1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25 /* alu/jmp fields */ 26 #define BPF_MOV 0xb0 /* mov reg to reg */ 27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29 /* change endianness of a register */ 30 #define BPF_END 0xd0 /* flags for endianness conversion: */ 31 #define BPF_TO_LE 0x00 /* convert to little-endian */ 32 #define BPF_TO_BE 0x08 /* convert to big-endian */ 33 #define BPF_FROM_LE BPF_TO_LE 34 #define BPF_FROM_BE BPF_TO_BE 35 36 /* jmp encodings */ 37 #define BPF_JNE 0x50 /* jump != */ 38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44 #define BPF_CALL 0x80 /* function call */ 45 #define BPF_EXIT 0x90 /* function return */ 46 47 /* atomic op type fields (stored in immediate) */ 48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52 /* Register numbers */ 53 enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66 }; 67 68 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 69 #define MAX_BPF_REG __MAX_BPF_REG 70 71 struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77 }; 78 79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80 struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83 }; 84 85 struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88 }; 89 90 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 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 334 * 335 * Return 336 * Returns zero on success. On error, -1 is returned and *errno* 337 * is set appropriately. 338 * 339 * **ENOSPC** 340 * Either *data_size_out* or *ctx_size_out* is too small. 341 * **ENOTSUPP** 342 * This command is not supported by the program type of 343 * the program referred to by *prog_fd*. 344 * 345 * BPF_PROG_GET_NEXT_ID 346 * Description 347 * Fetch the next eBPF program currently loaded into the kernel. 348 * 349 * Looks for the eBPF program with an id greater than *start_id* 350 * and updates *next_id* on success. If no other eBPF programs 351 * remain with ids higher than *start_id*, returns -1 and sets 352 * *errno* to **ENOENT**. 353 * 354 * Return 355 * Returns zero on success. On error, or when no id remains, -1 356 * is returned and *errno* is set appropriately. 357 * 358 * BPF_MAP_GET_NEXT_ID 359 * Description 360 * Fetch the next eBPF map currently loaded into the kernel. 361 * 362 * Looks for the eBPF map with an id greater than *start_id* 363 * and updates *next_id* on success. If no other eBPF maps 364 * remain with ids higher than *start_id*, returns -1 and sets 365 * *errno* to **ENOENT**. 366 * 367 * Return 368 * Returns zero on success. On error, or when no id remains, -1 369 * is returned and *errno* is set appropriately. 370 * 371 * BPF_PROG_GET_FD_BY_ID 372 * Description 373 * Open a file descriptor for the eBPF program corresponding to 374 * *prog_id*. 375 * 376 * Return 377 * A new file descriptor (a nonnegative integer), or -1 if an 378 * error occurred (in which case, *errno* is set appropriately). 379 * 380 * BPF_MAP_GET_FD_BY_ID 381 * Description 382 * Open a file descriptor for the eBPF map corresponding to 383 * *map_id*. 384 * 385 * Return 386 * A new file descriptor (a nonnegative integer), or -1 if an 387 * error occurred (in which case, *errno* is set appropriately). 388 * 389 * BPF_OBJ_GET_INFO_BY_FD 390 * Description 391 * Obtain information about the eBPF object corresponding to 392 * *bpf_fd*. 393 * 394 * Populates up to *info_len* bytes of *info*, which will be in 395 * one of the following formats depending on the eBPF object type 396 * of *bpf_fd*: 397 * 398 * * **struct bpf_prog_info** 399 * * **struct bpf_map_info** 400 * * **struct bpf_btf_info** 401 * * **struct bpf_link_info** 402 * 403 * Return 404 * Returns zero on success. On error, -1 is returned and *errno* 405 * is set appropriately. 406 * 407 * BPF_PROG_QUERY 408 * Description 409 * Obtain information about eBPF programs associated with the 410 * specified *attach_type* hook. 411 * 412 * The *target_fd* must be a valid file descriptor for a kernel 413 * object which depends on the attach type of *attach_bpf_fd*: 414 * 415 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 416 * **BPF_PROG_TYPE_CGROUP_SKB**, 417 * **BPF_PROG_TYPE_CGROUP_SOCK**, 418 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 419 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 420 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 421 * **BPF_PROG_TYPE_SOCK_OPS** 422 * 423 * Control Group v2 hierarchy with the eBPF controller 424 * enabled. Requires the kernel to be compiled with 425 * **CONFIG_CGROUP_BPF**. 426 * 427 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 428 * 429 * Network namespace (eg /proc/self/ns/net). 430 * 431 * **BPF_PROG_TYPE_LIRC_MODE2** 432 * 433 * LIRC device path (eg /dev/lircN). Requires the kernel 434 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 435 * 436 * **BPF_PROG_QUERY** always fetches the number of programs 437 * attached and the *attach_flags* which were used to attach those 438 * programs. Additionally, if *prog_ids* is nonzero and the number 439 * of attached programs is less than *prog_cnt*, populates 440 * *prog_ids* with the eBPF program ids of the programs attached 441 * at *target_fd*. 442 * 443 * The following flags may alter the result: 444 * 445 * **BPF_F_QUERY_EFFECTIVE** 446 * Only return information regarding programs which are 447 * currently effective at the specified *target_fd*. 448 * 449 * Return 450 * Returns zero on success. On error, -1 is returned and *errno* 451 * is set appropriately. 452 * 453 * BPF_RAW_TRACEPOINT_OPEN 454 * Description 455 * Attach an eBPF program to a tracepoint *name* to access kernel 456 * internal arguments of the tracepoint in their raw form. 457 * 458 * The *prog_fd* must be a valid file descriptor associated with 459 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 460 * 461 * No ABI guarantees are made about the content of tracepoint 462 * arguments exposed to the corresponding eBPF program. 463 * 464 * Applying **close**\ (2) to the file descriptor returned by 465 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 466 * 467 * Return 468 * A new file descriptor (a nonnegative integer), or -1 if an 469 * error occurred (in which case, *errno* is set appropriately). 470 * 471 * BPF_BTF_LOAD 472 * Description 473 * Verify and load BPF Type Format (BTF) metadata into the kernel, 474 * returning a new file descriptor associated with the metadata. 475 * BTF is described in more detail at 476 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 477 * 478 * The *btf* parameter must point to valid memory providing 479 * *btf_size* bytes of BTF binary metadata. 480 * 481 * The returned file descriptor can be passed to other **bpf**\ () 482 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 483 * associate the BTF with those objects. 484 * 485 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 486 * parameters to specify a *btf_log_buf*, *btf_log_size* and 487 * *btf_log_level* which allow the kernel to return freeform log 488 * output regarding the BTF verification process. 489 * 490 * Return 491 * A new file descriptor (a nonnegative integer), or -1 if an 492 * error occurred (in which case, *errno* is set appropriately). 493 * 494 * BPF_BTF_GET_FD_BY_ID 495 * Description 496 * Open a file descriptor for the BPF Type Format (BTF) 497 * corresponding to *btf_id*. 498 * 499 * Return 500 * A new file descriptor (a nonnegative integer), or -1 if an 501 * error occurred (in which case, *errno* is set appropriately). 502 * 503 * BPF_TASK_FD_QUERY 504 * Description 505 * Obtain information about eBPF programs associated with the 506 * target process identified by *pid* and *fd*. 507 * 508 * If the *pid* and *fd* are associated with a tracepoint, kprobe 509 * or uprobe perf event, then the *prog_id* and *fd_type* will 510 * be populated with the eBPF program id and file descriptor type 511 * of type **bpf_task_fd_type**. If associated with a kprobe or 512 * uprobe, the *probe_offset* and *probe_addr* will also be 513 * populated. Optionally, if *buf* is provided, then up to 514 * *buf_len* bytes of *buf* will be populated with the name of 515 * the tracepoint, kprobe or uprobe. 516 * 517 * The resulting *prog_id* may be introspected in deeper detail 518 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 519 * 520 * Return 521 * Returns zero on success. On error, -1 is returned and *errno* 522 * is set appropriately. 523 * 524 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 525 * Description 526 * Look up an element with the given *key* in the map referred to 527 * by the file descriptor *fd*, and if found, delete the element. 528 * 529 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 530 * types, the *flags* argument needs to be set to 0, but for other 531 * map types, it may be specified as: 532 * 533 * **BPF_F_LOCK** 534 * Look up and delete the value of a spin-locked map 535 * without returning the lock. This must be specified if 536 * the elements contain a spinlock. 537 * 538 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 539 * implement this command as a "pop" operation, deleting the top 540 * element rather than one corresponding to *key*. 541 * The *key* and *key_len* parameters should be zeroed when 542 * issuing this operation for these map types. 543 * 544 * This command is only valid for the following map types: 545 * * **BPF_MAP_TYPE_QUEUE** 546 * * **BPF_MAP_TYPE_STACK** 547 * * **BPF_MAP_TYPE_HASH** 548 * * **BPF_MAP_TYPE_PERCPU_HASH** 549 * * **BPF_MAP_TYPE_LRU_HASH** 550 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 551 * 552 * Return 553 * Returns zero on success. On error, -1 is returned and *errno* 554 * is set appropriately. 555 * 556 * BPF_MAP_FREEZE 557 * Description 558 * Freeze the permissions of the specified map. 559 * 560 * Write permissions may be frozen by passing zero *flags*. 561 * Upon success, no future syscall invocations may alter the 562 * map state of *map_fd*. Write operations from eBPF programs 563 * are still possible for a frozen map. 564 * 565 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 566 * 567 * Return 568 * Returns zero on success. On error, -1 is returned and *errno* 569 * is set appropriately. 570 * 571 * BPF_BTF_GET_NEXT_ID 572 * Description 573 * Fetch the next BPF Type Format (BTF) object currently loaded 574 * into the kernel. 575 * 576 * Looks for the BTF object with an id greater than *start_id* 577 * and updates *next_id* on success. If no other BTF objects 578 * remain with ids higher than *start_id*, returns -1 and sets 579 * *errno* to **ENOENT**. 580 * 581 * Return 582 * Returns zero on success. On error, or when no id remains, -1 583 * is returned and *errno* is set appropriately. 584 * 585 * BPF_MAP_LOOKUP_BATCH 586 * Description 587 * Iterate and fetch multiple elements in a map. 588 * 589 * Two opaque values are used to manage batch operations, 590 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 591 * to NULL to begin the batched operation. After each subsequent 592 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 593 * *out_batch* as the *in_batch* for the next operation to 594 * continue iteration from the current point. 595 * 596 * The *keys* and *values* are output parameters which must point 597 * to memory large enough to hold *count* items based on the key 598 * and value size of the map *map_fd*. The *keys* buffer must be 599 * of *key_size* * *count*. The *values* buffer must be of 600 * *value_size* * *count*. 601 * 602 * The *elem_flags* argument may be specified as one of the 603 * following: 604 * 605 * **BPF_F_LOCK** 606 * Look up the value of a spin-locked map without 607 * returning the lock. This must be specified if the 608 * elements contain a spinlock. 609 * 610 * On success, *count* elements from the map are copied into the 611 * user buffer, with the keys copied into *keys* and the values 612 * copied into the corresponding indices in *values*. 613 * 614 * If an error is returned and *errno* is not **EFAULT**, *count* 615 * is set to the number of successfully processed elements. 616 * 617 * Return 618 * Returns zero on success. On error, -1 is returned and *errno* 619 * is set appropriately. 620 * 621 * May set *errno* to **ENOSPC** to indicate that *keys* or 622 * *values* is too small to dump an entire bucket during 623 * iteration of a hash-based map type. 624 * 625 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 626 * Description 627 * Iterate and delete all elements in a map. 628 * 629 * This operation has the same behavior as 630 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 631 * 632 * * Every element that is successfully returned is also deleted 633 * from the map. This is at least *count* elements. Note that 634 * *count* is both an input and an output parameter. 635 * * Upon returning with *errno* set to **EFAULT**, up to 636 * *count* elements may be deleted without returning the keys 637 * and values of the deleted elements. 638 * 639 * Return 640 * Returns zero on success. On error, -1 is returned and *errno* 641 * is set appropriately. 642 * 643 * BPF_MAP_UPDATE_BATCH 644 * Description 645 * Update multiple elements in a map by *key*. 646 * 647 * The *keys* and *values* are input parameters which must point 648 * to memory large enough to hold *count* items based on the key 649 * and value size of the map *map_fd*. The *keys* buffer must be 650 * of *key_size* * *count*. The *values* buffer must be of 651 * *value_size* * *count*. 652 * 653 * Each element specified in *keys* is sequentially updated to the 654 * value in the corresponding index in *values*. The *in_batch* 655 * and *out_batch* parameters are ignored and should be zeroed. 656 * 657 * The *elem_flags* argument should be specified as one of the 658 * following: 659 * 660 * **BPF_ANY** 661 * Create new elements or update a existing elements. 662 * **BPF_NOEXIST** 663 * Create new elements only if they do not exist. 664 * **BPF_EXIST** 665 * Update existing elements. 666 * **BPF_F_LOCK** 667 * Update spin_lock-ed map elements. This must be 668 * specified if the map value contains a spinlock. 669 * 670 * On success, *count* elements from the map are updated. 671 * 672 * If an error is returned and *errno* is not **EFAULT**, *count* 673 * is set to the number of successfully processed elements. 674 * 675 * Return 676 * Returns zero on success. On error, -1 is returned and *errno* 677 * is set appropriately. 678 * 679 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 680 * **E2BIG**. **E2BIG** indicates that the number of elements in 681 * the map reached the *max_entries* limit specified at map 682 * creation time. 683 * 684 * May set *errno* to one of the following error codes under 685 * specific circumstances: 686 * 687 * **EEXIST** 688 * If *flags* specifies **BPF_NOEXIST** and the element 689 * with *key* already exists in the map. 690 * **ENOENT** 691 * If *flags* specifies **BPF_EXIST** and the element with 692 * *key* does not exist in the map. 693 * 694 * BPF_MAP_DELETE_BATCH 695 * Description 696 * Delete multiple elements in a map by *key*. 697 * 698 * The *keys* parameter is an input parameter which must point 699 * to memory large enough to hold *count* items based on the key 700 * size of the map *map_fd*, that is, *key_size* * *count*. 701 * 702 * Each element specified in *keys* is sequentially deleted. The 703 * *in_batch*, *out_batch*, and *values* parameters are ignored 704 * and should be zeroed. 705 * 706 * The *elem_flags* argument may be specified as one of the 707 * following: 708 * 709 * **BPF_F_LOCK** 710 * Look up the value of a spin-locked map without 711 * returning the lock. This must be specified if the 712 * elements contain a spinlock. 713 * 714 * On success, *count* elements from the map are updated. 715 * 716 * If an error is returned and *errno* is not **EFAULT**, *count* 717 * is set to the number of successfully processed elements. If 718 * *errno* is **EFAULT**, up to *count* elements may be been 719 * deleted. 720 * 721 * Return 722 * Returns zero on success. On error, -1 is returned and *errno* 723 * is set appropriately. 724 * 725 * BPF_LINK_CREATE 726 * Description 727 * Attach an eBPF program to a *target_fd* at the specified 728 * *attach_type* hook and return a file descriptor handle for 729 * managing the link. 730 * 731 * Return 732 * A new file descriptor (a nonnegative integer), or -1 if an 733 * error occurred (in which case, *errno* is set appropriately). 734 * 735 * BPF_LINK_UPDATE 736 * Description 737 * Update the eBPF program in the specified *link_fd* to 738 * *new_prog_fd*. 739 * 740 * Return 741 * Returns zero on success. On error, -1 is returned and *errno* 742 * is set appropriately. 743 * 744 * BPF_LINK_GET_FD_BY_ID 745 * Description 746 * Open a file descriptor for the eBPF Link corresponding to 747 * *link_id*. 748 * 749 * Return 750 * A new file descriptor (a nonnegative integer), or -1 if an 751 * error occurred (in which case, *errno* is set appropriately). 752 * 753 * BPF_LINK_GET_NEXT_ID 754 * Description 755 * Fetch the next eBPF link currently loaded into the kernel. 756 * 757 * Looks for the eBPF link with an id greater than *start_id* 758 * and updates *next_id* on success. If no other eBPF links 759 * remain with ids higher than *start_id*, returns -1 and sets 760 * *errno* to **ENOENT**. 761 * 762 * Return 763 * Returns zero on success. On error, or when no id remains, -1 764 * is returned and *errno* is set appropriately. 765 * 766 * BPF_ENABLE_STATS 767 * Description 768 * Enable eBPF runtime statistics gathering. 769 * 770 * Runtime statistics gathering for the eBPF runtime is disabled 771 * by default to minimize the corresponding performance overhead. 772 * This command enables statistics globally. 773 * 774 * Multiple programs may independently enable statistics. 775 * After gathering the desired statistics, eBPF runtime statistics 776 * may be disabled again by calling **close**\ (2) for the file 777 * descriptor returned by this function. Statistics will only be 778 * disabled system-wide when all outstanding file descriptors 779 * returned by prior calls for this subcommand are closed. 780 * 781 * Return 782 * A new file descriptor (a nonnegative integer), or -1 if an 783 * error occurred (in which case, *errno* is set appropriately). 784 * 785 * BPF_ITER_CREATE 786 * Description 787 * Create an iterator on top of the specified *link_fd* (as 788 * previously created using **BPF_LINK_CREATE**) and return a 789 * file descriptor that can be used to trigger the iteration. 790 * 791 * If the resulting file descriptor is pinned to the filesystem 792 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 793 * for that path will trigger the iterator to read kernel state 794 * using the eBPF program attached to *link_fd*. 795 * 796 * Return 797 * A new file descriptor (a nonnegative integer), or -1 if an 798 * error occurred (in which case, *errno* is set appropriately). 799 * 800 * BPF_LINK_DETACH 801 * Description 802 * Forcefully detach the specified *link_fd* from its 803 * corresponding attachment point. 804 * 805 * Return 806 * Returns zero on success. On error, -1 is returned and *errno* 807 * is set appropriately. 808 * 809 * BPF_PROG_BIND_MAP 810 * Description 811 * Bind a map to the lifetime of an eBPF program. 812 * 813 * The map identified by *map_fd* is bound to the program 814 * identified by *prog_fd* and only released when *prog_fd* is 815 * released. This may be used in cases where metadata should be 816 * associated with a program which otherwise does not contain any 817 * references to the map (for example, embedded in the eBPF 818 * program instructions). 819 * 820 * Return 821 * Returns zero on success. On error, -1 is returned and *errno* 822 * is set appropriately. 823 * 824 * NOTES 825 * eBPF objects (maps and programs) can be shared between processes. 826 * 827 * * After **fork**\ (2), the child inherits file descriptors 828 * referring to the same eBPF objects. 829 * * File descriptors referring to eBPF objects can be transferred over 830 * **unix**\ (7) domain sockets. 831 * * File descriptors referring to eBPF objects can be duplicated in the 832 * usual way, using **dup**\ (2) and similar calls. 833 * * File descriptors referring to eBPF objects can be pinned to the 834 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 835 * 836 * An eBPF object is deallocated only after all file descriptors referring 837 * to the object have been closed and no references remain pinned to the 838 * filesystem or attached (for example, bound to a program or device). 839 */ 840 enum bpf_cmd { 841 BPF_MAP_CREATE, 842 BPF_MAP_LOOKUP_ELEM, 843 BPF_MAP_UPDATE_ELEM, 844 BPF_MAP_DELETE_ELEM, 845 BPF_MAP_GET_NEXT_KEY, 846 BPF_PROG_LOAD, 847 BPF_OBJ_PIN, 848 BPF_OBJ_GET, 849 BPF_PROG_ATTACH, 850 BPF_PROG_DETACH, 851 BPF_PROG_TEST_RUN, 852 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 853 BPF_PROG_GET_NEXT_ID, 854 BPF_MAP_GET_NEXT_ID, 855 BPF_PROG_GET_FD_BY_ID, 856 BPF_MAP_GET_FD_BY_ID, 857 BPF_OBJ_GET_INFO_BY_FD, 858 BPF_PROG_QUERY, 859 BPF_RAW_TRACEPOINT_OPEN, 860 BPF_BTF_LOAD, 861 BPF_BTF_GET_FD_BY_ID, 862 BPF_TASK_FD_QUERY, 863 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 864 BPF_MAP_FREEZE, 865 BPF_BTF_GET_NEXT_ID, 866 BPF_MAP_LOOKUP_BATCH, 867 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 868 BPF_MAP_UPDATE_BATCH, 869 BPF_MAP_DELETE_BATCH, 870 BPF_LINK_CREATE, 871 BPF_LINK_UPDATE, 872 BPF_LINK_GET_FD_BY_ID, 873 BPF_LINK_GET_NEXT_ID, 874 BPF_ENABLE_STATS, 875 BPF_ITER_CREATE, 876 BPF_LINK_DETACH, 877 BPF_PROG_BIND_MAP, 878 }; 879 880 enum bpf_map_type { 881 BPF_MAP_TYPE_UNSPEC, 882 BPF_MAP_TYPE_HASH, 883 BPF_MAP_TYPE_ARRAY, 884 BPF_MAP_TYPE_PROG_ARRAY, 885 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 886 BPF_MAP_TYPE_PERCPU_HASH, 887 BPF_MAP_TYPE_PERCPU_ARRAY, 888 BPF_MAP_TYPE_STACK_TRACE, 889 BPF_MAP_TYPE_CGROUP_ARRAY, 890 BPF_MAP_TYPE_LRU_HASH, 891 BPF_MAP_TYPE_LRU_PERCPU_HASH, 892 BPF_MAP_TYPE_LPM_TRIE, 893 BPF_MAP_TYPE_ARRAY_OF_MAPS, 894 BPF_MAP_TYPE_HASH_OF_MAPS, 895 BPF_MAP_TYPE_DEVMAP, 896 BPF_MAP_TYPE_SOCKMAP, 897 BPF_MAP_TYPE_CPUMAP, 898 BPF_MAP_TYPE_XSKMAP, 899 BPF_MAP_TYPE_SOCKHASH, 900 BPF_MAP_TYPE_CGROUP_STORAGE, 901 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 902 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 903 BPF_MAP_TYPE_QUEUE, 904 BPF_MAP_TYPE_STACK, 905 BPF_MAP_TYPE_SK_STORAGE, 906 BPF_MAP_TYPE_DEVMAP_HASH, 907 BPF_MAP_TYPE_STRUCT_OPS, 908 BPF_MAP_TYPE_RINGBUF, 909 BPF_MAP_TYPE_INODE_STORAGE, 910 BPF_MAP_TYPE_TASK_STORAGE, 911 BPF_MAP_TYPE_BLOOM_FILTER, 912 }; 913 914 /* Note that tracing related programs such as 915 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 916 * are not subject to a stable API since kernel internal data 917 * structures can change from release to release and may 918 * therefore break existing tracing BPF programs. Tracing BPF 919 * programs correspond to /a/ specific kernel which is to be 920 * analyzed, and not /a/ specific kernel /and/ all future ones. 921 */ 922 enum bpf_prog_type { 923 BPF_PROG_TYPE_UNSPEC, 924 BPF_PROG_TYPE_SOCKET_FILTER, 925 BPF_PROG_TYPE_KPROBE, 926 BPF_PROG_TYPE_SCHED_CLS, 927 BPF_PROG_TYPE_SCHED_ACT, 928 BPF_PROG_TYPE_TRACEPOINT, 929 BPF_PROG_TYPE_XDP, 930 BPF_PROG_TYPE_PERF_EVENT, 931 BPF_PROG_TYPE_CGROUP_SKB, 932 BPF_PROG_TYPE_CGROUP_SOCK, 933 BPF_PROG_TYPE_LWT_IN, 934 BPF_PROG_TYPE_LWT_OUT, 935 BPF_PROG_TYPE_LWT_XMIT, 936 BPF_PROG_TYPE_SOCK_OPS, 937 BPF_PROG_TYPE_SK_SKB, 938 BPF_PROG_TYPE_CGROUP_DEVICE, 939 BPF_PROG_TYPE_SK_MSG, 940 BPF_PROG_TYPE_RAW_TRACEPOINT, 941 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 942 BPF_PROG_TYPE_LWT_SEG6LOCAL, 943 BPF_PROG_TYPE_LIRC_MODE2, 944 BPF_PROG_TYPE_SK_REUSEPORT, 945 BPF_PROG_TYPE_FLOW_DISSECTOR, 946 BPF_PROG_TYPE_CGROUP_SYSCTL, 947 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 948 BPF_PROG_TYPE_CGROUP_SOCKOPT, 949 BPF_PROG_TYPE_TRACING, 950 BPF_PROG_TYPE_STRUCT_OPS, 951 BPF_PROG_TYPE_EXT, 952 BPF_PROG_TYPE_LSM, 953 BPF_PROG_TYPE_SK_LOOKUP, 954 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 955 }; 956 957 enum bpf_attach_type { 958 BPF_CGROUP_INET_INGRESS, 959 BPF_CGROUP_INET_EGRESS, 960 BPF_CGROUP_INET_SOCK_CREATE, 961 BPF_CGROUP_SOCK_OPS, 962 BPF_SK_SKB_STREAM_PARSER, 963 BPF_SK_SKB_STREAM_VERDICT, 964 BPF_CGROUP_DEVICE, 965 BPF_SK_MSG_VERDICT, 966 BPF_CGROUP_INET4_BIND, 967 BPF_CGROUP_INET6_BIND, 968 BPF_CGROUP_INET4_CONNECT, 969 BPF_CGROUP_INET6_CONNECT, 970 BPF_CGROUP_INET4_POST_BIND, 971 BPF_CGROUP_INET6_POST_BIND, 972 BPF_CGROUP_UDP4_SENDMSG, 973 BPF_CGROUP_UDP6_SENDMSG, 974 BPF_LIRC_MODE2, 975 BPF_FLOW_DISSECTOR, 976 BPF_CGROUP_SYSCTL, 977 BPF_CGROUP_UDP4_RECVMSG, 978 BPF_CGROUP_UDP6_RECVMSG, 979 BPF_CGROUP_GETSOCKOPT, 980 BPF_CGROUP_SETSOCKOPT, 981 BPF_TRACE_RAW_TP, 982 BPF_TRACE_FENTRY, 983 BPF_TRACE_FEXIT, 984 BPF_MODIFY_RETURN, 985 BPF_LSM_MAC, 986 BPF_TRACE_ITER, 987 BPF_CGROUP_INET4_GETPEERNAME, 988 BPF_CGROUP_INET6_GETPEERNAME, 989 BPF_CGROUP_INET4_GETSOCKNAME, 990 BPF_CGROUP_INET6_GETSOCKNAME, 991 BPF_XDP_DEVMAP, 992 BPF_CGROUP_INET_SOCK_RELEASE, 993 BPF_XDP_CPUMAP, 994 BPF_SK_LOOKUP, 995 BPF_XDP, 996 BPF_SK_SKB_VERDICT, 997 BPF_SK_REUSEPORT_SELECT, 998 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 999 BPF_PERF_EVENT, 1000 __MAX_BPF_ATTACH_TYPE 1001 }; 1002 1003 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1004 1005 enum bpf_link_type { 1006 BPF_LINK_TYPE_UNSPEC = 0, 1007 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1008 BPF_LINK_TYPE_TRACING = 2, 1009 BPF_LINK_TYPE_CGROUP = 3, 1010 BPF_LINK_TYPE_ITER = 4, 1011 BPF_LINK_TYPE_NETNS = 5, 1012 BPF_LINK_TYPE_XDP = 6, 1013 BPF_LINK_TYPE_PERF_EVENT = 7, 1014 1015 MAX_BPF_LINK_TYPE, 1016 }; 1017 1018 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1019 * 1020 * NONE(default): No further bpf programs allowed in the subtree. 1021 * 1022 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1023 * the program in this cgroup yields to sub-cgroup program. 1024 * 1025 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1026 * that cgroup program gets run in addition to the program in this cgroup. 1027 * 1028 * Only one program is allowed to be attached to a cgroup with 1029 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1030 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1031 * release old program and attach the new one. Attach flags has to match. 1032 * 1033 * Multiple programs are allowed to be attached to a cgroup with 1034 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1035 * (those that were attached first, run first) 1036 * The programs of sub-cgroup are executed first, then programs of 1037 * this cgroup and then programs of parent cgroup. 1038 * When children program makes decision (like picking TCP CA or sock bind) 1039 * parent program has a chance to override it. 1040 * 1041 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1042 * programs for a cgroup. Though it's possible to replace an old program at 1043 * any position by also specifying BPF_F_REPLACE flag and position itself in 1044 * replace_bpf_fd attribute. Old program at this position will be released. 1045 * 1046 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1047 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1048 * Ex1: 1049 * cgrp1 (MULTI progs A, B) -> 1050 * cgrp2 (OVERRIDE prog C) -> 1051 * cgrp3 (MULTI prog D) -> 1052 * cgrp4 (OVERRIDE prog E) -> 1053 * cgrp5 (NONE prog F) 1054 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1055 * if prog F is detached, the execution is E,D,A,B 1056 * if prog F and D are detached, the execution is E,A,B 1057 * if prog F, E and D are detached, the execution is C,A,B 1058 * 1059 * All eligible programs are executed regardless of return code from 1060 * earlier programs. 1061 */ 1062 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1063 #define BPF_F_ALLOW_MULTI (1U << 1) 1064 #define BPF_F_REPLACE (1U << 2) 1065 1066 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1067 * verifier will perform strict alignment checking as if the kernel 1068 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1069 * and NET_IP_ALIGN defined to 2. 1070 */ 1071 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1072 1073 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1074 * verifier will allow any alignment whatsoever. On platforms 1075 * with strict alignment requirements for loads ands stores (such 1076 * as sparc and mips) the verifier validates that all loads and 1077 * stores provably follow this requirement. This flag turns that 1078 * checking and enforcement off. 1079 * 1080 * It is mostly used for testing when we want to validate the 1081 * context and memory access aspects of the verifier, but because 1082 * of an unaligned access the alignment check would trigger before 1083 * the one we are interested in. 1084 */ 1085 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1086 1087 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1088 * Verifier does sub-register def/use analysis and identifies instructions whose 1089 * def only matters for low 32-bit, high 32-bit is never referenced later 1090 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1091 * that it is safe to ignore clearing high 32-bit for these instructions. This 1092 * saves some back-ends a lot of code-gen. However such optimization is not 1093 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1094 * hence hasn't used verifier's analysis result. But, we really want to have a 1095 * way to be able to verify the correctness of the described optimization on 1096 * x86_64 on which testsuites are frequently exercised. 1097 * 1098 * So, this flag is introduced. Once it is set, verifier will randomize high 1099 * 32-bit for those instructions who has been identified as safe to ignore them. 1100 * Then, if verifier is not doing correct analysis, such randomization will 1101 * regress tests to expose bugs. 1102 */ 1103 #define BPF_F_TEST_RND_HI32 (1U << 2) 1104 1105 /* The verifier internal test flag. Behavior is undefined */ 1106 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1107 1108 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1109 * restrict map and helper usage for such programs. Sleepable BPF programs can 1110 * only be attached to hooks where kernel execution context allows sleeping. 1111 * Such programs are allowed to use helpers that may sleep like 1112 * bpf_copy_from_user(). 1113 */ 1114 #define BPF_F_SLEEPABLE (1U << 4) 1115 1116 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1117 * fully support xdp frags. 