ctx) \ 6004 FN(msg_redirect_hash, 71, ##ctx) \ 6005 FN(sk_redirect_hash, 72, ##ctx) \ 6006 FN(lwt_push_encap, 73, ##ctx) \ 6007 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 6008 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 6009 FN(lwt_seg6_action, 76, ##ctx) \ 6010 FN(rc_repeat, 77, ##ctx) \ 6011 FN(rc_keydown, 78, ##ctx) \ 6012 FN(skb_cgroup_id, 79, ##ctx) \ 6013 FN(get_current_cgroup_id, 80, ##ctx) \ 6014 FN(get_local_storage, 81, ##ctx) \ 6015 FN(sk_select_reuseport, 82, ##ctx) \ 6016 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 6017 FN(sk_lookup_tcp, 84, ##ctx) \ 6018 FN(sk_lookup_udp, 85, ##ctx) \ 6019 FN(sk_release, 86, ##ctx) \ 6020 FN(map_push_elem, 87, ##ctx) \ 6021 FN(map_pop_elem, 88, ##ctx) \ 6022 FN(map_peek_elem, 89, ##ctx) \ 6023 FN(msg_push_data, 90, ##ctx) \ 6024 FN(msg_pop_data, 91, ##ctx) \ 6025 FN(rc_pointer_rel, 92, ##ctx) \ 6026 FN(spin_lock, 93, ##ctx) \ 6027 FN(spin_unlock, 94, ##ctx) \ 6028 FN(sk_fullsock, 95, ##ctx) \ 6029 FN(tcp_sock, 96, ##ctx) \ 6030 FN(skb_ecn_set_ce, 97, ##ctx) \ 6031 FN(get_listener_sock, 98, ##ctx) \ 6032 FN(skc_lookup_tcp, 99, ##ctx) \ 6033 FN(tcp_check_syncookie, 100, ##ctx) \ 6034 FN(sysctl_get_name, 101, ##ctx) \ 6035 FN(sysctl_get_current_value, 102, ##ctx) \ 6036 FN(sysctl_get_new_value, 103, ##ctx) \ 6037 FN(sysctl_set_new_value, 104, ##ctx) \ 6038 FN(strtol, 105, ##ctx) \ 6039 FN(strtoul, 106, ##ctx) \ 6040 FN(sk_storage_get, 107, ##ctx) \ 6041 FN(sk_storage_delete, 108, ##ctx) \ 6042 FN(send_signal, 109, ##ctx) \ 6043 FN(tcp_gen_syncookie, 110, ##ctx) \ 6044 FN(skb_output, 111, ##ctx) \ 6045 FN(probe_read_user, 112, ##ctx) \ 6046 FN(probe_read_kernel, 113, ##ctx) \ 6047 FN(probe_read_user_str, 114, ##ctx) \ 6048 FN(probe_read_kernel_str, 115, ##ctx) \ 6049 FN(tcp_send_ack, 116, ##ctx) \ 6050 FN(send_signal_thread, 117, ##ctx) \ 6051 FN(jiffies64, 118, ##ctx) \ 6052 FN(read_branch_records, 119, ##ctx) \ 6053 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 6054 FN(xdp_output, 121, ##ctx) \ 6055 FN(get_netns_cookie, 122, ##ctx) \ 6056 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 6057 FN(sk_assign, 124, ##ctx) \ 6058 FN(ktime_get_boot_ns, 125, ##ctx) \ 6059 FN(seq_printf, 126, ##ctx) \ 6060 FN(seq_write, 127, ##ctx) \ 6061 FN(sk_cgroup_id, 128, ##ctx) \ 6062 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 6063 FN(ringbuf_output, 130, ##ctx) \ 6064 FN(ringbuf_reserve, 131, ##ctx) \ 6065 FN(ringbuf_submit, 132, ##ctx) \ 6066 FN(ringbuf_discard, 133, ##ctx) \ 6067 FN(ringbuf_query, 134, ##ctx) \ 6068 FN(csum_level, 135, ##ctx) \ 6069 FN(skc_to_tcp6_sock, 136, ##ctx) \ 6070 FN(skc_to_tcp_sock, 137, ##ctx) \ 6071 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 6072 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 6073 FN(skc_to_udp6_sock, 140, ##ctx) \ 6074 FN(get_task_stack, 141, ##ctx) \ 6075 FN(load_hdr_opt, 142, ##ctx) \ 6076 FN(store_hdr_opt, 143, ##ctx) \ 6077 FN(reserve_hdr_opt, 144, ##ctx) \ 6078 FN(inode_storage_get, 145, ##ctx) \ 6079 FN(inode_storage_delete, 146, ##ctx) \ 6080 FN(d_path, 147, ##ctx) \ 6081 FN(copy_from_user, 148, ##ctx) \ 6082 FN(snprintf_btf, 149, ##ctx) \ 6083 FN(seq_printf_btf, 150, ##ctx) \ 6084 FN(skb_cgroup_classid, 151, ##ctx) \ 6085 FN(redirect_neigh, 152, ##ctx) \ 6086 FN(per_cpu_ptr, 153, ##ctx) \ 6087 FN(this_cpu_ptr, 154, ##ctx) \ 6088 FN(redirect_peer, 155, ##ctx) \ 6089 FN(task_storage_get, 156, ##ctx) \ 6090 FN(task_storage_delete, 157, ##ctx) \ 6091 FN(get_current_task_btf, 158, ##ctx) \ 6092 FN(bprm_opts_set, 159, ##ctx) \ 6093 FN(ktime_get_coarse_ns, 160, ##ctx) \ 6094 FN(ima_inode_hash, 161, ##ctx) \ 6095 FN(sock_from_file, 162, ##ctx) \ 6096 FN(check_mtu, 163, ##ctx) \ 6097 FN(for_each_map_elem, 164, ##ctx) \ 6098 FN(snprintf, 165, ##ctx) \ 6099 FN(sys_bpf, 166, ##ctx) \ 6100 FN(btf_find_by_name_kind, 167, ##ctx) \ 6101 FN(sys_close, 168, ##ctx) \ 6102 FN(timer_init, 169, ##ctx) \ 6103 FN(timer_set_callback, 170, ##ctx) \ 6104 FN(timer_start, 171, ##ctx) \ 6105 FN(timer_cancel, 172, ##ctx) \ 6106 FN(get_func_ip, 173, ##ctx) \ 6107 FN(get_attach_cookie, 174, ##ctx) \ 6108 FN(task_pt_regs, 175, ##ctx) \ 6109 FN(get_branch_snapshot, 176, ##ctx) \ 6110 FN(trace_vprintk, 177, ##ctx) \ 6111 FN(skc_to_unix_sock, 178, ##ctx) \ 6112 FN(kallsyms_lookup_name, 179, ##ctx) \ 6113 FN(find_vma, 180, ##ctx) \ 6114 FN(loop, 181, ##ctx) \ 6115 FN(strncmp, 182, ##ctx) \ 6116 FN(get_func_arg, 183, ##ctx) \ 6117 FN(get_func_ret, 184, ##ctx) \ 6118 FN(get_func_arg_cnt, 185, ##ctx) \ 6119 FN(get_retval, 186, ##ctx) \ 6120 FN(set_retval, 187, ##ctx) \ 6121 FN(xdp_get_buff_len, 188, ##ctx) \ 6122 FN(xdp_load_bytes, 189, ##ctx) \ 6123 FN(xdp_store_bytes, 190, ##ctx) \ 6124 FN(copy_from_user_task, 191, ##ctx) \ 6125 FN(skb_set_tstamp, 192, ##ctx) \ 6126 FN(ima_file_hash, 193, ##ctx) \ 6127 FN(kptr_xchg, 194, ##ctx) \ 6128 FN(map_lookup_percpu_elem, 195, ##ctx) \ 6129 FN(skc_to_mptcp_sock, 196, ##ctx) \ 6130 FN(dynptr_from_mem, 197, ##ctx) \ 6131 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 6132 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 6133 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 6134 FN(dynptr_read, 201, ##ctx) \ 6135 FN(dynptr_write, 202, ##ctx) \ 6136 FN(dynptr_data, 203, ##ctx) \ 6137 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 6138 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 6139 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 6140 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 6141 FN(ktime_get_tai_ns, 208, ##ctx) \ 6142 FN(user_ringbuf_drain, 209, ##ctx) \ 6143 FN(cgrp_storage_get, 210, ##ctx) \ 6144 FN(cgrp_storage_delete, 211, ##ctx) \ 6145 /* This helper list is effectively frozen. If you are trying to \ 6146 * add a new helper, you should add a kfunc instead which has \ 6147 * less stability guarantees. See Documentation/bpf/kfuncs.rst \ 6148 */ 6149 6150 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 6151 * know or care about integer value that is now passed as second argument 6152 */ 6153 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 6154 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 6155 6156 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 6157 * function eBPF program intends to call 6158 */ 6159 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 6160 enum bpf_func_id { 6161 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 6162 __BPF_FUNC_MAX_ID, 6163 }; 6164 #undef __BPF_ENUM_FN 6165 6166 /* All flags used by eBPF helper functions, placed here. */ 6167 6168 /* BPF_FUNC_skb_store_bytes flags. */ 6169 enum { 6170 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 6171 BPF_F_INVALIDATE_HASH = (1ULL << 1), 6172 }; 6173 6174 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 6175 * First 4 bits are for passing the header field size. 6176 */ 6177 enum { 6178 BPF_F_HDR_FIELD_MASK = 0xfULL, 6179 }; 6180 6181 /* BPF_FUNC_l4_csum_replace flags. */ 6182 enum { 6183 BPF_F_PSEUDO_HDR = (1ULL << 4), 6184 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 6185 BPF_F_MARK_ENFORCE = (1ULL << 6), 6186 BPF_F_IPV6 = (1ULL << 7), 6187 }; 6188 6189 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6190 enum { 6191 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6192 }; 6193 6194 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6195 enum { 6196 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6197 BPF_F_USER_STACK = (1ULL << 8), 6198 /* flags used by BPF_FUNC_get_stackid only. */ 6199 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6200 BPF_F_REUSE_STACKID = (1ULL << 10), 6201 /* flags used by BPF_FUNC_get_stack only. */ 6202 BPF_F_USER_BUILD_ID = (1ULL << 11), 6203 }; 6204 6205 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6206 enum { 6207 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6208 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6209 BPF_F_SEQ_NUMBER = (1ULL << 3), 6210 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6211 }; 6212 6213 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6214 enum { 6215 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6216 }; 6217 6218 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6219 * BPF_FUNC_perf_event_read_value flags. 6220 */ 6221 enum { 6222 BPF_F_INDEX_MASK = 0xffffffffULL, 6223 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6224 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6225 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6226 }; 6227 6228 /* Current network namespace */ 6229 enum { 6230 BPF_F_CURRENT_NETNS = (-1L), 6231 }; 6232 6233 /* BPF_FUNC_csum_level level values. */ 6234 enum { 6235 BPF_CSUM_LEVEL_QUERY, 6236 BPF_CSUM_LEVEL_INC, 6237 BPF_CSUM_LEVEL_DEC, 6238 BPF_CSUM_LEVEL_RESET, 6239 }; 6240 6241 /* BPF_FUNC_skb_adjust_room flags. */ 6242 enum { 6243 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6244 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6245 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6246 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6247 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6248 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6249 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6250 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6251 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6252 }; 6253 6254 enum { 6255 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6256 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6257 }; 6258 6259 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6260 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6261 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6262 6263 /* BPF_FUNC_sysctl_get_name flags. */ 6264 enum { 6265 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6266 }; 6267 6268 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6269 enum { 6270 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6271 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6272 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6273 */ 6274 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6275 }; 6276 6277 /* BPF_FUNC_read_branch_records flags. */ 6278 enum { 6279 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6280 }; 6281 6282 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6283 * BPF_FUNC_bpf_ringbuf_output flags. 6284 */ 6285 enum { 6286 BPF_RB_NO_WAKEUP = (1ULL << 0), 6287 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6288 }; 6289 6290 /* BPF_FUNC_bpf_ringbuf_query flags */ 6291 enum { 6292 BPF_RB_AVAIL_DATA = 0, 6293 BPF_RB_RING_SIZE = 1, 6294 BPF_RB_CONS_POS = 2, 6295 BPF_RB_PROD_POS = 3, 6296 BPF_RB_OVERWRITE_POS = 4, 6297 }; 6298 6299 /* BPF ring buffer constants */ 6300 enum { 6301 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6302 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6303 BPF_RINGBUF_HDR_SZ = 8, 6304 }; 6305 6306 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6307 enum { 6308 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6309 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6310 }; 6311 6312 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6313 enum bpf_adj_room_mode { 6314 BPF_ADJ_ROOM_NET, 6315 BPF_ADJ_ROOM_MAC, 6316 }; 6317 6318 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6319 enum bpf_hdr_start_off { 6320 BPF_HDR_START_MAC, 6321 BPF_HDR_START_NET, 6322 }; 6323 6324 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6325 enum bpf_lwt_encap_mode { 6326 BPF_LWT_ENCAP_SEG6, 6327 BPF_LWT_ENCAP_SEG6_INLINE, 6328 BPF_LWT_ENCAP_IP, 6329 }; 6330 6331 /* Flags for bpf_bprm_opts_set helper */ 6332 enum { 6333 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6334 }; 6335 6336 /* Flags for bpf_redirect and bpf_redirect_map helpers */ 6337 enum { 6338 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */ 6339 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */ 6340 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */ 6341 #define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS) 6342 }; 6343 6344 #define __bpf_md_ptr(type, name) \ 6345 union { \ 6346 type name; \ 6347 __u64 :64; \ 6348 } __attribute__((aligned(8))) 6349 6350 /* The enum used in skb->tstamp_type. It specifies the clock type 6351 * of the time stored in the skb->tstamp. 6352 */ 6353 enum { 6354 BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */ 6355 BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */ 6356 BPF_SKB_CLOCK_REALTIME = 0, 6357 BPF_SKB_CLOCK_MONOTONIC = 1, 6358 BPF_SKB_CLOCK_TAI = 2, 6359 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle, 6360 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid. 6361 */ 6362 }; 6363 6364 /* user accessible mirror of in-kernel sk_buff. 6365 * new fields can only be added to the end of this structure 6366 */ 6367 struct __sk_buff { 6368 __u32 len; 6369 __u32 pkt_type; 6370 __u32 mark; 6371 __u32 queue_mapping; 6372 __u32 protocol; 6373 __u32 vlan_present; 6374 __u32 vlan_tci; 6375 __u32 vlan_proto; 6376 __u32 priority; 6377 __u32 ingress_ifindex; 6378 __u32 ifindex; 6379 __u32 tc_index; 6380 __u32 cb[5]; 6381 __u32 hash; 6382 __u32 tc_classid; 6383 __u32 data; 6384 __u32 data_end; 6385 __u32 napi_id; 6386 6387 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6388 __u32 family; 6389 __u32 remote_ip4; /* Stored in network byte order */ 6390 __u32 local_ip4; /* Stored in network byte order */ 6391 __u32 remote_ip6[4]; /* Stored in network byte order */ 6392 __u32 local_ip6[4]; /* Stored in network byte order */ 6393 __u32 remote_port; /* Stored in network byte order */ 6394 __u32 local_port; /* stored in host byte order */ 6395 /* ... here. */ 6396 6397 __u32 data_meta; 6398 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6399 __u64 tstamp; 6400 __u32 wire_len; 6401 __u32 gso_segs; 6402 __bpf_md_ptr(struct bpf_sock *, sk); 6403 __u32 gso_size; 6404 __u8 tstamp_type; 6405 __u32 :24; /* Padding, future use. */ 6406 __u64 hwtstamp; 6407 }; 6408 6409 struct bpf_tunnel_key { 6410 __u32 tunnel_id; 6411 union { 6412 __u32 remote_ipv4; 6413 __u32 remote_ipv6[4]; 6414 }; 6415 __u8 tunnel_tos; 6416 __u8 tunnel_ttl; 6417 union { 6418 __u16 tunnel_ext; /* compat */ 6419 __be16 tunnel_flags; 6420 }; 6421 __u32 tunnel_label; 6422 union { 6423 __u32 local_ipv4; 6424 __u32 local_ipv6[4]; 6425 }; 6426 }; 6427 6428 /* user accessible mirror of in-kernel xfrm_state. 