1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25 /* alu/jmp fields */ 26 #define BPF_MOV 0xb0 /* mov reg to reg */ 27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29 /* change endianness of a register */ 30 #define BPF_END 0xd0 /* flags for endianness conversion: */ 31 #define BPF_TO_LE 0x00 /* convert to little-endian */ 32 #define BPF_TO_BE 0x08 /* convert to big-endian */ 33 #define BPF_FROM_LE BPF_TO_LE 34 #define BPF_FROM_BE BPF_TO_BE 35 36 /* jmp encodings */ 37 #define BPF_JNE 0x50 /* jump != */ 38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44 #define BPF_CALL 0x80 /* function call */ 45 #define BPF_EXIT 0x90 /* function return */ 46 47 /* atomic op type fields (stored in immediate) */ 48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52 /* Register numbers */ 53 enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66 }; 67 68 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 69 #define MAX_BPF_REG __MAX_BPF_REG 70 71 struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77 }; 78 79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80 struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83 }; 84 85 struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 88 }; 89 90 enum bpf_cgroup_iter_order { 91 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 92 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 93 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 94 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 95 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 96 }; 97 98 union bpf_iter_link_info { 99 struct { 100 __u32 map_fd; 101 } map; 102 struct { 103 enum bpf_cgroup_iter_order order; 104 105 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 106 * both are zero, the walk starts from the default cgroup v2 107 * root. For walking v1 hierarchy, one should always explicitly 108 * specify cgroup_fd. 109 */ 110 __u32 cgroup_fd; 111 __u64 cgroup_id; 112 } cgroup; 113 }; 114 115 /* BPF syscall commands, see bpf(2) man-page for more details. */ 116 /** 117 * DOC: eBPF Syscall Preamble 118 * 119 * The operation to be performed by the **bpf**\ () system call is determined 120 * by the *cmd* argument. Each operation takes an accompanying argument, 121 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 122 * below). The size argument is the size of the union pointed to by *attr*. 123 */ 124 /** 125 * DOC: eBPF Syscall Commands 126 * 127 * BPF_MAP_CREATE 128 * Description 129 * Create a map and return a file descriptor that refers to the 130 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 131 * is automatically enabled for the new file descriptor. 132 * 133 * Applying **close**\ (2) to the file descriptor returned by 134 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 135 * 136 * Return 137 * A new file descriptor (a nonnegative integer), or -1 if an 138 * error occurred (in which case, *errno* is set appropriately). 139 * 140 * BPF_MAP_LOOKUP_ELEM 141 * Description 142 * Look up an element with a given *key* in the map referred to 143 * by the file descriptor *map_fd*. 144 * 145 * The *flags* argument may be specified as one of the 146 * following: 147 * 148 * **BPF_F_LOCK** 149 * Look up the value of a spin-locked map without 150 * returning the lock. This must be specified if the 151 * elements contain a spinlock. 152 * 153 * Return 154 * Returns zero on success. On error, -1 is returned and *errno* 155 * is set appropriately. 156 * 157 * BPF_MAP_UPDATE_ELEM 158 * Description 159 * Create or update an element (key/value pair) in a specified map. 160 * 161 * The *flags* argument should be specified as one of the 162 * following: 163 * 164 * **BPF_ANY** 165 * Create a new element or update an existing element. 166 * **BPF_NOEXIST** 167 * Create a new element only if it did not exist. 168 * **BPF_EXIST** 169 * Update an existing element. 170 * **BPF_F_LOCK** 171 * Update a spin_lock-ed map element. 172 * 173 * Return 174 * Returns zero on success. On error, -1 is returned and *errno* 175 * is set appropriately. 176 * 177 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 178 * **E2BIG**, **EEXIST**, or **ENOENT**. 179 * 180 * **E2BIG** 181 * The number of elements in the map reached the 182 * *max_entries* limit specified at map creation time. 183 * **EEXIST** 184 * If *flags* specifies **BPF_NOEXIST** and the element 185 * with *key* already exists in the map. 186 * **ENOENT** 187 * If *flags* specifies **BPF_EXIST** and the element with 188 * *key* does not exist in the map. 189 * 190 * BPF_MAP_DELETE_ELEM 191 * Description 192 * Look up and delete an element by key in a specified map. 193 * 194 * Return 195 * Returns zero on success. On error, -1 is returned and *errno* 196 * is set appropriately. 197 * 198 * BPF_MAP_GET_NEXT_KEY 199 * Description 200 * Look up an element by key in a specified map and return the key 201 * of the next element. Can be used to iterate over all elements 202 * in the map. 203 * 204 * Return 205 * Returns zero on success. On error, -1 is returned and *errno* 206 * is set appropriately. 207 * 208 * The following cases can be used to iterate over all elements of 209 * the map: 210 * 211 * * If *key* is not found, the operation returns zero and sets 212 * the *next_key* pointer to the key of the first element. 213 * * If *key* is found, the operation returns zero and sets the 214 * *next_key* pointer to the key of the next element. 215 * * If *key* is the last element, returns -1 and *errno* is set 216 * to **ENOENT**. 217 * 218 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 219 * **EINVAL** on error. 220 * 221 * BPF_PROG_LOAD 222 * Description 223 * Verify and load an eBPF program, returning a new file 224 * descriptor associated with the program. 225 * 226 * Applying **close**\ (2) to the file descriptor returned by 227 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 228 * 229 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 230 * automatically enabled for the new file descriptor. 231 * 232 * Return 233 * A new file descriptor (a nonnegative integer), or -1 if an 234 * error occurred (in which case, *errno* is set appropriately). 235 * 236 * BPF_OBJ_PIN 237 * Description 238 * Pin an eBPF program or map referred by the specified *bpf_fd* 239 * to the provided *pathname* on the filesystem. 240 * 241 * The *pathname* argument must not contain a dot ("."). 242 * 243 * On success, *pathname* retains a reference to the eBPF object, 244 * preventing deallocation of the object when the original 245 * *bpf_fd* is closed. This allow the eBPF object to live beyond 246 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 247 * process. 248 * 249 * Applying **unlink**\ (2) or similar calls to the *pathname* 250 * unpins the object from the filesystem, removing the reference. 251 * If no other file descriptors or filesystem nodes refer to the 252 * same object, it will be deallocated (see NOTES). 253 * 254 * The filesystem type for the parent directory of *pathname* must 255 * be **BPF_FS_MAGIC**. 256 * 257 * Return 258 * Returns zero on success. On error, -1 is returned and *errno* 259 * is set appropriately. 260 * 261 * BPF_OBJ_GET 262 * Description 263 * Open a file descriptor for the eBPF object pinned to the 264 * specified *pathname*. 265 * 266 * Return 267 * A new file descriptor (a nonnegative integer), or -1 if an 268 * error occurred (in which case, *errno* is set appropriately). 269 * 270 * BPF_PROG_ATTACH 271 * Description 272 * Attach an eBPF program to a *target_fd* at the specified 273 * *attach_type* hook. 274 * 275 * The *attach_type* specifies the eBPF attachment point to 276 * attach the program to, and must be one of *bpf_attach_type* 277 * (see below). 278 * 279 * The *attach_bpf_fd* must be a valid file descriptor for a 280 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 281 * or sock_ops type corresponding to the specified *attach_type*. 282 * 283 * The *target_fd* must be a valid file descriptor for a kernel 284 * object which depends on the attach type of *attach_bpf_fd*: 285 * 286 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 287 * **BPF_PROG_TYPE_CGROUP_SKB**, 288 * **BPF_PROG_TYPE_CGROUP_SOCK**, 289 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 290 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 291 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 292 * **BPF_PROG_TYPE_SOCK_OPS** 293 * 294 * Control Group v2 hierarchy with the eBPF controller 295 * enabled. Requires the kernel to be compiled with 296 * **CONFIG_CGROUP_BPF**. 297 * 298 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 299 * 300 * Network namespace (eg /proc/self/ns/net). 301 * 302 * **BPF_PROG_TYPE_LIRC_MODE2** 303 * 304 * LIRC device path (eg /dev/lircN). Requires the kernel 305 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 306 * 307 * **BPF_PROG_TYPE_SK_SKB**, 308 * **BPF_PROG_TYPE_SK_MSG** 309 * 310 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 311 * 312 * Return 313 * Returns zero on success. On error, -1 is returned and *errno* 314 * is set appropriately. 315 * 316 * BPF_PROG_DETACH 317 * Description 318 * Detach the eBPF program associated with the *target_fd* at the 319 * hook specified by *attach_type*. The program must have been 320 * previously attached using **BPF_PROG_ATTACH**. 321 * 322 * Return 323 * Returns zero on success. On error, -1 is returned and *errno* 324 * is set appropriately. 325 * 326 * BPF_PROG_TEST_RUN 327 * Description 328 * Run the eBPF program associated with the *prog_fd* a *repeat* 329 * number of times against a provided program context *ctx_in* and 330 * data *data_in*, and return the modified program context 331 * *ctx_out*, *data_out* (for example, packet data), result of the 332 * execution *retval*, and *duration* of the test run. 333 * 334 * The sizes of the buffers provided as input and output 335 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 336 * be provided in the corresponding variables *ctx_size_in*, 337 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 338 * of these parameters are not provided (ie set to NULL), the 339 * corresponding size field must be zero. 340 * 341 * Some program types have particular requirements: 342 * 343 * **BPF_PROG_TYPE_SK_LOOKUP** 344 * *data_in* and *data_out* must be NULL. 345 * 346 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 347 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 348 * 349 * *ctx_out*, *data_in* and *data_out* must be NULL. 350 * *repeat* must be zero. 351 * 352 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 353 * 354 * Return 355 * Returns zero on success. On error, -1 is returned and *errno* 356 * is set appropriately. 357 * 358 * **ENOSPC** 359 * Either *data_size_out* or *ctx_size_out* is too small. 360 * **ENOTSUPP** 361 * This command is not supported by the program type of 362 * the program referred to by *prog_fd*. 363 * 364 * BPF_PROG_GET_NEXT_ID 365 * Description 366 * Fetch the next eBPF program currently loaded into the kernel. 367 * 368 * Looks for the eBPF program with an id greater than *start_id* 369 * and updates *next_id* on success. If no other eBPF programs 370 * remain with ids higher than *start_id*, returns -1 and sets 371 * *errno* to **ENOENT**. 372 * 373 * Return 374 * Returns zero on success. On error, or when no id remains, -1 375 * is returned and *errno* is set appropriately. 376 * 377 * BPF_MAP_GET_NEXT_ID 378 * Description 379 * Fetch the next eBPF map currently loaded into the kernel. 380 * 381 * Looks for the eBPF map with an id greater than *start_id* 382 * and updates *next_id* on success. If no other eBPF maps 383 * remain with ids higher than *start_id*, returns -1 and sets 384 * *errno* to **ENOENT**. 385 * 386 * Return 387 * Returns zero on success. On error, or when no id remains, -1 388 * is returned and *errno* is set appropriately. 389 * 390 * BPF_PROG_GET_FD_BY_ID 391 * Description 392 * Open a file descriptor for the eBPF program corresponding to 393 * *prog_id*. 394 * 395 * Return 396 * A new file descriptor (a nonnegative integer), or -1 if an 397 * error occurred (in which case, *errno* is set appropriately). 398 * 399 * BPF_MAP_GET_FD_BY_ID 400 * Description 401 * Open a file descriptor for the eBPF map corresponding to 402 * *map_id*. 403 * 404 * Return 405 * A new file descriptor (a nonnegative integer), or -1 if an 406 * error occurred (in which case, *errno* is set appropriately). 407 * 408 * BPF_OBJ_GET_INFO_BY_FD 409 * Description 410 * Obtain information about the eBPF object corresponding to 411 * *bpf_fd*. 412 * 413 * Populates up to *info_len* bytes of *info*, which will be in 414 * one of the following formats depending on the eBPF object type 415 * of *bpf_fd*: 416 * 417 * * **struct bpf_prog_info** 418 * * **struct bpf_map_info** 419 * * **struct bpf_btf_info** 420 * * **struct bpf_link_info** 421 * 422 * Return 423 * Returns zero on success. On error, -1 is returned and *errno* 424 * is set appropriately. 425 * 426 * BPF_PROG_QUERY 427 * Description 428 * Obtain information about eBPF programs associated with the 429 * specified *attach_type* hook. 430 * 431 * The *target_fd* must be a valid file descriptor for a kernel 432 * object which depends on the attach type of *attach_bpf_fd*: 433 * 434 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 435 * **BPF_PROG_TYPE_CGROUP_SKB**, 436 * **BPF_PROG_TYPE_CGROUP_SOCK**, 437 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 438 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 439 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 440 * **BPF_PROG_TYPE_SOCK_OPS** 441 * 442 * Control Group v2 hierarchy with the eBPF controller 443 * enabled. Requires the kernel to be compiled with 444 * **CONFIG_CGROUP_BPF**. 445 * 446 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 447 * 448 * Network namespace (eg /proc/self/ns/net). 449 * 450 * **BPF_PROG_TYPE_LIRC_MODE2** 451 * 452 * LIRC device path (eg /dev/lircN). Requires the kernel 453 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 454 * 455 * **BPF_PROG_QUERY** always fetches the number of programs 456 * attached and the *attach_flags* which were used to attach those 457 * programs. Additionally, if *prog_ids* is nonzero and the number 458 * of attached programs is less than *prog_cnt*, populates 459 * *prog_ids* with the eBPF program ids of the programs attached 460 * at *target_fd*. 461 * 462 * The following flags may alter the result: 463 * 464 * **BPF_F_QUERY_EFFECTIVE** 465 * Only return information regarding programs which are 466 * currently effective at the specified *target_fd*. 467 * 468 * Return 469 * Returns zero on success. On error, -1 is returned and *errno* 470 * is set appropriately. 471 * 472 * BPF_RAW_TRACEPOINT_OPEN 473 * Description 474 * Attach an eBPF program to a tracepoint *name* to access kernel 475 * internal arguments of the tracepoint in their raw form. 476 * 477 * The *prog_fd* must be a valid file descriptor associated with 478 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 479 * 480 * No ABI guarantees are made about the content of tracepoint 481 * arguments exposed to the corresponding eBPF program. 482 * 483 * Applying **close**\ (2) to the file descriptor returned by 484 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 485 * 486 * Return 487 * A new file descriptor (a nonnegative integer), or -1 if an 488 * error occurred (in which case, *errno* is set appropriately). 489 * 490 * BPF_BTF_LOAD 491 * Description 492 * Verify and load BPF Type Format (BTF) metadata into the kernel, 493 * returning a new file descriptor associated with the metadata. 494 * BTF is described in more detail at 495 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 496 * 497 * The *btf* parameter must point to valid memory providing 498 * *btf_size* bytes of BTF binary metadata. 499 * 500 * The returned file descriptor can be passed to other **bpf**\ () 501 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 502 * associate the BTF with those objects. 503 * 504 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 505 * parameters to specify a *btf_log_buf*, *btf_log_size* and 506 * *btf_log_level* which allow the kernel to return freeform log 507 * output regarding the BTF verification process. 508 * 509 * Return 510 * A new file descriptor (a nonnegative integer), or -1 if an 511 * error occurred (in which case, *errno* is set appropriately). 512 * 513 * BPF_BTF_GET_FD_BY_ID 514 * Description 515 * Open a file descriptor for the BPF Type Format (BTF) 516 * corresponding to *btf_id*. 517 * 518 * Return 519 * A new file descriptor (a nonnegative integer), or -1 if an 520 * error occurred (in which case, *errno* is set appropriately). 521 * 522 * BPF_TASK_FD_QUERY 523 * Description 524 * Obtain information about eBPF programs associated with the 525 * target process identified by *pid* and *fd*. 526 * 527 * If the *pid* and *fd* are associated with a tracepoint, kprobe 528 * or uprobe perf event, then the *prog_id* and *fd_type* will 529 * be populated with the eBPF program id and file descriptor type 530 * of type **bpf_task_fd_type**. If associated with a kprobe or 531 * uprobe, the *probe_offset* and *probe_addr* will also be 532 * populated. Optionally, if *buf* is provided, then up to 533 * *buf_len* bytes of *buf* will be populated with the name of 534 * the tracepoint, kprobe or uprobe. 535 * 536 * The resulting *prog_id* may be introspected in deeper detail 537 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 538 * 539 * Return 540 * Returns zero on success. On error, -1 is returned and *errno* 541 * is set appropriately. 542 * 543 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 544 * Description 545 * Look up an element with the given *key* in the map referred to 546 * by the file descriptor *fd*, and if found, delete the element. 547 * 548 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 549 * types, the *flags* argument needs to be set to 0, but for other 550 * map types, it may be specified as: 551 * 552 * **BPF_F_LOCK** 553 * Look up and delete the value of a spin-locked map 554 * without returning the lock. This must be specified if 555 * the elements contain a spinlock. 556 * 557 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 558 * implement this command as a "pop" operation, deleting the top 559 * element rather than one corresponding to *key*. 560 * The *key* and *key_len* parameters should be zeroed when 561 * issuing this operation for these map types. 562 * 563 * This command is only valid for the following map types: 564 * * **BPF_MAP_TYPE_QUEUE** 565 * * **BPF_MAP_TYPE_STACK** 566 * * **BPF_MAP_TYPE_HASH** 567 * * **BPF_MAP_TYPE_PERCPU_HASH** 568 * * **BPF_MAP_TYPE_LRU_HASH** 569 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 570 * 571 * Return 572 * Returns zero on success. On error, -1 is returned and *errno* 573 * is set appropriately. 574 * 575 * BPF_MAP_FREEZE 576 * Description 577 * Freeze the permissions of the specified map. 578 * 579 * Write permissions may be frozen by passing zero *flags*. 580 * Upon success, no future syscall invocations may alter the 581 * map state of *map_fd*. Write operations from eBPF programs 582 * are still possible for a frozen map. 583 * 584 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 585 * 586 * Return 587 * Returns zero on success. On error, -1 is returned and *errno* 588 * is set appropriately. 589 * 590 * BPF_BTF_GET_NEXT_ID 591 * Description 592 * Fetch the next BPF Type Format (BTF) object currently loaded 593 * into the kernel. 594 * 595 * Looks for the BTF object with an id greater than *start_id* 596 * and updates *next_id* on success. If no other BTF objects 597 * remain with ids higher than *start_id*, returns -1 and sets 598 * *errno* to **ENOENT**. 599 * 600 * Return 601 * Returns zero on success. On error, or when no id remains, -1 602 * is returned and *errno* is set appropriately. 603 * 604 * BPF_MAP_LOOKUP_BATCH 605 * Description 606 * Iterate and fetch multiple elements in a map. 607 * 608 * Two opaque values are used to manage batch operations, 609 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 610 * to NULL to begin the batched operation. After each subsequent 611 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 612 * *out_batch* as the *in_batch* for the next operation to 613 * continue iteration from the current point. 614 * 615 * The *keys* and *values* are output parameters which must point 616 * to memory large enough to hold *count* items based on the key 617 * and value size of the map *map_fd*. The *keys* buffer must be 618 * of *key_size* * *count*. The *values* buffer must be of 619 * *value_size* * *count*. 620 * 621 * The *elem_flags* argument may be specified as one of the 622 * following: 623 * 624 * **BPF_F_LOCK** 625 * Look up the value of a spin-locked map without 626 * returning the lock. This must be specified if the 627 * elements contain a spinlock. 628 * 629 * On success, *count* elements from the map are copied into the 630 * user buffer, with the keys copied into *keys* and the values 631 * copied into the corresponding indices in *values*. 632 * 633 * If an error is returned and *errno* is not **EFAULT**, *count* 634 * is set to the number of successfully processed elements. 635 * 636 * Return 637 * Returns zero on success. On error, -1 is returned and *errno* 638 * is set appropriately. 639 * 640 * May set *errno* to **ENOSPC** to indicate that *keys* or 641 * *values* is too small to dump an entire bucket during 642 * iteration of a hash-based map type. 643 * 644 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 645 * Description 646 * Iterate and delete all elements in a map. 647 * 648 * This operation has the same behavior as 649 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 650 * 651 * * Every element that is successfully returned is also deleted 652 * from the map. This is at least *count* elements. Note that 653 * *count* is both an input and an output parameter. 654 * * Upon returning with *errno* set to **EFAULT**, up to 655 * *count* elements may be deleted without returning the keys 656 * and values of the deleted elements. 657 * 658 * Return 659 * Returns zero on success. On error, -1 is returned and *errno* 660 * is set appropriately. 661 * 662 * BPF_MAP_UPDATE_BATCH 663 * Description 664 * Update multiple elements in a map by *key*. 665 * 666 * The *keys* and *values* are input parameters which must point 667 * to memory large enough to hold *count* items based on the key 668 * and value size of the map *map_fd*. The *keys* buffer must be 669 * of *key_size* * *count*. The *values* buffer must be of 670 * *value_size* * *count*. 671 * 672 * Each element specified in *keys* is sequentially updated to the 673 * value in the corresponding index in *values*. The *in_batch* 674 * and *out_batch* parameters are ignored and should be zeroed. 675 * 676 * The *elem_flags* argument should be specified as one of the 677 * following: 678 * 679 * **BPF_ANY** 680 * Create new elements or update a existing elements. 681 * **BPF_NOEXIST** 682 * Create new elements only if they do not exist. 683 * **BPF_EXIST** 684 * Update existing elements. 685 * **BPF_F_LOCK** 686 * Update spin_lock-ed map elements. This must be 687 * specified if the map value contains a spinlock. 688 * 689 * On success, *count* elements from the map are updated. 690 * 691 * If an error is returned and *errno* is not **EFAULT**, *count* 692 * is set to the number of successfully processed elements. 693 * 694 * Return 695 * Returns zero on success. On error, -1 is returned and *errno* 696 * is set appropriately. 697 * 698 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 699 * **E2BIG**. **E2BIG** indicates that the number of elements in 700 * the map reached the *max_entries* limit specified at map 701 * creation time. 702 * 703 * May set *errno* to one of the following error codes under 704 * specific circumstances: 705 * 706 * **EEXIST** 707 * If *flags* specifies **BPF_NOEXIST** and the element 708 * with *key* already exists in the map. 709 * **ENOENT** 710 * If *flags* specifies **BPF_EXIST** and the element with 711 * *key* does not exist in the map. 712 * 713 * BPF_MAP_DELETE_BATCH 714 * Description 715 * Delete multiple elements in a map by *key*. 716 * 717 * The *keys* parameter is an input parameter which must point 718 * to memory large enough to hold *count* items based on the key 719 * size of the map *map_fd*, that is, *key_size* * *count*. 720 * 721 * Each element specified in *keys* is sequentially deleted. The 722 * *in_batch*, *out_batch*, and *values* parameters are ignored 723 * and should be zeroed. 724 * 725 * The *elem_flags* argument may be specified as one of the 726 * following: 727 * 728 * **BPF_F_LOCK** 729 * Look up the value of a spin-locked map without 730 * returning the lock. This must be specified if the 731 * elements contain a spinlock. 732 * 733 * On success, *count* elements from the map are updated. 734 * 735 * If an error is returned and *errno* is not **EFAULT**, *count* 736 * is set to the number of successfully processed elements. If 737 * *errno* is **EFAULT**, up to *count* elements may be been 738 * deleted. 739 * 740 * Return 741 * Returns zero on success. On error, -1 is returned and *errno* 742 * is set appropriately. 743 * 744 * BPF_LINK_CREATE 745 * Description 746 * Attach an eBPF program to a *target_fd* at the specified 747 * *attach_type* hook and return a file descriptor handle for 748 * managing the link. 749 * 750 * Return 751 * A new file descriptor (a nonnegative integer), or -1 if an 752 * error occurred (in which case, *errno* is set appropriately). 753 * 754 * BPF_LINK_UPDATE 755 * Description 756 * Update the eBPF program in the specified *link_fd* to 757 * *new_prog_fd*. 758 * 759 * Return 760 * Returns zero on success. On error, -1 is returned and *errno* 761 * is set appropriately. 762 * 763 * BPF_LINK_GET_FD_BY_ID 764 * Description 765 * Open a file descriptor for the eBPF Link corresponding to 766 * *link_id*. 767 * 768 * Return 769 * A new file descriptor (a nonnegative integer), or -1 if an 770 * error occurred (in which case, *errno* is set appropriately). 771 * 772 * BPF_LINK_GET_NEXT_ID 773 * Description 774 * Fetch the next eBPF link currently loaded into the kernel. 775 * 776 * Looks for the eBPF link with an id greater than *start_id* 777 * and updates *next_id* on success. If no other eBPF links 778 * remain with ids higher than *start_id*, returns -1 and sets 779 * *errno* to **ENOENT**. 780 * 781 * Return 782 * Returns zero on success. On error, or when no id remains, -1 783 * is returned and *errno* is set appropriately. 784 * 785 * BPF_ENABLE_STATS 786 * Description 787 * Enable eBPF runtime statistics gathering. 788 * 789 * Runtime statistics gathering for the eBPF runtime is disabled 790 * by default to minimize the corresponding performance overhead. 791 * This command enables statistics globally. 792 * 793 * Multiple programs may independently enable statistics. 794 * After gathering the desired statistics, eBPF runtime statistics 795 * may be disabled again by calling **close**\ (2) for the file 796 * descriptor returned by this function. Statistics will only be 797 * disabled system-wide when all outstanding file descriptors 798 * returned by prior calls for this subcommand are closed. 799 * 800 * Return 801 * A new file descriptor (a nonnegative integer), or -1 if an 802 * error occurred (in which case, *errno* is set appropriately). 803 * 804 * BPF_ITER_CREATE 805 * Description 806 * Create an iterator on top of the specified *link_fd* (as 807 * previously created using **BPF_LINK_CREATE**) and return a 808 * file descriptor that can be used to trigger the iteration. 809 * 810 * If the resulting file descriptor is pinned to the filesystem 811 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 812 * for that path will trigger the iterator to read kernel state 813 * using the eBPF program attached to *link_fd*. 814 * 815 * Return 816 * A new file descriptor (a nonnegative integer), or -1 if an 817 * error occurred (in which case, *errno* is set appropriately). 818 * 819 * BPF_LINK_DETACH 820 * Description 821 * Forcefully detach the specified *link_fd* from its 822 * corresponding attachment point. 823 * 824 * Return 825 * Returns zero on success. On error, -1 is returned and *errno* 826 * is set appropriately. 827 * 828 * BPF_PROG_BIND_MAP 829 * Description 830 * Bind a map to the lifetime of an eBPF program. 831 * 832 * The map identified by *map_fd* is bound to the program 833 * identified by *prog_fd* and only released when *prog_fd* is 834 * released. This may be used in cases where metadata should be 835 * associated with a program which otherwise does not contain any 836 * references to the map (for example, embedded in the eBPF 837 * program instructions). 838 * 839 * Return 840 * Returns zero on success. On error, -1 is returned and *errno* 841 * is set appropriately. 842 * 843 * NOTES 844 * eBPF objects (maps and programs) can be shared between processes. 845 * 846 * * After **fork**\ (2), the child inherits file descriptors 847 * referring to the same eBPF objects. 848 * * File descriptors referring to eBPF objects can be transferred over 849 * **unix**\ (7) domain sockets. 850 * * File descriptors referring to eBPF objects can be duplicated in the 851 * usual way, using **dup**\ (2) and similar calls. 852 * * File descriptors referring to eBPF objects can be pinned to the 853 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 854 * 855 * An eBPF object is deallocated only after all file descriptors referring 856 * to the object have been closed and no references remain pinned to the 857 * filesystem or attached (for example, bound to a program or device). 858 */ 859 enum bpf_cmd { 860 BPF_MAP_CREATE, 861 BPF_MAP_LOOKUP_ELEM, 862 BPF_MAP_UPDATE_ELEM, 863 BPF_MAP_DELETE_ELEM, 864 BPF_MAP_GET_NEXT_KEY, 865 BPF_PROG_LOAD, 866 BPF_OBJ_PIN, 867 BPF_OBJ_GET, 868 BPF_PROG_ATTACH, 869 BPF_PROG_DETACH, 870 BPF_PROG_TEST_RUN, 871 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 872 BPF_PROG_GET_NEXT_ID, 873 BPF_MAP_GET_NEXT_ID, 874 BPF_PROG_GET_FD_BY_ID, 875 BPF_MAP_GET_FD_BY_ID, 876 BPF_OBJ_GET_INFO_BY_FD, 877 BPF_PROG_QUERY, 878 BPF_RAW_TRACEPOINT_OPEN, 879 BPF_BTF_LOAD, 880 BPF_BTF_GET_FD_BY_ID, 881 BPF_TASK_FD_QUERY, 882 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 883 BPF_MAP_FREEZE, 884 BPF_BTF_GET_NEXT_ID, 885 BPF_MAP_LOOKUP_BATCH, 886 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 887 BPF_MAP_UPDATE_BATCH, 888 BPF_MAP_DELETE_BATCH, 889 BPF_LINK_CREATE, 890 BPF_LINK_UPDATE, 891 BPF_LINK_GET_FD_BY_ID, 892 BPF_LINK_GET_NEXT_ID, 893 BPF_ENABLE_STATS, 894 BPF_ITER_CREATE, 895 BPF_LINK_DETACH, 896 BPF_PROG_BIND_MAP, 897 }; 898 899 enum bpf_map_type { 900 BPF_MAP_TYPE_UNSPEC, 901 BPF_MAP_TYPE_HASH, 902 BPF_MAP_TYPE_ARRAY, 903 BPF_MAP_TYPE_PROG_ARRAY, 904 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 905 BPF_MAP_TYPE_PERCPU_HASH, 906 BPF_MAP_TYPE_PERCPU_ARRAY, 907 BPF_MAP_TYPE_STACK_TRACE, 908 BPF_MAP_TYPE_CGROUP_ARRAY, 909 BPF_MAP_TYPE_LRU_HASH, 910 BPF_MAP_TYPE_LRU_PERCPU_HASH, 911 BPF_MAP_TYPE_LPM_TRIE, 912 BPF_MAP_TYPE_ARRAY_OF_MAPS, 913 BPF_MAP_TYPE_HASH_OF_MAPS, 914 BPF_MAP_TYPE_DEVMAP, 915 BPF_MAP_TYPE_SOCKMAP, 916 BPF_MAP_TYPE_CPUMAP, 917 BPF_MAP_TYPE_XSKMAP, 918 BPF_MAP_TYPE_SOCKHASH, 919 BPF_MAP_TYPE_CGROUP_STORAGE, 920 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 921 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 922 BPF_MAP_TYPE_QUEUE, 923 BPF_MAP_TYPE_STACK, 924 BPF_MAP_TYPE_SK_STORAGE, 925 BPF_MAP_TYPE_DEVMAP_HASH, 926 BPF_MAP_TYPE_STRUCT_OPS, 927 BPF_MAP_TYPE_RINGBUF, 928 BPF_MAP_TYPE_INODE_STORAGE, 929 BPF_MAP_TYPE_TASK_STORAGE, 930 BPF_MAP_TYPE_BLOOM_FILTER, 931 }; 932 933 /* Note that tracing related programs such as 934 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 935 * are not subject to a stable API since kernel internal data 936 * structures can change from release to release and may 937 * therefore break existing tracing BPF programs. Tracing BPF 938 * programs correspond to /a/ specific kernel which is to be 939 * analyzed, and not /a/ specific kernel /and/ all future ones. 940 */ 941 enum bpf_prog_type { 942 BPF_PROG_TYPE_UNSPEC, 943 BPF_PROG_TYPE_SOCKET_FILTER, 944 BPF_PROG_TYPE_KPROBE, 945 BPF_PROG_TYPE_SCHED_CLS, 946 BPF_PROG_TYPE_SCHED_ACT, 947 BPF_PROG_TYPE_TRACEPOINT, 948 BPF_PROG_TYPE_XDP, 949 BPF_PROG_TYPE_PERF_EVENT, 950 BPF_PROG_TYPE_CGROUP_SKB, 951 BPF_PROG_TYPE_CGROUP_SOCK, 952 BPF_PROG_TYPE_LWT_IN, 953 BPF_PROG_TYPE_LWT_OUT, 954 BPF_PROG_TYPE_LWT_XMIT, 955 BPF_PROG_TYPE_SOCK_OPS, 956 BPF_PROG_TYPE_SK_SKB, 957 BPF_PROG_TYPE_CGROUP_DEVICE, 958 BPF_PROG_TYPE_SK_MSG, 959 BPF_PROG_TYPE_RAW_TRACEPOINT, 960 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 961 BPF_PROG_TYPE_LWT_SEG6LOCAL, 962 BPF_PROG_TYPE_LIRC_MODE2, 963 BPF_PROG_TYPE_SK_REUSEPORT, 964 BPF_PROG_TYPE_FLOW_DISSECTOR, 965 BPF_PROG_TYPE_CGROUP_SYSCTL, 966 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 967 BPF_PROG_TYPE_CGROUP_SOCKOPT, 968 BPF_PROG_TYPE_TRACING, 969 BPF_PROG_TYPE_STRUCT_OPS, 970 BPF_PROG_TYPE_EXT, 971 BPF_PROG_TYPE_LSM, 972 BPF_PROG_TYPE_SK_LOOKUP, 973 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 974 }; 975 976 enum bpf_attach_type { 977 BPF_CGROUP_INET_INGRESS, 978 BPF_CGROUP_INET_EGRESS, 979 BPF_CGROUP_INET_SOCK_CREATE, 980 BPF_CGROUP_SOCK_OPS, 981 BPF_SK_SKB_STREAM_PARSER, 982 BPF_SK_SKB_STREAM_VERDICT, 983 BPF_CGROUP_DEVICE, 984 BPF_SK_MSG_VERDICT, 985 BPF_CGROUP_INET4_BIND, 986 BPF_CGROUP_INET6_BIND, 987 BPF_CGROUP_INET4_CONNECT, 988 BPF_CGROUP_INET6_CONNECT, 989 BPF_CGROUP_INET4_POST_BIND, 990 BPF_CGROUP_INET6_POST_BIND, 991 BPF_CGROUP_UDP4_SENDMSG, 992 BPF_CGROUP_UDP6_SENDMSG, 993 BPF_LIRC_MODE2, 994 BPF_FLOW_DISSECTOR, 995 BPF_CGROUP_SYSCTL, 996 BPF_CGROUP_UDP4_RECVMSG, 997 BPF_CGROUP_UDP6_RECVMSG, 998 BPF_CGROUP_GETSOCKOPT, 999 BPF_CGROUP_SETSOCKOPT, 1000 BPF_TRACE_RAW_TP, 1001 BPF_TRACE_FENTRY, 1002 BPF_TRACE_FEXIT, 1003 BPF_MODIFY_RETURN, 1004 BPF_LSM_MAC, 1005 BPF_TRACE_ITER, 1006 BPF_CGROUP_INET4_GETPEERNAME, 1007 BPF_CGROUP_INET6_GETPEERNAME, 1008 BPF_CGROUP_INET4_GETSOCKNAME, 1009 BPF_CGROUP_INET6_GETSOCKNAME, 1010 BPF_XDP_DEVMAP, 1011 BPF_CGROUP_INET_SOCK_RELEASE, 1012 BPF_XDP_CPUMAP, 1013 BPF_SK_LOOKUP, 1014 BPF_XDP, 1015 BPF_SK_SKB_VERDICT, 1016 BPF_SK_REUSEPORT_SELECT, 1017 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1018 BPF_PERF_EVENT, 1019 BPF_TRACE_KPROBE_MULTI, 1020 BPF_LSM_CGROUP, 1021 __MAX_BPF_ATTACH_TYPE 1022 }; 1023 1024 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1025 1026 enum bpf_link_type { 1027 BPF_LINK_TYPE_UNSPEC = 0, 1028 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1029 BPF_LINK_TYPE_TRACING = 2, 1030 BPF_LINK_TYPE_CGROUP = 3, 1031 BPF_LINK_TYPE_ITER = 4, 1032 BPF_LINK_TYPE_NETNS = 5, 1033 BPF_LINK_TYPE_XDP = 6, 1034 BPF_LINK_TYPE_PERF_EVENT = 7, 1035 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1036 BPF_LINK_TYPE_STRUCT_OPS = 9, 1037 1038 MAX_BPF_LINK_TYPE, 1039 }; 1040 1041 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1042 * 1043 * NONE(default): No further bpf programs allowed in the subtree. 