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_MEMSX 0x80 /* load with sign extension */ 23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 25 26 /* alu/jmp fields */ 27 #define BPF_MOV 0xb0 /* mov reg to reg */ 28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 29 30 /* change endianness of a register */ 31 #define BPF_END 0xd0 /* flags for endianness conversion: */ 32 #define BPF_TO_LE 0x00 /* convert to little-endian */ 33 #define BPF_TO_BE 0x08 /* convert to big-endian */ 34 #define BPF_FROM_LE BPF_TO_LE 35 #define BPF_FROM_BE BPF_TO_BE 36 37 /* jmp encodings */ 38 #define BPF_JNE 0x50 /* jump != */ 39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 45 #define BPF_CALL 0x80 /* function call */ 46 #define BPF_EXIT 0x90 /* function return */ 47 48 /* atomic op type fields (stored in immediate) */ 49 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 50 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 51 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 52 53 /* Register numbers */ 54 enum { 55 BPF_REG_0 = 0, 56 BPF_REG_1, 57 BPF_REG_2, 58 BPF_REG_3, 59 BPF_REG_4, 60 BPF_REG_5, 61 BPF_REG_6, 62 BPF_REG_7, 63 BPF_REG_8, 64 BPF_REG_9, 65 BPF_REG_10, 66 __MAX_BPF_REG, 67 }; 68 69 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 70 #define MAX_BPF_REG __MAX_BPF_REG 71 72 struct bpf_insn { 73 __u8 code; /* opcode */ 74 __u8 dst_reg:4; /* dest register */ 75 __u8 src_reg:4; /* source register */ 76 __s16 off; /* signed offset */ 77 __s32 imm; /* signed immediate constant */ 78 }; 79 80 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 81 struct bpf_lpm_trie_key { 82 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 83 __u8 data[0]; /* Arbitrary size */ 84 }; 85 86 struct bpf_cgroup_storage_key { 87 __u64 cgroup_inode_id; /* cgroup inode id */ 88 __u32 attach_type; /* program attach type (enum bpf_attach_type) */ 89 }; 90 91 enum bpf_cgroup_iter_order { 92 BPF_CGROUP_ITER_ORDER_UNSPEC = 0, 93 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */ 94 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */ 95 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */ 96 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */ 97 }; 98 99 union bpf_iter_link_info { 100 struct { 101 __u32 map_fd; 102 } map; 103 struct { 104 enum bpf_cgroup_iter_order order; 105 106 /* At most one of cgroup_fd and cgroup_id can be non-zero. If 107 * both are zero, the walk starts from the default cgroup v2 108 * root. For walking v1 hierarchy, one should always explicitly 109 * specify cgroup_fd. 110 */ 111 __u32 cgroup_fd; 112 __u64 cgroup_id; 113 } cgroup; 114 /* Parameters of task iterators. */ 115 struct { 116 __u32 tid; 117 __u32 pid; 118 __u32 pid_fd; 119 } task; 120 }; 121 122 /* BPF syscall commands, see bpf(2) man-page for more details. */ 123 /** 124 * DOC: eBPF Syscall Preamble 125 * 126 * The operation to be performed by the **bpf**\ () system call is determined 127 * by the *cmd* argument. Each operation takes an accompanying argument, 128 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 129 * below). The size argument is the size of the union pointed to by *attr*. 130 */ 131 /** 132 * DOC: eBPF Syscall Commands 133 * 134 * BPF_MAP_CREATE 135 * Description 136 * Create a map and return a file descriptor that refers to the 137 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 138 * is automatically enabled for the new file descriptor. 139 * 140 * Applying **close**\ (2) to the file descriptor returned by 141 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 142 * 143 * Return 144 * A new file descriptor (a nonnegative integer), or -1 if an 145 * error occurred (in which case, *errno* is set appropriately). 146 * 147 * BPF_MAP_LOOKUP_ELEM 148 * Description 149 * Look up an element with a given *key* in the map referred to 150 * by the file descriptor *map_fd*. 151 * 152 * The *flags* argument may be specified as one of the 153 * following: 154 * 155 * **BPF_F_LOCK** 156 * Look up the value of a spin-locked map without 157 * returning the lock. This must be specified if the 158 * elements contain a spinlock. 159 * 160 * Return 161 * Returns zero on success. On error, -1 is returned and *errno* 162 * is set appropriately. 163 * 164 * BPF_MAP_UPDATE_ELEM 165 * Description 166 * Create or update an element (key/value pair) in a specified map. 167 * 168 * The *flags* argument should be specified as one of the 169 * following: 170 * 171 * **BPF_ANY** 172 * Create a new element or update an existing element. 173 * **BPF_NOEXIST** 174 * Create a new element only if it did not exist. 175 * **BPF_EXIST** 176 * Update an existing element. 177 * **BPF_F_LOCK** 178 * Update a spin_lock-ed map element. 179 * 180 * Return 181 * Returns zero on success. On error, -1 is returned and *errno* 182 * is set appropriately. 183 * 184 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 185 * **E2BIG**, **EEXIST**, or **ENOENT**. 186 * 187 * **E2BIG** 188 * The number of elements in the map reached the 189 * *max_entries* limit specified at map creation time. 190 * **EEXIST** 191 * If *flags* specifies **BPF_NOEXIST** and the element 192 * with *key* already exists in the map. 193 * **ENOENT** 194 * If *flags* specifies **BPF_EXIST** and the element with 195 * *key* does not exist in the map. 196 * 197 * BPF_MAP_DELETE_ELEM 198 * Description 199 * Look up and delete an element by key in a specified map. 200 * 201 * Return 202 * Returns zero on success. On error, -1 is returned and *errno* 203 * is set appropriately. 