6429 * new fields can only be added to the end of this structure 6430 */ 6431 struct bpf_xfrm_state { 6432 __u32 reqid; 6433 __u32 spi; /* Stored in network byte order */ 6434 __u16 family; 6435 __u16 ext; /* Padding, future use. */ 6436 union { 6437 __u32 remote_ipv4; /* Stored in network byte order */ 6438 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6439 }; 6440 }; 6441 6442 /* Generic BPF return codes which all BPF program types may support. 6443 * The values are binary compatible with their TC_ACT_* counter-part to 6444 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6445 * programs. 6446 * 6447 * XDP is handled seprately, see XDP_*. 6448 */ 6449 enum bpf_ret_code { 6450 BPF_OK = 0, 6451 /* 1 reserved */ 6452 BPF_DROP = 2, 6453 /* 3-6 reserved */ 6454 BPF_REDIRECT = 7, 6455 /* >127 are reserved for prog type specific return codes. 6456 * 6457 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6458 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6459 * changed and should be routed based on its new L3 header. 6460 * (This is an L3 redirect, as opposed to L2 redirect 6461 * represented by BPF_REDIRECT above). 6462 */ 6463 BPF_LWT_REROUTE = 128, 6464 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6465 * to indicate that no custom dissection was performed, and 6466 * fallback to standard dissector is requested. 6467 */ 6468 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6469 }; 6470 6471 struct bpf_sock { 6472 __u32 bound_dev_if; 6473 __u32 family; 6474 __u32 type; 6475 __u32 protocol; 6476 __u32 mark; 6477 __u32 priority; 6478 /* IP address also allows 1 and 2 bytes access */ 6479 __u32 src_ip4; 6480 __u32 src_ip6[4]; 6481 __u32 src_port; /* host byte order */ 6482 __be16 dst_port; /* network byte order */ 6483 __u16 :16; /* zero padding */ 6484 __u32 dst_ip4; 6485 __u32 dst_ip6[4]; 6486 __u32 state; 6487 __s32 rx_queue_mapping; 6488 }; 6489 6490 struct bpf_tcp_sock { 6491 __u32 snd_cwnd; /* Sending congestion window */ 6492 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6493 __u32 rtt_min; 6494 __u32 snd_ssthresh; /* Slow start size threshold */ 6495 __u32 rcv_nxt; /* What we want to receive next */ 6496 __u32 snd_nxt; /* Next sequence we send */ 6497 __u32 snd_una; /* First byte we want an ack for */ 6498 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6499 __u32 ecn_flags; /* ECN status bits. */ 6500 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6501 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6502 __u32 packets_out; /* Packets which are "in flight" */ 6503 __u32 retrans_out; /* Retransmitted packets out */ 6504 __u32 total_retrans; /* Total retransmits for entire connection */ 6505 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6506 * total number of segments in. 6507 */ 6508 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6509 * total number of data segments in. 6510 */ 6511 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6512 * The total number of segments sent. 6513 */ 6514 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6515 * total number of data segments sent. 6516 */ 6517 __u32 lost_out; /* Lost packets */ 6518 __u32 sacked_out; /* SACK'd packets */ 6519 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6520 * sum(delta(rcv_nxt)), or how many bytes 6521 * were acked. 6522 */ 6523 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6524 * sum(delta(snd_una)), or how many bytes 6525 * were acked. 6526 */ 6527 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6528 * total number of DSACK blocks received 6529 */ 6530 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6531 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6532 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6533 }; 6534 6535 struct bpf_sock_tuple { 6536 union { 6537 struct { 6538 __be32 saddr; 6539 __be32 daddr; 6540 __be16 sport; 6541 __be16 dport; 6542 } ipv4; 6543 struct { 6544 __be32 saddr[4]; 6545 __be32 daddr[4]; 6546 __be16 sport; 6547 __be16 dport; 6548 } ipv6; 6549 }; 6550 }; 6551 6552 /* (Simplified) user return codes for tcx prog type. 6553 * A valid tcx program must return one of these defined values. All other 6554 * return codes are reserved for future use. Must remain compatible with 6555 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6556 * return codes are mapped to TCX_NEXT. 6557 */ 6558 enum tcx_action_base { 6559 TCX_NEXT = -1, 6560 TCX_PASS = 0, 6561 TCX_DROP = 2, 6562 TCX_REDIRECT = 7, 6563 }; 6564 6565 struct bpf_xdp_sock { 6566 __u32 queue_id; 6567 }; 6568 6569 #define XDP_PACKET_HEADROOM 256 6570 6571 /* User return codes for XDP prog type. 6572 * A valid XDP program must return one of these defined values. All other 6573 * return codes are reserved for future use. Unknown return codes will 6574 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6575 */ 6576 enum xdp_action { 6577 XDP_ABORTED = 0, 6578 XDP_DROP, 6579 XDP_PASS, 6580 XDP_TX, 6581 XDP_REDIRECT, 6582 }; 6583 6584 /* user accessible metadata for XDP packet hook 6585 * new fields must be added to the end of this structure 6586 */ 6587 struct xdp_md { 6588 __u32 data; 6589 __u32 data_end; 6590 __u32 data_meta; 6591 /* Below access go through struct xdp_rxq_info */ 6592 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6593 __u32 rx_queue_index; /* rxq->queue_index */ 6594 6595 __u32 egress_ifindex; /* txq->dev->ifindex */ 6596 }; 6597 6598 /* DEVMAP map-value layout 6599 * 6600 * The struct data-layout of map-value is a configuration interface. 6601 * New members can only be added to the end of this structure. 6602 */ 6603 struct bpf_devmap_val { 6604 __u32 ifindex; /* device index */ 6605 union { 6606 int fd; /* prog fd on map write */ 6607 __u32 id; /* prog id on map read */ 6608 } bpf_prog; 6609 }; 6610 6611 /* CPUMAP map-value layout 6612 * 6613 * The struct data-layout of map-value is a configuration interface. 6614 * New members can only be added to the end of this structure. 6615 */ 6616 struct bpf_cpumap_val { 6617 __u32 qsize; /* queue size to remote target CPU */ 6618 union { 6619 int fd; /* prog fd on map write */ 6620 __u32 id; /* prog id on map read */ 6621 } bpf_prog; 6622 }; 6623 6624 enum sk_action { 6625 SK_DROP = 0, 6626 SK_PASS, 6627 }; 6628 6629 /* user accessible metadata for SK_MSG packet hook, new fields must 6630 * be added to the end of this structure 6631 */ 6632 struct sk_msg_md { 6633 __bpf_md_ptr(void *, data); 6634 __bpf_md_ptr(void *, data_end); 6635 6636 __u32 family; 6637 __u32 remote_ip4; /* Stored in network byte order */ 6638 __u32 local_ip4; /* Stored in network byte order */ 6639 __u32 remote_ip6[4]; /* Stored in network byte order */ 6640 __u32 local_ip6[4]; /* Stored in network byte order */ 6641 __u32 remote_port; /* Stored in network byte order */ 6642 __u32 local_port; /* stored in host byte order */ 6643 __u32 size; /* Total size of sk_msg */ 6644 6645 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6646 }; 6647 6648 struct sk_reuseport_md { 6649 /* 6650 * Start of directly accessible data. It begins from 6651 * the tcp/udp header. 