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