1118 */ 1119 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1120 1121 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1122 * the following extensions: 1123 * 1124 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1125 * insn[0].imm: map fd or fd_idx 1126 * insn[1].imm: 0 1127 * insn[0].off: 0 1128 * insn[1].off: 0 1129 * ldimm64 rewrite: address of map 1130 * verifier type: CONST_PTR_TO_MAP 1131 */ 1132 #define BPF_PSEUDO_MAP_FD 1 1133 #define BPF_PSEUDO_MAP_IDX 5 1134 1135 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1136 * insn[0].imm: map fd or fd_idx 1137 * insn[1].imm: offset into value 1138 * insn[0].off: 0 1139 * insn[1].off: 0 1140 * ldimm64 rewrite: address of map[0]+offset 1141 * verifier type: PTR_TO_MAP_VALUE 1142 */ 1143 #define BPF_PSEUDO_MAP_VALUE 2 1144 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1145 1146 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1147 * insn[0].imm: kernel btd id of VAR 1148 * insn[1].imm: 0 1149 * insn[0].off: 0 1150 * insn[1].off: 0 1151 * ldimm64 rewrite: address of the kernel variable 1152 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1153 * is struct/union. 1154 */ 1155 #define BPF_PSEUDO_BTF_ID 3 1156 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1157 * insn[0].imm: insn offset to the func 1158 * insn[1].imm: 0 1159 * insn[0].off: 0 1160 * insn[1].off: 0 1161 * ldimm64 rewrite: address of the function 1162 * verifier type: PTR_TO_FUNC. 1163 */ 1164 #define BPF_PSEUDO_FUNC 4 1165 1166 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1167 * offset to another bpf function 1168 */ 1169 #define BPF_PSEUDO_CALL 1 1170 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1171 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1172 */ 1173 #define BPF_PSEUDO_KFUNC_CALL 2 1174 1175 /* flags for BPF_MAP_UPDATE_ELEM command */ 1176 enum { 1177 BPF_ANY = 0, /* create new element or update existing */ 1178 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1179 BPF_EXIST = 2, /* update existing element */ 1180 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1181 }; 1182 1183 /* flags for BPF_MAP_CREATE command */ 1184 enum { 1185 BPF_F_NO_PREALLOC = (1U << 0), 1186 /* Instead of having one common LRU list in the 1187 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1188 * which can scale and perform better. 1189 * Note, the LRU nodes (including free nodes) cannot be moved 1190 * across different LRU lists. 1191 */ 1192 BPF_F_NO_COMMON_LRU = (1U << 1), 1193 /* Specify numa node during map creation */ 1194 BPF_F_NUMA_NODE = (1U << 2), 1195 1196 /* Flags for accessing BPF object from syscall side. */ 1197 BPF_F_RDONLY = (1U << 3), 1198 BPF_F_WRONLY = (1U << 4), 1199 1200 /* Flag for stack_map, store build_id+offset instead of pointer */ 1201 BPF_F_STACK_BUILD_ID = (1U << 5), 1202 1203 /* Zero-initialize hash function seed. This should only be used for testing. */ 1204 BPF_F_ZERO_SEED = (1U << 6), 1205 1206 /* Flags for accessing BPF object from program side. */ 1207 BPF_F_RDONLY_PROG = (1U << 7), 1208 BPF_F_WRONLY_PROG = (1U << 8), 1209 1210 /* Clone map from listener for newly accepted socket */ 1211 BPF_F_CLONE = (1U << 9), 1212 1213 /* Enable memory-mapping BPF map */ 1214 BPF_F_MMAPABLE = (1U << 10), 1215 1216 /* Share perf_event among processes */ 1217 BPF_F_PRESERVE_ELEMS = (1U << 11), 1218 1219 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1220 BPF_F_INNER_MAP = (1U << 12), 1221 }; 1222 1223 /* Flags for BPF_PROG_QUERY. */ 1224 1225 /* Query effective (directly attached + inherited from ancestor cgroups) 1226 * programs that will be executed for events within a cgroup. 1227 * attach_flags with this flag are returned only for directly attached programs. 1228 */ 1229 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1230 1231 /* Flags for BPF_PROG_TEST_RUN */ 1232 1233 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1234 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1235 /* If set, XDP frames will be transmitted after processing */ 1236 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1237 1238 /* type for BPF_ENABLE_STATS */ 1239 enum bpf_stats_type { 1240 /* enabled run_time_ns and run_cnt */ 1241 BPF_STATS_RUN_TIME = 0, 1242 }; 1243 1244 enum bpf_stack_build_id_status { 1245 /* user space need an empty entry to identify end of a trace */ 1246 BPF_STACK_BUILD_ID_EMPTY = 0, 1247 /* with valid build_id and offset */ 1248 BPF_STACK_BUILD_ID_VALID = 1, 1249 /* couldn't get build_id, fallback to ip */ 1250 BPF_STACK_BUILD_ID_IP = 2, 1251 }; 1252 1253 #define BPF_BUILD_ID_SIZE 20 1254 struct bpf_stack_build_id { 1255 __s32 status; 1256 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1257 union { 1258 __u64 offset; 1259 __u64 ip; 1260 }; 1261 }; 1262 1263 #define BPF_OBJ_NAME_LEN 16U 1264 1265 union bpf_attr { 1266 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1267 __u32 map_type; /* one of enum bpf_map_type */ 1268 __u32 key_size; /* size of key in bytes */ 1269 __u32 value_size; /* size of value in bytes */ 1270 __u32 max_entries; /* max number of entries in a map */ 1271 __u32 map_flags; /* BPF_MAP_CREATE related 1272 * flags defined above. 1273 */ 1274 __u32 inner_map_fd; /* fd pointing to the inner map */ 1275 __u32 numa_node; /* numa node (effective only if 1276 * BPF_F_NUMA_NODE is set). 1277 */ 1278 char map_name[BPF_OBJ_NAME_LEN]; 1279 __u32 map_ifindex; /* ifindex of netdev to create on */ 1280 __u32 btf_fd; /* fd pointing to a BTF type data */ 1281 __u32 btf_key_type_id; /* BTF type_id of the key */ 1282 __u32 btf_value_type_id; /* BTF type_id of the value */ 1283 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1284 * struct stored as the 1285 * map value 1286 */ 1287 /* Any per-map-type extra fields 1288 * 1289 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1290 * number of hash functions (if 0, the bloom filter will default 1291 * to using 5 hash functions). 1292 */ 1293 __u64 map_extra; 1294 }; 1295 1296 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1297 __u32 map_fd; 1298 __aligned_u64 key; 1299 union { 1300 __aligned_u64 value; 1301 __aligned_u64 next_key; 1302 }; 1303 __u64 flags; 1304 }; 1305 1306 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1307 __aligned_u64 in_batch; /* start batch, 1308 * NULL to start from beginning 1309 */ 1310 __aligned_u64 out_batch; /* output: next start batch */ 1311 __aligned_u64 keys; 1312 __aligned_u64 values; 1313 __u32 count; /* input/output: 1314 * input: # of key/value 1315 * elements 1316 * output: # of filled elements 1317 */ 1318 __u32 map_fd; 1319 __u64 elem_flags; 1320 __u64 flags; 1321 } batch; 1322 1323 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1324 __u32 prog_type; /* one of enum bpf_prog_type */ 1325 __u32 insn_cnt; 1326 __aligned_u64 insns; 1327 __aligned_u64 license; 1328 __u32 log_level; /* verbosity level of verifier */ 1329 __u32 log_size; /* size of user buffer */ 1330 __aligned_u64 log_buf; /* user supplied buffer */ 1331 __u32 kern_version; /* not used */ 1332 __u32 prog_flags; 1333 char prog_name[BPF_OBJ_NAME_LEN]; 1334 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1335 /* For some prog types expected attach type must be known at 1336 * load time to verify attach type specific parts of prog 1337 * (context accesses, allowed helpers, etc). 1338 */ 1339 __u32 expected_attach_type; 1340 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1341 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1342 __aligned_u64 func_info; /* func info */ 1343 __u32 func_info_cnt; /* number of bpf_func_info records */ 1344 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1345 __aligned_u64 line_info; /* line info */ 1346 __u32 line_info_cnt; /* number of bpf_line_info records */ 1347 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1348 union { 1349 /* valid prog_fd to attach to bpf prog */ 1350 __u32 attach_prog_fd; 1351 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1352 __u32 attach_btf_obj_fd; 1353 }; 1354 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1355 __aligned_u64 fd_array; /* array of FDs */ 1356 __aligned_u64 core_relos; 1357 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1358 }; 1359 1360 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1361 __aligned_u64 pathname; 1362 __u32 bpf_fd; 1363 __u32 file_flags; 1364 }; 1365 1366 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1367 __u32 target_fd; /* container object to attach to */ 1368 __u32 attach_bpf_fd; /* eBPF program to attach */ 1369 __u32 attach_type; 1370 __u32 attach_flags; 1371 __u32 replace_bpf_fd; /* previously attached eBPF 1372 * program to replace if 1373 * BPF_F_REPLACE is used 1374 */ 1375 }; 1376 1377 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1378 __u32 prog_fd; 1379 __u32 retval; 1380 __u32 data_size_in; /* input: len of data_in */ 1381 __u32 data_size_out; /* input/output: len of data_out 1382 * returns ENOSPC if data_out 1383 * is too small. 1384 */ 1385 __aligned_u64 data_in; 1386 __aligned_u64 data_out; 1387 __u32 repeat; 1388 __u32 duration; 1389 __u32 ctx_size_in; /* input: len of ctx_in */ 1390 __u32 ctx_size_out; /* input/output: len of ctx_out 1391 * returns ENOSPC if ctx_out 1392 * is too small. 1393 */ 1394 __aligned_u64 ctx_in; 1395 __aligned_u64 ctx_out; 1396 __u32 flags; 1397 __u32 cpu; 1398 __u32 batch_size; 1399 } test; 1400 1401 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1402 union { 1403 __u32 start_id; 1404 __u32 prog_id; 1405 __u32 map_id; 1406 __u32 btf_id; 1407 __u32 link_id; 1408 }; 1409 __u32 next_id; 1410 __u32 open_flags; 1411 }; 1412 1413 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1414 __u32 bpf_fd; 1415 __u32 info_len; 1416 __aligned_u64 info; 1417 } info; 1418 1419 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1420 __u32 target_fd; /* container object to query */ 1421 __u32 attach_type; 1422 __u32 query_flags; 1423 __u32 attach_flags; 1424 __aligned_u64 prog_ids; 1425 __u32 prog_cnt; 1426 } query; 1427 1428 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1429 __u64 name; 1430 __u32 prog_fd; 1431 } raw_tracepoint; 1432 1433 struct { /* anonymous struct for BPF_BTF_LOAD */ 1434 __aligned_u64 btf; 1435 __aligned_u64 btf_log_buf; 1436 __u32 btf_size; 1437 __u32 btf_log_size; 1438 __u32 btf_log_level; 1439 }; 1440 1441 struct { 1442 __u32 pid; /* input: pid */ 1443 __u32 fd; /* input: fd */ 1444 __u32 flags; /* input: flags */ 1445 __u32 buf_len; /* input/output: buf len */ 1446 __aligned_u64 buf; /* input/output: 1447 * tp_name for tracepoint 1448 * symbol for kprobe 1449 * filename for uprobe 1450 */ 1451 __u32 prog_id; /* output: prod_id */ 1452 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1453 __u64 probe_offset; /* output: probe_offset */ 1454 __u64 probe_addr; /* output: probe_addr */ 1455 } task_fd_query; 1456 1457 struct { /* struct used by BPF_LINK_CREATE command */ 1458 __u32 prog_fd; /* eBPF program to attach */ 1459 union { 1460 __u32 target_fd; /* object to attach to */ 1461 __u32 target_ifindex; /* target ifindex */ 1462 }; 1463 __u32 attach_type; /* attach type */ 1464 __u32 flags; /* extra flags */ 1465 union { 1466 __u32 target_btf_id; /* btf_id of target to attach to */ 1467 struct { 1468 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1469 __u32 iter_info_len; /* iter_info length */ 1470 }; 1471 struct { 1472 /* black box user-provided value passed through 1473 * to BPF program at the execution time and 1474 * accessible through bpf_get_attach_cookie() BPF helper 1475 */ 1476 __u64 bpf_cookie; 1477 } perf_event; 1478 }; 1479 } link_create; 1480 1481 struct { /* struct used by BPF_LINK_UPDATE command */ 1482 __u32 link_fd; /* link fd */ 1483 /* new program fd to update link with */ 1484 __u32 new_prog_fd; 1485 __u32 flags; /* extra flags */ 1486 /* expected link's program fd; is specified only if 1487 * BPF_F_REPLACE flag is set in flags */ 1488 __u32 old_prog_fd; 1489 } link_update; 1490 1491 struct { 1492 __u32 link_fd; 1493 } link_detach; 1494 1495 struct { /* struct used by BPF_ENABLE_STATS command */ 1496 __u32 type; 1497 } enable_stats; 1498 1499 struct { /* struct used by BPF_ITER_CREATE command */ 1500 __u32 link_fd; 1501 __u32 flags; 1502 } iter_create; 1503 1504 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1505 __u32 prog_fd; 1506 __u32 map_fd; 1507 __u32 flags; /* extra flags */ 1508 } prog_bind_map; 1509 1510 } __attribute__((aligned(8))); 1511 1512 /* The description below is an attempt at providing documentation to eBPF 1513 * developers about the multiple available eBPF helper functions. It can be 1514 * parsed and used to produce a manual page. The workflow is the following, 1515 * and requires the rst2man utility: 1516 * 1517 * $ ./scripts/bpf_doc.py \ 1518 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1519 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1520 * $ man /tmp/bpf-helpers.7 1521 * 1522 * Note that in order to produce this external documentation, some RST 1523 * formatting is used in the descriptions to get "bold" and "italics" in 1524 * manual pages. Also note that the few trailing white spaces are 1525 * intentional, removing them would break paragraphs for rst2man. 1526 * 1527 * Start of BPF helper function descriptions: 1528 * 1529 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1530 * Description 1531 * Perform a lookup in *map* for an entry associated to *key*. 1532 * Return 1533 * Map value associated to *key*, or **NULL** if no entry was 1534 * found. 1535 * 1536 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1537 * Description 1538 * Add or update the value of the entry associated to *key* in 1539 * *map* with *value*. *flags* is one of: 1540 * 1541 * **BPF_NOEXIST** 1542 * The entry for *key* must not exist in the map. 1543 * **BPF_EXIST** 1544 * The entry for *key* must already exist in the map. 1545 * **BPF_ANY** 1546 * No condition on the existence of the entry for *key*. 1547 * 1548 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1549 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1550 * elements always exist), the helper would return an error. 1551 * Return 1552 * 0 on success, or a negative error in case of failure. 1553 * 1554 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1555 * Description 1556 * Delete entry with *key* from *map*. 1557 * Return 1558 * 0 on success, or a negative error in case of failure. 1559 * 1560 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1561 * Description 1562 * For tracing programs, safely attempt to read *size* bytes from 1563 * kernel space address *unsafe_ptr* and store the data in *dst*. 1564 * 1565 * Generally, use **bpf_probe_read_user**\ () or 1566 * **bpf_probe_read_kernel**\ () instead. 1567 * Return 1568 * 0 on success, or a negative error in case of failure. 1569 * 1570 * u64 bpf_ktime_get_ns(void) 1571 * Description 1572 * Return the time elapsed since system boot, in nanoseconds. 1573 * Does not include time the system was suspended. 1574 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1575 * Return 1576 * Current *ktime*. 1577 * 1578 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1579 * Description 1580 * This helper is a "printk()-like" facility for debugging. It 1581 * prints a message defined by format *fmt* (of size *fmt_size*) 1582 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1583 * available. It can take up to three additional **u64** 1584 * arguments (as an eBPF helpers, the total number of arguments is 1585 * limited to five). 1586 * 1587 * Each time the helper is called, it appends a line to the trace. 1588 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 1589 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1590 * The format of the trace is customizable, and the exact output 1591 * one will get depends on the options set in 1592 * *\/sys/kernel/debug/tracing/trace_options* (see also the 1593 * *README* file under the same directory). However, it usually 1594 * defaults to something like: 1595 * 1596 * :: 1597 * 1598 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1599 * 1600 * In the above: 1601 * 1602 * * ``telnet`` is the name of the current task. 1603 * * ``470`` is the PID of the current task. 1604 * * ``001`` is the CPU number on which the task is 1605 * running. 1606 * * In ``.N..``, each character refers to a set of 1607 * options (whether irqs are enabled, scheduling 1608 * options, whether hard/softirqs are running, level of 1609 * preempt_disabled respectively). **N** means that 1610 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1611 * are set. 1612 * * ``419421.045894`` is a timestamp. 1613 * * ``0x00000001`` is a fake value used by BPF for the 1614 * instruction pointer register. 1615 * * ``<formatted msg>`` is the message formatted with 1616 * *fmt*. 1617 * 1618 * The conversion specifiers supported by *fmt* are similar, but 1619 * more limited than for printk(). They are **%d**, **%i**, 1620 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1621 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1622 * of field, padding with zeroes, etc.) is available, and the 1623 * helper will return **-EINVAL** (but print nothing) if it 1624 * encounters an unknown specifier. 1625 * 1626 * Also, note that **bpf_trace_printk**\ () is slow, and should 1627 * only be used for debugging purposes. For this reason, a notice 1628 * block (spanning several lines) is printed to kernel logs and 1629 * states that the helper should not be used "for production use" 1630 * the first time this helper is used (or more precisely, when 1631 * **trace_printk**\ () buffers are allocated). For passing values 1632 * to user space, perf events should be preferred. 1633 * Return 1634 * The number of bytes written to the buffer, or a negative error 1635 * in case of failure. 1636 * 1637 * u32 bpf_get_prandom_u32(void) 1638 * Description 1639 * Get a pseudo-random number. 1640 * 1641 * From a security point of view, this helper uses its own 1642 * pseudo-random internal state, and cannot be used to infer the 1643 * seed of other random functions in the kernel. However, it is 1644 * essential to note that the generator used by the helper is not 1645 * cryptographically secure. 1646 * Return 1647 * A random 32-bit unsigned value. 1648 * 1649 * u32 bpf_get_smp_processor_id(void) 1650 * Description 1651 * Get the SMP (symmetric multiprocessing) processor id. Note that 1652 * all programs run with migration disabled, which means that the 1653 * SMP processor id is stable during all the execution of the 1654 * program. 1655 * Return 1656 * The SMP id of the processor running the program. 1657 * 1658 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1659 * Description 1660 * Store *len* bytes from address *from* into the packet 1661 * associated to *skb*, at *offset*. *flags* are a combination of 1662 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1663 * checksum for the packet after storing the bytes) and 1664 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1665 * **->swhash** and *skb*\ **->l4hash** to 0). 1666 * 1667 * A call to this helper is susceptible to change the underlying 1668 * packet buffer. Therefore, at load time, all checks on pointers 1669 * previously done by the verifier are invalidated and must be 1670 * performed again, if the helper is used in combination with 1671 * direct packet access. 1672 * Return 1673 * 0 on success, or a negative error in case of failure. 1674 * 1675 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1676 * Description 1677 * Recompute the layer 3 (e.g. IP) checksum for the packet 1678 * associated to *skb*. Computation is incremental, so the helper 1679 * must know the former value of the header field that was 1680 * modified (*from*), the new value of this field (*to*), and the 1681 * number of bytes (2 or 4) for this field, stored in *size*. 1682 * Alternatively, it is possible to store the difference between 1683 * the previous and the new values of the header field in *to*, by 1684 * setting *from* and *size* to 0. For both methods, *offset* 1685 * indicates the location of the IP checksum within the packet. 1686 * 1687 * This helper works in combination with **bpf_csum_diff**\ (), 1688 * which does not update the checksum in-place, but offers more 1689 * flexibility and can handle sizes larger than 2 or 4 for the 1690 * checksum to update. 1691 * 1692 * A call to this helper is susceptible to change the underlying 1693 * packet buffer. Therefore, at load time, all checks on pointers 1694 * previously done by the verifier are invalidated and must be 1695 * performed again, if the helper is used in combination with 1696 * direct packet access. 1697 * Return 1698 * 0 on success, or a negative error in case of failure. 1699 * 1700 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1701 * Description 1702 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1703 * packet associated to *skb*. Computation is incremental, so the 1704 * helper must know the former value of the header field that was 1705 * modified (*from*), the new value of this field (*to*), and the 1706 * number of bytes (2 or 4) for this field, stored on the lowest 1707 * four bits of *flags*. Alternatively, it is possible to store 1708 * the difference between the previous and the new values of the 1709 * header field in *to*, by setting *from* and the four lowest 1710 * bits of *flags* to 0. For both methods, *offset* indicates the 1711 * location of the IP checksum within the packet. In addition to 1712 * the size of the field, *flags* can be added (bitwise OR) actual 1713 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1714 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1715 * for updates resulting in a null checksum the value is set to 1716 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1717 * the checksum is to be computed against a pseudo-header. 1718 * 1719 * This helper works in combination with **bpf_csum_diff**\ (), 1720 * which does not update the checksum in-place, but offers more 1721 * flexibility and can handle sizes larger than 2 or 4 for the 1722 * checksum to update. 1723 * 1724 * A call to this helper is susceptible to change the underlying 1725 * packet buffer. Therefore, at load time, all checks on pointers 1726 * previously done by the verifier are invalidated and must be 1727 * performed again, if the helper is used in combination with 1728 * direct packet access. 1729 * Return 1730 * 0 on success, or a negative error in case of failure. 1731 * 1732 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1733 * Description 1734 * This special helper is used to trigger a "tail call", or in 1735 * other words, to jump into another eBPF program. The same stack 1736 * frame is used (but values on stack and in registers for the 1737 * caller are not accessible to the callee). This mechanism allows 1738 * for program chaining, either for raising the maximum number of 1739 * available eBPF instructions, or to execute given programs in 1740 * conditional blocks. For security reasons, there is an upper 1741 * limit to the number of successive tail calls that can be 1742 * performed. 1743 * 1744 * Upon call of this helper, the program attempts to jump into a 1745 * program referenced at index *index* in *prog_array_map*, a 1746 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1747 * *ctx*, a pointer to the context. 1748 * 1749 * If the call succeeds, the kernel immediately runs the first 1750 * instruction of the new program. This is not a function call, 1751 * and it never returns to the previous program. If the call 1752 * fails, then the helper has no effect, and the caller continues 1753 * to run its subsequent instructions. A call can fail if the 1754 * destination program for the jump does not exist (i.e. *index* 1755 * is superior to the number of entries in *prog_array_map*), or 1756 * if the maximum number of tail calls has been reached for this 1757 * chain of programs. This limit is defined in the kernel by the 1758 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1759 * which is currently set to 33. 1760 * Return 1761 * 0 on success, or a negative error in case of failure. 1762 * 1763 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1764 * Description 1765 * Clone and redirect the packet associated to *skb* to another 1766 * net device of index *ifindex*. Both ingress and egress 1767 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1768 * value in *flags* is used to make the distinction (ingress path 1769 * is selected if the flag is present, egress path otherwise). 1770 * This is the only flag supported for now. 1771 * 1772 * In comparison with **bpf_redirect**\ () helper, 1773 * **bpf_clone_redirect**\ () has the associated cost of 1774 * duplicating the packet buffer, but this can be executed out of 1775 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1776 * efficient, but it is handled through an action code where the 1777 * redirection happens only after the eBPF program has returned. 1778 * 1779 * A call to this helper is susceptible to change the underlying 1780 * packet buffer. Therefore, at load time, all checks on pointers 1781 * previously done by the verifier are invalidated and must be 1782 * performed again, if the helper is used in combination with 1783 * direct packet access. 1784 * Return 1785 * 0 on success, or a negative error in case of failure. 1786 * 1787 * u64 bpf_get_current_pid_tgid(void) 1788 * Description 1789 * Get the current pid and tgid. 1790 * Return 1791 * A 64-bit integer containing the current tgid and pid, and 1792 * created as such: 1793 * *current_task*\ **->tgid << 32 \|** 1794 * *current_task*\ **->pid**. 1795 * 1796 * u64 bpf_get_current_uid_gid(void) 1797 * Description 1798 * Get the current uid and gid. 1799 * Return 1800 * A 64-bit integer containing the current GID and UID, and 1801 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1802 * 1803 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1804 * Description 1805 * Copy the **comm** attribute of the current task into *buf* of 1806 * *size_of_buf*. The **comm** attribute contains the name of 1807 * the executable (excluding the path) for the current task. The 1808 * *size_of_buf* must be strictly positive. On success, the 1809 * helper makes sure that the *buf* is NUL-terminated. On failure, 1810 * it is filled with zeroes. 1811 * Return 1812 * 0 on success, or a negative error in case of failure. 1813 * 1814 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1815 * Description 1816 * Retrieve the classid for the current task, i.e. for the net_cls 1817 * cgroup to which *skb* belongs. 1818 * 1819 * This helper can be used on TC egress path, but not on ingress. 1820 * 1821 * The net_cls cgroup provides an interface to tag network packets 1822 * based on a user-provided identifier for all traffic coming from 1823 * the tasks belonging to the related cgroup. See also the related 1824 * kernel documentation, available from the Linux sources in file 1825 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1826 * 1827 * The Linux kernel has two versions for cgroups: there are 1828 * cgroups v1 and cgroups v2. Both are available to users, who can 1829 * use a mixture of them, but note that the net_cls cgroup is for 1830 * cgroup v1 only. This makes it incompatible with BPF programs 1831 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1832 * only hold data for one version of cgroups at a time). 1833 * 1834 * This helper is only available is the kernel was compiled with 1835 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1836 * "**y**" or to "**m**". 1837 * Return 1838 * The classid, or 0 for the default unconfigured classid. 1839 * 1840 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1841 * Description 1842 * Push a *vlan_tci* (VLAN tag control information) of protocol 1843 * *vlan_proto* to the packet associated to *skb*, then update 1844 * the checksum. Note that if *vlan_proto* is different from 1845 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1846 * be **ETH_P_8021Q**. 1847 * 1848 * A call to this helper is susceptible to change the underlying 1849 * packet buffer. Therefore, at load time, all checks on pointers 1850 * previously done by the verifier are invalidated and must be 1851 * performed again, if the helper is used in combination with 1852 * direct packet access. 1853 * Return 1854 * 0 on success, or a negative error in case of failure. 1855 * 1856 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1857 * Description 1858 * Pop a VLAN header from the packet associated to *skb*. 1859 * 1860 * A call to this helper is susceptible to change the underlying 1861 * packet buffer. Therefore, at load time, all checks on pointers 1862 * previously done by the verifier are invalidated and must be 1863 * performed again, if the helper is used in combination with 1864 * direct packet access. 1865 * Return 1866 * 0 on success, or a negative error in case of failure. 1867 * 1868 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1869 * Description 1870 * Get tunnel metadata. This helper takes a pointer *key* to an 1871 * empty **struct bpf_tunnel_key** of **size**, that will be 1872 * filled with tunnel metadata for the packet associated to *skb*. 1873 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1874 * indicates that the tunnel is based on IPv6 protocol instead of 1875 * IPv4. 1876 * 1877 * The **struct bpf_tunnel_key** is an object that generalizes the 1878 * principal parameters used by various tunneling protocols into a 1879 * single struct. This way, it can be used to easily make a 1880 * decision based on the contents of the encapsulation header, 1881 * "summarized" in this struct. In particular, it holds the IP 1882 * address of the remote end (IPv4 or IPv6, depending on the case) 1883 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1884 * this struct exposes the *key*\ **->tunnel_id**, which is 1885 * generally mapped to a VNI (Virtual Network Identifier), making 1886 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1887 * () helper. 1888 * 1889 * Let's imagine that the following code is part of a program 1890 * attached to the TC ingress interface, on one end of a GRE 1891 * tunnel, and is supposed to filter out all messages coming from 1892 * remote ends with IPv4 address other than 10.0.0.1: 1893 * 1894 * :: 1895 * 1896 * int ret; 1897 * struct bpf_tunnel_key key = {}; 1898 * 1899 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1900 * if (ret < 0) 1901 * return TC_ACT_SHOT; // drop packet 1902 * 1903 * if (key.remote_ipv4 != 0x0a000001) 1904 * return TC_ACT_SHOT; // drop packet 1905 * 1906 * return TC_ACT_OK; // accept packet 1907 * 1908 * This interface can also be used with all encapsulation devices 1909 * that can operate in "collect metadata" mode: instead of having 1910 * one network device per specific configuration, the "collect 1911 * metadata" mode only requires a single device where the 1912 * configuration can be extracted from this helper. 1913 * 1914 * This can be used together with various tunnels such as VXLan, 1915 * Geneve, GRE or IP in IP (IPIP). 1916 * Return 1917 * 0 on success, or a negative error in case of failure. 1918 * 1919 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1920 * Description 1921 * Populate tunnel metadata for packet associated to *skb.