1044 * 1045 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1046 * the program in this cgroup yields to sub-cgroup program. 1047 * 1048 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1049 * that cgroup program gets run in addition to the program in this cgroup. 1050 * 1051 * Only one program is allowed to be attached to a cgroup with 1052 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1053 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1054 * release old program and attach the new one. Attach flags has to match. 1055 * 1056 * Multiple programs are allowed to be attached to a cgroup with 1057 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1058 * (those that were attached first, run first) 1059 * The programs of sub-cgroup are executed first, then programs of 1060 * this cgroup and then programs of parent cgroup. 1061 * When children program makes decision (like picking TCP CA or sock bind) 1062 * parent program has a chance to override it. 1063 * 1064 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1065 * programs for a cgroup. Though it's possible to replace an old program at 1066 * any position by also specifying BPF_F_REPLACE flag and position itself in 1067 * replace_bpf_fd attribute. Old program at this position will be released. 1068 * 1069 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1070 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1071 * Ex1: 1072 * cgrp1 (MULTI progs A, B) -> 1073 * cgrp2 (OVERRIDE prog C) -> 1074 * cgrp3 (MULTI prog D) -> 1075 * cgrp4 (OVERRIDE prog E) -> 1076 * cgrp5 (NONE prog F) 1077 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1078 * if prog F is detached, the execution is E,D,A,B 1079 * if prog F and D are detached, the execution is E,A,B 1080 * if prog F, E and D are detached, the execution is C,A,B 1081 * 1082 * All eligible programs are executed regardless of return code from 1083 * earlier programs. 1084 */ 1085 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1086 #define BPF_F_ALLOW_MULTI (1U << 1) 1087 #define BPF_F_REPLACE (1U << 2) 1088 1089 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1090 * verifier will perform strict alignment checking as if the kernel 1091 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1092 * and NET_IP_ALIGN defined to 2. 1093 */ 1094 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1095 1096 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1097 * verifier will allow any alignment whatsoever. On platforms 1098 * with strict alignment requirements for loads ands stores (such 1099 * as sparc and mips) the verifier validates that all loads and 1100 * stores provably follow this requirement. This flag turns that 1101 * checking and enforcement off. 1102 * 1103 * It is mostly used for testing when we want to validate the 1104 * context and memory access aspects of the verifier, but because 1105 * of an unaligned access the alignment check would trigger before 1106 * the one we are interested in. 1107 */ 1108 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1109 1110 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1111 * Verifier does sub-register def/use analysis and identifies instructions whose 1112 * def only matters for low 32-bit, high 32-bit is never referenced later 1113 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1114 * that it is safe to ignore clearing high 32-bit for these instructions. This 1115 * saves some back-ends a lot of code-gen. However such optimization is not 1116 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1117 * hence hasn't used verifier's analysis result. But, we really want to have a 1118 * way to be able to verify the correctness of the described optimization on 1119 * x86_64 on which testsuites are frequently exercised. 1120 * 1121 * So, this flag is introduced. Once it is set, verifier will randomize high 1122 * 32-bit for those instructions who has been identified as safe to ignore them. 1123 * Then, if verifier is not doing correct analysis, such randomization will 1124 * regress tests to expose bugs. 1125 */ 1126 #define BPF_F_TEST_RND_HI32 (1U << 2) 1127 1128 /* The verifier internal test flag. Behavior is undefined */ 1129 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1130 1131 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1132 * restrict map and helper usage for such programs. Sleepable BPF programs can 1133 * only be attached to hooks where kernel execution context allows sleeping. 1134 * Such programs are allowed to use helpers that may sleep like 1135 * bpf_copy_from_user(). 1136 */ 1137 #define BPF_F_SLEEPABLE (1U << 4) 1138 1139 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1140 * fully support xdp frags. 1141 */ 1142 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1143 1144 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1145 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1146 */ 1147 #define BPF_F_KPROBE_MULTI_RETURN (1U << 0) 1148 1149 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1150 * the following extensions: 1151 * 1152 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1153 * insn[0].imm: map fd or fd_idx 1154 * insn[1].imm: 0 1155 * insn[0].off: 0 1156 * insn[1].off: 0 1157 * ldimm64 rewrite: address of map 1158 * verifier type: CONST_PTR_TO_MAP 1159 */ 1160 #define BPF_PSEUDO_MAP_FD 1 1161 #define BPF_PSEUDO_MAP_IDX 5 1162 1163 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1164 * insn[0].imm: map fd or fd_idx 1165 * insn[1].imm: offset into value 1166 * insn[0].off: 0 1167 * insn[1].off: 0 1168 * ldimm64 rewrite: address of map[0]+offset 1169 * verifier type: PTR_TO_MAP_VALUE 1170 */ 1171 #define BPF_PSEUDO_MAP_VALUE 2 1172 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1173 1174 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1175 * insn[0].imm: kernel btd id of VAR 1176 * insn[1].imm: 0 1177 * insn[0].off: 0 1178 * insn[1].off: 0 1179 * ldimm64 rewrite: address of the kernel variable 1180 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1181 * is struct/union. 1182 */ 1183 #define BPF_PSEUDO_BTF_ID 3 1184 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1185 * insn[0].imm: insn offset to the func 1186 * insn[1].imm: 0 1187 * insn[0].off: 0 1188 * insn[1].off: 0 1189 * ldimm64 rewrite: address of the function 1190 * verifier type: PTR_TO_FUNC. 1191 */ 1192 #define BPF_PSEUDO_FUNC 4 1193 1194 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1195 * offset to another bpf function 1196 */ 1197 #define BPF_PSEUDO_CALL 1 1198 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1199 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1200 */ 1201 #define BPF_PSEUDO_KFUNC_CALL 2 1202 1203 /* flags for BPF_MAP_UPDATE_ELEM command */ 1204 enum { 1205 BPF_ANY = 0, /* create new element or update existing */ 1206 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1207 BPF_EXIST = 2, /* update existing element */ 1208 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1209 }; 1210 1211 /* flags for BPF_MAP_CREATE command */ 1212 enum { 1213 BPF_F_NO_PREALLOC = (1U << 0), 1214 /* Instead of having one common LRU list in the 1215 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1216 * which can scale and perform better. 1217 * Note, the LRU nodes (including free nodes) cannot be moved 1218 * across different LRU lists. 1219 */ 1220 BPF_F_NO_COMMON_LRU = (1U << 1), 1221 /* Specify numa node during map creation */ 1222 BPF_F_NUMA_NODE = (1U << 2), 1223 1224 /* Flags for accessing BPF object from syscall side. */ 1225 BPF_F_RDONLY = (1U << 3), 1226 BPF_F_WRONLY = (1U << 4), 1227 1228 /* Flag for stack_map, store build_id+offset instead of pointer */ 1229 BPF_F_STACK_BUILD_ID = (1U << 5), 1230 1231 /* Zero-initialize hash function seed. This should only be used for testing. */ 1232 BPF_F_ZERO_SEED = (1U << 6), 1233 1234 /* Flags for accessing BPF object from program side. */ 1235 BPF_F_RDONLY_PROG = (1U << 7), 1236 BPF_F_WRONLY_PROG = (1U << 8), 1237 1238 /* Clone map from listener for newly accepted socket */ 1239 BPF_F_CLONE = (1U << 9), 1240 1241 /* Enable memory-mapping BPF map */ 1242 BPF_F_MMAPABLE = (1U << 10), 1243 1244 /* Share perf_event among processes */ 1245 BPF_F_PRESERVE_ELEMS = (1U << 11), 1246 1247 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1248 BPF_F_INNER_MAP = (1U << 12), 1249 }; 1250 1251 /* Flags for BPF_PROG_QUERY. */ 1252 1253 /* Query effective (directly attached + inherited from ancestor cgroups) 1254 * programs that will be executed for events within a cgroup. 1255 * attach_flags with this flag are returned only for directly attached programs. 1256 */ 1257 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1258 1259 /* Flags for BPF_PROG_TEST_RUN */ 1260 1261 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1262 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1263 /* If set, XDP frames will be transmitted after processing */ 1264 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1265 1266 /* type for BPF_ENABLE_STATS */ 1267 enum bpf_stats_type { 1268 /* enabled run_time_ns and run_cnt */ 1269 BPF_STATS_RUN_TIME = 0, 1270 }; 1271 1272 enum bpf_stack_build_id_status { 1273 /* user space need an empty entry to identify end of a trace */ 1274 BPF_STACK_BUILD_ID_EMPTY = 0, 1275 /* with valid build_id and offset */ 1276 BPF_STACK_BUILD_ID_VALID = 1, 1277 /* couldn't get build_id, fallback to ip */ 1278 BPF_STACK_BUILD_ID_IP = 2, 1279 }; 1280 1281 #define BPF_BUILD_ID_SIZE 20 1282 struct bpf_stack_build_id { 1283 __s32 status; 1284 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1285 union { 1286 __u64 offset; 1287 __u64 ip; 1288 }; 1289 }; 1290 1291 #define BPF_OBJ_NAME_LEN 16U 1292 1293 union bpf_attr { 1294 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1295 __u32 map_type; /* one of enum bpf_map_type */ 1296 __u32 key_size; /* size of key in bytes */ 1297 __u32 value_size; /* size of value in bytes */ 1298 __u32 max_entries; /* max number of entries in a map */ 1299 __u32 map_flags; /* BPF_MAP_CREATE related 1300 * flags defined above. 1301 */ 1302 __u32 inner_map_fd; /* fd pointing to the inner map */ 1303 __u32 numa_node; /* numa node (effective only if 1304 * BPF_F_NUMA_NODE is set). 1305 */ 1306 char map_name[BPF_OBJ_NAME_LEN]; 1307 __u32 map_ifindex; /* ifindex of netdev to create on */ 1308 __u32 btf_fd; /* fd pointing to a BTF type data */ 1309 __u32 btf_key_type_id; /* BTF type_id of the key */ 1310 __u32 btf_value_type_id; /* BTF type_id of the value */ 1311 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1312 * struct stored as the 1313 * map value 1314 */ 1315 /* Any per-map-type extra fields 1316 * 1317 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1318 * number of hash functions (if 0, the bloom filter will default 1319 * to using 5 hash functions). 1320 */ 1321 __u64 map_extra; 1322 }; 1323 1324 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1325 __u32 map_fd; 1326 __aligned_u64 key; 1327 union { 1328 __aligned_u64 value; 1329 __aligned_u64 next_key; 1330 }; 1331 __u64 flags; 1332 }; 1333 1334 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1335 __aligned_u64 in_batch; /* start batch, 1336 * NULL to start from beginning 1337 */ 1338 __aligned_u64 out_batch; /* output: next start batch */ 1339 __aligned_u64 keys; 1340 __aligned_u64 values; 1341 __u32 count; /* input/output: 1342 * input: # of key/value 1343 * elements 1344 * output: # of filled elements 1345 */ 1346 __u32 map_fd; 1347 __u64 elem_flags; 1348 __u64 flags; 1349 } batch; 1350 1351 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1352 __u32 prog_type; /* one of enum bpf_prog_type */ 1353 __u32 insn_cnt; 1354 __aligned_u64 insns; 1355 __aligned_u64 license; 1356 __u32 log_level; /* verbosity level of verifier */ 1357 __u32 log_size; /* size of user buffer */ 1358 __aligned_u64 log_buf; /* user supplied buffer */ 1359 __u32 kern_version; /* not used */ 1360 __u32 prog_flags; 1361 char prog_name[BPF_OBJ_NAME_LEN]; 1362 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1363 /* For some prog types expected attach type must be known at 1364 * load time to verify attach type specific parts of prog 1365 * (context accesses, allowed helpers, etc). 1366 */ 1367 __u32 expected_attach_type; 1368 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1369 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1370 __aligned_u64 func_info; /* func info */ 1371 __u32 func_info_cnt; /* number of bpf_func_info records */ 1372 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1373 __aligned_u64 line_info; /* line info */ 1374 __u32 line_info_cnt; /* number of bpf_line_info records */ 1375 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1376 union { 1377 /* valid prog_fd to attach to bpf prog */ 1378 __u32 attach_prog_fd; 1379 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1380 __u32 attach_btf_obj_fd; 1381 }; 1382 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1383 __aligned_u64 fd_array; /* array of FDs */ 1384 __aligned_u64 core_relos; 1385 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1386 }; 1387 1388 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1389 __aligned_u64 pathname; 1390 __u32 bpf_fd; 1391 __u32 file_flags; 1392 }; 1393 1394 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1395 __u32 target_fd; /* container object to attach to */ 1396 __u32 attach_bpf_fd; /* eBPF program to attach */ 1397 __u32 attach_type; 1398 __u32 attach_flags; 1399 __u32 replace_bpf_fd; /* previously attached eBPF 1400 * program to replace if 1401 * BPF_F_REPLACE is used 1402 */ 1403 }; 1404 1405 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1406 __u32 prog_fd; 1407 __u32 retval; 1408 __u32 data_size_in; /* input: len of data_in */ 1409 __u32 data_size_out; /* input/output: len of data_out 1410 * returns ENOSPC if data_out 1411 * is too small. 1412 */ 1413 __aligned_u64 data_in; 1414 __aligned_u64 data_out; 1415 __u32 repeat; 1416 __u32 duration; 1417 __u32 ctx_size_in; /* input: len of ctx_in */ 1418 __u32 ctx_size_out; /* input/output: len of ctx_out 1419 * returns ENOSPC if ctx_out 1420 * is too small. 1421 */ 1422 __aligned_u64 ctx_in; 1423 __aligned_u64 ctx_out; 1424 __u32 flags; 1425 __u32 cpu; 1426 __u32 batch_size; 1427 } test; 1428 1429 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1430 union { 1431 __u32 start_id; 1432 __u32 prog_id; 1433 __u32 map_id; 1434 __u32 btf_id; 1435 __u32 link_id; 1436 }; 1437 __u32 next_id; 1438 __u32 open_flags; 1439 }; 1440 1441 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1442 __u32 bpf_fd; 1443 __u32 info_len; 1444 __aligned_u64 info; 1445 } info; 1446 1447 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1448 __u32 target_fd; /* container object to query */ 1449 __u32 attach_type; 1450 __u32 query_flags; 1451 __u32 attach_flags; 1452 __aligned_u64 prog_ids; 1453 __u32 prog_cnt; 1454 __aligned_u64 prog_attach_flags; /* output: per-program attach_flags */ 1455 } query; 1456 1457 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1458 __u64 name; 1459 __u32 prog_fd; 1460 } raw_tracepoint; 1461 1462 struct { /* anonymous struct for BPF_BTF_LOAD */ 1463 __aligned_u64 btf; 1464 __aligned_u64 btf_log_buf; 1465 __u32 btf_size; 1466 __u32 btf_log_size; 1467 __u32 btf_log_level; 1468 }; 1469 1470 struct { 1471 __u32 pid; /* input: pid */ 1472 __u32 fd; /* input: fd */ 1473 __u32 flags; /* input: flags */ 1474 __u32 buf_len; /* input/output: buf len */ 1475 __aligned_u64 buf; /* input/output: 1476 * tp_name for tracepoint 1477 * symbol for kprobe 1478 * filename for uprobe 1479 */ 1480 __u32 prog_id; /* output: prod_id */ 1481 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1482 __u64 probe_offset; /* output: probe_offset */ 1483 __u64 probe_addr; /* output: probe_addr */ 1484 } task_fd_query; 1485 1486 struct { /* struct used by BPF_LINK_CREATE command */ 1487 __u32 prog_fd; /* eBPF program to attach */ 1488 union { 1489 __u32 target_fd; /* object to attach to */ 1490 __u32 target_ifindex; /* target ifindex */ 1491 }; 1492 __u32 attach_type; /* attach type */ 1493 __u32 flags; /* extra flags */ 1494 union { 1495 __u32 target_btf_id; /* btf_id of target to attach to */ 1496 struct { 1497 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1498 __u32 iter_info_len; /* iter_info length */ 1499 }; 1500 struct { 1501 /* black box user-provided value passed through 1502 * to BPF program at the execution time and 1503 * accessible through bpf_get_attach_cookie() BPF helper 1504 */ 1505 __u64 bpf_cookie; 1506 } perf_event; 1507 struct { 1508 __u32 flags; 1509 __u32 cnt; 1510 __aligned_u64 syms; 1511 __aligned_u64 addrs; 1512 __aligned_u64 cookies; 1513 } kprobe_multi; 1514 struct { 1515 /* this is overlaid with the target_btf_id above. */ 1516 __u32 target_btf_id; 1517 /* black box user-provided value passed through 1518 * to BPF program at the execution time and 1519 * accessible through bpf_get_attach_cookie() BPF helper 1520 */ 1521 __u64 cookie; 1522 } tracing; 1523 }; 1524 } link_create; 1525 1526 struct { /* struct used by BPF_LINK_UPDATE command */ 1527 __u32 link_fd; /* link fd */ 1528 /* new program fd to update link with */ 1529 __u32 new_prog_fd; 1530 __u32 flags; /* extra flags */ 1531 /* expected link's program fd; is specified only if 1532 * BPF_F_REPLACE flag is set in flags */ 1533 __u32 old_prog_fd; 1534 } link_update; 1535 1536 struct { 1537 __u32 link_fd; 1538 } link_detach; 1539 1540 struct { /* struct used by BPF_ENABLE_STATS command */ 1541 __u32 type; 1542 } enable_stats; 1543 1544 struct { /* struct used by BPF_ITER_CREATE command */ 1545 __u32 link_fd; 1546 __u32 flags; 1547 } iter_create; 1548 1549 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1550 __u32 prog_fd; 1551 __u32 map_fd; 1552 __u32 flags; /* extra flags */ 1553 } prog_bind_map; 1554 1555 } __attribute__((aligned(8))); 1556 1557 /* The description below is an attempt at providing documentation to eBPF 1558 * developers about the multiple available eBPF helper functions. It can be 1559 * parsed and used to produce a manual page. The workflow is the following, 1560 * and requires the rst2man utility: 1561 * 1562 * $ ./scripts/bpf_doc.py \ 1563 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1564 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1565 * $ man /tmp/bpf-helpers.7 1566 * 1567 * Note that in order to produce this external documentation, some RST 1568 * formatting is used in the descriptions to get "bold" and "italics" in 1569 * manual pages. Also note that the few trailing white spaces are 1570 * intentional, removing them would break paragraphs for rst2man. 1571 * 1572 * Start of BPF helper function descriptions: 1573 * 1574 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1575 * Description 1576 * Perform a lookup in *map* for an entry associated to *key*. 1577 * Return 1578 * Map value associated to *key*, or **NULL** if no entry was 1579 * found. 1580 * 1581 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1582 * Description 1583 * Add or update the value of the entry associated to *key* in 1584 * *map* with *value*. *flags* is one of: 1585 * 1586 * **BPF_NOEXIST** 1587 * The entry for *key* must not exist in the map. 1588 * **BPF_EXIST** 1589 * The entry for *key* must already exist in the map. 1590 * **BPF_ANY** 1591 * No condition on the existence of the entry for *key*. 1592 * 1593 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1594 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1595 * elements always exist), the helper would return an error. 1596 * Return 1597 * 0 on success, or a negative error in case of failure. 1598 * 1599 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1600 * Description 1601 * Delete entry with *key* from *map*. 1602 * Return 1603 * 0 on success, or a negative error in case of failure. 1604 * 1605 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1606 * Description 1607 * For tracing programs, safely attempt to read *size* bytes from 1608 * kernel space address *unsafe_ptr* and store the data in *dst*. 1609 * 1610 * Generally, use **bpf_probe_read_user**\ () or 1611 * **bpf_probe_read_kernel**\ () instead. 1612 * Return 1613 * 0 on success, or a negative error in case of failure. 1614 * 1615 * u64 bpf_ktime_get_ns(void) 1616 * Description 1617 * Return the time elapsed since system boot, in nanoseconds. 1618 * Does not include time the system was suspended. 1619 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1620 * Return 1621 * Current *ktime*. 1622 * 1623 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1624 * Description 1625 * This helper is a "printk()-like" facility for debugging. It 1626 * prints a message defined by format *fmt* (of size *fmt_size*) 1627 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1628 * available. It can take up to three additional **u64** 1629 * arguments (as an eBPF helpers, the total number of arguments is 1630 * limited to five). 1631 * 1632 * Each time the helper is called, it appends a line to the trace. 1633 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 1634 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1635 * The format of the trace is customizable, and the exact output 1636 * one will get depends on the options set in 1637 * *\/sys/kernel/debug/tracing/trace_options* (see also the 1638 * *README* file under the same directory). However, it usually 1639 * defaults to something like: 1640 * 1641 * :: 1642 * 1643 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1644 * 1645 * In the above: 1646 * 1647 * * ``telnet`` is the name of the current task. 1648 * * ``470`` is the PID of the current task. 1649 * * ``001`` is the CPU number on which the task is 1650 * running. 1651 * * In ``.N..``, each character refers to a set of 1652 * options (whether irqs are enabled, scheduling 1653 * options, whether hard/softirqs are running, level of 1654 * preempt_disabled respectively). **N** means that 1655 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1656 * are set. 1657 * * ``419421.045894`` is a timestamp. 1658 * * ``0x00000001`` is a fake value used by BPF for the 1659 * instruction pointer register. 1660 * * ``<formatted msg>`` is the message formatted with 1661 * *fmt*. 1662 * 1663 * The conversion specifiers supported by *fmt* are similar, but 1664 * more limited than for printk(). They are **%d**, **%i**, 1665 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1666 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1667 * of field, padding with zeroes, etc.) is available, and the 1668 * helper will return **-EINVAL** (but print nothing) if it 1669 * encounters an unknown specifier. 1670 * 1671 * Also, note that **bpf_trace_printk**\ () is slow, and should 1672 * only be used for debugging purposes. For this reason, a notice 1673 * block (spanning several lines) is printed to kernel logs and 1674 * states that the helper should not be used "for production use" 1675 * the first time this helper is used (or more precisely, when 1676 * **trace_printk**\ () buffers are allocated). For passing values 1677 * to user space, perf events should be preferred. 1678 * Return 1679 * The number of bytes written to the buffer, or a negative error 1680 * in case of failure. 1681 * 1682 * u32 bpf_get_prandom_u32(void) 1683 * Description 1684 * Get a pseudo-random number. 1685 * 1686 * From a security point of view, this helper uses its own 1687 * pseudo-random internal state, and cannot be used to infer the 1688 * seed of other random functions in the kernel. However, it is 1689 * essential to note that the generator used by the helper is not 1690 * cryptographically secure. 1691 * Return 1692 * A random 32-bit unsigned value. 1693 * 1694 * u32 bpf_get_smp_processor_id(void) 1695 * Description 1696 * Get the SMP (symmetric multiprocessing) processor id. Note that 1697 * all programs run with migration disabled, which means that the 1698 * SMP processor id is stable during all the execution of the 1699 * program. 1700 * Return 1701 * The SMP id of the processor running the program. 1702 * 1703 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1704 * Description 1705 * Store *len* bytes from address *from* into the packet 1706 * associated to *skb*, at *offset*. *flags* are a combination of 1707 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1708 * checksum for the packet after storing the bytes) and 1709 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1710 * **->swhash** and *skb*\ **->l4hash** to 0). 1711 * 1712 * A call to this helper is susceptible to change the underlying 1713 * packet buffer. Therefore, at load time, all checks on pointers 1714 * previously done by the verifier are invalidated and must be 1715 * performed again, if the helper is used in combination with 1716 * direct packet access. 1717 * Return 1718 * 0 on success, or a negative error in case of failure. 1719 * 1720 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1721 * Description 1722 * Recompute the layer 3 (e.g. IP) checksum for the packet 1723 * associated to *skb*. Computation is incremental, so the helper 1724 * must know the former value of the header field that was 1725 * modified (*from*), the new value of this field (*to*), and the 1726 * number of bytes (2 or 4) for this field, stored in *size*. 1727 * Alternatively, it is possible to store the difference between 1728 * the previous and the new values of the header field in *to*, by 1729 * setting *from* and *size* to 0. For both methods, *offset* 1730 * indicates the location of the IP checksum within the packet. 1731 * 1732 * This helper works in combination with **bpf_csum_diff**\ (), 1733 * which does not update the checksum in-place, but offers more 1734 * flexibility and can handle sizes larger than 2 or 4 for the 1735 * checksum to update. 1736 * 1737 * A call to this helper is susceptible to change the underlying 1738 * packet buffer. Therefore, at load time, all checks on pointers 1739 * previously done by the verifier are invalidated and must be 1740 * performed again, if the helper is used in combination with 1741 * direct packet access. 1742 * Return 1743 * 0 on success, or a negative error in case of failure. 1744 * 1745 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1746 * Description 1747 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1748 * packet associated to *skb*. Computation is incremental, so the 1749 * helper must know the former value of the header field that was 1750 * modified (*from*), the new value of this field (*to*), and the 1751 * number of bytes (2 or 4) for this field, stored on the lowest 1752 * four bits of *flags*. Alternatively, it is possible to store 1753 * the difference between the previous and the new values of the 1754 * header field in *to*, by setting *from* and the four lowest 1755 * bits of *flags* to 0. For both methods, *offset* indicates the 1756 * location of the IP checksum within the packet. In addition to 1757 * the size of the field, *flags* can be added (bitwise OR) actual 1758 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1759 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1760 * for updates resulting in a null checksum the value is set to 1761 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1762 * the checksum is to be computed against a pseudo-header. 1763 * 1764 * This helper works in combination with **bpf_csum_diff**\ (), 1765 * which does not update the checksum in-place, but offers more 1766 * flexibility and can handle sizes larger than 2 or 4 for the 1767 * checksum to update. 1768 * 1769 * A call to this helper is susceptible to change the underlying 1770 * packet buffer. Therefore, at load time, all checks on pointers 1771 * previously done by the verifier are invalidated and must be 1772 * performed again, if the helper is used in combination with 1773 * direct packet access. 1774 * Return 1775 * 0 on success, or a negative error in case of failure. 1776 * 1777 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1778 * Description 1779 * This special helper is used to trigger a "tail call", or in 1780 * other words, to jump into another eBPF program. The same stack 1781 * frame is used (but values on stack and in registers for the 1782 * caller are not accessible to the callee). This mechanism allows 1783 * for program chaining, either for raising the maximum number of 1784 * available eBPF instructions, or to execute given programs in 1785 * conditional blocks. For security reasons, there is an upper 1786 * limit to the number of successive tail calls that can be 1787 * performed. 1788 * 1789 * Upon call of this helper, the program attempts to jump into a 1790 * program referenced at index *index* in *prog_array_map*, a 1791 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1792 * *ctx*, a pointer to the context. 1793 * 1794 * If the call succeeds, the kernel immediately runs the first 1795 * instruction of the new program. This is not a function call, 1796 * and it never returns to the previous program. If the call 1797 * fails, then the helper has no effect, and the caller continues 1798 * to run its subsequent instructions. A call can fail if the 1799 * destination program for the jump does not exist (i.e. *index* 1800 * is superior to the number of entries in *prog_array_map*), or 1801 * if the maximum number of tail calls has been reached for this 1802 * chain of programs. This limit is defined in the kernel by the 1803 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1804 * which is currently set to 33. 1805 * Return 1806 * 0 on success, or a negative error in case of failure. 1807 * 1808 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1809 * Description 1810 * Clone and redirect the packet associated to *skb* to another 1811 * net device of index *ifindex*. Both ingress and egress 1812 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1813 * value in *flags* is used to make the distinction (ingress path 1814 * is selected if the flag is present, egress path otherwise). 1815 * This is the only flag supported for now. 1816 * 1817 * In comparison with **bpf_redirect**\ () helper, 1818 * **bpf_clone_redirect**\ () has the associated cost of 1819 * duplicating the packet buffer, but this can be executed out of 1820 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1821 * efficient, but it is handled through an action code where the 1822 * redirection happens only after the eBPF program has returned. 1823 * 1824 * A call to this helper is susceptible to change the underlying 1825 * packet buffer. Therefore, at load time, all checks on pointers 1826 * previously done by the verifier are invalidated and must be 1827 * performed again, if the helper is used in combination with 1828 * direct packet access. 1829 * Return 1830 * 0 on success, or a negative error in case of failure. 1831 * 1832 * u64 bpf_get_current_pid_tgid(void) 1833 * Description 1834 * Get the current pid and tgid. 1835 * Return 1836 * A 64-bit integer containing the current tgid and pid, and 1837 * created as such: 1838 * *current_task*\ **->tgid << 32 \|** 1839 * *current_task*\ **->pid**. 1840 * 1841 * u64 bpf_get_current_uid_gid(void) 1842 * Description 1843 * Get the current uid and gid. 1844 * Return 1845 * A 64-bit integer containing the current GID and UID, and 1846 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1847 * 1848 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1849 * Description 1850 * Copy the **comm** attribute of the current task into *buf* of 1851 * *size_of_buf*. The **comm** attribute contains the name of 1852 * the executable (excluding the path) for the current task. The 1853 * *size_of_buf* must be strictly positive. On success, the 1854 * helper makes sure that the *buf* is NUL-terminated. On failure, 1855 * it is filled with zeroes. 