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 */ 88 }; 89 90 union bpf_iter_link_info { 91 struct { 92 __u32 map_fd; 93 } map; 94 }; 95 96 /* BPF syscall commands, see bpf(2) man-page for more details. */ 97 /** 98 * DOC: eBPF Syscall Preamble 99 * 100 * The operation to be performed by the **bpf**\ () system call is determined 101 * by the *cmd* argument. Each operation takes an accompanying argument, 102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 103 * below). The size argument is the size of the union pointed to by *attr*. 104 */ 105 /** 106 * DOC: eBPF Syscall Commands 107 * 108 * BPF_MAP_CREATE 109 * Description 110 * Create a map and return a file descriptor that refers to the 111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 112 * is automatically enabled for the new file descriptor. 113 * 114 * Applying **close**\ (2) to the file descriptor returned by 115 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 116 * 117 * Return 118 * A new file descriptor (a nonnegative integer), or -1 if an 119 * error occurred (in which case, *errno* is set appropriately). 120 * 121 * BPF_MAP_LOOKUP_ELEM 122 * Description 123 * Look up an element with a given *key* in the map referred to 124 * by the file descriptor *map_fd*. 125 * 126 * The *flags* argument may be specified as one of the 127 * following: 128 * 129 * **BPF_F_LOCK** 130 * Look up the value of a spin-locked map without 131 * returning the lock. This must be specified if the 132 * elements contain a spinlock. 133 * 134 * Return 135 * Returns zero on success. On error, -1 is returned and *errno* 136 * is set appropriately. 137 * 138 * BPF_MAP_UPDATE_ELEM 139 * Description 140 * Create or update an element (key/value pair) in a specified map. 141 * 142 * The *flags* argument should be specified as one of the 143 * following: 144 * 145 * **BPF_ANY** 146 * Create a new element or update an existing element. 147 * **BPF_NOEXIST** 148 * Create a new element only if it did not exist. 149 * **BPF_EXIST** 150 * Update an existing element. 151 * **BPF_F_LOCK** 152 * Update a spin_lock-ed map element. 153 * 154 * Return 155 * Returns zero on success. On error, -1 is returned and *errno* 156 * is set appropriately. 157 * 158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 159 * **E2BIG**, **EEXIST**, or **ENOENT**. 160 * 161 * **E2BIG** 162 * The number of elements in the map reached the 163 * *max_entries* limit specified at map creation time. 164 * **EEXIST** 165 * If *flags* specifies **BPF_NOEXIST** and the element 166 * with *key* already exists in the map. 167 * **ENOENT** 168 * If *flags* specifies **BPF_EXIST** and the element with 169 * *key* does not exist in the map. 170 * 171 * BPF_MAP_DELETE_ELEM 172 * Description 173 * Look up and delete an element by key in a specified map. 174 * 175 * Return 176 * Returns zero on success. On error, -1 is returned and *errno* 177 * is set appropriately. 178 * 179 * BPF_MAP_GET_NEXT_KEY 180 * Description 181 * Look up an element by key in a specified map and return the key 182 * of the next element. Can be used to iterate over all elements 183 * in the map. 184 * 185 * Return 186 * Returns zero on success. On error, -1 is returned and *errno* 187 * is set appropriately. 188 * 189 * The following cases can be used to iterate over all elements of 190 * the map: 191 * 192 * * If *key* is not found, the operation returns zero and sets 193 * the *next_key* pointer to the key of the first element. 194 * * If *key* is found, the operation returns zero and sets the 195 * *next_key* pointer to the key of the next element. 196 * * If *key* is the last element, returns -1 and *errno* is set 197 * to **ENOENT**. 198 * 199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 200 * **EINVAL** on error. 201 * 202 * BPF_PROG_LOAD 203 * Description 204 * Verify and load an eBPF program, returning a new file 205 * descriptor associated with the program. 206 * 207 * Applying **close**\ (2) to the file descriptor returned by 208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 209 * 210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 211 * automatically enabled for the new file descriptor. 212 * 213 * Return 214 * A new file descriptor (a nonnegative integer), or -1 if an 215 * error occurred (in which case, *errno* is set appropriately). 216 * 217 * BPF_OBJ_PIN 218 * Description 219 * Pin an eBPF program or map referred by the specified *bpf_fd* 220 * to the provided *pathname* on the filesystem. 221 * 222 * The *pathname* argument must not contain a dot ("."). 223 * 224 * On success, *pathname* retains a reference to the eBPF object, 225 * preventing deallocation of the object when the original 226 * *bpf_fd* is closed. This allow the eBPF object to live beyond 227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 228 * process. 229 * 230 * Applying **unlink**\ (2) or similar calls to the *pathname* 231 * unpins the object from the filesystem, removing the reference. 232 * If no other file descriptors or filesystem nodes refer to the 233 * same object, it will be deallocated (see NOTES). 234 * 235 * The filesystem type for the parent directory of *pathname* must 236 * be **BPF_FS_MAGIC**. 237 * 238 * Return 239 * Returns zero on success. On error, -1 is returned and *errno* 240 * is set appropriately. 241 * 242 * BPF_OBJ_GET 243 * Description 244 * Open a file descriptor for the eBPF object pinned to the 245 * specified *pathname*. 246 * 247 * Return 248 * A new file descriptor (a nonnegative integer), or -1 if an 249 * error occurred (in which case, *errno* is set appropriately). 250 * 251 * BPF_PROG_ATTACH 252 * Description 253 * Attach an eBPF program to a *target_fd* at the specified 254 * *attach_type* hook. 255 * 256 * The *attach_type* specifies the eBPF attachment point to 257 * attach the program to, and must be one of *bpf_attach_type* 258 * (see below). 259 * 260 * The *attach_bpf_fd* must be a valid file descriptor for a 261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 262 * or sock_ops type corresponding to the specified *attach_type*. 263 * 264 * The *target_fd* must be a valid file descriptor for a kernel 265 * object which depends on the attach type of *attach_bpf_fd*: 266 * 267 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 268 * **BPF_PROG_TYPE_CGROUP_SKB**, 269 * **BPF_PROG_TYPE_CGROUP_SOCK**, 270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 273 * **BPF_PROG_TYPE_SOCK_OPS** 274 * 275 * Control Group v2 hierarchy with the eBPF controller 276 * enabled. Requires the kernel to be compiled with 277 * **CONFIG_CGROUP_BPF**. 278 * 279 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 280 * 281 * Network namespace (eg /proc/self/ns/net). 282 * 283 * **BPF_PROG_TYPE_LIRC_MODE2** 284 * 285 * LIRC device path (eg /dev/lircN). Requires the kernel 286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 287 * 288 * **BPF_PROG_TYPE_SK_SKB**, 289 * **BPF_PROG_TYPE_SK_MSG** 290 * 291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 292 * 293 * Return 294 * Returns zero on success. On error, -1 is returned and *errno* 295 * is set appropriately. 296 * 297 * BPF_PROG_DETACH 298 * Description 299 * Detach the eBPF program associated with the *target_fd* at the 300 * hook specified by *attach_type*. The program must have been 301 * previously attached using **BPF_PROG_ATTACH**. 302 * 303 * Return 304 * Returns zero on success. On error, -1 is returned and *errno* 305 * is set appropriately. 306 * 307 * BPF_PROG_TEST_RUN 308 * Description 309 * Run the eBPF program associated with the *prog_fd* a *repeat* 310 * number of times against a provided program context *ctx_in* and 311 * data *data_in*, and return the modified program context 312 * *ctx_out*, *data_out* (for example, packet data), result of the 313 * execution *retval*, and *duration* of the test run. 314 * 315 * The sizes of the buffers provided as input and output 316 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 317 * be provided in the corresponding variables *ctx_size_in*, 318 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 319 * of these parameters are not provided (ie set to NULL), the 320 * corresponding size field must be zero. 321 * 322 * Some program types have particular requirements: 323 * 324 * **BPF_PROG_TYPE_SK_LOOKUP** 325 * *data_in* and *data_out* must be NULL. 326 * 327 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 328 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 329 * 330 * *ctx_out*, *data_in* and *data_out* must be NULL. 331 * *repeat* must be zero. 332 * 333 * Return 334 * Returns zero on success. On error, -1 is returned and *errno* 335 * is set appropriately. 336 * 337 * **ENOSPC** 338 * Either *data_size_out* or *ctx_size_out* is too small. 339 * **ENOTSUPP** 340 * This command is not supported by the program type of 341 * the program referred to by *prog_fd*. 342 * 343 * BPF_PROG_GET_NEXT_ID 344 * Description 345 * Fetch the next eBPF program currently loaded into the kernel. 346 * 347 * Looks for the eBPF program with an id greater than *start_id* 348 * and updates *next_id* on success. If no other eBPF programs 349 * remain with ids higher than *start_id*, returns -1 and sets 350 * *errno* to **ENOENT**. 351 * 352 * Return 353 * Returns zero on success. On error, or when no id remains, -1 354 * is returned and *errno* is set appropriately. 355 * 356 * BPF_MAP_GET_NEXT_ID 357 * Description 358 * Fetch the next eBPF map currently loaded into the kernel. 359 * 360 * Looks for the eBPF map with an id greater than *start_id* 361 * and updates *next_id* on success. If no other eBPF maps 362 * remain with ids higher than *start_id*, returns -1 and sets 363 * *errno* to **ENOENT**. 364 * 365 * Return 366 * Returns zero on success. On error, or when no id remains, -1 367 * is returned and *errno* is set appropriately. 368 * 369 * BPF_PROG_GET_FD_BY_ID 370 * Description 371 * Open a file descriptor for the eBPF program corresponding to 372 * *prog_id*. 373 * 374 * Return 375 * A new file descriptor (a nonnegative integer), or -1 if an 376 * error occurred (in which case, *errno* is set appropriately). 377 * 378 * BPF_MAP_GET_FD_BY_ID 379 * Description 380 * Open a file descriptor for the eBPF map corresponding to 381 * *map_id*. 382 * 383 * Return 384 * A new file descriptor (a nonnegative integer), or -1 if an 385 * error occurred (in which case, *errno* is set appropriately). 386 * 387 * BPF_OBJ_GET_INFO_BY_FD 388 * Description 389 * Obtain information about the eBPF object corresponding to 390 * *bpf_fd*. 391 * 392 * Populates up to *info_len* bytes of *info*, which will be in 393 * one of the following formats depending on the eBPF object type 394 * of *bpf_fd*: 395 * 396 * * **struct bpf_prog_info** 397 * * **struct bpf_map_info** 398 * * **struct bpf_btf_info** 399 * * **struct bpf_link_info** 400 * 401 * Return 402 * Returns zero on success. On error, -1 is returned and *errno* 403 * is set appropriately. 404 * 405 * BPF_PROG_QUERY 406 * Description 407 * Obtain information about eBPF programs associated with the 408 * specified *attach_type* hook. 409 * 410 * The *target_fd* must be a valid file descriptor for a kernel 411 * object which depends on the attach type of *attach_bpf_fd*: 412 * 413 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 414 * **BPF_PROG_TYPE_CGROUP_SKB**, 415 * **BPF_PROG_TYPE_CGROUP_SOCK**, 416 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 417 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 418 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 419 * **BPF_PROG_TYPE_SOCK_OPS** 420 * 421 * Control Group v2 hierarchy with the eBPF controller 422 * enabled. Requires the kernel to be compiled with 423 * **CONFIG_CGROUP_BPF**. 424 * 425 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 426 * 427 * Network namespace (eg /proc/self/ns/net). 428 * 429 * **BPF_PROG_TYPE_LIRC_MODE2** 430 * 431 * LIRC device path (eg /dev/lircN). Requires the kernel 432 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 433 * 434 * **BPF_PROG_QUERY** always fetches the number of programs 435 * attached and the *attach_flags* which were used to attach those 436 * programs. Additionally, if *prog_ids* is nonzero and the number 437 * of attached programs is less than *prog_cnt*, populates 438 * *prog_ids* with the eBPF program ids of the programs attached 439 * at *target_fd*. 440 * 441 * The following flags may alter the result: 442 * 443 * **BPF_F_QUERY_EFFECTIVE** 444 * Only return information regarding programs which are 445 * currently effective at the specified *target_fd*. 446 * 447 * Return 448 * Returns zero on success. On error, -1 is returned and *errno* 449 * is set appropriately. 450 * 451 * BPF_RAW_TRACEPOINT_OPEN 452 * Description 453 * Attach an eBPF program to a tracepoint *name* to access kernel 454 * internal arguments of the tracepoint in their raw form. 455 * 456 * The *prog_fd* must be a valid file descriptor associated with 457 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 458 * 459 * No ABI guarantees are made about the content of tracepoint 460 * arguments exposed to the corresponding eBPF program. 461 * 462 * Applying **close**\ (2) to the file descriptor returned by 463 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 464 * 465 * Return 466 * A new file descriptor (a nonnegative integer), or -1 if an 467 * error occurred (in which case, *errno* is set appropriately). 468 * 469 * BPF_BTF_LOAD 470 * Description 471 * Verify and load BPF Type Format (BTF) metadata into the kernel, 472 * returning a new file descriptor associated with the metadata. 473 * BTF is described in more detail at 474 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 475 * 476 * The *btf* parameter must point to valid memory providing 477 * *btf_size* bytes of BTF binary metadata. 478 * 479 * The returned file descriptor can be passed to other **bpf**\ () 480 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 481 * associate the BTF with those objects. 482 * 483 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 484 * parameters to specify a *btf_log_buf*, *btf_log_size* and 485 * *btf_log_level* which allow the kernel to return freeform log 486 * output regarding the BTF verification process. 487 * 488 * Return 489 * A new file descriptor (a nonnegative integer), or -1 if an 490 * error occurred (in which case, *errno* is set appropriately). 491 * 492 * BPF_BTF_GET_FD_BY_ID 493 * Description 494 * Open a file descriptor for the BPF Type Format (BTF) 495 * corresponding to *btf_id*. 496 * 497 * Return 498 * A new file descriptor (a nonnegative integer), or -1 if an 499 * error occurred (in which case, *errno* is set appropriately). 500 * 501 * BPF_TASK_FD_QUERY 502 * Description 503 * Obtain information about eBPF programs associated with the 504 * target process identified by *pid* and *fd*. 505 * 506 * If the *pid* and *fd* are associated with a tracepoint, kprobe 507 * or uprobe perf event, then the *prog_id* and *fd_type* will 508 * be populated with the eBPF program id and file descriptor type 509 * of type **bpf_task_fd_type**. If associated with a kprobe or 510 * uprobe, the *probe_offset* and *probe_addr* will also be 511 * populated. Optionally, if *buf* is provided, then up to 512 * *buf_len* bytes of *buf* will be populated with the name of 513 * the tracepoint, kprobe or uprobe. 514 * 515 * The resulting *prog_id* may be introspected in deeper detail 516 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 517 * 518 * Return 519 * Returns zero on success. On error, -1 is returned and *errno* 520 * is set appropriately. 521 * 522 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 523 * Description 524 * Look up an element with the given *key* in the map referred to 525 * by the file descriptor *fd*, and if found, delete the element. 526 * 527 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 528 * types, the *flags* argument needs to be set to 0, but for other 529 * map types, it may be specified as: 530 * 531 * **BPF_F_LOCK** 532 * Look up and delete the value of a spin-locked map 533 * without returning the lock. This must be specified if 534 * the elements contain a spinlock. 535 * 536 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 537 * implement this command as a "pop" operation, deleting the top 538 * element rather than one corresponding to *key*. 539 * The *key* and *key_len* parameters should be zeroed when 540 * issuing this operation for these map types. 541 * 542 * This command is only valid for the following map types: 543 * * **BPF_MAP_TYPE_QUEUE** 544 * * **BPF_MAP_TYPE_STACK** 545 * * **BPF_MAP_TYPE_HASH** 546 * * **BPF_MAP_TYPE_PERCPU_HASH** 547 * * **BPF_MAP_TYPE_LRU_HASH** 548 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 549 * 550 * Return 551 * Returns zero on success. On error, -1 is returned and *errno* 552 * is set appropriately. 553 * 554 * BPF_MAP_FREEZE 555 * Description 556 * Freeze the permissions of the specified map. 557 * 558 * Write permissions may be frozen by passing zero *flags*. 559 * Upon success, no future syscall invocations may alter the 560 * map state of *map_fd*. Write operations from eBPF programs 561 * are still possible for a frozen map. 562 * 563 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 564 * 565 * Return 566 * Returns zero on success. On error, -1 is returned and *errno* 567 * is set appropriately. 568 * 569 * BPF_BTF_GET_NEXT_ID 570 * Description 571 * Fetch the next BPF Type Format (BTF) object currently loaded 572 * into the kernel. 573 * 574 * Looks for the BTF object with an id greater than *start_id* 575 * and updates *next_id* on success. If no other BTF objects 576 * remain with ids higher than *start_id*, returns -1 and sets 577 * *errno* to **ENOENT**. 578 * 579 * Return 580 * Returns zero on success. On error, or when no id remains, -1 581 * is returned and *errno* is set appropriately. 582 * 583 * BPF_MAP_LOOKUP_BATCH 584 * Description 585 * Iterate and fetch multiple elements in a map. 586 * 587 * Two opaque values are used to manage batch operations, 588 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 589 * to NULL to begin the batched operation. After each subsequent 590 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 591 * *out_batch* as the *in_batch* for the next operation to 592 * continue iteration from the current point. 593 * 594 * The *keys* and *values* are output parameters which must point 595 * to memory large enough to hold *count* items based on the key 596 * and value size of the map *map_fd*. The *keys* buffer must be 597 * of *key_size* * *count*. The *values* buffer must be of 598 * *value_size* * *count*. 599 * 600 * The *elem_flags* argument may be specified as one of the 601 * following: 602 * 603 * **BPF_F_LOCK** 604 * Look up the value of a spin-locked map without 605 * returning the lock. This must be specified if the 606 * elements contain a spinlock. 607 * 608 * On success, *count* elements from the map are copied into the 609 * user buffer, with the keys copied into *keys* and the values 610 * copied into the corresponding indices in *values*. 611 * 612 * If an error is returned and *errno* is not **EFAULT**, *count* 613 * is set to the number of successfully processed elements. 614 * 615 * Return 616 * Returns zero on success. On error, -1 is returned and *errno* 617 * is set appropriately. 618 * 619 * May set *errno* to **ENOSPC** to indicate that *keys* or 620 * *values* is too small to dump an entire bucket during 621 * iteration of a hash-based map type. 622 * 623 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 624 * Description 625 * Iterate and delete all elements in a map. 626 * 627 * This operation has the same behavior as 628 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 629 * 630 * * Every element that is successfully returned is also deleted 631 * from the map. This is at least *count* elements. Note that 632 * *count* is both an input and an output parameter. 633 * * Upon returning with *errno* set to **EFAULT**, up to 634 * *count* elements may be deleted without returning the keys 635 * and values of the deleted elements. 636 * 637 * Return 638 * Returns zero on success. On error, -1 is returned and *errno* 639 * is set appropriately. 640 * 641 * BPF_MAP_UPDATE_BATCH 642 * Description 643 * Update multiple elements in a map by *key*. 644 * 645 * The *keys* and *values* are input parameters which must point 646 * to memory large enough to hold *count* items based on the key 647 * and value size of the map *map_fd*. The *keys* buffer must be 648 * of *key_size* * *count*. The *values* buffer must be of 649 * *value_size* * *count*. 650 * 651 * Each element specified in *keys* is sequentially updated to the 652 * value in the corresponding index in *values*. The *in_batch* 653 * and *out_batch* parameters are ignored and should be zeroed. 654 * 655 * The *elem_flags* argument should be specified as one of the 656 * following: 657 * 658 * **BPF_ANY** 659 * Create new elements or update a existing elements. 660 * **BPF_NOEXIST** 661 * Create new elements only if they do not exist. 662 * **BPF_EXIST** 663 * Update existing elements. 664 * **BPF_F_LOCK** 665 * Update spin_lock-ed map elements. This must be 666 * specified if the map value contains a spinlock. 667 * 668 * On success, *count* elements from the map are updated. 669 * 670 * If an error is returned and *errno* is not **EFAULT**, *count* 671 * is set to the number of successfully processed elements. 672 * 673 * Return 674 * Returns zero on success. On error, -1 is returned and *errno* 675 * is set appropriately. 676 * 677 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 678 * **E2BIG**. **E2BIG** indicates that the number of elements in 679 * the map reached the *max_entries* limit specified at map 680 * creation time. 681 * 682 * May set *errno* to one of the following error codes under 683 * specific circumstances: 684 * 685 * **EEXIST** 686 * If *flags* specifies **BPF_NOEXIST** and the element 687 * with *key* already exists in the map. 688 * **ENOENT** 689 * If *flags* specifies **BPF_EXIST** and the element with 690 * *key* does not exist in the map. 691 * 692 * BPF_MAP_DELETE_BATCH 693 * Description 694 * Delete multiple elements in a map by *key*. 695 * 696 * The *keys* parameter is an input parameter which must point 697 * to memory large enough to hold *count* items based on the key 698 * size of the map *map_fd*, that is, *key_size* * *count*. 699 * 700 * Each element specified in *keys* is sequentially deleted. The 701 * *in_batch*, *out_batch*, and *values* parameters are ignored 702 * and should be zeroed. 703 * 704 * The *elem_flags* argument may be specified as one of the 705 * following: 706 * 707 * **BPF_F_LOCK** 708 * Look up the value of a spin-locked map without 709 * returning the lock. This must be specified if the 710 * elements contain a spinlock. 711 * 712 * On success, *count* elements from the map are updated. 713 * 714 * If an error is returned and *errno* is not **EFAULT**, *count* 715 * is set to the number of successfully processed elements. If 716 * *errno* is **EFAULT**, up to *count* elements may be been 717 * deleted. 718 * 719 * Return 720 * Returns zero on success. On error, -1 is returned and *errno* 721 * is set appropriately. 722 * 723 * BPF_LINK_CREATE 724 * Description 725 * Attach an eBPF program to a *target_fd* at the specified 726 * *attach_type* hook and return a file descriptor handle for 727 * managing the link. 728 * 729 * Return 730 * A new file descriptor (a nonnegative integer), or -1 if an 731 * error occurred (in which case, *errno* is set appropriately). 732 * 733 * BPF_LINK_UPDATE 734 * Description 735 * Update the eBPF program in the specified *link_fd* to 736 * *new_prog_fd*. 737 * 738 * Return 739 * Returns zero on success. On error, -1 is returned and *errno* 740 * is set appropriately. 741 * 742 * BPF_LINK_GET_FD_BY_ID 743 * Description 744 * Open a file descriptor for the eBPF Link corresponding to 745 * *link_id*. 746 * 747 * Return 748 * A new file descriptor (a nonnegative integer), or -1 if an 749 * error occurred (in which case, *errno* is set appropriately). 750 * 751 * BPF_LINK_GET_NEXT_ID 752 * Description 753 * Fetch the next eBPF link currently loaded into the kernel. 754 * 755 * Looks for the eBPF link with an id greater than *start_id* 756 * and updates *next_id* on success. If no other eBPF links 757 * remain with ids higher than *start_id*, returns -1 and sets 758 * *errno* to **ENOENT**. 759 * 760 * Return 761 * Returns zero on success. On error, or when no id remains, -1 762 * is returned and *errno* is set appropriately. 763 * 764 * BPF_ENABLE_STATS 765 * Description 766 * Enable eBPF runtime statistics gathering. 767 * 768 * Runtime statistics gathering for the eBPF runtime is disabled 769 * by default to minimize the corresponding performance overhead. 770 * This command enables statistics globally. 771 * 772 * Multiple programs may independently enable statistics. 773 * After gathering the desired statistics, eBPF runtime statistics 774 * may be disabled again by calling **close**\ (2) for the file 775 * descriptor returned by this function. Statistics will only be 776 * disabled system-wide when all outstanding file descriptors 777 * returned by prior calls for this subcommand are closed. 778 * 779 * Return 780 * A new file descriptor (a nonnegative integer), or -1 if an 781 * error occurred (in which case, *errno* is set appropriately). 782 * 783 * BPF_ITER_CREATE 784 * Description 785 * Create an iterator on top of the specified *link_fd* (as 786 * previously created using **BPF_LINK_CREATE**) and return a 787 * file descriptor that can be used to trigger the iteration. 788 * 789 * If the resulting file descriptor is pinned to the filesystem 790 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 791 * for that path will trigger the iterator to read kernel state 792 * using the eBPF program attached to *link_fd*. 793 * 794 * Return 795 * A new file descriptor (a nonnegative integer), or -1 if an 796 * error occurred (in which case, *errno* is set appropriately). 797 * 798 * BPF_LINK_DETACH 799 * Description 800 * Forcefully detach the specified *link_fd* from its 801 * corresponding attachment point. 802 * 803 * Return 804 * Returns zero on success. On error, -1 is returned and *errno* 805 * is set appropriately. 806 * 807 * BPF_PROG_BIND_MAP 808 * Description 809 * Bind a map to the lifetime of an eBPF program. 810 * 811 * The map identified by *map_fd* is bound to the program 812 * identified by *prog_fd* and only released when *prog_fd* is 813 * released. This may be used in cases where metadata should be 814 * associated with a program which otherwise does not contain any 815 * references to the map (for example, embedded in the eBPF 816 * program instructions). 817 * 818 * Return 819 * Returns zero on success. On error, -1 is returned and *errno* 820 * is set appropriately. 821 * 822 * NOTES 823 * eBPF objects (maps and programs) can be shared between processes. 824 * 825 * * After **fork**\ (2), the child inherits file descriptors 826 * referring to the same eBPF objects. 827 * * File descriptors referring to eBPF objects can be transferred over 828 * **unix**\ (7) domain sockets. 829 * * File descriptors referring to eBPF objects can be duplicated in the 830 * usual way, using **dup**\ (2) and similar calls. 831 * * File descriptors referring to eBPF objects can be pinned to the 832 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 833 * 834 * An eBPF object is deallocated only after all file descriptors referring 835 * to the object have been closed and no references remain pinned to the 836 * filesystem or attached (for example, bound to a program or device). 