* The 1922 * tunnel metadata is set to the contents of *key*, of *size*. The 1923 * *flags* can be set to a combination of the following values: 1924 * 1925 * **BPF_F_TUNINFO_IPV6** 1926 * Indicate that the tunnel is based on IPv6 protocol 1927 * instead of IPv4. 1928 * **BPF_F_ZERO_CSUM_TX** 1929 * For IPv4 packets, add a flag to tunnel metadata 1930 * indicating that checksum computation should be skipped 1931 * and checksum set to zeroes. 1932 * **BPF_F_DONT_FRAGMENT** 1933 * Add a flag to tunnel metadata indicating that the 1934 * packet should not be fragmented. 1935 * **BPF_F_SEQ_NUMBER** 1936 * Add a flag to tunnel metadata indicating that a 1937 * sequence number should be added to tunnel header before 1938 * sending the packet. This flag was added for GRE 1939 * encapsulation, but might be used with other protocols 1940 * as well in the future. 1941 * 1942 * Here is a typical usage on the transmit path: 1943 * 1944 * :: 1945 * 1946 * struct bpf_tunnel_key key; 1947 * populate key ... 1948 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1949 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1950 * 1951 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1952 * helper for additional information. 1953 * Return 1954 * 0 on success, or a negative error in case of failure. 1955 * 1956 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1957 * Description 1958 * Read the value of a perf event counter. This helper relies on a 1959 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1960 * the perf event counter is selected when *map* is updated with 1961 * perf event file descriptors. The *map* is an array whose size 1962 * is the number of available CPUs, and each cell contains a value 1963 * relative to one CPU. The value to retrieve is indicated by 1964 * *flags*, that contains the index of the CPU to look up, masked 1965 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1966 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1967 * current CPU should be retrieved. 1968 * 1969 * Note that before Linux 4.13, only hardware perf event can be 1970 * retrieved. 1971 * 1972 * Also, be aware that the newer helper 1973 * **bpf_perf_event_read_value**\ () is recommended over 1974 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1975 * quirks where error and counter value are used as a return code 1976 * (which is wrong to do since ranges may overlap). This issue is 1977 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1978 * time provides more features over the **bpf_perf_event_read**\ 1979 * () interface. Please refer to the description of 1980 * **bpf_perf_event_read_value**\ () for details. 1981 * Return 1982 * The value of the perf event counter read from the map, or a 1983 * negative error code in case of failure. 1984 * 1985 * long bpf_redirect(u32 ifindex, u64 flags) 1986 * Description 1987 * Redirect the packet to another net device of index *ifindex*. 1988 * This helper is somewhat similar to **bpf_clone_redirect**\ 1989 * (), except that the packet is not cloned, which provides 1990 * increased performance. 1991 * 1992 * Except for XDP, both ingress and egress interfaces can be used 1993 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1994 * to make the distinction (ingress path is selected if the flag 1995 * is present, egress path otherwise). Currently, XDP only 1996 * supports redirection to the egress interface, and accepts no 1997 * flag at all. 1998 * 1999 * The same effect can also be attained with the more generic 2000 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2001 * redirect target instead of providing it directly to the helper. 2002 * Return 2003 * For XDP, the helper returns **XDP_REDIRECT** on success or 2004 * **XDP_ABORTED** on error. For other program types, the values 2005 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2006 * error. 2007 * 2008 * u32 bpf_get_route_realm(struct sk_buff *skb) 2009 * Description 2010 * Retrieve the realm or the route, that is to say the 2011 * **tclassid** field of the destination for the *skb*. The 2012 * identifier retrieved is a user-provided tag, similar to the 2013 * one used with the net_cls cgroup (see description for 2014 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2015 * held by a route (a destination entry), not by a task. 2016 * 2017 * Retrieving this identifier works with the clsact TC egress hook 2018 * (see also **tc-bpf(8)**), or alternatively on conventional 2019 * classful egress qdiscs, but not on TC ingress path. In case of 2020 * clsact TC egress hook, this has the advantage that, internally, 2021 * the destination entry has not been dropped yet in the transmit 2022 * path. Therefore, the destination entry does not need to be 2023 * artificially held via **netif_keep_dst**\ () for a classful 2024 * qdisc until the *skb* is freed. 2025 * 2026 * This helper is available only if the kernel was compiled with 2027 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2028 * Return 2029 * The realm of the route for the packet associated to *skb*, or 0 2030 * if none was found. 2031 * 2032 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2033 * Description 2034 * Write raw *data* blob into a special BPF perf event held by 2035 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2036 * event must have the following attributes: **PERF_SAMPLE_RAW** 2037 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2038 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2039 * 2040 * The *flags* are used to indicate the index in *map* for which 2041 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2042 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2043 * to indicate that the index of the current CPU core should be 2044 * used. 2045 * 2046 * The value to write, of *size*, is passed through eBPF stack and 2047 * pointed by *data*. 2048 * 2049 * The context of the program *ctx* needs also be passed to the 2050 * helper. 2051 * 2052 * On user space, a program willing to read the values needs to 2053 * call **perf_event_open**\ () on the perf event (either for 2054 * one or for all CPUs) and to store the file descriptor into the 2055 * *map*. This must be done before the eBPF program can send data 2056 * into it. An example is available in file 2057 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2058 * tree (the eBPF program counterpart is in 2059 * *samples/bpf/trace_output_kern.c*). 2060 * 2061 * **bpf_perf_event_output**\ () achieves better performance 2062 * than **bpf_trace_printk**\ () for sharing data with user 2063 * space, and is much better suitable for streaming data from eBPF 2064 * programs. 2065 * 2066 * Note that this helper is not restricted to tracing use cases 2067 * and can be used with programs attached to TC or XDP as well, 2068 * where it allows for passing data to user space listeners. Data 2069 * can be: 2070 * 2071 * * Only custom structs, 2072 * * Only the packet payload, or 2073 * * A combination of both. 2074 * Return 2075 * 0 on success, or a negative error in case of failure. 2076 * 2077 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2078 * Description 2079 * This helper was provided as an easy way to load data from a 2080 * packet. It can be used to load *len* bytes from *offset* from 2081 * the packet associated to *skb*, into the buffer pointed by 2082 * *to*. 2083 * 2084 * Since Linux 4.7, usage of this helper has mostly been replaced 2085 * by "direct packet access", enabling packet data to be 2086 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2087 * pointing respectively to the first byte of packet data and to 2088 * the byte after the last byte of packet data. However, it 2089 * remains useful if one wishes to read large quantities of data 2090 * at once from a packet into the eBPF stack. 2091 * Return 2092 * 0 on success, or a negative error in case of failure. 2093 * 2094 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2095 * Description 2096 * Walk a user or a kernel stack and return its id. To achieve 2097 * this, the helper needs *ctx*, which is a pointer to the context 2098 * on which the tracing program is executed, and a pointer to a 2099 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2100 * 2101 * The last argument, *flags*, holds the number of stack frames to 2102 * skip (from 0 to 255), masked with 2103 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2104 * a combination of the following flags: 2105 * 2106 * **BPF_F_USER_STACK** 2107 * Collect a user space stack instead of a kernel stack. 2108 * **BPF_F_FAST_STACK_CMP** 2109 * Compare stacks by hash only. 2110 * **BPF_F_REUSE_STACKID** 2111 * If two different stacks hash into the same *stackid*, 2112 * discard the old one. 2113 * 2114 * The stack id retrieved is a 32 bit long integer handle which 2115 * can be further combined with other data (including other stack 2116 * ids) and used as a key into maps. This can be useful for 2117 * generating a variety of graphs (such as flame graphs or off-cpu 2118 * graphs). 2119 * 2120 * For walking a stack, this helper is an improvement over 2121 * **bpf_probe_read**\ (), which can be used with unrolled loops 2122 * but is not efficient and consumes a lot of eBPF instructions. 2123 * Instead, **bpf_get_stackid**\ () can collect up to 2124 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2125 * this limit can be controlled with the **sysctl** program, and 2126 * that it should be manually increased in order to profile long 2127 * user stacks (such as stacks for Java programs). To do so, use: 2128 * 2129 * :: 2130 * 2131 * # sysctl kernel.perf_event_max_stack=<new value> 2132 * Return 2133 * The positive or null stack id on success, or a negative error 2134 * in case of failure. 2135 * 2136 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2137 * Description 2138 * Compute a checksum difference, from the raw buffer pointed by 2139 * *from*, of length *from_size* (that must be a multiple of 4), 2140 * towards the raw buffer pointed by *to*, of size *to_size* 2141 * (same remark). An optional *seed* can be added to the value 2142 * (this can be cascaded, the seed may come from a previous call 2143 * to the helper). 2144 * 2145 * This is flexible enough to be used in several ways: 2146 * 2147 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2148 * checksum, it can be used when pushing new data. 2149 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2150 * checksum, it can be used when removing data from a packet. 2151 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2152 * can be used to compute a diff. Note that *from_size* and 2153 * *to_size* do not need to be equal. 2154 * 2155 * This helper can be used in combination with 2156 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2157 * which one can feed in the difference computed with 2158 * **bpf_csum_diff**\ (). 2159 * Return 2160 * The checksum result, or a negative error code in case of 2161 * failure. 2162 * 2163 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2164 * Description 2165 * Retrieve tunnel options metadata for the packet associated to 2166 * *skb*, and store the raw tunnel option data to the buffer *opt* 2167 * of *size*. 2168 * 2169 * This helper can be used with encapsulation devices that can 2170 * operate in "collect metadata" mode (please refer to the related 2171 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2172 * more details). A particular example where this can be used is 2173 * in combination with the Geneve encapsulation protocol, where it 2174 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2175 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2176 * the eBPF program. This allows for full customization of these 2177 * headers. 2178 * Return 2179 * The size of the option data retrieved. 2180 * 2181 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2182 * Description 2183 * Set tunnel options metadata for the packet associated to *skb* 2184 * to the option data contained in the raw buffer *opt* of *size*. 2185 * 2186 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2187 * helper for additional information. 2188 * Return 2189 * 0 on success, or a negative error in case of failure. 2190 * 2191 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2192 * Description 2193 * Change the protocol of the *skb* to *proto*. Currently 2194 * supported are transition from IPv4 to IPv6, and from IPv6 to 2195 * IPv4. The helper takes care of the groundwork for the 2196 * transition, including resizing the socket buffer. The eBPF 2197 * program is expected to fill the new headers, if any, via 2198 * **skb_store_bytes**\ () and to recompute the checksums with 2199 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2200 * (). The main case for this helper is to perform NAT64 2201 * operations out of an eBPF program. 2202 * 2203 * Internally, the GSO type is marked as dodgy so that headers are 2204 * checked and segments are recalculated by the GSO/GRO engine. 2205 * The size for GSO target is adapted as well. 2206 * 2207 * All values for *flags* are reserved for future usage, and must 2208 * be left at zero. 2209 * 2210 * A call to this helper is susceptible to change the underlying 2211 * packet buffer. Therefore, at load time, all checks on pointers 2212 * previously done by the verifier are invalidated and must be 2213 * performed again, if the helper is used in combination with 2214 * direct packet access. 2215 * Return 2216 * 0 on success, or a negative error in case of failure. 2217 * 2218 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2219 * Description 2220 * Change the packet type for the packet associated to *skb*. This 2221 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2222 * the eBPF program does not have a write access to *skb*\ 2223 * **->pkt_type** beside this helper. Using a helper here allows 2224 * for graceful handling of errors. 2225 * 2226 * The major use case is to change incoming *skb*s to 2227 * **PACKET_HOST** in a programmatic way instead of having to 2228 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2229 * example. 2230 * 2231 * Note that *type* only allows certain values. At this time, they 2232 * are: 2233 * 2234 * **PACKET_HOST** 2235 * Packet is for us. 2236 * **PACKET_BROADCAST** 2237 * Send packet to all. 2238 * **PACKET_MULTICAST** 2239 * Send packet to group. 2240 * **PACKET_OTHERHOST** 2241 * Send packet to someone else. 2242 * Return 2243 * 0 on success, or a negative error in case of failure. 2244 * 2245 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2246 * Description 2247 * Check whether *skb* is a descendant of the cgroup2 held by 2248 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2249 * Return 2250 * The return value depends on the result of the test, and can be: 2251 * 2252 * * 0, if the *skb* failed the cgroup2 descendant test. 2253 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2254 * * A negative error code, if an error occurred. 2255 * 2256 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2257 * Description 2258 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2259 * not set, in particular if the hash was cleared due to mangling, 2260 * recompute this hash. Later accesses to the hash can be done 2261 * directly with *skb*\ **->hash**. 2262 * 2263 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2264 * prototype with **bpf_skb_change_proto**\ (), or calling 2265 * **bpf_skb_store_bytes**\ () with the 2266 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2267 * the hash and to trigger a new computation for the next call to 2268 * **bpf_get_hash_recalc**\ (). 2269 * Return 2270 * The 32-bit hash. 2271 * 2272 * u64 bpf_get_current_task(void) 2273 * Description 2274 * Get the current task. 2275 * Return 2276 * A pointer to the current task struct. 2277 * 2278 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2279 * Description 2280 * Attempt in a safe way to write *len* bytes from the buffer 2281 * *src* to *dst* in memory. It only works for threads that are in 2282 * user context, and *dst* must be a valid user space address. 2283 * 2284 * This helper should not be used to implement any kind of 2285 * security mechanism because of TOC-TOU attacks, but rather to 2286 * debug, divert, and manipulate execution of semi-cooperative 2287 * processes. 2288 * 2289 * Keep in mind that this feature is meant for experiments, and it 2290 * has a risk of crashing the system and running programs. 2291 * Therefore, when an eBPF program using this helper is attached, 2292 * a warning including PID and process name is printed to kernel 2293 * logs. 2294 * Return 2295 * 0 on success, or a negative error in case of failure. 2296 * 2297 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2298 * Description 2299 * Check whether the probe is being run is the context of a given 2300 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2301 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2302 * Return 2303 * The return value depends on the result of the test, and can be: 2304 * 2305 * * 1, if current task belongs to the cgroup2. 2306 * * 0, if current task does not belong to the cgroup2. 2307 * * A negative error code, if an error occurred. 2308 * 2309 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2310 * Description 2311 * Resize (trim or grow) the packet associated to *skb* to the 2312 * new *len*. The *flags* are reserved for future usage, and must 2313 * be left at zero. 2314 * 2315 * The basic idea is that the helper performs the needed work to 2316 * change the size of the packet, then the eBPF program rewrites 2317 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2318 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2319 * and others. This helper is a slow path utility intended for 2320 * replies with control messages. And because it is targeted for 2321 * slow path, the helper itself can afford to be slow: it 2322 * implicitly linearizes, unclones and drops offloads from the 2323 * *skb*. 2324 * 2325 * A call to this helper is susceptible to change the underlying 2326 * packet buffer. Therefore, at load time, all checks on pointers 2327 * previously done by the verifier are invalidated and must be 2328 * performed again, if the helper is used in combination with 2329 * direct packet access. 2330 * Return 2331 * 0 on success, or a negative error in case of failure. 2332 * 2333 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2334 * Description 2335 * Pull in non-linear data in case the *skb* is non-linear and not 2336 * all of *len* are part of the linear section. Make *len* bytes 2337 * from *skb* readable and writable. If a zero value is passed for 2338 * *len*, then the whole length of the *skb* is pulled. 2339 * 2340 * This helper is only needed for reading and writing with direct 2341 * packet access. 2342 * 2343 * For direct packet access, testing that offsets to access 2344 * are within packet boundaries (test on *skb*\ **->data_end**) is 2345 * susceptible to fail if offsets are invalid, or if the requested 2346 * data is in non-linear parts of the *skb*. On failure the 2347 * program can just bail out, or in the case of a non-linear 2348 * buffer, use a helper to make the data available. The 2349 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2350 * the data. Another one consists in using **bpf_skb_pull_data** 2351 * to pull in once the non-linear parts, then retesting and 2352 * eventually access the data. 2353 * 2354 * At the same time, this also makes sure the *skb* is uncloned, 2355 * which is a necessary condition for direct write. As this needs 2356 * to be an invariant for the write part only, the verifier 2357 * detects writes and adds a prologue that is calling 2358 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2359 * the very beginning in case it is indeed cloned. 2360 * 2361 * A call to this helper is susceptible to change the underlying 2362 * packet buffer. Therefore, at load time, all checks on pointers 2363 * previously done by the verifier are invalidated and must be 2364 * performed again, if the helper is used in combination with 2365 * direct packet access. 2366 * Return 2367 * 0 on success, or a negative error in case of failure. 2368 * 2369 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2370 * Description 2371 * Add the checksum *csum* into *skb*\ **->csum** in case the 2372 * driver has supplied a checksum for the entire packet into that 2373 * field. Return an error otherwise. This helper is intended to be 2374 * used in combination with **bpf_csum_diff**\ (), in particular 2375 * when the checksum needs to be updated after data has been 2376 * written into the packet through direct packet access. 2377 * Return 2378 * The checksum on success, or a negative error code in case of 2379 * failure. 2380 * 2381 * void bpf_set_hash_invalid(struct sk_buff *skb) 2382 * Description 2383 * Invalidate the current *skb*\ **->hash**. It can be used after 2384 * mangling on headers through direct packet access, in order to 2385 * indicate that the hash is outdated and to trigger a 2386 * recalculation the next time the kernel tries to access this 2387 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2388 * Return 2389 * void. 2390 * 2391 * long bpf_get_numa_node_id(void) 2392 * Description 2393 * Return the id of the current NUMA node. The primary use case 2394 * for this helper is the selection of sockets for the local NUMA 2395 * node, when the program is attached to sockets using the 2396 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2397 * but the helper is also available to other eBPF program types, 2398 * similarly to **bpf_get_smp_processor_id**\ (). 2399 * Return 2400 * The id of current NUMA node. 2401 * 2402 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2403 * Description 2404 * Grows headroom of packet associated to *skb* and adjusts the 2405 * offset of the MAC header accordingly, adding *len* bytes of 2406 * space. It automatically extends and reallocates memory as 2407 * required. 2408 * 2409 * This helper can be used on a layer 3 *skb* to push a MAC header 2410 * for redirection into a layer 2 device. 2411 * 2412 * All values for *flags* are reserved for future usage, and must 2413 * be left at zero. 2414 * 2415 * A call to this helper is susceptible to change the underlying 2416 * packet buffer. Therefore, at load time, all checks on pointers 2417 * previously done by the verifier are invalidated and must be 2418 * performed again, if the helper is used in combination with 2419 * direct packet access. 2420 * Return 2421 * 0 on success, or a negative error in case of failure. 2422 * 2423 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2424 * Description 2425 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2426 * it is possible to use a negative value for *delta*. This helper 2427 * can be used to prepare the packet for pushing or popping 2428 * headers. 2429 * 2430 * A call to this helper is susceptible to change the underlying 2431 * packet buffer. Therefore, at load time, all checks on pointers 2432 * previously done by the verifier are invalidated and must be 2433 * performed again, if the helper is used in combination with 2434 * direct packet access. 2435 * Return 2436 * 0 on success, or a negative error in case of failure. 2437 * 2438 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2439 * Description 2440 * Copy a NUL terminated string from an unsafe kernel address 2441 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2442 * more details. 2443 * 2444 * Generally, use **bpf_probe_read_user_str**\ () or 2445 * **bpf_probe_read_kernel_str**\ () instead. 2446 * Return 2447 * On success, the strictly positive length of the string, 2448 * including the trailing NUL character. On error, a negative 2449 * value. 2450 * 2451 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2452 * Description 2453 * If the **struct sk_buff** pointed by *skb* has a known socket, 2454 * retrieve the cookie (generated by the kernel) of this socket. 2455 * If no cookie has been set yet, generate a new cookie. Once 2456 * generated, the socket cookie remains stable for the life of the 2457 * socket. This helper can be useful for monitoring per socket 2458 * networking traffic statistics as it provides a global socket 2459 * identifier that can be assumed unique. 2460 * Return 2461 * A 8-byte long unique number on success, or 0 if the socket 2462 * field is missing inside *skb*. 2463 * 2464 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2465 * Description 2466 * Equivalent to bpf_get_socket_cookie() helper that accepts 2467 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2468 * Return 2469 * A 8-byte long unique number. 2470 * 2471 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2472 * Description 2473 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2474 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2475 * Return 2476 * A 8-byte long unique number. 2477 * 2478 * u64 bpf_get_socket_cookie(struct sock *sk) 2479 * Description 2480 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2481 * *sk*, but gets socket from a BTF **struct sock**. This helper 2482 * also works for sleepable programs. 2483 * Return 2484 * A 8-byte long unique number or 0 if *sk* is NULL. 2485 * 2486 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2487 * Description 2488 * Get the owner UID of the socked associated to *skb*. 2489 * Return 2490 * The owner UID of the socket associated to *skb*. If the socket 2491 * is **NULL**, or if it is not a full socket (i.e. if it is a 2492 * time-wait or a request socket instead), **overflowuid** value 2493 * is returned (note that **overflowuid** might also be the actual 2494 * UID value for the socket). 2495 * 2496 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2497 * Description 2498 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2499 * to value *hash*. 2500 * Return 2501 * 0 2502 * 2503 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2504 * Description 2505 * Emulate a call to **setsockopt()** on the socket associated to 2506 * *bpf_socket*, which must be a full socket. The *level* at 2507 * which the option resides and the name *optname* of the option 2508 * must be specified, see **setsockopt(2)** for more information. 2509 * The option value of length *optlen* is pointed by *optval*. 2510 * 2511 * *bpf_socket* should be one of the following: 2512 * 2513 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2514 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2515 * and **BPF_CGROUP_INET6_CONNECT**. 2516 * 2517 * This helper actually implements a subset of **setsockopt()**. 2518 * It supports the following *level*\ s: 2519 * 2520 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2521 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2522 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2523 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 2524 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2525 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2526 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2527 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2528 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 2529 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2530 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2531 * Return 2532 * 0 on success, or a negative error in case of failure. 2533 * 2534 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2535 * Description 2536 * Grow or shrink the room for data in the packet associated to 2537 * *skb* by *len_diff*, and according to the selected *mode*. 2538 * 2539 * By default, the helper will reset any offloaded checksum 2540 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2541 * by the following flag: 2542 * 2543 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2544 * checksum data of the skb to CHECKSUM_NONE. 2545 * 2546 * There are two supported modes at this time: 2547 * 2548 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2549 * (room space is added or removed below the layer 2 header). 2550 * 2551 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2552 * (room space is added or removed below the layer 3 header). 2553 * 2554 * The following flags are supported at this time: 2555 * 2556 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2557 * Adjusting mss in this way is not allowed for datagrams. 2558 * 2559 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2560 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2561 * Any new space is reserved to hold a tunnel header. 2562 * Configure skb offsets and other fields accordingly. 2563 * 2564 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2565 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2566 * Use with ENCAP_L3 flags to further specify the tunnel type. 2567 * 2568 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2569 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2570 * type; *len* is the length of the inner MAC header. 2571 * 2572 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2573 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2574 * L2 type as Ethernet. 2575 * 2576 * A call to this helper is susceptible to change the underlying 2577 * packet buffer. Therefore, at load time, all checks on pointers 2578 * previously done by the verifier are invalidated and must be 2579 * performed again, if the helper is used in combination with 2580 * direct packet access. 2581 * Return 2582 * 0 on success, or a negative error in case of failure. 2583 * 2584 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2585 * Description 2586 * Redirect the packet to the endpoint referenced by *map* at 2587 * index *key*. Depending on its type, this *map* can contain 2588 * references to net devices (for forwarding packets through other 2589 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2590 * but this is only implemented for native XDP (with driver 2591 * support) as of this writing). 2592 * 2593 * The lower two bits of *flags* are used as the return code if 2594 * the map lookup fails. This is so that the return value can be 2595 * one of the XDP program return codes up to **XDP_TX**, as chosen 2596 * by the caller. The higher bits of *flags* can be set to 2597 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2598 * 2599 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2600 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2601 * interface will be excluded when do broadcasting. 2602 * 2603 * See also **bpf_redirect**\ (), which only supports redirecting 2604 * to an ifindex, but doesn't require a map to do so. 2605 * Return 2606 * **XDP_REDIRECT** on success, or the value of the two lower bits 2607 * of the *flags* argument on error. 