1856 * Return 1857 * 0 on success, or a negative error in case of failure. 1858 * 1859 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1860 * Description 1861 * Retrieve the classid for the current task, i.e. for the net_cls 1862 * cgroup to which *skb* belongs. 1863 * 1864 * This helper can be used on TC egress path, but not on ingress. 1865 * 1866 * The net_cls cgroup provides an interface to tag network packets 1867 * based on a user-provided identifier for all traffic coming from 1868 * the tasks belonging to the related cgroup. See also the related 1869 * kernel documentation, available from the Linux sources in file 1870 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1871 * 1872 * The Linux kernel has two versions for cgroups: there are 1873 * cgroups v1 and cgroups v2. Both are available to users, who can 1874 * use a mixture of them, but note that the net_cls cgroup is for 1875 * cgroup v1 only. This makes it incompatible with BPF programs 1876 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1877 * only hold data for one version of cgroups at a time). 1878 * 1879 * This helper is only available is the kernel was compiled with 1880 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1881 * "**y**" or to "**m**". 1882 * Return 1883 * The classid, or 0 for the default unconfigured classid. 1884 * 1885 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1886 * Description 1887 * Push a *vlan_tci* (VLAN tag control information) of protocol 1888 * *vlan_proto* to the packet associated to *skb*, then update 1889 * the checksum. Note that if *vlan_proto* is different from 1890 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1891 * be **ETH_P_8021Q**. 1892 * 1893 * A call to this helper is susceptible to change the underlying 1894 * packet buffer. Therefore, at load time, all checks on pointers 1895 * previously done by the verifier are invalidated and must be 1896 * performed again, if the helper is used in combination with 1897 * direct packet access. 1898 * Return 1899 * 0 on success, or a negative error in case of failure. 1900 * 1901 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1902 * Description 1903 * Pop a VLAN header from the packet associated to *skb*. 1904 * 1905 * A call to this helper is susceptible to change the underlying 1906 * packet buffer. Therefore, at load time, all checks on pointers 1907 * previously done by the verifier are invalidated and must be 1908 * performed again, if the helper is used in combination with 1909 * direct packet access. 1910 * Return 1911 * 0 on success, or a negative error in case of failure. 1912 * 1913 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1914 * Description 1915 * Get tunnel metadata. This helper takes a pointer *key* to an 1916 * empty **struct bpf_tunnel_key** of **size**, that will be 1917 * filled with tunnel metadata for the packet associated to *skb*. 1918 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1919 * indicates that the tunnel is based on IPv6 protocol instead of 1920 * IPv4. 1921 * 1922 * The **struct bpf_tunnel_key** is an object that generalizes the 1923 * principal parameters used by various tunneling protocols into a 1924 * single struct. This way, it can be used to easily make a 1925 * decision based on the contents of the encapsulation header, 1926 * "summarized" in this struct. In particular, it holds the IP 1927 * address of the remote end (IPv4 or IPv6, depending on the case) 1928 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1929 * this struct exposes the *key*\ **->tunnel_id**, which is 1930 * generally mapped to a VNI (Virtual Network Identifier), making 1931 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1932 * () helper. 1933 * 1934 * Let's imagine that the following code is part of a program 1935 * attached to the TC ingress interface, on one end of a GRE 1936 * tunnel, and is supposed to filter out all messages coming from 1937 * remote ends with IPv4 address other than 10.0.0.1: 1938 * 1939 * :: 1940 * 1941 * int ret; 1942 * struct bpf_tunnel_key key = {}; 1943 * 1944 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1945 * if (ret < 0) 1946 * return TC_ACT_SHOT; // drop packet 1947 * 1948 * if (key.remote_ipv4 != 0x0a000001) 1949 * return TC_ACT_SHOT; // drop packet 1950 * 1951 * return TC_ACT_OK; // accept packet 1952 * 1953 * This interface can also be used with all encapsulation devices 1954 * that can operate in "collect metadata" mode: instead of having 1955 * one network device per specific configuration, the "collect 1956 * metadata" mode only requires a single device where the 1957 * configuration can be extracted from this helper. 1958 * 1959 * This can be used together with various tunnels such as VXLan, 1960 * Geneve, GRE or IP in IP (IPIP). 1961 * Return 1962 * 0 on success, or a negative error in case of failure. 1963 * 1964 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1965 * Description 1966 * Populate tunnel metadata for packet associated to *skb.* The 1967 * tunnel metadata is set to the contents of *key*, of *size*. The 1968 * *flags* can be set to a combination of the following values: 1969 * 1970 * **BPF_F_TUNINFO_IPV6** 1971 * Indicate that the tunnel is based on IPv6 protocol 1972 * instead of IPv4. 1973 * **BPF_F_ZERO_CSUM_TX** 1974 * For IPv4 packets, add a flag to tunnel metadata 1975 * indicating that checksum computation should be skipped 1976 * and checksum set to zeroes. 1977 * **BPF_F_DONT_FRAGMENT** 1978 * Add a flag to tunnel metadata indicating that the 1979 * packet should not be fragmented. 1980 * **BPF_F_SEQ_NUMBER** 1981 * Add a flag to tunnel metadata indicating that a 1982 * sequence number should be added to tunnel header before 1983 * sending the packet. This flag was added for GRE 1984 * encapsulation, but might be used with other protocols 1985 * as well in the future. 1986 * 1987 * Here is a typical usage on the transmit path: 1988 * 1989 * :: 1990 * 1991 * struct bpf_tunnel_key key; 1992 * populate key ... 1993 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1994 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1995 * 1996 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1997 * helper for additional information. 1998 * Return 1999 * 0 on success, or a negative error in case of failure. 2000 * 2001 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2002 * Description 2003 * Read the value of a perf event counter. This helper relies on a 2004 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2005 * the perf event counter is selected when *map* is updated with 2006 * perf event file descriptors. The *map* is an array whose size 2007 * is the number of available CPUs, and each cell contains a value 2008 * relative to one CPU. The value to retrieve is indicated by 2009 * *flags*, that contains the index of the CPU to look up, masked 2010 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2011 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2012 * current CPU should be retrieved. 2013 * 2014 * Note that before Linux 4.13, only hardware perf event can be 2015 * retrieved. 2016 * 2017 * Also, be aware that the newer helper 2018 * **bpf_perf_event_read_value**\ () is recommended over 2019 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2020 * quirks where error and counter value are used as a return code 2021 * (which is wrong to do since ranges may overlap). This issue is 2022 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2023 * time provides more features over the **bpf_perf_event_read**\ 2024 * () interface. Please refer to the description of 2025 * **bpf_perf_event_read_value**\ () for details. 2026 * Return 2027 * The value of the perf event counter read from the map, or a 2028 * negative error code in case of failure. 2029 * 2030 * long bpf_redirect(u32 ifindex, u64 flags) 2031 * Description 2032 * Redirect the packet to another net device of index *ifindex*. 2033 * This helper is somewhat similar to **bpf_clone_redirect**\ 2034 * (), except that the packet is not cloned, which provides 2035 * increased performance. 2036 * 2037 * Except for XDP, both ingress and egress interfaces can be used 2038 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2039 * to make the distinction (ingress path is selected if the flag 2040 * is present, egress path otherwise). Currently, XDP only 2041 * supports redirection to the egress interface, and accepts no 2042 * flag at all. 2043 * 2044 * The same effect can also be attained with the more generic 2045 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2046 * redirect target instead of providing it directly to the helper. 2047 * Return 2048 * For XDP, the helper returns **XDP_REDIRECT** on success or 2049 * **XDP_ABORTED** on error. For other program types, the values 2050 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2051 * error. 2052 * 2053 * u32 bpf_get_route_realm(struct sk_buff *skb) 2054 * Description 2055 * Retrieve the realm or the route, that is to say the 2056 * **tclassid** field of the destination for the *skb*. The 2057 * identifier retrieved is a user-provided tag, similar to the 2058 * one used with the net_cls cgroup (see description for 2059 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2060 * held by a route (a destination entry), not by a task. 2061 * 2062 * Retrieving this identifier works with the clsact TC egress hook 2063 * (see also **tc-bpf(8)**), or alternatively on conventional 2064 * classful egress qdiscs, but not on TC ingress path. In case of 2065 * clsact TC egress hook, this has the advantage that, internally, 2066 * the destination entry has not been dropped yet in the transmit 2067 * path. Therefore, the destination entry does not need to be 2068 * artificially held via **netif_keep_dst**\ () for a classful 2069 * qdisc until the *skb* is freed. 2070 * 2071 * This helper is available only if the kernel was compiled with 2072 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2073 * Return 2074 * The realm of the route for the packet associated to *skb*, or 0 2075 * if none was found. 2076 * 2077 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2078 * Description 2079 * Write raw *data* blob into a special BPF perf event held by 2080 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2081 * event must have the following attributes: **PERF_SAMPLE_RAW** 2082 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2083 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2084 * 2085 * The *flags* are used to indicate the index in *map* for which 2086 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2087 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2088 * to indicate that the index of the current CPU core should be 2089 * used. 2090 * 2091 * The value to write, of *size*, is passed through eBPF stack and 2092 * pointed by *data*. 2093 * 2094 * The context of the program *ctx* needs also be passed to the 2095 * helper. 2096 * 2097 * On user space, a program willing to read the values needs to 2098 * call **perf_event_open**\ () on the perf event (either for 2099 * one or for all CPUs) and to store the file descriptor into the 2100 * *map*. This must be done before the eBPF program can send data 2101 * into it. An example is available in file 2102 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2103 * tree (the eBPF program counterpart is in 2104 * *samples/bpf/trace_output_kern.c*). 2105 * 2106 * **bpf_perf_event_output**\ () achieves better performance 2107 * than **bpf_trace_printk**\ () for sharing data with user 2108 * space, and is much better suitable for streaming data from eBPF 2109 * programs. 2110 * 2111 * Note that this helper is not restricted to tracing use cases 2112 * and can be used with programs attached to TC or XDP as well, 2113 * where it allows for passing data to user space listeners. Data 2114 * can be: 2115 * 2116 * * Only custom structs, 2117 * * Only the packet payload, or 2118 * * A combination of both. 2119 * Return 2120 * 0 on success, or a negative error in case of failure. 2121 * 2122 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2123 * Description 2124 * This helper was provided as an easy way to load data from a 2125 * packet. It can be used to load *len* bytes from *offset* from 2126 * the packet associated to *skb*, into the buffer pointed by 2127 * *to*. 2128 * 2129 * Since Linux 4.7, usage of this helper has mostly been replaced 2130 * by "direct packet access", enabling packet data to be 2131 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2132 * pointing respectively to the first byte of packet data and to 2133 * the byte after the last byte of packet data. However, it 2134 * remains useful if one wishes to read large quantities of data 2135 * at once from a packet into the eBPF stack. 2136 * Return 2137 * 0 on success, or a negative error in case of failure. 2138 * 2139 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2140 * Description 2141 * Walk a user or a kernel stack and return its id. To achieve 2142 * this, the helper needs *ctx*, which is a pointer to the context 2143 * on which the tracing program is executed, and a pointer to a 2144 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2145 * 2146 * The last argument, *flags*, holds the number of stack frames to 2147 * skip (from 0 to 255), masked with 2148 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2149 * a combination of the following flags: 2150 * 2151 * **BPF_F_USER_STACK** 2152 * Collect a user space stack instead of a kernel stack. 2153 * **BPF_F_FAST_STACK_CMP** 2154 * Compare stacks by hash only. 2155 * **BPF_F_REUSE_STACKID** 2156 * If two different stacks hash into the same *stackid*, 2157 * discard the old one. 2158 * 2159 * The stack id retrieved is a 32 bit long integer handle which 2160 * can be further combined with other data (including other stack 2161 * ids) and used as a key into maps. This can be useful for 2162 * generating a variety of graphs (such as flame graphs or off-cpu 2163 * graphs). 2164 * 2165 * For walking a stack, this helper is an improvement over 2166 * **bpf_probe_read**\ (), which can be used with unrolled loops 2167 * but is not efficient and consumes a lot of eBPF instructions. 2168 * Instead, **bpf_get_stackid**\ () can collect up to 2169 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2170 * this limit can be controlled with the **sysctl** program, and 2171 * that it should be manually increased in order to profile long 2172 * user stacks (such as stacks for Java programs). To do so, use: 2173 * 2174 * :: 2175 * 2176 * # sysctl kernel.perf_event_max_stack=<new value> 2177 * Return 2178 * The positive or null stack id on success, or a negative error 2179 * in case of failure. 2180 * 2181 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2182 * Description 2183 * Compute a checksum difference, from the raw buffer pointed by 2184 * *from*, of length *from_size* (that must be a multiple of 4), 2185 * towards the raw buffer pointed by *to*, of size *to_size* 2186 * (same remark). An optional *seed* can be added to the value 2187 * (this can be cascaded, the seed may come from a previous call 2188 * to the helper). 2189 * 2190 * This is flexible enough to be used in several ways: 2191 * 2192 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2193 * checksum, it can be used when pushing new data. 2194 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2195 * checksum, it can be used when removing data from a packet. 2196 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2197 * can be used to compute a diff. Note that *from_size* and 2198 * *to_size* do not need to be equal. 2199 * 2200 * This helper can be used in combination with 2201 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2202 * which one can feed in the difference computed with 2203 * **bpf_csum_diff**\ (). 2204 * Return 2205 * The checksum result, or a negative error code in case of 2206 * failure. 2207 * 2208 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2209 * Description 2210 * Retrieve tunnel options metadata for the packet associated to 2211 * *skb*, and store the raw tunnel option data to the buffer *opt* 2212 * of *size*. 2213 * 2214 * This helper can be used with encapsulation devices that can 2215 * operate in "collect metadata" mode (please refer to the related 2216 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2217 * more details). A particular example where this can be used is 2218 * in combination with the Geneve encapsulation protocol, where it 2219 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2220 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2221 * the eBPF program. This allows for full customization of these 2222 * headers. 2223 * Return 2224 * The size of the option data retrieved. 2225 * 2226 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2227 * Description 2228 * Set tunnel options metadata for the packet associated to *skb* 2229 * to the option data contained in the raw buffer *opt* of *size*. 2230 * 2231 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2232 * helper for additional information. 2233 * Return 2234 * 0 on success, or a negative error in case of failure. 2235 * 2236 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2237 * Description 2238 * Change the protocol of the *skb* to *proto*. Currently 2239 * supported are transition from IPv4 to IPv6, and from IPv6 to 2240 * IPv4. The helper takes care of the groundwork for the 2241 * transition, including resizing the socket buffer. The eBPF 2242 * program is expected to fill the new headers, if any, via 2243 * **skb_store_bytes**\ () and to recompute the checksums with 2244 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2245 * (). The main case for this helper is to perform NAT64 2246 * operations out of an eBPF program. 2247 * 2248 * Internally, the GSO type is marked as dodgy so that headers are 2249 * checked and segments are recalculated by the GSO/GRO engine. 2250 * The size for GSO target is adapted as well. 2251 * 2252 * All values for *flags* are reserved for future usage, and must 2253 * be left at zero. 2254 * 2255 * A call to this helper is susceptible to change the underlying 2256 * packet buffer. Therefore, at load time, all checks on pointers 2257 * previously done by the verifier are invalidated and must be 2258 * performed again, if the helper is used in combination with 2259 * direct packet access. 2260 * Return 2261 * 0 on success, or a negative error in case of failure. 2262 * 2263 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2264 * Description 2265 * Change the packet type for the packet associated to *skb*. This 2266 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2267 * the eBPF program does not have a write access to *skb*\ 2268 * **->pkt_type** beside this helper. Using a helper here allows 2269 * for graceful handling of errors. 2270 * 2271 * The major use case is to change incoming *skb*s to 2272 * **PACKET_HOST** in a programmatic way instead of having to 2273 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2274 * example. 2275 * 2276 * Note that *type* only allows certain values. At this time, they 2277 * are: 2278 * 2279 * **PACKET_HOST** 2280 * Packet is for us. 2281 * **PACKET_BROADCAST** 2282 * Send packet to all. 2283 * **PACKET_MULTICAST** 2284 * Send packet to group. 2285 * **PACKET_OTHERHOST** 2286 * Send packet to someone else. 2287 * Return 2288 * 0 on success, or a negative error in case of failure. 2289 * 2290 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2291 * Description 2292 * Check whether *skb* is a descendant of the cgroup2 held by 2293 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2294 * Return 2295 * The return value depends on the result of the test, and can be: 2296 * 2297 * * 0, if the *skb* failed the cgroup2 descendant test. 2298 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2299 * * A negative error code, if an error occurred. 2300 * 2301 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2302 * Description 2303 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2304 * not set, in particular if the hash was cleared due to mangling, 2305 * recompute this hash. Later accesses to the hash can be done 2306 * directly with *skb*\ **->hash**. 2307 * 2308 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2309 * prototype with **bpf_skb_change_proto**\ (), or calling 2310 * **bpf_skb_store_bytes**\ () with the 2311 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2312 * the hash and to trigger a new computation for the next call to 2313 * **bpf_get_hash_recalc**\ (). 2314 * Return 2315 * The 32-bit hash. 2316 * 2317 * u64 bpf_get_current_task(void) 2318 * Description 2319 * Get the current task. 2320 * Return 2321 * A pointer to the current task struct. 2322 * 2323 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2324 * Description 2325 * Attempt in a safe way to write *len* bytes from the buffer 2326 * *src* to *dst* in memory. It only works for threads that are in 2327 * user context, and *dst* must be a valid user space address. 2328 * 2329 * This helper should not be used to implement any kind of 2330 * security mechanism because of TOC-TOU attacks, but rather to 2331 * debug, divert, and manipulate execution of semi-cooperative 2332 * processes. 2333 * 2334 * Keep in mind that this feature is meant for experiments, and it 2335 * has a risk of crashing the system and running programs. 2336 * Therefore, when an eBPF program using this helper is attached, 2337 * a warning including PID and process name is printed to kernel 2338 * logs. 2339 * Return 2340 * 0 on success, or a negative error in case of failure. 2341 * 2342 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2343 * Description 2344 * Check whether the probe is being run is the context of a given 2345 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2346 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2347 * Return 2348 * The return value depends on the result of the test, and can be: 2349 * 2350 * * 1, if current task belongs to the cgroup2. 2351 * * 0, if current task does not belong to the cgroup2. 2352 * * A negative error code, if an error occurred. 2353 * 2354 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2355 * Description 2356 * Resize (trim or grow) the packet associated to *skb* to the 2357 * new *len*. The *flags* are reserved for future usage, and must 2358 * be left at zero. 2359 * 2360 * The basic idea is that the helper performs the needed work to 2361 * change the size of the packet, then the eBPF program rewrites 2362 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2363 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2364 * and others. This helper is a slow path utility intended for 2365 * replies with control messages. And because it is targeted for 2366 * slow path, the helper itself can afford to be slow: it 2367 * implicitly linearizes, unclones and drops offloads from the 2368 * *skb*. 2369 * 2370 * A call to this helper is susceptible to change the underlying 2371 * packet buffer. Therefore, at load time, all checks on pointers 2372 * previously done by the verifier are invalidated and must be 2373 * performed again, if the helper is used in combination with 2374 * direct packet access. 2375 * Return 2376 * 0 on success, or a negative error in case of failure. 2377 * 2378 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2379 * Description 2380 * Pull in non-linear data in case the *skb* is non-linear and not 2381 * all of *len* are part of the linear section. Make *len* bytes 2382 * from *skb* readable and writable. If a zero value is passed for 2383 * *len*, then all bytes in the linear part of *skb* will be made 2384 * readable and writable. 2385 * 2386 * This helper is only needed for reading and writing with direct 2387 * packet access. 2388 * 2389 * For direct packet access, testing that offsets to access 2390 * are within packet boundaries (test on *skb*\ **->data_end**) is 2391 * susceptible to fail if offsets are invalid, or if the requested 2392 * data is in non-linear parts of the *skb*. On failure the 2393 * program can just bail out, or in the case of a non-linear 2394 * buffer, use a helper to make the data available. The 2395 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2396 * the data. Another one consists in using **bpf_skb_pull_data** 2397 * to pull in once the non-linear parts, then retesting and 2398 * eventually access the data. 2399 * 2400 * At the same time, this also makes sure the *skb* is uncloned, 2401 * which is a necessary condition for direct write. As this needs 2402 * to be an invariant for the write part only, the verifier 2403 * detects writes and adds a prologue that is calling 2404 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2405 * the very beginning in case it is indeed cloned. 2406 * 2407 * A call to this helper is susceptible to change the underlying 2408 * packet buffer. Therefore, at load time, all checks on pointers 2409 * previously done by the verifier are invalidated and must be 2410 * performed again, if the helper is used in combination with 2411 * direct packet access. 2412 * Return 2413 * 0 on success, or a negative error in case of failure. 2414 * 2415 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2416 * Description 2417 * Add the checksum *csum* into *skb*\ **->csum** in case the 2418 * driver has supplied a checksum for the entire packet into that 2419 * field. Return an error otherwise. This helper is intended to be 2420 * used in combination with **bpf_csum_diff**\ (), in particular 2421 * when the checksum needs to be updated after data has been 2422 * written into the packet through direct packet access. 2423 * Return 2424 * The checksum on success, or a negative error code in case of 2425 * failure. 2426 * 2427 * void bpf_set_hash_invalid(struct sk_buff *skb) 2428 * Description 2429 * Invalidate the current *skb*\ **->hash**. It can be used after 2430 * mangling on headers through direct packet access, in order to 2431 * indicate that the hash is outdated and to trigger a 2432 * recalculation the next time the kernel tries to access this 2433 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2434 * Return 2435 * void. 2436 * 2437 * long bpf_get_numa_node_id(void) 2438 * Description 2439 * Return the id of the current NUMA node. The primary use case 2440 * for this helper is the selection of sockets for the local NUMA 2441 * node, when the program is attached to sockets using the 2442 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2443 * but the helper is also available to other eBPF program types, 2444 * similarly to **bpf_get_smp_processor_id**\ (). 2445 * Return 2446 * The id of current NUMA node. 2447 * 2448 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2449 * Description 2450 * Grows headroom of packet associated to *skb* and adjusts the 2451 * offset of the MAC header accordingly, adding *len* bytes of 2452 * space. It automatically extends and reallocates memory as 2453 * required. 2454 * 2455 * This helper can be used on a layer 3 *skb* to push a MAC header 2456 * for redirection into a layer 2 device. 2457 * 2458 * All values for *flags* are reserved for future usage, and must 2459 * be left at zero. 2460 * 2461 * A call to this helper is susceptible to change the underlying 2462 * packet buffer. Therefore, at load time, all checks on pointers 2463 * previously done by the verifier are invalidated and must be 2464 * performed again, if the helper is used in combination with 2465 * direct packet access. 2466 * Return 2467 * 0 on success, or a negative error in case of failure. 2468 * 2469 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2470 * Description 2471 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2472 * it is possible to use a negative value for *delta*. This helper 2473 * can be used to prepare the packet for pushing or popping 2474 * headers. 2475 * 2476 * A call to this helper is susceptible to change the underlying 2477 * packet buffer. Therefore, at load time, all checks on pointers 2478 * previously done by the verifier are invalidated and must be 2479 * performed again, if the helper is used in combination with 2480 * direct packet access. 2481 * Return 2482 * 0 on success, or a negative error in case of failure. 2483 * 2484 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2485 * Description 2486 * Copy a NUL terminated string from an unsafe kernel address 2487 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2488 * more details. 2489 * 2490 * Generally, use **bpf_probe_read_user_str**\ () or 2491 * **bpf_probe_read_kernel_str**\ () instead. 2492 * Return 2493 * On success, the strictly positive length of the string, 2494 * including the trailing NUL character. On error, a negative 2495 * value. 2496 * 2497 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2498 * Description 2499 * If the **struct sk_buff** pointed by *skb* has a known socket, 2500 * retrieve the cookie (generated by the kernel) of this socket. 2501 * If no cookie has been set yet, generate a new cookie. Once 2502 * generated, the socket cookie remains stable for the life of the 2503 * socket. This helper can be useful for monitoring per socket 2504 * networking traffic statistics as it provides a global socket 2505 * identifier that can be assumed unique. 2506 * Return 2507 * A 8-byte long unique number on success, or 0 if the socket 2508 * field is missing inside *skb*. 2509 * 2510 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2511 * Description 2512 * Equivalent to bpf_get_socket_cookie() helper that accepts 2513 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2514 * Return 2515 * A 8-byte long unique number. 2516 * 2517 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2518 * Description 2519 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2520 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2521 * Return 2522 * A 8-byte long unique number. 2523 * 2524 * u64 bpf_get_socket_cookie(struct sock *sk) 2525 * Description 2526 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2527 * *sk*, but gets socket from a BTF **struct sock**. This helper 2528 * also works for sleepable programs. 2529 * Return 2530 * A 8-byte long unique number or 0 if *sk* is NULL. 2531 * 2532 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2533 * Description 2534 * Get the owner UID of the socked associated to *skb*. 2535 * Return 2536 * The owner UID of the socket associated to *skb*. If the socket 2537 * is **NULL**, or if it is not a full socket (i.e. if it is a 2538 * time-wait or a request socket instead), **overflowuid** value 2539 * is returned (note that **overflowuid** might also be the actual 2540 * UID value for the socket). 2541 * 2542 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2543 * Description 2544 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2545 * to value *hash*. 2546 * Return 2547 * 0 2548 * 2549 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2550 * Description 2551 * Emulate a call to **setsockopt()** on the socket associated to 2552 * *bpf_socket*, which must be a full socket. The *level* at 2553 * which the option resides and the name *optname* of the option 2554 * must be specified, see **setsockopt(2)** for more information. 2555 * The option value of length *optlen* is pointed by *optval*. 2556 * 2557 * *bpf_socket* should be one of the following: 2558 * 2559 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2560 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2561 * and **BPF_CGROUP_INET6_CONNECT**. 2562 * 2563 * This helper actually implements a subset of **setsockopt()**. 2564 * It supports the following *level*\ s: 2565 * 2566 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2567 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2568 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2569 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 2570 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2571 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2572 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2573 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2574 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 2575 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2576 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2577 * Return 2578 * 0 on success, or a negative error in case of failure. 2579 * 2580 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2581 * Description 2582 * Grow or shrink the room for data in the packet associated to 2583 * *skb* by *len_diff*, and according to the selected *mode*. 2584 * 2585 * By default, the helper will reset any offloaded checksum 2586 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2587 * by the following flag: 2588 * 2589 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2590 * checksum data of the skb to CHECKSUM_NONE. 2591 * 2592 * There are two supported modes at this time: 2593 * 2594 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2595 * (room space is added or removed between the layer 2 and 2596 * layer 3 headers). 