6652 */ 6653 __bpf_md_ptr(void *, data); 6654 /* End of directly accessible data */ 6655 __bpf_md_ptr(void *, data_end); 6656 /* 6657 * Total length of packet (starting from the tcp/udp header). 6658 * Note that the directly accessible bytes (data_end - data) 6659 * could be less than this "len". Those bytes could be 6660 * indirectly read by a helper "bpf_skb_load_bytes()". 6661 */ 6662 __u32 len; 6663 /* 6664 * Eth protocol in the mac header (network byte order). e.g. 6665 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6666 */ 6667 __u32 eth_protocol; 6668 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6669 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6670 __u32 hash; /* A hash of the packet 4 tuples */ 6671 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6672 * new incoming connection request (e.g. selecting a listen sk for 6673 * the received SYN in the TCP case). reuse->sk is one of the sk 6674 * in the reuseport group. The bpf prog can use reuse->sk to learn 6675 * the local listening ip/port without looking into the skb. 6676 * 6677 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6678 * reuse->migrating_sk is the socket that needs to be migrated 6679 * to another listening socket. migrating_sk could be a fullsock 6680 * sk that is fully established or a reqsk that is in-the-middle 6681 * of 3-way handshake. 6682 */ 6683 __bpf_md_ptr(struct bpf_sock *, sk); 6684 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6685 }; 6686 6687 #define BPF_TAG_SIZE 8 6688 6689 struct bpf_prog_info { 6690 __u32 type; 6691 __u32 id; 6692 __u8 tag[BPF_TAG_SIZE]; 6693 __u32 jited_prog_len; 6694 __u32 xlated_prog_len; 6695 __aligned_u64 jited_prog_insns; 6696 __aligned_u64 xlated_prog_insns; 6697 __u64 load_time; /* ns since boottime */ 6698 __u32 created_by_uid; 6699 __u32 nr_map_ids; 6700 __aligned_u64 map_ids; 6701 char name[BPF_OBJ_NAME_LEN]; 6702 __u32 ifindex; 6703 __u32 gpl_compatible:1; 6704 __u32 :31; /* alignment pad */ 6705 __u64 netns_dev; 6706 __u64 netns_ino; 6707 __u32 nr_jited_ksyms; 6708 __u32 nr_jited_func_lens; 6709 __aligned_u64 jited_ksyms; 6710 __aligned_u64 jited_func_lens; 6711 __u32 btf_id; 6712 __u32 func_info_rec_size; 6713 __aligned_u64 func_info; 6714 __u32 nr_func_info; 6715 __u32 nr_line_info; 6716 __aligned_u64 line_info; 6717 __aligned_u64 jited_line_info; 6718 __u32 nr_jited_line_info; 6719 __u32 line_info_rec_size; 6720 __u32 jited_line_info_rec_size; 6721 __u32 nr_prog_tags; 6722 __aligned_u64 prog_tags; 6723 __u64 run_time_ns; 6724 __u64 run_cnt; 6725 __u64 recursion_misses; 6726 __u32 verified_insns; 6727 __u32 attach_btf_obj_id; 6728 __u32 attach_btf_id; 6729 __u32 :32; 6730 } __attribute__((aligned(8))); 6731 6732 struct bpf_map_info { 6733 __u32 type; 6734 __u32 id; 6735 __u32 key_size; 6736 __u32 value_size; 6737 __u32 max_entries; 6738 __u32 map_flags; 6739 char name[BPF_OBJ_NAME_LEN]; 6740 __u32 ifindex; 6741 __u32 btf_vmlinux_value_type_id; 6742 __u64 netns_dev; 6743 __u64 netns_ino; 6744 __u32 btf_id; 6745 __u32 btf_key_type_id; 6746 __u32 btf_value_type_id; 6747 __u32 btf_vmlinux_id; 6748 __u64 map_extra; 6749 __aligned_u64 hash; 6750 __u32 hash_size; 6751 __u32 :32; 6752 } __attribute__((aligned(8))); 6753 6754 struct bpf_btf_info { 6755 __aligned_u64 btf; 6756 __u32 btf_size; 6757 __u32 id; 6758 __aligned_u64 name; 6759 __u32 name_len; 6760 __u32 kernel_btf; 6761 } __attribute__((aligned(8))); 6762 6763 struct bpf_link_info { 6764 __u32 type; 6765 __u32 id; 6766 __u32 prog_id; 6767 union { 6768 struct { 6769 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6770 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6771 __u32 :32; 6772 __u64 cookie; 6773 } raw_tracepoint; 6774 struct { 6775 __u32 attach_type; 6776 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6777 __u32 target_btf_id; /* BTF type id inside the object */ 6778 __u32 :32; 6779 __u64 cookie; 6780 } tracing; 6781 struct { 6782 __u64 cgroup_id; 6783 __u32 attach_type; 6784 } cgroup; 6785 struct { 6786 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6787 __u32 target_name_len; /* in/out: target_name buffer len */ 6788 6789 /* If the iter specific field is 32 bits, it can be put 6790 * in the first or second union. Otherwise it should be 6791 * put in the second union. 6792 */ 6793 union { 6794 struct { 6795 __u32 map_id; 6796 } map; 6797 }; 6798 union { 6799 struct { 6800 __u64 cgroup_id; 6801 __u32 order; 6802 } cgroup; 6803 struct { 6804 __u32 tid; 6805 __u32 pid; 6806 } task; 6807 }; 6808 } iter; 6809 struct { 6810 __u32 netns_ino; 6811 __u32 attach_type; 6812 } netns; 6813 struct { 6814 __u32 ifindex; 6815 } xdp; 6816 struct { 6817 __u32 map_id; 6818 } struct_ops; 6819 struct { 6820 __u32 pf; 6821 __u32 hooknum; 6822 __s32 priority; 6823 __u32 flags; 6824 } netfilter; 6825 struct { 6826 __aligned_u64 addrs; 6827 __u32 count; /* in/out: kprobe_multi function count */ 6828 __u32 flags; 6829 __u64 missed; 6830 __aligned_u64 cookies; 6831 } kprobe_multi; 6832 struct { 6833 __aligned_u64 path; 6834 __aligned_u64 offsets; 6835 __aligned_u64 ref_ctr_offsets; 6836 __aligned_u64 cookies; 6837 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6838 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6839 __u32 flags; 6840 __u32 pid; 6841 } uprobe_multi; 6842 struct { 6843 __u32 type; /* enum bpf_perf_event_type */ 6844 __u32 :32; 6845 union { 6846 struct { 6847 __aligned_u64 file_name; /* in/out */ 6848 __u32 name_len; 6849 __u32 offset; /* offset from file_name */ 6850 __u64 cookie; 6851 __u64 ref_ctr_offset; 6852 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6853 struct { 6854 __aligned_u64 func_name; /* in/out */ 6855 __u32 name_len; 6856 __u32 offset; /* offset from func_name */ 6857 __u64 addr; 6858 __u64 missed; 6859 __u64 cookie; 6860 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6861 struct { 6862 __aligned_u64 tp_name; /* in/out */ 6863 __u32 name_len; 6864 __u32 :32; 6865 __u64 cookie; 6866 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6867 struct { 6868 __u64 config; 6869 __u32 type; 6870 __u32 :32; 6871 __u64 cookie; 6872 } event; /* BPF_PERF_EVENT_EVENT */ 6873 }; 6874 } perf_event; 6875 struct { 6876 __u32 ifindex; 6877 __u32 attach_type; 6878 } tcx; 6879 struct { 6880 __u32 ifindex; 6881 __u32 attach_type; 6882 } netkit; 6883 struct { 6884 __u32 map_id; 6885 __u32 attach_type; 6886 } sockmap; 6887 }; 6888 } __attribute__((aligned(8))); 6889 6890 struct bpf_token_info { 6891 __u64 allowed_cmds; 6892 __u64 allowed_maps; 6893 __u64 allowed_progs; 6894 __u64 allowed_attachs; 6895 } __attribute__((aligned(8))); 6896 6897 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6898 * by user and intended to be used by socket (e.g. to bind to, depends on 6899 * attach type). 