204 * 205 * BPF_MAP_GET_NEXT_KEY 206 * Description 207 * Look up an element by key in a specified map and return the key 208 * of the next element. Can be used to iterate over all elements 209 * in the map. 210 * 211 * Return 212 * Returns zero on success. On error, -1 is returned and *errno* 213 * is set appropriately. 214 * 215 * The following cases can be used to iterate over all elements of 216 * the map: 217 * 218 * * If *key* is not found, the operation returns zero and sets 219 * the *next_key* pointer to the key of the first element. 220 * * If *key* is found, the operation returns zero and sets the 221 * *next_key* pointer to the key of the next element. 222 * * If *key* is the last element, returns -1 and *errno* is set 223 * to **ENOENT**. 224 * 225 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 226 * **EINVAL** on error. 227 * 228 * BPF_PROG_LOAD 229 * Description 230 * Verify and load an eBPF program, returning a new file 231 * descriptor associated with the program. 232 * 233 * Applying **close**\ (2) to the file descriptor returned by 234 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 235 * 236 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 237 * automatically enabled for the new file descriptor. 238 * 239 * Return 240 * A new file descriptor (a nonnegative integer), or -1 if an 241 * error occurred (in which case, *errno* is set appropriately). 242 * 243 * BPF_OBJ_PIN 244 * Description 245 * Pin an eBPF program or map referred by the specified *bpf_fd* 246 * to the provided *pathname* on the filesystem. 247 * 248 * The *pathname* argument must not contain a dot ("."). 249 * 250 * On success, *pathname* retains a reference to the eBPF object, 251 * preventing deallocation of the object when the original 252 * *bpf_fd* is closed. This allow the eBPF object to live beyond 253 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 254 * process. 255 * 256 * Applying **unlink**\ (2) or similar calls to the *pathname* 257 * unpins the object from the filesystem, removing the reference. 258 * If no other file descriptors or filesystem nodes refer to the 259 * same object, it will be deallocated (see NOTES). 260 * 261 * The filesystem type for the parent directory of *pathname* must 262 * be **BPF_FS_MAGIC**. 263 * 264 * Return 265 * Returns zero on success. On error, -1 is returned and *errno* 266 * is set appropriately. 267 * 268 * BPF_OBJ_GET 269 * Description 270 * Open a file descriptor for the eBPF object pinned to the 271 * specified *pathname*. 272 * 273 * Return 274 * A new file descriptor (a nonnegative integer), or -1 if an 275 * error occurred (in which case, *errno* is set appropriately). 276 * 277 * BPF_PROG_ATTACH 278 * Description 279 * Attach an eBPF program to a *target_fd* at the specified 280 * *attach_type* hook. 281 * 282 * The *attach_type* specifies the eBPF attachment point to 283 * attach the program to, and must be one of *bpf_attach_type* 284 * (see below). 285 * 286 * The *attach_bpf_fd* must be a valid file descriptor for a 287 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 288 * or sock_ops type corresponding to the specified *attach_type*. 289 * 290 * The *target_fd* must be a valid file descriptor for a kernel 291 * object which depends on the attach type of *attach_bpf_fd*: 292 * 293 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 294 * **BPF_PROG_TYPE_CGROUP_SKB**, 295 * **BPF_PROG_TYPE_CGROUP_SOCK**, 296 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 297 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 298 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 299 * **BPF_PROG_TYPE_SOCK_OPS** 300 * 301 * Control Group v2 hierarchy with the eBPF controller 302 * enabled. Requires the kernel to be compiled with 303 * **CONFIG_CGROUP_BPF**. 304 * 305 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 306 * 307 * Network namespace (eg /proc/self/ns/net). 308 * 309 * **BPF_PROG_TYPE_LIRC_MODE2** 310 * 311 * LIRC device path (eg /dev/lircN). Requires the kernel 312 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 313 * 314 * **BPF_PROG_TYPE_SK_SKB**, 315 * **BPF_PROG_TYPE_SK_MSG** 316 * 317 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 318 * 319 * Return 320 * Returns zero on success. On error, -1 is returned and *errno* 321 * is set appropriately. 322 * 323 * BPF_PROG_DETACH 324 * Description 325 * Detach the eBPF program associated with the *target_fd* at the 326 * hook specified by *attach_type*. The program must have been 327 * previously attached using **BPF_PROG_ATTACH**. 328 * 329 * Return 330 * Returns zero on success. On error, -1 is returned and *errno* 331 * is set appropriately. 