2608 * 2609 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2610 * Description 2611 * Redirect the packet to the socket referenced by *map* (of type 2612 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2613 * egress interfaces can be used for redirection. The 2614 * **BPF_F_INGRESS** value in *flags* is used to make the 2615 * distinction (ingress path is selected if the flag is present, 2616 * egress path otherwise). This is the only flag supported for now. 2617 * Return 2618 * **SK_PASS** on success, or **SK_DROP** on error. 2619 * 2620 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2621 * Description 2622 * Add an entry to, or update a *map* referencing sockets. The 2623 * *skops* is used as a new value for the entry associated to 2624 * *key*. *flags* is one of: 2625 * 2626 * **BPF_NOEXIST** 2627 * The entry for *key* must not exist in the map. 2628 * **BPF_EXIST** 2629 * The entry for *key* must already exist in the map. 2630 * **BPF_ANY** 2631 * No condition on the existence of the entry for *key*. 2632 * 2633 * If the *map* has eBPF programs (parser and verdict), those will 2634 * be inherited by the socket being added. If the socket is 2635 * already attached to eBPF programs, this results in an error. 2636 * Return 2637 * 0 on success, or a negative error in case of failure. 2638 * 2639 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2640 * Description 2641 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2642 * *delta* (which can be positive or negative). Note that this 2643 * operation modifies the address stored in *xdp_md*\ **->data**, 2644 * so the latter must be loaded only after the helper has been 2645 * called. 2646 * 2647 * The use of *xdp_md*\ **->data_meta** is optional and programs 2648 * are not required to use it. The rationale is that when the 2649 * packet is processed with XDP (e.g. as DoS filter), it is 2650 * possible to push further meta data along with it before passing 2651 * to the stack, and to give the guarantee that an ingress eBPF 2652 * program attached as a TC classifier on the same device can pick 2653 * this up for further post-processing. Since TC works with socket 2654 * buffers, it remains possible to set from XDP the **mark** or 2655 * **priority** pointers, or other pointers for the socket buffer. 2656 * Having this scratch space generic and programmable allows for 2657 * more flexibility as the user is free to store whatever meta 2658 * data they need. 2659 * 2660 * A call to this helper is susceptible to change the underlying 2661 * packet buffer. Therefore, at load time, all checks on pointers 2662 * previously done by the verifier are invalidated and must be 2663 * performed again, if the helper is used in combination with 2664 * direct packet access. 2665 * Return 2666 * 0 on success, or a negative error in case of failure. 2667 * 2668 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2669 * Description 2670 * Read the value of a perf event counter, and store it into *buf* 2671 * of size *buf_size*. This helper relies on a *map* of type 2672 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2673 * counter is selected when *map* is updated with perf event file 2674 * descriptors. The *map* is an array whose size is the number of 2675 * available CPUs, and each cell contains a value relative to one 2676 * CPU. The value to retrieve is indicated by *flags*, that 2677 * contains the index of the CPU to look up, masked with 2678 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2679 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2680 * current CPU should be retrieved. 2681 * 2682 * This helper behaves in a way close to 2683 * **bpf_perf_event_read**\ () helper, save that instead of 2684 * just returning the value observed, it fills the *buf* 2685 * structure. This allows for additional data to be retrieved: in 2686 * particular, the enabled and running times (in *buf*\ 2687 * **->enabled** and *buf*\ **->running**, respectively) are 2688 * copied. In general, **bpf_perf_event_read_value**\ () is 2689 * recommended over **bpf_perf_event_read**\ (), which has some 2690 * ABI issues and provides fewer functionalities. 2691 * 2692 * These values are interesting, because hardware PMU (Performance 2693 * Monitoring Unit) counters are limited resources. When there are 2694 * more PMU based perf events opened than available counters, 2695 * kernel will multiplex these events so each event gets certain 2696 * percentage (but not all) of the PMU time. In case that 2697 * multiplexing happens, the number of samples or counter value 2698 * will not reflect the case compared to when no multiplexing 2699 * occurs. This makes comparison between different runs difficult. 2700 * Typically, the counter value should be normalized before 2701 * comparing to other experiments. The usual normalization is done 2702 * as follows. 2703 * 2704 * :: 2705 * 2706 * normalized_counter = counter * t_enabled / t_running 2707 * 2708 * Where t_enabled is the time enabled for event and t_running is 2709 * the time running for event since last normalization. The 2710 * enabled and running times are accumulated since the perf event 2711 * open. To achieve scaling factor between two invocations of an 2712 * eBPF program, users can use CPU id as the key (which is 2713 * typical for perf array usage model) to remember the previous 2714 * value and do the calculation inside the eBPF program. 2715 * Return 2716 * 0 on success, or a negative error in case of failure. 2717 * 2718 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2719 * Description 2720 * For en eBPF program attached to a perf event, retrieve the 2721 * value of the event counter associated to *ctx* and store it in 2722 * the structure pointed by *buf* and of size *buf_size*. Enabled 2723 * and running times are also stored in the structure (see 2724 * description of helper **bpf_perf_event_read_value**\ () for 2725 * more details). 2726 * Return 2727 * 0 on success, or a negative error in case of failure. 2728 * 2729 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2730 * Description 2731 * Emulate a call to **getsockopt()** on the socket associated to 2732 * *bpf_socket*, which must be a full socket. The *level* at 2733 * which the option resides and the name *optname* of the option 2734 * must be specified, see **getsockopt(2)** for more information. 2735 * The retrieved value is stored in the structure pointed by 2736 * *opval* and of length *optlen*. 2737 * 2738 * *bpf_socket* should be one of the following: 2739 * 2740 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2741 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2742 * and **BPF_CGROUP_INET6_CONNECT**. 2743 * 2744 * This helper actually implements a subset of **getsockopt()**. 2745 * It supports the following *level*\ s: 2746 * 2747 * * **IPPROTO_TCP**, which supports *optname* 2748 * **TCP_CONGESTION**. 2749 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2750 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2751 * Return 2752 * 0 on success, or a negative error in case of failure. 2753 * 2754 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2755 * Description 2756 * Used for error injection, this helper uses kprobes to override 2757 * the return value of the probed function, and to set it to *rc*. 2758 * The first argument is the context *regs* on which the kprobe 2759 * works. 2760 * 2761 * This helper works by setting the PC (program counter) 2762 * to an override function which is run in place of the original 2763 * probed function. This means the probed function is not run at 2764 * all. The replacement function just returns with the required 2765 * value. 2766 * 2767 * This helper has security implications, and thus is subject to 2768 * restrictions. It is only available if the kernel was compiled 2769 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2770 * option, and in this case it only works on functions tagged with 2771 * **ALLOW_ERROR_INJECTION** in the kernel code. 2772 * 2773 * Also, the helper is only available for the architectures having 2774 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2775 * x86 architecture is the only one to support this feature. 2776 * Return 2777 * 0 2778 * 2779 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2780 * Description 2781 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2782 * for the full TCP socket associated to *bpf_sock_ops* to 2783 * *argval*. 2784 * 2785 * The primary use of this field is to determine if there should 2786 * be calls to eBPF programs of type 2787 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2788 * code. A program of the same type can change its value, per 2789 * connection and as necessary, when the connection is 2790 * established. This field is directly accessible for reading, but 2791 * this helper must be used for updates in order to return an 2792 * error if an eBPF program tries to set a callback that is not 2793 * supported in the current kernel. 2794 * 2795 * *argval* is a flag array which can combine these flags: 2796 * 2797 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2798 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2799 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2800 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2801 * 2802 * Therefore, this function can be used to clear a callback flag by 2803 * setting the appropriate bit to zero. e.g. to disable the RTO 2804 * callback: 2805 * 2806 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2807 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2808 * 2809 * Here are some examples of where one could call such eBPF 2810 * program: 2811 * 2812 * * When RTO fires. 2813 * * When a packet is retransmitted. 2814 * * When the connection terminates. 2815 * * When a packet is sent. 2816 * * When a packet is received. 2817 * Return 2818 * Code **-EINVAL** if the socket is not a full TCP socket; 2819 * otherwise, a positive number containing the bits that could not 2820 * be set is returned (which comes down to 0 if all bits were set 2821 * as required). 2822 * 2823 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2824 * Description 2825 * This helper is used in programs implementing policies at the 2826 * socket level. If the message *msg* is allowed to pass (i.e. if 2827 * the verdict eBPF program returns **SK_PASS**), redirect it to 2828 * the socket referenced by *map* (of type 2829 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2830 * egress interfaces can be used for redirection. The 2831 * **BPF_F_INGRESS** value in *flags* is used to make the 2832 * distinction (ingress path is selected if the flag is present, 2833 * egress path otherwise). This is the only flag supported for now. 2834 * Return 2835 * **SK_PASS** on success, or **SK_DROP** on error. 2836 * 2837 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2838 * Description 2839 * For socket policies, apply the verdict of the eBPF program to 2840 * the next *bytes* (number of bytes) of message *msg*. 2841 * 2842 * For example, this helper can be used in the following cases: 2843 * 2844 * * A single **sendmsg**\ () or **sendfile**\ () system call 2845 * contains multiple logical messages that the eBPF program is 2846 * supposed to read and for which it should apply a verdict. 2847 * * An eBPF program only cares to read the first *bytes* of a 2848 * *msg*. If the message has a large payload, then setting up 2849 * and calling the eBPF program repeatedly for all bytes, even 2850 * though the verdict is already known, would create unnecessary 2851 * overhead. 2852 * 2853 * When called from within an eBPF program, the helper sets a 2854 * counter internal to the BPF infrastructure, that is used to 2855 * apply the last verdict to the next *bytes*. If *bytes* is 2856 * smaller than the current data being processed from a 2857 * **sendmsg**\ () or **sendfile**\ () system call, the first 2858 * *bytes* will be sent and the eBPF program will be re-run with 2859 * the pointer for start of data pointing to byte number *bytes* 2860 * **+ 1**. If *bytes* is larger than the current data being 2861 * processed, then the eBPF verdict will be applied to multiple 2862 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2863 * consumed. 2864 * 2865 * Note that if a socket closes with the internal counter holding 2866 * a non-zero value, this is not a problem because data is not 2867 * being buffered for *bytes* and is sent as it is received. 2868 * Return 2869 * 0 2870 * 2871 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2872 * Description 2873 * For socket policies, prevent the execution of the verdict eBPF 2874 * program for message *msg* until *bytes* (byte number) have been 2875 * accumulated. 2876 * 2877 * This can be used when one needs a specific number of bytes 2878 * before a verdict can be assigned, even if the data spans 2879 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2880 * case would be a user calling **sendmsg**\ () repeatedly with 2881 * 1-byte long message segments. Obviously, this is bad for 2882 * performance, but it is still valid. If the eBPF program needs 2883 * *bytes* bytes to validate a header, this helper can be used to 2884 * prevent the eBPF program to be called again until *bytes* have 2885 * been accumulated. 2886 * Return 2887 * 0 2888 * 2889 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2890 * Description 2891 * For socket policies, pull in non-linear data from user space 2892 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2893 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2894 * respectively. 2895 * 2896 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2897 * *msg* it can only parse data that the (**data**, **data_end**) 2898 * pointers have already consumed. For **sendmsg**\ () hooks this 2899 * is likely the first scatterlist element. But for calls relying 2900 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2901 * be the range (**0**, **0**) because the data is shared with 2902 * user space and by default the objective is to avoid allowing 2903 * user space to modify data while (or after) eBPF verdict is 2904 * being decided. This helper can be used to pull in data and to 2905 * set the start and end pointer to given values. Data will be 2906 * copied if necessary (i.e. if data was not linear and if start 2907 * and end pointers do not point to the same chunk). 2908 * 2909 * A call to this helper is susceptible to change the underlying 2910 * packet buffer. Therefore, at load time, all checks on pointers 2911 * previously done by the verifier are invalidated and must be 2912 * performed again, if the helper is used in combination with 2913 * direct packet access. 2914 * 2915 * All values for *flags* are reserved for future usage, and must 2916 * be left at zero. 2917 * Return 2918 * 0 on success, or a negative error in case of failure. 2919 * 2920 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2921 * Description 2922 * Bind the socket associated to *ctx* to the address pointed by 2923 * *addr*, of length *addr_len*. This allows for making outgoing 2924 * connection from the desired IP address, which can be useful for 2925 * example when all processes inside a cgroup should use one 2926 * single IP address on a host that has multiple IP configured. 2927 * 2928 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2929 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2930 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2931 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2932 * behavior and lets the kernel efficiently pick up an unused 2933 * port as long as 4-tuple is unique. Passing non-zero port might 2934 * lead to degraded performance. 2935 * Return 2936 * 0 on success, or a negative error in case of failure. 2937 * 2938 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2939 * Description 2940 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2941 * possible to both shrink and grow the packet tail. 2942 * Shrink done via *delta* being a negative integer. 2943 * 2944 * A call to this helper is susceptible to change the underlying 2945 * packet buffer. Therefore, at load time, all checks on pointers 2946 * previously done by the verifier are invalidated and must be 2947 * performed again, if the helper is used in combination with 2948 * direct packet access. 2949 * Return 2950 * 0 on success, or a negative error in case of failure. 2951 * 2952 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2953 * Description 2954 * Retrieve the XFRM state (IP transform framework, see also 2955 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2956 * 2957 * The retrieved value is stored in the **struct bpf_xfrm_state** 2958 * pointed by *xfrm_state* and of length *size*. 2959 * 2960 * All values for *flags* are reserved for future usage, and must 2961 * be left at zero. 2962 * 2963 * This helper is available only if the kernel was compiled with 2964 * **CONFIG_XFRM** configuration option. 2965 * Return 2966 * 0 on success, or a negative error in case of failure. 2967 * 2968 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2969 * Description 2970 * Return a user or a kernel stack in bpf program provided buffer. 2971 * To achieve this, the helper needs *ctx*, which is a pointer 2972 * to the context on which the tracing program is executed. 2973 * To store the stacktrace, the bpf program provides *buf* with 2974 * a nonnegative *size*. 2975 * 2976 * The last argument, *flags*, holds the number of stack frames to 2977 * skip (from 0 to 255), masked with 2978 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2979 * the following flags: 2980 * 2981 * **BPF_F_USER_STACK** 2982 * Collect a user space stack instead of a kernel stack. 2983 * **BPF_F_USER_BUILD_ID** 2984 * Collect buildid+offset instead of ips for user stack, 2985 * only valid if **BPF_F_USER_STACK** is also specified. 2986 * 2987 * **bpf_get_stack**\ () can collect up to 2988 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2989 * to sufficient large buffer size. Note that 2990 * this limit can be controlled with the **sysctl** program, and 2991 * that it should be manually increased in order to profile long 2992 * user stacks (such as stacks for Java programs). To do so, use: 2993 * 2994 * :: 2995 * 2996 * # sysctl kernel.perf_event_max_stack=<new value> 2997 * Return 2998 * A non-negative value equal to or less than *size* on success, 2999 * or a negative error in case of failure. 3000 * 3001 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3002 * Description 3003 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3004 * it provides an easy way to load *len* bytes from *offset* 3005 * from the packet associated to *skb*, into the buffer pointed 3006 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3007 * a fifth argument *start_header* exists in order to select a 3008 * base offset to start from. *start_header* can be one of: 3009 * 3010 * **BPF_HDR_START_MAC** 3011 * Base offset to load data from is *skb*'s mac header. 3012 * **BPF_HDR_START_NET** 3013 * Base offset to load data from is *skb*'s network header. 3014 * 3015 * In general, "direct packet access" is the preferred method to 3016 * access packet data, however, this helper is in particular useful 3017 * in socket filters where *skb*\ **->data** does not always point 3018 * to the start of the mac header and where "direct packet access" 3019 * is not available. 3020 * Return 3021 * 0 on success, or a negative error in case of failure. 3022 * 3023 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3024 * Description 3025 * Do FIB lookup in kernel tables using parameters in *params*. 3026 * If lookup is successful and result shows packet is to be 3027 * forwarded, the neighbor tables are searched for the nexthop. 3028 * If successful (ie., FIB lookup shows forwarding and nexthop 3029 * is resolved), the nexthop address is returned in ipv4_dst 3030 * or ipv6_dst based on family, smac is set to mac address of 3031 * egress device, dmac is set to nexthop mac address, rt_metric 3032 * is set to metric from route (IPv4/IPv6 only), and ifindex 3033 * is set to the device index of the nexthop from the FIB lookup. 3034 * 3035 * *plen* argument is the size of the passed in struct. 3036 * *flags* argument can be a combination of one or more of the 3037 * following values: 3038 * 3039 * **BPF_FIB_LOOKUP_DIRECT** 3040 * Do a direct table lookup vs full lookup using FIB 3041 * rules. 3042 * **BPF_FIB_LOOKUP_OUTPUT** 3043 * Perform lookup from an egress perspective (default is 3044 * ingress). 3045 * 3046 * *ctx* is either **struct xdp_md** for XDP programs or 3047 * **struct sk_buff** tc cls_act programs. 3048 * Return 3049 * * < 0 if any input argument is invalid 3050 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3051 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3052 * packet is not forwarded or needs assist from full stack 3053 * 3054 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3055 * was exceeded and output params->mtu_result contains the MTU. 3056 * 3057 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3058 * Description 3059 * Add an entry to, or update a sockhash *map* referencing sockets. 3060 * The *skops* is used as a new value for the entry associated to 3061 * *key*. *flags* is one of: 3062 * 3063 * **BPF_NOEXIST** 3064 * The entry for *key* must not exist in the map. 3065 * **BPF_EXIST** 3066 * The entry for *key* must already exist in the map. 3067 * **BPF_ANY** 3068 * No condition on the existence of the entry for *key*. 3069 * 3070 * If the *map* has eBPF programs (parser and verdict), those will 3071 * be inherited by the socket being added. If the socket is 3072 * already attached to eBPF programs, this results in an error. 3073 * Return 3074 * 0 on success, or a negative error in case of failure. 3075 * 3076 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3077 * Description 3078 * This helper is used in programs implementing policies at the 3079 * socket level. If the message *msg* is allowed to pass (i.e. if 3080 * the verdict eBPF program returns **SK_PASS**), redirect it to 3081 * the socket referenced by *map* (of type 3082 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3083 * egress interfaces can be used for redirection. The 3084 * **BPF_F_INGRESS** value in *flags* is used to make the 3085 * distinction (ingress path is selected if the flag is present, 3086 * egress path otherwise). This is the only flag supported for now. 3087 * Return 3088 * **SK_PASS** on success, or **SK_DROP** on error. 3089 * 3090 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3091 * Description 3092 * This helper is used in programs implementing policies at the 3093 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3094 * if the verdict eBPF program returns **SK_PASS**), redirect it 3095 * to the socket referenced by *map* (of type 3096 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3097 * egress interfaces can be used for redirection. The 3098 * **BPF_F_INGRESS** value in *flags* is used to make the 3099 * distinction (ingress path is selected if the flag is present, 3100 * egress otherwise). This is the only flag supported for now. 3101 * Return 3102 * **SK_PASS** on success, or **SK_DROP** on error. 3103 * 3104 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3105 * Description 3106 * Encapsulate the packet associated to *skb* within a Layer 3 3107 * protocol header. This header is provided in the buffer at 3108 * address *hdr*, with *len* its size in bytes. *type* indicates 3109 * the protocol of the header and can be one of: 3110 * 3111 * **BPF_LWT_ENCAP_SEG6** 3112 * IPv6 encapsulation with Segment Routing Header 3113 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3114 * the IPv6 header is computed by the kernel. 3115 * **BPF_LWT_ENCAP_SEG6_INLINE** 3116 * Only works if *skb* contains an IPv6 packet. Insert a 3117 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3118 * the IPv6 header. 3119 * **BPF_LWT_ENCAP_IP** 3120 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3121 * must be IPv4 or IPv6, followed by zero or more 3122 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3123 * total bytes in all prepended headers. Please note that 3124 * if **skb_is_gso**\ (*skb*) is true, no more than two 3125 * headers can be prepended, and the inner header, if 3126 * present, should be either GRE or UDP/GUE. 3127 * 3128 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3129 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3130 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3131 * **BPF_PROG_TYPE_LWT_XMIT**. 3132 * 3133 * A call to this helper is susceptible to change the underlying 3134 * packet buffer. Therefore, at load time, all checks on pointers 3135 * previously done by the verifier are invalidated and must be 3136 * performed again, if the helper is used in combination with 3137 * direct packet access. 3138 * Return 3139 * 0 on success, or a negative error in case of failure. 3140 * 3141 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3142 * Description 3143 * Store *len* bytes from address *from* into the packet 3144 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3145 * inside the outermost IPv6 Segment Routing Header can be 3146 * modified through this helper. 3147 * 3148 * A call to this helper is susceptible to change the underlying 3149 * packet buffer. Therefore, at load time, all checks on pointers 3150 * previously done by the verifier are invalidated and must be 3151 * performed again, if the helper is used in combination with 3152 * direct packet access. 3153 * Return 3154 * 0 on success, or a negative error in case of failure. 3155 * 3156 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3157 * Description 3158 * Adjust the size allocated to TLVs in the outermost IPv6 3159 * Segment Routing Header contained in the packet associated to 3160 * *skb*, at position *offset* by *delta* bytes. Only offsets 3161 * after the segments are accepted. *delta* can be as well 3162 * positive (growing) as negative (shrinking). 3163 * 3164 * A call to this helper is susceptible to change the underlying 3165 * packet buffer. Therefore, at load time, all checks on pointers 3166 * previously done by the verifier are invalidated and must be 3167 * performed again, if the helper is used in combination with 3168 * direct packet access. 3169 * Return 3170 * 0 on success, or a negative error in case of failure. 3171 * 3172 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3173 * Description 3174 * Apply an IPv6 Segment Routing action of type *action* to the 3175 * packet associated to *skb*. Each action takes a parameter 3176 * contained at address *param*, and of length *param_len* bytes. 3177 * *action* can be one of: 3178 * 3179 * **SEG6_LOCAL_ACTION_END_X** 3180 * End.X action: Endpoint with Layer-3 cross-connect. 3181 * Type of *param*: **struct in6_addr**. 3182 * **SEG6_LOCAL_ACTION_END_T** 3183 * End.T action: Endpoint with specific IPv6 table lookup. 3184 * Type of *param*: **int**. 3185 * **SEG6_LOCAL_ACTION_END_B6** 3186 * End.B6 action: Endpoint bound to an SRv6 policy. 3187 * Type of *param*: **struct ipv6_sr_hdr**. 3188 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3189 * End.B6.Encap action: Endpoint bound to an SRv6 3190 * encapsulation policy. 3191 * Type of *param*: **struct ipv6_sr_hdr**. 3192 * 3193 * A call to this helper is susceptible to change the underlying 3194 * packet buffer. Therefore, at load time, all checks on pointers 3195 * previously done by the verifier are invalidated and must be 3196 * performed again, if the helper is used in combination with 3197 * direct packet access. 3198 * Return 3199 * 0 on success, or a negative error in case of failure. 3200 * 3201 * long bpf_rc_repeat(void *ctx) 3202 * Description 3203 * This helper is used in programs implementing IR decoding, to 3204 * report a successfully decoded repeat key message. This delays 3205 * the generation of a key up event for previously generated 3206 * key down event. 3207 * 3208 * Some IR protocols like NEC have a special IR message for 3209 * repeating last button, for when a button is held down. 3210 * 3211 * The *ctx* should point to the lirc sample as passed into 3212 * the program. 3213 * 3214 * This helper is only available is the kernel was compiled with 3215 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3216 * "**y**". 3217 * Return 3218 * 0 3219 * 3220 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3221 * Description 3222 * This helper is used in programs implementing IR decoding, to 3223 * report a successfully decoded key press with *scancode*, 3224 * *toggle* value in the given *protocol*. The scancode will be 3225 * translated to a keycode using the rc keymap, and reported as 3226 * an input key down event. After a period a key up event is 3227 * generated. This period can be extended by calling either 3228 * **bpf_rc_keydown**\ () again with the same values, or calling 3229 * **bpf_rc_repeat**\ (). 3230 * 3231 * Some protocols include a toggle bit, in case the button was 3232 * released and pressed again between consecutive scancodes. 3233 * 3234 * The *ctx* should point to the lirc sample as passed into 3235 * the program. 3236 * 3237 * The *protocol* is the decoded protocol number (see 3238 * **enum rc_proto** for some predefined values). 3239 * 3240 * This helper is only available is the kernel was compiled with 3241 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3242 * "**y**". 3243 * Return 3244 * 0 3245 * 3246 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3247 * Description 3248 * Return the cgroup v2 id of the socket associated with the *skb*. 3249 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3250 * helper for cgroup v1 by providing a tag resp. identifier that 3251 * can be matched on or used for map lookups e.g. to implement 3252 * policy. The cgroup v2 id of a given path in the hierarchy is 3253 * exposed in user space through the f_handle API in order to get 3254 * to the same 64-bit id. 3255 * 3256 * This helper can be used on TC egress path, but not on ingress, 3257 * and is available only if the kernel was compiled with the 3258 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3259 * Return 3260 * The id is returned or 0 in case the id could not be retrieved. 3261 * 3262 * u64 bpf_get_current_cgroup_id(void) 3263 * Description 3264 * Get the current cgroup id based on the cgroup within which 3265 * the current task is running. 3266 * Return 3267 * A 64-bit integer containing the current cgroup id based 3268 * on the cgroup within which the current task is running. 3269 * 3270 * void *bpf_get_local_storage(void *map, u64 flags) 3271 * Description 3272 * Get the pointer to the local storage area. 3273 * The type and the size of the local storage is defined 3274 * by the *map* argument. 3275 * The *flags* meaning is specific for each map type, 3276 * and has to be 0 for cgroup local storage. 