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