332 * 333 * BPF_PROG_TEST_RUN 334 * Description 335 * Run the eBPF program associated with the *prog_fd* a *repeat* 336 * number of times against a provided program context *ctx_in* and 337 * data *data_in*, and return the modified program context 338 * *ctx_out*, *data_out* (for example, packet data), result of the 339 * execution *retval*, and *duration* of the test run. 340 * 341 * The sizes of the buffers provided as input and output 342 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 343 * be provided in the corresponding variables *ctx_size_in*, 344 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 345 * of these parameters are not provided (ie set to NULL), the 346 * corresponding size field must be zero. 347 * 348 * Some program types have particular requirements: 349 * 350 * **BPF_PROG_TYPE_SK_LOOKUP** 351 * *data_in* and *data_out* must be NULL. 352 * 353 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 354 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 355 * 356 * *ctx_out*, *data_in* and *data_out* must be NULL. 357 * *repeat* must be zero. 358 * 359 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN. 360 * 361 * Return 362 * Returns zero on success. On error, -1 is returned and *errno* 363 * is set appropriately. 364 * 365 * **ENOSPC** 366 * Either *data_size_out* or *ctx_size_out* is too small. 367 * **ENOTSUPP** 368 * This command is not supported by the program type of 369 * the program referred to by *prog_fd*. 370 * 371 * BPF_PROG_GET_NEXT_ID 372 * Description 373 * Fetch the next eBPF program currently loaded into the kernel. 374 * 375 * Looks for the eBPF program with an id greater than *start_id* 376 * and updates *next_id* on success. If no other eBPF programs 377 * remain with ids higher than *start_id*, returns -1 and sets 378 * *errno* to **ENOENT**. 379 * 380 * Return 381 * Returns zero on success. On error, or when no id remains, -1 382 * is returned and *errno* is set appropriately. 383 * 384 * BPF_MAP_GET_NEXT_ID 385 * Description 386 * Fetch the next eBPF map currently loaded into the kernel. 387 * 388 * Looks for the eBPF map with an id greater than *start_id* 389 * and updates *next_id* on success. If no other eBPF maps 390 * remain with ids higher than *start_id*, returns -1 and sets 391 * *errno* to **ENOENT**. 392 * 393 * Return 394 * Returns zero on success. On error, or when no id remains, -1 395 * is returned and *errno* is set appropriately. 396 * 397 * BPF_PROG_GET_FD_BY_ID 398 * Description 399 * Open a file descriptor for the eBPF program corresponding to 400 * *prog_id*. 401 * 402 * Return 403 * A new file descriptor (a nonnegative integer), or -1 if an 404 * error occurred (in which case, *errno* is set appropriately). 405 * 406 * BPF_MAP_GET_FD_BY_ID 407 * Description 408 * Open a file descriptor for the eBPF map corresponding to 409 * *map_id*. 410 * 411 * Return 412 * A new file descriptor (a nonnegative integer), or -1 if an 413 * error occurred (in which case, *errno* is set appropriately). 414 * 415 * BPF_OBJ_GET_INFO_BY_FD 416 * Description 417 * Obtain information about the eBPF object corresponding to 418 * *bpf_fd*. 419 * 420 * Populates up to *info_len* bytes of *info*, which will be in 421 * one of the following formats depending on the eBPF object type 422 * of *bpf_fd*: 423 * 424 * * **struct bpf_prog_info** 425 * * **struct bpf_map_info** 426 * * **struct bpf_btf_info** 427 * * **struct bpf_link_info** 428 * 429 * Return 430 * Returns zero on success. On error, -1 is returned and *errno* 431 * is set appropriately. 432 * 433 * BPF_PROG_QUERY 434 * Description 435 * Obtain information about eBPF programs associated with the 436 * specified *attach_type* hook. 437 * 438 * The *target_fd* must be a valid file descriptor for a kernel 439 * object which depends on the attach type of *attach_bpf_fd*: 440 * 441 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 442 * **BPF_PROG_TYPE_CGROUP_SKB**, 443 * **BPF_PROG_TYPE_CGROUP_SOCK**, 444 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 445 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 446 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 447 * **BPF_PROG_TYPE_SOCK_OPS** 448 * 449 * Control Group v2 hierarchy with the eBPF controller 450 * enabled. Requires the kernel to be compiled with 451 * **CONFIG_CGROUP_BPF**. 452 * 453 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 454 * 455 * Network namespace (eg /proc/self/ns/net). 456 * 457 * **BPF_PROG_TYPE_LIRC_MODE2** 458 * 459 * LIRC device path (eg /dev/lircN). Requires the kernel 460 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 461 * 462 * **BPF_PROG_QUERY** always fetches the number of programs 463 * attached and the *attach_flags* which were used to attach those 464 * programs. Additionally, if *prog_ids* is nonzero and the number 465 * of attached programs is less than *prog_cnt*, populates 466 * *prog_ids* with the eBPF program ids of the programs attached 467 * at *target_fd*. 468 * 469 * The following flags may alter the result: 470 * 471 * **BPF_F_QUERY_EFFECTIVE** 472 * Only return information regarding programs which are 473 * currently effective at the specified *target_fd*. 474 * 475 * Return 476 * Returns zero on success. On error, -1 is returned and *errno* 477 * is set appropriately. 478 * 479 * BPF_RAW_TRACEPOINT_OPEN 480 * Description 481 * Attach an eBPF program to a tracepoint *name* to access kernel 482 * internal arguments of the tracepoint in their raw form. 483 * 484 * The *prog_fd* must be a valid file descriptor associated with 485 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 486 * 487 * No ABI guarantees are made about the content of tracepoint 488 * arguments exposed to the corresponding eBPF program. 