3277 * 3278 * Depending on the BPF program type, a local storage area 3279 * can be shared between multiple instances of the BPF program, 3280 * running simultaneously. 3281 * 3282 * A user should care about the synchronization by himself. 3283 * For example, by using the **BPF_ATOMIC** instructions to alter 3284 * the shared data. 3285 * Return 3286 * A pointer to the local storage area. 3287 * 3288 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3289 * Description 3290 * Select a **SO_REUSEPORT** socket from a 3291 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3292 * It checks the selected socket is matching the incoming 3293 * request in the socket buffer. 3294 * Return 3295 * 0 on success, or a negative error in case of failure. 3296 * 3297 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3298 * Description 3299 * Return id of cgroup v2 that is ancestor of cgroup associated 3300 * with the *skb* at the *ancestor_level*. The root cgroup is at 3301 * *ancestor_level* zero and each step down the hierarchy 3302 * increments the level. If *ancestor_level* == level of cgroup 3303 * associated with *skb*, then return value will be same as that 3304 * of **bpf_skb_cgroup_id**\ (). 3305 * 3306 * The helper is useful to implement policies based on cgroups 3307 * that are upper in hierarchy than immediate cgroup associated 3308 * with *skb*. 3309 * 3310 * The format of returned id and helper limitations are same as in 3311 * **bpf_skb_cgroup_id**\ (). 3312 * Return 3313 * The id is returned or 0 in case the id could not be retrieved. 3314 * 3315 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3316 * Description 3317 * Look for TCP socket matching *tuple*, optionally in a child 3318 * network namespace *netns*. The return value must be checked, 3319 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3320 * 3321 * The *ctx* should point to the context of the program, such as 3322 * the skb or socket (depending on the hook in use). This is used 3323 * to determine the base network namespace for the lookup. 3324 * 3325 * *tuple_size* must be one of: 3326 * 3327 * **sizeof**\ (*tuple*\ **->ipv4**) 3328 * Look for an IPv4 socket. 3329 * **sizeof**\ (*tuple*\ **->ipv6**) 3330 * Look for an IPv6 socket. 3331 * 3332 * If the *netns* is a negative signed 32-bit integer, then the 3333 * socket lookup table in the netns associated with the *ctx* 3334 * will be used. For the TC hooks, this is the netns of the device 3335 * in the skb. For socket hooks, this is the netns of the socket. 3336 * If *netns* is any other signed 32-bit value greater than or 3337 * equal to zero then it specifies the ID of the netns relative to 3338 * the netns associated with the *ctx*. *netns* values beyond the 3339 * range of 32-bit integers are reserved for future use. 3340 * 3341 * All values for *flags* are reserved for future usage, and must 3342 * be left at zero. 3343 * 3344 * This helper is available only if the kernel was compiled with 3345 * **CONFIG_NET** configuration option. 3346 * Return 3347 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3348 * For sockets with reuseport option, the **struct bpf_sock** 3349 * result is from *reuse*\ **->socks**\ [] using the hash of the 3350 * tuple. 3351 * 3352 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3353 * Description 3354 * Look for UDP socket matching *tuple*, optionally in a child 3355 * network namespace *netns*. The return value must be checked, 3356 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3357 * 3358 * The *ctx* should point to the context of the program, such as 3359 * the skb or socket (depending on the hook in use). This is used 3360 * to determine the base network namespace for the lookup. 3361 * 3362 * *tuple_size* must be one of: 3363 * 3364 * **sizeof**\ (*tuple*\ **->ipv4**) 3365 * Look for an IPv4 socket. 3366 * **sizeof**\ (*tuple*\ **->ipv6**) 3367 * Look for an IPv6 socket. 3368 * 3369 * If the *netns* is a negative signed 32-bit integer, then the 3370 * socket lookup table in the netns associated with the *ctx* 3371 * will be used. For the TC hooks, this is the netns of the device 3372 * in the skb. For socket hooks, this is the netns of the socket. 3373 * If *netns* is any other signed 32-bit value greater than or 3374 * equal to zero then it specifies the ID of the netns relative to 3375 * the netns associated with the *ctx*. *netns* values beyond the 3376 * range of 32-bit integers are reserved for future use. 3377 * 3378 * All values for *flags* are reserved for future usage, and must 3379 * be left at zero. 3380 * 3381 * This helper is available only if the kernel was compiled with 3382 * **CONFIG_NET** configuration option. 3383 * Return 3384 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3385 * For sockets with reuseport option, the **struct bpf_sock** 3386 * result is from *reuse*\ **->socks**\ [] using the hash of the 3387 * tuple. 3388 * 3389 * long bpf_sk_release(void *sock) 3390 * Description 3391 * Release the reference held by *sock*. *sock* must be a 3392 * non-**NULL** pointer that was returned from 3393 * **bpf_sk_lookup_xxx**\ (). 3394 * Return 3395 * 0 on success, or a negative error in case of failure. 3396 * 3397 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3398 * Description 3399 * Push an element *value* in *map*. *flags* is one of: 3400 * 3401 * **BPF_EXIST** 3402 * If the queue/stack is full, the oldest element is 3403 * removed to make room for this. 3404 * Return 3405 * 0 on success, or a negative error in case of failure. 3406 * 3407 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3408 * Description 3409 * Pop an element from *map*. 3410 * Return 3411 * 0 on success, or a negative error in case of failure. 3412 * 3413 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3414 * Description 3415 * Get an element from *map* without removing it. 3416 * Return 3417 * 0 on success, or a negative error in case of failure. 3418 * 3419 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3420 * Description 3421 * For socket policies, insert *len* bytes into *msg* at offset 3422 * *start*. 3423 * 3424 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3425 * *msg* it may want to insert metadata or options into the *msg*. 3426 * This can later be read and used by any of the lower layer BPF 3427 * hooks. 3428 * 3429 * This helper may fail if under memory pressure (a malloc 3430 * fails) in these cases BPF programs will get an appropriate 3431 * error and BPF programs will need to handle them. 3432 * Return 3433 * 0 on success, or a negative error in case of failure. 3434 * 3435 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3436 * Description 3437 * Will remove *len* bytes from a *msg* starting at byte *start*. 3438 * This may result in **ENOMEM** errors under certain situations if 3439 * an allocation and copy are required due to a full ring buffer. 3440 * However, the helper will try to avoid doing the allocation 3441 * if possible. Other errors can occur if input parameters are 3442 * invalid either due to *start* byte not being valid part of *msg* 3443 * payload and/or *pop* value being to large. 3444 * Return 3445 * 0 on success, or a negative error in case of failure. 3446 * 3447 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3448 * Description 3449 * This helper is used in programs implementing IR decoding, to 3450 * report a successfully decoded pointer movement. 3451 * 3452 * The *ctx* should point to the lirc sample as passed into 3453 * the program. 3454 * 3455 * This helper is only available is the kernel was compiled with 3456 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3457 * "**y**". 3458 * Return 3459 * 0 3460 * 3461 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3462 * Description 3463 * Acquire a spinlock represented by the pointer *lock*, which is 3464 * stored as part of a value of a map. Taking the lock allows to 3465 * safely update the rest of the fields in that value. The 3466 * spinlock can (and must) later be released with a call to 3467 * **bpf_spin_unlock**\ (\ *lock*\ ). 3468 * 3469 * Spinlocks in BPF programs come with a number of restrictions 3470 * and constraints: 3471 * 3472 * * **bpf_spin_lock** objects are only allowed inside maps of 3473 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3474 * list could be extended in the future). 3475 * * BTF description of the map is mandatory. 3476 * * The BPF program can take ONE lock at a time, since taking two 3477 * or more could cause dead locks. 3478 * * Only one **struct bpf_spin_lock** is allowed per map element. 3479 * * When the lock is taken, calls (either BPF to BPF or helpers) 3480 * are not allowed. 3481 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3482 * allowed inside a spinlock-ed region. 3483 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3484 * the lock, on all execution paths, before it returns. 3485 * * The BPF program can access **struct bpf_spin_lock** only via 3486 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3487 * helpers. Loading or storing data into the **struct 3488 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3489 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3490 * of the map value must be a struct and have **struct 3491 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3492 * Nested lock inside another struct is not allowed. 3493 * * The **struct bpf_spin_lock** *lock* field in a map value must 3494 * be aligned on a multiple of 4 bytes in that value. 3495 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3496 * the **bpf_spin_lock** field to user space. 3497 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3498 * a BPF program, do not update the **bpf_spin_lock** field. 3499 * * **bpf_spin_lock** cannot be on the stack or inside a 3500 * networking packet (it can only be inside of a map values). 3501 * * **bpf_spin_lock** is available to root only. 3502 * * Tracing programs and socket filter programs cannot use 3503 * **bpf_spin_lock**\ () due to insufficient preemption checks 3504 * (but this may change in the future). 3505 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3506 * Return 3507 * 0 3508 * 3509 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3510 * Description 3511 * Release the *lock* previously locked by a call to 3512 * **bpf_spin_lock**\ (\ *lock*\ ). 3513 * Return 3514 * 0 3515 * 3516 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3517 * Description 3518 * This helper gets a **struct bpf_sock** pointer such 3519 * that all the fields in this **bpf_sock** can be accessed. 3520 * Return 3521 * A **struct bpf_sock** pointer on success, or **NULL** in 3522 * case of failure. 3523 * 3524 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3525 * Description 3526 * This helper gets a **struct bpf_tcp_sock** pointer from a 3527 * **struct bpf_sock** pointer. 3528 * Return 3529 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3530 * case of failure. 3531 * 3532 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3533 * Description 3534 * Set ECN (Explicit Congestion Notification) field of IP header 3535 * to **CE** (Congestion Encountered) if current value is **ECT** 3536 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3537 * and IPv4. 3538 * Return 3539 * 1 if the **CE** flag is set (either by the current helper call 3540 * or because it was already present), 0 if it is not set. 3541 * 3542 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3543 * Description 3544 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3545 * **bpf_sk_release**\ () is unnecessary and not allowed. 3546 * Return 3547 * A **struct bpf_sock** pointer on success, or **NULL** in 3548 * case of failure. 3549 * 3550 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3551 * Description 3552 * Look for TCP socket matching *tuple*, optionally in a child 3553 * network namespace *netns*. The return value must be checked, 3554 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3555 * 3556 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3557 * that it also returns timewait or request sockets. Use 3558 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3559 * full structure. 3560 * 3561 * This helper is available only if the kernel was compiled with 3562 * **CONFIG_NET** configuration option. 3563 * Return 3564 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3565 * For sockets with reuseport option, the **struct bpf_sock** 3566 * result is from *reuse*\ **->socks**\ [] using the hash of the 3567 * tuple. 3568 * 3569 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3570 * Description 3571 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3572 * the listening socket in *sk*. 3573 * 3574 * *iph* points to the start of the IPv4 or IPv6 header, while 3575 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3576 * **sizeof**\ (**struct ip6hdr**). 3577 * 3578 * *th* points to the start of the TCP header, while *th_len* 3579 * contains **sizeof**\ (**struct tcphdr**). 3580 * Return 3581 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3582 * error otherwise. 3583 * 3584 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3585 * Description 3586 * Get name of sysctl in /proc/sys/ and copy it into provided by 3587 * program buffer *buf* of size *buf_len*. 3588 * 3589 * The buffer is always NUL terminated, unless it's zero-sized. 3590 * 3591 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3592 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3593 * only (e.g. "tcp_mem"). 3594 * Return 3595 * Number of character copied (not including the trailing NUL). 3596 * 3597 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3598 * truncated name in this case). 3599 * 3600 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3601 * Description 3602 * Get current value of sysctl as it is presented in /proc/sys 3603 * (incl. newline, etc), and copy it as a string into provided 3604 * by program buffer *buf* of size *buf_len*. 3605 * 3606 * The whole value is copied, no matter what file position user 3607 * space issued e.g. sys_read at. 3608 * 3609 * The buffer is always NUL terminated, unless it's zero-sized. 3610 * Return 3611 * Number of character copied (not including the trailing NUL). 3612 * 3613 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3614 * truncated name in this case). 3615 * 3616 * **-EINVAL** if current value was unavailable, e.g. because 3617 * sysctl is uninitialized and read returns -EIO for it. 3618 * 3619 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3620 * Description 3621 * Get new value being written by user space to sysctl (before 3622 * the actual write happens) and copy it as a string into 3623 * provided by program buffer *buf* of size *buf_len*. 3624 * 3625 * User space may write new value at file position > 0. 3626 * 3627 * The buffer is always NUL terminated, unless it's zero-sized. 3628 * Return 3629 * Number of character copied (not including the trailing NUL). 3630 * 3631 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3632 * truncated name in this case). 3633 * 3634 * **-EINVAL** if sysctl is being read. 3635 * 3636 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3637 * Description 3638 * Override new value being written by user space to sysctl with 3639 * value provided by program in buffer *buf* of size *buf_len*. 3640 * 3641 * *buf* should contain a string in same form as provided by user 3642 * space on sysctl write. 3643 * 3644 * User space may write new value at file position > 0. To override 3645 * the whole sysctl value file position should be set to zero. 3646 * Return 3647 * 0 on success. 3648 * 3649 * **-E2BIG** if the *buf_len* is too big. 3650 * 3651 * **-EINVAL** if sysctl is being read. 3652 * 3653 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3654 * Description 3655 * Convert the initial part of the string from buffer *buf* of 3656 * size *buf_len* to a long integer according to the given base 3657 * and save the result in *res*. 3658 * 3659 * The string may begin with an arbitrary amount of white space 3660 * (as determined by **isspace**\ (3)) followed by a single 3661 * optional '**-**' sign. 3662 * 3663 * Five least significant bits of *flags* encode base, other bits 3664 * are currently unused. 3665 * 3666 * Base must be either 8, 10, 16 or 0 to detect it automatically 3667 * similar to user space **strtol**\ (3). 3668 * Return 3669 * Number of characters consumed on success. Must be positive but 3670 * no more than *buf_len*. 3671 * 3672 * **-EINVAL** if no valid digits were found or unsupported base 3673 * was provided. 3674 * 3675 * **-ERANGE** if resulting value was out of range. 3676 * 3677 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3678 * Description 3679 * Convert the initial part of the string from buffer *buf* of 3680 * size *buf_len* to an unsigned long integer according to the 3681 * given base and save the result in *res*. 3682 * 3683 * The string may begin with an arbitrary amount of white space 3684 * (as determined by **isspace**\ (3)). 3685 * 3686 * Five least significant bits of *flags* encode base, other bits 3687 * are currently unused. 3688 * 3689 * Base must be either 8, 10, 16 or 0 to detect it automatically 3690 * similar to user space **strtoul**\ (3). 3691 * Return 3692 * Number of characters consumed on success. Must be positive but 3693 * no more than *buf_len*. 3694 * 3695 * **-EINVAL** if no valid digits were found or unsupported base 3696 * was provided. 3697 * 3698 * **-ERANGE** if resulting value was out of range. 3699 * 3700 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3701 * Description 3702 * Get a bpf-local-storage from a *sk*. 3703 * 3704 * Logically, it could be thought of getting the value from 3705 * a *map* with *sk* as the **key**. From this 3706 * perspective, the usage is not much different from 3707 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3708 * helper enforces the key must be a full socket and the map must 3709 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3710 * 3711 * Underneath, the value is stored locally at *sk* instead of 3712 * the *map*. The *map* is used as the bpf-local-storage 3713 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3714 * searched against all bpf-local-storages residing at *sk*. 3715 * 3716 * *sk* is a kernel **struct sock** pointer for LSM program. 3717 * *sk* is a **struct bpf_sock** pointer for other program types. 3718 * 3719 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3720 * used such that a new bpf-local-storage will be 3721 * created if one does not exist. *value* can be used 3722 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3723 * the initial value of a bpf-local-storage. If *value* is 3724 * **NULL**, the new bpf-local-storage will be zero initialized. 3725 * Return 3726 * A bpf-local-storage pointer is returned on success. 3727 * 3728 * **NULL** if not found or there was an error in adding 3729 * a new bpf-local-storage. 3730 * 3731 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3732 * Description 3733 * Delete a bpf-local-storage from a *sk*. 3734 * Return 3735 * 0 on success. 3736 * 3737 * **-ENOENT** if the bpf-local-storage cannot be found. 3738 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3739 * 3740 * long bpf_send_signal(u32 sig) 3741 * Description 3742 * Send signal *sig* to the process of the current task. 3743 * The signal may be delivered to any of this process's threads. 3744 * Return 3745 * 0 on success or successfully queued. 3746 * 3747 * **-EBUSY** if work queue under nmi is full. 3748 * 3749 * **-EINVAL** if *sig* is invalid. 3750 * 3751 * **-EPERM** if no permission to send the *sig*. 3752 * 3753 * **-EAGAIN** if bpf program can try again. 3754 * 3755 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3756 * Description 3757 * Try to issue a SYN cookie for the packet with corresponding 3758 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3759 * 3760 * *iph* points to the start of the IPv4 or IPv6 header, while 3761 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3762 * **sizeof**\ (**struct ip6hdr**). 3763 * 3764 * *th* points to the start of the TCP header, while *th_len* 3765 * contains the length of the TCP header. 3766 * Return 3767 * On success, lower 32 bits hold the generated SYN cookie in 3768 * followed by 16 bits which hold the MSS value for that cookie, 3769 * and the top 16 bits are unused. 3770 * 3771 * On failure, the returned value is one of the following: 3772 * 3773 * **-EINVAL** SYN cookie cannot be issued due to error 3774 * 3775 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3776 * 3777 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3778 * 3779 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3780 * 3781 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3782 * Description 3783 * Write raw *data* blob into a special BPF perf event held by 3784 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3785 * event must have the following attributes: **PERF_SAMPLE_RAW** 3786 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3787 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3788 * 3789 * The *flags* are used to indicate the index in *map* for which 3790 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3791 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3792 * to indicate that the index of the current CPU core should be 3793 * used. 3794 * 3795 * The value to write, of *size*, is passed through eBPF stack and 3796 * pointed by *data*. 3797 * 3798 * *ctx* is a pointer to in-kernel struct sk_buff. 3799 * 3800 * This helper is similar to **bpf_perf_event_output**\ () but 3801 * restricted to raw_tracepoint bpf programs. 3802 * Return 3803 * 0 on success, or a negative error in case of failure. 3804 * 3805 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3806 * Description 3807 * Safely attempt to read *size* bytes from user space address 3808 * *unsafe_ptr* and store the data in *dst*. 3809 * Return 3810 * 0 on success, or a negative error in case of failure. 3811 * 3812 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3813 * Description 3814 * Safely attempt to read *size* bytes from kernel space address 3815 * *unsafe_ptr* and store the data in *dst*. 3816 * Return 3817 * 0 on success, or a negative error in case of failure. 3818 * 3819 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3820 * Description 3821 * Copy a NUL terminated string from an unsafe user address 3822 * *unsafe_ptr* to *dst*. The *size* should include the 3823 * terminating NUL byte. In case the string length is smaller than 3824 * *size*, the target is not padded with further NUL bytes. If the 3825 * string length is larger than *size*, just *size*-1 bytes are 3826 * copied and the last byte is set to NUL. 3827 * 3828 * On success, returns the number of bytes that were written, 3829 * including the terminal NUL. This makes this helper useful in 3830 * tracing programs for reading strings, and more importantly to 3831 * get its length at runtime. See the following snippet: 3832 * 3833 * :: 3834 * 3835 * SEC("kprobe/sys_open") 3836 * void bpf_sys_open(struct pt_regs *ctx) 3837 * { 3838 * char buf[PATHLEN]; // PATHLEN is defined to 256 3839 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3840 * ctx->di); 3841 * 3842 * // Consume buf, for example push it to 3843 * // userspace via bpf_perf_event_output(); we 3844 * // can use res (the string length) as event 3845 * // size, after checking its boundaries. 3846 * } 3847 * 3848 * In comparison, using **bpf_probe_read_user**\ () helper here 3849 * instead to read the string would require to estimate the length 3850 * at compile time, and would often result in copying more memory 3851 * than necessary. 3852 * 3853 * Another useful use case is when parsing individual process 3854 * arguments or individual environment variables navigating 3855 * *current*\ **->mm->arg_start** and *current*\ 3856 * **->mm->env_start**: using this helper and the return value, 3857 * one can quickly iterate at the right offset of the memory area. 3858 * Return 3859 * On success, the strictly positive length of the output string, 3860 * including the trailing NUL character. On error, a negative 3861 * value. 3862 * 3863 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3864 * Description 3865 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3866 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3867 * Return 3868 * On success, the strictly positive length of the string, including 3869 * the trailing NUL character. On error, a negative value. 3870 * 3871 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3872 * Description 3873 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3874 * *rcv_nxt* is the ack_seq to be sent out. 3875 * Return 3876 * 0 on success, or a negative error in case of failure. 3877 * 3878 * long bpf_send_signal_thread(u32 sig) 3879 * Description 3880 * Send signal *sig* to the thread corresponding to the current task. 3881 * Return 3882 * 0 on success or successfully queued. 3883 * 3884 * **-EBUSY** if work queue under nmi is full. 3885 * 3886 * **-EINVAL** if *sig* is invalid. 3887 * 3888 * **-EPERM** if no permission to send the *sig*. 3889 * 3890 * **-EAGAIN** if bpf program can try again. 3891 * 3892 * u64 bpf_jiffies64(void) 3893 * Description 3894 * Obtain the 64bit jiffies 3895 * Return 3896 * The 64 bit jiffies 3897 * 3898 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3899 * Description 3900 * For an eBPF program attached to a perf event, retrieve the 3901 * branch records (**struct perf_branch_entry**) associated to *ctx* 3902 * and store it in the buffer pointed by *buf* up to size 3903 * *size* bytes. 3904 * Return 3905 * On success, number of bytes written to *buf*. On error, a 3906 * negative value. 3907 * 3908 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3909 * instead return the number of bytes required to store all the 3910 * branch entries. If this flag is set, *buf* may be NULL. 3911 * 3912 * **-EINVAL** if arguments invalid or **size** not a multiple 3913 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3914 * 3915 * **-ENOENT** if architecture does not support branch records. 3916 * 3917 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3918 * Description 3919 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3920 * *namespace* will be returned in *nsdata*. 3921 * Return 3922 * 0 on success, or one of the following in case of failure: 3923 * 3924 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3925 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3926 * 3927 * **-ENOENT** if pidns does not exists for the current task. 3928 * 3929 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3930 * Description 3931 * Write raw *data* blob into a special BPF perf event held by 3932 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3933 * event must have the following attributes: **PERF_SAMPLE_RAW** 3934 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3935 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3936 * 3937 * The *flags* are used to indicate the index in *map* for which 3938 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3939 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3940 * to indicate that the index of the current CPU core should be 3941 * used. 3942 * 3943 * The value to write, of *size*, is passed through eBPF stack and 3944 * pointed by *data*. 3945 * 3946 * *ctx* is a pointer to in-kernel struct xdp_buff. 3947 * 3948 * This helper is similar to **bpf_perf_eventoutput**\ () but 3949 * restricted to raw_tracepoint bpf programs. 3950 * Return 3951 * 0 on success, or a negative error in case of failure. 3952 * 3953 * u64 bpf_get_netns_cookie(void *ctx) 3954 * Description 3955 * Retrieve the cookie (generated by the kernel) of the network 3956 * namespace the input *ctx* is associated with. The network 3957 * namespace cookie remains stable for its lifetime and provides 3958 * a global identifier that can be assumed unique. If *ctx* is 3959 * NULL, then the helper returns the cookie for the initial 3960 * network namespace. The cookie itself is very similar to that 3961 * of **bpf_get_socket_cookie**\ () helper, but for network 3962 * namespaces instead of sockets. 3963 * Return 3964 * A 8-byte long opaque number. 3965 * 3966 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3967 * Description 3968 * Return id of cgroup v2 that is ancestor of the cgroup associated 3969 * with the current task at the *ancestor_level*. The root cgroup 3970 * is at *ancestor_level* zero and each step down the hierarchy 3971 * increments the level. If *ancestor_level* == level of cgroup 3972 * associated with the current task, then return value will be the 3973 * same as that of **bpf_get_current_cgroup_id**\ (). 3974 * 3975 * The helper is useful to implement policies based on cgroups 3976 * that are upper in hierarchy than immediate cgroup associated 3977 * with the current task. 3978 * 3979 * The format of returned id and helper limitations are same as in 3980 * **bpf_get_current_cgroup_id**\ (). 3981 * Return 3982 * The id is returned or 0 in case the id could not be retrieved. 3983 * 3984 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3985 * Description 3986 * Helper is overloaded depending on BPF program type. This 3987 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3988 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3989 * 3990 * Assign the *sk* to the *skb*. When combined with appropriate 3991 * routing configuration to receive the packet towards the socket, 3992 * will cause *skb* to be delivered to the specified socket. 3993 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3994 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3995 * interfere with successful delivery to the socket. 3996 * 3997 * This operation is only valid from TC ingress path. 3998 * 3999 * The *flags* argument must be zero. 4000 * Return 4001 * 0 on success, or a negative error in case of failure: 4002 * 4003 * **-EINVAL** if specified *flags* are not supported. 4004 * 4005 * **-ENOENT** if the socket is unavailable for assignment. 4006 * 4007 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4008 * 4009 * **-EOPNOTSUPP** if the operation is not supported, for example 4010 * a call from outside of TC ingress. 4011 * 4012 * **-ESOCKTNOSUPPORT** if the socket type is not supported 4013 * (reuseport). 4014 * 4015 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4016 * Description 4017 * Helper is overloaded depending on BPF program type. This 4018 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4019 * 4020 * Select the *sk* as a result of a socket lookup. 4021 * 4022 * For the operation to succeed passed socket must be compatible 4023 * with the packet description provided by the *ctx* object. 