837 */ 838 enum bpf_cmd { 839 BPF_MAP_CREATE, 840 BPF_MAP_LOOKUP_ELEM, 841 BPF_MAP_UPDATE_ELEM, 842 BPF_MAP_DELETE_ELEM, 843 BPF_MAP_GET_NEXT_KEY, 844 BPF_PROG_LOAD, 845 BPF_OBJ_PIN, 846 BPF_OBJ_GET, 847 BPF_PROG_ATTACH, 848 BPF_PROG_DETACH, 849 BPF_PROG_TEST_RUN, 850 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 851 BPF_PROG_GET_NEXT_ID, 852 BPF_MAP_GET_NEXT_ID, 853 BPF_PROG_GET_FD_BY_ID, 854 BPF_MAP_GET_FD_BY_ID, 855 BPF_OBJ_GET_INFO_BY_FD, 856 BPF_PROG_QUERY, 857 BPF_RAW_TRACEPOINT_OPEN, 858 BPF_BTF_LOAD, 859 BPF_BTF_GET_FD_BY_ID, 860 BPF_TASK_FD_QUERY, 861 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 862 BPF_MAP_FREEZE, 863 BPF_BTF_GET_NEXT_ID, 864 BPF_MAP_LOOKUP_BATCH, 865 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 866 BPF_MAP_UPDATE_BATCH, 867 BPF_MAP_DELETE_BATCH, 868 BPF_LINK_CREATE, 869 BPF_LINK_UPDATE, 870 BPF_LINK_GET_FD_BY_ID, 871 BPF_LINK_GET_NEXT_ID, 872 BPF_ENABLE_STATS, 873 BPF_ITER_CREATE, 874 BPF_LINK_DETACH, 875 BPF_PROG_BIND_MAP, 876 }; 877 878 enum bpf_map_type { 879 BPF_MAP_TYPE_UNSPEC, 880 BPF_MAP_TYPE_HASH, 881 BPF_MAP_TYPE_ARRAY, 882 BPF_MAP_TYPE_PROG_ARRAY, 883 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 884 BPF_MAP_TYPE_PERCPU_HASH, 885 BPF_MAP_TYPE_PERCPU_ARRAY, 886 BPF_MAP_TYPE_STACK_TRACE, 887 BPF_MAP_TYPE_CGROUP_ARRAY, 888 BPF_MAP_TYPE_LRU_HASH, 889 BPF_MAP_TYPE_LRU_PERCPU_HASH, 890 BPF_MAP_TYPE_LPM_TRIE, 891 BPF_MAP_TYPE_ARRAY_OF_MAPS, 892 BPF_MAP_TYPE_HASH_OF_MAPS, 893 BPF_MAP_TYPE_DEVMAP, 894 BPF_MAP_TYPE_SOCKMAP, 895 BPF_MAP_TYPE_CPUMAP, 896 BPF_MAP_TYPE_XSKMAP, 897 BPF_MAP_TYPE_SOCKHASH, 898 BPF_MAP_TYPE_CGROUP_STORAGE, 899 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 900 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 901 BPF_MAP_TYPE_QUEUE, 902 BPF_MAP_TYPE_STACK, 903 BPF_MAP_TYPE_SK_STORAGE, 904 BPF_MAP_TYPE_DEVMAP_HASH, 905 BPF_MAP_TYPE_STRUCT_OPS, 906 BPF_MAP_TYPE_RINGBUF, 907 BPF_MAP_TYPE_INODE_STORAGE, 908 BPF_MAP_TYPE_TASK_STORAGE, 909 }; 910 911 /* Note that tracing related programs such as 912 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 913 * are not subject to a stable API since kernel internal data 914 * structures can change from release to release and may 915 * therefore break existing tracing BPF programs. Tracing BPF 916 * programs correspond to /a/ specific kernel which is to be 917 * analyzed, and not /a/ specific kernel /and/ all future ones. 918 */ 919 enum bpf_prog_type { 920 BPF_PROG_TYPE_UNSPEC, 921 BPF_PROG_TYPE_SOCKET_FILTER, 922 BPF_PROG_TYPE_KPROBE, 923 BPF_PROG_TYPE_SCHED_CLS, 924 BPF_PROG_TYPE_SCHED_ACT, 925 BPF_PROG_TYPE_TRACEPOINT, 926 BPF_PROG_TYPE_XDP, 927 BPF_PROG_TYPE_PERF_EVENT, 928 BPF_PROG_TYPE_CGROUP_SKB, 929 BPF_PROG_TYPE_CGROUP_SOCK, 930 BPF_PROG_TYPE_LWT_IN, 931 BPF_PROG_TYPE_LWT_OUT, 932 BPF_PROG_TYPE_LWT_XMIT, 933 BPF_PROG_TYPE_SOCK_OPS, 934 BPF_PROG_TYPE_SK_SKB, 935 BPF_PROG_TYPE_CGROUP_DEVICE, 936 BPF_PROG_TYPE_SK_MSG, 937 BPF_PROG_TYPE_RAW_TRACEPOINT, 938 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 939 BPF_PROG_TYPE_LWT_SEG6LOCAL, 940 BPF_PROG_TYPE_LIRC_MODE2, 941 BPF_PROG_TYPE_SK_REUSEPORT, 942 BPF_PROG_TYPE_FLOW_DISSECTOR, 943 BPF_PROG_TYPE_CGROUP_SYSCTL, 944 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 945 BPF_PROG_TYPE_CGROUP_SOCKOPT, 946 BPF_PROG_TYPE_TRACING, 947 BPF_PROG_TYPE_STRUCT_OPS, 948 BPF_PROG_TYPE_EXT, 949 BPF_PROG_TYPE_LSM, 950 BPF_PROG_TYPE_SK_LOOKUP, 951 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 952 }; 953 954 enum bpf_attach_type { 955 BPF_CGROUP_INET_INGRESS, 956 BPF_CGROUP_INET_EGRESS, 957 BPF_CGROUP_INET_SOCK_CREATE, 958 BPF_CGROUP_SOCK_OPS, 959 BPF_SK_SKB_STREAM_PARSER, 960 BPF_SK_SKB_STREAM_VERDICT, 961 BPF_CGROUP_DEVICE, 962 BPF_SK_MSG_VERDICT, 963 BPF_CGROUP_INET4_BIND, 964 BPF_CGROUP_INET6_BIND, 965 BPF_CGROUP_INET4_CONNECT, 966 BPF_CGROUP_INET6_CONNECT, 967 BPF_CGROUP_INET4_POST_BIND, 968 BPF_CGROUP_INET6_POST_BIND, 969 BPF_CGROUP_UDP4_SENDMSG, 970 BPF_CGROUP_UDP6_SENDMSG, 971 BPF_LIRC_MODE2, 972 BPF_FLOW_DISSECTOR, 973 BPF_CGROUP_SYSCTL, 974 BPF_CGROUP_UDP4_RECVMSG, 975 BPF_CGROUP_UDP6_RECVMSG, 976 BPF_CGROUP_GETSOCKOPT, 977 BPF_CGROUP_SETSOCKOPT, 978 BPF_TRACE_RAW_TP, 979 BPF_TRACE_FENTRY, 980 BPF_TRACE_FEXIT, 981 BPF_MODIFY_RETURN, 982 BPF_LSM_MAC, 983 BPF_TRACE_ITER, 984 BPF_CGROUP_INET4_GETPEERNAME, 985 BPF_CGROUP_INET6_GETPEERNAME, 986 BPF_CGROUP_INET4_GETSOCKNAME, 987 BPF_CGROUP_INET6_GETSOCKNAME, 988 BPF_XDP_DEVMAP, 989 BPF_CGROUP_INET_SOCK_RELEASE, 990 BPF_XDP_CPUMAP, 991 BPF_SK_LOOKUP, 992 BPF_XDP, 993 BPF_SK_SKB_VERDICT, 994 BPF_SK_REUSEPORT_SELECT, 995 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 996 __MAX_BPF_ATTACH_TYPE 997 }; 998 999 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1000 1001 enum bpf_link_type { 1002 BPF_LINK_TYPE_UNSPEC = 0, 1003 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1004 BPF_LINK_TYPE_TRACING = 2, 1005 BPF_LINK_TYPE_CGROUP = 3, 1006 BPF_LINK_TYPE_ITER = 4, 1007 BPF_LINK_TYPE_NETNS = 5, 1008 BPF_LINK_TYPE_XDP = 6, 1009 1010 MAX_BPF_LINK_TYPE, 1011 }; 1012 1013 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1014 * 1015 * NONE(default): No further bpf programs allowed in the subtree. 1016 * 1017 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1018 * the program in this cgroup yields to sub-cgroup program. 1019 * 1020 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1021 * that cgroup program gets run in addition to the program in this cgroup. 1022 * 1023 * Only one program is allowed to be attached to a cgroup with 1024 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1025 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1026 * release old program and attach the new one. Attach flags has to match. 1027 * 1028 * Multiple programs are allowed to be attached to a cgroup with 1029 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1030 * (those that were attached first, run first) 1031 * The programs of sub-cgroup are executed first, then programs of 1032 * this cgroup and then programs of parent cgroup. 1033 * When children program makes decision (like picking TCP CA or sock bind) 1034 * parent program has a chance to override it. 1035 * 1036 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1037 * programs for a cgroup. Though it's possible to replace an old program at 1038 * any position by also specifying BPF_F_REPLACE flag and position itself in 1039 * replace_bpf_fd attribute. Old program at this position will be released. 1040 * 1041 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1042 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1043 * Ex1: 1044 * cgrp1 (MULTI progs A, B) -> 1045 * cgrp2 (OVERRIDE prog C) -> 1046 * cgrp3 (MULTI prog D) -> 1047 * cgrp4 (OVERRIDE prog E) -> 1048 * cgrp5 (NONE prog F) 1049 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1050 * if prog F is detached, the execution is E,D,A,B 1051 * if prog F and D are detached, the execution is E,A,B 1052 * if prog F, E and D are detached, the execution is C,A,B 1053 * 1054 * All eligible programs are executed regardless of return code from 1055 * earlier programs. 1056 */ 1057 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1058 #define BPF_F_ALLOW_MULTI (1U << 1) 1059 #define BPF_F_REPLACE (1U << 2) 1060 1061 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1062 * verifier will perform strict alignment checking as if the kernel 1063 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1064 * and NET_IP_ALIGN defined to 2. 1065 */ 1066 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1067 1068 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1069 * verifier will allow any alignment whatsoever. On platforms 1070 * with strict alignment requirements for loads ands stores (such 1071 * as sparc and mips) the verifier validates that all loads and 1072 * stores provably follow this requirement. This flag turns that 1073 * checking and enforcement off. 1074 * 1075 * It is mostly used for testing when we want to validate the 1076 * context and memory access aspects of the verifier, but because 1077 * of an unaligned access the alignment check would trigger before 1078 * the one we are interested in. 1079 */ 1080 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1081 1082 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1083 * Verifier does sub-register def/use analysis and identifies instructions whose 1084 * def only matters for low 32-bit, high 32-bit is never referenced later 1085 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1086 * that it is safe to ignore clearing high 32-bit for these instructions. This 1087 * saves some back-ends a lot of code-gen. However such optimization is not 1088 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1089 * hence hasn't used verifier's analysis result. But, we really want to have a 1090 * way to be able to verify the correctness of the described optimization on 1091 * x86_64 on which testsuites are frequently exercised. 1092 * 1093 * So, this flag is introduced. Once it is set, verifier will randomize high 1094 * 32-bit for those instructions who has been identified as safe to ignore them. 1095 * Then, if verifier is not doing correct analysis, such randomization will 1096 * regress tests to expose bugs. 1097 */ 1098 #define BPF_F_TEST_RND_HI32 (1U << 2) 1099 1100 /* The verifier internal test flag. Behavior is undefined */ 1101 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1102 1103 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1104 * restrict map and helper usage for such programs. Sleepable BPF programs can 1105 * only be attached to hooks where kernel execution context allows sleeping. 1106 * Such programs are allowed to use helpers that may sleep like 1107 * bpf_copy_from_user(). 1108 */ 1109 #define BPF_F_SLEEPABLE (1U << 4) 1110 1111 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1112 * the following extensions: 1113 * 1114 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1115 * insn[0].imm: map fd or fd_idx 1116 * insn[1].imm: 0 1117 * insn[0].off: 0 1118 * insn[1].off: 0 1119 * ldimm64 rewrite: address of map 1120 * verifier type: CONST_PTR_TO_MAP 1121 */ 1122 #define BPF_PSEUDO_MAP_FD 1 1123 #define BPF_PSEUDO_MAP_IDX 5 1124 1125 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1126 * insn[0].imm: map fd or fd_idx 1127 * insn[1].imm: offset into value 1128 * insn[0].off: 0 1129 * insn[1].off: 0 1130 * ldimm64 rewrite: address of map[0]+offset 1131 * verifier type: PTR_TO_MAP_VALUE 1132 */ 1133 #define BPF_PSEUDO_MAP_VALUE 2 1134 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1135 1136 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1137 * insn[0].imm: kernel btd id of VAR 1138 * insn[1].imm: 0 1139 * insn[0].off: 0 1140 * insn[1].off: 0 1141 * ldimm64 rewrite: address of the kernel variable 1142 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1143 * is struct/union. 1144 */ 1145 #define BPF_PSEUDO_BTF_ID 3 1146 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1147 * insn[0].imm: insn offset to the func 1148 * insn[1].imm: 0 1149 * insn[0].off: 0 1150 * insn[1].off: 0 1151 * ldimm64 rewrite: address of the function 1152 * verifier type: PTR_TO_FUNC. 1153 */ 1154 #define BPF_PSEUDO_FUNC 4 1155 1156 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1157 * offset to another bpf function 1158 */ 1159 #define BPF_PSEUDO_CALL 1 1160 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1161 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1162 */ 1163 #define BPF_PSEUDO_KFUNC_CALL 2 1164 1165 /* flags for BPF_MAP_UPDATE_ELEM command */ 1166 enum { 1167 BPF_ANY = 0, /* create new element or update existing */ 1168 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1169 BPF_EXIST = 2, /* update existing element */ 1170 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1171 }; 1172 1173 /* flags for BPF_MAP_CREATE command */ 1174 enum { 1175 BPF_F_NO_PREALLOC = (1U << 0), 1176 /* Instead of having one common LRU list in the 1177 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1178 * which can scale and perform better. 1179 * Note, the LRU nodes (including free nodes) cannot be moved 1180 * across different LRU lists. 1181 */ 1182 BPF_F_NO_COMMON_LRU = (1U << 1), 1183 /* Specify numa node during map creation */ 1184 BPF_F_NUMA_NODE = (1U << 2), 1185 1186 /* Flags for accessing BPF object from syscall side. */ 1187 BPF_F_RDONLY = (1U << 3), 1188 BPF_F_WRONLY = (1U << 4), 1189 1190 /* Flag for stack_map, store build_id+offset instead of pointer */ 1191 BPF_F_STACK_BUILD_ID = (1U << 5), 1192 1193 /* Zero-initialize hash function seed. This should only be used for testing. */ 1194 BPF_F_ZERO_SEED = (1U << 6), 1195 1196 /* Flags for accessing BPF object from program side. */ 1197 BPF_F_RDONLY_PROG = (1U << 7), 1198 BPF_F_WRONLY_PROG = (1U << 8), 1199 1200 /* Clone map from listener for newly accepted socket */ 1201 BPF_F_CLONE = (1U << 9), 1202 1203 /* Enable memory-mapping BPF map */ 1204 BPF_F_MMAPABLE = (1U << 10), 1205 1206 /* Share perf_event among processes */ 1207 BPF_F_PRESERVE_ELEMS = (1U << 11), 1208 1209 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1210 BPF_F_INNER_MAP = (1U << 12), 1211 }; 1212 1213 /* Flags for BPF_PROG_QUERY. */ 1214 1215 /* Query effective (directly attached + inherited from ancestor cgroups) 1216 * programs that will be executed for events within a cgroup. 1217 * attach_flags with this flag are returned only for directly attached programs. 1218 */ 1219 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1220 1221 /* Flags for BPF_PROG_TEST_RUN */ 1222 1223 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1224 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1225 1226 /* type for BPF_ENABLE_STATS */ 1227 enum bpf_stats_type { 1228 /* enabled run_time_ns and run_cnt */ 1229 BPF_STATS_RUN_TIME = 0, 1230 }; 1231 1232 enum bpf_stack_build_id_status { 1233 /* user space need an empty entry to identify end of a trace */ 1234 BPF_STACK_BUILD_ID_EMPTY = 0, 1235 /* with valid build_id and offset */ 1236 BPF_STACK_BUILD_ID_VALID = 1, 1237 /* couldn't get build_id, fallback to ip */ 1238 BPF_STACK_BUILD_ID_IP = 2, 1239 }; 1240 1241 #define BPF_BUILD_ID_SIZE 20 1242 struct bpf_stack_build_id { 1243 __s32 status; 1244 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1245 union { 1246 __u64 offset; 1247 __u64 ip; 1248 }; 1249 }; 1250 1251 #define BPF_OBJ_NAME_LEN 16U 1252 1253 union bpf_attr { 1254 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1255 __u32 map_type; /* one of enum bpf_map_type */ 1256 __u32 key_size; /* size of key in bytes */ 1257 __u32 value_size; /* size of value in bytes */ 1258 __u32 max_entries; /* max number of entries in a map */ 1259 __u32 map_flags; /* BPF_MAP_CREATE related 1260 * flags defined above. 1261 */ 1262 __u32 inner_map_fd; /* fd pointing to the inner map */ 1263 __u32 numa_node; /* numa node (effective only if 1264 * BPF_F_NUMA_NODE is set). 1265 */ 1266 char map_name[BPF_OBJ_NAME_LEN]; 1267 __u32 map_ifindex; /* ifindex of netdev to create on */ 1268 __u32 btf_fd; /* fd pointing to a BTF type data */ 1269 __u32 btf_key_type_id; /* BTF type_id of the key */ 1270 __u32 btf_value_type_id; /* BTF type_id of the value */ 1271 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1272 * struct stored as the 1273 * map value 1274 */ 1275 }; 1276 1277 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1278 __u32 map_fd; 1279 __aligned_u64 key; 1280 union { 1281 __aligned_u64 value; 1282 __aligned_u64 next_key; 1283 }; 1284 __u64 flags; 1285 }; 1286 1287 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1288 __aligned_u64 in_batch; /* start batch, 1289 * NULL to start from beginning 1290 */ 1291 __aligned_u64 out_batch; /* output: next start batch */ 1292 __aligned_u64 keys; 1293 __aligned_u64 values; 1294 __u32 count; /* input/output: 1295 * input: # of key/value 1296 * elements 1297 * output: # of filled elements 1298 */ 1299 __u32 map_fd; 1300 __u64 elem_flags; 1301 __u64 flags; 1302 } batch; 1303 1304 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1305 __u32 prog_type; /* one of enum bpf_prog_type */ 1306 __u32 insn_cnt; 1307 __aligned_u64 insns; 1308 __aligned_u64 license; 1309 __u32 log_level; /* verbosity level of verifier */ 1310 __u32 log_size; /* size of user buffer */ 1311 __aligned_u64 log_buf; /* user supplied buffer */ 1312 __u32 kern_version; /* not used */ 1313 __u32 prog_flags; 1314 char prog_name[BPF_OBJ_NAME_LEN]; 1315 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1316 /* For some prog types expected attach type must be known at 1317 * load time to verify attach type specific parts of prog 1318 * (context accesses, allowed helpers, etc). 1319 */ 1320 __u32 expected_attach_type; 1321 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1322 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1323 __aligned_u64 func_info; /* func info */ 1324 __u32 func_info_cnt; /* number of bpf_func_info records */ 1325 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1326 __aligned_u64 line_info; /* line info */ 1327 __u32 line_info_cnt; /* number of bpf_line_info records */ 1328 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1329 union { 1330 /* valid prog_fd to attach to bpf prog */ 1331 __u32 attach_prog_fd; 1332 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1333 __u32 attach_btf_obj_fd; 1334 }; 1335 __u32 :32; /* pad */ 1336 __aligned_u64 fd_array; /* array of FDs */ 1337 }; 1338 1339 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1340 __aligned_u64 pathname; 1341 __u32 bpf_fd; 1342 __u32 file_flags; 1343 }; 1344 1345 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1346 __u32 target_fd; /* container object to attach to */ 1347 __u32 attach_bpf_fd; /* eBPF program to attach */ 1348 __u32 attach_type; 1349 __u32 attach_flags; 1350 __u32 replace_bpf_fd; /* previously attached eBPF 1351 * program to replace if 1352 * BPF_F_REPLACE is used 1353 */ 1354 }; 1355 1356 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1357 __u32 prog_fd; 1358 __u32 retval; 1359 __u32 data_size_in; /* input: len of data_in */ 1360 __u32 data_size_out; /* input/output: len of data_out 1361 * returns ENOSPC if data_out 1362 * is too small. 1363 */ 1364 __aligned_u64 data_in; 1365 __aligned_u64 data_out; 1366 __u32 repeat; 1367 __u32 duration; 1368 __u32 ctx_size_in; /* input: len of ctx_in */ 1369 __u32 ctx_size_out; /* input/output: len of ctx_out 1370 * returns ENOSPC if ctx_out 1371 * is too small. 1372 */ 1373 __aligned_u64 ctx_in; 1374 __aligned_u64 ctx_out; 1375 __u32 flags; 1376 __u32 cpu; 1377 } test; 1378 1379 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1380 union { 1381 __u32 start_id; 1382 __u32 prog_id; 1383 __u32 map_id; 1384 __u32 btf_id; 1385 __u32 link_id; 1386 }; 1387 __u32 next_id; 1388 __u32 open_flags; 1389 }; 1390 1391 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1392 __u32 bpf_fd; 1393 __u32 info_len; 1394 __aligned_u64 info; 1395 } info; 1396 1397 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1398 __u32 target_fd; /* container object to query */ 1399 __u32 attach_type; 1400 __u32 query_flags; 1401 __u32 attach_flags; 1402 __aligned_u64 prog_ids; 1403 __u32 prog_cnt; 1404 } query; 1405 1406 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1407 __u64 name; 1408 __u32 prog_fd; 1409 } raw_tracepoint; 1410 1411 struct { /* anonymous struct for BPF_BTF_LOAD */ 1412 __aligned_u64 btf; 1413 __aligned_u64 btf_log_buf; 1414 __u32 btf_size; 1415 __u32 btf_log_size; 1416 __u32 btf_log_level; 1417 }; 1418 1419 struct { 1420 __u32 pid; /* input: pid */ 1421 __u32 fd; /* input: fd */ 1422 __u32 flags; /* input: flags */ 1423 __u32 buf_len; /* input/output: buf len */ 1424 __aligned_u64 buf; /* input/output: 1425 * tp_name for tracepoint 1426 * symbol for kprobe 1427 * filename for uprobe 1428 */ 1429 __u32 prog_id; /* output: prod_id */ 1430 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1431 __u64 probe_offset; /* output: probe_offset */ 1432 __u64 probe_addr; /* output: probe_addr */ 1433 } task_fd_query; 1434 1435 struct { /* struct used by BPF_LINK_CREATE command */ 1436 __u32 prog_fd; /* eBPF program to attach */ 1437 union { 1438 __u32 target_fd; /* object to attach to */ 1439 __u32 target_ifindex; /* target ifindex */ 1440 }; 1441 __u32 attach_type; /* attach type */ 1442 __u32 flags; /* extra flags */ 1443 union { 1444 __u32 target_btf_id; /* btf_id of target to attach to */ 1445 struct { 1446 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1447 __u32 iter_info_len; /* iter_info length */ 1448 }; 1449 }; 1450 } link_create; 1451 1452 struct { /* struct used by BPF_LINK_UPDATE command */ 1453 __u32 link_fd; /* link fd */ 1454 /* new program fd to update link with */ 1455 __u32 new_prog_fd; 1456 __u32 flags; /* extra flags */ 1457 /* expected link's program fd; is specified only if 1458 * BPF_F_REPLACE flag is set in flags */ 1459 __u32 old_prog_fd; 1460 } link_update; 1461 1462 struct { 1463 __u32 link_fd; 1464 } link_detach; 1465 1466 struct { /* struct used by BPF_ENABLE_STATS command */ 1467 __u32 type; 1468 } enable_stats; 1469 1470 struct { /* struct used by BPF_ITER_CREATE command */ 1471 __u32 link_fd; 1472 __u32 flags; 1473 } iter_create; 1474 1475 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1476 __u32 prog_fd; 1477 __u32 map_fd; 1478 __u32 flags; /* extra flags */ 1479 } prog_bind_map; 1480 1481 } __attribute__((aligned(8))); 1482 1483 /* The description below is an attempt at providing documentation to eBPF 1484 * developers about the multiple available eBPF helper functions. It can be 1485 * parsed and used to produce a manual page. The workflow is the following, 1486 * and requires the rst2man utility: 1487 * 1488 * $ ./scripts/bpf_doc.py \ 1489 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1490 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1491 * $ man /tmp/bpf-helpers.7 1492 * 1493 * Note that in order to produce this external documentation, some RST 1494 * formatting is used in the descriptions to get "bold" and "italics" in 1495 * manual pages. Also note that the few trailing white spaces are 1496 * intentional, removing them would break paragraphs for rst2man. 1497 * 1498 * Start of BPF helper function descriptions: 1499 * 1500 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1501 * Description 1502 * Perform a lookup in *map* for an entry associated to *key*. 1503 * Return 1504 * Map value associated to *key*, or **NULL** if no entry was 1505 * found. 1506 * 1507 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1508 * Description 1509 * Add or update the value of the entry associated to *key* in 1510 * *map* with *value*. *flags* is one of: 1511 * 1512 * **BPF_NOEXIST** 1513 * The entry for *key* must not exist in the map. 1514 * **BPF_EXIST** 1515 * The entry for *key* must already exist in the map. 1516 * **BPF_ANY** 1517 * No condition on the existence of the entry for *key*. 1518 * 1519 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1520 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1521 * elements always exist), the helper would return an error. 1522 * Return 1523 * 0 on success, or a negative error in case of failure. 1524 * 1525 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1526 * Description 1527 * Delete entry with *key* from *map*. 1528 * Return 1529 * 0 on success, or a negative error in case of failure. 1530 * 1531 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1532 * Description 1533 * For tracing programs, safely attempt to read *size* bytes from 1534 * kernel space address *unsafe_ptr* and store the data in *dst*. 1535 * 1536 * Generally, use **bpf_probe_read_user**\ () or 1537 * **bpf_probe_read_kernel**\ () instead. 1538 * Return 1539 * 0 on success, or a negative error in case of failure. 1540 * 1541 * u64 bpf_ktime_get_ns(void) 1542 * Description 1543 * Return the time elapsed since system boot, in nanoseconds. 1544 * Does not include time the system was suspended. 1545 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1546 * Return 1547 * Current *ktime*. 1548 * 1549 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1550 * Description 1551 * This helper is a "printk()-like" facility for debugging. It 1552 * prints a message defined by format *fmt* (of size *fmt_size*) 1553 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1554 * available. It can take up to three additional **u64** 1555 * arguments (as an eBPF helpers, the total number of arguments is 1556 * limited to five). 1557 * 1558 * Each time the helper is called, it appends a line to the trace. 1559 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 1560 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1561 * The format of the trace is customizable, and the exact output 1562 * one will get depends on the options set in 1563 * *\/sys/kernel/debug/tracing/trace_options* (see also the 1564 * *README* file under the same directory). However, it usually 1565 * defaults to something like: 1566 * 1567 * :: 1568 * 1569 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1570 * 1571 * In the above: 1572 * 1573 * * ``telnet`` is the name of the current task. 1574 * * ``470`` is the PID of the current task. 1575 * * ``001`` is the CPU number on which the task is 1576 * running. 1577 * * In ``.N..``, each character refers to a set of 1578 * options (whether irqs are enabled, scheduling 1579 * options, whether hard/softirqs are running, level of 1580 * preempt_disabled respectively). **N** means that 1581 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1582 * are set. 1583 * * ``419421.045894`` is a timestamp. 1584 * * ``0x00000001`` is a fake value used by BPF for the 1585 * instruction pointer register. 1586 * * ``<formatted msg>`` is the message formatted with 1587 * *fmt*. 1588 * 1589 * The conversion specifiers supported by *fmt* are similar, but 1590 * more limited than for printk(). They are **%d**, **%i**, 1591 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1592 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1593 * of field, padding with zeroes, etc.) is available, and the 1594 * helper will return **-EINVAL** (but print nothing) if it 1595 * encounters an unknown specifier. 1596 * 1597 * Also, note that **bpf_trace_printk**\ () is slow, and should 1598 * only be used for debugging purposes. For this reason, a notice 1599 * block (spanning several lines) is printed to kernel logs and 1600 * states that the helper should not be used "for production use" 1601 * the first time this helper is used (or more precisely, when 1602 * **trace_printk**\ () buffers are allocated). For passing values 1603 * to user space, perf events should be preferred. 1604 * Return 1605 * The number of bytes written to the buffer, or a negative error 1606 * in case of failure. 1607 * 1608 * u32 bpf_get_prandom_u32(void) 1609 * Description 1610 * Get a pseudo-random number. 1611 * 1612 * From a security point of view, this helper uses its own 1613 * pseudo-random internal state, and cannot be used to infer the 1614 * seed of other random functions in the kernel. However, it is 1615 * essential to note that the generator used by the helper is not 1616 * cryptographically secure. 1617 * Return 1618 * A random 32-bit unsigned value. 1619 * 1620 * u32 bpf_get_smp_processor_id(void) 1621 * Description 1622 * Get the SMP (symmetric multiprocessing) processor id. Note that 1623 * all programs run with preemption disabled, which means that the 1624 * SMP processor id is stable during all the execution of the 1625 * program. 1626 * Return 1627 * The SMP id of the processor running the program. 1628 * 1629 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1630 * Description 1631 * Store *len* bytes from address *from* into the packet 1632 * associated to *skb*, at *offset*. *flags* are a combination of 1633 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1634 * checksum for the packet after storing the bytes) and 1635 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1636 * **->swhash** and *skb*\ **->l4hash** to 0). 1637 * 1638 * A call to this helper is susceptible to change the underlying 1639 * packet buffer. Therefore, at load time, all checks on pointers 1640 * previously done by the verifier are invalidated and must be 1641 * performed again, if the helper is used in combination with 1642 * direct packet access. 1643 * Return 1644 * 0 on success, or a negative error in case of failure. 1645 * 1646 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1647 * Description 1648 * Recompute the layer 3 (e.g. IP) checksum for the packet 1649 * associated to *skb*. Computation is incremental, so the helper 1650 * must know the former value of the header field that was 1651 * modified (*from*), the new value of this field (*to*), and the 1652 * number of bytes (2 or 4) for this field, stored in *size*. 1653 * Alternatively, it is possible to store the difference between 1654 * the previous and the new values of the header field in *to*, by 1655 * setting *from* and *size* to 0. For both methods, *offset* 1656 * indicates the location of the IP checksum within the packet. 1657 * 1658 * This helper works in combination with **bpf_csum_diff**\ (), 1659 * which does not update the checksum in-place, but offers more 1660 * flexibility and can handle sizes larger than 2 or 4 for the 1661 * checksum to update. 1662 * 1663 * A call to this helper is susceptible to change the underlying 1664 * packet buffer. Therefore, at load time, all checks on pointers 1665 * previously done by the verifier are invalidated and must be 1666 * performed again, if the helper is used in combination with 1667 * direct packet access. 1668 * Return 1669 * 0 on success, or a negative error in case of failure. 1670 * 1671 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1672 * Description 1673 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1674 * packet associated to *skb*. Computation is incremental, so the 1675 * helper must know the former value of the header field that was 1676 * modified (*from*), the new value of this field (*to*), and the 1677 * number of bytes (2 or 4) for this field, stored on the lowest 1678 * four bits of *flags*. Alternatively, it is possible to store 1679 * the difference between the previous and the new values of the 1680 * header field in *to*, by setting *from* and the four lowest 1681 * bits of *flags* to 0. For both methods, *offset* indicates the 1682 * location of the IP checksum within the packet. In addition to 1683 * the size of the field, *flags* can be added (bitwise OR) actual 1684 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1685 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1686 * for updates resulting in a null checksum the value is set to 1687 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1688 * the checksum is to be computed against a pseudo-header. 1689 * 1690 * This helper works in combination with **bpf_csum_diff**\ (), 1691 * which does not update the checksum in-place, but offers more 1692 * flexibility and can handle sizes larger than 2 or 4 for the 1693 * checksum to update. 1694 * 1695 * A call to this helper is susceptible to change the underlying 1696 * packet buffer. Therefore, at load time, all checks on pointers 1697 * previously done by the verifier are invalidated and must be 1698 * performed again, if the helper is used in combination with 1699 * direct packet access. 1700 * Return 1701 * 0 on success, or a negative error in case of failure. 1702 * 1703 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1704 * Description 1705 * This special helper is used to trigger a "tail call", or in 1706 * other words, to jump into another eBPF program. The same stack 1707 * frame is used (but values on stack and in registers for the 1708 * caller are not accessible to the callee). This mechanism allows 1709 * for program chaining, either for raising the maximum number of 1710 * available eBPF instructions, or to execute given programs in 1711 * conditional blocks. For security reasons, there is an upper 1712 * limit to the number of successive tail calls that can be 1713 * performed. 