489 * 490 * Applying **close**\ (2) to the file descriptor returned by 491 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 492 * 493 * Return 494 * A new file descriptor (a nonnegative integer), or -1 if an 495 * error occurred (in which case, *errno* is set appropriately). 496 * 497 * BPF_BTF_LOAD 498 * Description 499 * Verify and load BPF Type Format (BTF) metadata into the kernel, 500 * returning a new file descriptor associated with the metadata. 501 * BTF is described in more detail at 502 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 503 * 504 * The *btf* parameter must point to valid memory providing 505 * *btf_size* bytes of BTF binary metadata. 506 * 507 * The returned file descriptor can be passed to other **bpf**\ () 508 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 509 * associate the BTF with those objects. 510 * 511 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 512 * parameters to specify a *btf_log_buf*, *btf_log_size* and 513 * *btf_log_level* which allow the kernel to return freeform log 514 * output regarding the BTF verification process. 515 * 516 * Return 517 * A new file descriptor (a nonnegative integer), or -1 if an 518 * error occurred (in which case, *errno* is set appropriately). 519 * 520 * BPF_BTF_GET_FD_BY_ID 521 * Description 522 * Open a file descriptor for the BPF Type Format (BTF) 523 * corresponding to *btf_id*. 524 * 525 * Return 526 * A new file descriptor (a nonnegative integer), or -1 if an 527 * error occurred (in which case, *errno* is set appropriately). 528 * 529 * BPF_TASK_FD_QUERY 530 * Description 531 * Obtain information about eBPF programs associated with the 532 * target process identified by *pid* and *fd*. 533 * 534 * If the *pid* and *fd* are associated with a tracepoint, kprobe 535 * or uprobe perf event, then the *prog_id* and *fd_type* will 536 * be populated with the eBPF program id and file descriptor type 537 * of type **bpf_task_fd_type**. If associated with a kprobe or 538 * uprobe, the *probe_offset* and *probe_addr* will also be 539 * populated. Optionally, if *buf* is provided, then up to 540 * *buf_len* bytes of *buf* will be populated with the name of 541 * the tracepoint, kprobe or uprobe. 542 * 543 * The resulting *prog_id* may be introspected in deeper detail 544 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 545 * 546 * Return 547 * Returns zero on success. On error, -1 is returned and *errno* 548 * is set appropriately. 549 * 550 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 551 * Description 552 * Look up an element with the given *key* in the map referred to 553 * by the file descriptor *fd*, and if found, delete the element. 554 * 555 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map 556 * types, the *flags* argument needs to be set to 0, but for other 557 * map types, it may be specified as: 558 * 559 * **BPF_F_LOCK** 560 * Look up and delete the value of a spin-locked map 561 * without returning the lock. This must be specified if 562 * the elements contain a spinlock. 563 * 564 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 565 * implement this command as a "pop" operation, deleting the top 566 * element rather than one corresponding to *key*. 567 * The *key* and *key_len* parameters should be zeroed when 568 * issuing this operation for these map types. 569 * 570 * This command is only valid for the following map types: 571 * * **BPF_MAP_TYPE_QUEUE** 572 * * **BPF_MAP_TYPE_STACK** 573 * * **BPF_MAP_TYPE_HASH** 574 * * **BPF_MAP_TYPE_PERCPU_HASH** 575 * * **BPF_MAP_TYPE_LRU_HASH** 576 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH** 577 * 578 * Return 579 * Returns zero on success. On error, -1 is returned and *errno* 580 * is set appropriately. 581 * 582 * BPF_MAP_FREEZE 583 * Description 584 * Freeze the permissions of the specified map. 585 * 586 * Write permissions may be frozen by passing zero *flags*. 587 * Upon success, no future syscall invocations may alter the 588 * map state of *map_fd*. Write operations from eBPF programs 589 * are still possible for a frozen map. 590 * 591 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 592 * 593 * Return 594 * Returns zero on success. On error, -1 is returned and *errno* 595 * is set appropriately. 596 * 597 * BPF_BTF_GET_NEXT_ID 598 * Description 599 * Fetch the next BPF Type Format (BTF) object currently loaded 600 * into the kernel. 601 * 602 * Looks for the BTF object with an id greater than *start_id* 603 * and updates *next_id* on success. If no other BTF objects 604 * remain with ids higher than *start_id*, returns -1 and sets 605 * *errno* to **ENOENT**. 606 * 607 * Return 608 * Returns zero on success. On error, or when no id remains, -1 609 * is returned and *errno* is set appropriately. 610 * 611 * BPF_MAP_LOOKUP_BATCH 612 * Description 613 * Iterate and fetch multiple elements in a map. 614 * 615 * Two opaque values are used to manage batch operations, 616 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 617 * to NULL to begin the batched operation. After each subsequent 618 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 619 * *out_batch* as the *in_batch* for the next operation to 620 * continue iteration from the current point. 