4024 * 4025 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4026 * be an exact match. While IP family (**AF_INET** or 4027 * **AF_INET6**) must be compatible, that is IPv6 sockets 4028 * that are not v6-only can be selected for IPv4 packets. 4029 * 4030 * Only TCP listeners and UDP unconnected sockets can be 4031 * selected. *sk* can also be NULL to reset any previous 4032 * selection. 4033 * 4034 * *flags* argument can combination of following values: 4035 * 4036 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4037 * socket selection, potentially done by a BPF program 4038 * that ran before us. 4039 * 4040 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4041 * load-balancing within reuseport group for the socket 4042 * being selected. 4043 * 4044 * On success *ctx->sk* will point to the selected socket. 4045 * 4046 * Return 4047 * 0 on success, or a negative errno in case of failure. 4048 * 4049 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4050 * not compatible with packet family (*ctx->family*). 4051 * 4052 * * **-EEXIST** if socket has been already selected, 4053 * potentially by another program, and 4054 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4055 * 4056 * * **-EINVAL** if unsupported flags were specified. 4057 * 4058 * * **-EPROTOTYPE** if socket L4 protocol 4059 * (*sk->protocol*) doesn't match packet protocol 4060 * (*ctx->protocol*). 4061 * 4062 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4063 * state (TCP listening or UDP unconnected). 4064 * 4065 * u64 bpf_ktime_get_boot_ns(void) 4066 * Description 4067 * Return the time elapsed since system boot, in nanoseconds. 4068 * Does include the time the system was suspended. 4069 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4070 * Return 4071 * Current *ktime*. 4072 * 4073 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4074 * Description 4075 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4076 * out the format string. 4077 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4078 * the format string itself. The *data* and *data_len* are format string 4079 * arguments. The *data* are a **u64** array and corresponding format string 4080 * values are stored in the array. For strings and pointers where pointees 4081 * are accessed, only the pointer values are stored in the *data* array. 4082 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4083 * 4084 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4085 * Reading kernel memory may fail due to either invalid address or 4086 * valid address but requiring a major memory fault. If reading kernel memory 4087 * fails, the string for **%s** will be an empty string, and the ip 4088 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4089 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4090 * Return 4091 * 0 on success, or a negative error in case of failure: 4092 * 4093 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4094 * by returning 1 from bpf program. 4095 * 4096 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4097 * 4098 * **-E2BIG** if *fmt* contains too many format specifiers. 4099 * 4100 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4101 * 4102 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4103 * Description 4104 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4105 * The *m* represents the seq_file. The *data* and *len* represent the 4106 * data to write in bytes. 4107 * Return 4108 * 0 on success, or a negative error in case of failure: 4109 * 4110 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4111 * 4112 * u64 bpf_sk_cgroup_id(void *sk) 4113 * Description 4114 * Return the cgroup v2 id of the socket *sk*. 4115 * 4116 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4117 * returned from **bpf_sk_lookup_xxx**\ (), 4118 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4119 * same as in **bpf_skb_cgroup_id**\ (). 4120 * 4121 * This helper is available only if the kernel was compiled with 4122 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4123 * Return 4124 * The id is returned or 0 in case the id could not be retrieved. 4125 * 4126 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4127 * Description 4128 * Return id of cgroup v2 that is ancestor of cgroup associated 4129 * with the *sk* at the *ancestor_level*. The root cgroup is at 4130 * *ancestor_level* zero and each step down the hierarchy 4131 * increments the level. If *ancestor_level* == level of cgroup 4132 * associated with *sk*, then return value will be same as that 4133 * of **bpf_sk_cgroup_id**\ (). 4134 * 4135 * The helper is useful to implement policies based on cgroups 4136 * that are upper in hierarchy than immediate cgroup associated 4137 * with *sk*. 4138 * 4139 * The format of returned id and helper limitations are same as in 4140 * **bpf_sk_cgroup_id**\ (). 4141 * Return 4142 * The id is returned or 0 in case the id could not be retrieved. 4143 * 4144 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4145 * Description 4146 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4147 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4148 * of new data availability is sent. 4149 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4150 * of new data availability is sent unconditionally. 4151 * If **0** is specified in *flags*, an adaptive notification 4152 * of new data availability is sent. 4153 * 4154 * An adaptive notification is a notification sent whenever the user-space 4155 * process has caught up and consumed all available payloads. In case the user-space 4156 * process is still processing a previous payload, then no notification is needed 4157 * as it will process the newly added payload automatically. 4158 * Return 4159 * 0 on success, or a negative error in case of failure. 4160 * 4161 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4162 * Description 4163 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4164 * *flags* must be 0. 4165 * Return 4166 * Valid pointer with *size* bytes of memory available; NULL, 4167 * otherwise. 4168 * 4169 * void bpf_ringbuf_submit(void *data, u64 flags) 4170 * Description 4171 * Submit reserved ring buffer sample, pointed to by *data*. 4172 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4173 * of new data availability is sent. 4174 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4175 * of new data availability is sent unconditionally. 4176 * If **0** is specified in *flags*, an adaptive notification 4177 * of new data availability is sent. 4178 * 4179 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4180 * Return 4181 * Nothing. Always succeeds. 4182 * 4183 * void bpf_ringbuf_discard(void *data, u64 flags) 4184 * Description 4185 * Discard reserved ring buffer sample, pointed to by *data*. 4186 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4187 * of new data availability is sent. 4188 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4189 * of new data availability is sent unconditionally. 4190 * If **0** is specified in *flags*, an adaptive notification 4191 * of new data availability is sent. 4192 * 4193 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4194 * Return 4195 * Nothing. Always succeeds. 4196 * 4197 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4198 * Description 4199 * Query various characteristics of provided ring buffer. What 4200 * exactly is queries is determined by *flags*: 4201 * 4202 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4203 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4204 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4205 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4206 * 4207 * Data returned is just a momentary snapshot of actual values 4208 * and could be inaccurate, so this facility should be used to 4209 * power heuristics and for reporting, not to make 100% correct 4210 * calculation. 4211 * Return 4212 * Requested value, or 0, if *flags* are not recognized. 4213 * 4214 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4215 * Description 4216 * Change the skbs checksum level by one layer up or down, or 4217 * reset it entirely to none in order to have the stack perform 4218 * checksum validation. The level is applicable to the following 4219 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4220 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4221 * through **bpf_skb_adjust_room**\ () helper with passing in 4222 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4223 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4224 * the UDP header is removed. Similarly, an encap of the latter 4225 * into the former could be accompanied by a helper call to 4226 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4227 * skb is still intended to be processed in higher layers of the 4228 * stack instead of just egressing at tc. 4229 * 4230 * There are three supported level settings at this time: 4231 * 4232 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4233 * with CHECKSUM_UNNECESSARY. 4234 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4235 * with CHECKSUM_UNNECESSARY. 4236 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4237 * sets CHECKSUM_NONE to force checksum validation by the stack. 4238 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4239 * skb->csum_level. 4240 * Return 4241 * 0 on success, or a negative error in case of failure. In the 4242 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4243 * is returned or the error code -EACCES in case the skb is not 4244 * subject to CHECKSUM_UNNECESSARY. 4245 * 4246 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4247 * Description 4248 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4249 * Return 4250 * *sk* if casting is valid, or **NULL** otherwise. 4251 * 4252 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4253 * Description 4254 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4255 * Return 4256 * *sk* if casting is valid, or **NULL** otherwise. 4257 * 4258 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4259 * Description 4260 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4261 * Return 4262 * *sk* if casting is valid, or **NULL** otherwise. 4263 * 4264 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4265 * Description 4266 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4267 * Return 4268 * *sk* if casting is valid, or **NULL** otherwise. 4269 * 4270 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4271 * Description 4272 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4273 * Return 4274 * *sk* if casting is valid, or **NULL** otherwise. 4275 * 4276 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4277 * Description 4278 * Return a user or a kernel stack in bpf program provided buffer. 4279 * To achieve this, the helper needs *task*, which is a valid 4280 * pointer to **struct task_struct**. To store the stacktrace, the 4281 * bpf program provides *buf* with a nonnegative *size*. 4282 * 4283 * The last argument, *flags*, holds the number of stack frames to 4284 * skip (from 0 to 255), masked with 4285 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4286 * the following flags: 4287 * 4288 * **BPF_F_USER_STACK** 4289 * Collect a user space stack instead of a kernel stack. 4290 * **BPF_F_USER_BUILD_ID** 4291 * Collect buildid+offset instead of ips for user stack, 4292 * only valid if **BPF_F_USER_STACK** is also specified. 4293 * 4294 * **bpf_get_task_stack**\ () can collect up to 4295 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4296 * to sufficient large buffer size. Note that 4297 * this limit can be controlled with the **sysctl** program, and 4298 * that it should be manually increased in order to profile long 4299 * user stacks (such as stacks for Java programs). To do so, use: 4300 * 4301 * :: 4302 * 4303 * # sysctl kernel.perf_event_max_stack=<new value> 4304 * Return 4305 * A non-negative value equal to or less than *size* on success, 4306 * or a negative error in case of failure. 4307 * 4308 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4309 * Description 4310 * Load header option. Support reading a particular TCP header 4311 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4312 * 4313 * If *flags* is 0, it will search the option from the 4314 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4315 * has details on what skb_data contains under different 4316 * *skops*\ **->op**. 4317 * 4318 * The first byte of the *searchby_res* specifies the 4319 * kind that it wants to search. 4320 * 4321 * If the searching kind is an experimental kind 4322 * (i.e. 253 or 254 according to RFC6994). It also 4323 * needs to specify the "magic" which is either 4324 * 2 bytes or 4 bytes. It then also needs to 4325 * specify the size of the magic by using 4326 * the 2nd byte which is "kind-length" of a TCP 4327 * header option and the "kind-length" also 4328 * includes the first 2 bytes "kind" and "kind-length" 4329 * itself as a normal TCP header option also does. 4330 * 4331 * For example, to search experimental kind 254 with 4332 * 2 byte magic 0xeB9F, the searchby_res should be 4333 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4334 * 4335 * To search for the standard window scale option (3), 4336 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4337 * Note, kind-length must be 0 for regular option. 4338 * 4339 * Searching for No-Op (0) and End-of-Option-List (1) are 4340 * not supported. 4341 * 4342 * *len* must be at least 2 bytes which is the minimal size 4343 * of a header option. 4344 * 4345 * Supported flags: 4346 * 4347 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4348 * saved_syn packet or the just-received syn packet. 4349 * 4350 * Return 4351 * > 0 when found, the header option is copied to *searchby_res*. 4352 * The return value is the total length copied. On failure, a 4353 * negative error code is returned: 4354 * 4355 * **-EINVAL** if a parameter is invalid. 4356 * 4357 * **-ENOMSG** if the option is not found. 4358 * 4359 * **-ENOENT** if no syn packet is available when 4360 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4361 * 4362 * **-ENOSPC** if there is not enough space. Only *len* number of 4363 * bytes are copied. 4364 * 4365 * **-EFAULT** on failure to parse the header options in the 4366 * packet. 4367 * 4368 * **-EPERM** if the helper cannot be used under the current 4369 * *skops*\ **->op**. 4370 * 4371 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4372 * Description 4373 * Store header option. The data will be copied 4374 * from buffer *from* with length *len* to the TCP header. 4375 * 4376 * The buffer *from* should have the whole option that 4377 * includes the kind, kind-length, and the actual 4378 * option data. The *len* must be at least kind-length 4379 * long. The kind-length does not have to be 4 byte 4380 * aligned. The kernel will take care of the padding 4381 * and setting the 4 bytes aligned value to th->doff. 4382 * 4383 * This helper will check for duplicated option 4384 * by searching the same option in the outgoing skb. 4385 * 4386 * This helper can only be called during 4387 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4388 * 4389 * Return 4390 * 0 on success, or negative error in case of failure: 4391 * 4392 * **-EINVAL** If param is invalid. 4393 * 4394 * **-ENOSPC** if there is not enough space in the header. 4395 * Nothing has been written 4396 * 4397 * **-EEXIST** if the option already exists. 4398 * 4399 * **-EFAULT** on failrue to parse the existing header options. 4400 * 4401 * **-EPERM** if the helper cannot be used under the current 4402 * *skops*\ **->op**. 4403 * 4404 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4405 * Description 4406 * Reserve *len* bytes for the bpf header option. The 4407 * space will be used by **bpf_store_hdr_opt**\ () later in 4408 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4409 * 4410 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4411 * the total number of bytes will be reserved. 4412 * 4413 * This helper can only be called during 4414 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4415 * 4416 * Return 4417 * 0 on success, or negative error in case of failure: 4418 * 4419 * **-EINVAL** if a parameter is invalid. 4420 * 4421 * **-ENOSPC** if there is not enough space in the header. 4422 * 4423 * **-EPERM** if the helper cannot be used under the current 4424 * *skops*\ **->op**. 4425 * 4426 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4427 * Description 4428 * Get a bpf_local_storage from an *inode*. 4429 * 4430 * Logically, it could be thought of as getting the value from 4431 * a *map* with *inode* as the **key**. From this 4432 * perspective, the usage is not much different from 4433 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4434 * helper enforces the key must be an inode and the map must also 4435 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4436 * 4437 * Underneath, the value is stored locally at *inode* instead of 4438 * the *map*. The *map* is used as the bpf-local-storage 4439 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4440 * searched against all bpf_local_storage residing at *inode*. 4441 * 4442 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4443 * used such that a new bpf_local_storage will be 4444 * created if one does not exist. *value* can be used 4445 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4446 * the initial value of a bpf_local_storage. If *value* is 4447 * **NULL**, the new bpf_local_storage will be zero initialized. 4448 * Return 4449 * A bpf_local_storage pointer is returned on success. 4450 * 4451 * **NULL** if not found or there was an error in adding 4452 * a new bpf_local_storage. 4453 * 4454 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4455 * Description 4456 * Delete a bpf_local_storage from an *inode*. 4457 * Return 4458 * 0 on success. 4459 * 4460 * **-ENOENT** if the bpf_local_storage cannot be found. 4461 * 4462 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4463 * Description 4464 * Return full path for given **struct path** object, which 4465 * needs to be the kernel BTF *path* object. The path is 4466 * returned in the provided buffer *buf* of size *sz* and 4467 * is zero terminated. 4468 * 4469 * Return 4470 * On success, the strictly positive length of the string, 4471 * including the trailing NUL character. On error, a negative 4472 * value. 4473 * 4474 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4475 * Description 4476 * Read *size* bytes from user space address *user_ptr* and store 4477 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4478 * Return 4479 * 0 on success, or a negative error in case of failure. 4480 * 4481 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4482 * Description 4483 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4484 * using *ptr*->type_id. This value should specify the type 4485 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4486 * can be used to look up vmlinux BTF type ids. Traversing the 4487 * data structure using BTF, the type information and values are 4488 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4489 * the pointer data is carried out to avoid kernel crashes during 4490 * operation. Smaller types can use string space on the stack; 4491 * larger programs can use map data to store the string 4492 * representation. 4493 * 4494 * The string can be subsequently shared with userspace via 4495 * bpf_perf_event_output() or ring buffer interfaces. 4496 * bpf_trace_printk() is to be avoided as it places too small 4497 * a limit on string size to be useful. 4498 * 4499 * *flags* is a combination of 4500 * 4501 * **BTF_F_COMPACT** 4502 * no formatting around type information 4503 * **BTF_F_NONAME** 4504 * no struct/union member names/types 4505 * **BTF_F_PTR_RAW** 4506 * show raw (unobfuscated) pointer values; 4507 * equivalent to printk specifier %px. 4508 * **BTF_F_ZERO** 4509 * show zero-valued struct/union members; they 4510 * are not displayed by default 4511 * 4512 * Return 4513 * The number of bytes that were written (or would have been 4514 * written if output had to be truncated due to string size), 4515 * or a negative error in cases of failure. 4516 * 4517 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4518 * Description 4519 * Use BTF to write to seq_write a string representation of 4520 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4521 * *flags* are identical to those used for bpf_snprintf_btf. 4522 * Return 4523 * 0 on success or a negative error in case of failure. 4524 * 4525 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4526 * Description 4527 * See **bpf_get_cgroup_classid**\ () for the main description. 4528 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4529 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4530 * associated socket instead of the current process. 4531 * Return 4532 * The id is returned or 0 in case the id could not be retrieved. 4533 * 4534 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4535 * Description 4536 * Redirect the packet to another net device of index *ifindex* 4537 * and fill in L2 addresses from neighboring subsystem. This helper 4538 * is somewhat similar to **bpf_redirect**\ (), except that it 4539 * populates L2 addresses as well, meaning, internally, the helper 4540 * relies on the neighbor lookup for the L2 address of the nexthop. 4541 * 4542 * The helper will perform a FIB lookup based on the skb's 4543 * networking header to get the address of the next hop, unless 4544 * this is supplied by the caller in the *params* argument. The 4545 * *plen* argument indicates the len of *params* and should be set 4546 * to 0 if *params* is NULL. 4547 * 4548 * The *flags* argument is reserved and must be 0. The helper is 4549 * currently only supported for tc BPF program types, and enabled 4550 * for IPv4 and IPv6 protocols. 4551 * Return 4552 * The helper returns **TC_ACT_REDIRECT** on success or 4553 * **TC_ACT_SHOT** on error. 4554 * 4555 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4556 * Description 4557 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4558 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4559 * extern variable decorated with '__ksym'. For ksym, there is a 4560 * global var (either static or global) defined of the same name 4561 * in the kernel. The ksym is percpu if the global var is percpu. 4562 * The returned pointer points to the global percpu var on *cpu*. 4563 * 4564 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4565 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4566 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4567 * bpf_per_cpu_ptr() must check the returned value. 4568 * Return 4569 * A pointer pointing to the kernel percpu variable on *cpu*, or 4570 * NULL, if *cpu* is invalid. 4571 * 4572 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4573 * Description 4574 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4575 * pointer to the percpu kernel variable on this cpu. See the 4576 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4577 * 4578 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4579 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4580 * never return NULL. 4581 * Return 4582 * A pointer pointing to the kernel percpu variable on this cpu. 4583 * 4584 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4585 * Description 4586 * Redirect the packet to another net device of index *ifindex*. 4587 * This helper is somewhat similar to **bpf_redirect**\ (), except 4588 * that the redirection happens to the *ifindex*' peer device and 4589 * the netns switch takes place from ingress to ingress without 4590 * going through the CPU's backlog queue. 4591 * 4592 * The *flags* argument is reserved and must be 0. The helper is 4593 * currently only supported for tc BPF program types at the ingress 4594 * hook and for veth device types. The peer device must reside in a 4595 * different network namespace. 4596 * Return 4597 * The helper returns **TC_ACT_REDIRECT** on success or 4598 * **TC_ACT_SHOT** on error. 4599 * 4600 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4601 * Description 4602 * Get a bpf_local_storage from the *task*. 4603 * 4604 * Logically, it could be thought of as getting the value from 4605 * a *map* with *task* as the **key**. From this 4606 * perspective, the usage is not much different from 4607 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4608 * helper enforces the key must be an task_struct and the map must also 4609 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4610 * 4611 * Underneath, the value is stored locally at *task* instead of 4612 * the *map*. The *map* is used as the bpf-local-storage 4613 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4614 * searched against all bpf_local_storage residing at *task*. 4615 * 4616 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4617 * used such that a new bpf_local_storage will be 4618 * created if one does not exist. *value* can be used 4619 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4620 * the initial value of a bpf_local_storage. If *value* is 4621 * **NULL**, the new bpf_local_storage will be zero initialized. 4622 * Return 4623 * A bpf_local_storage pointer is returned on success. 4624 * 4625 * **NULL** if not found or there was an error in adding 4626 * a new bpf_local_storage. 4627 * 4628 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4629 * Description 4630 * Delete a bpf_local_storage from a *task*. 4631 * Return 4632 * 0 on success. 4633 * 4634 * **-ENOENT** if the bpf_local_storage cannot be found. 4635 * 4636 * struct task_struct *bpf_get_current_task_btf(void) 4637 * Description 4638 * Return a BTF pointer to the "current" task. 4639 * This pointer can also be used in helpers that accept an 4640 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4641 * Return 4642 * Pointer to the current task. 4643 * 4644 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4645 * Description 4646 * Set or clear certain options on *bprm*: 4647 * 4648 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4649 * which sets the **AT_SECURE** auxv for glibc. The bit 4650 * is cleared if the flag is not specified. 4651 * Return 4652 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4653 * 4654 * u64 bpf_ktime_get_coarse_ns(void) 4655 * Description 4656 * Return a coarse-grained version of the time elapsed since 4657 * system boot, in nanoseconds. Does not include time the system 4658 * was suspended. 4659 * 4660 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4661 * Return 4662 * Current *ktime*. 4663 * 4664 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4665 * Description 4666 * Returns the stored IMA hash of the *inode* (if it's avaialable). 4667 * If the hash is larger than *size*, then only *size* 4668 * bytes will be copied to *dst* 4669 * Return 4670 * The **hash_algo** is returned on success, 4671 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4672 * invalid arguments are passed. 4673 * 4674 * struct socket *bpf_sock_from_file(struct file *file) 4675 * Description 4676 * If the given file represents a socket, returns the associated 4677 * socket. 4678 * Return 4679 * A pointer to a struct socket on success or NULL if the file is 4680 * not a socket. 4681 * 4682 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4683 * Description 4684 * Check packet size against exceeding MTU of net device (based 4685 * on *ifindex*). This helper will likely be used in combination 4686 * with helpers that adjust/change the packet size. 4687 * 4688 * The argument *len_diff* can be used for querying with a planned 4689 * size change. This allows to check MTU prior to changing packet 4690 * ctx. Providing an *len_diff* adjustment that is larger than the 4691 * actual packet size (resulting in negative packet size) will in 4692 * principle not exceed the MTU, why it is not considered a 4693 * failure. Other BPF-helpers are needed for performing the 4694 * planned size change, why the responsability for catch a negative 4695 * packet size belong in those helpers. 4696 * 4697 * Specifying *ifindex* zero means the MTU check is performed 4698 * against the current net device. This is practical if this isn't 4699 * used prior to redirect. 4700 * 4701 * On input *mtu_len* must be a valid pointer, else verifier will 4702 * reject BPF program. If the value *mtu_len* is initialized to 4703 * zero then the ctx packet size is use. When value *mtu_len* is 4704 * provided as input this specify the L3 length that the MTU check 4705 * is done against. Remember XDP and TC length operate at L2, but 4706 * this value is L3 as this correlate to MTU and IP-header tot_len 4707 * values which are L3 (similar behavior as bpf_fib_lookup). 4708 * 4709 * The Linux kernel route table can configure MTUs on a more 4710 * specific per route level, which is not provided by this helper. 4711 * For route level MTU checks use the **bpf_fib_lookup**\ () 4712 * helper. 4713 * 4714 * *ctx* is either **struct xdp_md** for XDP programs or 4715 * **struct sk_buff** for tc cls_act programs. 4716 * 4717 * The *flags* argument can be a combination of one or more of the 4718 * following values: 4719 * 4720 * **BPF_MTU_CHK_SEGS** 4721 * This flag will only works for *ctx* **struct sk_buff**. 4722 * If packet context contains extra packet segment buffers 4723 * (often knows as GSO skb), then MTU check is harder to 4724 * check at this point, because in transmit path it is 4725 * possible for the skb packet to get re-segmented 4726 * (depending on net device features). This could still be 4727 * a MTU violation, so this flag enables performing MTU 4728 * check against segments, with a different violation 4729 * return code to tell it apart. Check cannot use len_diff. 4730 * 4731 * On return *mtu_len* pointer contains the MTU value of the net 4732 * device. Remember the net device configured MTU is the L3 size, 4733 * which is returned here and XDP and TC length operate at L2. 4734 * Helper take this into account for you, but remember when using 4735 * MTU value in your BPF-code. 4736 * 4737 * Return 4738 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4739 * 4740 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4741 * 4742 * MTU violations return positive values, but also populate MTU 4743 * value in *mtu_len* pointer, as this can be needed for 4744 * implementing PMTU handing: 4745 * 4746 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4747 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4748 * 4749 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4750 * Description 4751 * For each element in **map**, call **callback_fn** function with 4752 * **map**, **callback_ctx** and other map-specific parameters. 4753 * The **callback_fn** should be a static function and 4754 * the **callback_ctx** should be a pointer to the stack. 4755 * The **flags** is used to control certain aspects of the helper. 