1714 * 1715 * Upon call of this helper, the program attempts to jump into a 1716 * program referenced at index *index* in *prog_array_map*, a 1717 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1718 * *ctx*, a pointer to the context. 1719 * 1720 * If the call succeeds, the kernel immediately runs the first 1721 * instruction of the new program. This is not a function call, 1722 * and it never returns to the previous program. If the call 1723 * fails, then the helper has no effect, and the caller continues 1724 * to run its subsequent instructions. A call can fail if the 1725 * destination program for the jump does not exist (i.e. *index* 1726 * is superior to the number of entries in *prog_array_map*), or 1727 * if the maximum number of tail calls has been reached for this 1728 * chain of programs. This limit is defined in the kernel by the 1729 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1730 * which is currently set to 32. 1731 * Return 1732 * 0 on success, or a negative error in case of failure. 1733 * 1734 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1735 * Description 1736 * Clone and redirect the packet associated to *skb* to another 1737 * net device of index *ifindex*. Both ingress and egress 1738 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1739 * value in *flags* is used to make the distinction (ingress path 1740 * is selected if the flag is present, egress path otherwise). 1741 * This is the only flag supported for now. 1742 * 1743 * In comparison with **bpf_redirect**\ () helper, 1744 * **bpf_clone_redirect**\ () has the associated cost of 1745 * duplicating the packet buffer, but this can be executed out of 1746 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1747 * efficient, but it is handled through an action code where the 1748 * redirection happens only after the eBPF program has returned. 1749 * 1750 * A call to this helper is susceptible to change the underlying 1751 * packet buffer. Therefore, at load time, all checks on pointers 1752 * previously done by the verifier are invalidated and must be 1753 * performed again, if the helper is used in combination with 1754 * direct packet access. 1755 * Return 1756 * 0 on success, or a negative error in case of failure. 1757 * 1758 * u64 bpf_get_current_pid_tgid(void) 1759 * Return 1760 * A 64-bit integer containing the current tgid and pid, and 1761 * created as such: 1762 * *current_task*\ **->tgid << 32 \|** 1763 * *current_task*\ **->pid**. 1764 * 1765 * u64 bpf_get_current_uid_gid(void) 1766 * Return 1767 * A 64-bit integer containing the current GID and UID, and 1768 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1769 * 1770 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1771 * Description 1772 * Copy the **comm** attribute of the current task into *buf* of 1773 * *size_of_buf*. The **comm** attribute contains the name of 1774 * the executable (excluding the path) for the current task. The 1775 * *size_of_buf* must be strictly positive. On success, the 1776 * helper makes sure that the *buf* is NUL-terminated. On failure, 1777 * it is filled with zeroes. 1778 * Return 1779 * 0 on success, or a negative error in case of failure. 1780 * 1781 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1782 * Description 1783 * Retrieve the classid for the current task, i.e. for the net_cls 1784 * cgroup to which *skb* belongs. 1785 * 1786 * This helper can be used on TC egress path, but not on ingress. 1787 * 1788 * The net_cls cgroup provides an interface to tag network packets 1789 * based on a user-provided identifier for all traffic coming from 1790 * the tasks belonging to the related cgroup. See also the related 1791 * kernel documentation, available from the Linux sources in file 1792 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1793 * 1794 * The Linux kernel has two versions for cgroups: there are 1795 * cgroups v1 and cgroups v2. Both are available to users, who can 1796 * use a mixture of them, but note that the net_cls cgroup is for 1797 * cgroup v1 only. This makes it incompatible with BPF programs 1798 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1799 * only hold data for one version of cgroups at a time). 1800 * 1801 * This helper is only available is the kernel was compiled with 1802 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1803 * "**y**" or to "**m**". 1804 * Return 1805 * The classid, or 0 for the default unconfigured classid. 1806 * 1807 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1808 * Description 1809 * Push a *vlan_tci* (VLAN tag control information) of protocol 1810 * *vlan_proto* to the packet associated to *skb*, then update 1811 * the checksum. Note that if *vlan_proto* is different from 1812 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1813 * be **ETH_P_8021Q**. 1814 * 1815 * A call to this helper is susceptible to change the underlying 1816 * packet buffer. Therefore, at load time, all checks on pointers 1817 * previously done by the verifier are invalidated and must be 1818 * performed again, if the helper is used in combination with 1819 * direct packet access. 1820 * Return 1821 * 0 on success, or a negative error in case of failure. 1822 * 1823 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1824 * Description 1825 * Pop a VLAN header from the packet associated to *skb*. 1826 * 1827 * A call to this helper is susceptible to change the underlying 1828 * packet buffer. Therefore, at load time, all checks on pointers 1829 * previously done by the verifier are invalidated and must be 1830 * performed again, if the helper is used in combination with 1831 * direct packet access. 1832 * Return 1833 * 0 on success, or a negative error in case of failure. 1834 * 1835 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1836 * Description 1837 * Get tunnel metadata. This helper takes a pointer *key* to an 1838 * empty **struct bpf_tunnel_key** of **size**, that will be 1839 * filled with tunnel metadata for the packet associated to *skb*. 1840 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1841 * indicates that the tunnel is based on IPv6 protocol instead of 1842 * IPv4. 1843 * 1844 * The **struct bpf_tunnel_key** is an object that generalizes the 1845 * principal parameters used by various tunneling protocols into a 1846 * single struct. This way, it can be used to easily make a 1847 * decision based on the contents of the encapsulation header, 1848 * "summarized" in this struct. In particular, it holds the IP 1849 * address of the remote end (IPv4 or IPv6, depending on the case) 1850 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1851 * this struct exposes the *key*\ **->tunnel_id**, which is 1852 * generally mapped to a VNI (Virtual Network Identifier), making 1853 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1854 * () helper. 1855 * 1856 * Let's imagine that the following code is part of a program 1857 * attached to the TC ingress interface, on one end of a GRE 1858 * tunnel, and is supposed to filter out all messages coming from 1859 * remote ends with IPv4 address other than 10.0.0.1: 1860 * 1861 * :: 1862 * 1863 * int ret; 1864 * struct bpf_tunnel_key key = {}; 1865 * 1866 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1867 * if (ret < 0) 1868 * return TC_ACT_SHOT; // drop packet 1869 * 1870 * if (key.remote_ipv4 != 0x0a000001) 1871 * return TC_ACT_SHOT; // drop packet 1872 * 1873 * return TC_ACT_OK; // accept packet 1874 * 1875 * This interface can also be used with all encapsulation devices 1876 * that can operate in "collect metadata" mode: instead of having 1877 * one network device per specific configuration, the "collect 1878 * metadata" mode only requires a single device where the 1879 * configuration can be extracted from this helper. 1880 * 1881 * This can be used together with various tunnels such as VXLan, 1882 * Geneve, GRE or IP in IP (IPIP). 1883 * Return 1884 * 0 on success, or a negative error in case of failure. 1885 * 1886 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1887 * Description 1888 * Populate tunnel metadata for packet associated to *skb.* The 1889 * tunnel metadata is set to the contents of *key*, of *size*. The 1890 * *flags* can be set to a combination of the following values: 1891 * 1892 * **BPF_F_TUNINFO_IPV6** 1893 * Indicate that the tunnel is based on IPv6 protocol 1894 * instead of IPv4. 1895 * **BPF_F_ZERO_CSUM_TX** 1896 * For IPv4 packets, add a flag to tunnel metadata 1897 * indicating that checksum computation should be skipped 1898 * and checksum set to zeroes. 1899 * **BPF_F_DONT_FRAGMENT** 1900 * Add a flag to tunnel metadata indicating that the 1901 * packet should not be fragmented. 1902 * **BPF_F_SEQ_NUMBER** 1903 * Add a flag to tunnel metadata indicating that a 1904 * sequence number should be added to tunnel header before 1905 * sending the packet. This flag was added for GRE 1906 * encapsulation, but might be used with other protocols 1907 * as well in the future. 1908 * 1909 * Here is a typical usage on the transmit path: 1910 * 1911 * :: 1912 * 1913 * struct bpf_tunnel_key key; 1914 * populate key ... 1915 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1916 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1917 * 1918 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1919 * helper for additional information. 1920 * Return 1921 * 0 on success, or a negative error in case of failure. 1922 * 1923 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1924 * Description 1925 * Read the value of a perf event counter. This helper relies on a 1926 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1927 * the perf event counter is selected when *map* is updated with 1928 * perf event file descriptors. The *map* is an array whose size 1929 * is the number of available CPUs, and each cell contains a value 1930 * relative to one CPU. The value to retrieve is indicated by 1931 * *flags*, that contains the index of the CPU to look up, masked 1932 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1933 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1934 * current CPU should be retrieved. 1935 * 1936 * Note that before Linux 4.13, only hardware perf event can be 1937 * retrieved. 1938 * 1939 * Also, be aware that the newer helper 1940 * **bpf_perf_event_read_value**\ () is recommended over 1941 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1942 * quirks where error and counter value are used as a return code 1943 * (which is wrong to do since ranges may overlap). This issue is 1944 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1945 * time provides more features over the **bpf_perf_event_read**\ 1946 * () interface. Please refer to the description of 1947 * **bpf_perf_event_read_value**\ () for details. 1948 * Return 1949 * The value of the perf event counter read from the map, or a 1950 * negative error code in case of failure. 1951 * 1952 * long bpf_redirect(u32 ifindex, u64 flags) 1953 * Description 1954 * Redirect the packet to another net device of index *ifindex*. 1955 * This helper is somewhat similar to **bpf_clone_redirect**\ 1956 * (), except that the packet is not cloned, which provides 1957 * increased performance. 1958 * 1959 * Except for XDP, both ingress and egress interfaces can be used 1960 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1961 * to make the distinction (ingress path is selected if the flag 1962 * is present, egress path otherwise). Currently, XDP only 1963 * supports redirection to the egress interface, and accepts no 1964 * flag at all. 1965 * 1966 * The same effect can also be attained with the more generic 1967 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1968 * redirect target instead of providing it directly to the helper. 1969 * Return 1970 * For XDP, the helper returns **XDP_REDIRECT** on success or 1971 * **XDP_ABORTED** on error. For other program types, the values 1972 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1973 * error. 1974 * 1975 * u32 bpf_get_route_realm(struct sk_buff *skb) 1976 * Description 1977 * Retrieve the realm or the route, that is to say the 1978 * **tclassid** field of the destination for the *skb*. The 1979 * identifier retrieved is a user-provided tag, similar to the 1980 * one used with the net_cls cgroup (see description for 1981 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1982 * held by a route (a destination entry), not by a task. 1983 * 1984 * Retrieving this identifier works with the clsact TC egress hook 1985 * (see also **tc-bpf(8)**), or alternatively on conventional 1986 * classful egress qdiscs, but not on TC ingress path. In case of 1987 * clsact TC egress hook, this has the advantage that, internally, 1988 * the destination entry has not been dropped yet in the transmit 1989 * path. Therefore, the destination entry does not need to be 1990 * artificially held via **netif_keep_dst**\ () for a classful 1991 * qdisc until the *skb* is freed. 1992 * 1993 * This helper is available only if the kernel was compiled with 1994 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1995 * Return 1996 * The realm of the route for the packet associated to *skb*, or 0 1997 * if none was found. 1998 * 1999 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2000 * Description 2001 * Write raw *data* blob into a special BPF perf event held by 2002 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2003 * event must have the following attributes: **PERF_SAMPLE_RAW** 2004 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2005 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2006 * 2007 * The *flags* are used to indicate the index in *map* for which 2008 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2009 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2010 * to indicate that the index of the current CPU core should be 2011 * used. 2012 * 2013 * The value to write, of *size*, is passed through eBPF stack and 2014 * pointed by *data*. 2015 * 2016 * The context of the program *ctx* needs also be passed to the 2017 * helper. 2018 * 2019 * On user space, a program willing to read the values needs to 2020 * call **perf_event_open**\ () on the perf event (either for 2021 * one or for all CPUs) and to store the file descriptor into the 2022 * *map*. This must be done before the eBPF program can send data 2023 * into it. An example is available in file 2024 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2025 * tree (the eBPF program counterpart is in 2026 * *samples/bpf/trace_output_kern.c*). 2027 * 2028 * **bpf_perf_event_output**\ () achieves better performance 2029 * than **bpf_trace_printk**\ () for sharing data with user 2030 * space, and is much better suitable for streaming data from eBPF 2031 * programs. 2032 * 2033 * Note that this helper is not restricted to tracing use cases 2034 * and can be used with programs attached to TC or XDP as well, 2035 * where it allows for passing data to user space listeners. Data 2036 * can be: 2037 * 2038 * * Only custom structs, 2039 * * Only the packet payload, or 2040 * * A combination of both. 2041 * Return 2042 * 0 on success, or a negative error in case of failure. 2043 * 2044 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2045 * Description 2046 * This helper was provided as an easy way to load data from a 2047 * packet. It can be used to load *len* bytes from *offset* from 2048 * the packet associated to *skb*, into the buffer pointed by 2049 * *to*. 2050 * 2051 * Since Linux 4.7, usage of this helper has mostly been replaced 2052 * by "direct packet access", enabling packet data to be 2053 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2054 * pointing respectively to the first byte of packet data and to 2055 * the byte after the last byte of packet data. However, it 2056 * remains useful if one wishes to read large quantities of data 2057 * at once from a packet into the eBPF stack. 2058 * Return 2059 * 0 on success, or a negative error in case of failure. 2060 * 2061 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2062 * Description 2063 * Walk a user or a kernel stack and return its id. To achieve 2064 * this, the helper needs *ctx*, which is a pointer to the context 2065 * on which the tracing program is executed, and a pointer to a 2066 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2067 * 2068 * The last argument, *flags*, holds the number of stack frames to 2069 * skip (from 0 to 255), masked with 2070 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2071 * a combination of the following flags: 2072 * 2073 * **BPF_F_USER_STACK** 2074 * Collect a user space stack instead of a kernel stack. 2075 * **BPF_F_FAST_STACK_CMP** 2076 * Compare stacks by hash only. 2077 * **BPF_F_REUSE_STACKID** 2078 * If two different stacks hash into the same *stackid*, 2079 * discard the old one. 2080 * 2081 * The stack id retrieved is a 32 bit long integer handle which 2082 * can be further combined with other data (including other stack 2083 * ids) and used as a key into maps. This can be useful for 2084 * generating a variety of graphs (such as flame graphs or off-cpu 2085 * graphs). 2086 * 2087 * For walking a stack, this helper is an improvement over 2088 * **bpf_probe_read**\ (), which can be used with unrolled loops 2089 * but is not efficient and consumes a lot of eBPF instructions. 2090 * Instead, **bpf_get_stackid**\ () can collect up to 2091 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2092 * this limit can be controlled with the **sysctl** program, and 2093 * that it should be manually increased in order to profile long 2094 * user stacks (such as stacks for Java programs). To do so, use: 2095 * 2096 * :: 2097 * 2098 * # sysctl kernel.perf_event_max_stack=<new value> 2099 * Return 2100 * The positive or null stack id on success, or a negative error 2101 * in case of failure. 2102 * 2103 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2104 * Description 2105 * Compute a checksum difference, from the raw buffer pointed by 2106 * *from*, of length *from_size* (that must be a multiple of 4), 2107 * towards the raw buffer pointed by *to*, of size *to_size* 2108 * (same remark). An optional *seed* can be added to the value 2109 * (this can be cascaded, the seed may come from a previous call 2110 * to the helper). 2111 * 2112 * This is flexible enough to be used in several ways: 2113 * 2114 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2115 * checksum, it can be used when pushing new data. 2116 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2117 * checksum, it can be used when removing data from a packet. 2118 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2119 * can be used to compute a diff. Note that *from_size* and 2120 * *to_size* do not need to be equal. 2121 * 2122 * This helper can be used in combination with 2123 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2124 * which one can feed in the difference computed with 2125 * **bpf_csum_diff**\ (). 2126 * Return 2127 * The checksum result, or a negative error code in case of 2128 * failure. 2129 * 2130 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2131 * Description 2132 * Retrieve tunnel options metadata for the packet associated to 2133 * *skb*, and store the raw tunnel option data to the buffer *opt* 2134 * of *size*. 2135 * 2136 * This helper can be used with encapsulation devices that can 2137 * operate in "collect metadata" mode (please refer to the related 2138 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2139 * more details). A particular example where this can be used is 2140 * in combination with the Geneve encapsulation protocol, where it 2141 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2142 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2143 * the eBPF program. This allows for full customization of these 2144 * headers. 2145 * Return 2146 * The size of the option data retrieved. 2147 * 2148 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2149 * Description 2150 * Set tunnel options metadata for the packet associated to *skb* 2151 * to the option data contained in the raw buffer *opt* of *size*. 2152 * 2153 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2154 * helper for additional information. 2155 * Return 2156 * 0 on success, or a negative error in case of failure. 2157 * 2158 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2159 * Description 2160 * Change the protocol of the *skb* to *proto*. Currently 2161 * supported are transition from IPv4 to IPv6, and from IPv6 to 2162 * IPv4. The helper takes care of the groundwork for the 2163 * transition, including resizing the socket buffer. The eBPF 2164 * program is expected to fill the new headers, if any, via 2165 * **skb_store_bytes**\ () and to recompute the checksums with 2166 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2167 * (). The main case for this helper is to perform NAT64 2168 * operations out of an eBPF program. 2169 * 2170 * Internally, the GSO type is marked as dodgy so that headers are 2171 * checked and segments are recalculated by the GSO/GRO engine. 2172 * The size for GSO target is adapted as well. 2173 * 2174 * All values for *flags* are reserved for future usage, and must 2175 * be left at zero. 2176 * 2177 * A call to this helper is susceptible to change the underlying 2178 * packet buffer. Therefore, at load time, all checks on pointers 2179 * previously done by the verifier are invalidated and must be 2180 * performed again, if the helper is used in combination with 2181 * direct packet access. 2182 * Return 2183 * 0 on success, or a negative error in case of failure. 2184 * 2185 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2186 * Description 2187 * Change the packet type for the packet associated to *skb*. This 2188 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2189 * the eBPF program does not have a write access to *skb*\ 2190 * **->pkt_type** beside this helper. Using a helper here allows 2191 * for graceful handling of errors. 2192 * 2193 * The major use case is to change incoming *skb*s to 2194 * **PACKET_HOST** in a programmatic way instead of having to 2195 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2196 * example. 2197 * 2198 * Note that *type* only allows certain values. At this time, they 2199 * are: 2200 * 2201 * **PACKET_HOST** 2202 * Packet is for us. 2203 * **PACKET_BROADCAST** 2204 * Send packet to all. 2205 * **PACKET_MULTICAST** 2206 * Send packet to group. 2207 * **PACKET_OTHERHOST** 2208 * Send packet to someone else. 2209 * Return 2210 * 0 on success, or a negative error in case of failure. 2211 * 2212 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2213 * Description 2214 * Check whether *skb* is a descendant of the cgroup2 held by 2215 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2216 * Return 2217 * The return value depends on the result of the test, and can be: 2218 * 2219 * * 0, if the *skb* failed the cgroup2 descendant test. 2220 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2221 * * A negative error code, if an error occurred. 2222 * 2223 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2224 * Description 2225 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2226 * not set, in particular if the hash was cleared due to mangling, 2227 * recompute this hash. Later accesses to the hash can be done 2228 * directly with *skb*\ **->hash**. 2229 * 2230 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2231 * prototype with **bpf_skb_change_proto**\ (), or calling 2232 * **bpf_skb_store_bytes**\ () with the 2233 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2234 * the hash and to trigger a new computation for the next call to 2235 * **bpf_get_hash_recalc**\ (). 2236 * Return 2237 * The 32-bit hash. 2238 * 2239 * u64 bpf_get_current_task(void) 2240 * Return 2241 * A pointer to the current task struct. 2242 * 2243 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2244 * Description 2245 * Attempt in a safe way to write *len* bytes from the buffer 2246 * *src* to *dst* in memory. It only works for threads that are in 2247 * user context, and *dst* must be a valid user space address. 2248 * 2249 * This helper should not be used to implement any kind of 2250 * security mechanism because of TOC-TOU attacks, but rather to 2251 * debug, divert, and manipulate execution of semi-cooperative 2252 * processes. 2253 * 2254 * Keep in mind that this feature is meant for experiments, and it 2255 * has a risk of crashing the system and running programs. 2256 * Therefore, when an eBPF program using this helper is attached, 2257 * a warning including PID and process name is printed to kernel 2258 * logs. 2259 * Return 2260 * 0 on success, or a negative error in case of failure. 2261 * 2262 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2263 * Description 2264 * Check whether the probe is being run is the context of a given 2265 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2266 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2267 * Return 2268 * The return value depends on the result of the test, and can be: 2269 * 2270 * * 0, if current task belongs to the cgroup2. 2271 * * 1, if current task does not belong to the cgroup2. 2272 * * A negative error code, if an error occurred. 2273 * 2274 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2275 * Description 2276 * Resize (trim or grow) the packet associated to *skb* to the 2277 * new *len*. The *flags* are reserved for future usage, and must 2278 * be left at zero. 2279 * 2280 * The basic idea is that the helper performs the needed work to 2281 * change the size of the packet, then the eBPF program rewrites 2282 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2283 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2284 * and others. This helper is a slow path utility intended for 2285 * replies with control messages. And because it is targeted for 2286 * slow path, the helper itself can afford to be slow: it 2287 * implicitly linearizes, unclones and drops offloads from the 2288 * *skb*. 2289 * 2290 * A call to this helper is susceptible to change the underlying 2291 * packet buffer. Therefore, at load time, all checks on pointers 2292 * previously done by the verifier are invalidated and must be 2293 * performed again, if the helper is used in combination with 2294 * direct packet access. 2295 * Return 2296 * 0 on success, or a negative error in case of failure. 2297 * 2298 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2299 * Description 2300 * Pull in non-linear data in case the *skb* is non-linear and not 2301 * all of *len* are part of the linear section. Make *len* bytes 2302 * from *skb* readable and writable. If a zero value is passed for 2303 * *len*, then the whole length of the *skb* is pulled. 2304 * 2305 * This helper is only needed for reading and writing with direct 2306 * packet access. 2307 * 2308 * For direct packet access, testing that offsets to access 2309 * are within packet boundaries (test on *skb*\ **->data_end**) is 2310 * susceptible to fail if offsets are invalid, or if the requested 2311 * data is in non-linear parts of the *skb*. On failure the 2312 * program can just bail out, or in the case of a non-linear 2313 * buffer, use a helper to make the data available. The 2314 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2315 * the data. Another one consists in using **bpf_skb_pull_data** 2316 * to pull in once the non-linear parts, then retesting and 2317 * eventually access the data. 2318 * 2319 * At the same time, this also makes sure the *skb* is uncloned, 2320 * which is a necessary condition for direct write. As this needs 2321 * to be an invariant for the write part only, the verifier 2322 * detects writes and adds a prologue that is calling 2323 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2324 * the very beginning in case it is indeed cloned. 2325 * 2326 * A call to this helper is susceptible to change the underlying 2327 * packet buffer. Therefore, at load time, all checks on pointers 2328 * previously done by the verifier are invalidated and must be 2329 * performed again, if the helper is used in combination with 2330 * direct packet access. 2331 * Return 2332 * 0 on success, or a negative error in case of failure. 2333 * 2334 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2335 * Description 2336 * Add the checksum *csum* into *skb*\ **->csum** in case the 2337 * driver has supplied a checksum for the entire packet into that 2338 * field. Return an error otherwise. This helper is intended to be 2339 * used in combination with **bpf_csum_diff**\ (), in particular 2340 * when the checksum needs to be updated after data has been 2341 * written into the packet through direct packet access. 2342 * Return 2343 * The checksum on success, or a negative error code in case of 2344 * failure. 2345 * 2346 * void bpf_set_hash_invalid(struct sk_buff *skb) 2347 * Description 2348 * Invalidate the current *skb*\ **->hash**. It can be used after 2349 * mangling on headers through direct packet access, in order to 2350 * indicate that the hash is outdated and to trigger a 2351 * recalculation the next time the kernel tries to access this 2352 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2353 * 2354 * long bpf_get_numa_node_id(void) 2355 * Description 2356 * Return the id of the current NUMA node. The primary use case 2357 * for this helper is the selection of sockets for the local NUMA 2358 * node, when the program is attached to sockets using the 2359 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2360 * but the helper is also available to other eBPF program types, 2361 * similarly to **bpf_get_smp_processor_id**\ (). 2362 * Return 2363 * The id of current NUMA node. 2364 * 2365 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2366 * Description 2367 * Grows headroom of packet associated to *skb* and adjusts the 2368 * offset of the MAC header accordingly, adding *len* bytes of 2369 * space. It automatically extends and reallocates memory as 2370 * required. 