621 * 622 * The *keys* and *values* are output parameters which must point 623 * to memory large enough to hold *count* items based on the key 624 * and value size of the map *map_fd*. The *keys* buffer must be 625 * of *key_size* * *count*. The *values* buffer must be of 626 * *value_size* * *count*. 627 * 628 * The *elem_flags* argument may be specified as one of the 629 * following: 630 * 631 * **BPF_F_LOCK** 632 * Look up the value of a spin-locked map without 633 * returning the lock. This must be specified if the 634 * elements contain a spinlock. 635 * 636 * On success, *count* elements from the map are copied into the 637 * user buffer, with the keys copied into *keys* and the values 638 * copied into the corresponding indices in *values*. 639 * 640 * If an error is returned and *errno* is not **EFAULT**, *count* 641 * is set to the number of successfully processed elements. 642 * 643 * Return 644 * Returns zero on success. On error, -1 is returned and *errno* 645 * is set appropriately. 646 * 647 * May set *errno* to **ENOSPC** to indicate that *keys* or 648 * *values* is too small to dump an entire bucket during 649 * iteration of a hash-based map type. 650 * 651 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 652 * Description 653 * Iterate and delete all elements in a map. 654 * 655 * This operation has the same behavior as 656 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 657 * 658 * * Every element that is successfully returned is also deleted 659 * from the map. This is at least *count* elements. Note that 660 * *count* is both an input and an output parameter. 661 * * Upon returning with *errno* set to **EFAULT**, up to 662 * *count* elements may be deleted without returning the keys 663 * and values of the deleted elements. 664 * 665 * Return 666 * Returns zero on success. On error, -1 is returned and *errno* 667 * is set appropriately. 668 * 669 * BPF_MAP_UPDATE_BATCH 670 * Description 671 * Update multiple elements in a map by *key*. 672 * 673 * The *keys* and *values* are input parameters which must point 674 * to memory large enough to hold *count* items based on the key 675 * and value size of the map *map_fd*. The *keys* buffer must be 676 * of *key_size* * *count*. The *values* buffer must be of 677 * *value_size* * *count*. 678 * 679 * Each element specified in *keys* is sequentially updated to the 680 * value in the corresponding index in *values*. The *in_batch* 681 * and *out_batch* parameters are ignored and should be zeroed. 682 * 683 * The *elem_flags* argument should be specified as one of the 684 * following: 685 * 686 * **BPF_ANY** 687 * Create new elements or update a existing elements. 688 * **BPF_NOEXIST** 689 * Create new elements only if they do not exist. 690 * **BPF_EXIST** 691 * Update existing elements. 692 * **BPF_F_LOCK** 693 * Update spin_lock-ed map elements. This must be 694 * specified if the map value contains a spinlock. 695 * 696 * On success, *count* elements from the map are updated. 697 * 698 * If an error is returned and *errno* is not **EFAULT**, *count* 699 * is set to the number of successfully processed elements. 700 * 701 * Return 702 * Returns zero on success. On error, -1 is returned and *errno* 703 * is set appropriately. 704 * 705 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 706 * **E2BIG**. **E2BIG** indicates that the number of elements in 707 * the map reached the *max_entries* limit specified at map 708 * creation time. 709 * 710 * May set *errno* to one of the following error codes under 711 * specific circumstances: 712 * 713 * **EEXIST** 714 * If *flags* specifies **BPF_NOEXIST** and the element 715 * with *key* already exists in the map. 716 * **ENOENT** 717 * If *flags* specifies **BPF_EXIST** and the element with 718 * *key* does not exist in the map. 719 * 720 * BPF_MAP_DELETE_BATCH 721 * Description 722 * Delete multiple elements in a map by *key*. 723 * 724 * The *keys* parameter is an input parameter which must point 725 * to memory large enough to hold *count* items based on the key 726 * size of the map *map_fd*, that is, *key_size* * *count*. 727 * 728 * Each element specified in *keys* is sequentially deleted. The 729 * *in_batch*, *out_batch*, and *values* parameters are ignored 730 * and should be zeroed. 731 * 732 * The *elem_flags* argument may be specified as one of the 733 * following: 734 * 735 * **BPF_F_LOCK** 736 * Look up the value of a spin-locked map without 737 * returning the lock. This must be specified if the 738 * elements contain a spinlock. 739 * 740 * On success, *count* elements from the map are updated. 741 * 742 * If an error is returned and *errno* is not **EFAULT**, *count* 743 * is set to the number of successfully processed elements. If 744 * *errno* is **EFAULT**, up to *count* elements may be been 745 * deleted. 746 * 747 * Return 748 * Returns zero on success. On error, -1 is returned and *errno* 749 * is set appropriately. 