4756 * Currently, the **flags** must be 0. 4757 * 4758 * The following are a list of supported map types and their 4759 * respective expected callback signatures: 4760 * 4761 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4762 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4763 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4764 * 4765 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4766 * 4767 * For per_cpu maps, the map_value is the value on the cpu where the 4768 * bpf_prog is running. 4769 * 4770 * If **callback_fn** return 0, the helper will continue to the next 4771 * element. If return value is 1, the helper will skip the rest of 4772 * elements and return. Other return values are not used now. 4773 * 4774 * Return 4775 * The number of traversed map elements for success, **-EINVAL** for 4776 * invalid **flags**. 4777 * 4778 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4779 * Description 4780 * Outputs a string into the **str** buffer of size **str_size** 4781 * based on a format string stored in a read-only map pointed by 4782 * **fmt**. 4783 * 4784 * Each format specifier in **fmt** corresponds to one u64 element 4785 * in the **data** array. For strings and pointers where pointees 4786 * are accessed, only the pointer values are stored in the *data* 4787 * array. The *data_len* is the size of *data* in bytes - must be 4788 * a multiple of 8. 4789 * 4790 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4791 * memory. Reading kernel memory may fail due to either invalid 4792 * address or valid address but requiring a major memory fault. If 4793 * reading kernel memory fails, the string for **%s** will be an 4794 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4795 * Not returning error to bpf program is consistent with what 4796 * **bpf_trace_printk**\ () does for now. 4797 * 4798 * Return 4799 * The strictly positive length of the formatted string, including 4800 * the trailing zero character. If the return value is greater than 4801 * **str_size**, **str** contains a truncated string, guaranteed to 4802 * be zero-terminated except when **str_size** is 0. 4803 * 4804 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4805 * 4806 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4807 * Description 4808 * Execute bpf syscall with given arguments. 4809 * Return 4810 * A syscall result. 4811 * 4812 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4813 * Description 4814 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4815 * Return 4816 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4817 * 4818 * long bpf_sys_close(u32 fd) 4819 * Description 4820 * Execute close syscall for given FD. 4821 * Return 4822 * A syscall result. 4823 * 4824 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4825 * Description 4826 * Initialize the timer. 4827 * First 4 bits of *flags* specify clockid. 4828 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4829 * All other bits of *flags* are reserved. 4830 * The verifier will reject the program if *timer* is not from 4831 * the same *map*. 4832 * Return 4833 * 0 on success. 4834 * **-EBUSY** if *timer* is already initialized. 4835 * **-EINVAL** if invalid *flags* are passed. 4836 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4837 * The user space should either hold a file descriptor to a map with timers 4838 * or pin such map in bpffs. When map is unpinned or file descriptor is 4839 * closed all timers in the map will be cancelled and freed. 4840 * 4841 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4842 * Description 4843 * Configure the timer to call *callback_fn* static function. 4844 * Return 4845 * 0 on success. 4846 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4847 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4848 * The user space should either hold a file descriptor to a map with timers 4849 * or pin such map in bpffs. When map is unpinned or file descriptor is 4850 * closed all timers in the map will be cancelled and freed. 4851 * 4852 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4853 * Description 4854 * Set timer expiration N nanoseconds from the current time. The 4855 * configured callback will be invoked in soft irq context on some cpu 4856 * and will not repeat unless another bpf_timer_start() is made. 4857 * In such case the next invocation can migrate to a different cpu. 4858 * Since struct bpf_timer is a field inside map element the map 4859 * owns the timer. The bpf_timer_set_callback() will increment refcnt 4860 * of BPF program to make sure that callback_fn code stays valid. 4861 * When user space reference to a map reaches zero all timers 4862 * in a map are cancelled and corresponding program's refcnts are 4863 * decremented. This is done to make sure that Ctrl-C of a user 4864 * process doesn't leave any timers running. If map is pinned in 4865 * bpffs the callback_fn can re-arm itself indefinitely. 4866 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4867 * cancel and free the timer in the given map element. 4868 * The map can contain timers that invoke callback_fn-s from different 4869 * programs. The same callback_fn can serve different timers from 4870 * different maps if key/value layout matches across maps. 4871 * Every bpf_timer_set_callback() can have different callback_fn. 4872 * 4873 * Return 4874 * 0 on success. 4875 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4876 * or invalid *flags* are passed. 4877 * 4878 * long bpf_timer_cancel(struct bpf_timer *timer) 4879 * Description 4880 * Cancel the timer and wait for callback_fn to finish if it was running. 4881 * Return 4882 * 0 if the timer was not active. 4883 * 1 if the timer was active. 4884 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4885 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4886 * own timer which would have led to a deadlock otherwise. 4887 * 4888 * u64 bpf_get_func_ip(void *ctx) 4889 * Description 4890 * Get address of the traced function (for tracing and kprobe programs). 4891 * Return 4892 * Address of the traced function. 4893 * 4894 * u64 bpf_get_attach_cookie(void *ctx) 4895 * Description 4896 * Get bpf_cookie value provided (optionally) during the program 4897 * attachment. It might be different for each individual 4898 * attachment, even if BPF program itself is the same. 4899 * Expects BPF program context *ctx* as a first argument. 4900 * 4901 * Supported for the following program types: 4902 * - kprobe/uprobe; 4903 * - tracepoint; 4904 * - perf_event. 4905 * Return 4906 * Value specified by user at BPF link creation/attachment time 4907 * or 0, if it was not specified. 4908 * 4909 * long bpf_task_pt_regs(struct task_struct *task) 4910 * Description 4911 * Get the struct pt_regs associated with **task**. 4912 * Return 4913 * A pointer to struct pt_regs. 4914 * 4915 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 4916 * Description 4917 * Get branch trace from hardware engines like Intel LBR. The 4918 * hardware engine is stopped shortly after the helper is 4919 * called. Therefore, the user need to filter branch entries 4920 * based on the actual use case. To capture branch trace 4921 * before the trigger point of the BPF program, the helper 4922 * should be called at the beginning of the BPF program. 4923 * 4924 * The data is stored as struct perf_branch_entry into output 4925 * buffer *entries*. *size* is the size of *entries* in bytes. 4926 * *flags* is reserved for now and must be zero. 4927 * 4928 * Return 4929 * On success, number of bytes written to *buf*. On error, a 4930 * negative value. 4931 * 4932 * **-EINVAL** if *flags* is not zero. 4933 * 4934 * **-ENOENT** if architecture does not support branch records. 4935 * 4936 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4937 * Description 4938 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 4939 * to format and can handle more format args as a result. 4940 * 4941 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 4942 * Return 4943 * The number of bytes written to the buffer, or a negative error 4944 * in case of failure. 4945 * 4946 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 4947 * Description 4948 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 4949 * Return 4950 * *sk* if casting is valid, or **NULL** otherwise. 4951 * 4952 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 4953 * Description 4954 * Get the address of a kernel symbol, returned in *res*. *res* is 4955 * set to 0 if the symbol is not found. 4956 * Return 4957 * On success, zero. On error, a negative value. 4958 * 4959 * **-EINVAL** if *flags* is not zero. 4960 * 4961 * **-EINVAL** if string *name* is not the same size as *name_sz*. 4962 * 4963 * **-ENOENT** if symbol is not found. 4964 * 4965 * **-EPERM** if caller does not have permission to obtain kernel address. 4966 * 4967 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 4968 * Description 4969 * Find vma of *task* that contains *addr*, call *callback_fn* 4970 * function with *task*, *vma*, and *callback_ctx*. 4971 * The *callback_fn* should be a static function and 4972 * the *callback_ctx* should be a pointer to the stack. 4973 * The *flags* is used to control certain aspects of the helper. 4974 * Currently, the *flags* must be 0. 4975 * 4976 * The expected callback signature is 4977 * 4978 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 4979 * 4980 * Return 4981 * 0 on success. 4982 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 4983 * **-EBUSY** if failed to try lock mmap_lock. 4984 * **-EINVAL** for invalid **flags**. 4985 * 4986 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 4987 * Description 4988 * For **nr_loops**, call **callback_fn** function 4989 * with **callback_ctx** as the context parameter. 4990 * The **callback_fn** should be a static function and 4991 * the **callback_ctx** should be a pointer to the stack. 4992 * The **flags** is used to control certain aspects of the helper. 4993 * Currently, the **flags** must be 0. Currently, nr_loops is 4994 * limited to 1 << 23 (~8 million) loops. 4995 * 4996 * long (\*callback_fn)(u32 index, void \*ctx); 4997 * 4998 * where **index** is the current index in the loop. The index 4999 * is zero-indexed. 5000 * 5001 * If **callback_fn** returns 0, the helper will continue to the next 5002 * loop. If return value is 1, the helper will skip the rest of 5003 * the loops and return. Other return values are not used now, 5004 * and will be rejected by the verifier. 5005 * 5006 * Return 5007 * The number of loops performed, **-EINVAL** for invalid **flags**, 5008 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5009 * 5010 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5011 * Description 5012 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5013 * to be null-terminated and **s1_sz** is the maximum storage 5014 * size of **s1**. **s2** must be a read-only string. 5015 * Return 5016 * An integer less than, equal to, or greater than zero 5017 * if the first **s1_sz** bytes of **s1** is found to be 5018 * less than, to match, or be greater than **s2**. 5019 * 5020 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5021 * Description 5022 * Get **n**-th argument (zero based) of the traced function (for tracing programs) 5023 * returned in **value**. 5024 * 5025 * Return 5026 * 0 on success. 5027 * **-EINVAL** if n >= arguments count of traced function. 5028 * 5029 * long bpf_get_func_ret(void *ctx, u64 *value) 5030 * Description 5031 * Get return value of the traced function (for tracing programs) 5032 * in **value**. 5033 * 5034 * Return 5035 * 0 on success. 5036 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5037 * 5038 * long bpf_get_func_arg_cnt(void *ctx) 5039 * Description 5040 * Get number of arguments of the traced function (for tracing programs). 5041 * 5042 * Return 5043 * The number of arguments of the traced function. 5044 * 5045 * int bpf_get_retval(void) 5046 * Description 5047 * Get the syscall's return value that will be returned to userspace. 5048 * 5049 * This helper is currently supported by cgroup programs only. 5050 * Return 5051 * The syscall's return value. 5052 * 5053 * int bpf_set_retval(int retval) 5054 * Description 5055 * Set the syscall's return value that will be returned to userspace. 5056 * 5057 * This helper is currently supported by cgroup programs only. 5058 * Return 5059 * 0 on success, or a negative error in case of failure. 5060 * 5061 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5062 * Description 5063 * Get the total size of a given xdp buff (linear and paged area) 5064 * Return 5065 * The total size of a given xdp buffer. 5066 * 5067 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5068 * Description 5069 * This helper is provided as an easy way to load data from a 5070 * xdp buffer. It can be used to load *len* bytes from *offset* from 5071 * the frame associated to *xdp_md*, into the buffer pointed by 5072 * *buf*. 5073 * Return 5074 * 0 on success, or a negative error in case of failure. 5075 * 5076 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5077 * Description 5078 * Store *len* bytes from buffer *buf* into the frame 5079 * associated to *xdp_md*, at *offset*. 5080 * Return 5081 * 0 on success, or a negative error in case of failure. 5082 * 5083 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5084 * Description 5085 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5086 * address space, and stores the data in *dst*. *flags* is not 5087 * used yet and is provided for future extensibility. This helper 5088 * can only be used by sleepable programs. 5089 * Return 5090 * 0 on success, or a negative error in case of failure. On error 5091 * *dst* buffer is zeroed out. 5092 * 5093 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5094 * Description 5095 * Change the __sk_buff->tstamp_type to *tstamp_type* 5096 * and set *tstamp* to the __sk_buff->tstamp together. 5097 * 5098 * If there is no need to change the __sk_buff->tstamp_type, 5099 * the tstamp value can be directly written to __sk_buff->tstamp 5100 * instead. 5101 * 5102 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5103 * will be kept during bpf_redirect_*(). A non zero 5104 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5105 * *tstamp_type*. 5106 * 5107 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5108 * with a zero *tstamp*. 5109 * 5110 * Only IPv4 and IPv6 skb->protocol are supported. 5111 * 5112 * This function is most useful when it needs to set a 5113 * mono delivery time to __sk_buff->tstamp and then 5114 * bpf_redirect_*() to the egress of an iface. For example, 5115 * changing the (rcv) timestamp in __sk_buff->tstamp at 5116 * ingress to a mono delivery time and then bpf_redirect_*() 5117 * to sch_fq@phy-dev. 5118 * Return 5119 * 0 on success. 5120 * **-EINVAL** for invalid input 5121 * **-EOPNOTSUPP** for unsupported protocol 5122 * 5123 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5124 * Description 5125 * Returns a calculated IMA hash of the *file*. 5126 * If the hash is larger than *size*, then only *size* 5127 * bytes will be copied to *dst* 5128 * Return 5129 * The **hash_algo** is returned on success, 5130 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5131 * invalid arguments are passed. 5132 */ 5133 #define __BPF_FUNC_MAPPER(FN) \ 5134 FN(unspec), \ 5135 FN(map_lookup_elem), \ 5136 FN(map_update_elem), \ 5137 FN(map_delete_elem), \ 5138 FN(probe_read), \ 5139 FN(ktime_get_ns), \ 5140 FN(trace_printk), \ 5141 FN(get_prandom_u32), \ 5142 FN(get_smp_processor_id), \ 5143 FN(skb_store_bytes), \ 5144 FN(l3_csum_replace), \ 5145 FN(l4_csum_replace), \ 5146 FN(tail_call), \ 5147 FN(clone_redirect), \ 5148 FN(get_current_pid_tgid), \ 5149 FN(get_current_uid_gid), \ 5150 FN(get_current_comm), \ 5151 FN(get_cgroup_classid), \ 5152 FN(skb_vlan_push), \ 5153 FN(skb_vlan_pop), \ 5154 FN(skb_get_tunnel_key), \ 5155 FN(skb_set_tunnel_key), \ 5156 FN(perf_event_read), \ 5157 FN(redirect), \ 5158 FN(get_route_realm), \ 5159 FN(perf_event_output), \ 5160 FN(skb_load_bytes), \ 5161 FN(get_stackid), \ 5162 FN(csum_diff), \ 5163 FN(skb_get_tunnel_opt), \ 5164 FN(skb_set_tunnel_opt), \ 5165 FN(skb_change_proto), \ 5166 FN(skb_change_type), \ 5167 FN(skb_under_cgroup), \ 5168 FN(get_hash_recalc), \ 5169 FN(get_current_task), \ 5170 FN(probe_write_user), \ 5171 FN(current_task_under_cgroup), \ 5172 FN(skb_change_tail), \ 5173 FN(skb_pull_data), \ 5174 FN(csum_update), \ 5175 FN(set_hash_invalid), \ 5176 FN(get_numa_node_id), \ 5177 FN(skb_change_head), \ 5178 FN(xdp_adjust_head), \ 5179 FN(probe_read_str), \ 5180 FN(get_socket_cookie), \ 5181 FN(get_socket_uid), \ 5182 FN(set_hash), \ 5183 FN(setsockopt), \ 5184 FN(skb_adjust_room), \ 5185 FN(redirect_map), \ 5186 FN(sk_redirect_map), \ 5187 FN(sock_map_update), \ 5188 FN(xdp_adjust_meta), \ 5189 FN(perf_event_read_value), \ 5190 FN(perf_prog_read_value), \ 5191 FN(getsockopt), \ 5192 FN(override_return), \ 5193 FN(sock_ops_cb_flags_set), \ 5194 FN(msg_redirect_map), \ 5195 FN(msg_apply_bytes), \ 5196 FN(msg_cork_bytes), \ 5197 FN(msg_pull_data), \ 5198 FN(bind), \ 5199 FN(xdp_adjust_tail), \ 5200 FN(skb_get_xfrm_state), \ 5201 FN(get_stack), \ 5202 FN(skb_load_bytes_relative), \ 5203 FN(fib_lookup), \ 5204 FN(sock_hash_update), \ 5205 FN(msg_redirect_hash), \ 5206 FN(sk_redirect_hash), \ 5207 FN(lwt_push_encap), \ 5208 FN(lwt_seg6_store_bytes), \ 5209 FN(lwt_seg6_adjust_srh), \ 5210 FN(lwt_seg6_action), \ 5211 FN(rc_repeat), \ 5212 FN(rc_keydown), \ 5213 FN(skb_cgroup_id), \ 5214 FN(get_current_cgroup_id), \ 5215 FN(get_local_storage), \ 5216 FN(sk_select_reuseport), \ 5217 FN(skb_ancestor_cgroup_id), \ 5218 FN(sk_lookup_tcp), \ 5219 FN(sk_lookup_udp), \ 5220 FN(sk_release), \ 5221 FN(map_push_elem), \ 5222 FN(map_pop_elem), \ 5223 FN(map_peek_elem), \ 5224 FN(msg_push_data), \ 5225 FN(msg_pop_data), \ 5226 FN(rc_pointer_rel), \ 5227 FN(spin_lock), \ 5228 FN(spin_unlock), \ 5229 FN(sk_fullsock), \ 5230 FN(tcp_sock), \ 5231 FN(skb_ecn_set_ce), \ 5232 FN(get_listener_sock), \ 5233 FN(skc_lookup_tcp), \ 5234 FN(tcp_check_syncookie), \ 5235 FN(sysctl_get_name), \ 5236 FN(sysctl_get_current_value), \ 5237 FN(sysctl_get_new_value), \ 5238 FN(sysctl_set_new_value), \ 5239 FN(strtol), \ 5240 FN(strtoul), \ 5241 FN(sk_storage_get), \ 5242 FN(sk_storage_delete), \ 5243 FN(send_signal), \ 5244 FN(tcp_gen_syncookie), \ 5245 FN(skb_output), \ 5246 FN(probe_read_user), \ 5247 FN(probe_read_kernel), \ 5248 FN(probe_read_user_str), \ 5249 FN(probe_read_kernel_str), \ 5250 FN(tcp_send_ack), \ 5251 FN(send_signal_thread), \ 5252 FN(jiffies64), \ 5253 FN(read_branch_records), \ 5254 FN(get_ns_current_pid_tgid), \ 5255 FN(xdp_output), \ 5256 FN(get_netns_cookie), \ 5257 FN(get_current_ancestor_cgroup_id), \ 5258 FN(sk_assign), \ 5259 FN(ktime_get_boot_ns), \ 5260 FN(seq_printf), \ 5261 FN(seq_write), \ 5262 FN(sk_cgroup_id), \ 5263 FN(sk_ancestor_cgroup_id), \ 5264 FN(ringbuf_output), \ 5265 FN(ringbuf_reserve), \ 5266 FN(ringbuf_submit), \ 5267 FN(ringbuf_discard), \ 5268 FN(ringbuf_query), \ 5269 FN(csum_level), \ 5270 FN(skc_to_tcp6_sock), \ 5271 FN(skc_to_tcp_sock), \ 5272 FN(skc_to_tcp_timewait_sock), \ 5273 FN(skc_to_tcp_request_sock), \ 5274 FN(skc_to_udp6_sock), \ 5275 FN(get_task_stack), \ 5276 FN(load_hdr_opt), \ 5277 FN(store_hdr_opt), \ 5278 FN(reserve_hdr_opt), \ 5279 FN(inode_storage_get), \ 5280 FN(inode_storage_delete), \ 5281 FN(d_path), \ 5282 FN(copy_from_user), \ 5283 FN(snprintf_btf), \ 5284 FN(seq_printf_btf), \ 5285 FN(skb_cgroup_classid), \ 5286 FN(redirect_neigh), \ 5287 FN(per_cpu_ptr), \ 5288 FN(this_cpu_ptr), \ 5289 FN(redirect_peer), \ 5290 FN(task_storage_get), \ 5291 FN(task_storage_delete), \ 5292 FN(get_current_task_btf), \ 5293 FN(bprm_opts_set), \ 5294 FN(ktime_get_coarse_ns), \ 5295 FN(ima_inode_hash), \ 5296 FN(sock_from_file), \ 5297 FN(check_mtu), \ 5298 FN(for_each_map_elem), \ 5299 FN(snprintf), \ 5300 FN(sys_bpf), \ 5301 FN(btf_find_by_name_kind), \ 5302 FN(sys_close), \ 5303 FN(timer_init), \ 5304 FN(timer_set_callback), \ 5305 FN(timer_start), \ 5306 FN(timer_cancel), \ 5307 FN(get_func_ip), \ 5308 FN(get_attach_cookie), \ 5309 FN(task_pt_regs), \ 5310 FN(get_branch_snapshot), \ 5311 FN(trace_vprintk), \ 5312 FN(skc_to_unix_sock), \ 5313 FN(kallsyms_lookup_name), \ 5314 FN(find_vma), \ 5315 FN(loop), \ 5316 FN(strncmp), \ 5317 FN(get_func_arg), \ 5318 FN(get_func_ret), \ 5319 FN(get_func_arg_cnt), \ 5320 FN(get_retval), \ 5321 FN(set_retval), \ 5322 FN(xdp_get_buff_len), \ 5323 FN(xdp_load_bytes), \ 5324 FN(xdp_store_bytes), \ 5325 FN(copy_from_user_task), \ 5326 FN(skb_set_tstamp), \ 5327 FN(ima_file_hash), \ 5328 /* */ 5329 5330 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5331 * function eBPF program intends to call 5332 */ 5333 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5334 enum bpf_func_id { 5335 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5336 __BPF_FUNC_MAX_ID, 5337 }; 5338 #undef __BPF_ENUM_FN 5339 5340 /* All flags used by eBPF helper functions, placed here. */ 5341 5342 /* BPF_FUNC_skb_store_bytes flags. */ 5343 enum { 5344 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5345 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5346 }; 5347 5348 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5349 * First 4 bits are for passing the header field size. 5350 */ 5351 enum { 5352 BPF_F_HDR_FIELD_MASK = 0xfULL, 5353 }; 5354 5355 /* BPF_FUNC_l4_csum_replace flags. */ 5356 enum { 5357 BPF_F_PSEUDO_HDR = (1ULL << 4), 5358 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5359 BPF_F_MARK_ENFORCE = (1ULL << 6), 5360 }; 5361 5362 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5363 enum { 5364 BPF_F_INGRESS = (1ULL << 0), 5365 }; 5366 5367 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5368 enum { 5369 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5370 }; 5371 5372 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5373 enum { 5374 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5375 BPF_F_USER_STACK = (1ULL << 8), 5376 /* flags used by BPF_FUNC_get_stackid only. */ 5377 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5378 BPF_F_REUSE_STACKID = (1ULL << 10), 5379 /* flags used by BPF_FUNC_get_stack only. */ 5380 BPF_F_USER_BUILD_ID = (1ULL << 11), 5381 }; 5382 5383 /* BPF_FUNC_skb_set_tunnel_key flags. */ 5384 enum { 5385 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5386 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5387 BPF_F_SEQ_NUMBER = (1ULL << 3), 5388 }; 5389 5390 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5391 * BPF_FUNC_perf_event_read_value flags. 5392 */ 5393 enum { 5394 BPF_F_INDEX_MASK = 0xffffffffULL, 5395 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5396 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5397 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5398 }; 5399 5400 /* Current network namespace */ 5401 enum { 5402 BPF_F_CURRENT_NETNS = (-1L), 5403 }; 5404 5405 /* BPF_FUNC_csum_level level values. */ 5406 enum { 5407 BPF_CSUM_LEVEL_QUERY, 5408 BPF_CSUM_LEVEL_INC, 5409 BPF_CSUM_LEVEL_DEC, 5410 BPF_CSUM_LEVEL_RESET, 5411 }; 5412 5413 /* BPF_FUNC_skb_adjust_room flags. */ 5414 enum { 5415 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5416 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5417 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5418 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5419 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5420 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5421 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5422 }; 5423 5424 enum { 5425 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5426 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5427 }; 5428 5429 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5430 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5431 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5432 5433 /* BPF_FUNC_sysctl_get_name flags. */ 5434 enum { 5435 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5436 }; 5437 5438 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 5439 enum { 5440 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5441 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5442 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5443 */ 5444 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5445 }; 5446 5447 /* BPF_FUNC_read_branch_records flags. */ 5448 enum { 5449 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5450 }; 5451 5452 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5453 * BPF_FUNC_bpf_ringbuf_output flags. 5454 */ 5455 enum { 5456 BPF_RB_NO_WAKEUP = (1ULL << 0), 5457 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5458 }; 5459 5460 /* BPF_FUNC_bpf_ringbuf_query flags */ 5461 enum { 5462 BPF_RB_AVAIL_DATA = 0, 5463 BPF_RB_RING_SIZE = 1, 5464 BPF_RB_CONS_POS = 2, 5465 BPF_RB_PROD_POS = 3, 5466 }; 5467 5468 /* BPF ring buffer constants */ 5469 enum { 5470 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5471 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5472 BPF_RINGBUF_HDR_SZ = 8, 5473 }; 5474 5475 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5476 enum { 5477 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5478 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5479 }; 5480 5481 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 5482 enum bpf_adj_room_mode { 5483 BPF_ADJ_ROOM_NET, 5484 BPF_ADJ_ROOM_MAC, 5485 }; 5486 5487 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5488 enum bpf_hdr_start_off { 5489 BPF_HDR_START_MAC, 5490 BPF_HDR_START_NET, 5491 }; 5492 5493 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5494 enum bpf_lwt_encap_mode { 5495 BPF_LWT_ENCAP_SEG6, 5496 BPF_LWT_ENCAP_SEG6_INLINE, 5497 BPF_LWT_ENCAP_IP, 5498 }; 5499 5500 /* Flags for bpf_bprm_opts_set helper */ 5501 enum { 5502 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5503 }; 5504 5505 /* Flags for bpf_redirect_map helper */ 5506 enum { 5507 BPF_F_BROADCAST = (1ULL << 3), 5508 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5509 }; 5510 5511 #define __bpf_md_ptr(type, name) \ 5512 union { \ 5513 type name; \ 5514 __u64 :64; \ 5515 } __attribute__((aligned(8))) 5516 5517 enum { 5518 BPF_SKB_TSTAMP_UNSPEC, 5519 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 5520 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 5521 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 5522 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 5523 */ 5524 }; 5525 5526 /* user accessible mirror of in-kernel sk_buff. 5527 * new fields can only be added to the end of this structure 5528 */ 5529 struct __sk_buff { 5530 __u32 len; 5531 __u32 pkt_type; 5532 __u32 mark; 5533 __u32 queue_mapping; 5534 __u32 protocol; 5535 __u32 vlan_present; 5536 __u32 vlan_tci; 5537 __u32 vlan_proto; 5538 __u32 priority; 5539 __u32 ingress_ifindex; 5540 __u32 ifindex; 5541 __u32 tc_index; 5542 __u32 cb[5]; 5543 __u32 hash; 5544 __u32 tc_classid; 5545 __u32 data; 5546 __u32 data_end; 5547 __u32 napi_id; 5548 5549 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 5550 __u32 family; 5551 __u32 remote_ip4; /* Stored in network byte order */ 5552 __u32 local_ip4; /* Stored in network byte order */ 5553 __u32 remote_ip6[4]; /* Stored in network byte order */ 5554 __u32 local_ip6[4]; /* Stored in network byte order */ 5555 __u32 remote_port; /* Stored in network byte order */ 5556 __u32 local_port; /* stored in host byte order */ 5557 /* ... here. */ 5558 5559 __u32 data_meta; 5560 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5561 __u64 tstamp; 5562 __u32 wire_len; 5563 __u32 gso_segs; 5564 __bpf_md_ptr(struct bpf_sock *, sk); 5565 __u32 gso_size; 5566 __u8 tstamp_type; 5567 __u32 :24; /* Padding, future use. */ 5568 __u64 hwtstamp; 5569 }; 5570 5571 struct bpf_tunnel_key { 5572 __u32 tunnel_id; 5573 union { 5574 __u32 remote_ipv4; 5575 __u32 remote_ipv6[4]; 5576 }; 5577 __u8 tunnel_tos; 5578 __u8 tunnel_ttl; 5579 __u16 tunnel_ext; /* Padding, future use. */ 5580 __u32 tunnel_label; 5581 }; 5582 5583 /* user accessible mirror of in-kernel xfrm_state. 5584 * new fields can only be added to the end of this structure 5585 */ 5586 struct bpf_xfrm_state { 5587 __u32 reqid; 5588 __u32 spi; /* Stored in network byte order */ 5589 __u16 family; 5590 __u16 ext; /* Padding, future use. */ 5591 union { 5592 __u32 remote_ipv4; /* Stored in network byte order */ 5593 __u32 remote_ipv6[4]; /* Stored in network byte order */ 5594 }; 5595 }; 5596 5597 /* Generic BPF return codes which all BPF program types may support. 5598 * The values are binary compatible with their TC_ACT_* counter-part to 5599 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 5600 * programs. 5601 * 5602 * XDP is handled seprately, see XDP_*. 5603 */ 5604 enum bpf_ret_code { 5605 BPF_OK = 0, 5606 /* 1 reserved */ 5607 BPF_DROP = 2, 5608 /* 3-6 reserved */ 5609 BPF_REDIRECT = 7, 5610 /* >127 are reserved for prog type specific return codes. 5611 * 5612 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 5613 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 5614 * changed and should be routed based on its new L3 header. 5615 * (This is an L3 redirect, as opposed to L2 redirect 5616 * represented by BPF_REDIRECT above). 5617 */ 5618 BPF_LWT_REROUTE = 128, 5619 }; 5620 5621 struct bpf_sock { 5622 __u32 bound_dev_if; 5623 __u32 family; 5624 __u32 type; 5625 __u32 protocol; 5626 __u32 mark; 5627 __u32 priority; 5628 /* IP address also allows 1 and 2 bytes access */ 5629 __u32 src_ip4; 5630 __u32 src_ip6[4]; 5631 __u32 src_port; /* host byte order */ 5632 __be16 dst_port; /* network byte order */ 5633 __u16 :16; /* zero padding */ 5634 __u32 dst_ip4; 5635 __u32 dst_ip6[4]; 5636 __u32 state; 5637 __s32 rx_queue_mapping; 5638 }; 5639 5640 struct bpf_tcp_sock { 5641 __u32 snd_cwnd; /* Sending congestion window */ 5642 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5643 __u32 rtt_min; 5644 __u32 snd_ssthresh; /* Slow start size threshold */ 5645 __u32 rcv_nxt; /* What we want to receive next */ 5646 __u32 snd_nxt; /* Next sequence we send */ 5647 __u32 snd_una; /* First byte we want an ack for */ 5648 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5649 __u32 ecn_flags; /* ECN status bits. */ 5650 __u32 rate_delivered; /* saved rate sample: packets delivered */ 5651 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5652 __u32 packets_out; /* Packets which are "in flight" */ 5653 __u32 retrans_out; /* Retransmitted packets out */ 5654 __u32 total_retrans; /* Total retransmits for entire connection */ 5655 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5656 * total number of segments in. 5657 */ 5658 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5659 * total number of data segments in. 5660 */ 5661 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5662 * The total number of segments sent. 5663 */ 5664 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5665 * total number of data segments sent. 5666 */ 5667 __u32 lost_out; /* Lost packets */ 5668 __u32 sacked_out; /* SACK'd packets */ 5669 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5670 * sum(delta(rcv_nxt)), or how many bytes 5671 * were acked. 