2597 * 2598 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2599 * (room space is added or removed between the layer 3 and 2600 * layer 4 headers). 2601 * 2602 * The following flags are supported at this time: 2603 * 2604 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2605 * Adjusting mss in this way is not allowed for datagrams. 2606 * 2607 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2608 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2609 * Any new space is reserved to hold a tunnel header. 2610 * Configure skb offsets and other fields accordingly. 2611 * 2612 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2613 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2614 * Use with ENCAP_L3 flags to further specify the tunnel type. 2615 * 2616 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2617 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2618 * type; *len* is the length of the inner MAC header. 2619 * 2620 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2621 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2622 * L2 type as Ethernet. 2623 * 2624 * A call to this helper is susceptible to change the underlying 2625 * packet buffer. Therefore, at load time, all checks on pointers 2626 * previously done by the verifier are invalidated and must be 2627 * performed again, if the helper is used in combination with 2628 * direct packet access. 2629 * Return 2630 * 0 on success, or a negative error in case of failure. 2631 * 2632 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2633 * Description 2634 * Redirect the packet to the endpoint referenced by *map* at 2635 * index *key*. Depending on its type, this *map* can contain 2636 * references to net devices (for forwarding packets through other 2637 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2638 * but this is only implemented for native XDP (with driver 2639 * support) as of this writing). 2640 * 2641 * The lower two bits of *flags* are used as the return code if 2642 * the map lookup fails. This is so that the return value can be 2643 * one of the XDP program return codes up to **XDP_TX**, as chosen 2644 * by the caller. The higher bits of *flags* can be set to 2645 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2646 * 2647 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2648 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2649 * interface will be excluded when do broadcasting. 2650 * 2651 * See also **bpf_redirect**\ (), which only supports redirecting 2652 * to an ifindex, but doesn't require a map to do so. 2653 * Return 2654 * **XDP_REDIRECT** on success, or the value of the two lower bits 2655 * of the *flags* argument on error. 2656 * 2657 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2658 * Description 2659 * Redirect the packet to the socket referenced by *map* (of type 2660 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2661 * egress interfaces can be used for redirection. The 2662 * **BPF_F_INGRESS** value in *flags* is used to make the 2663 * distinction (ingress path is selected if the flag is present, 2664 * egress path otherwise). This is the only flag supported for now. 2665 * Return 2666 * **SK_PASS** on success, or **SK_DROP** on error. 2667 * 2668 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2669 * Description 2670 * Add an entry to, or update a *map* referencing sockets. The 2671 * *skops* is used as a new value for the entry associated to 2672 * *key*. *flags* is one of: 2673 * 2674 * **BPF_NOEXIST** 2675 * The entry for *key* must not exist in the map. 2676 * **BPF_EXIST** 2677 * The entry for *key* must already exist in the map. 2678 * **BPF_ANY** 2679 * No condition on the existence of the entry for *key*. 2680 * 2681 * If the *map* has eBPF programs (parser and verdict), those will 2682 * be inherited by the socket being added. If the socket is 2683 * already attached to eBPF programs, this results in an error. 2684 * Return 2685 * 0 on success, or a negative error in case of failure. 2686 * 2687 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2688 * Description 2689 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2690 * *delta* (which can be positive or negative). Note that this 2691 * operation modifies the address stored in *xdp_md*\ **->data**, 2692 * so the latter must be loaded only after the helper has been 2693 * called. 2694 * 2695 * The use of *xdp_md*\ **->data_meta** is optional and programs 2696 * are not required to use it. The rationale is that when the 2697 * packet is processed with XDP (e.g. as DoS filter), it is 2698 * possible to push further meta data along with it before passing 2699 * to the stack, and to give the guarantee that an ingress eBPF 2700 * program attached as a TC classifier on the same device can pick 2701 * this up for further post-processing. Since TC works with socket 2702 * buffers, it remains possible to set from XDP the **mark** or 2703 * **priority** pointers, or other pointers for the socket buffer. 2704 * Having this scratch space generic and programmable allows for 2705 * more flexibility as the user is free to store whatever meta 2706 * data they need. 2707 * 2708 * A call to this helper is susceptible to change the underlying 2709 * packet buffer. Therefore, at load time, all checks on pointers 2710 * previously done by the verifier are invalidated and must be 2711 * performed again, if the helper is used in combination with 2712 * direct packet access. 2713 * Return 2714 * 0 on success, or a negative error in case of failure. 2715 * 2716 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2717 * Description 2718 * Read the value of a perf event counter, and store it into *buf* 2719 * of size *buf_size*. This helper relies on a *map* of type 2720 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2721 * counter is selected when *map* is updated with perf event file 2722 * descriptors. The *map* is an array whose size is the number of 2723 * available CPUs, and each cell contains a value relative to one 2724 * CPU. The value to retrieve is indicated by *flags*, that 2725 * contains the index of the CPU to look up, masked with 2726 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2727 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2728 * current CPU should be retrieved. 2729 * 2730 * This helper behaves in a way close to 2731 * **bpf_perf_event_read**\ () helper, save that instead of 2732 * just returning the value observed, it fills the *buf* 2733 * structure. This allows for additional data to be retrieved: in 2734 * particular, the enabled and running times (in *buf*\ 2735 * **->enabled** and *buf*\ **->running**, respectively) are 2736 * copied. In general, **bpf_perf_event_read_value**\ () is 2737 * recommended over **bpf_perf_event_read**\ (), which has some 2738 * ABI issues and provides fewer functionalities. 2739 * 2740 * These values are interesting, because hardware PMU (Performance 2741 * Monitoring Unit) counters are limited resources. When there are 2742 * more PMU based perf events opened than available counters, 2743 * kernel will multiplex these events so each event gets certain 2744 * percentage (but not all) of the PMU time. In case that 2745 * multiplexing happens, the number of samples or counter value 2746 * will not reflect the case compared to when no multiplexing 2747 * occurs. This makes comparison between different runs difficult. 2748 * Typically, the counter value should be normalized before 2749 * comparing to other experiments. The usual normalization is done 2750 * as follows. 2751 * 2752 * :: 2753 * 2754 * normalized_counter = counter * t_enabled / t_running 2755 * 2756 * Where t_enabled is the time enabled for event and t_running is 2757 * the time running for event since last normalization. The 2758 * enabled and running times are accumulated since the perf event 2759 * open. To achieve scaling factor between two invocations of an 2760 * eBPF program, users can use CPU id as the key (which is 2761 * typical for perf array usage model) to remember the previous 2762 * value and do the calculation inside the eBPF program. 2763 * Return 2764 * 0 on success, or a negative error in case of failure. 2765 * 2766 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2767 * Description 2768 * For en eBPF program attached to a perf event, retrieve the 2769 * value of the event counter associated to *ctx* and store it in 2770 * the structure pointed by *buf* and of size *buf_size*. Enabled 2771 * and running times are also stored in the structure (see 2772 * description of helper **bpf_perf_event_read_value**\ () for 2773 * more details). 2774 * Return 2775 * 0 on success, or a negative error in case of failure. 2776 * 2777 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2778 * Description 2779 * Emulate a call to **getsockopt()** on the socket associated to 2780 * *bpf_socket*, which must be a full socket. The *level* at 2781 * which the option resides and the name *optname* of the option 2782 * must be specified, see **getsockopt(2)** for more information. 2783 * The retrieved value is stored in the structure pointed by 2784 * *opval* and of length *optlen*. 2785 * 2786 * *bpf_socket* should be one of the following: 2787 * 2788 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2789 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2790 * and **BPF_CGROUP_INET6_CONNECT**. 2791 * 2792 * This helper actually implements a subset of **getsockopt()**. 2793 * It supports the following *level*\ s: 2794 * 2795 * * **IPPROTO_TCP**, which supports *optname* 2796 * **TCP_CONGESTION**. 2797 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2798 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2799 * Return 2800 * 0 on success, or a negative error in case of failure. 2801 * 2802 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2803 * Description 2804 * Used for error injection, this helper uses kprobes to override 2805 * the return value of the probed function, and to set it to *rc*. 2806 * The first argument is the context *regs* on which the kprobe 2807 * works. 2808 * 2809 * This helper works by setting the PC (program counter) 2810 * to an override function which is run in place of the original 2811 * probed function. This means the probed function is not run at 2812 * all. The replacement function just returns with the required 2813 * value. 2814 * 2815 * This helper has security implications, and thus is subject to 2816 * restrictions. It is only available if the kernel was compiled 2817 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2818 * option, and in this case it only works on functions tagged with 2819 * **ALLOW_ERROR_INJECTION** in the kernel code. 2820 * 2821 * Also, the helper is only available for the architectures having 2822 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2823 * x86 architecture is the only one to support this feature. 2824 * Return 2825 * 0 2826 * 2827 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2828 * Description 2829 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2830 * for the full TCP socket associated to *bpf_sock_ops* to 2831 * *argval*. 2832 * 2833 * The primary use of this field is to determine if there should 2834 * be calls to eBPF programs of type 2835 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2836 * code. A program of the same type can change its value, per 2837 * connection and as necessary, when the connection is 2838 * established. This field is directly accessible for reading, but 2839 * this helper must be used for updates in order to return an 2840 * error if an eBPF program tries to set a callback that is not 2841 * supported in the current kernel. 2842 * 2843 * *argval* is a flag array which can combine these flags: 2844 * 2845 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2846 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2847 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2848 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2849 * 2850 * Therefore, this function can be used to clear a callback flag by 2851 * setting the appropriate bit to zero. e.g. to disable the RTO 2852 * callback: 2853 * 2854 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2855 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2856 * 2857 * Here are some examples of where one could call such eBPF 2858 * program: 2859 * 2860 * * When RTO fires. 2861 * * When a packet is retransmitted. 2862 * * When the connection terminates. 2863 * * When a packet is sent. 2864 * * When a packet is received. 2865 * Return 2866 * Code **-EINVAL** if the socket is not a full TCP socket; 2867 * otherwise, a positive number containing the bits that could not 2868 * be set is returned (which comes down to 0 if all bits were set 2869 * as required). 2870 * 2871 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2872 * Description 2873 * This helper is used in programs implementing policies at the 2874 * socket level. If the message *msg* is allowed to pass (i.e. if 2875 * the verdict eBPF program returns **SK_PASS**), redirect it to 2876 * the socket referenced by *map* (of type 2877 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2878 * egress interfaces can be used for redirection. The 2879 * **BPF_F_INGRESS** value in *flags* is used to make the 2880 * distinction (ingress path is selected if the flag is present, 2881 * egress path otherwise). This is the only flag supported for now. 2882 * Return 2883 * **SK_PASS** on success, or **SK_DROP** on error. 2884 * 2885 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2886 * Description 2887 * For socket policies, apply the verdict of the eBPF program to 2888 * the next *bytes* (number of bytes) of message *msg*. 2889 * 2890 * For example, this helper can be used in the following cases: 2891 * 2892 * * A single **sendmsg**\ () or **sendfile**\ () system call 2893 * contains multiple logical messages that the eBPF program is 2894 * supposed to read and for which it should apply a verdict. 2895 * * An eBPF program only cares to read the first *bytes* of a 2896 * *msg*. If the message has a large payload, then setting up 2897 * and calling the eBPF program repeatedly for all bytes, even 2898 * though the verdict is already known, would create unnecessary 2899 * overhead. 2900 * 2901 * When called from within an eBPF program, the helper sets a 2902 * counter internal to the BPF infrastructure, that is used to 2903 * apply the last verdict to the next *bytes*. If *bytes* is 2904 * smaller than the current data being processed from a 2905 * **sendmsg**\ () or **sendfile**\ () system call, the first 2906 * *bytes* will be sent and the eBPF program will be re-run with 2907 * the pointer for start of data pointing to byte number *bytes* 2908 * **+ 1**. If *bytes* is larger than the current data being 2909 * processed, then the eBPF verdict will be applied to multiple 2910 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2911 * consumed. 2912 * 2913 * Note that if a socket closes with the internal counter holding 2914 * a non-zero value, this is not a problem because data is not 2915 * being buffered for *bytes* and is sent as it is received. 2916 * Return 2917 * 0 2918 * 2919 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2920 * Description 2921 * For socket policies, prevent the execution of the verdict eBPF 2922 * program for message *msg* until *bytes* (byte number) have been 2923 * accumulated. 2924 * 2925 * This can be used when one needs a specific number of bytes 2926 * before a verdict can be assigned, even if the data spans 2927 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2928 * case would be a user calling **sendmsg**\ () repeatedly with 2929 * 1-byte long message segments. Obviously, this is bad for 2930 * performance, but it is still valid. If the eBPF program needs 2931 * *bytes* bytes to validate a header, this helper can be used to 2932 * prevent the eBPF program to be called again until *bytes* have 2933 * been accumulated. 2934 * Return 2935 * 0 2936 * 2937 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2938 * Description 2939 * For socket policies, pull in non-linear data from user space 2940 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2941 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2942 * respectively. 2943 * 2944 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2945 * *msg* it can only parse data that the (**data**, **data_end**) 2946 * pointers have already consumed. For **sendmsg**\ () hooks this 2947 * is likely the first scatterlist element. But for calls relying 2948 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2949 * be the range (**0**, **0**) because the data is shared with 2950 * user space and by default the objective is to avoid allowing 2951 * user space to modify data while (or after) eBPF verdict is 2952 * being decided. This helper can be used to pull in data and to 2953 * set the start and end pointer to given values. Data will be 2954 * copied if necessary (i.e. if data was not linear and if start 2955 * and end pointers do not point to the same chunk). 2956 * 2957 * A call to this helper is susceptible to change the underlying 2958 * packet buffer. Therefore, at load time, all checks on pointers 2959 * previously done by the verifier are invalidated and must be 2960 * performed again, if the helper is used in combination with 2961 * direct packet access. 2962 * 2963 * All values for *flags* are reserved for future usage, and must 2964 * be left at zero. 2965 * Return 2966 * 0 on success, or a negative error in case of failure. 2967 * 2968 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2969 * Description 2970 * Bind the socket associated to *ctx* to the address pointed by 2971 * *addr*, of length *addr_len*. This allows for making outgoing 2972 * connection from the desired IP address, which can be useful for 2973 * example when all processes inside a cgroup should use one 2974 * single IP address on a host that has multiple IP configured. 2975 * 2976 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2977 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2978 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2979 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2980 * behavior and lets the kernel efficiently pick up an unused 2981 * port as long as 4-tuple is unique. Passing non-zero port might 2982 * lead to degraded performance. 2983 * Return 2984 * 0 on success, or a negative error in case of failure. 2985 * 2986 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2987 * Description 2988 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2989 * possible to both shrink and grow the packet tail. 2990 * Shrink done via *delta* being a negative integer. 2991 * 2992 * A call to this helper is susceptible to change the underlying 2993 * packet buffer. Therefore, at load time, all checks on pointers 2994 * previously done by the verifier are invalidated and must be 2995 * performed again, if the helper is used in combination with 2996 * direct packet access. 2997 * Return 2998 * 0 on success, or a negative error in case of failure. 2999 * 3000 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3001 * Description 3002 * Retrieve the XFRM state (IP transform framework, see also 3003 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3004 * 3005 * The retrieved value is stored in the **struct bpf_xfrm_state** 3006 * pointed by *xfrm_state* and of length *size*. 3007 * 3008 * All values for *flags* are reserved for future usage, and must 3009 * be left at zero. 3010 * 3011 * This helper is available only if the kernel was compiled with 3012 * **CONFIG_XFRM** configuration option. 3013 * Return 3014 * 0 on success, or a negative error in case of failure. 3015 * 3016 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3017 * Description 3018 * Return a user or a kernel stack in bpf program provided buffer. 3019 * To achieve this, the helper needs *ctx*, which is a pointer 3020 * to the context on which the tracing program is executed. 3021 * To store the stacktrace, the bpf program provides *buf* with 3022 * a nonnegative *size*. 3023 * 3024 * The last argument, *flags*, holds the number of stack frames to 3025 * skip (from 0 to 255), masked with 3026 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3027 * the following flags: 3028 * 3029 * **BPF_F_USER_STACK** 3030 * Collect a user space stack instead of a kernel stack. 3031 * **BPF_F_USER_BUILD_ID** 3032 * Collect (build_id, file_offset) instead of ips for user 3033 * stack, only valid if **BPF_F_USER_STACK** is also 3034 * specified. 3035 * 3036 * *file_offset* is an offset relative to the beginning 3037 * of the executable or shared object file backing the vma 3038 * which the *ip* falls in. It is *not* an offset relative 3039 * to that object's base address. Accordingly, it must be 3040 * adjusted by adding (sh_addr - sh_offset), where 3041 * sh_{addr,offset} correspond to the executable section 3042 * containing *file_offset* in the object, for comparisons 3043 * to symbols' st_value to be valid. 3044 * 3045 * **bpf_get_stack**\ () can collect up to 3046 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3047 * to sufficient large buffer size. Note that 3048 * this limit can be controlled with the **sysctl** program, and 3049 * that it should be manually increased in order to profile long 3050 * user stacks (such as stacks for Java programs). To do so, use: 3051 * 3052 * :: 3053 * 3054 * # sysctl kernel.perf_event_max_stack=<new value> 3055 * Return 3056 * The non-negative copied *buf* length equal to or less than 3057 * *size* on success, or a negative error in case of failure. 3058 * 3059 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3060 * Description 3061 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3062 * it provides an easy way to load *len* bytes from *offset* 3063 * from the packet associated to *skb*, into the buffer pointed 3064 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3065 * a fifth argument *start_header* exists in order to select a 3066 * base offset to start from. *start_header* can be one of: 3067 * 3068 * **BPF_HDR_START_MAC** 3069 * Base offset to load data from is *skb*'s mac header. 3070 * **BPF_HDR_START_NET** 3071 * Base offset to load data from is *skb*'s network header. 3072 * 3073 * In general, "direct packet access" is the preferred method to 3074 * access packet data, however, this helper is in particular useful 3075 * in socket filters where *skb*\ **->data** does not always point 3076 * to the start of the mac header and where "direct packet access" 3077 * is not available. 3078 * Return 3079 * 0 on success, or a negative error in case of failure. 3080 * 3081 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3082 * Description 3083 * Do FIB lookup in kernel tables using parameters in *params*. 3084 * If lookup is successful and result shows packet is to be 3085 * forwarded, the neighbor tables are searched for the nexthop. 3086 * If successful (ie., FIB lookup shows forwarding and nexthop 3087 * is resolved), the nexthop address is returned in ipv4_dst 3088 * or ipv6_dst based on family, smac is set to mac address of 3089 * egress device, dmac is set to nexthop mac address, rt_metric 3090 * is set to metric from route (IPv4/IPv6 only), and ifindex 3091 * is set to the device index of the nexthop from the FIB lookup. 3092 * 3093 * *plen* argument is the size of the passed in struct. 3094 * *flags* argument can be a combination of one or more of the 3095 * following values: 3096 * 3097 * **BPF_FIB_LOOKUP_DIRECT** 3098 * Do a direct table lookup vs full lookup using FIB 3099 * rules. 3100 * **BPF_FIB_LOOKUP_OUTPUT** 3101 * Perform lookup from an egress perspective (default is 3102 * ingress). 3103 * 3104 * *ctx* is either **struct xdp_md** for XDP programs or 3105 * **struct sk_buff** tc cls_act programs. 3106 * Return 3107 * * < 0 if any input argument is invalid 3108 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3109 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3110 * packet is not forwarded or needs assist from full stack 3111 * 3112 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3113 * was exceeded and output params->mtu_result contains the MTU. 3114 * 3115 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3116 * Description 3117 * Add an entry to, or update a sockhash *map* referencing sockets. 3118 * The *skops* is used as a new value for the entry associated to 3119 * *key*. *flags* is one of: 3120 * 3121 * **BPF_NOEXIST** 3122 * The entry for *key* must not exist in the map. 3123 * **BPF_EXIST** 3124 * The entry for *key* must already exist in the map. 3125 * **BPF_ANY** 3126 * No condition on the existence of the entry for *key*. 3127 * 3128 * If the *map* has eBPF programs (parser and verdict), those will 3129 * be inherited by the socket being added. If the socket is 3130 * already attached to eBPF programs, this results in an error. 3131 * Return 3132 * 0 on success, or a negative error in case of failure. 3133 * 3134 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3135 * Description 3136 * This helper is used in programs implementing policies at the 3137 * socket level. If the message *msg* is allowed to pass (i.e. if 3138 * the verdict eBPF program returns **SK_PASS**), redirect it to 3139 * the socket referenced by *map* (of type 3140 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3141 * egress interfaces can be used for redirection. The 3142 * **BPF_F_INGRESS** value in *flags* is used to make the 3143 * distinction (ingress path is selected if the flag is present, 3144 * egress path otherwise). This is the only flag supported for now. 3145 * Return 3146 * **SK_PASS** on success, or **SK_DROP** on error. 3147 * 3148 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3149 * Description 3150 * This helper is used in programs implementing policies at the 3151 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3152 * if the verdict eBPF program returns **SK_PASS**), redirect it 3153 * to the socket referenced by *map* (of type 3154 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3155 * egress interfaces can be used for redirection. The 3156 * **BPF_F_INGRESS** value in *flags* is used to make the 3157 * distinction (ingress path is selected if the flag is present, 3158 * egress otherwise). This is the only flag supported for now. 3159 * Return 3160 * **SK_PASS** on success, or **SK_DROP** on error. 3161 * 3162 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3163 * Description 3164 * Encapsulate the packet associated to *skb* within a Layer 3 3165 * protocol header. This header is provided in the buffer at 3166 * address *hdr*, with *len* its size in bytes. *type* indicates 3167 * the protocol of the header and can be one of: 3168 * 3169 * **BPF_LWT_ENCAP_SEG6** 3170 * IPv6 encapsulation with Segment Routing Header 3171 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3172 * the IPv6 header is computed by the kernel. 3173 * **BPF_LWT_ENCAP_SEG6_INLINE** 3174 * Only works if *skb* contains an IPv6 packet. Insert a 3175 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3176 * the IPv6 header. 3177 * **BPF_LWT_ENCAP_IP** 3178 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3179 * must be IPv4 or IPv6, followed by zero or more 3180 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3181 * total bytes in all prepended headers. Please note that 3182 * if **skb_is_gso**\ (*skb*) is true, no more than two 3183 * headers can be prepended, and the inner header, if 3184 * present, should be either GRE or UDP/GUE. 3185 * 3186 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3187 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3188 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3189 * **BPF_PROG_TYPE_LWT_XMIT**. 3190 * 3191 * A call to this helper is susceptible to change the underlying 3192 * packet buffer. Therefore, at load time, all checks on pointers 3193 * previously done by the verifier are invalidated and must be 3194 * performed again, if the helper is used in combination with 3195 * direct packet access. 3196 * Return 3197 * 0 on success, or a negative error in case of failure. 3198 * 3199 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3200 * Description 3201 * Store *len* bytes from address *from* into the packet 3202 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3203 * inside the outermost IPv6 Segment Routing Header can be 3204 * modified through this helper. 3205 * 3206 * A call to this helper is susceptible to change the underlying 3207 * packet buffer. Therefore, at load time, all checks on pointers 3208 * previously done by the verifier are invalidated and must be 3209 * performed again, if the helper is used in combination with 3210 * direct packet access. 3211 * Return 3212 * 0 on success, or a negative error in case of failure. 3213 * 3214 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3215 * Description 3216 * Adjust the size allocated to TLVs in the outermost IPv6 3217 * Segment Routing Header contained in the packet associated to 3218 * *skb*, at position *offset* by *delta* bytes. Only offsets 3219 * after the segments are accepted. *delta* can be as well 3220 * positive (growing) as negative (shrinking). 3221 * 3222 * A call to this helper is susceptible to change the underlying 3223 * packet buffer. Therefore, at load time, all checks on pointers 3224 * previously done by the verifier are invalidated and must be 3225 * performed again, if the helper is used in combination with 3226 * direct packet access. 3227 * Return 3228 * 0 on success, or a negative error in case of failure. 3229 * 3230 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3231 * Description 3232 * Apply an IPv6 Segment Routing action of type *action* to the 3233 * packet associated to *skb*. Each action takes a parameter 3234 * contained at address *param*, and of length *param_len* bytes. 3235 * *action* can be one of: 3236 * 3237 * **SEG6_LOCAL_ACTION_END_X** 3238 * End.X action: Endpoint with Layer-3 cross-connect. 3239 * Type of *param*: **struct in6_addr**. 3240 * **SEG6_LOCAL_ACTION_END_T** 3241 * End.T action: Endpoint with specific IPv6 table lookup. 3242 * Type of *param*: **int**. 3243 * **SEG6_LOCAL_ACTION_END_B6** 3244 * End.B6 action: Endpoint bound to an SRv6 policy. 3245 * Type of *param*: **struct ipv6_sr_hdr**. 3246 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3247 * End.B6.Encap action: Endpoint bound to an SRv6 3248 * encapsulation policy. 3249 * Type of *param*: **struct ipv6_sr_hdr**. 3250 * 3251 * A call to this helper is susceptible to change the underlying 3252 * packet buffer. Therefore, at load time, all checks on pointers 3253 * previously done by the verifier are invalidated and must be 3254 * performed again, if the helper is used in combination with 3255 * direct packet access. 3256 * Return 3257 * 0 on success, or a negative error in case of failure. 3258 * 3259 * long bpf_rc_repeat(void *ctx) 3260 * Description 3261 * This helper is used in programs implementing IR decoding, to 3262 * report a successfully decoded repeat key message. This delays 3263 * the generation of a key up event for previously generated 3264 * key down event. 3265 * 3266 * Some IR protocols like NEC have a special IR message for 3267 * repeating last button, for when a button is held down. 3268 * 3269 * The *ctx* should point to the lirc sample as passed into 3270 * the program. 3271 * 3272 * This helper is only available is the kernel was compiled with 3273 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3274 * "**y**". 3275 * Return 3276 * 0 3277 * 3278 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3279 * Description 3280 * This helper is used in programs implementing IR decoding, to 3281 * report a successfully decoded key press with *scancode*, 3282 * *toggle* value in the given *protocol*. The scancode will be 3283 * translated to a keycode using the rc keymap, and reported as 3284 * an input key down event. After a period a key up event is 3285 * generated. This period can be extended by calling either 3286 * **bpf_rc_keydown**\ () again with the same values, or calling 3287 * **bpf_rc_repeat**\ (). 3288 * 3289 * Some protocols include a toggle bit, in case the button was 3290 * released and pressed again between consecutive scancodes. 3291 * 3292 * The *ctx* should point to the lirc sample as passed into 3293 * the program. 3294 * 3295 * The *protocol* is the decoded protocol number (see 3296 * **enum rc_proto** for some predefined values). 3297 * 3298 * This helper is only available is the kernel was compiled with 3299 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3300 * "**y**". 3301 * Return 3302 * 0 3303 * 3304 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3305 * Description 3306 * Return the cgroup v2 id of the socket associated with the *skb*. 3307 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3308 * helper for cgroup v1 by providing a tag resp. identifier that 3309 * can be matched on or used for map lookups e.g. to implement 3310 * policy. The cgroup v2 id of a given path in the hierarchy is 3311 * exposed in user space through the f_handle API in order to get 3312 * to the same 64-bit id. 3313 * 3314 * This helper can be used on TC egress path, but not on ingress, 3315 * and is available only if the kernel was compiled with the 3316 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3317 * Return 3318 * The id is returned or 0 in case the id could not be retrieved. 