750 * 751 * BPF_LINK_CREATE 752 * Description 753 * Attach an eBPF program to a *target_fd* at the specified 754 * *attach_type* hook and return a file descriptor handle for 755 * managing the link. 756 * 757 * Return 758 * A new file descriptor (a nonnegative integer), or -1 if an 759 * error occurred (in which case, *errno* is set appropriately). 760 * 761 * BPF_LINK_UPDATE 762 * Description 763 * Update the eBPF program in the specified *link_fd* to 764 * *new_prog_fd*. 765 * 766 * Return 767 * Returns zero on success. On error, -1 is returned and *errno* 768 * is set appropriately. 769 * 770 * BPF_LINK_GET_FD_BY_ID 771 * Description 772 * Open a file descriptor for the eBPF Link corresponding to 773 * *link_id*. 774 * 775 * Return 776 * A new file descriptor (a nonnegative integer), or -1 if an 777 * error occurred (in which case, *errno* is set appropriately). 778 * 779 * BPF_LINK_GET_NEXT_ID 780 * Description 781 * Fetch the next eBPF link currently loaded into the kernel. 782 * 783 * Looks for the eBPF link with an id greater than *start_id* 784 * and updates *next_id* on success. If no other eBPF links 785 * remain with ids higher than *start_id*, returns -1 and sets 786 * *errno* to **ENOENT**. 787 * 788 * Return 789 * Returns zero on success. On error, or when no id remains, -1 790 * is returned and *errno* is set appropriately. 791 * 792 * BPF_ENABLE_STATS 793 * Description 794 * Enable eBPF runtime statistics gathering. 795 * 796 * Runtime statistics gathering for the eBPF runtime is disabled 797 * by default to minimize the corresponding performance overhead. 798 * This command enables statistics globally. 799 * 800 * Multiple programs may independently enable statistics. 801 * After gathering the desired statistics, eBPF runtime statistics 802 * may be disabled again by calling **close**\ (2) for the file 803 * descriptor returned by this function. Statistics will only be 804 * disabled system-wide when all outstanding file descriptors 805 * returned by prior calls for this subcommand are closed. 806 * 807 * Return 808 * A new file descriptor (a nonnegative integer), or -1 if an 809 * error occurred (in which case, *errno* is set appropriately). 810 * 811 * BPF_ITER_CREATE 812 * Description 813 * Create an iterator on top of the specified *link_fd* (as 814 * previously created using **BPF_LINK_CREATE**) and return a 815 * file descriptor that can be used to trigger the iteration. 816 * 817 * If the resulting file descriptor is pinned to the filesystem 818 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 819 * for that path will trigger the iterator to read kernel state 820 * using the eBPF program attached to *link_fd*. 821 * 822 * Return 823 * A new file descriptor (a nonnegative integer), or -1 if an 824 * error occurred (in which case, *errno* is set appropriately). 825 * 826 * BPF_LINK_DETACH 827 * Description 828 * Forcefully detach the specified *link_fd* from its 829 * corresponding attachment point. 830 * 831 * Return 832 * Returns zero on success. On error, -1 is returned and *errno* 833 * is set appropriately. 834 * 835 * BPF_PROG_BIND_MAP 836 * Description 837 * Bind a map to the lifetime of an eBPF program. 838 * 839 * The map identified by *map_fd* is bound to the program 840 * identified by *prog_fd* and only released when *prog_fd* is 841 * released. This may be used in cases where metadata should be 842 * associated with a program which otherwise does not contain any 843 * references to the map (for example, embedded in the eBPF 844 * program instructions). 845 * 846 * Return 847 * Returns zero on success. On error, -1 is returned and *errno* 848 * is set appropriately. 849 * 850 * BPF_TOKEN_CREATE 851 * Description 852 * Create BPF token with embedded information about what 853 * BPF-related functionality it allows: 854 * - a set of allowed bpf() syscall commands; 855 * - a set of allowed BPF map types to be created with 856 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed; 857 * - a set of allowed BPF program types and BPF program attach 858 * types to be loaded with BPF_PROG_LOAD command, if 859 * BPF_PROG_LOAD itself is allowed. 860 * 861 * BPF token is created (derived) from an instance of BPF FS, 862 * assuming it has necessary delegation mount options specified. 863 * This BPF token can be passed as an extra parameter to various 864 * bpf() syscall commands to grant BPF subsystem functionality to 865 * unprivileged processes. 866 * 867 * When created, BPF token is "associated" with the owning 868 * user namespace of BPF FS instance (super block) that it was 869 * derived from, and subsequent BPF operations performed with 870 * BPF token would be performing capabilities checks (i.e., 871 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within 872 * that user namespace. Without BPF token, such capabilities 873 * have to be granted in init user namespace, making bpf() 874 * syscall incompatible with user namespace, for the most part. 875 * 876 * Return 877 * A new file descriptor (a nonnegative integer), or -1 if an 878 * error occurred (in which case, *errno* is set appropriately). 879 * 880 * NOTES 881 * eBPF objects (maps and programs) can be shared between processes. 882 * 883 * * After **fork**\ (2), the child inherits file descriptors 884 * referring to the same eBPF objects. 885 * * File descriptors referring to eBPF objects can be transferred over 886 * **unix**\ (7) domain sockets. 