2371 * 2372 * This helper can be used on a layer 3 *skb* to push a MAC header 2373 * for redirection into a layer 2 device. 2374 * 2375 * All values for *flags* are reserved for future usage, and must 2376 * be left at zero. 2377 * 2378 * A call to this helper is susceptible to change the underlying 2379 * packet buffer. Therefore, at load time, all checks on pointers 2380 * previously done by the verifier are invalidated and must be 2381 * performed again, if the helper is used in combination with 2382 * direct packet access. 2383 * Return 2384 * 0 on success, or a negative error in case of failure. 2385 * 2386 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2387 * Description 2388 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2389 * it is possible to use a negative value for *delta*. This helper 2390 * can be used to prepare the packet for pushing or popping 2391 * headers. 2392 * 2393 * A call to this helper is susceptible to change the underlying 2394 * packet buffer. Therefore, at load time, all checks on pointers 2395 * previously done by the verifier are invalidated and must be 2396 * performed again, if the helper is used in combination with 2397 * direct packet access. 2398 * Return 2399 * 0 on success, or a negative error in case of failure. 2400 * 2401 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2402 * Description 2403 * Copy a NUL terminated string from an unsafe kernel address 2404 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2405 * more details. 2406 * 2407 * Generally, use **bpf_probe_read_user_str**\ () or 2408 * **bpf_probe_read_kernel_str**\ () instead. 2409 * Return 2410 * On success, the strictly positive length of the string, 2411 * including the trailing NUL character. On error, a negative 2412 * value. 2413 * 2414 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2415 * Description 2416 * If the **struct sk_buff** pointed by *skb* has a known socket, 2417 * retrieve the cookie (generated by the kernel) of this socket. 2418 * If no cookie has been set yet, generate a new cookie. Once 2419 * generated, the socket cookie remains stable for the life of the 2420 * socket. This helper can be useful for monitoring per socket 2421 * networking traffic statistics as it provides a global socket 2422 * identifier that can be assumed unique. 2423 * Return 2424 * A 8-byte long unique number on success, or 0 if the socket 2425 * field is missing inside *skb*. 2426 * 2427 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2428 * Description 2429 * Equivalent to bpf_get_socket_cookie() helper that accepts 2430 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2431 * Return 2432 * A 8-byte long unique number. 2433 * 2434 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2435 * Description 2436 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2437 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2438 * Return 2439 * A 8-byte long unique number. 2440 * 2441 * u64 bpf_get_socket_cookie(struct sock *sk) 2442 * Description 2443 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2444 * *sk*, but gets socket from a BTF **struct sock**. This helper 2445 * also works for sleepable programs. 2446 * Return 2447 * A 8-byte long unique number or 0 if *sk* is NULL. 2448 * 2449 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2450 * Return 2451 * The owner UID of the socket associated to *skb*. If the socket 2452 * is **NULL**, or if it is not a full socket (i.e. if it is a 2453 * time-wait or a request socket instead), **overflowuid** value 2454 * is returned (note that **overflowuid** might also be the actual 2455 * UID value for the socket). 2456 * 2457 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2458 * Description 2459 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2460 * to value *hash*. 2461 * Return 2462 * 0 2463 * 2464 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2465 * Description 2466 * Emulate a call to **setsockopt()** on the socket associated to 2467 * *bpf_socket*, which must be a full socket. The *level* at 2468 * which the option resides and the name *optname* of the option 2469 * must be specified, see **setsockopt(2)** for more information. 2470 * The option value of length *optlen* is pointed by *optval*. 2471 * 2472 * *bpf_socket* should be one of the following: 2473 * 2474 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2475 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2476 * and **BPF_CGROUP_INET6_CONNECT**. 2477 * 2478 * This helper actually implements a subset of **setsockopt()**. 2479 * It supports the following *level*\ s: 2480 * 2481 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2482 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2483 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2484 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 2485 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2486 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2487 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2488 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2489 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 2490 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2491 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2492 * Return 2493 * 0 on success, or a negative error in case of failure. 2494 * 2495 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2496 * Description 2497 * Grow or shrink the room for data in the packet associated to 2498 * *skb* by *len_diff*, and according to the selected *mode*. 2499 * 2500 * By default, the helper will reset any offloaded checksum 2501 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2502 * by the following flag: 2503 * 2504 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2505 * checksum data of the skb to CHECKSUM_NONE. 2506 * 2507 * There are two supported modes at this time: 2508 * 2509 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2510 * (room space is added or removed below the layer 2 header). 2511 * 2512 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2513 * (room space is added or removed below the layer 3 header). 2514 * 2515 * The following flags are supported at this time: 2516 * 2517 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2518 * Adjusting mss in this way is not allowed for datagrams. 2519 * 2520 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2521 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2522 * Any new space is reserved to hold a tunnel header. 2523 * Configure skb offsets and other fields accordingly. 2524 * 2525 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2526 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2527 * Use with ENCAP_L3 flags to further specify the tunnel type. 2528 * 2529 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2530 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2531 * type; *len* is the length of the inner MAC header. 2532 * 2533 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2534 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2535 * L2 type as Ethernet. 2536 * 2537 * A call to this helper is susceptible to change the underlying 2538 * packet buffer. Therefore, at load time, all checks on pointers 2539 * previously done by the verifier are invalidated and must be 2540 * performed again, if the helper is used in combination with 2541 * direct packet access. 2542 * Return 2543 * 0 on success, or a negative error in case of failure. 2544 * 2545 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2546 * Description 2547 * Redirect the packet to the endpoint referenced by *map* at 2548 * index *key*. Depending on its type, this *map* can contain 2549 * references to net devices (for forwarding packets through other 2550 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2551 * but this is only implemented for native XDP (with driver 2552 * support) as of this writing). 2553 * 2554 * The lower two bits of *flags* are used as the return code if 2555 * the map lookup fails. This is so that the return value can be 2556 * one of the XDP program return codes up to **XDP_TX**, as chosen 2557 * by the caller. The higher bits of *flags* can be set to 2558 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2559 * 2560 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2561 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2562 * interface will be excluded when do broadcasting. 2563 * 2564 * See also **bpf_redirect**\ (), which only supports redirecting 2565 * to an ifindex, but doesn't require a map to do so. 2566 * Return 2567 * **XDP_REDIRECT** on success, or the value of the two lower bits 2568 * of the *flags* argument on error. 2569 * 2570 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2571 * Description 2572 * Redirect the packet to the socket referenced by *map* (of type 2573 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2574 * egress interfaces can be used for redirection. The 2575 * **BPF_F_INGRESS** value in *flags* is used to make the 2576 * distinction (ingress path is selected if the flag is present, 2577 * egress path otherwise). This is the only flag supported for now. 2578 * Return 2579 * **SK_PASS** on success, or **SK_DROP** on error. 2580 * 2581 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2582 * Description 2583 * Add an entry to, or update a *map* referencing sockets. The 2584 * *skops* is used as a new value for the entry associated to 2585 * *key*. *flags* is one of: 2586 * 2587 * **BPF_NOEXIST** 2588 * The entry for *key* must not exist in the map. 2589 * **BPF_EXIST** 2590 * The entry for *key* must already exist in the map. 2591 * **BPF_ANY** 2592 * No condition on the existence of the entry for *key*. 2593 * 2594 * If the *map* has eBPF programs (parser and verdict), those will 2595 * be inherited by the socket being added. If the socket is 2596 * already attached to eBPF programs, this results in an error. 2597 * Return 2598 * 0 on success, or a negative error in case of failure. 2599 * 2600 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2601 * Description 2602 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2603 * *delta* (which can be positive or negative). Note that this 2604 * operation modifies the address stored in *xdp_md*\ **->data**, 2605 * so the latter must be loaded only after the helper has been 2606 * called. 2607 * 2608 * The use of *xdp_md*\ **->data_meta** is optional and programs 2609 * are not required to use it. The rationale is that when the 2610 * packet is processed with XDP (e.g. as DoS filter), it is 2611 * possible to push further meta data along with it before passing 2612 * to the stack, and to give the guarantee that an ingress eBPF 2613 * program attached as a TC classifier on the same device can pick 2614 * this up for further post-processing. Since TC works with socket 2615 * buffers, it remains possible to set from XDP the **mark** or 2616 * **priority** pointers, or other pointers for the socket buffer. 2617 * Having this scratch space generic and programmable allows for 2618 * more flexibility as the user is free to store whatever meta 2619 * data they need. 2620 * 2621 * A call to this helper is susceptible to change the underlying 2622 * packet buffer. Therefore, at load time, all checks on pointers 2623 * previously done by the verifier are invalidated and must be 2624 * performed again, if the helper is used in combination with 2625 * direct packet access. 2626 * Return 2627 * 0 on success, or a negative error in case of failure. 2628 * 2629 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2630 * Description 2631 * Read the value of a perf event counter, and store it into *buf* 2632 * of size *buf_size*. This helper relies on a *map* of type 2633 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2634 * counter is selected when *map* is updated with perf event file 2635 * descriptors. The *map* is an array whose size is the number of 2636 * available CPUs, and each cell contains a value relative to one 2637 * CPU. The value to retrieve is indicated by *flags*, that 2638 * contains the index of the CPU to look up, masked with 2639 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2640 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2641 * current CPU should be retrieved. 2642 * 2643 * This helper behaves in a way close to 2644 * **bpf_perf_event_read**\ () helper, save that instead of 2645 * just returning the value observed, it fills the *buf* 2646 * structure. This allows for additional data to be retrieved: in 2647 * particular, the enabled and running times (in *buf*\ 2648 * **->enabled** and *buf*\ **->running**, respectively) are 2649 * copied. In general, **bpf_perf_event_read_value**\ () is 2650 * recommended over **bpf_perf_event_read**\ (), which has some 2651 * ABI issues and provides fewer functionalities. 2652 * 2653 * These values are interesting, because hardware PMU (Performance 2654 * Monitoring Unit) counters are limited resources. When there are 2655 * more PMU based perf events opened than available counters, 2656 * kernel will multiplex these events so each event gets certain 2657 * percentage (but not all) of the PMU time. In case that 2658 * multiplexing happens, the number of samples or counter value 2659 * will not reflect the case compared to when no multiplexing 2660 * occurs. This makes comparison between different runs difficult. 2661 * Typically, the counter value should be normalized before 2662 * comparing to other experiments. The usual normalization is done 2663 * as follows. 2664 * 2665 * :: 2666 * 2667 * normalized_counter = counter * t_enabled / t_running 2668 * 2669 * Where t_enabled is the time enabled for event and t_running is 2670 * the time running for event since last normalization. The 2671 * enabled and running times are accumulated since the perf event 2672 * open. To achieve scaling factor between two invocations of an 2673 * eBPF program, users can use CPU id as the key (which is 2674 * typical for perf array usage model) to remember the previous 2675 * value and do the calculation inside the eBPF program. 2676 * Return 2677 * 0 on success, or a negative error in case of failure. 2678 * 2679 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2680 * Description 2681 * For en eBPF program attached to a perf event, retrieve the 2682 * value of the event counter associated to *ctx* and store it in 2683 * the structure pointed by *buf* and of size *buf_size*. Enabled 2684 * and running times are also stored in the structure (see 2685 * description of helper **bpf_perf_event_read_value**\ () for 2686 * more details). 2687 * Return 2688 * 0 on success, or a negative error in case of failure. 2689 * 2690 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2691 * Description 2692 * Emulate a call to **getsockopt()** on the socket associated to 2693 * *bpf_socket*, which must be a full socket. The *level* at 2694 * which the option resides and the name *optname* of the option 2695 * must be specified, see **getsockopt(2)** for more information. 2696 * The retrieved value is stored in the structure pointed by 2697 * *opval* and of length *optlen*. 2698 * 2699 * *bpf_socket* should be one of the following: 2700 * 2701 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2702 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2703 * and **BPF_CGROUP_INET6_CONNECT**. 2704 * 2705 * This helper actually implements a subset of **getsockopt()**. 2706 * It supports the following *level*\ s: 2707 * 2708 * * **IPPROTO_TCP**, which supports *optname* 2709 * **TCP_CONGESTION**. 2710 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2711 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2712 * Return 2713 * 0 on success, or a negative error in case of failure. 2714 * 2715 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2716 * Description 2717 * Used for error injection, this helper uses kprobes to override 2718 * the return value of the probed function, and to set it to *rc*. 2719 * The first argument is the context *regs* on which the kprobe 2720 * works. 2721 * 2722 * This helper works by setting the PC (program counter) 2723 * to an override function which is run in place of the original 2724 * probed function. This means the probed function is not run at 2725 * all. The replacement function just returns with the required 2726 * value. 2727 * 2728 * This helper has security implications, and thus is subject to 2729 * restrictions. It is only available if the kernel was compiled 2730 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2731 * option, and in this case it only works on functions tagged with 2732 * **ALLOW_ERROR_INJECTION** in the kernel code. 2733 * 2734 * Also, the helper is only available for the architectures having 2735 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2736 * x86 architecture is the only one to support this feature. 2737 * Return 2738 * 0 2739 * 2740 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2741 * Description 2742 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2743 * for the full TCP socket associated to *bpf_sock_ops* to 2744 * *argval*. 2745 * 2746 * The primary use of this field is to determine if there should 2747 * be calls to eBPF programs of type 2748 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2749 * code. A program of the same type can change its value, per 2750 * connection and as necessary, when the connection is 2751 * established. This field is directly accessible for reading, but 2752 * this helper must be used for updates in order to return an 2753 * error if an eBPF program tries to set a callback that is not 2754 * supported in the current kernel. 2755 * 2756 * *argval* is a flag array which can combine these flags: 2757 * 2758 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2759 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2760 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2761 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2762 * 2763 * Therefore, this function can be used to clear a callback flag by 2764 * setting the appropriate bit to zero. e.g. to disable the RTO 2765 * callback: 2766 * 2767 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2768 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2769 * 2770 * Here are some examples of where one could call such eBPF 2771 * program: 2772 * 2773 * * When RTO fires. 2774 * * When a packet is retransmitted. 2775 * * When the connection terminates. 2776 * * When a packet is sent. 2777 * * When a packet is received. 2778 * Return 2779 * Code **-EINVAL** if the socket is not a full TCP socket; 2780 * otherwise, a positive number containing the bits that could not 2781 * be set is returned (which comes down to 0 if all bits were set 2782 * as required). 2783 * 2784 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2785 * Description 2786 * This helper is used in programs implementing policies at the 2787 * socket level. If the message *msg* is allowed to pass (i.e. if 2788 * the verdict eBPF program returns **SK_PASS**), redirect it to 2789 * the socket referenced by *map* (of type 2790 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2791 * egress interfaces can be used for redirection. The 2792 * **BPF_F_INGRESS** value in *flags* is used to make the 2793 * distinction (ingress path is selected if the flag is present, 2794 * egress path otherwise). This is the only flag supported for now. 2795 * Return 2796 * **SK_PASS** on success, or **SK_DROP** on error. 2797 * 2798 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2799 * Description 2800 * For socket policies, apply the verdict of the eBPF program to 2801 * the next *bytes* (number of bytes) of message *msg*. 2802 * 2803 * For example, this helper can be used in the following cases: 2804 * 2805 * * A single **sendmsg**\ () or **sendfile**\ () system call 2806 * contains multiple logical messages that the eBPF program is 2807 * supposed to read and for which it should apply a verdict. 2808 * * An eBPF program only cares to read the first *bytes* of a 2809 * *msg*. If the message has a large payload, then setting up 2810 * and calling the eBPF program repeatedly for all bytes, even 2811 * though the verdict is already known, would create unnecessary 2812 * overhead. 2813 * 2814 * When called from within an eBPF program, the helper sets a 2815 * counter internal to the BPF infrastructure, that is used to 2816 * apply the last verdict to the next *bytes*. If *bytes* is 2817 * smaller than the current data being processed from a 2818 * **sendmsg**\ () or **sendfile**\ () system call, the first 2819 * *bytes* will be sent and the eBPF program will be re-run with 2820 * the pointer for start of data pointing to byte number *bytes* 2821 * **+ 1**. If *bytes* is larger than the current data being 2822 * processed, then the eBPF verdict will be applied to multiple 2823 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2824 * consumed. 2825 * 2826 * Note that if a socket closes with the internal counter holding 2827 * a non-zero value, this is not a problem because data is not 2828 * being buffered for *bytes* and is sent as it is received. 2829 * Return 2830 * 0 2831 * 2832 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2833 * Description 2834 * For socket policies, prevent the execution of the verdict eBPF 2835 * program for message *msg* until *bytes* (byte number) have been 2836 * accumulated. 2837 * 2838 * This can be used when one needs a specific number of bytes 2839 * before a verdict can be assigned, even if the data spans 2840 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2841 * case would be a user calling **sendmsg**\ () repeatedly with 2842 * 1-byte long message segments. Obviously, this is bad for 2843 * performance, but it is still valid. If the eBPF program needs 2844 * *bytes* bytes to validate a header, this helper can be used to 2845 * prevent the eBPF program to be called again until *bytes* have 2846 * been accumulated. 2847 * Return 2848 * 0 2849 * 2850 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2851 * Description 2852 * For socket policies, pull in non-linear data from user space 2853 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2854 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2855 * respectively. 2856 * 2857 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2858 * *msg* it can only parse data that the (**data**, **data_end**) 2859 * pointers have already consumed. For **sendmsg**\ () hooks this 2860 * is likely the first scatterlist element. But for calls relying 2861 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2862 * be the range (**0**, **0**) because the data is shared with 2863 * user space and by default the objective is to avoid allowing 2864 * user space to modify data while (or after) eBPF verdict is 2865 * being decided. This helper can be used to pull in data and to 2866 * set the start and end pointer to given values. Data will be 2867 * copied if necessary (i.e. if data was not linear and if start 2868 * and end pointers do not point to the same chunk). 2869 * 2870 * A call to this helper is susceptible to change the underlying 2871 * packet buffer. Therefore, at load time, all checks on pointers 2872 * previously done by the verifier are invalidated and must be 2873 * performed again, if the helper is used in combination with 2874 * direct packet access. 2875 * 2876 * All values for *flags* are reserved for future usage, and must 2877 * be left at zero. 2878 * Return 2879 * 0 on success, or a negative error in case of failure. 2880 * 2881 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2882 * Description 2883 * Bind the socket associated to *ctx* to the address pointed by 2884 * *addr*, of length *addr_len*. This allows for making outgoing 2885 * connection from the desired IP address, which can be useful for 2886 * example when all processes inside a cgroup should use one 2887 * single IP address on a host that has multiple IP configured. 2888 * 2889 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2890 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2891 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2892 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2893 * behavior and lets the kernel efficiently pick up an unused 2894 * port as long as 4-tuple is unique. Passing non-zero port might 2895 * lead to degraded performance. 2896 * Return 2897 * 0 on success, or a negative error in case of failure. 2898 * 2899 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2900 * Description 2901 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2902 * possible to both shrink and grow the packet tail. 2903 * Shrink done via *delta* being a negative integer. 2904 * 2905 * A call to this helper is susceptible to change the underlying 2906 * packet buffer. Therefore, at load time, all checks on pointers 2907 * previously done by the verifier are invalidated and must be 2908 * performed again, if the helper is used in combination with 2909 * direct packet access. 2910 * Return 2911 * 0 on success, or a negative error in case of failure. 2912 * 2913 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2914 * Description 2915 * Retrieve the XFRM state (IP transform framework, see also 2916 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2917 * 2918 * The retrieved value is stored in the **struct bpf_xfrm_state** 2919 * pointed by *xfrm_state* and of length *size*. 2920 * 2921 * All values for *flags* are reserved for future usage, and must 2922 * be left at zero. 2923 * 2924 * This helper is available only if the kernel was compiled with 2925 * **CONFIG_XFRM** configuration option. 2926 * Return 2927 * 0 on success, or a negative error in case of failure. 2928 * 2929 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2930 * Description 2931 * Return a user or a kernel stack in bpf program provided buffer. 2932 * To achieve this, the helper needs *ctx*, which is a pointer 2933 * to the context on which the tracing program is executed. 2934 * To store the stacktrace, the bpf program provides *buf* with 2935 * a nonnegative *size*. 2936 * 2937 * The last argument, *flags*, holds the number of stack frames to 2938 * skip (from 0 to 255), masked with 2939 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2940 * the following flags: 2941 * 2942 * **BPF_F_USER_STACK** 2943 * Collect a user space stack instead of a kernel stack. 2944 * **BPF_F_USER_BUILD_ID** 2945 * Collect buildid+offset instead of ips for user stack, 2946 * only valid if **BPF_F_USER_STACK** is also specified. 2947 * 2948 * **bpf_get_stack**\ () can collect up to 2949 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2950 * to sufficient large buffer size. Note that 2951 * this limit can be controlled with the **sysctl** program, and 2952 * that it should be manually increased in order to profile long 2953 * user stacks (such as stacks for Java programs). To do so, use: 2954 * 2955 * :: 2956 * 2957 * # sysctl kernel.perf_event_max_stack=<new value> 2958 * Return 2959 * A non-negative value equal to or less than *size* on success, 2960 * or a negative error in case of failure. 2961 * 2962 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2963 * Description 2964 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2965 * it provides an easy way to load *len* bytes from *offset* 2966 * from the packet associated to *skb*, into the buffer pointed 2967 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2968 * a fifth argument *start_header* exists in order to select a 2969 * base offset to start from. *start_header* can be one of: 2970 * 2971 * **BPF_HDR_START_MAC** 2972 * Base offset to load data from is *skb*'s mac header. 2973 * **BPF_HDR_START_NET** 2974 * Base offset to load data from is *skb*'s network header. 2975 * 2976 * In general, "direct packet access" is the preferred method to 2977 * access packet data, however, this helper is in particular useful 2978 * in socket filters where *skb*\ **->data** does not always point 2979 * to the start of the mac header and where "direct packet access" 2980 * is not available. 2981 * Return 2982 * 0 on success, or a negative error in case of failure. 2983 * 2984 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2985 * Description 2986 * Do FIB lookup in kernel tables using parameters in *params*. 2987 * If lookup is successful and result shows packet is to be 2988 * forwarded, the neighbor tables are searched for the nexthop. 2989 * If successful (ie., FIB lookup shows forwarding and nexthop 2990 * is resolved), the nexthop address is returned in ipv4_dst 2991 * or ipv6_dst based on family, smac is set to mac address of 2992 * egress device, dmac is set to nexthop mac address, rt_metric 2993 * is set to metric from route (IPv4/IPv6 only), and ifindex 2994 * is set to the device index of the nexthop from the FIB lookup. 2995 * 2996 * *plen* argument is the size of the passed in struct. 2997 * *flags* argument can be a combination of one or more of the 2998 * following values: 2999 * 3000 * **BPF_FIB_LOOKUP_DIRECT** 3001 * Do a direct table lookup vs full lookup using FIB 3002 * rules. 3003 * **BPF_FIB_LOOKUP_OUTPUT** 3004 * Perform lookup from an egress perspective (default is 3005 * ingress). 3006 * 3007 * *ctx* is either **struct xdp_md** for XDP programs or 3008 * **struct sk_buff** tc cls_act programs. 3009 * Return 3010 * * < 0 if any input argument is invalid 3011 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3012 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3013 * packet is not forwarded or needs assist from full stack 3014 * 3015 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3016 * was exceeded and output params->mtu_result contains the MTU. 3017 * 3018 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3019 * Description 3020 * Add an entry to, or update a sockhash *map* referencing sockets. 