6900 */ 6901 struct bpf_sock_addr { 6902 __u32 user_family; /* Allows 4-byte read, but no write. */ 6903 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6904 * Stored in network byte order. 6905 */ 6906 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6907 * Stored in network byte order. 6908 */ 6909 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6910 * Stored in network byte order 6911 */ 6912 __u32 family; /* Allows 4-byte read, but no write */ 6913 __u32 type; /* Allows 4-byte read, but no write */ 6914 __u32 protocol; /* Allows 4-byte read, but no write */ 6915 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6916 * Stored in network byte order. 6917 */ 6918 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6919 * Stored in network byte order. 6920 */ 6921 __bpf_md_ptr(struct bpf_sock *, sk); 6922 }; 6923 6924 /* User bpf_sock_ops struct to access socket values and specify request ops 6925 * and their replies. 6926 * Some of this fields are in network (bigendian) byte order and may need 6927 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6928 * New fields can only be added at the end of this structure 6929 */ 6930 struct bpf_sock_ops { 6931 __u32 op; 6932 union { 6933 __u32 args[4]; /* Optionally passed to bpf program */ 6934 __u32 reply; /* Returned by bpf program */ 6935 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6936 }; 6937 __u32 family; 6938 __u32 remote_ip4; /* Stored in network byte order */ 6939 __u32 local_ip4; /* Stored in network byte order */ 6940 __u32 remote_ip6[4]; /* Stored in network byte order */ 6941 __u32 local_ip6[4]; /* Stored in network byte order */ 6942 __u32 remote_port; /* Stored in network byte order */ 6943 __u32 local_port; /* stored in host byte order */ 6944 __u32 is_fullsock; /* Some TCP fields are only valid if 6945 * there is a full socket. If not, the 6946 * fields read as zero. 6947 */ 6948 __u32 snd_cwnd; 6949 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6950 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6951 __u32 state; 6952 __u32 rtt_min; 6953 __u32 snd_ssthresh; 6954 __u32 rcv_nxt; 6955 __u32 snd_nxt; 6956 __u32 snd_una; 6957 __u32 mss_cache; 6958 __u32 ecn_flags; 6959 __u32 rate_delivered; 6960 __u32 rate_interval_us; 6961 __u32 packets_out; 6962 __u32 retrans_out; 6963 __u32 total_retrans; 6964 __u32 segs_in; 6965 __u32 data_segs_in; 6966 __u32 segs_out; 6967 __u32 data_segs_out; 6968 __u32 lost_out; 6969 __u32 sacked_out; 6970 __u32 sk_txhash; 6971 __u64 bytes_received; 6972 __u64 bytes_acked; 6973 __bpf_md_ptr(struct bpf_sock *, sk); 6974 /* [skb_data, skb_data_end) covers the whole TCP header. 6975 * 6976 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6977 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6978 * header has not been written. 6979 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6980 * been written so far. 6981 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6982 * the 3WHS. 6983 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6984 * the 3WHS. 6985 * 6986 * bpf_load_hdr_opt() can also be used to read a particular option. 6987 */ 6988 __bpf_md_ptr(void *, skb_data); 6989 __bpf_md_ptr(void *, skb_data_end); 6990 __u32 skb_len; /* The total length of a packet. 6991 * It includes the header, options, 6992 * and payload. 6993 */ 6994 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6995 * an easy way to check for tcp_flags 6996 * without parsing skb_data. 6997 * 6998 * In particular, the skb_tcp_flags 6999 * will still be available in 7000 * BPF_SOCK_OPS_HDR_OPT_LEN even though 7001 * the outgoing header has not 7002 * been written yet. 7003 */ 7004 __u64 skb_hwtstamp; 7005 }; 7006 7007 /* Definitions for bpf_sock_ops_cb_flags */ 7008 enum { 7009 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 7010 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 7011 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 7012 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 7013 /* Call bpf for all received TCP headers. The bpf prog will be 7014 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 7015 * 7016 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 7017 * for the header option related helpers that will be useful 7018 * to the bpf programs. 7019 * 7020 * It could be used at the client/active side (i.e. connect() side) 7021 * when the server told it that the server was in syncookie 7022 * mode and required the active side to resend the bpf-written 7023 * options. The active side can keep writing the bpf-options until 7024 * it received a valid packet from the server side to confirm 7025 * the earlier packet (and options) has been received. The later 7026 * example patch is using it like this at the active side when the 7027 * server is in syncookie mode. 7028 * 7029 * The bpf prog will usually turn this off in the common cases. 7030 */ 7031 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 7032 /* Call bpf when kernel has received a header option that 7033 * the kernel cannot handle. The bpf prog will be called under 7034 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 7035 * 7036 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 7037 * for the header option related helpers that will be useful 7038 * to the bpf programs. 7039 */ 7040 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 7041 /* Call bpf when the kernel is writing header options for the 7042 * outgoing packet. The bpf prog will first be called 7043 * to reserve space in a skb under 7044 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 7045 * the bpf prog will be called to write the header option(s) 7046 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7047 * 7048 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 7049 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 7050 * related helpers that will be useful to the bpf programs. 7051 * 7052 * The kernel gets its chance to reserve space and write 7053 * options first before the BPF program does. 7054 */ 7055 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 7056 /* Mask of all currently supported cb flags */ 7057 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 7058 }; 7059 7060 enum { 7061 SK_BPF_CB_TX_TIMESTAMPING = 1<<0, 7062 SK_BPF_CB_MASK = (SK_BPF_CB_TX_TIMESTAMPING - 1) | 7063 SK_BPF_CB_TX_TIMESTAMPING 7064 }; 7065 7066 /* List of known BPF sock_ops operators. 