3319 * 3320 * u64 bpf_get_current_cgroup_id(void) 3321 * Description 3322 * Get the current cgroup id based on the cgroup within which 3323 * the current task is running. 3324 * Return 3325 * A 64-bit integer containing the current cgroup id based 3326 * on the cgroup within which the current task is running. 3327 * 3328 * void *bpf_get_local_storage(void *map, u64 flags) 3329 * Description 3330 * Get the pointer to the local storage area. 3331 * The type and the size of the local storage is defined 3332 * by the *map* argument. 3333 * The *flags* meaning is specific for each map type, 3334 * and has to be 0 for cgroup local storage. 3335 * 3336 * Depending on the BPF program type, a local storage area 3337 * can be shared between multiple instances of the BPF program, 3338 * running simultaneously. 3339 * 3340 * A user should care about the synchronization by himself. 3341 * For example, by using the **BPF_ATOMIC** instructions to alter 3342 * the shared data. 3343 * Return 3344 * A pointer to the local storage area. 3345 * 3346 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3347 * Description 3348 * Select a **SO_REUSEPORT** socket from a 3349 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3350 * It checks the selected socket is matching the incoming 3351 * request in the socket buffer. 3352 * Return 3353 * 0 on success, or a negative error in case of failure. 3354 * 3355 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3356 * Description 3357 * Return id of cgroup v2 that is ancestor of cgroup associated 3358 * with the *skb* at the *ancestor_level*. The root cgroup is at 3359 * *ancestor_level* zero and each step down the hierarchy 3360 * increments the level. If *ancestor_level* == level of cgroup 3361 * associated with *skb*, then return value will be same as that 3362 * of **bpf_skb_cgroup_id**\ (). 3363 * 3364 * The helper is useful to implement policies based on cgroups 3365 * that are upper in hierarchy than immediate cgroup associated 3366 * with *skb*. 3367 * 3368 * The format of returned id and helper limitations are same as in 3369 * **bpf_skb_cgroup_id**\ (). 3370 * Return 3371 * The id is returned or 0 in case the id could not be retrieved. 3372 * 3373 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3374 * Description 3375 * Look for TCP socket matching *tuple*, optionally in a child 3376 * network namespace *netns*. The return value must be checked, 3377 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3378 * 3379 * The *ctx* should point to the context of the program, such as 3380 * the skb or socket (depending on the hook in use). This is used 3381 * to determine the base network namespace for the lookup. 3382 * 3383 * *tuple_size* must be one of: 3384 * 3385 * **sizeof**\ (*tuple*\ **->ipv4**) 3386 * Look for an IPv4 socket. 3387 * **sizeof**\ (*tuple*\ **->ipv6**) 3388 * Look for an IPv6 socket. 3389 * 3390 * If the *netns* is a negative signed 32-bit integer, then the 3391 * socket lookup table in the netns associated with the *ctx* 3392 * will be used. For the TC hooks, this is the netns of the device 3393 * in the skb. For socket hooks, this is the netns of the socket. 3394 * If *netns* is any other signed 32-bit value greater than or 3395 * equal to zero then it specifies the ID of the netns relative to 3396 * the netns associated with the *ctx*. *netns* values beyond the 3397 * range of 32-bit integers are reserved for future use. 3398 * 3399 * All values for *flags* are reserved for future usage, and must 3400 * be left at zero. 3401 * 3402 * This helper is available only if the kernel was compiled with 3403 * **CONFIG_NET** configuration option. 3404 * Return 3405 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3406 * For sockets with reuseport option, the **struct bpf_sock** 3407 * result is from *reuse*\ **->socks**\ [] using the hash of the 3408 * tuple. 3409 * 3410 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3411 * Description 3412 * Look for UDP socket matching *tuple*, optionally in a child 3413 * network namespace *netns*. The return value must be checked, 3414 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3415 * 3416 * The *ctx* should point to the context of the program, such as 3417 * the skb or socket (depending on the hook in use). This is used 3418 * to determine the base network namespace for the lookup. 3419 * 3420 * *tuple_size* must be one of: 3421 * 3422 * **sizeof**\ (*tuple*\ **->ipv4**) 3423 * Look for an IPv4 socket. 3424 * **sizeof**\ (*tuple*\ **->ipv6**) 3425 * Look for an IPv6 socket. 3426 * 3427 * If the *netns* is a negative signed 32-bit integer, then the 3428 * socket lookup table in the netns associated with the *ctx* 3429 * will be used. For the TC hooks, this is the netns of the device 3430 * in the skb. For socket hooks, this is the netns of the socket. 3431 * If *netns* is any other signed 32-bit value greater than or 3432 * equal to zero then it specifies the ID of the netns relative to 3433 * the netns associated with the *ctx*. *netns* values beyond the 3434 * range of 32-bit integers are reserved for future use. 3435 * 3436 * All values for *flags* are reserved for future usage, and must 3437 * be left at zero. 3438 * 3439 * This helper is available only if the kernel was compiled with 3440 * **CONFIG_NET** configuration option. 3441 * Return 3442 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3443 * For sockets with reuseport option, the **struct bpf_sock** 3444 * result is from *reuse*\ **->socks**\ [] using the hash of the 3445 * tuple. 3446 * 3447 * long bpf_sk_release(void *sock) 3448 * Description 3449 * Release the reference held by *sock*. *sock* must be a 3450 * non-**NULL** pointer that was returned from 3451 * **bpf_sk_lookup_xxx**\ (). 3452 * Return 3453 * 0 on success, or a negative error in case of failure. 3454 * 3455 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3456 * Description 3457 * Push an element *value* in *map*. *flags* is one of: 3458 * 3459 * **BPF_EXIST** 3460 * If the queue/stack is full, the oldest element is 3461 * removed to make room for this. 3462 * Return 3463 * 0 on success, or a negative error in case of failure. 3464 * 3465 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3466 * Description 3467 * Pop an element from *map*. 3468 * Return 3469 * 0 on success, or a negative error in case of failure. 3470 * 3471 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3472 * Description 3473 * Get an element from *map* without removing it. 3474 * Return 3475 * 0 on success, or a negative error in case of failure. 3476 * 3477 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3478 * Description 3479 * For socket policies, insert *len* bytes into *msg* at offset 3480 * *start*. 3481 * 3482 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3483 * *msg* it may want to insert metadata or options into the *msg*. 3484 * This can later be read and used by any of the lower layer BPF 3485 * hooks. 3486 * 3487 * This helper may fail if under memory pressure (a malloc 3488 * fails) in these cases BPF programs will get an appropriate 3489 * error and BPF programs will need to handle them. 3490 * Return 3491 * 0 on success, or a negative error in case of failure. 3492 * 3493 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3494 * Description 3495 * Will remove *len* bytes from a *msg* starting at byte *start*. 3496 * This may result in **ENOMEM** errors under certain situations if 3497 * an allocation and copy are required due to a full ring buffer. 3498 * However, the helper will try to avoid doing the allocation 3499 * if possible. Other errors can occur if input parameters are 3500 * invalid either due to *start* byte not being valid part of *msg* 3501 * payload and/or *pop* value being to large. 3502 * Return 3503 * 0 on success, or a negative error in case of failure. 3504 * 3505 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3506 * Description 3507 * This helper is used in programs implementing IR decoding, to 3508 * report a successfully decoded pointer movement. 3509 * 3510 * The *ctx* should point to the lirc sample as passed into 3511 * the program. 3512 * 3513 * This helper is only available is the kernel was compiled with 3514 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3515 * "**y**". 3516 * Return 3517 * 0 3518 * 3519 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3520 * Description 3521 * Acquire a spinlock represented by the pointer *lock*, which is 3522 * stored as part of a value of a map. Taking the lock allows to 3523 * safely update the rest of the fields in that value. The 3524 * spinlock can (and must) later be released with a call to 3525 * **bpf_spin_unlock**\ (\ *lock*\ ). 3526 * 3527 * Spinlocks in BPF programs come with a number of restrictions 3528 * and constraints: 3529 * 3530 * * **bpf_spin_lock** objects are only allowed inside maps of 3531 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3532 * list could be extended in the future). 3533 * * BTF description of the map is mandatory. 3534 * * The BPF program can take ONE lock at a time, since taking two 3535 * or more could cause dead locks. 3536 * * Only one **struct bpf_spin_lock** is allowed per map element. 3537 * * When the lock is taken, calls (either BPF to BPF or helpers) 3538 * are not allowed. 3539 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3540 * allowed inside a spinlock-ed region. 3541 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3542 * the lock, on all execution paths, before it returns. 3543 * * The BPF program can access **struct bpf_spin_lock** only via 3544 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3545 * helpers. Loading or storing data into the **struct 3546 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3547 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3548 * of the map value must be a struct and have **struct 3549 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3550 * Nested lock inside another struct is not allowed. 3551 * * The **struct bpf_spin_lock** *lock* field in a map value must 3552 * be aligned on a multiple of 4 bytes in that value. 3553 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3554 * the **bpf_spin_lock** field to user space. 3555 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3556 * a BPF program, do not update the **bpf_spin_lock** field. 3557 * * **bpf_spin_lock** cannot be on the stack or inside a 3558 * networking packet (it can only be inside of a map values). 3559 * * **bpf_spin_lock** is available to root only. 3560 * * Tracing programs and socket filter programs cannot use 3561 * **bpf_spin_lock**\ () due to insufficient preemption checks 3562 * (but this may change in the future). 3563 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3564 * Return 3565 * 0 3566 * 3567 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3568 * Description 3569 * Release the *lock* previously locked by a call to 3570 * **bpf_spin_lock**\ (\ *lock*\ ). 3571 * Return 3572 * 0 3573 * 3574 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3575 * Description 3576 * This helper gets a **struct bpf_sock** pointer such 3577 * that all the fields in this **bpf_sock** can be accessed. 3578 * Return 3579 * A **struct bpf_sock** pointer on success, or **NULL** in 3580 * case of failure. 3581 * 3582 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3583 * Description 3584 * This helper gets a **struct bpf_tcp_sock** pointer from a 3585 * **struct bpf_sock** pointer. 3586 * Return 3587 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3588 * case of failure. 3589 * 3590 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3591 * Description 3592 * Set ECN (Explicit Congestion Notification) field of IP header 3593 * to **CE** (Congestion Encountered) if current value is **ECT** 3594 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3595 * and IPv4. 3596 * Return 3597 * 1 if the **CE** flag is set (either by the current helper call 3598 * or because it was already present), 0 if it is not set. 3599 * 3600 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3601 * Description 3602 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3603 * **bpf_sk_release**\ () is unnecessary and not allowed. 3604 * Return 3605 * A **struct bpf_sock** pointer on success, or **NULL** in 3606 * case of failure. 3607 * 3608 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3609 * Description 3610 * Look for TCP socket matching *tuple*, optionally in a child 3611 * network namespace *netns*. The return value must be checked, 3612 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3613 * 3614 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3615 * that it also returns timewait or request sockets. Use 3616 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3617 * full structure. 3618 * 3619 * This helper is available only if the kernel was compiled with 3620 * **CONFIG_NET** configuration option. 3621 * Return 3622 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3623 * For sockets with reuseport option, the **struct bpf_sock** 3624 * result is from *reuse*\ **->socks**\ [] using the hash of the 3625 * tuple. 3626 * 3627 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3628 * Description 3629 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3630 * the listening socket in *sk*. 3631 * 3632 * *iph* points to the start of the IPv4 or IPv6 header, while 3633 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3634 * **sizeof**\ (**struct ipv6hdr**). 3635 * 3636 * *th* points to the start of the TCP header, while *th_len* 3637 * contains the length of the TCP header (at least 3638 * **sizeof**\ (**struct tcphdr**)). 3639 * Return 3640 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3641 * error otherwise. 3642 * 3643 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3644 * Description 3645 * Get name of sysctl in /proc/sys/ and copy it into provided by 3646 * program buffer *buf* of size *buf_len*. 3647 * 3648 * The buffer is always NUL terminated, unless it's zero-sized. 3649 * 3650 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3651 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3652 * only (e.g. "tcp_mem"). 3653 * Return 3654 * Number of character copied (not including the trailing NUL). 3655 * 3656 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3657 * truncated name in this case). 3658 * 3659 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3660 * Description 3661 * Get current value of sysctl as it is presented in /proc/sys 3662 * (incl. newline, etc), and copy it as a string into provided 3663 * by program buffer *buf* of size *buf_len*. 3664 * 3665 * The whole value is copied, no matter what file position user 3666 * space issued e.g. sys_read at. 3667 * 3668 * The buffer is always NUL terminated, unless it's zero-sized. 3669 * Return 3670 * Number of character copied (not including the trailing NUL). 3671 * 3672 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3673 * truncated name in this case). 3674 * 3675 * **-EINVAL** if current value was unavailable, e.g. because 3676 * sysctl is uninitialized and read returns -EIO for it. 3677 * 3678 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3679 * Description 3680 * Get new value being written by user space to sysctl (before 3681 * the actual write happens) and copy it as a string into 3682 * provided by program buffer *buf* of size *buf_len*. 3683 * 3684 * User space may write new value at file position > 0. 3685 * 3686 * The buffer is always NUL terminated, unless it's zero-sized. 3687 * Return 3688 * Number of character copied (not including the trailing NUL). 3689 * 3690 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3691 * truncated name in this case). 3692 * 3693 * **-EINVAL** if sysctl is being read. 3694 * 3695 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3696 * Description 3697 * Override new value being written by user space to sysctl with 3698 * value provided by program in buffer *buf* of size *buf_len*. 3699 * 3700 * *buf* should contain a string in same form as provided by user 3701 * space on sysctl write. 3702 * 3703 * User space may write new value at file position > 0. To override 3704 * the whole sysctl value file position should be set to zero. 3705 * Return 3706 * 0 on success. 3707 * 3708 * **-E2BIG** if the *buf_len* is too big. 3709 * 3710 * **-EINVAL** if sysctl is being read. 3711 * 3712 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3713 * Description 3714 * Convert the initial part of the string from buffer *buf* of 3715 * size *buf_len* to a long integer according to the given base 3716 * and save the result in *res*. 3717 * 3718 * The string may begin with an arbitrary amount of white space 3719 * (as determined by **isspace**\ (3)) followed by a single 3720 * optional '**-**' sign. 3721 * 3722 * Five least significant bits of *flags* encode base, other bits 3723 * are currently unused. 3724 * 3725 * Base must be either 8, 10, 16 or 0 to detect it automatically 3726 * similar to user space **strtol**\ (3). 3727 * Return 3728 * Number of characters consumed on success. Must be positive but 3729 * no more than *buf_len*. 3730 * 3731 * **-EINVAL** if no valid digits were found or unsupported base 3732 * was provided. 3733 * 3734 * **-ERANGE** if resulting value was out of range. 3735 * 3736 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3737 * Description 3738 * Convert the initial part of the string from buffer *buf* of 3739 * size *buf_len* to an unsigned long integer according to the 3740 * given base and save the result in *res*. 3741 * 3742 * The string may begin with an arbitrary amount of white space 3743 * (as determined by **isspace**\ (3)). 3744 * 3745 * Five least significant bits of *flags* encode base, other bits 3746 * are currently unused. 3747 * 3748 * Base must be either 8, 10, 16 or 0 to detect it automatically 3749 * similar to user space **strtoul**\ (3). 3750 * Return 3751 * Number of characters consumed on success. Must be positive but 3752 * no more than *buf_len*. 3753 * 3754 * **-EINVAL** if no valid digits were found or unsupported base 3755 * was provided. 3756 * 3757 * **-ERANGE** if resulting value was out of range. 3758 * 3759 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3760 * Description 3761 * Get a bpf-local-storage from a *sk*. 3762 * 3763 * Logically, it could be thought of getting the value from 3764 * a *map* with *sk* as the **key**. From this 3765 * perspective, the usage is not much different from 3766 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3767 * helper enforces the key must be a full socket and the map must 3768 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3769 * 3770 * Underneath, the value is stored locally at *sk* instead of 3771 * the *map*. The *map* is used as the bpf-local-storage 3772 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3773 * searched against all bpf-local-storages residing at *sk*. 3774 * 3775 * *sk* is a kernel **struct sock** pointer for LSM program. 3776 * *sk* is a **struct bpf_sock** pointer for other program types. 3777 * 3778 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3779 * used such that a new bpf-local-storage will be 3780 * created if one does not exist. *value* can be used 3781 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3782 * the initial value of a bpf-local-storage. If *value* is 3783 * **NULL**, the new bpf-local-storage will be zero initialized. 3784 * Return 3785 * A bpf-local-storage pointer is returned on success. 3786 * 3787 * **NULL** if not found or there was an error in adding 3788 * a new bpf-local-storage. 3789 * 3790 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3791 * Description 3792 * Delete a bpf-local-storage from a *sk*. 3793 * Return 3794 * 0 on success. 3795 * 3796 * **-ENOENT** if the bpf-local-storage cannot be found. 3797 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3798 * 3799 * long bpf_send_signal(u32 sig) 3800 * Description 3801 * Send signal *sig* to the process of the current task. 3802 * The signal may be delivered to any of this process's threads. 3803 * Return 3804 * 0 on success or successfully queued. 3805 * 3806 * **-EBUSY** if work queue under nmi is full. 3807 * 3808 * **-EINVAL** if *sig* is invalid. 3809 * 3810 * **-EPERM** if no permission to send the *sig*. 3811 * 3812 * **-EAGAIN** if bpf program can try again. 3813 * 3814 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3815 * Description 3816 * Try to issue a SYN cookie for the packet with corresponding 3817 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3818 * 3819 * *iph* points to the start of the IPv4 or IPv6 header, while 3820 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3821 * **sizeof**\ (**struct ipv6hdr**). 3822 * 3823 * *th* points to the start of the TCP header, while *th_len* 3824 * contains the length of the TCP header with options (at least 3825 * **sizeof**\ (**struct tcphdr**)). 3826 * Return 3827 * On success, lower 32 bits hold the generated SYN cookie in 3828 * followed by 16 bits which hold the MSS value for that cookie, 3829 * and the top 16 bits are unused. 3830 * 3831 * On failure, the returned value is one of the following: 3832 * 3833 * **-EINVAL** SYN cookie cannot be issued due to error 3834 * 3835 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3836 * 3837 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3838 * 3839 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3840 * 3841 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3842 * Description 3843 * Write raw *data* blob into a special BPF perf event held by 3844 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3845 * event must have the following attributes: **PERF_SAMPLE_RAW** 3846 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3847 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3848 * 3849 * The *flags* are used to indicate the index in *map* for which 3850 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3851 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3852 * to indicate that the index of the current CPU core should be 3853 * used. 3854 * 3855 * The value to write, of *size*, is passed through eBPF stack and 3856 * pointed by *data*. 3857 * 3858 * *ctx* is a pointer to in-kernel struct sk_buff. 3859 * 3860 * This helper is similar to **bpf_perf_event_output**\ () but 3861 * restricted to raw_tracepoint bpf programs. 3862 * Return 3863 * 0 on success, or a negative error in case of failure. 3864 * 3865 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3866 * Description 3867 * Safely attempt to read *size* bytes from user space address 3868 * *unsafe_ptr* and store the data in *dst*. 3869 * Return 3870 * 0 on success, or a negative error in case of failure. 3871 * 3872 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3873 * Description 3874 * Safely attempt to read *size* bytes from kernel space address 3875 * *unsafe_ptr* and store the data in *dst*. 3876 * Return 3877 * 0 on success, or a negative error in case of failure. 3878 * 3879 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3880 * Description 3881 * Copy a NUL terminated string from an unsafe user address 3882 * *unsafe_ptr* to *dst*. The *size* should include the 3883 * terminating NUL byte. In case the string length is smaller than 3884 * *size*, the target is not padded with further NUL bytes. If the 3885 * string length is larger than *size*, just *size*-1 bytes are 3886 * copied and the last byte is set to NUL. 3887 * 3888 * On success, returns the number of bytes that were written, 3889 * including the terminal NUL. This makes this helper useful in 3890 * tracing programs for reading strings, and more importantly to 3891 * get its length at runtime. See the following snippet: 3892 * 3893 * :: 3894 * 3895 * SEC("kprobe/sys_open") 3896 * void bpf_sys_open(struct pt_regs *ctx) 3897 * { 3898 * char buf[PATHLEN]; // PATHLEN is defined to 256 3899 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3900 * ctx->di); 3901 * 3902 * // Consume buf, for example push it to 3903 * // userspace via bpf_perf_event_output(); we 3904 * // can use res (the string length) as event 3905 * // size, after checking its boundaries. 3906 * } 3907 * 3908 * In comparison, using **bpf_probe_read_user**\ () helper here 3909 * instead to read the string would require to estimate the length 3910 * at compile time, and would often result in copying more memory 3911 * than necessary. 3912 * 3913 * Another useful use case is when parsing individual process 3914 * arguments or individual environment variables navigating 3915 * *current*\ **->mm->arg_start** and *current*\ 3916 * **->mm->env_start**: using this helper and the return value, 3917 * one can quickly iterate at the right offset of the memory area. 3918 * Return 3919 * On success, the strictly positive length of the output string, 3920 * including the trailing NUL character. On error, a negative 3921 * value. 3922 * 3923 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3924 * Description 3925 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3926 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3927 * Return 3928 * On success, the strictly positive length of the string, including 3929 * the trailing NUL character. On error, a negative value. 3930 * 3931 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3932 * Description 3933 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3934 * *rcv_nxt* is the ack_seq to be sent out. 3935 * Return 3936 * 0 on success, or a negative error in case of failure. 3937 * 3938 * long bpf_send_signal_thread(u32 sig) 3939 * Description 3940 * Send signal *sig* to the thread corresponding to the current task. 3941 * Return 3942 * 0 on success or successfully queued. 3943 * 3944 * **-EBUSY** if work queue under nmi is full. 3945 * 3946 * **-EINVAL** if *sig* is invalid. 3947 * 3948 * **-EPERM** if no permission to send the *sig*. 3949 * 3950 * **-EAGAIN** if bpf program can try again. 3951 * 3952 * u64 bpf_jiffies64(void) 3953 * Description 3954 * Obtain the 64bit jiffies 3955 * Return 3956 * The 64 bit jiffies 3957 * 3958 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3959 * Description 3960 * For an eBPF program attached to a perf event, retrieve the 3961 * branch records (**struct perf_branch_entry**) associated to *ctx* 3962 * and store it in the buffer pointed by *buf* up to size 3963 * *size* bytes. 3964 * Return 3965 * On success, number of bytes written to *buf*. On error, a 3966 * negative value. 3967 * 3968 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3969 * instead return the number of bytes required to store all the 3970 * branch entries. If this flag is set, *buf* may be NULL. 3971 * 3972 * **-EINVAL** if arguments invalid or **size** not a multiple 3973 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3974 * 3975 * **-ENOENT** if architecture does not support branch records. 3976 * 3977 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3978 * Description 3979 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3980 * *namespace* will be returned in *nsdata*. 3981 * Return 3982 * 0 on success, or one of the following in case of failure: 3983 * 3984 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3985 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3986 * 3987 * **-ENOENT** if pidns does not exists for the current task. 3988 * 3989 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3990 * Description 3991 * Write raw *data* blob into a special BPF perf event held by 3992 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3993 * event must have the following attributes: **PERF_SAMPLE_RAW** 3994 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3995 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3996 * 3997 * The *flags* are used to indicate the index in *map* for which 3998 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3999 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4000 * to indicate that the index of the current CPU core should be 4001 * used. 4002 * 4003 * The value to write, of *size*, is passed through eBPF stack and 4004 * pointed by *data*. 4005 * 4006 * *ctx* is a pointer to in-kernel struct xdp_buff. 4007 * 4008 * This helper is similar to **bpf_perf_eventoutput**\ () but 4009 * restricted to raw_tracepoint bpf programs. 4010 * Return 4011 * 0 on success, or a negative error in case of failure. 4012 * 4013 * u64 bpf_get_netns_cookie(void *ctx) 4014 * Description 4015 * Retrieve the cookie (generated by the kernel) of the network 4016 * namespace the input *ctx* is associated with. The network 4017 * namespace cookie remains stable for its lifetime and provides 4018 * a global identifier that can be assumed unique. If *ctx* is 4019 * NULL, then the helper returns the cookie for the initial 4020 * network namespace. The cookie itself is very similar to that 4021 * of **bpf_get_socket_cookie**\ () helper, but for network 4022 * namespaces instead of sockets. 4023 * Return 4024 * A 8-byte long opaque number. 4025 * 4026 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4027 * Description 4028 * Return id of cgroup v2 that is ancestor of the cgroup associated 4029 * with the current task at the *ancestor_level*. The root cgroup 4030 * is at *ancestor_level* zero and each step down the hierarchy 4031 * increments the level. If *ancestor_level* == level of cgroup 4032 * associated with the current task, then return value will be the 4033 * same as that of **bpf_get_current_cgroup_id**\ (). 4034 * 4035 * The helper is useful to implement policies based on cgroups 4036 * that are upper in hierarchy than immediate cgroup associated 4037 * with the current task. 4038 * 4039 * The format of returned id and helper limitations are same as in 4040 * **bpf_get_current_cgroup_id**\ (). 4041 * Return 4042 * The id is returned or 0 in case the id could not be retrieved. 4043 * 4044 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4045 * Description 4046 * Helper is overloaded depending on BPF program type. This 4047 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4048 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4049 * 4050 * Assign the *sk* to the *skb*. When combined with appropriate 4051 * routing configuration to receive the packet towards the socket, 4052 * will cause *skb* to be delivered to the specified socket. 4053 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4054 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4055 * interfere with successful delivery to the socket. 4056 * 4057 * This operation is only valid from TC ingress path. 4058 * 4059 * The *flags* argument must be zero. 4060 * Return 4061 * 0 on success, or a negative error in case of failure: 4062 * 4063 * **-EINVAL** if specified *flags* are not supported. 4064 * 4065 * **-ENOENT** if the socket is unavailable for assignment. 4066 * 4067 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4068 * 4069 * **-EOPNOTSUPP** if the operation is not supported, for example 4070 * a call from outside of TC ingress. 4071 * 4072 * **-ESOCKTNOSUPPORT** if the socket type is not supported 4073 * (reuseport). 4074 * 4075 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4076 * Description 4077 * Helper is overloaded depending on BPF program type. This 4078 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4079 * 4080 * Select the *sk* as a result of a socket lookup. 4081 * 4082 * For the operation to succeed passed socket must be compatible 4083 * with the packet description provided by the *ctx* object. 4084 * 4085 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4086 * be an exact match. While IP family (**AF_INET** or 4087 * **AF_INET6**) must be compatible, that is IPv6 sockets 4088 * that are not v6-only can be selected for IPv4 packets. 4089 * 4090 * Only TCP listeners and UDP unconnected sockets can be 4091 * selected. *sk* can also be NULL to reset any previous 4092 * selection. 4093 * 4094 * *flags* argument can combination of following values: 4095 * 4096 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4097 * socket selection, potentially done by a BPF program 4098 * that ran before us. 4099 * 4100 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4101 * load-balancing within reuseport group for the socket 4102 * being selected. 4103 * 4104 * On success *ctx->sk* will point to the selected socket. 4105 * 4106 * Return 4107 * 0 on success, or a negative errno in case of failure. 4108 * 4109 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4110 * not compatible with packet family (*ctx->family*). 4111 * 4112 * * **-EEXIST** if socket has been already selected, 4113 * potentially by another program, and 4114 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4115 * 4116 * * **-EINVAL** if unsupported flags were specified. 4117 * 4118 * * **-EPROTOTYPE** if socket L4 protocol 4119 * (*sk->protocol*) doesn't match packet protocol 4120 * (*ctx->protocol*). 4121 * 4122 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4123 * state (TCP listening or UDP unconnected). 4124 * 4125 * u64 bpf_ktime_get_boot_ns(void) 4126 * Description 4127 * Return the time elapsed since system boot, in nanoseconds. 