3021 * The *skops* is used as a new value for the entry associated to 3022 * *key*. *flags* is one of: 3023 * 3024 * **BPF_NOEXIST** 3025 * The entry for *key* must not exist in the map. 3026 * **BPF_EXIST** 3027 * The entry for *key* must already exist in the map. 3028 * **BPF_ANY** 3029 * No condition on the existence of the entry for *key*. 3030 * 3031 * If the *map* has eBPF programs (parser and verdict), those will 3032 * be inherited by the socket being added. If the socket is 3033 * already attached to eBPF programs, this results in an error. 3034 * Return 3035 * 0 on success, or a negative error in case of failure. 3036 * 3037 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3038 * Description 3039 * This helper is used in programs implementing policies at the 3040 * socket level. If the message *msg* is allowed to pass (i.e. if 3041 * the verdict eBPF program returns **SK_PASS**), redirect it to 3042 * the socket referenced by *map* (of type 3043 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3044 * egress interfaces can be used for redirection. The 3045 * **BPF_F_INGRESS** value in *flags* is used to make the 3046 * distinction (ingress path is selected if the flag is present, 3047 * egress path otherwise). This is the only flag supported for now. 3048 * Return 3049 * **SK_PASS** on success, or **SK_DROP** on error. 3050 * 3051 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3052 * Description 3053 * This helper is used in programs implementing policies at the 3054 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3055 * if the verdict eBPF program returns **SK_PASS**), redirect it 3056 * to the socket referenced by *map* (of type 3057 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3058 * egress interfaces can be used for redirection. The 3059 * **BPF_F_INGRESS** value in *flags* is used to make the 3060 * distinction (ingress path is selected if the flag is present, 3061 * egress otherwise). This is the only flag supported for now. 3062 * Return 3063 * **SK_PASS** on success, or **SK_DROP** on error. 3064 * 3065 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3066 * Description 3067 * Encapsulate the packet associated to *skb* within a Layer 3 3068 * protocol header. This header is provided in the buffer at 3069 * address *hdr*, with *len* its size in bytes. *type* indicates 3070 * the protocol of the header and can be one of: 3071 * 3072 * **BPF_LWT_ENCAP_SEG6** 3073 * IPv6 encapsulation with Segment Routing Header 3074 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3075 * the IPv6 header is computed by the kernel. 3076 * **BPF_LWT_ENCAP_SEG6_INLINE** 3077 * Only works if *skb* contains an IPv6 packet. Insert a 3078 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3079 * the IPv6 header. 3080 * **BPF_LWT_ENCAP_IP** 3081 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3082 * must be IPv4 or IPv6, followed by zero or more 3083 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3084 * total bytes in all prepended headers. Please note that 3085 * if **skb_is_gso**\ (*skb*) is true, no more than two 3086 * headers can be prepended, and the inner header, if 3087 * present, should be either GRE or UDP/GUE. 3088 * 3089 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3090 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3091 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3092 * **BPF_PROG_TYPE_LWT_XMIT**. 3093 * 3094 * A call to this helper is susceptible to change the underlying 3095 * packet buffer. Therefore, at load time, all checks on pointers 3096 * previously done by the verifier are invalidated and must be 3097 * performed again, if the helper is used in combination with 3098 * direct packet access. 3099 * Return 3100 * 0 on success, or a negative error in case of failure. 3101 * 3102 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3103 * Description 3104 * Store *len* bytes from address *from* into the packet 3105 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3106 * inside the outermost IPv6 Segment Routing Header can be 3107 * modified through this helper. 3108 * 3109 * A call to this helper is susceptible to change the underlying 3110 * packet buffer. Therefore, at load time, all checks on pointers 3111 * previously done by the verifier are invalidated and must be 3112 * performed again, if the helper is used in combination with 3113 * direct packet access. 3114 * Return 3115 * 0 on success, or a negative error in case of failure. 3116 * 3117 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3118 * Description 3119 * Adjust the size allocated to TLVs in the outermost IPv6 3120 * Segment Routing Header contained in the packet associated to 3121 * *skb*, at position *offset* by *delta* bytes. Only offsets 3122 * after the segments are accepted. *delta* can be as well 3123 * positive (growing) as negative (shrinking). 3124 * 3125 * A call to this helper is susceptible to change the underlying 3126 * packet buffer. Therefore, at load time, all checks on pointers 3127 * previously done by the verifier are invalidated and must be 3128 * performed again, if the helper is used in combination with 3129 * direct packet access. 3130 * Return 3131 * 0 on success, or a negative error in case of failure. 3132 * 3133 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3134 * Description 3135 * Apply an IPv6 Segment Routing action of type *action* to the 3136 * packet associated to *skb*. Each action takes a parameter 3137 * contained at address *param*, and of length *param_len* bytes. 3138 * *action* can be one of: 3139 * 3140 * **SEG6_LOCAL_ACTION_END_X** 3141 * End.X action: Endpoint with Layer-3 cross-connect. 3142 * Type of *param*: **struct in6_addr**. 3143 * **SEG6_LOCAL_ACTION_END_T** 3144 * End.T action: Endpoint with specific IPv6 table lookup. 3145 * Type of *param*: **int**. 3146 * **SEG6_LOCAL_ACTION_END_B6** 3147 * End.B6 action: Endpoint bound to an SRv6 policy. 3148 * Type of *param*: **struct ipv6_sr_hdr**. 3149 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3150 * End.B6.Encap action: Endpoint bound to an SRv6 3151 * encapsulation policy. 3152 * Type of *param*: **struct ipv6_sr_hdr**. 3153 * 3154 * A call to this helper is susceptible to change the underlying 3155 * packet buffer. Therefore, at load time, all checks on pointers 3156 * previously done by the verifier are invalidated and must be 3157 * performed again, if the helper is used in combination with 3158 * direct packet access. 3159 * Return 3160 * 0 on success, or a negative error in case of failure. 3161 * 3162 * long bpf_rc_repeat(void *ctx) 3163 * Description 3164 * This helper is used in programs implementing IR decoding, to 3165 * report a successfully decoded repeat key message. This delays 3166 * the generation of a key up event for previously generated 3167 * key down event. 3168 * 3169 * Some IR protocols like NEC have a special IR message for 3170 * repeating last button, for when a button is held down. 3171 * 3172 * The *ctx* should point to the lirc sample as passed into 3173 * the program. 3174 * 3175 * This helper is only available is the kernel was compiled with 3176 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3177 * "**y**". 3178 * Return 3179 * 0 3180 * 3181 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3182 * Description 3183 * This helper is used in programs implementing IR decoding, to 3184 * report a successfully decoded key press with *scancode*, 3185 * *toggle* value in the given *protocol*. The scancode will be 3186 * translated to a keycode using the rc keymap, and reported as 3187 * an input key down event. After a period a key up event is 3188 * generated. This period can be extended by calling either 3189 * **bpf_rc_keydown**\ () again with the same values, or calling 3190 * **bpf_rc_repeat**\ (). 3191 * 3192 * Some protocols include a toggle bit, in case the button was 3193 * released and pressed again between consecutive scancodes. 3194 * 3195 * The *ctx* should point to the lirc sample as passed into 3196 * the program. 3197 * 3198 * The *protocol* is the decoded protocol number (see 3199 * **enum rc_proto** for some predefined values). 3200 * 3201 * This helper is only available is the kernel was compiled with 3202 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3203 * "**y**". 3204 * Return 3205 * 0 3206 * 3207 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3208 * Description 3209 * Return the cgroup v2 id of the socket associated with the *skb*. 3210 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3211 * helper for cgroup v1 by providing a tag resp. identifier that 3212 * can be matched on or used for map lookups e.g. to implement 3213 * policy. The cgroup v2 id of a given path in the hierarchy is 3214 * exposed in user space through the f_handle API in order to get 3215 * to the same 64-bit id. 3216 * 3217 * This helper can be used on TC egress path, but not on ingress, 3218 * and is available only if the kernel was compiled with the 3219 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3220 * Return 3221 * The id is returned or 0 in case the id could not be retrieved. 3222 * 3223 * u64 bpf_get_current_cgroup_id(void) 3224 * Return 3225 * A 64-bit integer containing the current cgroup id based 3226 * on the cgroup within which the current task is running. 3227 * 3228 * void *bpf_get_local_storage(void *map, u64 flags) 3229 * Description 3230 * Get the pointer to the local storage area. 3231 * The type and the size of the local storage is defined 3232 * by the *map* argument. 3233 * The *flags* meaning is specific for each map type, 3234 * and has to be 0 for cgroup local storage. 3235 * 3236 * Depending on the BPF program type, a local storage area 3237 * can be shared between multiple instances of the BPF program, 3238 * running simultaneously. 3239 * 3240 * A user should care about the synchronization by himself. 3241 * For example, by using the **BPF_ATOMIC** instructions to alter 3242 * the shared data. 3243 * Return 3244 * A pointer to the local storage area. 3245 * 3246 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3247 * Description 3248 * Select a **SO_REUSEPORT** socket from a 3249 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3250 * It checks the selected socket is matching the incoming 3251 * request in the socket buffer. 3252 * Return 3253 * 0 on success, or a negative error in case of failure. 3254 * 3255 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3256 * Description 3257 * Return id of cgroup v2 that is ancestor of cgroup associated 3258 * with the *skb* at the *ancestor_level*. The root cgroup is at 3259 * *ancestor_level* zero and each step down the hierarchy 3260 * increments the level. If *ancestor_level* == level of cgroup 3261 * associated with *skb*, then return value will be same as that 3262 * of **bpf_skb_cgroup_id**\ (). 3263 * 3264 * The helper is useful to implement policies based on cgroups 3265 * that are upper in hierarchy than immediate cgroup associated 3266 * with *skb*. 3267 * 3268 * The format of returned id and helper limitations are same as in 3269 * **bpf_skb_cgroup_id**\ (). 3270 * Return 3271 * The id is returned or 0 in case the id could not be retrieved. 3272 * 3273 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3274 * Description 3275 * Look for TCP socket matching *tuple*, optionally in a child 3276 * network namespace *netns*. The return value must be checked, 3277 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3278 * 3279 * The *ctx* should point to the context of the program, such as 3280 * the skb or socket (depending on the hook in use). This is used 3281 * to determine the base network namespace for the lookup. 3282 * 3283 * *tuple_size* must be one of: 3284 * 3285 * **sizeof**\ (*tuple*\ **->ipv4**) 3286 * Look for an IPv4 socket. 3287 * **sizeof**\ (*tuple*\ **->ipv6**) 3288 * Look for an IPv6 socket. 3289 * 3290 * If the *netns* is a negative signed 32-bit integer, then the 3291 * socket lookup table in the netns associated with the *ctx* 3292 * will be used. For the TC hooks, this is the netns of the device 3293 * in the skb. For socket hooks, this is the netns of the socket. 3294 * If *netns* is any other signed 32-bit value greater than or 3295 * equal to zero then it specifies the ID of the netns relative to 3296 * the netns associated with the *ctx*. *netns* values beyond the 3297 * range of 32-bit integers are reserved for future use. 3298 * 3299 * All values for *flags* are reserved for future usage, and must 3300 * be left at zero. 3301 * 3302 * This helper is available only if the kernel was compiled with 3303 * **CONFIG_NET** configuration option. 3304 * Return 3305 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3306 * For sockets with reuseport option, the **struct bpf_sock** 3307 * result is from *reuse*\ **->socks**\ [] using the hash of the 3308 * tuple. 3309 * 3310 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3311 * Description 3312 * Look for UDP socket matching *tuple*, optionally in a child 3313 * network namespace *netns*. The return value must be checked, 3314 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3315 * 3316 * The *ctx* should point to the context of the program, such as 3317 * the skb or socket (depending on the hook in use). This is used 3318 * to determine the base network namespace for the lookup. 3319 * 3320 * *tuple_size* must be one of: 3321 * 3322 * **sizeof**\ (*tuple*\ **->ipv4**) 3323 * Look for an IPv4 socket. 3324 * **sizeof**\ (*tuple*\ **->ipv6**) 3325 * Look for an IPv6 socket. 3326 * 3327 * If the *netns* is a negative signed 32-bit integer, then the 3328 * socket lookup table in the netns associated with the *ctx* 3329 * will be used. For the TC hooks, this is the netns of the device 3330 * in the skb. For socket hooks, this is the netns of the socket. 3331 * If *netns* is any other signed 32-bit value greater than or 3332 * equal to zero then it specifies the ID of the netns relative to 3333 * the netns associated with the *ctx*. *netns* values beyond the 3334 * range of 32-bit integers are reserved for future use. 3335 * 3336 * All values for *flags* are reserved for future usage, and must 3337 * be left at zero. 3338 * 3339 * This helper is available only if the kernel was compiled with 3340 * **CONFIG_NET** configuration option. 3341 * Return 3342 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3343 * For sockets with reuseport option, the **struct bpf_sock** 3344 * result is from *reuse*\ **->socks**\ [] using the hash of the 3345 * tuple. 3346 * 3347 * long bpf_sk_release(void *sock) 3348 * Description 3349 * Release the reference held by *sock*. *sock* must be a 3350 * non-**NULL** pointer that was returned from 3351 * **bpf_sk_lookup_xxx**\ (). 3352 * Return 3353 * 0 on success, or a negative error in case of failure. 3354 * 3355 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3356 * Description 3357 * Push an element *value* in *map*. *flags* is one of: 3358 * 3359 * **BPF_EXIST** 3360 * If the queue/stack is full, the oldest element is 3361 * removed to make room for this. 3362 * Return 3363 * 0 on success, or a negative error in case of failure. 3364 * 3365 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3366 * Description 3367 * Pop an element from *map*. 3368 * Return 3369 * 0 on success, or a negative error in case of failure. 3370 * 3371 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3372 * Description 3373 * Get an element from *map* without removing it. 3374 * Return 3375 * 0 on success, or a negative error in case of failure. 3376 * 3377 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3378 * Description 3379 * For socket policies, insert *len* bytes into *msg* at offset 3380 * *start*. 3381 * 3382 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3383 * *msg* it may want to insert metadata or options into the *msg*. 3384 * This can later be read and used by any of the lower layer BPF 3385 * hooks. 3386 * 3387 * This helper may fail if under memory pressure (a malloc 3388 * fails) in these cases BPF programs will get an appropriate 3389 * error and BPF programs will need to handle them. 3390 * Return 3391 * 0 on success, or a negative error in case of failure. 3392 * 3393 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3394 * Description 3395 * Will remove *len* bytes from a *msg* starting at byte *start*. 3396 * This may result in **ENOMEM** errors under certain situations if 3397 * an allocation and copy are required due to a full ring buffer. 3398 * However, the helper will try to avoid doing the allocation 3399 * if possible. Other errors can occur if input parameters are 3400 * invalid either due to *start* byte not being valid part of *msg* 3401 * payload and/or *pop* value being to large. 3402 * Return 3403 * 0 on success, or a negative error in case of failure. 3404 * 3405 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3406 * Description 3407 * This helper is used in programs implementing IR decoding, to 3408 * report a successfully decoded pointer movement. 3409 * 3410 * The *ctx* should point to the lirc sample as passed into 3411 * the program. 3412 * 3413 * This helper is only available is the kernel was compiled with 3414 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3415 * "**y**". 3416 * Return 3417 * 0 3418 * 3419 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3420 * Description 3421 * Acquire a spinlock represented by the pointer *lock*, which is 3422 * stored as part of a value of a map. Taking the lock allows to 3423 * safely update the rest of the fields in that value. The 3424 * spinlock can (and must) later be released with a call to 3425 * **bpf_spin_unlock**\ (\ *lock*\ ). 3426 * 3427 * Spinlocks in BPF programs come with a number of restrictions 3428 * and constraints: 3429 * 3430 * * **bpf_spin_lock** objects are only allowed inside maps of 3431 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3432 * list could be extended in the future). 3433 * * BTF description of the map is mandatory. 3434 * * The BPF program can take ONE lock at a time, since taking two 3435 * or more could cause dead locks. 3436 * * Only one **struct bpf_spin_lock** is allowed per map element. 3437 * * When the lock is taken, calls (either BPF to BPF or helpers) 3438 * are not allowed. 3439 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3440 * allowed inside a spinlock-ed region. 3441 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3442 * the lock, on all execution paths, before it returns. 3443 * * The BPF program can access **struct bpf_spin_lock** only via 3444 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3445 * helpers. Loading or storing data into the **struct 3446 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3447 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3448 * of the map value must be a struct and have **struct 3449 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3450 * Nested lock inside another struct is not allowed. 3451 * * The **struct bpf_spin_lock** *lock* field in a map value must 3452 * be aligned on a multiple of 4 bytes in that value. 3453 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3454 * the **bpf_spin_lock** field to user space. 3455 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3456 * a BPF program, do not update the **bpf_spin_lock** field. 3457 * * **bpf_spin_lock** cannot be on the stack or inside a 3458 * networking packet (it can only be inside of a map values). 3459 * * **bpf_spin_lock** is available to root only. 3460 * * Tracing programs and socket filter programs cannot use 3461 * **bpf_spin_lock**\ () due to insufficient preemption checks 3462 * (but this may change in the future). 3463 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3464 * Return 3465 * 0 3466 * 3467 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3468 * Description 3469 * Release the *lock* previously locked by a call to 3470 * **bpf_spin_lock**\ (\ *lock*\ ). 3471 * Return 3472 * 0 3473 * 3474 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3475 * Description 3476 * This helper gets a **struct bpf_sock** pointer such 3477 * that all the fields in this **bpf_sock** can be accessed. 3478 * Return 3479 * A **struct bpf_sock** pointer on success, or **NULL** in 3480 * case of failure. 3481 * 3482 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3483 * Description 3484 * This helper gets a **struct bpf_tcp_sock** pointer from a 3485 * **struct bpf_sock** pointer. 3486 * Return 3487 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3488 * case of failure. 3489 * 3490 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3491 * Description 3492 * Set ECN (Explicit Congestion Notification) field of IP header 3493 * to **CE** (Congestion Encountered) if current value is **ECT** 3494 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3495 * and IPv4. 3496 * Return 3497 * 1 if the **CE** flag is set (either by the current helper call 3498 * or because it was already present), 0 if it is not set. 3499 * 3500 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3501 * Description 3502 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3503 * **bpf_sk_release**\ () is unnecessary and not allowed. 3504 * Return 3505 * A **struct bpf_sock** pointer on success, or **NULL** in 3506 * case of failure. 3507 * 3508 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3509 * Description 3510 * Look for TCP socket matching *tuple*, optionally in a child 3511 * network namespace *netns*. The return value must be checked, 3512 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3513 * 3514 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3515 * that it also returns timewait or request sockets. Use 3516 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3517 * full structure. 3518 * 3519 * This helper is available only if the kernel was compiled with 3520 * **CONFIG_NET** configuration option. 3521 * Return 3522 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3523 * For sockets with reuseport option, the **struct bpf_sock** 3524 * result is from *reuse*\ **->socks**\ [] using the hash of the 3525 * tuple. 3526 * 3527 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3528 * Description 3529 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3530 * the listening socket in *sk*. 3531 * 3532 * *iph* points to the start of the IPv4 or IPv6 header, while 3533 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3534 * **sizeof**\ (**struct ip6hdr**). 3535 * 3536 * *th* points to the start of the TCP header, while *th_len* 3537 * contains **sizeof**\ (**struct tcphdr**). 3538 * Return 3539 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3540 * error otherwise. 3541 * 3542 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3543 * Description 3544 * Get name of sysctl in /proc/sys/ and copy it into provided by 3545 * program buffer *buf* of size *buf_len*. 3546 * 3547 * The buffer is always NUL terminated, unless it's zero-sized. 3548 * 3549 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3550 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3551 * only (e.g. "tcp_mem"). 3552 * Return 3553 * Number of character copied (not including the trailing NUL). 3554 * 3555 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3556 * truncated name in this case). 3557 * 3558 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3559 * Description 3560 * Get current value of sysctl as it is presented in /proc/sys 3561 * (incl. newline, etc), and copy it as a string into provided 3562 * by program buffer *buf* of size *buf_len*. 3563 * 3564 * The whole value is copied, no matter what file position user 3565 * space issued e.g. sys_read at. 3566 * 3567 * The buffer is always NUL terminated, unless it's zero-sized. 3568 * Return 3569 * Number of character copied (not including the trailing NUL). 3570 * 3571 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3572 * truncated name in this case). 3573 * 3574 * **-EINVAL** if current value was unavailable, e.g. because 3575 * sysctl is uninitialized and read returns -EIO for it. 3576 * 3577 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3578 * Description 3579 * Get new value being written by user space to sysctl (before 3580 * the actual write happens) and copy it as a string into 3581 * provided by program buffer *buf* of size *buf_len*. 3582 * 3583 * User space may write new value at file position > 0. 3584 * 3585 * The buffer is always NUL terminated, unless it's zero-sized. 3586 * Return 3587 * Number of character copied (not including the trailing NUL). 3588 * 3589 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3590 * truncated name in this case). 3591 * 3592 * **-EINVAL** if sysctl is being read. 3593 * 3594 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3595 * Description 3596 * Override new value being written by user space to sysctl with 3597 * value provided by program in buffer *buf* of size *buf_len*. 3598 * 3599 * *buf* should contain a string in same form as provided by user 3600 * space on sysctl write. 3601 * 3602 * User space may write new value at file position > 0. To override 3603 * the whole sysctl value file position should be set to zero. 3604 * Return 3605 * 0 on success. 3606 * 3607 * **-E2BIG** if the *buf_len* is too big. 3608 * 3609 * **-EINVAL** if sysctl is being read. 3610 * 3611 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3612 * Description 3613 * Convert the initial part of the string from buffer *buf* of 3614 * size *buf_len* to a long integer according to the given base 3615 * and save the result in *res*. 3616 * 3617 * The string may begin with an arbitrary amount of white space 3618 * (as determined by **isspace**\ (3)) followed by a single 3619 * optional '**-**' sign. 3620 * 3621 * Five least significant bits of *flags* encode base, other bits 3622 * are currently unused. 3623 * 3624 * Base must be either 8, 10, 16 or 0 to detect it automatically 3625 * similar to user space **strtol**\ (3). 3626 * Return 3627 * Number of characters consumed on success. Must be positive but 3628 * no more than *buf_len*. 3629 * 3630 * **-EINVAL** if no valid digits were found or unsupported base 3631 * was provided. 3632 * 3633 * **-ERANGE** if resulting value was out of range. 3634 * 3635 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3636 * Description 3637 * Convert the initial part of the string from buffer *buf* of 3638 * size *buf_len* to an unsigned long integer according to the 3639 * given base and save the result in *res*. 3640 * 3641 * The string may begin with an arbitrary amount of white space 3642 * (as determined by **isspace**\ (3)). 3643 * 3644 * Five least significant bits of *flags* encode base, other bits 3645 * are currently unused. 3646 * 3647 * Base must be either 8, 10, 16 or 0 to detect it automatically 3648 * similar to user space **strtoul**\ (3). 3649 * Return 3650 * Number of characters consumed on success. Must be positive but 3651 * no more than *buf_len*. 3652 * 3653 * **-EINVAL** if no valid digits were found or unsupported base 3654 * was provided. 3655 * 3656 * **-ERANGE** if resulting value was out of range. 3657 * 3658 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3659 * Description 3660 * Get a bpf-local-storage from a *sk*. 3661 * 3662 * Logically, it could be thought of getting the value from 3663 * a *map* with *sk* as the **key**. From this 3664 * perspective, the usage is not much different from 3665 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3666 * helper enforces the key must be a full socket and the map must 3667 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3668 * 3669 * Underneath, the value is stored locally at *sk* instead of 3670 * the *map*. The *map* is used as the bpf-local-storage 3671 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3672 * searched against all bpf-local-storages residing at *sk*. 3673 * 3674 * *sk* is a kernel **struct sock** pointer for LSM program. 3675 * *sk* is a **struct bpf_sock** pointer for other program types. 3676 * 3677 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3678 * used such that a new bpf-local-storage will be 3679 * created if one does not exist. *value* can be used 3680 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3681 * the initial value of a bpf-local-storage. If *value* is 3682 * **NULL**, the new bpf-local-storage will be zero initialized. 3683 * Return 3684 * A bpf-local-storage pointer is returned on success. 3685 * 3686 * **NULL** if not found or there was an error in adding 3687 * a new bpf-local-storage. 3688 * 3689 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3690 * Description 3691 * Delete a bpf-local-storage from a *sk*. 3692 * Return 3693 * 0 on success. 3694 * 3695 * **-ENOENT** if the bpf-local-storage cannot be found. 3696 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3697 * 3698 * long bpf_send_signal(u32 sig) 3699 * Description 3700 * Send signal *sig* to the process of the current task. 3701 * The signal may be delivered to any of this process's threads. 3702 * Return 3703 * 0 on success or successfully queued. 3704 * 3705 * **-EBUSY** if work queue under nmi is full. 3706 * 3707 * **-EINVAL** if *sig* is invalid. 3708 * 3709 * **-EPERM** if no permission to send the *sig*. 3710 * 3711 * **-EAGAIN** if bpf program can try again. 3712 * 3713 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3714 * Description 3715 * Try to issue a SYN cookie for the packet with corresponding 3716 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3717 * 3718 * *iph* points to the start of the IPv4 or IPv6 header, while 3719 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3720 * **sizeof**\ (**struct ip6hdr**). 3721 * 3722 * *th* points to the start of the TCP header, while *th_len* 3723 * contains the length of the TCP header. 3724 * Return 3725 * On success, lower 32 bits hold the generated SYN cookie in 3726 * followed by 16 bits which hold the MSS value for that cookie, 3727 * and the top 16 bits are unused. 