7067 * New entries can only be added at the end 7068 */ 7069 enum { 7070 BPF_SOCK_OPS_VOID, 7071 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 7072 * -1 if default value should be used 7073 */ 7074 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 7075 * window (in packets) or -1 if default 7076 * value should be used 7077 */ 7078 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 7079 * active connection is initialized 7080 */ 7081 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 7082 * active connection is 7083 * established 7084 */ 7085 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 7086 * passive connection is 7087 * established 7088 */ 7089 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 7090 * needs ECN 7091 */ 7092 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 7093 * based on the path and may be 7094 * dependent on the congestion control 7095 * algorithm. In general it indicates 7096 * a congestion threshold. RTTs above 7097 * this indicate congestion 7098 */ 7099 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 7100 * Arg1: value of icsk_retransmits 7101 * Arg2: value of icsk_rto 7102 * Arg3: whether RTO has expired 7103 */ 7104 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 7105 * Arg1: sequence number of 1st byte 7106 * Arg2: # segments 7107 * Arg3: return value of 7108 * tcp_transmit_skb (0 => success) 7109 */ 7110 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 7111 * Arg1: old_state 7112 * Arg2: new_state 7113 */ 7114 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 7115 * socket transition to LISTEN state. 7116 */ 7117 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 7118 * Arg1: measured RTT input (mrtt) 7119 * Arg2: updated srtt 7120 */ 7121 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 7122 * It will be called to handle 7123 * the packets received at 7124 * an already established 7125 * connection. 7126 * 7127 * sock_ops->skb_data: 7128 * Referring to the received skb. 7129 * It covers the TCP header only. 7130 * 7131 * bpf_load_hdr_opt() can also 7132 * be used to search for a 7133 * particular option. 7134 */ 7135 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 7136 * header option later in 7137 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7138 * Arg1: bool want_cookie. (in 7139 * writing SYNACK only) 7140 * 7141 * sock_ops->skb_data: 7142 * Not available because no header has 7143 * been written yet. 7144 * 7145 * sock_ops->skb_tcp_flags: 7146 * The tcp_flags of the 7147 * outgoing skb. (e.g. SYN, ACK, FIN). 7148 * 7149 * bpf_reserve_hdr_opt() should 7150 * be used to reserve space. 7151 */ 7152 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 7153 * Arg1: bool want_cookie. (in 7154 * writing SYNACK only) 7155 * 7156 * sock_ops->skb_data: 7157 * Referring to the outgoing skb. 7158 * It covers the TCP header 7159 * that has already been written 7160 * by the kernel and the 7161 * earlier bpf-progs. 7162 * 7163 * sock_ops->skb_tcp_flags: 7164 * The tcp_flags of the outgoing 7165 * skb. (e.g. SYN, ACK, FIN). 7166 * 7167 * bpf_store_hdr_opt() should 7168 * be used to write the 7169 * option. 7170 * 7171 * bpf_load_hdr_opt() can also 7172 * be used to search for a 7173 * particular option that 7174 * has already been written 7175 * by the kernel or the 7176 * earlier bpf-progs. 7177 */ 7178 BPF_SOCK_OPS_TSTAMP_SCHED_CB, /* Called when skb is passing 7179 * through dev layer when 7180 * SK_BPF_CB_TX_TIMESTAMPING 7181 * feature is on. 7182 */ 7183 BPF_SOCK_OPS_TSTAMP_SND_SW_CB, /* Called when skb is about to send 7184 * to the nic when SK_BPF_CB_TX_TIMESTAMPING 7185 * feature is on. 7186 */ 7187 BPF_SOCK_OPS_TSTAMP_SND_HW_CB, /* Called in hardware phase when 7188 * SK_BPF_CB_TX_TIMESTAMPING feature 7189 * is on. 7190 */ 7191 BPF_SOCK_OPS_TSTAMP_ACK_CB, /* Called when all the skbs in the 7192 * same sendmsg call are acked 7193 * when SK_BPF_CB_TX_TIMESTAMPING 7194 * feature is on. 7195 */ 7196 BPF_SOCK_OPS_TSTAMP_SENDMSG_CB, /* Called when every sendmsg syscall 7197 * is triggered. It's used to correlate 7198 * sendmsg timestamp with corresponding 7199 * tskey. 7200 */ 7201 }; 7202 7203 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 7204 * changes between the TCP and BPF versions. Ideally this should never happen. 7205 * If it does, we need to add code to convert them before calling 7206 * the BPF sock_ops function. 7207 */ 7208 enum { 7209 BPF_TCP_ESTABLISHED = 1, 7210 BPF_TCP_SYN_SENT, 7211 BPF_TCP_SYN_RECV, 7212 BPF_TCP_FIN_WAIT1, 7213 BPF_TCP_FIN_WAIT2, 7214 BPF_TCP_TIME_WAIT, 7215 BPF_TCP_CLOSE, 7216 BPF_TCP_CLOSE_WAIT, 7217 BPF_TCP_LAST_ACK, 7218 BPF_TCP_LISTEN, 7219 BPF_TCP_CLOSING, /* Now a valid state */ 7220 BPF_TCP_NEW_SYN_RECV, 7221 BPF_TCP_BOUND_INACTIVE, 7222 7223 BPF_TCP_MAX_STATES /* Leave at the end! */ 7224 }; 7225 7226 enum { 7227 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 7228 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 7229 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 7230 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 7231 /* Copy the SYN pkt to optval 7232 * 7233 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 7234 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 7235 * to only getting from the saved_syn. It can either get the 7236 * syn packet from: 7237 * 7238 * 1. the just-received SYN packet (only available when writing the 7239 * SYNACK). It will be useful when it is not necessary to 7240 * save the SYN packet for latter use. It is also the only way 7241 * to get the SYN during syncookie mode because the syn 7242 * packet cannot be saved during syncookie. 7243 * 7244 * OR 7245 * 7246 * 2. the earlier saved syn which was done by 7247 * bpf_setsockopt(TCP_SAVE_SYN). 7248 * 7249 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7250 * SYN packet is obtained. 7251 * 7252 * If the bpf-prog does not need the IP[46] header, the 7253 * bpf-prog can avoid parsing the IP header by using 7254 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7255 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7256 * 7257 * >0: Total number of bytes copied 7258 * -ENOSPC: Not enough space in optval. Only optlen number of 7259 * bytes is copied. 7260 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7261 * is not saved by setsockopt(TCP_SAVE_SYN). 