4128 * Does include the time the system was suspended. 4129 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4130 * Return 4131 * Current *ktime*. 4132 * 4133 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4134 * Description 4135 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4136 * out the format string. 4137 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4138 * the format string itself. The *data* and *data_len* are format string 4139 * arguments. The *data* are a **u64** array and corresponding format string 4140 * values are stored in the array. For strings and pointers where pointees 4141 * are accessed, only the pointer values are stored in the *data* array. 4142 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4143 * 4144 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4145 * Reading kernel memory may fail due to either invalid address or 4146 * valid address but requiring a major memory fault. If reading kernel memory 4147 * fails, the string for **%s** will be an empty string, and the ip 4148 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4149 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4150 * Return 4151 * 0 on success, or a negative error in case of failure: 4152 * 4153 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4154 * by returning 1 from bpf program. 4155 * 4156 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4157 * 4158 * **-E2BIG** if *fmt* contains too many format specifiers. 4159 * 4160 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4161 * 4162 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4163 * Description 4164 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4165 * The *m* represents the seq_file. The *data* and *len* represent the 4166 * data to write in bytes. 4167 * Return 4168 * 0 on success, or a negative error in case of failure: 4169 * 4170 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4171 * 4172 * u64 bpf_sk_cgroup_id(void *sk) 4173 * Description 4174 * Return the cgroup v2 id of the socket *sk*. 4175 * 4176 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4177 * returned from **bpf_sk_lookup_xxx**\ (), 4178 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4179 * same as in **bpf_skb_cgroup_id**\ (). 4180 * 4181 * This helper is available only if the kernel was compiled with 4182 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4183 * Return 4184 * The id is returned or 0 in case the id could not be retrieved. 4185 * 4186 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4187 * Description 4188 * Return id of cgroup v2 that is ancestor of cgroup associated 4189 * with the *sk* at the *ancestor_level*. The root cgroup is at 4190 * *ancestor_level* zero and each step down the hierarchy 4191 * increments the level. If *ancestor_level* == level of cgroup 4192 * associated with *sk*, then return value will be same as that 4193 * of **bpf_sk_cgroup_id**\ (). 4194 * 4195 * The helper is useful to implement policies based on cgroups 4196 * that are upper in hierarchy than immediate cgroup associated 4197 * with *sk*. 4198 * 4199 * The format of returned id and helper limitations are same as in 4200 * **bpf_sk_cgroup_id**\ (). 4201 * Return 4202 * The id is returned or 0 in case the id could not be retrieved. 4203 * 4204 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4205 * Description 4206 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4207 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4208 * of new data availability is sent. 4209 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4210 * of new data availability is sent unconditionally. 4211 * If **0** is specified in *flags*, an adaptive notification 4212 * of new data availability is sent. 4213 * 4214 * An adaptive notification is a notification sent whenever the user-space 4215 * process has caught up and consumed all available payloads. In case the user-space 4216 * process is still processing a previous payload, then no notification is needed 4217 * as it will process the newly added payload automatically. 4218 * Return 4219 * 0 on success, or a negative error in case of failure. 4220 * 4221 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4222 * Description 4223 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4224 * *flags* must be 0. 4225 * Return 4226 * Valid pointer with *size* bytes of memory available; NULL, 4227 * otherwise. 4228 * 4229 * void bpf_ringbuf_submit(void *data, u64 flags) 4230 * Description 4231 * Submit reserved ring buffer sample, pointed to by *data*. 4232 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4233 * of new data availability is sent. 4234 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4235 * of new data availability is sent unconditionally. 4236 * If **0** is specified in *flags*, an adaptive notification 4237 * of new data availability is sent. 4238 * 4239 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4240 * Return 4241 * Nothing. Always succeeds. 4242 * 4243 * void bpf_ringbuf_discard(void *data, u64 flags) 4244 * Description 4245 * Discard reserved ring buffer sample, pointed to by *data*. 4246 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4247 * of new data availability is sent. 4248 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4249 * of new data availability is sent unconditionally. 4250 * If **0** is specified in *flags*, an adaptive notification 4251 * of new data availability is sent. 4252 * 4253 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4254 * Return 4255 * Nothing. Always succeeds. 4256 * 4257 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4258 * Description 4259 * Query various characteristics of provided ring buffer. What 4260 * exactly is queries is determined by *flags*: 4261 * 4262 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4263 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4264 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4265 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4266 * 4267 * Data returned is just a momentary snapshot of actual values 4268 * and could be inaccurate, so this facility should be used to 4269 * power heuristics and for reporting, not to make 100% correct 4270 * calculation. 4271 * Return 4272 * Requested value, or 0, if *flags* are not recognized. 4273 * 4274 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4275 * Description 4276 * Change the skbs checksum level by one layer up or down, or 4277 * reset it entirely to none in order to have the stack perform 4278 * checksum validation. The level is applicable to the following 4279 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4280 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4281 * through **bpf_skb_adjust_room**\ () helper with passing in 4282 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4283 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4284 * the UDP header is removed. Similarly, an encap of the latter 4285 * into the former could be accompanied by a helper call to 4286 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4287 * skb is still intended to be processed in higher layers of the 4288 * stack instead of just egressing at tc. 4289 * 4290 * There are three supported level settings at this time: 4291 * 4292 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4293 * with CHECKSUM_UNNECESSARY. 4294 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4295 * with CHECKSUM_UNNECESSARY. 4296 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4297 * sets CHECKSUM_NONE to force checksum validation by the stack. 4298 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4299 * skb->csum_level. 4300 * Return 4301 * 0 on success, or a negative error in case of failure. In the 4302 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4303 * is returned or the error code -EACCES in case the skb is not 4304 * subject to CHECKSUM_UNNECESSARY. 4305 * 4306 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4307 * Description 4308 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4309 * Return 4310 * *sk* if casting is valid, or **NULL** otherwise. 4311 * 4312 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4313 * Description 4314 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4315 * Return 4316 * *sk* if casting is valid, or **NULL** otherwise. 4317 * 4318 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4319 * Description 4320 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4321 * Return 4322 * *sk* if casting is valid, or **NULL** otherwise. 4323 * 4324 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4325 * Description 4326 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4327 * Return 4328 * *sk* if casting is valid, or **NULL** otherwise. 4329 * 4330 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4331 * Description 4332 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4333 * Return 4334 * *sk* if casting is valid, or **NULL** otherwise. 4335 * 4336 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4337 * Description 4338 * Return a user or a kernel stack in bpf program provided buffer. 4339 * To achieve this, the helper needs *task*, which is a valid 4340 * pointer to **struct task_struct**. To store the stacktrace, the 4341 * bpf program provides *buf* with a nonnegative *size*. 4342 * 4343 * The last argument, *flags*, holds the number of stack frames to 4344 * skip (from 0 to 255), masked with 4345 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4346 * the following flags: 4347 * 4348 * **BPF_F_USER_STACK** 4349 * Collect a user space stack instead of a kernel stack. 4350 * **BPF_F_USER_BUILD_ID** 4351 * Collect buildid+offset instead of ips for user stack, 4352 * only valid if **BPF_F_USER_STACK** is also specified. 4353 * 4354 * **bpf_get_task_stack**\ () can collect up to 4355 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4356 * to sufficient large buffer size. Note that 4357 * this limit can be controlled with the **sysctl** program, and 4358 * that it should be manually increased in order to profile long 4359 * user stacks (such as stacks for Java programs). To do so, use: 4360 * 4361 * :: 4362 * 4363 * # sysctl kernel.perf_event_max_stack=<new value> 4364 * Return 4365 * The non-negative copied *buf* length equal to or less than 4366 * *size* on success, or a negative error in case of failure. 4367 * 4368 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4369 * Description 4370 * Load header option. Support reading a particular TCP header 4371 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4372 * 4373 * If *flags* is 0, it will search the option from the 4374 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4375 * has details on what skb_data contains under different 4376 * *skops*\ **->op**. 4377 * 4378 * The first byte of the *searchby_res* specifies the 4379 * kind that it wants to search. 4380 * 4381 * If the searching kind is an experimental kind 4382 * (i.e. 253 or 254 according to RFC6994). It also 4383 * needs to specify the "magic" which is either 4384 * 2 bytes or 4 bytes. It then also needs to 4385 * specify the size of the magic by using 4386 * the 2nd byte which is "kind-length" of a TCP 4387 * header option and the "kind-length" also 4388 * includes the first 2 bytes "kind" and "kind-length" 4389 * itself as a normal TCP header option also does. 4390 * 4391 * For example, to search experimental kind 254 with 4392 * 2 byte magic 0xeB9F, the searchby_res should be 4393 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4394 * 4395 * To search for the standard window scale option (3), 4396 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4397 * Note, kind-length must be 0 for regular option. 4398 * 4399 * Searching for No-Op (0) and End-of-Option-List (1) are 4400 * not supported. 4401 * 4402 * *len* must be at least 2 bytes which is the minimal size 4403 * of a header option. 4404 * 4405 * Supported flags: 4406 * 4407 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4408 * saved_syn packet or the just-received syn packet. 4409 * 4410 * Return 4411 * > 0 when found, the header option is copied to *searchby_res*. 4412 * The return value is the total length copied. On failure, a 4413 * negative error code is returned: 4414 * 4415 * **-EINVAL** if a parameter is invalid. 4416 * 4417 * **-ENOMSG** if the option is not found. 4418 * 4419 * **-ENOENT** if no syn packet is available when 4420 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4421 * 4422 * **-ENOSPC** if there is not enough space. Only *len* number of 4423 * bytes are copied. 4424 * 4425 * **-EFAULT** on failure to parse the header options in the 4426 * packet. 4427 * 4428 * **-EPERM** if the helper cannot be used under the current 4429 * *skops*\ **->op**. 4430 * 4431 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4432 * Description 4433 * Store header option. The data will be copied 4434 * from buffer *from* with length *len* to the TCP header. 4435 * 4436 * The buffer *from* should have the whole option that 4437 * includes the kind, kind-length, and the actual 4438 * option data. The *len* must be at least kind-length 4439 * long. The kind-length does not have to be 4 byte 4440 * aligned. The kernel will take care of the padding 4441 * and setting the 4 bytes aligned value to th->doff. 4442 * 4443 * This helper will check for duplicated option 4444 * by searching the same option in the outgoing skb. 4445 * 4446 * This helper can only be called during 4447 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4448 * 4449 * Return 4450 * 0 on success, or negative error in case of failure: 4451 * 4452 * **-EINVAL** If param is invalid. 4453 * 4454 * **-ENOSPC** if there is not enough space in the header. 4455 * Nothing has been written 4456 * 4457 * **-EEXIST** if the option already exists. 4458 * 4459 * **-EFAULT** on failure to parse the existing header options. 4460 * 4461 * **-EPERM** if the helper cannot be used under the current 4462 * *skops*\ **->op**. 4463 * 4464 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4465 * Description 4466 * Reserve *len* bytes for the bpf header option. The 4467 * space will be used by **bpf_store_hdr_opt**\ () later in 4468 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4469 * 4470 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4471 * the total number of bytes will be reserved. 4472 * 4473 * This helper can only be called during 4474 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4475 * 4476 * Return 4477 * 0 on success, or negative error in case of failure: 4478 * 4479 * **-EINVAL** if a parameter is invalid. 4480 * 4481 * **-ENOSPC** if there is not enough space in the header. 4482 * 4483 * **-EPERM** if the helper cannot be used under the current 4484 * *skops*\ **->op**. 4485 * 4486 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4487 * Description 4488 * Get a bpf_local_storage from an *inode*. 4489 * 4490 * Logically, it could be thought of as getting the value from 4491 * a *map* with *inode* as the **key**. From this 4492 * perspective, the usage is not much different from 4493 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4494 * helper enforces the key must be an inode and the map must also 4495 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4496 * 4497 * Underneath, the value is stored locally at *inode* instead of 4498 * the *map*. The *map* is used as the bpf-local-storage 4499 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4500 * searched against all bpf_local_storage residing at *inode*. 4501 * 4502 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4503 * used such that a new bpf_local_storage will be 4504 * created if one does not exist. *value* can be used 4505 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4506 * the initial value of a bpf_local_storage. If *value* is 4507 * **NULL**, the new bpf_local_storage will be zero initialized. 4508 * Return 4509 * A bpf_local_storage pointer is returned on success. 4510 * 4511 * **NULL** if not found or there was an error in adding 4512 * a new bpf_local_storage. 4513 * 4514 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4515 * Description 4516 * Delete a bpf_local_storage from an *inode*. 4517 * Return 4518 * 0 on success. 4519 * 4520 * **-ENOENT** if the bpf_local_storage cannot be found. 4521 * 4522 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4523 * Description 4524 * Return full path for given **struct path** object, which 4525 * needs to be the kernel BTF *path* object. The path is 4526 * returned in the provided buffer *buf* of size *sz* and 4527 * is zero terminated. 4528 * 4529 * Return 4530 * On success, the strictly positive length of the string, 4531 * including the trailing NUL character. On error, a negative 4532 * value. 4533 * 4534 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4535 * Description 4536 * Read *size* bytes from user space address *user_ptr* and store 4537 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4538 * Return 4539 * 0 on success, or a negative error in case of failure. 4540 * 4541 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4542 * Description 4543 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4544 * using *ptr*->type_id. This value should specify the type 4545 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4546 * can be used to look up vmlinux BTF type ids. Traversing the 4547 * data structure using BTF, the type information and values are 4548 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4549 * the pointer data is carried out to avoid kernel crashes during 4550 * operation. Smaller types can use string space on the stack; 4551 * larger programs can use map data to store the string 4552 * representation. 4553 * 4554 * The string can be subsequently shared with userspace via 4555 * bpf_perf_event_output() or ring buffer interfaces. 4556 * bpf_trace_printk() is to be avoided as it places too small 4557 * a limit on string size to be useful. 4558 * 4559 * *flags* is a combination of 4560 * 4561 * **BTF_F_COMPACT** 4562 * no formatting around type information 4563 * **BTF_F_NONAME** 4564 * no struct/union member names/types 4565 * **BTF_F_PTR_RAW** 4566 * show raw (unobfuscated) pointer values; 4567 * equivalent to printk specifier %px. 4568 * **BTF_F_ZERO** 4569 * show zero-valued struct/union members; they 4570 * are not displayed by default 4571 * 4572 * Return 4573 * The number of bytes that were written (or would have been 4574 * written if output had to be truncated due to string size), 4575 * or a negative error in cases of failure. 4576 * 4577 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4578 * Description 4579 * Use BTF to write to seq_write a string representation of 4580 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4581 * *flags* are identical to those used for bpf_snprintf_btf. 4582 * Return 4583 * 0 on success or a negative error in case of failure. 4584 * 4585 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4586 * Description 4587 * See **bpf_get_cgroup_classid**\ () for the main description. 4588 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4589 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4590 * associated socket instead of the current process. 4591 * Return 4592 * The id is returned or 0 in case the id could not be retrieved. 4593 * 4594 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4595 * Description 4596 * Redirect the packet to another net device of index *ifindex* 4597 * and fill in L2 addresses from neighboring subsystem. This helper 4598 * is somewhat similar to **bpf_redirect**\ (), except that it 4599 * populates L2 addresses as well, meaning, internally, the helper 4600 * relies on the neighbor lookup for the L2 address of the nexthop. 4601 * 4602 * The helper will perform a FIB lookup based on the skb's 4603 * networking header to get the address of the next hop, unless 4604 * this is supplied by the caller in the *params* argument. The 4605 * *plen* argument indicates the len of *params* and should be set 4606 * to 0 if *params* is NULL. 4607 * 4608 * The *flags* argument is reserved and must be 0. The helper is 4609 * currently only supported for tc BPF program types, and enabled 4610 * for IPv4 and IPv6 protocols. 4611 * Return 4612 * The helper returns **TC_ACT_REDIRECT** on success or 4613 * **TC_ACT_SHOT** on error. 4614 * 4615 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4616 * Description 4617 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4618 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4619 * extern variable decorated with '__ksym'. For ksym, there is a 4620 * global var (either static or global) defined of the same name 4621 * in the kernel. The ksym is percpu if the global var is percpu. 4622 * The returned pointer points to the global percpu var on *cpu*. 4623 * 4624 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4625 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4626 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4627 * bpf_per_cpu_ptr() must check the returned value. 4628 * Return 4629 * A pointer pointing to the kernel percpu variable on *cpu*, or 4630 * NULL, if *cpu* is invalid. 4631 * 4632 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4633 * Description 4634 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4635 * pointer to the percpu kernel variable on this cpu. See the 4636 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4637 * 4638 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4639 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4640 * never return NULL. 4641 * Return 4642 * A pointer pointing to the kernel percpu variable on this cpu. 4643 * 4644 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4645 * Description 4646 * Redirect the packet to another net device of index *ifindex*. 4647 * This helper is somewhat similar to **bpf_redirect**\ (), except 4648 * that the redirection happens to the *ifindex*' peer device and 4649 * the netns switch takes place from ingress to ingress without 4650 * going through the CPU's backlog queue. 4651 * 4652 * The *flags* argument is reserved and must be 0. The helper is 4653 * currently only supported for tc BPF program types at the ingress 4654 * hook and for veth device types. The peer device must reside in a 4655 * different network namespace. 4656 * Return 4657 * The helper returns **TC_ACT_REDIRECT** on success or 4658 * **TC_ACT_SHOT** on error. 4659 * 4660 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4661 * Description 4662 * Get a bpf_local_storage from the *task*. 4663 * 4664 * Logically, it could be thought of as getting the value from 4665 * a *map* with *task* as the **key**. From this 4666 * perspective, the usage is not much different from 4667 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4668 * helper enforces the key must be a task_struct and the map must also 4669 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4670 * 4671 * Underneath, the value is stored locally at *task* instead of 4672 * the *map*. The *map* is used as the bpf-local-storage 4673 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4674 * searched against all bpf_local_storage residing at *task*. 4675 * 4676 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4677 * used such that a new bpf_local_storage will be 4678 * created if one does not exist. *value* can be used 4679 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4680 * the initial value of a bpf_local_storage. If *value* is 4681 * **NULL**, the new bpf_local_storage will be zero initialized. 4682 * Return 4683 * A bpf_local_storage pointer is returned on success. 4684 * 4685 * **NULL** if not found or there was an error in adding 4686 * a new bpf_local_storage. 4687 * 4688 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4689 * Description 4690 * Delete a bpf_local_storage from a *task*. 4691 * Return 4692 * 0 on success. 4693 * 4694 * **-ENOENT** if the bpf_local_storage cannot be found. 4695 * 4696 * struct task_struct *bpf_get_current_task_btf(void) 4697 * Description 4698 * Return a BTF pointer to the "current" task. 4699 * This pointer can also be used in helpers that accept an 4700 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4701 * Return 4702 * Pointer to the current task. 4703 * 4704 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4705 * Description 4706 * Set or clear certain options on *bprm*: 4707 * 4708 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4709 * which sets the **AT_SECURE** auxv for glibc. The bit 4710 * is cleared if the flag is not specified. 4711 * Return 4712 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4713 * 4714 * u64 bpf_ktime_get_coarse_ns(void) 4715 * Description 4716 * Return a coarse-grained version of the time elapsed since 4717 * system boot, in nanoseconds. Does not include time the system 4718 * was suspended. 4719 * 4720 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4721 * Return 4722 * Current *ktime*. 4723 * 4724 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4725 * Description 4726 * Returns the stored IMA hash of the *inode* (if it's available). 4727 * If the hash is larger than *size*, then only *size* 4728 * bytes will be copied to *dst* 4729 * Return 4730 * The **hash_algo** is returned on success, 4731 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4732 * invalid arguments are passed. 4733 * 4734 * struct socket *bpf_sock_from_file(struct file *file) 4735 * Description 4736 * If the given file represents a socket, returns the associated 4737 * socket. 4738 * Return 4739 * A pointer to a struct socket on success or NULL if the file is 4740 * not a socket. 4741 * 4742 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4743 * Description 4744 * Check packet size against exceeding MTU of net device (based 4745 * on *ifindex*). This helper will likely be used in combination 4746 * with helpers that adjust/change the packet size. 4747 * 4748 * The argument *len_diff* can be used for querying with a planned 4749 * size change. This allows to check MTU prior to changing packet 4750 * ctx. Providing a *len_diff* adjustment that is larger than the 4751 * actual packet size (resulting in negative packet size) will in 4752 * principle not exceed the MTU, which is why it is not considered 4753 * a failure. Other BPF helpers are needed for performing the 4754 * planned size change; therefore the responsibility for catching 4755 * a negative packet size belongs in those helpers. 4756 * 4757 * Specifying *ifindex* zero means the MTU check is performed 4758 * against the current net device. This is practical if this isn't 4759 * used prior to redirect. 4760 * 4761 * On input *mtu_len* must be a valid pointer, else verifier will 4762 * reject BPF program. If the value *mtu_len* is initialized to 4763 * zero then the ctx packet size is use. When value *mtu_len* is 4764 * provided as input this specify the L3 length that the MTU check 4765 * is done against. Remember XDP and TC length operate at L2, but 4766 * this value is L3 as this correlate to MTU and IP-header tot_len 4767 * values which are L3 (similar behavior as bpf_fib_lookup). 4768 * 4769 * The Linux kernel route table can configure MTUs on a more 4770 * specific per route level, which is not provided by this helper. 4771 * For route level MTU checks use the **bpf_fib_lookup**\ () 4772 * helper. 4773 * 4774 * *ctx* is either **struct xdp_md** for XDP programs or 4775 * **struct sk_buff** for tc cls_act programs. 4776 * 4777 * The *flags* argument can be a combination of one or more of the 4778 * following values: 4779 * 4780 * **BPF_MTU_CHK_SEGS** 4781 * This flag will only works for *ctx* **struct sk_buff**. 4782 * If packet context contains extra packet segment buffers 4783 * (often knows as GSO skb), then MTU check is harder to 4784 * check at this point, because in transmit path it is 4785 * possible for the skb packet to get re-segmented 4786 * (depending on net device features). This could still be 4787 * a MTU violation, so this flag enables performing MTU 4788 * check against segments, with a different violation 4789 * return code to tell it apart. Check cannot use len_diff. 4790 * 4791 * On return *mtu_len* pointer contains the MTU value of the net 4792 * device. Remember the net device configured MTU is the L3 size, 4793 * which is returned here and XDP and TC length operate at L2. 4794 * Helper take this into account for you, but remember when using 4795 * MTU value in your BPF-code. 4796 * 4797 * Return 4798 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4799 * 4800 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4801 * 4802 * MTU violations return positive values, but also populate MTU 4803 * value in *mtu_len* pointer, as this can be needed for 4804 * implementing PMTU handing: 4805 * 4806 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4807 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4808 * 4809 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4810 * Description 4811 * For each element in **map**, call **callback_fn** function with 4812 * **map**, **callback_ctx** and other map-specific parameters. 4813 * The **callback_fn** should be a static function and 4814 * the **callback_ctx** should be a pointer to the stack. 4815 * The **flags** is used to control certain aspects of the helper. 4816 * Currently, the **flags** must be 0. 4817 * 4818 * The following are a list of supported map types and their 4819 * respective expected callback signatures: 4820 * 4821 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4822 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4823 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4824 * 4825 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4826 * 4827 * For per_cpu maps, the map_value is the value on the cpu where the 4828 * bpf_prog is running. 4829 * 4830 * If **callback_fn** return 0, the helper will continue to the next 4831 * element. If return value is 1, the helper will skip the rest of 4832 * elements and return. Other return values are not used now. 4833 * 4834 * Return 4835 * The number of traversed map elements for success, **-EINVAL** for 4836 * invalid **flags**. 4837 * 4838 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4839 * Description 4840 * Outputs a string into the **str** buffer of size **str_size** 4841 * based on a format string stored in a read-only map pointed by 4842 * **fmt**. 4843 * 4844 * Each format specifier in **fmt** corresponds to one u64 element 4845 * in the **data** array. For strings and pointers where pointees 4846 * are accessed, only the pointer values are stored in the *data* 4847 * array. The *data_len* is the size of *data* in bytes - must be 4848 * a multiple of 8. 4849 * 4850 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4851 * memory. Reading kernel memory may fail due to either invalid 4852 * address or valid address but requiring a major memory fault. If 4853 * reading kernel memory fails, the string for **%s** will be an 4854 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4855 * Not returning error to bpf program is consistent with what 4856 * **bpf_trace_printk**\ () does for now. 4857 * 4858 * Return 4859 * The strictly positive length of the formatted string, including 4860 * the trailing zero character. If the return value is greater than 4861 * **str_size**, **str** contains a truncated string, guaranteed to 4862 * be zero-terminated except when **str_size** is 0. 4863 * 4864 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4865 * 4866 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4867 * Description 4868 * Execute bpf syscall with given arguments. 4869 * Return 4870 * A syscall result. 4871 * 4872 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4873 * Description 4874 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4875 * Return 4876 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4877 * 4878 * long bpf_sys_close(u32 fd) 4879 * Description 4880 * Execute close syscall for given FD. 4881 * Return 4882 * A syscall result. 4883 * 4884 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 4885 * Description 4886 * Initialize the timer. 4887 * First 4 bits of *flags* specify clockid. 4888 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 4889 * All other bits of *flags* are reserved. 4890 * The verifier will reject the program if *timer* is not from 4891 * the same *map*. 4892 * Return 4893 * 0 on success. 4894 * **-EBUSY** if *timer* is already initialized. 4895 * **-EINVAL** if invalid *flags* are passed. 4896 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4897 * The user space should either hold a file descriptor to a map with timers 4898 * or pin such map in bpffs. When map is unpinned or file descriptor is 4899 * closed all timers in the map will be cancelled and freed. 4900 * 4901 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 4902 * Description 4903 * Configure the timer to call *callback_fn* static function. 4904 * Return 4905 * 0 on success. 4906 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4907 * **-EPERM** if *timer* is in a map that doesn't have any user references. 4908 * The user space should either hold a file descriptor to a map with timers 4909 * or pin such map in bpffs. When map is unpinned or file descriptor is 4910 * closed all timers in the map will be cancelled and freed. 