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