7262 */ 7263 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7264 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7265 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7266 TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */ 7267 SK_BPF_CB_FLAGS = 1009, /* Get or set sock ops flags in socket */ 7268 SK_BPF_BYPASS_PROT_MEM = 1010, /* Get or Set sk->sk_bypass_prot_mem */ 7269 7270 }; 7271 7272 enum { 7273 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7274 }; 7275 7276 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7277 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7278 */ 7279 enum { 7280 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7281 * total option spaces 7282 * required for an established 7283 * sk in order to calculate the 7284 * MSS. No skb is actually 7285 * sent. 7286 */ 7287 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7288 * when sending a SYN. 7289 */ 7290 }; 7291 7292 struct bpf_perf_event_value { 7293 __u64 counter; 7294 __u64 enabled; 7295 __u64 running; 7296 }; 7297 7298 enum { 7299 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7300 BPF_DEVCG_ACC_READ = (1ULL << 1), 7301 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7302 }; 7303 7304 enum { 7305 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7306 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7307 }; 7308 7309 struct bpf_cgroup_dev_ctx { 7310 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7311 __u32 access_type; 7312 __u32 major; 7313 __u32 minor; 7314 }; 7315 7316 struct bpf_raw_tracepoint_args { 7317 __u64 args[0]; 7318 }; 7319 7320 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7321 * OUTPUT: Do lookup from egress perspective; default is ingress 7322 */ 7323 enum { 7324 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7325 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7326 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7327 BPF_FIB_LOOKUP_TBID = (1U << 3), 7328 BPF_FIB_LOOKUP_SRC = (1U << 4), 7329 BPF_FIB_LOOKUP_MARK = (1U << 5), 7330 }; 7331 7332 enum { 7333 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7334 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7335 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7336 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7337 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7338 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7339 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7340 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7341 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7342 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7343 }; 7344 7345 struct bpf_fib_lookup { 7346 /* input: network family for lookup (AF_INET, AF_INET6) 7347 * output: network family of egress nexthop 7348 */ 7349 __u8 family; 7350 7351 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7352 __u8 l4_protocol; 7353 __be16 sport; 7354 __be16 dport; 7355 7356 union { /* used for MTU check */ 7357 /* input to lookup */ 7358 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7359 7360 /* output: MTU value */ 7361 __u16 mtu_result; 7362 } __attribute__((packed, aligned(2))); 7363 /* input: L3 device index for lookup 7364 * output: device index from FIB lookup 7365 */ 7366 __u32 ifindex; 7367 7368 union { 7369 /* inputs to lookup */ 7370 __u8 tos; /* AF_INET */ 7371 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7372 7373 /* output: metric of fib result (IPv4/IPv6 only) */ 7374 __u32 rt_metric; 7375 }; 7376 7377 /* input: source address to consider for lookup 7378 * output: source address result from lookup 7379 */ 7380 union { 7381 __be32 ipv4_src; 7382 __u32 ipv6_src[4]; /* in6_addr; network order */ 7383 }; 7384 7385 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7386 * network header. output: bpf_fib_lookup sets to gateway address 7387 * if FIB lookup returns gateway route 7388 */ 7389 union { 7390 __be32 ipv4_dst; 7391 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7392 }; 7393 7394 union { 7395 struct { 7396 /* output */ 7397 __be16 h_vlan_proto; 7398 __be16 h_vlan_TCI; 7399 }; 7400 /* input: when accompanied with the 7401 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7402 * specific routing table to use for the fib lookup. 7403 */ 7404 __u32 tbid; 7405 }; 7406 7407 union { 7408 /* input */ 7409 struct { 7410 __u32 mark; /* policy routing */ 7411 /* 2 4-byte holes for input */ 7412 }; 7413 7414 /* output: source and dest mac */ 7415 struct { 7416 __u8 smac[6]; /* ETH_ALEN */ 7417 __u8 dmac[6]; /* ETH_ALEN */ 7418 }; 7419 }; 7420 }; 7421 7422 struct bpf_redir_neigh { 7423 /* network family for lookup (AF_INET, AF_INET6) */ 7424 __u32 nh_family; 7425 /* network address of nexthop; skips fib lookup to find gateway */ 7426 union { 7427 __be32 ipv4_nh; 7428 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7429 }; 7430 }; 7431 7432 /* bpf_check_mtu flags*/ 7433 enum bpf_check_mtu_flags { 7434 BPF_MTU_CHK_SEGS = (1U << 0), 7435 }; 7436 7437 enum bpf_check_mtu_ret { 7438 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7439 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7440 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7441 }; 7442 7443 enum bpf_task_fd_type { 7444 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7445 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7446 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7447 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7448 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7449 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7450 }; 7451 7452 enum { 7453 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7454 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7455 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7456 }; 7457 7458 struct bpf_flow_keys { 7459 __u16 nhoff; 7460 __u16 thoff; 7461 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7462 __u8 is_frag; 7463 __u8 is_first_frag; 7464 __u8 is_encap; 7465 __u8 ip_proto; 7466 __be16 n_proto; 7467 __be16 sport; 7468 __be16 dport; 7469 union { 7470 struct { 7471 __be32 ipv4_src; 7472 __be32 ipv4_dst; 7473 }; 7474 struct { 7475 __u32 ipv6_src[4]; /* in6_addr; network order */ 7476 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7477 }; 7478 }; 7479 __u32 flags; 7480 __be32 flow_label; 7481 }; 7482 7483 struct bpf_func_info { 7484 __u32 insn_off; 7485 __u32 type_id; 7486 }; 7487 7488 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7489 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7490 7491 struct bpf_line_info { 7492 __u32 insn_off; 7493 __u32 file_name_off; 7494 __u32 line_off; 7495 __u32 line_col; 7496 }; 7497 7498 struct bpf_spin_lock { 7499 __u32 val; 7500 }; 7501 7502 struct bpf_timer { 7503 __u64 __opaque[2]; 7504 } __attribute__((aligned(8))); 7505 7506 struct bpf_task_work { 7507 __u64 __opaque; 7508 } __attribute__((aligned(8))); 7509 7510 struct bpf_wq { 7511 __u64 __opaque[2]; 7512 } __attribute__((aligned(8))); 7513 7514 struct bpf_dynptr { 7515 __u64 __opaque[2]; 7516 } __attribute__((aligned(8))); 7517 7518 struct bpf_list_head { 7519 __u64 __opaque[2]; 7520 } __attribute__((aligned(8))); 7521 7522 struct bpf_list_node { 7523 __u64 __opaque[3]; 7524 } __attribute__((aligned(8))); 7525 7526 struct bpf_rb_root { 7527 __u64 __opaque[2]; 7528 } __attribute__((aligned(8))); 7529 7530 struct bpf_rb_node { 7531 __u64 __opaque[4]; 7532 } __attribute__((aligned(8))); 7533 7534 struct bpf_refcount { 7535 __u32 __opaque[1]; 7536 } __attribute__((aligned(4))); 7537 7538 struct bpf_sysctl { 7539 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7540 * Allows 1,2,4-byte read, but no write. 7541 */ 7542 __u32 file_pos; /* Sysctl file position to read from, write to. 7543 * Allows 1,2,4-byte read an 4-byte write. 