4911 * 4912 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 4913 * Description 4914 * Set timer expiration N nanoseconds from the current time. The 4915 * configured callback will be invoked in soft irq context on some cpu 4916 * and will not repeat unless another bpf_timer_start() is made. 4917 * In such case the next invocation can migrate to a different cpu. 4918 * Since struct bpf_timer is a field inside map element the map 4919 * owns the timer. The bpf_timer_set_callback() will increment refcnt 4920 * of BPF program to make sure that callback_fn code stays valid. 4921 * When user space reference to a map reaches zero all timers 4922 * in a map are cancelled and corresponding program's refcnts are 4923 * decremented. This is done to make sure that Ctrl-C of a user 4924 * process doesn't leave any timers running. If map is pinned in 4925 * bpffs the callback_fn can re-arm itself indefinitely. 4926 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 4927 * cancel and free the timer in the given map element. 4928 * The map can contain timers that invoke callback_fn-s from different 4929 * programs. The same callback_fn can serve different timers from 4930 * different maps if key/value layout matches across maps. 4931 * Every bpf_timer_set_callback() can have different callback_fn. 4932 * 4933 * Return 4934 * 0 on success. 4935 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 4936 * or invalid *flags* are passed. 4937 * 4938 * long bpf_timer_cancel(struct bpf_timer *timer) 4939 * Description 4940 * Cancel the timer and wait for callback_fn to finish if it was running. 4941 * Return 4942 * 0 if the timer was not active. 4943 * 1 if the timer was active. 4944 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 4945 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 4946 * own timer which would have led to a deadlock otherwise. 4947 * 4948 * u64 bpf_get_func_ip(void *ctx) 4949 * Description 4950 * Get address of the traced function (for tracing and kprobe programs). 4951 * Return 4952 * Address of the traced function. 4953 * 4954 * u64 bpf_get_attach_cookie(void *ctx) 4955 * Description 4956 * Get bpf_cookie value provided (optionally) during the program 4957 * attachment. It might be different for each individual 4958 * attachment, even if BPF program itself is the same. 4959 * Expects BPF program context *ctx* as a first argument. 4960 * 4961 * Supported for the following program types: 4962 * - kprobe/uprobe; 4963 * - tracepoint; 4964 * - perf_event. 4965 * Return 4966 * Value specified by user at BPF link creation/attachment time 4967 * or 0, if it was not specified. 4968 * 4969 * long bpf_task_pt_regs(struct task_struct *task) 4970 * Description 4971 * Get the struct pt_regs associated with **task**. 4972 * Return 4973 * A pointer to struct pt_regs. 4974 * 4975 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 4976 * Description 4977 * Get branch trace from hardware engines like Intel LBR. The 4978 * hardware engine is stopped shortly after the helper is 4979 * called. Therefore, the user need to filter branch entries 4980 * based on the actual use case. To capture branch trace 4981 * before the trigger point of the BPF program, the helper 4982 * should be called at the beginning of the BPF program. 4983 * 4984 * The data is stored as struct perf_branch_entry into output 4985 * buffer *entries*. *size* is the size of *entries* in bytes. 4986 * *flags* is reserved for now and must be zero. 4987 * 4988 * Return 4989 * On success, number of bytes written to *buf*. On error, a 4990 * negative value. 4991 * 4992 * **-EINVAL** if *flags* is not zero. 4993 * 4994 * **-ENOENT** if architecture does not support branch records. 4995 * 4996 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4997 * Description 4998 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 4999 * to format and can handle more format args as a result. 5000 * 5001 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5002 * Return 5003 * The number of bytes written to the buffer, or a negative error 5004 * in case of failure. 5005 * 5006 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5007 * Description 5008 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5009 * Return 5010 * *sk* if casting is valid, or **NULL** otherwise. 5011 * 5012 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5013 * Description 5014 * Get the address of a kernel symbol, returned in *res*. *res* is 5015 * set to 0 if the symbol is not found. 5016 * Return 5017 * On success, zero. On error, a negative value. 5018 * 5019 * **-EINVAL** if *flags* is not zero. 5020 * 5021 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5022 * 5023 * **-ENOENT** if symbol is not found. 5024 * 5025 * **-EPERM** if caller does not have permission to obtain kernel address. 5026 * 5027 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5028 * Description 5029 * Find vma of *task* that contains *addr*, call *callback_fn* 5030 * function with *task*, *vma*, and *callback_ctx*. 5031 * The *callback_fn* should be a static function and 5032 * the *callback_ctx* should be a pointer to the stack. 5033 * The *flags* is used to control certain aspects of the helper. 5034 * Currently, the *flags* must be 0. 5035 * 5036 * The expected callback signature is 5037 * 5038 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5039 * 5040 * Return 5041 * 0 on success. 5042 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5043 * **-EBUSY** if failed to try lock mmap_lock. 5044 * **-EINVAL** for invalid **flags**. 5045 * 5046 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5047 * Description 5048 * For **nr_loops**, call **callback_fn** function 5049 * with **callback_ctx** as the context parameter. 5050 * The **callback_fn** should be a static function and 5051 * the **callback_ctx** should be a pointer to the stack. 5052 * The **flags** is used to control certain aspects of the helper. 5053 * Currently, the **flags** must be 0. Currently, nr_loops is 5054 * limited to 1 << 23 (~8 million) loops. 5055 * 5056 * long (\*callback_fn)(u32 index, void \*ctx); 5057 * 5058 * where **index** is the current index in the loop. The index 5059 * is zero-indexed. 5060 * 5061 * If **callback_fn** returns 0, the helper will continue to the next 5062 * loop. If return value is 1, the helper will skip the rest of 5063 * the loops and return. Other return values are not used now, 5064 * and will be rejected by the verifier. 5065 * 5066 * Return 5067 * The number of loops performed, **-EINVAL** for invalid **flags**, 5068 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5069 * 5070 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5071 * Description 5072 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5073 * to be null-terminated and **s1_sz** is the maximum storage 5074 * size of **s1**. **s2** must be a read-only string. 5075 * Return 5076 * An integer less than, equal to, or greater than zero 5077 * if the first **s1_sz** bytes of **s1** is found to be 5078 * less than, to match, or be greater than **s2**. 5079 * 5080 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5081 * Description 5082 * Get **n**-th argument (zero based) of the traced function (for tracing programs) 5083 * returned in **value**. 5084 * 5085 * Return 5086 * 0 on success. 5087 * **-EINVAL** if n >= arguments count of traced function. 5088 * 5089 * long bpf_get_func_ret(void *ctx, u64 *value) 5090 * Description 5091 * Get return value of the traced function (for tracing programs) 5092 * in **value**. 5093 * 5094 * Return 5095 * 0 on success. 5096 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5097 * 5098 * long bpf_get_func_arg_cnt(void *ctx) 5099 * Description 5100 * Get number of arguments of the traced function (for tracing programs). 5101 * 5102 * Return 5103 * The number of arguments of the traced function. 5104 * 5105 * int bpf_get_retval(void) 5106 * Description 5107 * Get the BPF program's return value that will be returned to the upper layers. 5108 * 5109 * This helper is currently supported by cgroup programs and only by the hooks 5110 * where BPF program's return value is returned to the userspace via errno. 5111 * Return 5112 * The BPF program's return value. 5113 * 5114 * int bpf_set_retval(int retval) 5115 * Description 5116 * Set the BPF program's return value that will be returned to the upper layers. 5117 * 5118 * This helper is currently supported by cgroup programs and only by the hooks 5119 * where BPF program's return value is returned to the userspace via errno. 5120 * 5121 * Note that there is the following corner case where the program exports an error 5122 * via bpf_set_retval but signals success via 'return 1': 5123 * 5124 * bpf_set_retval(-EPERM); 5125 * return 1; 5126 * 5127 * In this case, the BPF program's return value will use helper's -EPERM. This 5128 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5129 * 5130 * Return 5131 * 0 on success, or a negative error in case of failure. 5132 * 5133 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5134 * Description 5135 * Get the total size of a given xdp buff (linear and paged area) 5136 * Return 5137 * The total size of a given xdp buffer. 5138 * 5139 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5140 * Description 5141 * This helper is provided as an easy way to load data from a 5142 * xdp buffer. It can be used to load *len* bytes from *offset* from 5143 * the frame associated to *xdp_md*, into the buffer pointed by 5144 * *buf*. 5145 * Return 5146 * 0 on success, or a negative error in case of failure. 5147 * 5148 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5149 * Description 5150 * Store *len* bytes from buffer *buf* into the frame 5151 * associated to *xdp_md*, at *offset*. 5152 * Return 5153 * 0 on success, or a negative error in case of failure. 5154 * 5155 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5156 * Description 5157 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5158 * address space, and stores the data in *dst*. *flags* is not 5159 * used yet and is provided for future extensibility. This helper 5160 * can only be used by sleepable programs. 5161 * Return 5162 * 0 on success, or a negative error in case of failure. On error 5163 * *dst* buffer is zeroed out. 5164 * 5165 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5166 * Description 5167 * Change the __sk_buff->tstamp_type to *tstamp_type* 5168 * and set *tstamp* to the __sk_buff->tstamp together. 5169 * 5170 * If there is no need to change the __sk_buff->tstamp_type, 5171 * the tstamp value can be directly written to __sk_buff->tstamp 5172 * instead. 5173 * 5174 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5175 * will be kept during bpf_redirect_*(). A non zero 5176 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5177 * *tstamp_type*. 5178 * 5179 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5180 * with a zero *tstamp*. 5181 * 5182 * Only IPv4 and IPv6 skb->protocol are supported. 5183 * 5184 * This function is most useful when it needs to set a 5185 * mono delivery time to __sk_buff->tstamp and then 5186 * bpf_redirect_*() to the egress of an iface. For example, 5187 * changing the (rcv) timestamp in __sk_buff->tstamp at 5188 * ingress to a mono delivery time and then bpf_redirect_*() 5189 * to sch_fq@phy-dev. 5190 * Return 5191 * 0 on success. 5192 * **-EINVAL** for invalid input 5193 * **-EOPNOTSUPP** for unsupported protocol 5194 * 5195 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5196 * Description 5197 * Returns a calculated IMA hash of the *file*. 5198 * If the hash is larger than *size*, then only *size* 5199 * bytes will be copied to *dst* 5200 * Return 5201 * The **hash_algo** is returned on success, 5202 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5203 * invalid arguments are passed. 5204 * 5205 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5206 * Description 5207 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5208 * old value. *ptr* can be NULL, otherwise it must be a referenced 5209 * pointer which will be released when this helper is called. 5210 * Return 5211 * The old value of kptr (which can be NULL). The returned pointer 5212 * if not NULL, is a reference which must be released using its 5213 * corresponding release function, or moved into a BPF map before 5214 * program exit. 5215 * 5216 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5217 * Description 5218 * Perform a lookup in *percpu map* for an entry associated to 5219 * *key* on *cpu*. 5220 * Return 5221 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5222 * was found or *cpu* is invalid. 5223 * 5224 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5225 * Description 5226 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5227 * Return 5228 * *sk* if casting is valid, or **NULL** otherwise. 5229 * 5230 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5231 * Description 5232 * Get a dynptr to local memory *data*. 5233 * 5234 * *data* must be a ptr to a map value. 5235 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5236 * *flags* is currently unused. 5237 * Return 5238 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5239 * -EINVAL if flags is not 0. 5240 * 5241 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5242 * Description 5243 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5244 * through the dynptr interface. *flags* must be 0. 5245 * 5246 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5247 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5248 * reservation fails. This is enforced by the verifier. 5249 * Return 5250 * 0 on success, or a negative error in case of failure. 5251 * 5252 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5253 * Description 5254 * Submit reserved ring buffer sample, pointed to by *data*, 5255 * through the dynptr interface. This is a no-op if the dynptr is 5256 * invalid/null. 5257 * 5258 * For more information on *flags*, please see 5259 * 'bpf_ringbuf_submit'. 5260 * Return 5261 * Nothing. Always succeeds. 5262 * 5263 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5264 * Description 5265 * Discard reserved ring buffer sample through the dynptr 5266 * interface. This is a no-op if the dynptr is invalid/null. 5267 * 5268 * For more information on *flags*, please see 5269 * 'bpf_ringbuf_discard'. 5270 * Return 5271 * Nothing. Always succeeds. 5272 * 5273 * long bpf_dynptr_read(void *dst, u32 len, struct bpf_dynptr *src, u32 offset, u64 flags) 5274 * Description 5275 * Read *len* bytes from *src* into *dst*, starting from *offset* 5276 * into *src*. 5277 * *flags* is currently unused. 5278 * Return 5279 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5280 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5281 * *flags* is not 0. 5282 * 5283 * long bpf_dynptr_write(struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5284 * Description 5285 * Write *len* bytes from *src* into *dst*, starting from *offset* 5286 * into *dst*. 5287 * *flags* is currently unused. 5288 * Return 5289 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5290 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5291 * is a read-only dynptr or if *flags* is not 0. 5292 * 5293 * void *bpf_dynptr_data(struct bpf_dynptr *ptr, u32 offset, u32 len) 5294 * Description 5295 * Get a pointer to the underlying dynptr data. 5296 * 5297 * *len* must be a statically known value. The returned data slice 5298 * is invalidated whenever the dynptr is invalidated. 5299 * Return 5300 * Pointer to the underlying dynptr data, NULL if the dynptr is 5301 * read-only, if the dynptr is invalid, or if the offset and length 5302 * is out of bounds. 5303 * 5304 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5305 * Description 5306 * Try to issue a SYN cookie for the packet with corresponding 5307 * IPv4/TCP headers, *iph* and *th*, without depending on a 5308 * listening socket. 5309 * 5310 * *iph* points to the IPv4 header. 5311 * 5312 * *th* points to the start of the TCP header, while *th_len* 5313 * contains the length of the TCP header (at least 5314 * **sizeof**\ (**struct tcphdr**)). 5315 * Return 5316 * On success, lower 32 bits hold the generated SYN cookie in 5317 * followed by 16 bits which hold the MSS value for that cookie, 5318 * and the top 16 bits are unused. 5319 * 5320 * On failure, the returned value is one of the following: 5321 * 5322 * **-EINVAL** if *th_len* is invalid. 5323 * 5324 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5325 * Description 5326 * Try to issue a SYN cookie for the packet with corresponding 5327 * IPv6/TCP headers, *iph* and *th*, without depending on a 5328 * listening socket. 5329 * 5330 * *iph* points to the IPv6 header. 5331 * 5332 * *th* points to the start of the TCP header, while *th_len* 5333 * contains the length of the TCP header (at least 5334 * **sizeof**\ (**struct tcphdr**)). 5335 * Return 5336 * On success, lower 32 bits hold the generated SYN cookie in 5337 * followed by 16 bits which hold the MSS value for that cookie, 5338 * and the top 16 bits are unused. 5339 * 5340 * On failure, the returned value is one of the following: 5341 * 5342 * **-EINVAL** if *th_len* is invalid. 5343 * 5344 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5345 * 5346 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5347 * Description 5348 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5349 * without depending on a listening socket. 5350 * 5351 * *iph* points to the IPv4 header. 5352 * 5353 * *th* points to the TCP header. 5354 * Return 5355 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5356 * 5357 * On failure, the returned value is one of the following: 5358 * 5359 * **-EACCES** if the SYN cookie is not valid. 5360 * 5361 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5362 * Description 5363 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5364 * without depending on a listening socket. 5365 * 5366 * *iph* points to the IPv6 header. 5367 * 5368 * *th* points to the TCP header. 5369 * Return 5370 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5371 * 5372 * On failure, the returned value is one of the following: 5373 * 5374 * **-EACCES** if the SYN cookie is not valid. 5375 * 5376 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5377 * 5378 * u64 bpf_ktime_get_tai_ns(void) 5379 * Description 5380 * A nonsettable system-wide clock derived from wall-clock time but 5381 * ignoring leap seconds. This clock does not experience 5382 * discontinuities and backwards jumps caused by NTP inserting leap 5383 * seconds as CLOCK_REALTIME does. 5384 * 5385 * See: **clock_gettime**\ (**CLOCK_TAI**) 5386 * Return 5387 * Current *ktime*. 5388 * 5389 */ 5390 #define __BPF_FUNC_MAPPER(FN) \ 5391 FN(unspec), \ 5392 FN(map_lookup_elem), \ 5393 FN(map_update_elem), \ 5394 FN(map_delete_elem), \ 5395 FN(probe_read), \ 5396 FN(ktime_get_ns), \ 5397 FN(trace_printk), \ 5398 FN(get_prandom_u32), \ 5399 FN(get_smp_processor_id), \ 5400 FN(skb_store_bytes), \ 5401 FN(l3_csum_replace), \ 5402 FN(l4_csum_replace), \ 5403 FN(tail_call), \ 5404 FN(clone_redirect), \ 5405 FN(get_current_pid_tgid), \ 5406 FN(get_current_uid_gid), \ 5407 FN(get_current_comm), \ 5408 FN(get_cgroup_classid), \ 5409 FN(skb_vlan_push), \ 5410 FN(skb_vlan_pop), \ 5411 FN(skb_get_tunnel_key), \ 5412 FN(skb_set_tunnel_key), \ 5413 FN(perf_event_read), \ 5414 FN(redirect), \ 5415 FN(get_route_realm), \ 5416 FN(perf_event_output), \ 5417 FN(skb_load_bytes), \ 5418 FN(get_stackid), \ 5419 FN(csum_diff), \ 5420 FN(skb_get_tunnel_opt), \ 5421 FN(skb_set_tunnel_opt), \ 5422 FN(skb_change_proto), \ 5423 FN(skb_change_type), \ 5424 FN(skb_under_cgroup), \ 5425 FN(get_hash_recalc), \ 5426 FN(get_current_task), \ 5427 FN(probe_write_user), \ 5428 FN(current_task_under_cgroup), \ 5429 FN(skb_change_tail), \ 5430 FN(skb_pull_data), \ 5431 FN(csum_update), \ 5432 FN(set_hash_invalid), \ 5433 FN(get_numa_node_id), \ 5434 FN(skb_change_head), \ 5435 FN(xdp_adjust_head), \ 5436 FN(probe_read_str), \ 5437 FN(get_socket_cookie), \ 5438 FN(get_socket_uid), \ 5439 FN(set_hash), \ 5440 FN(setsockopt), \ 5441 FN(skb_adjust_room), \ 5442 FN(redirect_map), \ 5443 FN(sk_redirect_map), \ 5444 FN(sock_map_update), \ 5445 FN(xdp_adjust_meta), \ 5446 FN(perf_event_read_value), \ 5447 FN(perf_prog_read_value), \ 5448 FN(getsockopt), \ 5449 FN(override_return), \ 5450 FN(sock_ops_cb_flags_set), \ 5451 FN(msg_redirect_map), \ 5452 FN(msg_apply_bytes), \ 5453 FN(msg_cork_bytes), \ 5454 FN(msg_pull_data), \ 5455 FN(bind), \ 5456 FN(xdp_adjust_tail), \ 5457 FN(skb_get_xfrm_state), \ 5458 FN(get_stack), \ 5459 FN(skb_load_bytes_relative), \ 5460 FN(fib_lookup), \ 5461 FN(sock_hash_update), \ 5462 FN(msg_redirect_hash), \ 5463 FN(sk_redirect_hash), \ 5464 FN(lwt_push_encap), \ 5465 FN(lwt_seg6_store_bytes), \ 5466 FN(lwt_seg6_adjust_srh), \ 5467 FN(lwt_seg6_action), \ 5468 FN(rc_repeat), \ 5469 FN(rc_keydown), \ 5470 FN(skb_cgroup_id), \ 5471 FN(get_current_cgroup_id), \ 5472 FN(get_local_storage), \ 5473 FN(sk_select_reuseport), \ 5474 FN(skb_ancestor_cgroup_id), \ 5475 FN(sk_lookup_tcp), \ 5476 FN(sk_lookup_udp), \ 5477 FN(sk_release), \ 5478 FN(map_push_elem), \ 5479 FN(map_pop_elem), \ 5480 FN(map_peek_elem), \ 5481 FN(msg_push_data), \ 5482 FN(msg_pop_data), \ 5483 FN(rc_pointer_rel), \ 5484 FN(spin_lock), \ 5485 FN(spin_unlock), \ 5486 FN(sk_fullsock), \ 5487 FN(tcp_sock), \ 5488 FN(skb_ecn_set_ce), \ 5489 FN(get_listener_sock), \ 5490 FN(skc_lookup_tcp), \ 5491 FN(tcp_check_syncookie), \ 5492 FN(sysctl_get_name), \ 5493 FN(sysctl_get_current_value), \ 5494 FN(sysctl_get_new_value), \ 5495 FN(sysctl_set_new_value), \ 5496 FN(strtol), \ 5497 FN(strtoul), \ 5498 FN(sk_storage_get), \ 5499 FN(sk_storage_delete), \ 5500 FN(send_signal), \ 5501 FN(tcp_gen_syncookie), \ 5502 FN(skb_output), \ 5503 FN(probe_read_user), \ 5504 FN(probe_read_kernel), \ 5505 FN(probe_read_user_str), \ 5506 FN(probe_read_kernel_str), \ 5507 FN(tcp_send_ack), \ 5508 FN(send_signal_thread), \ 5509 FN(jiffies64), \ 5510 FN(read_branch_records), \ 5511 FN(get_ns_current_pid_tgid), \ 5512 FN(xdp_output), \ 5513 FN(get_netns_cookie), \ 5514 FN(get_current_ancestor_cgroup_id), \ 5515 FN(sk_assign), \ 5516 FN(ktime_get_boot_ns), \ 5517 FN(seq_printf), \ 5518 FN(seq_write), \ 5519 FN(sk_cgroup_id), \ 5520 FN(sk_ancestor_cgroup_id), \ 5521 FN(ringbuf_output), \ 5522 FN(ringbuf_reserve), \ 5523 FN(ringbuf_submit), \ 5524 FN(ringbuf_discard), \ 5525 FN(ringbuf_query), \ 5526 FN(csum_level), \ 5527 FN(skc_to_tcp6_sock), \ 5528 FN(skc_to_tcp_sock), \ 5529 FN(skc_to_tcp_timewait_sock), \ 5530 FN(skc_to_tcp_request_sock), \ 5531 FN(skc_to_udp6_sock), \ 5532 FN(get_task_stack), \ 5533 FN(load_hdr_opt), \ 5534 FN(store_hdr_opt), \ 5535 FN(reserve_hdr_opt), \ 5536 FN(inode_storage_get), \ 5537 FN(inode_storage_delete), \ 5538 FN(d_path), \ 5539 FN(copy_from_user), \ 5540 FN(snprintf_btf), \ 5541 FN(seq_printf_btf), \ 5542 FN(skb_cgroup_classid), \ 5543 FN(redirect_neigh), \ 5544 FN(per_cpu_ptr), \ 5545 FN(this_cpu_ptr), \ 5546 FN(redirect_peer), \ 5547 FN(task_storage_get), \ 5548 FN(task_storage_delete), \ 5549 FN(get_current_task_btf), \ 5550 FN(bprm_opts_set), \ 5551 FN(ktime_get_coarse_ns), \ 5552 FN(ima_inode_hash), \ 5553 FN(sock_from_file), \ 5554 FN(check_mtu), \ 5555 FN(for_each_map_elem), \ 5556 FN(snprintf), \ 5557 FN(sys_bpf), \ 5558 FN(btf_find_by_name_kind), \ 5559 FN(sys_close), \ 5560 FN(timer_init), \ 5561 FN(timer_set_callback), \ 5562 FN(timer_start), \ 5563 FN(timer_cancel), \ 5564 FN(get_func_ip), \ 5565 FN(get_attach_cookie), \ 5566 FN(task_pt_regs), \ 5567 FN(get_branch_snapshot), \ 5568 FN(trace_vprintk), \ 5569 FN(skc_to_unix_sock), \ 5570 FN(kallsyms_lookup_name), \ 5571 FN(find_vma), \ 5572 FN(loop), \ 5573 FN(strncmp), \ 5574 FN(get_func_arg), \ 5575 FN(get_func_ret), \ 5576 FN(get_func_arg_cnt), \ 5577 FN(get_retval), \ 5578 FN(set_retval), \ 5579 FN(xdp_get_buff_len), \ 5580 FN(xdp_load_bytes), \ 5581 FN(xdp_store_bytes), \ 5582 FN(copy_from_user_task), \ 5583 FN(skb_set_tstamp), \ 5584 FN(ima_file_hash), \ 5585 FN(kptr_xchg), \ 5586 FN(map_lookup_percpu_elem), \ 5587 FN(skc_to_mptcp_sock), \ 5588 FN(dynptr_from_mem), \ 5589 FN(ringbuf_reserve_dynptr), \ 5590 FN(ringbuf_submit_dynptr), \ 5591 FN(ringbuf_discard_dynptr), \ 5592 FN(dynptr_read), \ 5593 FN(dynptr_write), \ 5594 FN(dynptr_data), \ 5595 FN(tcp_raw_gen_syncookie_ipv4), \ 5596 FN(tcp_raw_gen_syncookie_ipv6), \ 5597 FN(tcp_raw_check_syncookie_ipv4), \ 5598 FN(tcp_raw_check_syncookie_ipv6), \ 5599 FN(ktime_get_tai_ns), \ 5600 /* */ 5601 5602 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5603 * function eBPF program intends to call 5604 */ 5605 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 5606 enum bpf_func_id { 5607 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5608 __BPF_FUNC_MAX_ID, 5609 }; 5610 #undef __BPF_ENUM_FN 5611 5612 /* All flags used by eBPF helper functions, placed here. */ 5613 5614 /* BPF_FUNC_skb_store_bytes flags. */ 5615 enum { 5616 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5617 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5618 }; 5619 5620 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5621 * First 4 bits are for passing the header field size. 5622 */ 5623 enum { 5624 BPF_F_HDR_FIELD_MASK = 0xfULL, 5625 }; 5626 5627 /* BPF_FUNC_l4_csum_replace flags. */ 5628 enum { 5629 BPF_F_PSEUDO_HDR = (1ULL << 4), 5630 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5631 BPF_F_MARK_ENFORCE = (1ULL << 6), 5632 }; 5633 5634 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 5635 enum { 5636 BPF_F_INGRESS = (1ULL << 0), 5637 }; 5638 5639 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 5640 enum { 5641 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 5642 }; 5643 5644 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 5645 enum { 5646 BPF_F_SKIP_FIELD_MASK = 0xffULL, 5647 BPF_F_USER_STACK = (1ULL << 8), 5648 /* flags used by BPF_FUNC_get_stackid only. */ 5649 BPF_F_FAST_STACK_CMP = (1ULL << 9), 5650 BPF_F_REUSE_STACKID = (1ULL << 10), 5651 /* flags used by BPF_FUNC_get_stack only. */ 5652 BPF_F_USER_BUILD_ID = (1ULL << 11), 5653 }; 5654 5655 /* BPF_FUNC_skb_set_tunnel_key flags. */ 5656 enum { 5657 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 5658 BPF_F_DONT_FRAGMENT = (1ULL << 2), 5659 BPF_F_SEQ_NUMBER = (1ULL << 3), 5660 }; 5661 5662 /* BPF_FUNC_skb_get_tunnel_key flags. */ 5663 enum { 5664 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 5665 }; 5666 5667 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 5668 * BPF_FUNC_perf_event_read_value flags. 5669 */ 5670 enum { 5671 BPF_F_INDEX_MASK = 0xffffffffULL, 5672 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5673 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5674 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5675 }; 5676 5677 /* Current network namespace */ 5678 enum { 5679 BPF_F_CURRENT_NETNS = (-1L), 5680 }; 5681 5682 /* BPF_FUNC_csum_level level values. */ 5683 enum { 5684 BPF_CSUM_LEVEL_QUERY, 5685 BPF_CSUM_LEVEL_INC, 5686 BPF_CSUM_LEVEL_DEC, 5687 BPF_CSUM_LEVEL_RESET, 5688 }; 5689 5690 /* BPF_FUNC_skb_adjust_room flags. */ 5691 enum { 5692 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5693 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5694 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5695 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5696 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5697 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5698 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5699 }; 5700 5701 enum { 5702 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5703 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5704 }; 5705 5706 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5707 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5708 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5709 5710 /* BPF_FUNC_sysctl_get_name flags. */ 5711 enum { 5712 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5713 }; 5714 5715 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 5716 enum { 5717 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5718 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5719 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5720 */ 5721 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5722 }; 5723 5724 /* BPF_FUNC_read_branch_records flags. */ 5725 enum { 5726 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5727 }; 5728 5729 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5730 * BPF_FUNC_bpf_ringbuf_output flags. 5731 */ 5732 enum { 5733 BPF_RB_NO_WAKEUP = (1ULL << 0), 5734 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5735 }; 5736 5737 /* BPF_FUNC_bpf_ringbuf_query flags */ 5738 enum { 5739 BPF_RB_AVAIL_DATA = 0, 5740 BPF_RB_RING_SIZE = 1, 5741 BPF_RB_CONS_POS = 2, 5742 BPF_RB_PROD_POS = 3, 5743 }; 5744 5745 /* BPF ring buffer constants */ 5746 enum { 5747 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5748 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5749 BPF_RINGBUF_HDR_SZ = 8, 5750 }; 5751 5752 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5753 enum { 5754 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5755 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5756 }; 5757 5758 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 5759 enum bpf_adj_room_mode { 5760 BPF_ADJ_ROOM_NET, 5761 BPF_ADJ_ROOM_MAC, 5762 }; 5763 5764 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5765 enum bpf_hdr_start_off { 5766 BPF_HDR_START_MAC, 5767 BPF_HDR_START_NET, 5768 }; 5769 5770 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5771 enum bpf_lwt_encap_mode { 5772 BPF_LWT_ENCAP_SEG6, 5773 BPF_LWT_ENCAP_SEG6_INLINE, 5774 BPF_LWT_ENCAP_IP, 5775 }; 5776 5777 /* Flags for bpf_bprm_opts_set helper */ 5778 enum { 5779 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5780 }; 5781 5782 /* Flags for bpf_redirect_map helper */ 5783 enum { 5784 BPF_F_BROADCAST = (1ULL << 3), 5785 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 5786 }; 5787 5788 #define __bpf_md_ptr(type, name) \ 5789 union { \ 5790 type name; \ 5791 __u64 :64; \ 5792 } __attribute__((aligned(8))) 5793 5794 enum { 5795 BPF_SKB_TSTAMP_UNSPEC, 5796 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 5797 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 5798 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 5799 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 5800 */ 5801 }; 5802 5803 /* user accessible mirror of in-kernel sk_buff. 5804 * new fields can only be added to the end of this structure 5805 */ 5806 struct __sk_buff { 5807 __u32 len; 5808 __u32 pkt_type; 5809 __u32 mark; 5810 __u32 queue_mapping; 5811 __u32 protocol; 5812 __u32 vlan_present; 5813 __u32 vlan_tci; 5814 __u32 vlan_proto; 5815 __u32 priority; 5816 __u32 ingress_ifindex; 5817 __u32 ifindex; 5818 __u32 tc_index; 5819 __u32 cb[5]; 5820 __u32 hash; 5821 __u32 tc_classid; 5822 __u32 data; 5823 __u32 data_end; 5824 __u32 napi_id; 5825 5826 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 5827 __u32 family; 5828 __u32 remote_ip4; /* Stored in network byte order */ 5829 __u32 local_ip4; /* Stored in network byte order */ 5830 __u32 remote_ip6[4]; /* Stored in network byte order */ 5831 __u32 local_ip6[4]; /* Stored in network byte order */ 5832 __u32 remote_port; /* Stored in network byte order */ 5833 __u32 local_port; /* stored in host byte order */ 5834 /* ... here. */ 5835 5836 __u32 data_meta; 5837 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5838 __u64 tstamp; 5839 __u32 wire_len; 5840 __u32 gso_segs; 5841 __bpf_md_ptr(struct bpf_sock *, sk); 5842 __u32 gso_size; 5843 __u8 tstamp_type; 5844 __u32 :24; /* Padding, future use. */ 5845 __u64 hwtstamp; 5846 }; 5847 5848 struct bpf_tunnel_key { 5849 __u32 tunnel_id; 5850 union { 5851 __u32 remote_ipv4; 5852 __u32 remote_ipv6[4]; 5853 }; 5854 __u8 tunnel_tos; 5855 __u8 tunnel_ttl; 5856 union { 5857 __u16 tunnel_ext; /* compat */ 5858 __be16 tunnel_flags; 5859 }; 5860 __u32 tunnel_label; 5861 union { 5862 __u32 local_ipv4; 5863 __u32 local_ipv6[4]; 5864 }; 5865 }; 5866 5867 /* user accessible mirror of in-kernel xfrm_state. 5868 * new fields can only be added to the end of this structure 5869 */ 5870 struct bpf_xfrm_state { 5871 __u32 reqid; 5872 __u32 spi; /* Stored in network byte order */ 5873 __u16 family; 5874 __u16 ext; /* Padding, future use. */ 5875 union { 5876 __u32 remote_ipv4; /* Stored in network byte order */ 5877 __u32 remote_ipv6[4]; /* Stored in network byte order */ 5878 }; 5879 }; 5880 5881 /* Generic BPF return codes which all BPF program types may support. 5882 * The values are binary compatible with their TC_ACT_* counter-part to 5883 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 5884 * programs. 5885 * 5886 * XDP is handled seprately, see XDP_*. 5887 */ 5888 enum bpf_ret_code { 5889 BPF_OK = 0, 5890 /* 1 reserved */ 5891 BPF_DROP = 2, 5892 /* 3-6 reserved */ 5893 BPF_REDIRECT = 7, 5894 /* >127 are reserved for prog type specific return codes. 5895 * 5896 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 5897 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 5898 * changed and should be routed based on its new L3 header. 