3728 * 3729 * On failure, the returned value is one of the following: 3730 * 3731 * **-EINVAL** SYN cookie cannot be issued due to error 3732 * 3733 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3734 * 3735 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3736 * 3737 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3738 * 3739 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3740 * Description 3741 * Write raw *data* blob into a special BPF perf event held by 3742 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3743 * event must have the following attributes: **PERF_SAMPLE_RAW** 3744 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3745 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3746 * 3747 * The *flags* are used to indicate the index in *map* for which 3748 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3749 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3750 * to indicate that the index of the current CPU core should be 3751 * used. 3752 * 3753 * The value to write, of *size*, is passed through eBPF stack and 3754 * pointed by *data*. 3755 * 3756 * *ctx* is a pointer to in-kernel struct sk_buff. 3757 * 3758 * This helper is similar to **bpf_perf_event_output**\ () but 3759 * restricted to raw_tracepoint bpf programs. 3760 * Return 3761 * 0 on success, or a negative error in case of failure. 3762 * 3763 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3764 * Description 3765 * Safely attempt to read *size* bytes from user space address 3766 * *unsafe_ptr* and store the data in *dst*. 3767 * Return 3768 * 0 on success, or a negative error in case of failure. 3769 * 3770 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3771 * Description 3772 * Safely attempt to read *size* bytes from kernel space address 3773 * *unsafe_ptr* and store the data in *dst*. 3774 * Return 3775 * 0 on success, or a negative error in case of failure. 3776 * 3777 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3778 * Description 3779 * Copy a NUL terminated string from an unsafe user address 3780 * *unsafe_ptr* to *dst*. The *size* should include the 3781 * terminating NUL byte. In case the string length is smaller than 3782 * *size*, the target is not padded with further NUL bytes. If the 3783 * string length is larger than *size*, just *size*-1 bytes are 3784 * copied and the last byte is set to NUL. 3785 * 3786 * On success, returns the number of bytes that were written, 3787 * including the terminal NUL. This makes this helper useful in 3788 * tracing programs for reading strings, and more importantly to 3789 * get its length at runtime. See the following snippet: 3790 * 3791 * :: 3792 * 3793 * SEC("kprobe/sys_open") 3794 * void bpf_sys_open(struct pt_regs *ctx) 3795 * { 3796 * char buf[PATHLEN]; // PATHLEN is defined to 256 3797 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3798 * ctx->di); 3799 * 3800 * // Consume buf, for example push it to 3801 * // userspace via bpf_perf_event_output(); we 3802 * // can use res (the string length) as event 3803 * // size, after checking its boundaries. 3804 * } 3805 * 3806 * In comparison, using **bpf_probe_read_user**\ () helper here 3807 * instead to read the string would require to estimate the length 3808 * at compile time, and would often result in copying more memory 3809 * than necessary. 3810 * 3811 * Another useful use case is when parsing individual process 3812 * arguments or individual environment variables navigating 3813 * *current*\ **->mm->arg_start** and *current*\ 3814 * **->mm->env_start**: using this helper and the return value, 3815 * one can quickly iterate at the right offset of the memory area. 3816 * Return 3817 * On success, the strictly positive length of the output string, 3818 * including the trailing NUL character. On error, a negative 3819 * value. 3820 * 3821 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3822 * Description 3823 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3824 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3825 * Return 3826 * On success, the strictly positive length of the string, including 3827 * the trailing NUL character. On error, a negative value. 3828 * 3829 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3830 * Description 3831 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3832 * *rcv_nxt* is the ack_seq to be sent out. 3833 * Return 3834 * 0 on success, or a negative error in case of failure. 3835 * 3836 * long bpf_send_signal_thread(u32 sig) 3837 * Description 3838 * Send signal *sig* to the thread corresponding to the current task. 3839 * Return 3840 * 0 on success or successfully queued. 3841 * 3842 * **-EBUSY** if work queue under nmi is full. 3843 * 3844 * **-EINVAL** if *sig* is invalid. 3845 * 3846 * **-EPERM** if no permission to send the *sig*. 3847 * 3848 * **-EAGAIN** if bpf program can try again. 3849 * 3850 * u64 bpf_jiffies64(void) 3851 * Description 3852 * Obtain the 64bit jiffies 3853 * Return 3854 * The 64 bit jiffies 3855 * 3856 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3857 * Description 3858 * For an eBPF program attached to a perf event, retrieve the 3859 * branch records (**struct perf_branch_entry**) associated to *ctx* 3860 * and store it in the buffer pointed by *buf* up to size 3861 * *size* bytes. 3862 * Return 3863 * On success, number of bytes written to *buf*. On error, a 3864 * negative value. 3865 * 3866 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3867 * instead return the number of bytes required to store all the 3868 * branch entries. If this flag is set, *buf* may be NULL. 3869 * 3870 * **-EINVAL** if arguments invalid or **size** not a multiple 3871 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3872 * 3873 * **-ENOENT** if architecture does not support branch records. 3874 * 3875 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3876 * Description 3877 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3878 * *namespace* will be returned in *nsdata*. 3879 * Return 3880 * 0 on success, or one of the following in case of failure: 3881 * 3882 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3883 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3884 * 3885 * **-ENOENT** if pidns does not exists for the current task. 3886 * 3887 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3888 * Description 3889 * Write raw *data* blob into a special BPF perf event held by 3890 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3891 * event must have the following attributes: **PERF_SAMPLE_RAW** 3892 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3893 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3894 * 3895 * The *flags* are used to indicate the index in *map* for which 3896 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3897 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3898 * to indicate that the index of the current CPU core should be 3899 * used. 3900 * 3901 * The value to write, of *size*, is passed through eBPF stack and 3902 * pointed by *data*. 3903 * 3904 * *ctx* is a pointer to in-kernel struct xdp_buff. 3905 * 3906 * This helper is similar to **bpf_perf_eventoutput**\ () but 3907 * restricted to raw_tracepoint bpf programs. 3908 * Return 3909 * 0 on success, or a negative error in case of failure. 3910 * 3911 * u64 bpf_get_netns_cookie(void *ctx) 3912 * Description 3913 * Retrieve the cookie (generated by the kernel) of the network 3914 * namespace the input *ctx* is associated with. The network 3915 * namespace cookie remains stable for its lifetime and provides 3916 * a global identifier that can be assumed unique. If *ctx* is 3917 * NULL, then the helper returns the cookie for the initial 3918 * network namespace. The cookie itself is very similar to that 3919 * of **bpf_get_socket_cookie**\ () helper, but for network 3920 * namespaces instead of sockets. 3921 * Return 3922 * A 8-byte long opaque number. 3923 * 3924 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3925 * Description 3926 * Return id of cgroup v2 that is ancestor of the cgroup associated 3927 * with the current task at the *ancestor_level*. The root cgroup 3928 * is at *ancestor_level* zero and each step down the hierarchy 3929 * increments the level. If *ancestor_level* == level of cgroup 3930 * associated with the current task, then return value will be the 3931 * same as that of **bpf_get_current_cgroup_id**\ (). 3932 * 3933 * The helper is useful to implement policies based on cgroups 3934 * that are upper in hierarchy than immediate cgroup associated 3935 * with the current task. 3936 * 3937 * The format of returned id and helper limitations are same as in 3938 * **bpf_get_current_cgroup_id**\ (). 3939 * Return 3940 * The id is returned or 0 in case the id could not be retrieved. 3941 * 3942 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3943 * Description 3944 * Helper is overloaded depending on BPF program type. This 3945 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3946 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3947 * 3948 * Assign the *sk* to the *skb*. When combined with appropriate 3949 * routing configuration to receive the packet towards the socket, 3950 * will cause *skb* to be delivered to the specified socket. 3951 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3952 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3953 * interfere with successful delivery to the socket. 3954 * 3955 * This operation is only valid from TC ingress path. 3956 * 3957 * The *flags* argument must be zero. 3958 * Return 3959 * 0 on success, or a negative error in case of failure: 3960 * 3961 * **-EINVAL** if specified *flags* are not supported. 3962 * 3963 * **-ENOENT** if the socket is unavailable for assignment. 3964 * 3965 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3966 * 3967 * **-EOPNOTSUPP** if the operation is not supported, for example 3968 * a call from outside of TC ingress. 3969 * 3970 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3971 * (reuseport). 3972 * 3973 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3974 * Description 3975 * Helper is overloaded depending on BPF program type. This 3976 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3977 * 3978 * Select the *sk* as a result of a socket lookup. 3979 * 3980 * For the operation to succeed passed socket must be compatible 3981 * with the packet description provided by the *ctx* object. 3982 * 3983 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3984 * be an exact match. While IP family (**AF_INET** or 3985 * **AF_INET6**) must be compatible, that is IPv6 sockets 3986 * that are not v6-only can be selected for IPv4 packets. 3987 * 3988 * Only TCP listeners and UDP unconnected sockets can be 3989 * selected. *sk* can also be NULL to reset any previous 3990 * selection. 3991 * 3992 * *flags* argument can combination of following values: 3993 * 3994 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3995 * socket selection, potentially done by a BPF program 3996 * that ran before us. 3997 * 3998 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3999 * load-balancing within reuseport group for the socket 4000 * being selected. 4001 * 4002 * On success *ctx->sk* will point to the selected socket. 4003 * 4004 * Return 4005 * 0 on success, or a negative errno in case of failure. 4006 * 4007 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4008 * not compatible with packet family (*ctx->family*). 4009 * 4010 * * **-EEXIST** if socket has been already selected, 4011 * potentially by another program, and 4012 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4013 * 4014 * * **-EINVAL** if unsupported flags were specified. 4015 * 4016 * * **-EPROTOTYPE** if socket L4 protocol 4017 * (*sk->protocol*) doesn't match packet protocol 4018 * (*ctx->protocol*). 4019 * 4020 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4021 * state (TCP listening or UDP unconnected). 4022 * 4023 * u64 bpf_ktime_get_boot_ns(void) 4024 * Description 4025 * Return the time elapsed since system boot, in nanoseconds. 4026 * Does include the time the system was suspended. 4027 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4028 * Return 4029 * Current *ktime*. 4030 * 4031 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4032 * Description 4033 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4034 * out the format string. 4035 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4036 * the format string itself. The *data* and *data_len* are format string 4037 * arguments. The *data* are a **u64** array and corresponding format string 4038 * values are stored in the array. For strings and pointers where pointees 4039 * are accessed, only the pointer values are stored in the *data* array. 4040 * The *data_len* is the size of *data* in bytes. 4041 * 4042 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4043 * Reading kernel memory may fail due to either invalid address or 4044 * valid address but requiring a major memory fault. If reading kernel memory 4045 * fails, the string for **%s** will be an empty string, and the ip 4046 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4047 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4048 * Return 4049 * 0 on success, or a negative error in case of failure: 4050 * 4051 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4052 * by returning 1 from bpf program. 4053 * 4054 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4055 * 4056 * **-E2BIG** if *fmt* contains too many format specifiers. 4057 * 4058 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4059 * 4060 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4061 * Description 4062 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4063 * The *m* represents the seq_file. The *data* and *len* represent the 4064 * data to write in bytes. 4065 * Return 4066 * 0 on success, or a negative error in case of failure: 4067 * 4068 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4069 * 4070 * u64 bpf_sk_cgroup_id(void *sk) 4071 * Description 4072 * Return the cgroup v2 id of the socket *sk*. 4073 * 4074 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4075 * returned from **bpf_sk_lookup_xxx**\ (), 4076 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4077 * same as in **bpf_skb_cgroup_id**\ (). 4078 * 4079 * This helper is available only if the kernel was compiled with 4080 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4081 * Return 4082 * The id is returned or 0 in case the id could not be retrieved. 4083 * 4084 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4085 * Description 4086 * Return id of cgroup v2 that is ancestor of cgroup associated 4087 * with the *sk* at the *ancestor_level*. The root cgroup is at 4088 * *ancestor_level* zero and each step down the hierarchy 4089 * increments the level. If *ancestor_level* == level of cgroup 4090 * associated with *sk*, then return value will be same as that 4091 * of **bpf_sk_cgroup_id**\ (). 4092 * 4093 * The helper is useful to implement policies based on cgroups 4094 * that are upper in hierarchy than immediate cgroup associated 4095 * with *sk*. 4096 * 4097 * The format of returned id and helper limitations are same as in 4098 * **bpf_sk_cgroup_id**\ (). 4099 * Return 4100 * The id is returned or 0 in case the id could not be retrieved. 4101 * 4102 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4103 * Description 4104 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4105 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4106 * of new data availability is sent. 4107 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4108 * of new data availability is sent unconditionally. 4109 * If **0** is specified in *flags*, an adaptive notification 4110 * of new data availability is sent. 4111 * 4112 * An adaptive notification is a notification sent whenever the user-space 4113 * process has caught up and consumed all available payloads. In case the user-space 4114 * process is still processing a previous payload, then no notification is needed 4115 * as it will process the newly added payload automatically. 4116 * Return 4117 * 0 on success, or a negative error in case of failure. 4118 * 4119 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4120 * Description 4121 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4122 * *flags* must be 0. 4123 * Return 4124 * Valid pointer with *size* bytes of memory available; NULL, 4125 * otherwise. 4126 * 4127 * void bpf_ringbuf_submit(void *data, u64 flags) 4128 * Description 4129 * Submit reserved ring buffer sample, pointed to by *data*. 4130 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4131 * of new data availability is sent. 4132 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4133 * of new data availability is sent unconditionally. 4134 * If **0** is specified in *flags*, an adaptive notification 4135 * of new data availability is sent. 4136 * 4137 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4138 * Return 4139 * Nothing. Always succeeds. 4140 * 4141 * void bpf_ringbuf_discard(void *data, u64 flags) 4142 * Description 4143 * Discard reserved ring buffer sample, pointed to by *data*. 4144 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4145 * of new data availability is sent. 4146 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4147 * of new data availability is sent unconditionally. 4148 * If **0** is specified in *flags*, an adaptive notification 4149 * of new data availability is sent. 4150 * 4151 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4152 * Return 4153 * Nothing. Always succeeds. 4154 * 4155 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4156 * Description 4157 * Query various characteristics of provided ring buffer. What 4158 * exactly is queries is determined by *flags*: 4159 * 4160 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4161 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4162 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4163 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4164 * 4165 * Data returned is just a momentary snapshot of actual values 4166 * and could be inaccurate, so this facility should be used to 4167 * power heuristics and for reporting, not to make 100% correct 4168 * calculation. 4169 * Return 4170 * Requested value, or 0, if *flags* are not recognized. 4171 * 4172 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4173 * Description 4174 * Change the skbs checksum level by one layer up or down, or 4175 * reset it entirely to none in order to have the stack perform 4176 * checksum validation. The level is applicable to the following 4177 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4178 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4179 * through **bpf_skb_adjust_room**\ () helper with passing in 4180 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4181 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4182 * the UDP header is removed. Similarly, an encap of the latter 4183 * into the former could be accompanied by a helper call to 4184 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4185 * skb is still intended to be processed in higher layers of the 4186 * stack instead of just egressing at tc. 4187 * 4188 * There are three supported level settings at this time: 4189 * 4190 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4191 * with CHECKSUM_UNNECESSARY. 4192 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4193 * with CHECKSUM_UNNECESSARY. 4194 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4195 * sets CHECKSUM_NONE to force checksum validation by the stack. 4196 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4197 * skb->csum_level. 4198 * Return 4199 * 0 on success, or a negative error in case of failure. In the 4200 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4201 * is returned or the error code -EACCES in case the skb is not 4202 * subject to CHECKSUM_UNNECESSARY. 4203 * 4204 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4205 * Description 4206 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4207 * Return 4208 * *sk* if casting is valid, or **NULL** otherwise. 4209 * 4210 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4211 * Description 4212 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4213 * Return 4214 * *sk* if casting is valid, or **NULL** otherwise. 4215 * 4216 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4217 * Description 4218 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4219 * Return 4220 * *sk* if casting is valid, or **NULL** otherwise. 4221 * 4222 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4223 * Description 4224 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4225 * Return 4226 * *sk* if casting is valid, or **NULL** otherwise. 4227 * 4228 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4229 * Description 4230 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4231 * Return 4232 * *sk* if casting is valid, or **NULL** otherwise. 4233 * 4234 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4235 * Description 4236 * Return a user or a kernel stack in bpf program provided buffer. 4237 * To achieve this, the helper needs *task*, which is a valid 4238 * pointer to **struct task_struct**. To store the stacktrace, the 4239 * bpf program provides *buf* with a nonnegative *size*. 4240 * 4241 * The last argument, *flags*, holds the number of stack frames to 4242 * skip (from 0 to 255), masked with 4243 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4244 * the following flags: 4245 * 4246 * **BPF_F_USER_STACK** 4247 * Collect a user space stack instead of a kernel stack. 4248 * **BPF_F_USER_BUILD_ID** 4249 * Collect buildid+offset instead of ips for user stack, 4250 * only valid if **BPF_F_USER_STACK** is also specified. 4251 * 4252 * **bpf_get_task_stack**\ () can collect up to 4253 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4254 * to sufficient large buffer size. Note that 4255 * this limit can be controlled with the **sysctl** program, and 4256 * that it should be manually increased in order to profile long 4257 * user stacks (such as stacks for Java programs). To do so, use: 4258 * 4259 * :: 4260 * 4261 * # sysctl kernel.perf_event_max_stack=<new value> 4262 * Return 4263 * A non-negative value equal to or less than *size* on success, 4264 * or a negative error in case of failure. 4265 * 4266 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4267 * Description 4268 * Load header option. Support reading a particular TCP header 4269 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4270 * 4271 * If *flags* is 0, it will search the option from the 4272 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4273 * has details on what skb_data contains under different 4274 * *skops*\ **->op**. 4275 * 4276 * The first byte of the *searchby_res* specifies the 4277 * kind that it wants to search. 4278 * 4279 * If the searching kind is an experimental kind 4280 * (i.e. 253 or 254 according to RFC6994). It also 4281 * needs to specify the "magic" which is either 4282 * 2 bytes or 4 bytes. It then also needs to 4283 * specify the size of the magic by using 4284 * the 2nd byte which is "kind-length" of a TCP 4285 * header option and the "kind-length" also 4286 * includes the first 2 bytes "kind" and "kind-length" 4287 * itself as a normal TCP header option also does. 4288 * 4289 * For example, to search experimental kind 254 with 4290 * 2 byte magic 0xeB9F, the searchby_res should be 4291 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4292 * 4293 * To search for the standard window scale option (3), 4294 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4295 * Note, kind-length must be 0 for regular option. 4296 * 4297 * Searching for No-Op (0) and End-of-Option-List (1) are 4298 * not supported. 4299 * 4300 * *len* must be at least 2 bytes which is the minimal size 4301 * of a header option. 4302 * 4303 * Supported flags: 4304 * 4305 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4306 * saved_syn packet or the just-received syn packet. 4307 * 4308 * Return 4309 * > 0 when found, the header option is copied to *searchby_res*. 4310 * The return value is the total length copied. On failure, a 4311 * negative error code is returned: 4312 * 4313 * **-EINVAL** if a parameter is invalid. 4314 * 4315 * **-ENOMSG** if the option is not found. 4316 * 4317 * **-ENOENT** if no syn packet is available when 4318 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4319 * 4320 * **-ENOSPC** if there is not enough space. Only *len* number of 4321 * bytes are copied. 4322 * 4323 * **-EFAULT** on failure to parse the header options in the 4324 * packet. 4325 * 4326 * **-EPERM** if the helper cannot be used under the current 4327 * *skops*\ **->op**. 4328 * 4329 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4330 * Description 4331 * Store header option. The data will be copied 4332 * from buffer *from* with length *len* to the TCP header. 4333 * 4334 * The buffer *from* should have the whole option that 4335 * includes the kind, kind-length, and the actual 4336 * option data. The *len* must be at least kind-length 4337 * long. The kind-length does not have to be 4 byte 4338 * aligned. The kernel will take care of the padding 4339 * and setting the 4 bytes aligned value to th->doff. 4340 * 4341 * This helper will check for duplicated option 4342 * by searching the same option in the outgoing skb. 4343 * 4344 * This helper can only be called during 4345 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4346 * 4347 * Return 4348 * 0 on success, or negative error in case of failure: 4349 * 4350 * **-EINVAL** If param is invalid. 4351 * 4352 * **-ENOSPC** if there is not enough space in the header. 4353 * Nothing has been written 4354 * 4355 * **-EEXIST** if the option already exists. 4356 * 4357 * **-EFAULT** on failrue to parse the existing header options. 4358 * 4359 * **-EPERM** if the helper cannot be used under the current 4360 * *skops*\ **->op**. 4361 * 4362 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4363 * Description 4364 * Reserve *len* bytes for the bpf header option. The 4365 * space will be used by **bpf_store_hdr_opt**\ () later in 4366 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4367 * 4368 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4369 * the total number of bytes will be reserved. 4370 * 4371 * This helper can only be called during 4372 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4373 * 4374 * Return 4375 * 0 on success, or negative error in case of failure: 4376 * 4377 * **-EINVAL** if a parameter is invalid. 4378 * 4379 * **-ENOSPC** if there is not enough space in the header. 4380 * 4381 * **-EPERM** if the helper cannot be used under the current 4382 * *skops*\ **->op**. 4383 * 4384 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4385 * Description 4386 * Get a bpf_local_storage from an *inode*. 4387 * 4388 * Logically, it could be thought of as getting the value from 4389 * a *map* with *inode* as the **key**. From this 4390 * perspective, the usage is not much different from 4391 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4392 * helper enforces the key must be an inode and the map must also 4393 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4394 * 4395 * Underneath, the value is stored locally at *inode* instead of 4396 * the *map*. The *map* is used as the bpf-local-storage 4397 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4398 * searched against all bpf_local_storage residing at *inode*. 4399 * 4400 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4401 * used such that a new bpf_local_storage will be 4402 * created if one does not exist. *value* can be used 4403 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4404 * the initial value of a bpf_local_storage. If *value* is 4405 * **NULL**, the new bpf_local_storage will be zero initialized. 4406 * Return 4407 * A bpf_local_storage pointer is returned on success. 4408 * 4409 * **NULL** if not found or there was an error in adding 4410 * a new bpf_local_storage. 4411 * 4412 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4413 * Description 4414 * Delete a bpf_local_storage from an *inode*. 4415 * Return 4416 * 0 on success. 4417 * 4418 * **-ENOENT** if the bpf_local_storage cannot be found. 4419 * 4420 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4421 * Description 4422 * Return full path for given **struct path** object, which 4423 * needs to be the kernel BTF *path* object. The path is 4424 * returned in the provided buffer *buf* of size *sz* and 4425 * is zero terminated. 4426 * 4427 * Return 4428 * On success, the strictly positive length of the string, 4429 * including the trailing NUL character. On error, a negative 4430 * value. 4431 * 4432 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4433 * Description 4434 * Read *size* bytes from user space address *user_ptr* and store 4435 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4436 * Return 4437 * 0 on success, or a negative error in case of failure. 4438 * 4439 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4440 * Description 4441 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4442 * using *ptr*->type_id. This value should specify the type 4443 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4444 * can be used to look up vmlinux BTF type ids. Traversing the 4445 * data structure using BTF, the type information and values are 4446 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4447 * the pointer data is carried out to avoid kernel crashes during 4448 * operation. Smaller types can use string space on the stack; 4449 * larger programs can use map data to store the string 4450 * representation. 4451 * 4452 * The string can be subsequently shared with userspace via 4453 * bpf_perf_event_output() or ring buffer interfaces. 