887 * * File descriptors referring to eBPF objects can be duplicated in the 888 * usual way, using **dup**\ (2) and similar calls. 889 * * File descriptors referring to eBPF objects can be pinned to the 890 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 891 * 892 * An eBPF object is deallocated only after all file descriptors referring 893 * to the object have been closed and no references remain pinned to the 894 * filesystem or attached (for example, bound to a program or device). 895 */ 896 enum bpf_cmd { 897 BPF_MAP_CREATE, 898 BPF_MAP_LOOKUP_ELEM, 899 BPF_MAP_UPDATE_ELEM, 900 BPF_MAP_DELETE_ELEM, 901 BPF_MAP_GET_NEXT_KEY, 902 BPF_PROG_LOAD, 903 BPF_OBJ_PIN, 904 BPF_OBJ_GET, 905 BPF_PROG_ATTACH, 906 BPF_PROG_DETACH, 907 BPF_PROG_TEST_RUN, 908 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 909 BPF_PROG_GET_NEXT_ID, 910 BPF_MAP_GET_NEXT_ID, 911 BPF_PROG_GET_FD_BY_ID, 912 BPF_MAP_GET_FD_BY_ID, 913 BPF_OBJ_GET_INFO_BY_FD, 914 BPF_PROG_QUERY, 915 BPF_RAW_TRACEPOINT_OPEN, 916 BPF_BTF_LOAD, 917 BPF_BTF_GET_FD_BY_ID, 918 BPF_TASK_FD_QUERY, 919 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 920 BPF_MAP_FREEZE, 921 BPF_BTF_GET_NEXT_ID, 922 BPF_MAP_LOOKUP_BATCH, 923 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 924 BPF_MAP_UPDATE_BATCH, 925 BPF_MAP_DELETE_BATCH, 926 BPF_LINK_CREATE, 927 BPF_LINK_UPDATE, 928 BPF_LINK_GET_FD_BY_ID, 929 BPF_LINK_GET_NEXT_ID, 930 BPF_ENABLE_STATS, 931 BPF_ITER_CREATE, 932 BPF_LINK_DETACH, 933 BPF_PROG_BIND_MAP, 934 BPF_TOKEN_CREATE, 935 __MAX_BPF_CMD, 936 }; 937 938 enum bpf_map_type { 939 BPF_MAP_TYPE_UNSPEC, 940 BPF_MAP_TYPE_HASH, 941 BPF_MAP_TYPE_ARRAY, 942 BPF_MAP_TYPE_PROG_ARRAY, 943 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 944 BPF_MAP_TYPE_PERCPU_HASH, 945 BPF_MAP_TYPE_PERCPU_ARRAY, 946 BPF_MAP_TYPE_STACK_TRACE, 947 BPF_MAP_TYPE_CGROUP_ARRAY, 948 BPF_MAP_TYPE_LRU_HASH, 949 BPF_MAP_TYPE_LRU_PERCPU_HASH, 950 BPF_MAP_TYPE_LPM_TRIE, 951 BPF_MAP_TYPE_ARRAY_OF_MAPS, 952 BPF_MAP_TYPE_HASH_OF_MAPS, 953 BPF_MAP_TYPE_DEVMAP, 954 BPF_MAP_TYPE_SOCKMAP, 955 BPF_MAP_TYPE_CPUMAP, 956 BPF_MAP_TYPE_XSKMAP, 957 BPF_MAP_TYPE_SOCKHASH, 958 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 959 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching 960 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to 961 * both cgroup-attached and other progs and supports all functionality 962 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark 963 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated. 964 */ 965 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED, 966 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 967 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 968 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs 969 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE + 970 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 971 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE 972 * deprecated. 973 */ 974 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED, 975 BPF_MAP_TYPE_QUEUE, 976 BPF_MAP_TYPE_STACK, 977 BPF_MAP_TYPE_SK_STORAGE, 978 BPF_MAP_TYPE_DEVMAP_HASH, 979 BPF_MAP_TYPE_STRUCT_OPS, 980 BPF_MAP_TYPE_RINGBUF, 981 BPF_MAP_TYPE_INODE_STORAGE, 982 BPF_MAP_TYPE_TASK_STORAGE, 983 BPF_MAP_TYPE_BLOOM_FILTER, 984 BPF_MAP_TYPE_USER_RINGBUF, 985 BPF_MAP_TYPE_CGRP_STORAGE, 986 __MAX_BPF_MAP_TYPE 987 }; 988 989 /* Note that tracing related programs such as 990 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 991 * are not subject to a stable API since kernel internal data 992 * structures can change from release to release and may 993 * therefore break existing tracing BPF programs. Tracing BPF 994 * programs correspond to /a/ specific kernel which is to be 995 * analyzed, and not /a/ specific kernel /and/ all future ones. 996 */ 997 enum bpf_prog_type { 998 BPF_PROG_TYPE_UNSPEC, 999 BPF_PROG_TYPE_SOCKET_FILTER, 1000 BPF_PROG_TYPE_KPROBE, 1001 BPF_PROG_TYPE_SCHED_CLS, 1002 BPF_PROG_TYPE_SCHED_ACT, 1003 BPF_PROG_TYPE_TRACEPOINT, 1004 BPF_PROG_TYPE_XDP, 1005 BPF_PROG_TYPE_PERF_EVENT, 1006 BPF_PROG_TYPE_CGROUP_SKB, 1007 BPF_PROG_TYPE_CGROUP_SOCK, 1008 BPF_PROG_TYPE_LWT_IN, 1009 BPF_PROG_TYPE_LWT_OUT, 1010 BPF_PROG_TYPE_LWT_XMIT, 1011 BPF_PROG_TYPE_SOCK_OPS, 1012 BPF_PROG_TYPE_SK_SKB, 1013 BPF_PROG_TYPE_CGROUP_DEVICE, 1014 BPF_PROG_TYPE_SK_MSG, 1015 BPF_PROG_TYPE_RAW_TRACEPOINT, 1016 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 1017 BPF_PROG_TYPE_LWT_SEG6LOCAL, 1018 BPF_PROG_TYPE_LIRC_MODE2, 1019 BPF_PROG_TYPE_SK_REUSEPORT, 1020 BPF_PROG_TYPE_FLOW_DISSECTOR, 1021 BPF_PROG_TYPE_CGROUP_SYSCTL, 1022 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1023 BPF_PROG_TYPE_CGROUP_SOCKOPT, 1024 BPF_PROG_TYPE_TRACING, 1025 BPF_PROG_TYPE_STRUCT_OPS, 1026 BPF_PROG_TYPE_EXT, 1027 BPF_PROG_TYPE_LSM, 1028 BPF_PROG_TYPE_SK_LOOKUP, 1029 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 1030 BPF_PROG_TYPE_NETFILTER, 1031 __MAX_BPF_PROG_TYPE 1032 }; 1033 1034 enum bpf_attach_type { 1035 BPF_CGROUP_INET_INGRESS, 1036 BPF_CGROUP_INET_EGRESS, 1037 BPF_CGROUP_INET_SOCK_CREATE, 1038 BPF_CGROUP_SOCK_OPS, 1039 BPF_SK_SKB_STREAM_PARSER, 1040 BPF_SK_SKB_STREAM_VERDICT, 1041 BPF_CGROUP_DEVICE, 1042 BPF_SK_MSG_VERDICT, 1043 BPF_CGROUP_INET4_BIND, 