7544 */ 7545 }; 7546 7547 struct bpf_sockopt { 7548 __bpf_md_ptr(struct bpf_sock *, sk); 7549 __bpf_md_ptr(void *, optval); 7550 __bpf_md_ptr(void *, optval_end); 7551 7552 __s32 level; 7553 __s32 optname; 7554 __s32 optlen; 7555 __s32 retval; 7556 }; 7557 7558 struct bpf_pidns_info { 7559 __u32 pid; 7560 __u32 tgid; 7561 }; 7562 7563 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7564 struct bpf_sk_lookup { 7565 union { 7566 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7567 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7568 }; 7569 7570 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7571 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7572 __u32 remote_ip4; /* Network byte order */ 7573 __u32 remote_ip6[4]; /* Network byte order */ 7574 __be16 remote_port; /* Network byte order */ 7575 __u16 :16; /* Zero padding */ 7576 __u32 local_ip4; /* Network byte order */ 7577 __u32 local_ip6[4]; /* Network byte order */ 7578 __u32 local_port; /* Host byte order */ 7579 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7580 }; 7581 7582 /* 7583 * struct btf_ptr is used for typed pointer representation; the 7584 * type id is used to render the pointer data as the appropriate type 7585 * via the bpf_snprintf_btf() helper described above. A flags field - 7586 * potentially to specify additional details about the BTF pointer 7587 * (rather than its mode of display) - is included for future use. 7588 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7589 */ 7590 struct btf_ptr { 7591 void *ptr; 7592 __u32 type_id; 7593 __u32 flags; /* BTF ptr flags; unused at present. */ 7594 }; 7595 7596 /* 7597 * Flags to control bpf_snprintf_btf() behaviour. 7598 * - BTF_F_COMPACT: no formatting around type information 7599 * - BTF_F_NONAME: no struct/union member names/types 7600 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7601 * equivalent to %px. 7602 * - BTF_F_ZERO: show zero-valued struct/union members; they 7603 * are not displayed by default 7604 */ 7605 enum { 7606 BTF_F_COMPACT = (1ULL << 0), 7607 BTF_F_NONAME = (1ULL << 1), 7608 BTF_F_PTR_RAW = (1ULL << 2), 7609 BTF_F_ZERO = (1ULL << 3), 7610 }; 7611 7612 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7613 * has to be adjusted by relocations. It is emitted by llvm and passed to 7614 * libbpf and later to the kernel. 7615 */ 7616 enum bpf_core_relo_kind { 7617 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7618 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7619 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7620 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7621 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7622 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7623 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7624 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7625 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7626 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7627 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7628 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7629 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7630 }; 7631 7632 /* 7633 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7634 * and from libbpf to the kernel. 7635 * 7636 * CO-RE relocation captures the following data: 7637 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7638 * its insn->imm field to be relocated with actual field info; 7639 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7640 * type or field; 7641 * - access_str_off - offset into corresponding .BTF string section. String 7642 * interpretation depends on specific relocation kind: 7643 * - for field-based relocations, string encodes an accessed field using 7644 * a sequence of field and array indices, separated by colon (:). It's 7645 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7646 * arguments for identifying offset to a field. 7647 * - for type-based relocations, strings is expected to be just "0"; 7648 * - for enum value-based relocations, string contains an index of enum 7649 * value within its enum type; 7650 * - kind - one of enum bpf_core_relo_kind; 7651 * 7652 * Example: 7653 * struct sample { 7654 * int a; 7655 * struct { 7656 * int b[10]; 7657 * }; 7658 * }; 7659 * 7660 * struct sample *s = ...; 7661 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7662 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7663 * // b is field #0 inside anon struct, accessing elem #5) 7664 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7665 * 7666 * type_id for all relocs in this example will capture BTF type id of 7667 * `struct sample`. 7668 * 7669 * Such relocation is emitted when using __builtin_preserve_access_index() 7670 * Clang built-in, passing expression that captures field address, e.g.: 7671 * 7672 * bpf_probe_read(&dst, sizeof(dst), 7673 * __builtin_preserve_access_index(&src->a.b.c)); 7674 * 7675 * In this case Clang will emit field relocation recording necessary data to 7676 * be able to find offset of embedded `a.b.c` field within `src` struct. 7677 * 7678 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7679 */ 7680 struct bpf_core_relo { 7681 __u32 insn_off; 7682 __u32 type_id; 7683 __u32 access_str_off; 7684 enum bpf_core_relo_kind kind; 7685 }; 7686 7687 /* 7688 * Flags to control bpf_timer_start() behaviour. 7689 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7690 * relative to current time. 7691 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7692 */ 7693 enum { 7694 BPF_F_TIMER_ABS = (1ULL << 0), 7695 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7696 }; 7697 7698 /* BPF numbers iterator state */ 7699 struct bpf_iter_num { 7700 /* opaque iterator state; having __u64 here allows to preserve correct 7701 * alignment requirements in vmlinux.h, generated from BTF 7702 */ 7703 __u64 __opaque[1]; 7704 } __attribute__((aligned(8))); 7705 7706 /* 7707 * Flags to control BPF kfunc behaviour. 7708 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective 7709 * helper documentation for details.) 7710 */ 7711 enum bpf_kfunc_flags { 7712 BPF_F_PAD_ZEROS = (1ULL << 0), 7713 }; 7714 7715 /* 7716 * Values of a BPF_MAP_TYPE_INSN_ARRAY entry must be of this type. 7717 * 7718 * Before the map is used the orig_off field should point to an 7719 * instruction inside the program being loaded. The other fields 7720 * must be set to 0. 7721 * 7722 * After the program is loaded, the xlated_off will be adjusted 7723 * by the verifier to point to the index of the original instruction 7724 * in the xlated program. If the instruction is deleted, it will 7725 * be set to (u32)-1. The jitted_off will be set to the corresponding 7726 * offset in the jitted image of the program. 7727 */ 7728 struct bpf_insn_array_value { 7729 __u32 orig_off; 7730 __u32 xlated_off; 7731 __u32 jitted_off; 7732 __u32 :32; 7733 }; 7734 7735 #endif /* _UAPI__LINUX_BPF_H__ */ 7736