4454 * bpf_trace_printk() is to be avoided as it places too small 4455 * a limit on string size to be useful. 4456 * 4457 * *flags* is a combination of 4458 * 4459 * **BTF_F_COMPACT** 4460 * no formatting around type information 4461 * **BTF_F_NONAME** 4462 * no struct/union member names/types 4463 * **BTF_F_PTR_RAW** 4464 * show raw (unobfuscated) pointer values; 4465 * equivalent to printk specifier %px. 4466 * **BTF_F_ZERO** 4467 * show zero-valued struct/union members; they 4468 * are not displayed by default 4469 * 4470 * Return 4471 * The number of bytes that were written (or would have been 4472 * written if output had to be truncated due to string size), 4473 * or a negative error in cases of failure. 4474 * 4475 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4476 * Description 4477 * Use BTF to write to seq_write a string representation of 4478 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4479 * *flags* are identical to those used for bpf_snprintf_btf. 4480 * Return 4481 * 0 on success or a negative error in case of failure. 4482 * 4483 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4484 * Description 4485 * See **bpf_get_cgroup_classid**\ () for the main description. 4486 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4487 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4488 * associated socket instead of the current process. 4489 * Return 4490 * The id is returned or 0 in case the id could not be retrieved. 4491 * 4492 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4493 * Description 4494 * Redirect the packet to another net device of index *ifindex* 4495 * and fill in L2 addresses from neighboring subsystem. This helper 4496 * is somewhat similar to **bpf_redirect**\ (), except that it 4497 * populates L2 addresses as well, meaning, internally, the helper 4498 * relies on the neighbor lookup for the L2 address of the nexthop. 4499 * 4500 * The helper will perform a FIB lookup based on the skb's 4501 * networking header to get the address of the next hop, unless 4502 * this is supplied by the caller in the *params* argument. The 4503 * *plen* argument indicates the len of *params* and should be set 4504 * to 0 if *params* is NULL. 4505 * 4506 * The *flags* argument is reserved and must be 0. The helper is 4507 * currently only supported for tc BPF program types, and enabled 4508 * for IPv4 and IPv6 protocols. 4509 * Return 4510 * The helper returns **TC_ACT_REDIRECT** on success or 4511 * **TC_ACT_SHOT** on error. 4512 * 4513 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4514 * Description 4515 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4516 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4517 * extern variable decorated with '__ksym'. For ksym, there is a 4518 * global var (either static or global) defined of the same name 4519 * in the kernel. The ksym is percpu if the global var is percpu. 4520 * The returned pointer points to the global percpu var on *cpu*. 4521 * 4522 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4523 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4524 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4525 * bpf_per_cpu_ptr() must check the returned value. 4526 * Return 4527 * A pointer pointing to the kernel percpu variable on *cpu*, or 4528 * NULL, if *cpu* is invalid. 4529 * 4530 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4531 * Description 4532 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4533 * pointer to the percpu kernel variable on this cpu. See the 4534 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4535 * 4536 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4537 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4538 * never return NULL. 4539 * Return 4540 * A pointer pointing to the kernel percpu variable on this cpu. 4541 * 4542 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4543 * Description 4544 * Redirect the packet to another net device of index *ifindex*. 4545 * This helper is somewhat similar to **bpf_redirect**\ (), except 4546 * that the redirection happens to the *ifindex*' peer device and 4547 * the netns switch takes place from ingress to ingress without 4548 * going through the CPU's backlog queue. 4549 * 4550 * The *flags* argument is reserved and must be 0. The helper is 4551 * currently only supported for tc BPF program types at the ingress 4552 * hook and for veth device types. The peer device must reside in a 4553 * different network namespace. 4554 * Return 4555 * The helper returns **TC_ACT_REDIRECT** on success or 4556 * **TC_ACT_SHOT** on error. 4557 * 4558 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4559 * Description 4560 * Get a bpf_local_storage from the *task*. 4561 * 4562 * Logically, it could be thought of as getting the value from 4563 * a *map* with *task* as the **key**. From this 4564 * perspective, the usage is not much different from 4565 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4566 * helper enforces the key must be an task_struct and the map must also 4567 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4568 * 4569 * Underneath, the value is stored locally at *task* instead of 4570 * the *map*. The *map* is used as the bpf-local-storage 4571 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4572 * searched against all bpf_local_storage residing at *task*. 4573 * 4574 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4575 * used such that a new bpf_local_storage will be 4576 * created if one does not exist. *value* can be used 4577 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4578 * the initial value of a bpf_local_storage. If *value* is 4579 * **NULL**, the new bpf_local_storage will be zero initialized. 4580 * Return 4581 * A bpf_local_storage pointer is returned on success. 4582 * 4583 * **NULL** if not found or there was an error in adding 4584 * a new bpf_local_storage. 4585 * 4586 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4587 * Description 4588 * Delete a bpf_local_storage from a *task*. 4589 * Return 4590 * 0 on success. 4591 * 4592 * **-ENOENT** if the bpf_local_storage cannot be found. 4593 * 4594 * struct task_struct *bpf_get_current_task_btf(void) 4595 * Description 4596 * Return a BTF pointer to the "current" task. 4597 * This pointer can also be used in helpers that accept an 4598 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4599 * Return 4600 * Pointer to the current task. 4601 * 4602 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4603 * Description 4604 * Set or clear certain options on *bprm*: 4605 * 4606 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4607 * which sets the **AT_SECURE** auxv for glibc. The bit 4608 * is cleared if the flag is not specified. 4609 * Return 4610 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4611 * 4612 * u64 bpf_ktime_get_coarse_ns(void) 4613 * Description 4614 * Return a coarse-grained version of the time elapsed since 4615 * system boot, in nanoseconds. Does not include time the system 4616 * was suspended. 4617 * 4618 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4619 * Return 4620 * Current *ktime*. 4621 * 4622 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4623 * Description 4624 * Returns the stored IMA hash of the *inode* (if it's avaialable). 4625 * If the hash is larger than *size*, then only *size* 4626 * bytes will be copied to *dst* 4627 * Return 4628 * The **hash_algo** is returned on success, 4629 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4630 * invalid arguments are passed. 4631 * 4632 * struct socket *bpf_sock_from_file(struct file *file) 4633 * Description 4634 * If the given file represents a socket, returns the associated 4635 * socket. 4636 * Return 4637 * A pointer to a struct socket on success or NULL if the file is 4638 * not a socket. 4639 * 4640 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4641 * Description 4642 * Check packet size against exceeding MTU of net device (based 4643 * on *ifindex*). This helper will likely be used in combination 4644 * with helpers that adjust/change the packet size. 4645 * 4646 * The argument *len_diff* can be used for querying with a planned 4647 * size change. This allows to check MTU prior to changing packet 4648 * ctx. Providing an *len_diff* adjustment that is larger than the 4649 * actual packet size (resulting in negative packet size) will in 4650 * principle not exceed the MTU, why it is not considered a 4651 * failure. Other BPF-helpers are needed for performing the 4652 * planned size change, why the responsability for catch a negative 4653 * packet size belong in those helpers. 4654 * 4655 * Specifying *ifindex* zero means the MTU check is performed 4656 * against the current net device. This is practical if this isn't 4657 * used prior to redirect. 4658 * 4659 * On input *mtu_len* must be a valid pointer, else verifier will 4660 * reject BPF program. If the value *mtu_len* is initialized to 4661 * zero then the ctx packet size is use. When value *mtu_len* is 4662 * provided as input this specify the L3 length that the MTU check 4663 * is done against. Remember XDP and TC length operate at L2, but 4664 * this value is L3 as this correlate to MTU and IP-header tot_len 4665 * values which are L3 (similar behavior as bpf_fib_lookup). 4666 * 4667 * The Linux kernel route table can configure MTUs on a more 4668 * specific per route level, which is not provided by this helper. 4669 * For route level MTU checks use the **bpf_fib_lookup**\ () 4670 * helper. 4671 * 4672 * *ctx* is either **struct xdp_md** for XDP programs or 4673 * **struct sk_buff** for tc cls_act programs. 4674 * 4675 * The *flags* argument can be a combination of one or more of the 4676 * following values: 4677 * 4678 * **BPF_MTU_CHK_SEGS** 4679 * This flag will only works for *ctx* **struct sk_buff**. 4680 * If packet context contains extra packet segment buffers 4681 * (often knows as GSO skb), then MTU check is harder to 4682 * check at this point, because in transmit path it is 4683 * possible for the skb packet to get re-segmented 4684 * (depending on net device features). This could still be 4685 * a MTU violation, so this flag enables performing MTU 4686 * check against segments, with a different violation 4687 * return code to tell it apart. Check cannot use len_diff. 4688 * 4689 * On return *mtu_len* pointer contains the MTU value of the net 4690 * device. Remember the net device configured MTU is the L3 size, 4691 * which is returned here and XDP and TC length operate at L2. 4692 * Helper take this into account for you, but remember when using 4693 * MTU value in your BPF-code. 4694 * 4695 * Return 4696 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4697 * 4698 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4699 * 4700 * MTU violations return positive values, but also populate MTU 4701 * value in *mtu_len* pointer, as this can be needed for 4702 * implementing PMTU handing: 4703 * 4704 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4705 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4706 * 4707 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4708 * Description 4709 * For each element in **map**, call **callback_fn** function with 4710 * **map**, **callback_ctx** and other map-specific parameters. 4711 * The **callback_fn** should be a static function and 4712 * the **callback_ctx** should be a pointer to the stack. 4713 * The **flags** is used to control certain aspects of the helper. 4714 * Currently, the **flags** must be 0. 4715 * 4716 * The following are a list of supported map types and their 4717 * respective expected callback signatures: 4718 * 4719 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4720 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4721 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4722 * 4723 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4724 * 4725 * For per_cpu maps, the map_value is the value on the cpu where the 4726 * bpf_prog is running. 4727 * 4728 * If **callback_fn** return 0, the helper will continue to the next 4729 * element. If return value is 1, the helper will skip the rest of 4730 * elements and return. Other return values are not used now. 4731 * 4732 * Return 4733 * The number of traversed map elements for success, **-EINVAL** for 4734 * invalid **flags**. 4735 * 4736 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4737 * Description 4738 * Outputs a string into the **str** buffer of size **str_size** 4739 * based on a format string stored in a read-only map pointed by 4740 * **fmt**. 4741 * 4742 * Each format specifier in **fmt** corresponds to one u64 element 4743 * in the **data** array. For strings and pointers where pointees 4744 * are accessed, only the pointer values are stored in the *data* 4745 * array. The *data_len* is the size of *data* in bytes. 4746 * 4747 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4748 * memory. Reading kernel memory may fail due to either invalid 4749 * address or valid address but requiring a major memory fault. If 4750 * reading kernel memory fails, the string for **%s** will be an 4751 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4752 * Not returning error to bpf program is consistent with what 4753 * **bpf_trace_printk**\ () does for now. 4754 * 4755 * Return 4756 * The strictly positive length of the formatted string, including 4757 * the trailing zero character. If the return value is greater than 4758 * **str_size**, **str** contains a truncated string, guaranteed to 4759 * be zero-terminated except when **str_size** is 0. 4760 * 4761 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4762 * 4763 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4764 * Description 4765 * Execute bpf syscall with given arguments. 4766 * Return 4767 * A syscall result. 4768 * 4769 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4770 * Description 4771 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4772 * Return 4773 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4774 * 4775 * long bpf_sys_close(u32 fd) 4776 * Description 4777 * Execute close syscall for given FD. 4778 * Return 4779 * A syscall result. 4780 * 4781 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4782 * Description 4783 * Initialize the timer. 4784 * First 4 bits of *flags* specify clockid. 4785 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4786 * All other bits of *flags* are reserved. 4787 * The verifier will reject the program if *timer* is not from 4788 * the same *map*. 4789 * Return 4790 * 0 on success. 4791 * **-EBUSY** if *timer* is already initialized. 4792 * **-EINVAL** if invalid *flags* are passed. 4793 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4794 * The user space should either hold a file descriptor to a map with timers 4795 * or pin such map in bpffs. When map is unpinned or file descriptor is 4796 * closed all timers in the map will be cancelled and freed. 4797 * 4798 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4799 * Description 4800 * Configure the timer to call *callback_fn* static function. 4801 * Return 4802 * 0 on success. 4803 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4804 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4805 * The user space should either hold a file descriptor to a map with timers 4806 * or pin such map in bpffs. When map is unpinned or file descriptor is 4807 * closed all timers in the map will be cancelled and freed. 4808 * 4809 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4810 * Description 4811 * Set timer expiration N nanoseconds from the current time. The 4812 * configured callback will be invoked in soft irq context on some cpu 4813 * and will not repeat unless another bpf_timer_start() is made. 4814 * In such case the next invocation can migrate to a different cpu. 4815 * Since struct bpf_timer is a field inside map element the map 4816 * owns the timer. The bpf_timer_set_callback() will increment refcnt 4817 * of BPF program to make sure that callback_fn code stays valid. 4818 * When user space reference to a map reaches zero all timers 4819 * in a map are cancelled and corresponding program's refcnts are 4820 * decremented. This is done to make sure that Ctrl-C of a user 4821 * process doesn't leave any timers running. If map is pinned in 4822 * bpffs the callback_fn can re-arm itself indefinitely. 4823 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4824 * cancel and free the timer in the given map element. 4825 * The map can contain timers that invoke callback_fn-s from different 4826 * programs. The same callback_fn can serve different timers from 4827 * different maps if key/value layout matches across maps. 4828 * Every bpf_timer_set_callback() can have different callback_fn. 4829 * 4830 * Return 4831 * 0 on success. 4832 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4833 * or invalid *flags* are passed. 4834 * 4835 * long bpf_timer_cancel(struct bpf_timer *timer) 4836 * Description 4837 * Cancel the timer and wait for callback_fn to finish if it was running. 4838 * Return 4839 * 0 if the timer was not active. 4840 * 1 if the timer was active. 4841 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4842 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4843 * own timer which would have led to a deadlock otherwise. 4844 * 4845 * u64 bpf_get_func_ip(void *ctx) 4846 * Description 4847 * Get address of the traced function (for tracing and kprobe programs). 4848 * Return 4849 * Address of the traced function. 4850 */ 4851 #define __BPF_FUNC_MAPPER(FN) \ 4852 FN(unspec), \ 4853 FN(map_lookup_elem), \ 4854 FN(map_update_elem), \ 4855 FN(map_delete_elem), \ 4856 FN(probe_read), \ 4857 FN(ktime_get_ns), \ 4858 FN(trace_printk), \ 4859 FN(get_prandom_u32), \ 4860 FN(get_smp_processor_id), \ 4861 FN(skb_store_bytes), \ 4862 FN(l3_csum_replace), \ 4863 FN(l4_csum_replace), \ 4864 FN(tail_call), \ 4865 FN(clone_redirect), \ 4866 FN(get_current_pid_tgid), \ 4867 FN(get_current_uid_gid), \ 4868 FN(get_current_comm), \ 4869 FN(get_cgroup_classid), \ 4870 FN(skb_vlan_push), \ 4871 FN(skb_vlan_pop), \ 4872 FN(skb_get_tunnel_key), \ 4873 FN(skb_set_tunnel_key), \ 4874 FN(perf_event_read), \ 4875 FN(redirect), \ 4876 FN(get_route_realm), \ 4877 FN(perf_event_output), \ 4878 FN(skb_load_bytes), \ 4879 FN(get_stackid), \ 4880 FN(csum_diff), \ 4881 FN(skb_get_tunnel_opt), \ 4882 FN(skb_set_tunnel_opt), \ 4883 FN(skb_change_proto), \ 4884 FN(skb_change_type), \ 4885 FN(skb_under_cgroup), \ 4886 FN(get_hash_recalc), \ 4887 FN(get_current_task), \ 4888 FN(probe_write_user), \ 4889 FN(current_task_under_cgroup), \ 4890 FN(skb_change_tail), \ 4891 FN(skb_pull_data), \ 4892 FN(csum_update), \ 4893 FN(set_hash_invalid), \ 4894 FN(get_numa_node_id), \ 4895 FN(skb_change_head), \ 4896 FN(xdp_adjust_head), \ 4897 FN(probe_read_str), \ 4898 FN(get_socket_cookie), \ 4899 FN(get_socket_uid), \ 4900 FN(set_hash), \ 4901 FN(setsockopt), \ 4902 FN(skb_adjust_room), \ 4903 FN(redirect_map), \ 4904 FN(sk_redirect_map), \ 4905 FN(sock_map_update), \ 4906 FN(xdp_adjust_meta), \ 4907 FN(perf_event_read_value), \ 4908 FN(perf_prog_read_value), \ 4909 FN(getsockopt), \ 4910 FN(override_return), \ 4911 FN(sock_ops_cb_flags_set), \ 4912 FN(msg_redirect_map), \ 4913 FN(msg_apply_bytes), \ 4914 FN(msg_cork_bytes), \ 4915 FN(msg_pull_data), \ 4916 FN(bind), \ 4917 FN(xdp_adjust_tail), \ 4918 FN(skb_get_xfrm_state), \ 4919 FN(get_stack), \ 4920 FN(skb_load_bytes_relative), \ 4921 FN(fib_lookup), \ 4922 FN(sock_hash_update), \ 4923 FN(msg_redirect_hash), \ 4924 FN(sk_redirect_hash), \ 4925 FN(lwt_push_encap), \ 4926 FN(lwt_seg6_store_bytes), \ 4927 FN(lwt_seg6_adjust_srh), \ 4928 FN(lwt_seg6_action), \ 4929 FN(rc_repeat), \ 4930 FN(rc_keydown), \ 4931 FN(skb_cgroup_id), \ 4932 FN(get_current_cgroup_id), \ 4933 FN(get_local_storage), \ 4934 FN(sk_select_reuseport), \ 4935 FN(skb_ancestor_cgroup_id), \ 4936 FN(sk_lookup_tcp), \ 4937 FN(sk_lookup_udp), \ 4938 FN(sk_release), \ 4939 FN(map_push_elem), \ 4940 FN(map_pop_elem), \ 4941 FN(map_peek_elem), \ 4942 FN(msg_push_data), \ 4943 FN(msg_pop_data), \ 4944 FN(rc_pointer_rel), \ 4945 FN(spin_lock), \ 4946 FN(spin_unlock), \ 4947 FN(sk_fullsock), \ 4948 FN(tcp_sock), \ 4949 FN(skb_ecn_set_ce), \ 4950 FN(get_listener_sock), \ 4951 FN(skc_lookup_tcp), \ 4952 FN(tcp_check_syncookie), \ 4953 FN(sysctl_get_name), \ 4954 FN(sysctl_get_current_value), \ 4955 FN(sysctl_get_new_value), \ 4956 FN(sysctl_set_new_value), \ 4957 FN(strtol), \ 4958 FN(strtoul), \ 4959 FN(sk_storage_get), \ 4960 FN(sk_storage_delete), \ 4961 FN(send_signal), \ 4962 FN(tcp_gen_syncookie), \ 4963 FN(skb_output), \ 4964 FN(probe_read_user), \ 4965 FN(probe_read_kernel), \ 4966 FN(probe_read_user_str), \ 4967 FN(probe_read_kernel_str), \ 4968 FN(tcp_send_ack), \ 4969 FN(send_signal_thread), \ 4970 FN(jiffies64), \ 4971 FN(read_branch_records), \ 4972 FN(get_ns_current_pid_tgid), \ 4973 FN(xdp_output), \ 4974 FN(get_netns_cookie), \ 4975 FN(get_current_ancestor_cgroup_id), \ 4976 FN(sk_assign), \ 4977 FN(ktime_get_boot_ns), \ 4978 FN(seq_printf), \ 4979 FN(seq_write), \ 4980 FN(sk_cgroup_id), \ 4981 FN(sk_ancestor_cgroup_id), \ 4982 FN(ringbuf_output), \ 4983 FN(ringbuf_reserve), \ 4984 FN(ringbuf_submit), \ 4985 FN(ringbuf_discard), \ 4986 FN(ringbuf_query), \ 4987 FN(csum_level), \ 4988 FN(skc_to_tcp6_sock), \ 4989 FN(skc_to_tcp_sock), \ 4990 FN(skc_to_tcp_timewait_sock), \ 4991 FN(skc_to_tcp_request_sock), \ 4992 FN(skc_to_udp6_sock), \ 4993 FN(get_task_stack), \ 4994 FN(load_hdr_opt), \ 4995 FN(store_hdr_opt), \ 4996 FN(reserve_hdr_opt), \ 4997 FN(inode_storage_get), \ 4998 FN(inode_storage_delete), \ 4999 FN(d_path), \ 5000 FN(copy_from_user), \ 5001 FN(snprintf_btf), \ 5002 FN(seq_printf_btf), \ 5003 FN(skb_cgroup_classid), \ 5004 FN(redirect_neigh), \ 5005 FN(per_cpu_ptr), \ 5006 FN(this_cpu_ptr), \ 5007 FN(redirect_peer), \ 5008 FN(task_storage_get), \ 5009 FN(task_storage_delete), \ 5010 FN(get_current_task_btf), \ 5011 FN(bprm_opts_set), \ 5012 FN(ktime_get_coarse_ns), \ 5013 FN(ima_inode_hash), \ 5014 FN(sock_from_file), \ 5015 FN(check_mtu), \ 5016 FN(for_each_map_elem), \ 5017 FN(snprintf), \ 5018 FN(sys_bpf), \ 5019 FN(btf_find_by_name_kind), \ 5020 FN(sys_close), \ 5021 FN(timer_init), \ 5022 FN(timer_set_callback), \ 5023 FN(timer_start), \ 5024 FN(timer_cancel), \ 5025 FN(get_func_ip), \ 5026 /* */ 5027 5028 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5029 * function eBPF program intends to call 5030 */ 5031 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5032 enum bpf_func_id { 5033 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5034 __BPF_FUNC_MAX_ID, 5035 }; 5036 #undef __BPF_ENUM_FN 5037 5038 /* All flags used by eBPF helper functions, placed here. */ 5039 5040 /* BPF_FUNC_skb_store_bytes flags. */ 5041 enum { 5042 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5043 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5044 }; 5045 5046 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5047 * First 4 bits are for passing the header field size. 5048 */ 5049 enum { 5050 BPF_F_HDR_FIELD_MASK = 0xfULL, 5051 }; 5052 5053 /* BPF_FUNC_l4_csum_replace flags. */ 5054 enum { 5055 BPF_F_PSEUDO_HDR = (1ULL << 4), 5056 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5057 BPF_F_MARK_ENFORCE = (1ULL << 6), 5058 }; 5059 5060 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5061 enum { 5062 BPF_F_INGRESS = (1ULL << 0), 5063 }; 5064 5065 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5066 enum { 5067 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5068 }; 5069 5070 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5071 enum { 5072 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5073 BPF_F_USER_STACK = (1ULL << 8), 5074 /* flags used by BPF_FUNC_get_stackid only. */ 5075 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5076 BPF_F_REUSE_STACKID = (1ULL << 10), 5077 /* flags used by BPF_FUNC_get_stack only. */ 5078 BPF_F_USER_BUILD_ID = (1ULL << 11), 5079 }; 5080 5081 /* BPF_FUNC_skb_set_tunnel_key flags. */ 5082 enum { 5083 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5084 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5085 BPF_F_SEQ_NUMBER = (1ULL << 3), 5086 }; 5087 5088 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5089 * BPF_FUNC_perf_event_read_value flags. 5090 */ 5091 enum { 5092 BPF_F_INDEX_MASK = 0xffffffffULL, 5093 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5094 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5095 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5096 }; 5097 5098 /* Current network namespace */ 5099 enum { 5100 BPF_F_CURRENT_NETNS = (-1L), 5101 }; 5102 5103 /* BPF_FUNC_csum_level level values. */ 5104 enum { 5105 BPF_CSUM_LEVEL_QUERY, 5106 BPF_CSUM_LEVEL_INC, 5107 BPF_CSUM_LEVEL_DEC, 5108 BPF_CSUM_LEVEL_RESET, 5109 }; 5110 5111 /* BPF_FUNC_skb_adjust_room flags. */ 5112 enum { 5113 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5114 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5115 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5116 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5117 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5118 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5119 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5120 }; 5121 5122 enum { 5123 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5124 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5125 }; 5126 5127 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5128 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5129 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5130 5131 /* BPF_FUNC_sysctl_get_name flags. */ 5132 enum { 5133 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5134 }; 5135 5136 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 5137 enum { 5138 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5139 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5140 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5141 */ 5142 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5143 }; 5144 5145 /* BPF_FUNC_read_branch_records flags. */ 5146 enum { 5147 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5148 }; 5149 5150 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5151 * BPF_FUNC_bpf_ringbuf_output flags. 5152 */ 5153 enum { 5154 BPF_RB_NO_WAKEUP = (1ULL << 0), 5155 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5156 }; 5157 5158 /* BPF_FUNC_bpf_ringbuf_query flags */ 5159 enum { 5160 BPF_RB_AVAIL_DATA = 0, 5161 BPF_RB_RING_SIZE = 1, 5162 BPF_RB_CONS_POS = 2, 5163 BPF_RB_PROD_POS = 3, 5164 }; 5165 5166 /* BPF ring buffer constants */ 5167 enum { 5168 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5169 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5170 BPF_RINGBUF_HDR_SZ = 8, 5171 }; 5172 5173 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5174 enum { 5175 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5176 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5177 }; 5178 5179 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 5180 enum bpf_adj_room_mode { 5181 BPF_ADJ_ROOM_NET, 5182 BPF_ADJ_ROOM_MAC, 5183 }; 5184 5185 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5186 enum bpf_hdr_start_off { 5187 BPF_HDR_START_MAC, 5188 BPF_HDR_START_NET, 5189 }; 5190 5191 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5192 enum bpf_lwt_encap_mode { 5193 BPF_LWT_ENCAP_SEG6, 5194 BPF_LWT_ENCAP_SEG6_INLINE, 5195 BPF_LWT_ENCAP_IP, 5196 }; 5197 5198 /* Flags for bpf_bprm_opts_set helper */ 5199 enum { 5200 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5201 }; 5202 5203 /* Flags for bpf_redirect_map helper */ 5204 enum { 5205 BPF_F_BROADCAST = (1ULL << 3), 5206 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5207 }; 5208 5209 #define __bpf_md_ptr(type, name) \ 5210 union { \ 5211 type name; \ 5212 __u64 :64; \ 5213 } __attribute__((aligned(8))) 5214 5215 /* user accessible mirror of in-kernel sk_buff. 5216 * new fields can only be added to the end of this structure 5217 */ 5218 struct __sk_buff { 5219 __u32 len; 5220 __u32 pkt_type; 5221 __u32 mark; 5222 __u32 queue_mapping; 5223 __u32 protocol; 5224 __u32 vlan_present; 5225 __u32 vlan_tci; 5226 __u32 vlan_proto; 5227 __u32 priority; 5228 __u32 ingress_ifindex; 5229 __u32 ifindex; 5230 __u32 tc_index; 5231 __u32 cb[5]; 5232 __u32 hash; 5233 __u32 tc_classid; 5234 __u32 data; 5235 __u32 data_end; 5236 __u32 napi_id; 5237 5238 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 5239 __u32 family; 5240 __u32 remote_ip4; /* Stored in network byte order */ 5241 __u32 local_ip4; /* Stored in network byte order */ 5242 __u32 remote_ip6[4]; /* Stored in network byte order */ 5243 __u32 local_ip6[4]; /* Stored in network byte order */ 5244 __u32 remote_port; /* Stored in network byte order */ 5245 __u32 local_port; /* stored in host byte order */ 5246 /* ... here. */ 5247 5248 __u32 data_meta; 5249 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5250 __u64 tstamp; 5251 __u32 wire_len; 5252 __u32 gso_segs; 5253 __bpf_md_ptr(struct bpf_sock *, sk); 5254 __u32 gso_size; 5255 }; 5256 5257 struct bpf_tunnel_key { 5258 __u32 tunnel_id; 5259 union { 5260 __u32 remote_ipv4; 5261 __u32 remote_ipv6[4]; 5262 }; 5263 __u8 tunnel_tos; 5264 __u8 tunnel_ttl; 5265 __u16 tunnel_ext; /* Padding, future use. */ 5266 __u32 tunnel_label; 5267 }; 5268 5269 /* user accessible mirror of in-kernel xfrm_state. 5270 * new fields can only be added to the end of this structure 5271 */ 5272 struct bpf_xfrm_state { 5273 __u32 reqid; 5274 __u32 spi; /* Stored in network byte order */ 5275 __u16 family; 5276 __u16 ext; /* Padding, future use. */ 5277 union { 5278 __u32 remote_ipv4; /* Stored in network byte order */ 5279 __u32 remote_ipv6[4]; /* Stored in network byte order */ 5280 }; 5281 }; 5282 5283 /* Generic BPF return codes which all BPF program types may support. 