5672 */ 5673 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5674 * sum(delta(snd_una)), or how many bytes 5675 * were acked. 5676 */ 5677 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 5678 * total number of DSACK blocks received 5679 */ 5680 __u32 delivered; /* Total data packets delivered incl. rexmits */ 5681 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5682 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5683 }; 5684 5685 struct bpf_sock_tuple { 5686 union { 5687 struct { 5688 __be32 saddr; 5689 __be32 daddr; 5690 __be16 sport; 5691 __be16 dport; 5692 } ipv4; 5693 struct { 5694 __be32 saddr[4]; 5695 __be32 daddr[4]; 5696 __be16 sport; 5697 __be16 dport; 5698 } ipv6; 5699 }; 5700 }; 5701 5702 struct bpf_xdp_sock { 5703 __u32 queue_id; 5704 }; 5705 5706 #define XDP_PACKET_HEADROOM 256 5707 5708 /* User return codes for XDP prog type. 5709 * A valid XDP program must return one of these defined values. All other 5710 * return codes are reserved for future use. Unknown return codes will 5711 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 5712 */ 5713 enum xdp_action { 5714 XDP_ABORTED = 0, 5715 XDP_DROP, 5716 XDP_PASS, 5717 XDP_TX, 5718 XDP_REDIRECT, 5719 }; 5720 5721 /* user accessible metadata for XDP packet hook 5722 * new fields must be added to the end of this structure 5723 */ 5724 struct xdp_md { 5725 __u32 data; 5726 __u32 data_end; 5727 __u32 data_meta; 5728 /* Below access go through struct xdp_rxq_info */ 5729 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 5730 __u32 rx_queue_index; /* rxq->queue_index */ 5731 5732 __u32 egress_ifindex; /* txq->dev->ifindex */ 5733 }; 5734 5735 /* DEVMAP map-value layout 5736 * 5737 * The struct data-layout of map-value is a configuration interface. 5738 * New members can only be added to the end of this structure. 5739 */ 5740 struct bpf_devmap_val { 5741 __u32 ifindex; /* device index */ 5742 union { 5743 int fd; /* prog fd on map write */ 5744 __u32 id; /* prog id on map read */ 5745 } bpf_prog; 5746 }; 5747 5748 /* CPUMAP map-value layout 5749 * 5750 * The struct data-layout of map-value is a configuration interface. 5751 * New members can only be added to the end of this structure. 5752 */ 5753 struct bpf_cpumap_val { 5754 __u32 qsize; /* queue size to remote target CPU */ 5755 union { 5756 int fd; /* prog fd on map write */ 5757 __u32 id; /* prog id on map read */ 5758 } bpf_prog; 5759 }; 5760 5761 enum sk_action { 5762 SK_DROP = 0, 5763 SK_PASS, 5764 }; 5765 5766 /* user accessible metadata for SK_MSG packet hook, new fields must 5767 * be added to the end of this structure 5768 */ 5769 struct sk_msg_md { 5770 __bpf_md_ptr(void *, data); 5771 __bpf_md_ptr(void *, data_end); 5772 5773 __u32 family; 5774 __u32 remote_ip4; /* Stored in network byte order */ 5775 __u32 local_ip4; /* Stored in network byte order */ 5776 __u32 remote_ip6[4]; /* Stored in network byte order */ 5777 __u32 local_ip6[4]; /* Stored in network byte order */ 5778 __u32 remote_port; /* Stored in network byte order */ 5779 __u32 local_port; /* stored in host byte order */ 5780 __u32 size; /* Total size of sk_msg */ 5781 5782 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 5783 }; 5784 5785 struct sk_reuseport_md { 5786 /* 5787 * Start of directly accessible data. It begins from 5788 * the tcp/udp header. 5789 */ 5790 __bpf_md_ptr(void *, data); 5791 /* End of directly accessible data */ 5792 __bpf_md_ptr(void *, data_end); 5793 /* 5794 * Total length of packet (starting from the tcp/udp header). 5795 * Note that the directly accessible bytes (data_end - data) 5796 * could be less than this "len". Those bytes could be 5797 * indirectly read by a helper "bpf_skb_load_bytes()". 5798 */ 5799 __u32 len; 5800 /* 5801 * Eth protocol in the mac header (network byte order). e.g. 5802 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 5803 */ 5804 __u32 eth_protocol; 5805 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 5806 __u32 bind_inany; /* Is sock bound to an INANY address? */ 5807 __u32 hash; /* A hash of the packet 4 tuples */ 5808 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 5809 * new incoming connection request (e.g. selecting a listen sk for 5810 * the received SYN in the TCP case). reuse->sk is one of the sk 5811 * in the reuseport group. The bpf prog can use reuse->sk to learn 5812 * the local listening ip/port without looking into the skb. 5813 * 5814 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 5815 * reuse->migrating_sk is the socket that needs to be migrated 5816 * to another listening socket. migrating_sk could be a fullsock 5817 * sk that is fully established or a reqsk that is in-the-middle 5818 * of 3-way handshake. 5819 */ 5820 __bpf_md_ptr(struct bpf_sock *, sk); 5821 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 5822 }; 5823 5824 #define BPF_TAG_SIZE 8 5825 5826 struct bpf_prog_info { 5827 __u32 type; 5828 __u32 id; 5829 __u8 tag[BPF_TAG_SIZE]; 5830 __u32 jited_prog_len; 5831 __u32 xlated_prog_len; 5832 __aligned_u64 jited_prog_insns; 5833 __aligned_u64 xlated_prog_insns; 5834 __u64 load_time; /* ns since boottime */ 5835 __u32 created_by_uid; 5836 __u32 nr_map_ids; 5837 __aligned_u64 map_ids; 5838 char name[BPF_OBJ_NAME_LEN]; 5839 __u32 ifindex; 5840 __u32 gpl_compatible:1; 5841 __u32 :31; /* alignment pad */ 5842 __u64 netns_dev; 5843 __u64 netns_ino; 5844 __u32 nr_jited_ksyms; 5845 __u32 nr_jited_func_lens; 5846 __aligned_u64 jited_ksyms; 5847 __aligned_u64 jited_func_lens; 5848 __u32 btf_id; 5849 __u32 func_info_rec_size; 5850 __aligned_u64 func_info; 5851 __u32 nr_func_info; 5852 __u32 nr_line_info; 5853 __aligned_u64 line_info; 5854 __aligned_u64 jited_line_info; 5855 __u32 nr_jited_line_info; 5856 __u32 line_info_rec_size; 5857 __u32 jited_line_info_rec_size; 5858 __u32 nr_prog_tags; 5859 __aligned_u64 prog_tags; 5860 __u64 run_time_ns; 5861 __u64 run_cnt; 5862 __u64 recursion_misses; 5863 __u32 verified_insns; 5864 } __attribute__((aligned(8))); 5865 5866 struct bpf_map_info { 5867 __u32 type; 5868 __u32 id; 5869 __u32 key_size; 5870 __u32 value_size; 5871 __u32 max_entries; 5872 __u32 map_flags; 5873 char name[BPF_OBJ_NAME_LEN]; 5874 __u32 ifindex; 5875 __u32 btf_vmlinux_value_type_id; 5876 __u64 netns_dev; 5877 __u64 netns_ino; 5878 __u32 btf_id; 5879 __u32 btf_key_type_id; 5880 __u32 btf_value_type_id; 5881 __u32 :32; /* alignment pad */ 5882 __u64 map_extra; 5883 } __attribute__((aligned(8))); 5884 5885 struct bpf_btf_info { 5886 __aligned_u64 btf; 5887 __u32 btf_size; 5888 __u32 id; 5889 __aligned_u64 name; 5890 __u32 name_len; 5891 __u32 kernel_btf; 5892 } __attribute__((aligned(8))); 5893 5894 struct bpf_link_info { 5895 __u32 type; 5896 __u32 id; 5897 __u32 prog_id; 5898 union { 5899 struct { 5900 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5901 __u32 tp_name_len; /* in/out: tp_name buffer len */ 5902 } raw_tracepoint; 5903 struct { 5904 __u32 attach_type; 5905 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5906 __u32 target_btf_id; /* BTF type id inside the object */ 5907 } tracing; 5908 struct { 5909 __u64 cgroup_id; 5910 __u32 attach_type; 5911 } cgroup; 5912 struct { 5913 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 5914 __u32 target_name_len; /* in/out: target_name buffer len */ 5915 union { 5916 struct { 5917 __u32 map_id; 5918 } map; 5919 }; 5920 } iter; 5921 struct { 5922 __u32 netns_ino; 5923 __u32 attach_type; 5924 } netns; 5925 struct { 5926 __u32 ifindex; 5927 } xdp; 5928 }; 5929 } __attribute__((aligned(8))); 5930 5931 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 5932 * by user and intended to be used by socket (e.g. to bind to, depends on 5933 * attach type). 5934 */ 5935 struct bpf_sock_addr { 5936 __u32 user_family; /* Allows 4-byte read, but no write. */ 5937 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5938 * Stored in network byte order. 5939 */ 5940 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5941 * Stored in network byte order. 5942 */ 5943 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 5944 * Stored in network byte order 5945 */ 5946 __u32 family; /* Allows 4-byte read, but no write */ 5947 __u32 type; /* Allows 4-byte read, but no write */ 5948 __u32 protocol; /* Allows 4-byte read, but no write */ 5949 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5950 * Stored in network byte order. 5951 */ 5952 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5953 * Stored in network byte order. 5954 */ 5955 __bpf_md_ptr(struct bpf_sock *, sk); 5956 }; 5957 5958 /* User bpf_sock_ops struct to access socket values and specify request ops 5959 * and their replies. 5960 * Some of this fields are in network (bigendian) byte order and may need 5961 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 5962 * New fields can only be added at the end of this structure 5963 */ 5964 struct bpf_sock_ops { 5965 __u32 op; 5966 union { 5967 __u32 args[4]; /* Optionally passed to bpf program */ 5968 __u32 reply; /* Returned by bpf program */ 5969 __u32 replylong[4]; /* Optionally returned by bpf prog */ 5970 }; 5971 __u32 family; 5972 __u32 remote_ip4; /* Stored in network byte order */ 5973 __u32 local_ip4; /* Stored in network byte order */ 5974 __u32 remote_ip6[4]; /* Stored in network byte order */ 5975 __u32 local_ip6[4]; /* Stored in network byte order */ 5976 __u32 remote_port; /* Stored in network byte order */ 5977 __u32 local_port; /* stored in host byte order */ 5978 __u32 is_fullsock; /* Some TCP fields are only valid if 5979 * there is a full socket. If not, the 5980 * fields read as zero. 5981 */ 5982 __u32 snd_cwnd; 5983 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5984 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 5985 __u32 state; 5986 __u32 rtt_min; 5987 __u32 snd_ssthresh; 5988 __u32 rcv_nxt; 5989 __u32 snd_nxt; 5990 __u32 snd_una; 5991 __u32 mss_cache; 5992 __u32 ecn_flags; 5993 __u32 rate_delivered; 5994 __u32 rate_interval_us; 5995 __u32 packets_out; 5996 __u32 retrans_out; 5997 __u32 total_retrans; 5998 __u32 segs_in; 5999 __u32 data_segs_in; 6000 __u32 segs_out; 6001 __u32 data_segs_out; 6002 __u32 lost_out; 6003 __u32 sacked_out; 6004 __u32 sk_txhash; 6005 __u64 bytes_received; 6006 __u64 bytes_acked; 6007 __bpf_md_ptr(struct bpf_sock *, sk); 6008 /* [skb_data, skb_data_end) covers the whole TCP header. 6009 * 6010 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6011 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6012 * header has not been written. 6013 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6014 * been written so far. 6015 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6016 * the 3WHS. 6017 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6018 * the 3WHS. 6019 * 6020 * bpf_load_hdr_opt() can also be used to read a particular option. 6021 */ 6022 __bpf_md_ptr(void *, skb_data); 6023 __bpf_md_ptr(void *, skb_data_end); 6024 __u32 skb_len; /* The total length of a packet. 6025 * It includes the header, options, 6026 * and payload. 6027 */ 6028 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6029 * an easy way to check for tcp_flags 6030 * without parsing skb_data. 6031 * 6032 * In particular, the skb_tcp_flags 6033 * will still be available in 6034 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6035 * the outgoing header has not 6036 * been written yet. 6037 */ 6038 }; 6039 6040 /* Definitions for bpf_sock_ops_cb_flags */ 6041 enum { 6042 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6043 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6044 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6045 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6046 /* Call bpf for all received TCP headers. The bpf prog will be 6047 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6048 * 6049 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6050 * for the header option related helpers that will be useful 6051 * to the bpf programs. 6052 * 6053 * It could be used at the client/active side (i.e. connect() side) 6054 * when the server told it that the server was in syncookie 6055 * mode and required the active side to resend the bpf-written 6056 * options. The active side can keep writing the bpf-options until 6057 * it received a valid packet from the server side to confirm 6058 * the earlier packet (and options) has been received. The later 6059 * example patch is using it like this at the active side when the 6060 * server is in syncookie mode. 6061 * 6062 * The bpf prog will usually turn this off in the common cases. 6063 */ 6064 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6065 /* Call bpf when kernel has received a header option that 6066 * the kernel cannot handle. The bpf prog will be called under 6067 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6068 * 6069 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6070 * for the header option related helpers that will be useful 6071 * to the bpf programs. 6072 */ 6073 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6074 /* Call bpf when the kernel is writing header options for the 6075 * outgoing packet. The bpf prog will first be called 6076 * to reserve space in a skb under 6077 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6078 * the bpf prog will be called to write the header option(s) 6079 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6080 * 6081 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6082 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6083 * related helpers that will be useful to the bpf programs. 6084 * 6085 * The kernel gets its chance to reserve space and write 6086 * options first before the BPF program does. 6087 */ 6088 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6089 /* Mask of all currently supported cb flags */ 6090 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6091 }; 6092 6093 /* List of known BPF sock_ops operators. 6094 * New entries can only be added at the end 6095 */ 6096 enum { 6097 BPF_SOCK_OPS_VOID, 6098 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6099 * -1 if default value should be used 6100 */ 6101 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6102 * window (in packets) or -1 if default 6103 * value should be used 6104 */ 6105 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6106 * active connection is initialized 6107 */ 6108 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6109 * active connection is 6110 * established 6111 */ 6112 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6113 * passive connection is 6114 * established 6115 */ 6116 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6117 * needs ECN 6118 */ 6119 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6120 * based on the path and may be 6121 * dependent on the congestion control 6122 * algorithm. In general it indicates 6123 * a congestion threshold. RTTs above 6124 * this indicate congestion 6125 */ 6126 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6127 * Arg1: value of icsk_retransmits 6128 * Arg2: value of icsk_rto 6129 * Arg3: whether RTO has expired 6130 */ 6131 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6132 * Arg1: sequence number of 1st byte 6133 * Arg2: # segments 6134 * Arg3: return value of 6135 * tcp_transmit_skb (0 => success) 6136 */ 6137 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6138 * Arg1: old_state 6139 * Arg2: new_state 6140 */ 6141 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6142 * socket transition to LISTEN state. 6143 */ 6144 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6145 */ 6146 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6147 * It will be called to handle 6148 * the packets received at 6149 * an already established 6150 * connection. 6151 * 6152 * sock_ops->skb_data: 6153 * Referring to the received skb. 6154 * It covers the TCP header only. 6155 * 6156 * bpf_load_hdr_opt() can also 6157 * be used to search for a 6158 * particular option. 6159 */ 6160 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6161 * header option later in 6162 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6163 * Arg1: bool want_cookie. (in 6164 * writing SYNACK only) 6165 * 6166 * sock_ops->skb_data: 6167 * Not available because no header has 6168 * been written yet. 6169 * 6170 * sock_ops->skb_tcp_flags: 6171 * The tcp_flags of the 6172 * outgoing skb. (e.g. SYN, ACK, FIN). 6173 * 6174 * bpf_reserve_hdr_opt() should 6175 * be used to reserve space. 6176 */ 6177 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6178 * Arg1: bool want_cookie. (in 6179 * writing SYNACK only) 6180 * 6181 * sock_ops->skb_data: 6182 * Referring to the outgoing skb. 6183 * It covers the TCP header 6184 * that has already been written 6185 * by the kernel and the 6186 * earlier bpf-progs. 6187 * 6188 * sock_ops->skb_tcp_flags: 6189 * The tcp_flags of the outgoing 6190 * skb. (e.g. SYN, ACK, FIN). 6191 * 6192 * bpf_store_hdr_opt() should 6193 * be used to write the 6194 * option. 6195 * 6196 * bpf_load_hdr_opt() can also 6197 * be used to search for a 6198 * particular option that 6199 * has already been written 6200 * by the kernel or the 6201 * earlier bpf-progs. 6202 */ 6203 }; 6204 6205 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6206 * changes between the TCP and BPF versions. Ideally this should never happen. 6207 * If it does, we need to add code to convert them before calling 6208 * the BPF sock_ops function. 6209 */ 6210 enum { 6211 BPF_TCP_ESTABLISHED = 1, 6212 BPF_TCP_SYN_SENT, 6213 BPF_TCP_SYN_RECV, 6214 BPF_TCP_FIN_WAIT1, 6215 BPF_TCP_FIN_WAIT2, 6216 BPF_TCP_TIME_WAIT, 6217 BPF_TCP_CLOSE, 6218 BPF_TCP_CLOSE_WAIT, 6219 BPF_TCP_LAST_ACK, 6220 BPF_TCP_LISTEN, 6221 BPF_TCP_CLOSING, /* Now a valid state */ 6222 BPF_TCP_NEW_SYN_RECV, 6223 6224 BPF_TCP_MAX_STATES /* Leave at the end! */ 6225 }; 6226 6227 enum { 6228 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6229 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6230 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6231 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6232 /* Copy the SYN pkt to optval 6233 * 6234 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6235 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6236 * to only getting from the saved_syn. It can either get the 6237 * syn packet from: 6238 * 6239 * 1. the just-received SYN packet (only available when writing the 6240 * SYNACK). It will be useful when it is not necessary to 6241 * save the SYN packet for latter use. It is also the only way 6242 * to get the SYN during syncookie mode because the syn 6243 * packet cannot be saved during syncookie. 6244 * 6245 * OR 6246 * 6247 * 2. the earlier saved syn which was done by 6248 * bpf_setsockopt(TCP_SAVE_SYN). 6249 * 6250 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6251 * SYN packet is obtained. 6252 * 6253 * If the bpf-prog does not need the IP[46] header, the 6254 * bpf-prog can avoid parsing the IP header by using 6255 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6256 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6257 * 6258 * >0: Total number of bytes copied 6259 * -ENOSPC: Not enough space in optval. Only optlen number of 6260 * bytes is copied. 6261 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6262 * is not saved by setsockopt(TCP_SAVE_SYN). 6263 */ 6264 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6265 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6266 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6267 }; 6268 6269 enum { 6270 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6271 }; 6272 6273 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6274 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6275 */ 6276 enum { 6277 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6278 * total option spaces 6279 * required for an established 6280 * sk in order to calculate the 6281 * MSS. No skb is actually 6282 * sent. 6283 */ 6284 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6285 * when sending a SYN. 6286 */ 6287 }; 6288 6289 struct bpf_perf_event_value { 6290 __u64 counter; 6291 __u64 enabled; 6292 __u64 running; 6293 }; 6294 6295 enum { 6296 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6297 BPF_DEVCG_ACC_READ = (1ULL << 1), 6298 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 6299 }; 6300 6301 enum { 6302 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 6303 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 6304 }; 6305 6306 struct bpf_cgroup_dev_ctx { 6307 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 6308 __u32 access_type; 6309 __u32 major; 6310 __u32 minor; 6311 }; 6312 6313 struct bpf_raw_tracepoint_args { 6314 __u64 args[0]; 6315 }; 6316 6317 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 6318 * OUTPUT: Do lookup from egress perspective; default is ingress 6319 */ 6320 enum { 6321 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6322 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6323 }; 6324 6325 enum { 6326 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6327 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6328 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6329 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6330 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6331 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6332 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6333 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6334 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6335 }; 6336 6337 struct bpf_fib_lookup { 6338 /* input: network family for lookup (AF_INET, AF_INET6) 6339 * output: network family of egress nexthop 6340 */ 6341 __u8 family; 6342 6343 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6344 __u8 l4_protocol; 6345 __be16 sport; 6346 __be16 dport; 6347 6348 union { /* used for MTU check */ 6349 /* input to lookup */ 6350 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6351 6352 /* output: MTU value */ 6353 __u16 mtu_result; 6354 }; 6355 /* input: L3 device index for lookup 6356 * output: device index from FIB lookup 6357 */ 6358 __u32 ifindex; 6359 6360 union { 6361 /* inputs to lookup */ 6362 __u8 tos; /* AF_INET */ 6363 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6364 6365 /* output: metric of fib result (IPv4/IPv6 only) */ 6366 __u32 rt_metric; 6367 }; 6368 6369 union { 6370 __be32 ipv4_src; 6371 __u32 ipv6_src[4]; /* in6_addr; network order */ 6372 }; 6373 6374 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6375 * network header. output: bpf_fib_lookup sets to gateway address 6376 * if FIB lookup returns gateway route 6377 */ 6378 union { 6379 __be32 ipv4_dst; 6380 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6381 }; 6382 6383 /* output */ 6384 __be16 h_vlan_proto; 6385 __be16 h_vlan_TCI; 6386 __u8 smac[6]; /* ETH_ALEN */ 6387 __u8 dmac[6]; /* ETH_ALEN */ 6388 }; 6389 6390 struct bpf_redir_neigh { 6391 /* network family for lookup (AF_INET, AF_INET6) */ 6392 __u32 nh_family; 6393 /* network address of nexthop; skips fib lookup to find gateway */ 6394 union { 6395 __be32 ipv4_nh; 6396 __u32 ipv6_nh[4]; /* in6_addr; network order */ 6397 }; 6398 }; 6399 6400 /* bpf_check_mtu flags*/ 6401 enum bpf_check_mtu_flags { 6402 BPF_MTU_CHK_SEGS = (1U << 0), 6403 }; 6404 6405 enum bpf_check_mtu_ret { 6406 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 6407 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6408 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 6409 }; 6410 6411 enum bpf_task_fd_type { 6412 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 6413 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 6414 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 6415 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 6416 BPF_FD_TYPE_UPROBE, /* filename + offset */ 6417 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 6418 }; 6419 6420 enum { 6421 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 6422 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 6423 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 6424 }; 6425 6426 struct bpf_flow_keys { 6427 __u16 nhoff; 6428 __u16 thoff; 6429 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6430 __u8 is_frag; 6431 __u8 is_first_frag; 6432 __u8 is_encap; 6433 __u8 ip_proto; 6434 __be16 n_proto; 6435 __be16 sport; 6436 __be16 dport; 6437 union { 6438 struct { 6439 __be32 ipv4_src; 6440 __be32 ipv4_dst; 6441 }; 6442 struct { 6443 __u32 ipv6_src[4]; /* in6_addr; network order */ 6444 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6445 }; 6446 }; 6447 __u32 flags; 6448 __be32 flow_label; 6449 }; 6450 6451 struct bpf_func_info { 6452 __u32 insn_off; 6453 __u32 type_id; 6454 }; 6455 6456 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6457 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6458 6459 struct bpf_line_info { 6460 __u32 insn_off; 6461 __u32 file_name_off; 6462 __u32 line_off; 6463 __u32 line_col; 6464 }; 6465 6466 struct bpf_spin_lock { 6467 __u32 val; 6468 }; 6469 6470 struct bpf_timer { 6471 __u64 :64; 6472 __u64 :64; 6473 } __attribute__((aligned(8))); 6474 6475 struct bpf_sysctl { 6476 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 6477 * Allows 1,2,4-byte read, but no write. 6478 */ 6479 __u32 file_pos; /* Sysctl file position to read from, write to. 6480 * Allows 1,2,4-byte read an 4-byte write. 6481 */ 6482 }; 6483 6484 struct bpf_sockopt { 6485 __bpf_md_ptr(struct bpf_sock *, sk); 6486 __bpf_md_ptr(void *, optval); 6487 __bpf_md_ptr(void *, optval_end); 6488 6489 __s32 level; 6490 __s32 optname; 6491 __s32 optlen; 6492 __s32 retval; 6493 }; 6494 6495 struct bpf_pidns_info { 6496 __u32 pid; 6497 __u32 tgid; 6498 }; 6499 6500 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6501 struct bpf_sk_lookup { 6502 union { 6503 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 6504 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 6505 }; 6506 6507 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6508 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6509 __u32 remote_ip4; /* Network byte order */ 6510 __u32 remote_ip6[4]; /* Network byte order */ 6511 __be16 remote_port; /* Network byte order */ 6512 __u16 :16; /* Zero padding */ 6513 __u32 local_ip4; /* Network byte order */ 6514 __u32 local_ip6[4]; /* Network byte order */ 6515 __u32 local_port; /* Host byte order */ 6516 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 6517 }; 6518 6519 /* 6520 * struct btf_ptr is used for typed pointer representation; the 6521 * type id is used to render the pointer data as the appropriate type 6522 * via the bpf_snprintf_btf() helper described above. A flags field - 6523 * potentially to specify additional details about the BTF pointer 6524 * (rather than its mode of display) - is included for future use. 6525 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6526 */ 6527 struct btf_ptr { 6528 void *ptr; 6529 __u32 type_id; 6530 __u32 flags; /* BTF ptr flags; unused at present. */ 6531 }; 6532 6533 /* 6534 * Flags to control bpf_snprintf_btf() behaviour. 6535 * - BTF_F_COMPACT: no formatting around type information 6536 * - BTF_F_NONAME: no struct/union member names/types 6537 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6538 * equivalent to %px. 6539 * - BTF_F_ZERO: show zero-valued struct/union members; they 6540 * are not displayed by default 6541 */ 6542 enum { 6543 BTF_F_COMPACT = (1ULL << 0), 6544 BTF_F_NONAME = (1ULL << 1), 6545 BTF_F_PTR_RAW = (1ULL << 2), 6546 BTF_F_ZERO = (1ULL << 3), 6547 }; 6548 6549 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 6550 * has to be adjusted by relocations. It is emitted by llvm and passed to 6551 * libbpf and later to the kernel. 6552 */ 6553 enum bpf_core_relo_kind { 6554 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 6555 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 6556 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 6557 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 6558 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 6559 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 6560 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 6561 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 6562 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 6563 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 6564 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 6565 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 6566 }; 6567 6568 /* 6569 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 6570 * and from libbpf to the kernel. 6571 * 6572 * CO-RE relocation captures the following data: 6573 * - insn_off - instruction offset (in bytes) within a BPF program that needs 6574 * its insn->imm field to be relocated with actual field info; 6575 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 6576 * type or field; 6577 * - access_str_off - offset into corresponding .BTF string section. String 6578 * interpretation depends on specific relocation kind: 6579 * - for field-based relocations, string encodes an accessed field using 6580 * a sequence of field and array indices, separated by colon (:). It's 6581 * conceptually very close to LLVM's getelementptr ([0]) instruction's 6582 * arguments for identifying offset to a field. 6583 * - for type-based relocations, strings is expected to be just "0"; 6584 * - for enum value-based relocations, string contains an index of enum 6585 * value within its enum type; 6586 * - kind - one of enum bpf_core_relo_kind; 6587 * 6588 * Example: 6589 * struct sample { 6590 * int a; 6591 * struct { 6592 * int b[10]; 6593 * }; 6594 * }; 6595 * 6596 * struct sample *s = ...; 6597 * int *x = &s->a; // encoded as "0:0" (a is field #0) 6598 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 6599 * // b is field #0 inside anon struct, accessing elem #5) 6600 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 6601 * 6602 * type_id for all relocs in this example will capture BTF type id of 6603 * `struct sample`. 6604 * 6605 * Such relocation is emitted when using __builtin_preserve_access_index() 6606 * Clang built-in, passing expression that captures field address, e.g.: 6607 * 6608 * bpf_probe_read(&dst, sizeof(dst), 6609 * __builtin_preserve_access_index(&src->a.b.c)); 6610 * 6611 * In this case Clang will emit field relocation recording necessary data to 6612 * be able to find offset of embedded `a.b.c` field within `src` struct. 6613 * 6614 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 6615 */ 6616 struct bpf_core_relo { 6617 __u32 insn_off; 6618 __u32 type_id; 6619 __u32 access_str_off; 6620 enum bpf_core_relo_kind kind; 6621 }; 6622 6623 #endif /* _UAPI__LINUX_BPF_H__ */ 6624