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