5284 * The values are binary compatible with their TC_ACT_* counter-part to 5285 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 5286 * programs. 5287 * 5288 * XDP is handled seprately, see XDP_*. 5289 */ 5290 enum bpf_ret_code { 5291 BPF_OK = 0, 5292 /* 1 reserved */ 5293 BPF_DROP = 2, 5294 /* 3-6 reserved */ 5295 BPF_REDIRECT = 7, 5296 /* >127 are reserved for prog type specific return codes. 5297 * 5298 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 5299 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 5300 * changed and should be routed based on its new L3 header. 5301 * (This is an L3 redirect, as opposed to L2 redirect 5302 * represented by BPF_REDIRECT above). 5303 */ 5304 BPF_LWT_REROUTE = 128, 5305 }; 5306 5307 struct bpf_sock { 5308 __u32 bound_dev_if; 5309 __u32 family; 5310 __u32 type; 5311 __u32 protocol; 5312 __u32 mark; 5313 __u32 priority; 5314 /* IP address also allows 1 and 2 bytes access */ 5315 __u32 src_ip4; 5316 __u32 src_ip6[4]; 5317 __u32 src_port; /* host byte order */ 5318 __u32 dst_port; /* network byte order */ 5319 __u32 dst_ip4; 5320 __u32 dst_ip6[4]; 5321 __u32 state; 5322 __s32 rx_queue_mapping; 5323 }; 5324 5325 struct bpf_tcp_sock { 5326 __u32 snd_cwnd; /* Sending congestion window */ 5327 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5328 __u32 rtt_min; 5329 __u32 snd_ssthresh; /* Slow start size threshold */ 5330 __u32 rcv_nxt; /* What we want to receive next */ 5331 __u32 snd_nxt; /* Next sequence we send */ 5332 __u32 snd_una; /* First byte we want an ack for */ 5333 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5334 __u32 ecn_flags; /* ECN status bits. */ 5335 __u32 rate_delivered; /* saved rate sample: packets delivered */ 5336 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5337 __u32 packets_out; /* Packets which are "in flight" */ 5338 __u32 retrans_out; /* Retransmitted packets out */ 5339 __u32 total_retrans; /* Total retransmits for entire connection */ 5340 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5341 * total number of segments in. 5342 */ 5343 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5344 * total number of data segments in. 5345 */ 5346 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5347 * The total number of segments sent. 5348 */ 5349 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5350 * total number of data segments sent. 5351 */ 5352 __u32 lost_out; /* Lost packets */ 5353 __u32 sacked_out; /* SACK'd packets */ 5354 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5355 * sum(delta(rcv_nxt)), or how many bytes 5356 * were acked. 5357 */ 5358 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5359 * sum(delta(snd_una)), or how many bytes 5360 * were acked. 5361 */ 5362 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 5363 * total number of DSACK blocks received 5364 */ 5365 __u32 delivered; /* Total data packets delivered incl. rexmits */ 5366 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5367 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5368 }; 5369 5370 struct bpf_sock_tuple { 5371 union { 5372 struct { 5373 __be32 saddr; 5374 __be32 daddr; 5375 __be16 sport; 5376 __be16 dport; 5377 } ipv4; 5378 struct { 5379 __be32 saddr[4]; 5380 __be32 daddr[4]; 5381 __be16 sport; 5382 __be16 dport; 5383 } ipv6; 5384 }; 5385 }; 5386 5387 struct bpf_xdp_sock { 5388 __u32 queue_id; 5389 }; 5390 5391 #define XDP_PACKET_HEADROOM 256 5392 5393 /* User return codes for XDP prog type. 5394 * A valid XDP program must return one of these defined values. All other 5395 * return codes are reserved for future use. Unknown return codes will 5396 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 5397 */ 5398 enum xdp_action { 5399 XDP_ABORTED = 0, 5400 XDP_DROP, 5401 XDP_PASS, 5402 XDP_TX, 5403 XDP_REDIRECT, 5404 }; 5405 5406 /* user accessible metadata for XDP packet hook 5407 * new fields must be added to the end of this structure 5408 */ 5409 struct xdp_md { 5410 __u32 data; 5411 __u32 data_end; 5412 __u32 data_meta; 5413 /* Below access go through struct xdp_rxq_info */ 5414 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 5415 __u32 rx_queue_index; /* rxq->queue_index */ 5416 5417 __u32 egress_ifindex; /* txq->dev->ifindex */ 5418 }; 5419 5420 /* DEVMAP map-value layout 5421 * 5422 * The struct data-layout of map-value is a configuration interface. 5423 * New members can only be added to the end of this structure. 5424 */ 5425 struct bpf_devmap_val { 5426 __u32 ifindex; /* device index */ 5427 union { 5428 int fd; /* prog fd on map write */ 5429 __u32 id; /* prog id on map read */ 5430 } bpf_prog; 5431 }; 5432 5433 /* CPUMAP map-value layout 5434 * 5435 * The struct data-layout of map-value is a configuration interface. 5436 * New members can only be added to the end of this structure. 5437 */ 5438 struct bpf_cpumap_val { 5439 __u32 qsize; /* queue size to remote target CPU */ 5440 union { 5441 int fd; /* prog fd on map write */ 5442 __u32 id; /* prog id on map read */ 5443 } bpf_prog; 5444 }; 5445 5446 enum sk_action { 5447 SK_DROP = 0, 5448 SK_PASS, 5449 }; 5450 5451 /* user accessible metadata for SK_MSG packet hook, new fields must 5452 * be added to the end of this structure 5453 */ 5454 struct sk_msg_md { 5455 __bpf_md_ptr(void *, data); 5456 __bpf_md_ptr(void *, data_end); 5457 5458 __u32 family; 5459 __u32 remote_ip4; /* Stored in network byte order */ 5460 __u32 local_ip4; /* Stored in network byte order */ 5461 __u32 remote_ip6[4]; /* Stored in network byte order */ 5462 __u32 local_ip6[4]; /* Stored in network byte order */ 5463 __u32 remote_port; /* Stored in network byte order */ 5464 __u32 local_port; /* stored in host byte order */ 5465 __u32 size; /* Total size of sk_msg */ 5466 5467 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 5468 }; 5469 5470 struct sk_reuseport_md { 5471 /* 5472 * Start of directly accessible data. It begins from 5473 * the tcp/udp header. 5474 */ 5475 __bpf_md_ptr(void *, data); 5476 /* End of directly accessible data */ 5477 __bpf_md_ptr(void *, data_end); 5478 /* 5479 * Total length of packet (starting from the tcp/udp header). 5480 * Note that the directly accessible bytes (data_end - data) 5481 * could be less than this "len". Those bytes could be 5482 * indirectly read by a helper "bpf_skb_load_bytes()". 5483 */ 5484 __u32 len; 5485 /* 5486 * Eth protocol in the mac header (network byte order). e.g. 5487 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 5488 */ 5489 __u32 eth_protocol; 5490 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 5491 __u32 bind_inany; /* Is sock bound to an INANY address? */ 5492 __u32 hash; /* A hash of the packet 4 tuples */ 5493 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 5494 * new incoming connection request (e.g. selecting a listen sk for 5495 * the received SYN in the TCP case). reuse->sk is one of the sk 5496 * in the reuseport group. The bpf prog can use reuse->sk to learn 5497 * the local listening ip/port without looking into the skb. 5498 * 5499 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 5500 * reuse->migrating_sk is the socket that needs to be migrated 5501 * to another listening socket. migrating_sk could be a fullsock 5502 * sk that is fully established or a reqsk that is in-the-middle 5503 * of 3-way handshake. 5504 */ 5505 __bpf_md_ptr(struct bpf_sock *, sk); 5506 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 5507 }; 5508 5509 #define BPF_TAG_SIZE 8 5510 5511 struct bpf_prog_info { 5512 __u32 type; 5513 __u32 id; 5514 __u8 tag[BPF_TAG_SIZE]; 5515 __u32 jited_prog_len; 5516 __u32 xlated_prog_len; 5517 __aligned_u64 jited_prog_insns; 5518 __aligned_u64 xlated_prog_insns; 5519 __u64 load_time; /* ns since boottime */ 5520 __u32 created_by_uid; 5521 __u32 nr_map_ids; 5522 __aligned_u64 map_ids; 5523 char name[BPF_OBJ_NAME_LEN]; 5524 __u32 ifindex; 5525 __u32 gpl_compatible:1; 5526 __u32 :31; /* alignment pad */ 5527 __u64 netns_dev; 5528 __u64 netns_ino; 5529 __u32 nr_jited_ksyms; 5530 __u32 nr_jited_func_lens; 5531 __aligned_u64 jited_ksyms; 5532 __aligned_u64 jited_func_lens; 5533 __u32 btf_id; 5534 __u32 func_info_rec_size; 5535 __aligned_u64 func_info; 5536 __u32 nr_func_info; 5537 __u32 nr_line_info; 5538 __aligned_u64 line_info; 5539 __aligned_u64 jited_line_info; 5540 __u32 nr_jited_line_info; 5541 __u32 line_info_rec_size; 5542 __u32 jited_line_info_rec_size; 5543 __u32 nr_prog_tags; 5544 __aligned_u64 prog_tags; 5545 __u64 run_time_ns; 5546 __u64 run_cnt; 5547 __u64 recursion_misses; 5548 } __attribute__((aligned(8))); 5549 5550 struct bpf_map_info { 5551 __u32 type; 5552 __u32 id; 5553 __u32 key_size; 5554 __u32 value_size; 5555 __u32 max_entries; 5556 __u32 map_flags; 5557 char name[BPF_OBJ_NAME_LEN]; 5558 __u32 ifindex; 5559 __u32 btf_vmlinux_value_type_id; 5560 __u64 netns_dev; 5561 __u64 netns_ino; 5562 __u32 btf_id; 5563 __u32 btf_key_type_id; 5564 __u32 btf_value_type_id; 5565 } __attribute__((aligned(8))); 5566 5567 struct bpf_btf_info { 5568 __aligned_u64 btf; 5569 __u32 btf_size; 5570 __u32 id; 5571 __aligned_u64 name; 5572 __u32 name_len; 5573 __u32 kernel_btf; 5574 } __attribute__((aligned(8))); 5575 5576 struct bpf_link_info { 5577 __u32 type; 5578 __u32 id; 5579 __u32 prog_id; 5580 union { 5581 struct { 5582 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5583 __u32 tp_name_len; /* in/out: tp_name buffer len */ 5584 } raw_tracepoint; 5585 struct { 5586 __u32 attach_type; 5587 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5588 __u32 target_btf_id; /* BTF type id inside the object */ 5589 } tracing; 5590 struct { 5591 __u64 cgroup_id; 5592 __u32 attach_type; 5593 } cgroup; 5594 struct { 5595 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 5596 __u32 target_name_len; /* in/out: target_name buffer len */ 5597 union { 5598 struct { 5599 __u32 map_id; 5600 } map; 5601 }; 5602 } iter; 5603 struct { 5604 __u32 netns_ino; 5605 __u32 attach_type; 5606 } netns; 5607 struct { 5608 __u32 ifindex; 5609 } xdp; 5610 }; 5611 } __attribute__((aligned(8))); 5612 5613 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 5614 * by user and intended to be used by socket (e.g. to bind to, depends on 5615 * attach type). 5616 */ 5617 struct bpf_sock_addr { 5618 __u32 user_family; /* Allows 4-byte read, but no write. */ 5619 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5620 * Stored in network byte order. 5621 */ 5622 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5623 * Stored in network byte order. 5624 */ 5625 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 5626 * Stored in network byte order 5627 */ 5628 __u32 family; /* Allows 4-byte read, but no write */ 5629 __u32 type; /* Allows 4-byte read, but no write */ 5630 __u32 protocol; /* Allows 4-byte read, but no write */ 5631 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5632 * Stored in network byte order. 5633 */ 5634 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5635 * Stored in network byte order. 5636 */ 5637 __bpf_md_ptr(struct bpf_sock *, sk); 5638 }; 5639 5640 /* User bpf_sock_ops struct to access socket values and specify request ops 5641 * and their replies. 5642 * Some of this fields are in network (bigendian) byte order and may need 5643 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 5644 * New fields can only be added at the end of this structure 5645 */ 5646 struct bpf_sock_ops { 5647 __u32 op; 5648 union { 5649 __u32 args[4]; /* Optionally passed to bpf program */ 5650 __u32 reply; /* Returned by bpf program */ 5651 __u32 replylong[4]; /* Optionally returned by bpf prog */ 5652 }; 5653 __u32 family; 5654 __u32 remote_ip4; /* Stored in network byte order */ 5655 __u32 local_ip4; /* Stored in network byte order */ 5656 __u32 remote_ip6[4]; /* Stored in network byte order */ 5657 __u32 local_ip6[4]; /* Stored in network byte order */ 5658 __u32 remote_port; /* Stored in network byte order */ 5659 __u32 local_port; /* stored in host byte order */ 5660 __u32 is_fullsock; /* Some TCP fields are only valid if 5661 * there is a full socket. If not, the 5662 * fields read as zero. 5663 */ 5664 __u32 snd_cwnd; 5665 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5666 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 5667 __u32 state; 5668 __u32 rtt_min; 5669 __u32 snd_ssthresh; 5670 __u32 rcv_nxt; 5671 __u32 snd_nxt; 5672 __u32 snd_una; 5673 __u32 mss_cache; 5674 __u32 ecn_flags; 5675 __u32 rate_delivered; 5676 __u32 rate_interval_us; 5677 __u32 packets_out; 5678 __u32 retrans_out; 5679 __u32 total_retrans; 5680 __u32 segs_in; 5681 __u32 data_segs_in; 5682 __u32 segs_out; 5683 __u32 data_segs_out; 5684 __u32 lost_out; 5685 __u32 sacked_out; 5686 __u32 sk_txhash; 5687 __u64 bytes_received; 5688 __u64 bytes_acked; 5689 __bpf_md_ptr(struct bpf_sock *, sk); 5690 /* [skb_data, skb_data_end) covers the whole TCP header. 5691 * 5692 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 5693 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 5694 * header has not been written. 5695 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 5696 * been written so far. 5697 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 5698 * the 3WHS. 5699 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 5700 * the 3WHS. 5701 * 5702 * bpf_load_hdr_opt() can also be used to read a particular option. 5703 */ 5704 __bpf_md_ptr(void *, skb_data); 5705 __bpf_md_ptr(void *, skb_data_end); 5706 __u32 skb_len; /* The total length of a packet. 5707 * It includes the header, options, 5708 * and payload. 5709 */ 5710 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 5711 * an easy way to check for tcp_flags 5712 * without parsing skb_data. 5713 * 5714 * In particular, the skb_tcp_flags 5715 * will still be available in 5716 * BPF_SOCK_OPS_HDR_OPT_LEN even though 5717 * the outgoing header has not 5718 * been written yet. 5719 */ 5720 }; 5721 5722 /* Definitions for bpf_sock_ops_cb_flags */ 5723 enum { 5724 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 5725 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 5726 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 5727 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 5728 /* Call bpf for all received TCP headers. The bpf prog will be 5729 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5730 * 5731 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5732 * for the header option related helpers that will be useful 5733 * to the bpf programs. 5734 * 5735 * It could be used at the client/active side (i.e. connect() side) 5736 * when the server told it that the server was in syncookie 5737 * mode and required the active side to resend the bpf-written 5738 * options. The active side can keep writing the bpf-options until 5739 * it received a valid packet from the server side to confirm 5740 * the earlier packet (and options) has been received. The later 5741 * example patch is using it like this at the active side when the 5742 * server is in syncookie mode. 5743 * 5744 * The bpf prog will usually turn this off in the common cases. 5745 */ 5746 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 5747 /* Call bpf when kernel has received a header option that 5748 * the kernel cannot handle. The bpf prog will be called under 5749 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 5750 * 5751 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5752 * for the header option related helpers that will be useful 5753 * to the bpf programs. 5754 */ 5755 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 5756 /* Call bpf when the kernel is writing header options for the 5757 * outgoing packet. The bpf prog will first be called 5758 * to reserve space in a skb under 5759 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 5760 * the bpf prog will be called to write the header option(s) 5761 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5762 * 5763 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 5764 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 5765 * related helpers that will be useful to the bpf programs. 5766 * 5767 * The kernel gets its chance to reserve space and write 5768 * options first before the BPF program does. 5769 */ 5770 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 5771 /* Mask of all currently supported cb flags */ 5772 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 5773 }; 5774 5775 /* List of known BPF sock_ops operators. 5776 * New entries can only be added at the end 5777 */ 5778 enum { 5779 BPF_SOCK_OPS_VOID, 5780 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 5781 * -1 if default value should be used 5782 */ 5783 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 5784 * window (in packets) or -1 if default 5785 * value should be used 5786 */ 5787 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 5788 * active connection is initialized 5789 */ 5790 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 5791 * active connection is 5792 * established 5793 */ 5794 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 5795 * passive connection is 5796 * established 5797 */ 5798 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 5799 * needs ECN 5800 */ 5801 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5802 * based on the path and may be 5803 * dependent on the congestion control 5804 * algorithm. In general it indicates 5805 * a congestion threshold. RTTs above 5806 * this indicate congestion 5807 */ 5808 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5809 * Arg1: value of icsk_retransmits 5810 * Arg2: value of icsk_rto 5811 * Arg3: whether RTO has expired 5812 */ 5813 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 5814 * Arg1: sequence number of 1st byte 5815 * Arg2: # segments 5816 * Arg3: return value of 5817 * tcp_transmit_skb (0 => success) 5818 */ 5819 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 5820 * Arg1: old_state 5821 * Arg2: new_state 5822 */ 5823 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 5824 * socket transition to LISTEN state. 5825 */ 5826 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 5827 */ 5828 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 5829 * It will be called to handle 5830 * the packets received at 5831 * an already established 5832 * connection. 5833 * 5834 * sock_ops->skb_data: 5835 * Referring to the received skb. 5836 * It covers the TCP header only. 5837 * 5838 * bpf_load_hdr_opt() can also 5839 * be used to search for a 5840 * particular option. 5841 */ 5842 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 5843 * header option later in 5844 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5845 * Arg1: bool want_cookie. (in 5846 * writing SYNACK only) 5847 * 5848 * sock_ops->skb_data: 5849 * Not available because no header has 5850 * been written yet. 5851 * 5852 * sock_ops->skb_tcp_flags: 5853 * The tcp_flags of the 5854 * outgoing skb. (e.g. SYN, ACK, FIN). 5855 * 5856 * bpf_reserve_hdr_opt() should 5857 * be used to reserve space. 5858 */ 5859 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 5860 * Arg1: bool want_cookie. (in 5861 * writing SYNACK only) 5862 * 5863 * sock_ops->skb_data: 5864 * Referring to the outgoing skb. 5865 * It covers the TCP header 5866 * that has already been written 5867 * by the kernel and the 5868 * earlier bpf-progs. 5869 * 5870 * sock_ops->skb_tcp_flags: 5871 * The tcp_flags of the outgoing 5872 * skb. (e.g. SYN, ACK, FIN). 5873 * 5874 * bpf_store_hdr_opt() should 5875 * be used to write the 5876 * option. 5877 * 5878 * bpf_load_hdr_opt() can also 5879 * be used to search for a 5880 * particular option that 5881 * has already been written 5882 * by the kernel or the 5883 * earlier bpf-progs. 5884 */ 5885 }; 5886 5887 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 5888 * changes between the TCP and BPF versions. Ideally this should never happen. 5889 * If it does, we need to add code to convert them before calling 5890 * the BPF sock_ops function. 5891 */ 5892 enum { 5893 BPF_TCP_ESTABLISHED = 1, 5894 BPF_TCP_SYN_SENT, 5895 BPF_TCP_SYN_RECV, 5896 BPF_TCP_FIN_WAIT1, 5897 BPF_TCP_FIN_WAIT2, 5898 BPF_TCP_TIME_WAIT, 5899 BPF_TCP_CLOSE, 5900 BPF_TCP_CLOSE_WAIT, 5901 BPF_TCP_LAST_ACK, 5902 BPF_TCP_LISTEN, 5903 BPF_TCP_CLOSING, /* Now a valid state */ 5904 BPF_TCP_NEW_SYN_RECV, 5905 5906 BPF_TCP_MAX_STATES /* Leave at the end! */ 5907 }; 5908 5909 enum { 5910 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 5911 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 5912 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 5913 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 5914 /* Copy the SYN pkt to optval 5915 * 5916 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 5917 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 5918 * to only getting from the saved_syn. It can either get the 5919 * syn packet from: 5920 * 5921 * 1. the just-received SYN packet (only available when writing the 5922 * SYNACK). It will be useful when it is not necessary to 5923 * save the SYN packet for latter use. It is also the only way 5924 * to get the SYN during syncookie mode because the syn 5925 * packet cannot be saved during syncookie. 5926 * 5927 * OR 5928 * 5929 * 2. the earlier saved syn which was done by 5930 * bpf_setsockopt(TCP_SAVE_SYN). 5931 * 5932 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 5933 * SYN packet is obtained. 5934 * 5935 * If the bpf-prog does not need the IP[46] header, the 5936 * bpf-prog can avoid parsing the IP header by using 5937 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 5938 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 5939 * 5940 * >0: Total number of bytes copied 5941 * -ENOSPC: Not enough space in optval. Only optlen number of 5942 * bytes is copied. 5943 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 5944 * is not saved by setsockopt(TCP_SAVE_SYN). 5945 */ 5946 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 5947 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 5948 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 5949 }; 5950 5951 enum { 5952 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 5953 }; 5954 5955 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 5956 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5957 */ 5958 enum { 5959 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 5960 * total option spaces 5961 * required for an established 5962 * sk in order to calculate the 5963 * MSS. No skb is actually 5964 * sent. 5965 */ 5966 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 5967 * when sending a SYN. 5968 */ 5969 }; 5970 5971 struct bpf_perf_event_value { 5972 __u64 counter; 5973 __u64 enabled; 5974 __u64 running; 5975 }; 5976 5977 enum { 5978 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 5979 BPF_DEVCG_ACC_READ = (1ULL << 1), 5980 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 5981 }; 5982 5983 enum { 5984 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 5985 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 5986 }; 5987 5988 struct bpf_cgroup_dev_ctx { 5989 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 5990 __u32 access_type; 5991 __u32 major; 5992 __u32 minor; 5993 }; 5994 5995 struct bpf_raw_tracepoint_args { 5996 __u64 args[0]; 5997 }; 5998 5999 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 6000 * OUTPUT: Do lookup from egress perspective; default is ingress 6001 */ 6002 enum { 6003 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6004 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6005 }; 6006 6007 enum { 6008 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6009 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6010 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6011 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6012 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6013 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6014 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6015 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6016 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6017 }; 6018 6019 struct bpf_fib_lookup { 6020 /* input: network family for lookup (AF_INET, AF_INET6) 6021 * output: network family of egress nexthop 6022 */ 6023 __u8 family; 6024 6025 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6026 __u8 l4_protocol; 6027 __be16 sport; 6028 __be16 dport; 6029 6030 union { /* used for MTU check */ 6031 /* input to lookup */ 6032 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6033 6034 /* output: MTU value */ 6035 __u16 mtu_result; 6036 }; 6037 /* input: L3 device index for lookup 6038 * output: device index from FIB lookup 6039 */ 6040 __u32 ifindex; 6041 6042 union { 6043 /* inputs to lookup */ 6044 __u8 tos; /* AF_INET */ 6045 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6046 6047 /* output: metric of fib result (IPv4/IPv6 only) */ 6048 __u32 rt_metric; 6049 }; 6050 6051 union { 6052 __be32 ipv4_src; 6053 __u32 ipv6_src[4]; /* in6_addr; network order */ 6054 }; 6055 6056 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6057 * network header. output: bpf_fib_lookup sets to gateway address 6058 * if FIB lookup returns gateway route 6059 */ 6060 union { 6061 __be32 ipv4_dst; 6062 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6063 }; 6064 6065 /* output */ 6066 __be16 h_vlan_proto; 6067 __be16 h_vlan_TCI; 6068 __u8 smac[6]; /* ETH_ALEN */ 6069 __u8 dmac[6]; /* ETH_ALEN */ 6070 }; 6071 6072 struct bpf_redir_neigh { 6073 /* network family for lookup (AF_INET, AF_INET6) */ 6074 __u32 nh_family; 6075 /* network address of nexthop; skips fib lookup to find gateway */ 6076 union { 6077 __be32 ipv4_nh; 6078 __u32 ipv6_nh[4]; /* in6_addr; network order */ 6079 }; 6080 }; 6081 6082 /* bpf_check_mtu flags*/ 6083 enum bpf_check_mtu_flags { 6084 BPF_MTU_CHK_SEGS = (1U << 0), 6085 }; 6086 6087 enum bpf_check_mtu_ret { 6088 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 6089 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6090 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 6091 }; 6092 6093 enum bpf_task_fd_type { 6094 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 6095 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 6096 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 6097 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 6098 BPF_FD_TYPE_UPROBE, /* filename + offset */ 6099 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 6100 }; 6101 6102 enum { 6103 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 6104 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 6105 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 6106 }; 6107 6108 struct bpf_flow_keys { 6109 __u16 nhoff; 6110 __u16 thoff; 6111 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6112 __u8 is_frag; 6113 __u8 is_first_frag; 6114 __u8 is_encap; 6115 __u8 ip_proto; 6116 __be16 n_proto; 6117 __be16 sport; 6118 __be16 dport; 6119 union { 6120 struct { 6121 __be32 ipv4_src; 6122 __be32 ipv4_dst; 6123 }; 6124 struct { 6125 __u32 ipv6_src[4]; /* in6_addr; network order */ 6126 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6127 }; 6128 }; 6129 __u32 flags; 6130 __be32 flow_label; 6131 }; 6132 6133 struct bpf_func_info { 6134 __u32 insn_off; 6135 __u32 type_id; 6136 }; 6137 6138 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6139 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6140 6141 struct bpf_line_info { 6142 __u32 insn_off; 6143 __u32 file_name_off; 6144 __u32 line_off; 6145 __u32 line_col; 6146 }; 6147 6148 struct bpf_spin_lock { 6149 __u32 val; 6150 }; 6151 6152 struct bpf_timer { 6153 __u64 :64; 6154 __u64 :64; 6155 } __attribute__((aligned(8))); 6156 6157 struct bpf_sysctl { 6158 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 6159 * Allows 1,2,4-byte read, but no write. 6160 */ 6161 __u32 file_pos; /* Sysctl file position to read from, write to. 6162 * Allows 1,2,4-byte read an 4-byte write. 6163 */ 6164 }; 6165 6166 struct bpf_sockopt { 6167 __bpf_md_ptr(struct bpf_sock *, sk); 6168 __bpf_md_ptr(void *, optval); 6169 __bpf_md_ptr(void *, optval_end); 6170 6171 __s32 level; 6172 __s32 optname; 6173 __s32 optlen; 6174 __s32 retval; 6175 }; 6176 6177 struct bpf_pidns_info { 6178 __u32 pid; 6179 __u32 tgid; 6180 }; 6181 6182 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6183 struct bpf_sk_lookup { 6184 union { 6185 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 6186 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 6187 }; 6188 6189 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6190 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6191 __u32 remote_ip4; /* Network byte order */ 6192 __u32 remote_ip6[4]; /* Network byte order */ 6193 __u32 remote_port; /* Network byte order */ 6194 __u32 local_ip4; /* Network byte order */ 6195 __u32 local_ip6[4]; /* Network byte order */ 6196 __u32 local_port; /* Host byte order */ 6197 }; 6198 6199 /* 6200 * struct btf_ptr is used for typed pointer representation; the 6201 * type id is used to render the pointer data as the appropriate type 6202 * via the bpf_snprintf_btf() helper described above. A flags field - 6203 * potentially to specify additional details about the BTF pointer 6204 * (rather than its mode of display) - is included for future use. 6205 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6206 */ 6207 struct btf_ptr { 6208 void *ptr; 6209 __u32 type_id; 6210 __u32 flags; /* BTF ptr flags; unused at present. */ 6211 }; 6212 6213 /* 6214 * Flags to control bpf_snprintf_btf() behaviour. 6215 * - BTF_F_COMPACT: no formatting around type information 6216 * - BTF_F_NONAME: no struct/union member names/types 6217 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6218 * equivalent to %px. 6219 * - BTF_F_ZERO: show zero-valued struct/union members; they 6220 * are not displayed by default 6221 */ 6222 enum { 6223 BTF_F_COMPACT = (1ULL << 0), 6224 BTF_F_NONAME = (1ULL << 1), 6225 BTF_F_PTR_RAW = (1ULL << 2), 6226 BTF_F_ZERO = (1ULL << 3), 6227 }; 6228 6229 #endif /* _UAPI__LINUX_BPF_H__ */ 6230