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