5899 * (This is an L3 redirect, as opposed to L2 redirect 5900 * represented by BPF_REDIRECT above). 5901 */ 5902 BPF_LWT_REROUTE = 128, 5903 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 5904 * to indicate that no custom dissection was performed, and 5905 * fallback to standard dissector is requested. 5906 */ 5907 BPF_FLOW_DISSECTOR_CONTINUE = 129, 5908 }; 5909 5910 struct bpf_sock { 5911 __u32 bound_dev_if; 5912 __u32 family; 5913 __u32 type; 5914 __u32 protocol; 5915 __u32 mark; 5916 __u32 priority; 5917 /* IP address also allows 1 and 2 bytes access */ 5918 __u32 src_ip4; 5919 __u32 src_ip6[4]; 5920 __u32 src_port; /* host byte order */ 5921 __be16 dst_port; /* network byte order */ 5922 __u16 :16; /* zero padding */ 5923 __u32 dst_ip4; 5924 __u32 dst_ip6[4]; 5925 __u32 state; 5926 __s32 rx_queue_mapping; 5927 }; 5928 5929 struct bpf_tcp_sock { 5930 __u32 snd_cwnd; /* Sending congestion window */ 5931 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5932 __u32 rtt_min; 5933 __u32 snd_ssthresh; /* Slow start size threshold */ 5934 __u32 rcv_nxt; /* What we want to receive next */ 5935 __u32 snd_nxt; /* Next sequence we send */ 5936 __u32 snd_una; /* First byte we want an ack for */ 5937 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5938 __u32 ecn_flags; /* ECN status bits. */ 5939 __u32 rate_delivered; /* saved rate sample: packets delivered */ 5940 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5941 __u32 packets_out; /* Packets which are "in flight" */ 5942 __u32 retrans_out; /* Retransmitted packets out */ 5943 __u32 total_retrans; /* Total retransmits for entire connection */ 5944 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5945 * total number of segments in. 5946 */ 5947 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5948 * total number of data segments in. 5949 */ 5950 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5951 * The total number of segments sent. 5952 */ 5953 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5954 * total number of data segments sent. 5955 */ 5956 __u32 lost_out; /* Lost packets */ 5957 __u32 sacked_out; /* SACK'd packets */ 5958 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5959 * sum(delta(rcv_nxt)), or how many bytes 5960 * were acked. 5961 */ 5962 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5963 * sum(delta(snd_una)), or how many bytes 5964 * were acked. 5965 */ 5966 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 5967 * total number of DSACK blocks received 5968 */ 5969 __u32 delivered; /* Total data packets delivered incl. rexmits */ 5970 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5971 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5972 }; 5973 5974 struct bpf_sock_tuple { 5975 union { 5976 struct { 5977 __be32 saddr; 5978 __be32 daddr; 5979 __be16 sport; 5980 __be16 dport; 5981 } ipv4; 5982 struct { 5983 __be32 saddr[4]; 5984 __be32 daddr[4]; 5985 __be16 sport; 5986 __be16 dport; 5987 } ipv6; 5988 }; 5989 }; 5990 5991 struct bpf_xdp_sock { 5992 __u32 queue_id; 5993 }; 5994 5995 #define XDP_PACKET_HEADROOM 256 5996 5997 /* User return codes for XDP prog type. 5998 * A valid XDP program must return one of these defined values. All other 5999 * return codes are reserved for future use. Unknown return codes will 6000 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6001 */ 6002 enum xdp_action { 6003 XDP_ABORTED = 0, 6004 XDP_DROP, 6005 XDP_PASS, 6006 XDP_TX, 6007 XDP_REDIRECT, 6008 }; 6009 6010 /* user accessible metadata for XDP packet hook 6011 * new fields must be added to the end of this structure 6012 */ 6013 struct xdp_md { 6014 __u32 data; 6015 __u32 data_end; 6016 __u32 data_meta; 6017 /* Below access go through struct xdp_rxq_info */ 6018 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6019 __u32 rx_queue_index; /* rxq->queue_index */ 6020 6021 __u32 egress_ifindex; /* txq->dev->ifindex */ 6022 }; 6023 6024 /* DEVMAP map-value layout 6025 * 6026 * The struct data-layout of map-value is a configuration interface. 6027 * New members can only be added to the end of this structure. 6028 */ 6029 struct bpf_devmap_val { 6030 __u32 ifindex; /* device index */ 6031 union { 6032 int fd; /* prog fd on map write */ 6033 __u32 id; /* prog id on map read */ 6034 } bpf_prog; 6035 }; 6036 6037 /* CPUMAP map-value layout 6038 * 6039 * The struct data-layout of map-value is a configuration interface. 6040 * New members can only be added to the end of this structure. 6041 */ 6042 struct bpf_cpumap_val { 6043 __u32 qsize; /* queue size to remote target CPU */ 6044 union { 6045 int fd; /* prog fd on map write */ 6046 __u32 id; /* prog id on map read */ 6047 } bpf_prog; 6048 }; 6049 6050 enum sk_action { 6051 SK_DROP = 0, 6052 SK_PASS, 6053 }; 6054 6055 /* user accessible metadata for SK_MSG packet hook, new fields must 6056 * be added to the end of this structure 6057 */ 6058 struct sk_msg_md { 6059 __bpf_md_ptr(void *, data); 6060 __bpf_md_ptr(void *, data_end); 6061 6062 __u32 family; 6063 __u32 remote_ip4; /* Stored in network byte order */ 6064 __u32 local_ip4; /* Stored in network byte order */ 6065 __u32 remote_ip6[4]; /* Stored in network byte order */ 6066 __u32 local_ip6[4]; /* Stored in network byte order */ 6067 __u32 remote_port; /* Stored in network byte order */ 6068 __u32 local_port; /* stored in host byte order */ 6069 __u32 size; /* Total size of sk_msg */ 6070 6071 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6072 }; 6073 6074 struct sk_reuseport_md { 6075 /* 6076 * Start of directly accessible data. It begins from 6077 * the tcp/udp header. 6078 */ 6079 __bpf_md_ptr(void *, data); 6080 /* End of directly accessible data */ 6081 __bpf_md_ptr(void *, data_end); 6082 /* 6083 * Total length of packet (starting from the tcp/udp header). 6084 * Note that the directly accessible bytes (data_end - data) 6085 * could be less than this "len". Those bytes could be 6086 * indirectly read by a helper "bpf_skb_load_bytes()". 6087 */ 6088 __u32 len; 6089 /* 6090 * Eth protocol in the mac header (network byte order). e.g. 6091 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6092 */ 6093 __u32 eth_protocol; 6094 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6095 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6096 __u32 hash; /* A hash of the packet 4 tuples */ 6097 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6098 * new incoming connection request (e.g. selecting a listen sk for 6099 * the received SYN in the TCP case). reuse->sk is one of the sk 6100 * in the reuseport group. The bpf prog can use reuse->sk to learn 6101 * the local listening ip/port without looking into the skb. 6102 * 6103 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6104 * reuse->migrating_sk is the socket that needs to be migrated 6105 * to another listening socket. migrating_sk could be a fullsock 6106 * sk that is fully established or a reqsk that is in-the-middle 6107 * of 3-way handshake. 6108 */ 6109 __bpf_md_ptr(struct bpf_sock *, sk); 6110 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6111 }; 6112 6113 #define BPF_TAG_SIZE 8 6114 6115 struct bpf_prog_info { 6116 __u32 type; 6117 __u32 id; 6118 __u8 tag[BPF_TAG_SIZE]; 6119 __u32 jited_prog_len; 6120 __u32 xlated_prog_len; 6121 __aligned_u64 jited_prog_insns; 6122 __aligned_u64 xlated_prog_insns; 6123 __u64 load_time; /* ns since boottime */ 6124 __u32 created_by_uid; 6125 __u32 nr_map_ids; 6126 __aligned_u64 map_ids; 6127 char name[BPF_OBJ_NAME_LEN]; 6128 __u32 ifindex; 6129 __u32 gpl_compatible:1; 6130 __u32 :31; /* alignment pad */ 6131 __u64 netns_dev; 6132 __u64 netns_ino; 6133 __u32 nr_jited_ksyms; 6134 __u32 nr_jited_func_lens; 6135 __aligned_u64 jited_ksyms; 6136 __aligned_u64 jited_func_lens; 6137 __u32 btf_id; 6138 __u32 func_info_rec_size; 6139 __aligned_u64 func_info; 6140 __u32 nr_func_info; 6141 __u32 nr_line_info; 6142 __aligned_u64 line_info; 6143 __aligned_u64 jited_line_info; 6144 __u32 nr_jited_line_info; 6145 __u32 line_info_rec_size; 6146 __u32 jited_line_info_rec_size; 6147 __u32 nr_prog_tags; 6148 __aligned_u64 prog_tags; 6149 __u64 run_time_ns; 6150 __u64 run_cnt; 6151 __u64 recursion_misses; 6152 __u32 verified_insns; 6153 __u32 attach_btf_obj_id; 6154 __u32 attach_btf_id; 6155 } __attribute__((aligned(8))); 6156 6157 struct bpf_map_info { 6158 __u32 type; 6159 __u32 id; 6160 __u32 key_size; 6161 __u32 value_size; 6162 __u32 max_entries; 6163 __u32 map_flags; 6164 char name[BPF_OBJ_NAME_LEN]; 6165 __u32 ifindex; 6166 __u32 btf_vmlinux_value_type_id; 6167 __u64 netns_dev; 6168 __u64 netns_ino; 6169 __u32 btf_id; 6170 __u32 btf_key_type_id; 6171 __u32 btf_value_type_id; 6172 __u32 :32; /* alignment pad */ 6173 __u64 map_extra; 6174 } __attribute__((aligned(8))); 6175 6176 struct bpf_btf_info { 6177 __aligned_u64 btf; 6178 __u32 btf_size; 6179 __u32 id; 6180 __aligned_u64 name; 6181 __u32 name_len; 6182 __u32 kernel_btf; 6183 } __attribute__((aligned(8))); 6184 6185 struct bpf_link_info { 6186 __u32 type; 6187 __u32 id; 6188 __u32 prog_id; 6189 union { 6190 struct { 6191 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6192 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6193 } raw_tracepoint; 6194 struct { 6195 __u32 attach_type; 6196 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6197 __u32 target_btf_id; /* BTF type id inside the object */ 6198 } tracing; 6199 struct { 6200 __u64 cgroup_id; 6201 __u32 attach_type; 6202 } cgroup; 6203 struct { 6204 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6205 __u32 target_name_len; /* in/out: target_name buffer len */ 6206 6207 /* If the iter specific field is 32 bits, it can be put 6208 * in the first or second union. Otherwise it should be 6209 * put in the second union. 6210 */ 6211 union { 6212 struct { 6213 __u32 map_id; 6214 } map; 6215 }; 6216 union { 6217 struct { 6218 __u64 cgroup_id; 6219 __u32 order; 6220 } cgroup; 6221 }; 6222 } iter; 6223 struct { 6224 __u32 netns_ino; 6225 __u32 attach_type; 6226 } netns; 6227 struct { 6228 __u32 ifindex; 6229 } xdp; 6230 }; 6231 } __attribute__((aligned(8))); 6232 6233 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6234 * by user and intended to be used by socket (e.g. to bind to, depends on 6235 * attach type). 6236 */ 6237 struct bpf_sock_addr { 6238 __u32 user_family; /* Allows 4-byte read, but no write. */ 6239 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6240 * Stored in network byte order. 6241 */ 6242 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6243 * Stored in network byte order. 6244 */ 6245 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6246 * Stored in network byte order 6247 */ 6248 __u32 family; /* Allows 4-byte read, but no write */ 6249 __u32 type; /* Allows 4-byte read, but no write */ 6250 __u32 protocol; /* Allows 4-byte read, but no write */ 6251 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6252 * Stored in network byte order. 6253 */ 6254 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6255 * Stored in network byte order. 6256 */ 6257 __bpf_md_ptr(struct bpf_sock *, sk); 6258 }; 6259 6260 /* User bpf_sock_ops struct to access socket values and specify request ops 6261 * and their replies. 6262 * Some of this fields are in network (bigendian) byte order and may need 6263 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6264 * New fields can only be added at the end of this structure 6265 */ 6266 struct bpf_sock_ops { 6267 __u32 op; 6268 union { 6269 __u32 args[4]; /* Optionally passed to bpf program */ 6270 __u32 reply; /* Returned by bpf program */ 6271 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6272 }; 6273 __u32 family; 6274 __u32 remote_ip4; /* Stored in network byte order */ 6275 __u32 local_ip4; /* Stored in network byte order */ 6276 __u32 remote_ip6[4]; /* Stored in network byte order */ 6277 __u32 local_ip6[4]; /* Stored in network byte order */ 6278 __u32 remote_port; /* Stored in network byte order */ 6279 __u32 local_port; /* stored in host byte order */ 6280 __u32 is_fullsock; /* Some TCP fields are only valid if 6281 * there is a full socket. If not, the 6282 * fields read as zero. 6283 */ 6284 __u32 snd_cwnd; 6285 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6286 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6287 __u32 state; 6288 __u32 rtt_min; 6289 __u32 snd_ssthresh; 6290 __u32 rcv_nxt; 6291 __u32 snd_nxt; 6292 __u32 snd_una; 6293 __u32 mss_cache; 6294 __u32 ecn_flags; 6295 __u32 rate_delivered; 6296 __u32 rate_interval_us; 6297 __u32 packets_out; 6298 __u32 retrans_out; 6299 __u32 total_retrans; 6300 __u32 segs_in; 6301 __u32 data_segs_in; 6302 __u32 segs_out; 6303 __u32 data_segs_out; 6304 __u32 lost_out; 6305 __u32 sacked_out; 6306 __u32 sk_txhash; 6307 __u64 bytes_received; 6308 __u64 bytes_acked; 6309 __bpf_md_ptr(struct bpf_sock *, sk); 6310 /* [skb_data, skb_data_end) covers the whole TCP header. 6311 * 6312 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6313 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6314 * header has not been written. 6315 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6316 * been written so far. 6317 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6318 * the 3WHS. 6319 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6320 * the 3WHS. 6321 * 6322 * bpf_load_hdr_opt() can also be used to read a particular option. 6323 */ 6324 __bpf_md_ptr(void *, skb_data); 6325 __bpf_md_ptr(void *, skb_data_end); 6326 __u32 skb_len; /* The total length of a packet. 6327 * It includes the header, options, 6328 * and payload. 6329 */ 6330 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6331 * an easy way to check for tcp_flags 6332 * without parsing skb_data. 6333 * 6334 * In particular, the skb_tcp_flags 6335 * will still be available in 6336 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6337 * the outgoing header has not 6338 * been written yet. 6339 */ 6340 }; 6341 6342 /* Definitions for bpf_sock_ops_cb_flags */ 6343 enum { 6344 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6345 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6346 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6347 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6348 /* Call bpf for all received TCP headers. The bpf prog will be 6349 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6350 * 6351 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6352 * for the header option related helpers that will be useful 6353 * to the bpf programs. 6354 * 6355 * It could be used at the client/active side (i.e. connect() side) 6356 * when the server told it that the server was in syncookie 6357 * mode and required the active side to resend the bpf-written 6358 * options. The active side can keep writing the bpf-options until 6359 * it received a valid packet from the server side to confirm 6360 * the earlier packet (and options) has been received. The later 6361 * example patch is using it like this at the active side when the 6362 * server is in syncookie mode. 6363 * 6364 * The bpf prog will usually turn this off in the common cases. 6365 */ 6366 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6367 /* Call bpf when kernel has received a header option that 6368 * the kernel cannot handle. The bpf prog will be called under 6369 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6370 * 6371 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6372 * for the header option related helpers that will be useful 6373 * to the bpf programs. 6374 */ 6375 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6376 /* Call bpf when the kernel is writing header options for the 6377 * outgoing packet. The bpf prog will first be called 6378 * to reserve space in a skb under 6379 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6380 * the bpf prog will be called to write the header option(s) 6381 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6382 * 6383 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6384 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6385 * related helpers that will be useful to the bpf programs. 6386 * 6387 * The kernel gets its chance to reserve space and write 6388 * options first before the BPF program does. 6389 */ 6390 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6391 /* Mask of all currently supported cb flags */ 6392 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6393 }; 6394 6395 /* List of known BPF sock_ops operators. 6396 * New entries can only be added at the end 6397 */ 6398 enum { 6399 BPF_SOCK_OPS_VOID, 6400 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6401 * -1 if default value should be used 6402 */ 6403 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6404 * window (in packets) or -1 if default 6405 * value should be used 6406 */ 6407 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6408 * active connection is initialized 6409 */ 6410 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6411 * active connection is 6412 * established 6413 */ 6414 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6415 * passive connection is 6416 * established 6417 */ 6418 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6419 * needs ECN 6420 */ 6421 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6422 * based on the path and may be 6423 * dependent on the congestion control 6424 * algorithm. In general it indicates 6425 * a congestion threshold. RTTs above 6426 * this indicate congestion 6427 */ 6428 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6429 * Arg1: value of icsk_retransmits 6430 * Arg2: value of icsk_rto 6431 * Arg3: whether RTO has expired 6432 */ 6433 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6434 * Arg1: sequence number of 1st byte 6435 * Arg2: # segments 6436 * Arg3: return value of 6437 * tcp_transmit_skb (0 => success) 6438 */ 6439 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6440 * Arg1: old_state 6441 * Arg2: new_state 6442 */ 6443 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6444 * socket transition to LISTEN state. 6445 */ 6446 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6447 */ 6448 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6449 * It will be called to handle 6450 * the packets received at 6451 * an already established 6452 * connection. 6453 * 6454 * sock_ops->skb_data: 6455 * Referring to the received skb. 6456 * It covers the TCP header only. 6457 * 6458 * bpf_load_hdr_opt() can also 6459 * be used to search for a 6460 * particular option. 6461 */ 6462 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6463 * header option later in 6464 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6465 * Arg1: bool want_cookie. (in 6466 * writing SYNACK only) 6467 * 6468 * sock_ops->skb_data: 6469 * Not available because no header has 6470 * been written yet. 6471 * 6472 * sock_ops->skb_tcp_flags: 6473 * The tcp_flags of the 6474 * outgoing skb. (e.g. SYN, ACK, FIN). 6475 * 6476 * bpf_reserve_hdr_opt() should 6477 * be used to reserve space. 6478 */ 6479 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6480 * Arg1: bool want_cookie. (in 6481 * writing SYNACK only) 6482 * 6483 * sock_ops->skb_data: 6484 * Referring to the outgoing skb. 6485 * It covers the TCP header 6486 * that has already been written 6487 * by the kernel and the 6488 * earlier bpf-progs. 6489 * 6490 * sock_ops->skb_tcp_flags: 6491 * The tcp_flags of the outgoing 6492 * skb. (e.g. SYN, ACK, FIN). 6493 * 6494 * bpf_store_hdr_opt() should 6495 * be used to write the 6496 * option. 6497 * 6498 * bpf_load_hdr_opt() can also 6499 * be used to search for a 6500 * particular option that 6501 * has already been written 6502 * by the kernel or the 6503 * earlier bpf-progs. 6504 */ 6505 }; 6506 6507 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6508 * changes between the TCP and BPF versions. Ideally this should never happen. 6509 * If it does, we need to add code to convert them before calling 6510 * the BPF sock_ops function. 6511 */ 6512 enum { 6513 BPF_TCP_ESTABLISHED = 1, 6514 BPF_TCP_SYN_SENT, 6515 BPF_TCP_SYN_RECV, 6516 BPF_TCP_FIN_WAIT1, 6517 BPF_TCP_FIN_WAIT2, 6518 BPF_TCP_TIME_WAIT, 6519 BPF_TCP_CLOSE, 6520 BPF_TCP_CLOSE_WAIT, 6521 BPF_TCP_LAST_ACK, 6522 BPF_TCP_LISTEN, 6523 BPF_TCP_CLOSING, /* Now a valid state */ 6524 BPF_TCP_NEW_SYN_RECV, 6525 6526 BPF_TCP_MAX_STATES /* Leave at the end! */ 6527 }; 6528 6529 enum { 6530 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6531 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6532 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6533 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6534 /* Copy the SYN pkt to optval 6535 * 6536 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6537 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6538 * to only getting from the saved_syn. It can either get the 6539 * syn packet from: 6540 * 6541 * 1. the just-received SYN packet (only available when writing the 6542 * SYNACK). It will be useful when it is not necessary to 6543 * save the SYN packet for latter use. It is also the only way 6544 * to get the SYN during syncookie mode because the syn 6545 * packet cannot be saved during syncookie. 6546 * 6547 * OR 6548 * 6549 * 2. the earlier saved syn which was done by 6550 * bpf_setsockopt(TCP_SAVE_SYN). 6551 * 6552 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 6553 * SYN packet is obtained. 6554 * 6555 * If the bpf-prog does not need the IP[46] header, the 6556 * bpf-prog can avoid parsing the IP header by using 6557 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 6558 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 6559 * 6560 * >0: Total number of bytes copied 6561 * -ENOSPC: Not enough space in optval. Only optlen number of 6562 * bytes is copied. 6563 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 6564 * is not saved by setsockopt(TCP_SAVE_SYN). 6565 */ 6566 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 6567 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 6568 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 6569 }; 6570 6571 enum { 6572 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 6573 }; 6574 6575 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 6576 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6577 */ 6578 enum { 6579 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 6580 * total option spaces 6581 * required for an established 6582 * sk in order to calculate the 6583 * MSS. No skb is actually 6584 * sent. 6585 */ 6586 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 6587 * when sending a SYN. 6588 */ 6589 }; 6590 6591 struct bpf_perf_event_value { 6592 __u64 counter; 6593 __u64 enabled; 6594 __u64 running; 6595 }; 6596 6597 enum { 6598 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 6599 BPF_DEVCG_ACC_READ = (1ULL << 1), 6600 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 6601 }; 6602 6603 enum { 6604 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 6605 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 6606 }; 6607 6608 struct bpf_cgroup_dev_ctx { 6609 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 6610 __u32 access_type; 6611 __u32 major; 6612 __u32 minor; 6613 }; 6614 6615 struct bpf_raw_tracepoint_args { 6616 __u64 args[0]; 6617 }; 6618 6619 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 6620 * OUTPUT: Do lookup from egress perspective; default is ingress 6621 */ 6622 enum { 6623 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 6624 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 6625 }; 6626 6627 enum { 6628 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 6629 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 6630 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 6631 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 6632 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 6633 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 6634 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 6635 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 6636 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6637 }; 6638 6639 struct bpf_fib_lookup { 6640 /* input: network family for lookup (AF_INET, AF_INET6) 6641 * output: network family of egress nexthop 6642 */ 6643 __u8 family; 6644 6645 /* set if lookup is to consider L4 data - e.g., FIB rules */ 6646 __u8 l4_protocol; 6647 __be16 sport; 6648 __be16 dport; 6649 6650 union { /* used for MTU check */ 6651 /* input to lookup */ 6652 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 6653 6654 /* output: MTU value */ 6655 __u16 mtu_result; 6656 }; 6657 /* input: L3 device index for lookup 6658 * output: device index from FIB lookup 6659 */ 6660 __u32 ifindex; 6661 6662 union { 6663 /* inputs to lookup */ 6664 __u8 tos; /* AF_INET */ 6665 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 6666 6667 /* output: metric of fib result (IPv4/IPv6 only) */ 6668 __u32 rt_metric; 6669 }; 6670 6671 union { 6672 __be32 ipv4_src; 6673 __u32 ipv6_src[4]; /* in6_addr; network order */ 6674 }; 6675 6676 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 6677 * network header. output: bpf_fib_lookup sets to gateway address 6678 * if FIB lookup returns gateway route 6679 */ 6680 union { 6681 __be32 ipv4_dst; 6682 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6683 }; 6684 6685 /* output */ 6686 __be16 h_vlan_proto; 6687 __be16 h_vlan_TCI; 6688 __u8 smac[6]; /* ETH_ALEN */ 6689 __u8 dmac[6]; /* ETH_ALEN */ 6690 }; 6691 6692 struct bpf_redir_neigh { 6693 /* network family for lookup (AF_INET, AF_INET6) */ 6694 __u32 nh_family; 6695 /* network address of nexthop; skips fib lookup to find gateway */ 6696 union { 6697 __be32 ipv4_nh; 6698 __u32 ipv6_nh[4]; /* in6_addr; network order */ 6699 }; 6700 }; 6701 6702 /* bpf_check_mtu flags*/ 6703 enum bpf_check_mtu_flags { 6704 BPF_MTU_CHK_SEGS = (1U << 0), 6705 }; 6706 6707 enum bpf_check_mtu_ret { 6708 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 6709 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 6710 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 6711 }; 6712 6713 enum bpf_task_fd_type { 6714 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 6715 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 6716 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 6717 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 6718 BPF_FD_TYPE_UPROBE, /* filename + offset */ 6719 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 6720 }; 6721 6722 enum { 6723 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 6724 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 6725 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 6726 }; 6727 6728 struct bpf_flow_keys { 6729 __u16 nhoff; 6730 __u16 thoff; 6731 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6732 __u8 is_frag; 6733 __u8 is_first_frag; 6734 __u8 is_encap; 6735 __u8 ip_proto; 6736 __be16 n_proto; 6737 __be16 sport; 6738 __be16 dport; 6739 union { 6740 struct { 6741 __be32 ipv4_src; 6742 __be32 ipv4_dst; 6743 }; 6744 struct { 6745 __u32 ipv6_src[4]; /* in6_addr; network order */ 6746 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6747 }; 6748 }; 6749 __u32 flags; 6750 __be32 flow_label; 6751 }; 6752 6753 struct bpf_func_info { 6754 __u32 insn_off; 6755 __u32 type_id; 6756 }; 6757 6758 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6759 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6760 6761 struct bpf_line_info { 6762 __u32 insn_off; 6763 __u32 file_name_off; 6764 __u32 line_off; 6765 __u32 line_col; 6766 }; 6767 6768 struct bpf_spin_lock { 6769 __u32 val; 6770 }; 6771 6772 struct bpf_timer { 6773 __u64 :64; 6774 __u64 :64; 6775 } __attribute__((aligned(8))); 6776 6777 struct bpf_dynptr { 6778 __u64 :64; 6779 __u64 :64; 6780 } __attribute__((aligned(8))); 6781 6782 struct bpf_sysctl { 6783 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 6784 * Allows 1,2,4-byte read, but no write. 6785 */ 6786 __u32 file_pos; /* Sysctl file position to read from, write to. 6787 * Allows 1,2,4-byte read an 4-byte write. 6788 */ 6789 }; 6790 6791 struct bpf_sockopt { 6792 __bpf_md_ptr(struct bpf_sock *, sk); 6793 __bpf_md_ptr(void *, optval); 6794 __bpf_md_ptr(void *, optval_end); 6795 6796 __s32 level; 6797 __s32 optname; 6798 __s32 optlen; 6799 __s32 retval; 6800 }; 6801 6802 struct bpf_pidns_info { 6803 __u32 pid; 6804 __u32 tgid; 6805 }; 6806 6807 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6808 struct bpf_sk_lookup { 6809 union { 6810 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 6811 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 6812 }; 6813 6814 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6815 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6816 __u32 remote_ip4; /* Network byte order */ 6817 __u32 remote_ip6[4]; /* Network byte order */ 6818 __be16 remote_port; /* Network byte order */ 6819 __u16 :16; /* Zero padding */ 6820 __u32 local_ip4; /* Network byte order */ 6821 __u32 local_ip6[4]; /* Network byte order */ 6822 __u32 local_port; /* Host byte order */ 6823 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 6824 }; 6825 6826 /* 6827 * struct btf_ptr is used for typed pointer representation; the 6828 * type id is used to render the pointer data as the appropriate type 6829 * via the bpf_snprintf_btf() helper described above. A flags field - 6830 * potentially to specify additional details about the BTF pointer 6831 * (rather than its mode of display) - is included for future use. 6832 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6833 */ 6834 struct btf_ptr { 6835 void *ptr; 6836 __u32 type_id; 6837 __u32 flags; /* BTF ptr flags; unused at present. */ 6838 }; 6839 6840 /* 6841 * Flags to control bpf_snprintf_btf() behaviour. 6842 * - BTF_F_COMPACT: no formatting around type information 6843 * - BTF_F_NONAME: no struct/union member names/types 6844 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6845 * equivalent to %px. 6846 * - BTF_F_ZERO: show zero-valued struct/union members; they 6847 * are not displayed by default 6848 */ 6849 enum { 6850 BTF_F_COMPACT = (1ULL << 0), 6851 BTF_F_NONAME = (1ULL << 1), 6852 BTF_F_PTR_RAW = (1ULL << 2), 6853 BTF_F_ZERO = (1ULL << 3), 6854 }; 6855 6856 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 6857 * has to be adjusted by relocations. It is emitted by llvm and passed to 6858 * libbpf and later to the kernel. 6859 */ 6860 enum bpf_core_relo_kind { 6861 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 6862 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 6863 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 6864 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 6865 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 6866 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 6867 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 6868 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 6869 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 6870 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 6871 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 6872 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 6873 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 6874 }; 6875 6876 /* 6877 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 6878 * and from libbpf to the kernel. 6879 * 6880 * CO-RE relocation captures the following data: 6881 * - insn_off - instruction offset (in bytes) within a BPF program that needs 6882 * its insn->imm field to be relocated with actual field info; 6883 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 6884 * type or field; 6885 * - access_str_off - offset into corresponding .BTF string section. String 6886 * interpretation depends on specific relocation kind: 6887 * - for field-based relocations, string encodes an accessed field using 6888 * a sequence of field and array indices, separated by colon (:). It's 6889 * conceptually very close to LLVM's getelementptr ([0]) instruction's 6890 * arguments for identifying offset to a field. 6891 * - for type-based relocations, strings is expected to be just "0"; 6892 * - for enum value-based relocations, string contains an index of enum 6893 * value within its enum type; 6894 * - kind - one of enum bpf_core_relo_kind; 6895 * 6896 * Example: 6897 * struct sample { 6898 * int a; 6899 * struct { 6900 * int b[10]; 6901 * }; 6902 * }; 6903 * 6904 * struct sample *s = ...; 6905 * int *x = &s->a; // encoded as "0:0" (a is field #0) 6906 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 6907 * // b is field #0 inside anon struct, accessing elem #5) 6908 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 6909 * 6910 * type_id for all relocs in this example will capture BTF type id of 6911 * `struct sample`. 6912 * 6913 * Such relocation is emitted when using __builtin_preserve_access_index() 6914 * Clang built-in, passing expression that captures field address, e.g.: 6915 * 6916 * bpf_probe_read(&dst, sizeof(dst), 6917 * __builtin_preserve_access_index(&src->a.b.c)); 6918 * 6919 * In this case Clang will emit field relocation recording necessary data to 6920 * be able to find offset of embedded `a.b.c` field within `src` struct. 6921 * 6922 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 6923 */ 6924 struct bpf_core_relo { 6925 __u32 insn_off; 6926 __u32 type_id; 6927 __u32 access_str_off; 6928 enum bpf_core_relo_kind kind; 6929 }; 6930 6931 #endif /* _UAPI__LINUX_BPF_H__ */ 6932