1044 BPF_CGROUP_INET6_BIND, 1045 BPF_CGROUP_INET4_CONNECT, 1046 BPF_CGROUP_INET6_CONNECT, 1047 BPF_CGROUP_INET4_POST_BIND, 1048 BPF_CGROUP_INET6_POST_BIND, 1049 BPF_CGROUP_UDP4_SENDMSG, 1050 BPF_CGROUP_UDP6_SENDMSG, 1051 BPF_LIRC_MODE2, 1052 BPF_FLOW_DISSECTOR, 1053 BPF_CGROUP_SYSCTL, 1054 BPF_CGROUP_UDP4_RECVMSG, 1055 BPF_CGROUP_UDP6_RECVMSG, 1056 BPF_CGROUP_GETSOCKOPT, 1057 BPF_CGROUP_SETSOCKOPT, 1058 BPF_TRACE_RAW_TP, 1059 BPF_TRACE_FENTRY, 1060 BPF_TRACE_FEXIT, 1061 BPF_MODIFY_RETURN, 1062 BPF_LSM_MAC, 1063 BPF_TRACE_ITER, 1064 BPF_CGROUP_INET4_GETPEERNAME, 1065 BPF_CGROUP_INET6_GETPEERNAME, 1066 BPF_CGROUP_INET4_GETSOCKNAME, 1067 BPF_CGROUP_INET6_GETSOCKNAME, 1068 BPF_XDP_DEVMAP, 1069 BPF_CGROUP_INET_SOCK_RELEASE, 1070 BPF_XDP_CPUMAP, 1071 BPF_SK_LOOKUP, 1072 BPF_XDP, 1073 BPF_SK_SKB_VERDICT, 1074 BPF_SK_REUSEPORT_SELECT, 1075 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE, 1076 BPF_PERF_EVENT, 1077 BPF_TRACE_KPROBE_MULTI, 1078 BPF_LSM_CGROUP, 1079 BPF_STRUCT_OPS, 1080 BPF_NETFILTER, 1081 BPF_TCX_INGRESS, 1082 BPF_TCX_EGRESS, 1083 BPF_TRACE_UPROBE_MULTI, 1084 BPF_CGROUP_UNIX_CONNECT, 1085 BPF_CGROUP_UNIX_SENDMSG, 1086 BPF_CGROUP_UNIX_RECVMSG, 1087 BPF_CGROUP_UNIX_GETPEERNAME, 1088 BPF_CGROUP_UNIX_GETSOCKNAME, 1089 BPF_NETKIT_PRIMARY, 1090 BPF_NETKIT_PEER, 1091 __MAX_BPF_ATTACH_TYPE 1092 }; 1093 1094 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1095 1096 enum bpf_link_type { 1097 BPF_LINK_TYPE_UNSPEC = 0, 1098 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 1099 BPF_LINK_TYPE_TRACING = 2, 1100 BPF_LINK_TYPE_CGROUP = 3, 1101 BPF_LINK_TYPE_ITER = 4, 1102 BPF_LINK_TYPE_NETNS = 5, 1103 BPF_LINK_TYPE_XDP = 6, 1104 BPF_LINK_TYPE_PERF_EVENT = 7, 1105 BPF_LINK_TYPE_KPROBE_MULTI = 8, 1106 BPF_LINK_TYPE_STRUCT_OPS = 9, 1107 BPF_LINK_TYPE_NETFILTER = 10, 1108 BPF_LINK_TYPE_TCX = 11, 1109 BPF_LINK_TYPE_UPROBE_MULTI = 12, 1110 BPF_LINK_TYPE_NETKIT = 13, 1111 __MAX_BPF_LINK_TYPE, 1112 }; 1113 1114 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE 1115 1116 enum bpf_perf_event_type { 1117 BPF_PERF_EVENT_UNSPEC = 0, 1118 BPF_PERF_EVENT_UPROBE = 1, 1119 BPF_PERF_EVENT_URETPROBE = 2, 1120 BPF_PERF_EVENT_KPROBE = 3, 1121 BPF_PERF_EVENT_KRETPROBE = 4, 1122 BPF_PERF_EVENT_TRACEPOINT = 5, 1123 BPF_PERF_EVENT_EVENT = 6, 1124 }; 1125 1126 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1127 * 1128 * NONE(default): No further bpf programs allowed in the subtree. 1129 * 1130 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1131 * the program in this cgroup yields to sub-cgroup program. 1132 * 1133 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1134 * that cgroup program gets run in addition to the program in this cgroup. 1135 * 1136 * Only one program is allowed to be attached to a cgroup with 1137 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1138 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1139 * release old program and attach the new one. Attach flags has to match. 1140 * 1141 * Multiple programs are allowed to be attached to a cgroup with 1142 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1143 * (those that were attached first, run first) 1144 * The programs of sub-cgroup are executed first, then programs of 1145 * this cgroup and then programs of parent cgroup. 1146 * When children program makes decision (like picking TCP CA or sock bind) 1147 * parent program has a chance to override it. 1148 * 1149 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1150 * programs for a cgroup. Though it's possible to replace an old program at 1151 * any position by also specifying BPF_F_REPLACE flag and position itself in 1152 * replace_bpf_fd attribute. Old program at this position will be released. 1153 * 1154 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1155 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1156 * Ex1: 1157 * cgrp1 (MULTI progs A, B) -> 1158 * cgrp2 (OVERRIDE prog C) -> 1159 * cgrp3 (MULTI prog D) -> 1160 * cgrp4 (OVERRIDE prog E) -> 1161 * cgrp5 (NONE prog F) 1162 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1163 * if prog F is detached, the execution is E,D,A,B 1164 * if prog F and D are detached, the execution is E,A,B 1165 * if prog F, E and D are detached, the execution is C,A,B 1166 * 1167 * All eligible programs are executed regardless of return code from 1168 * earlier programs. 1169 */ 1170 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1171 #define BPF_F_ALLOW_MULTI (1U << 1) 1172 /* Generic attachment flags. */ 1173 #define BPF_F_REPLACE (1U << 2) 1174 #define BPF_F_BEFORE (1U << 3) 1175 #define BPF_F_AFTER (1U << 4) 1176 #define BPF_F_ID (1U << 5) 1177 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */ 1178 1179 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1180 * verifier will perform strict alignment checking as if the kernel 1181 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1182 * and NET_IP_ALIGN defined to 2. 1183 */ 1184 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1185 1186 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the 1187 * verifier will allow any alignment whatsoever. On platforms 1188 * with strict alignment requirements for loads ands stores (such 1189 * as sparc and mips) the verifier validates that all loads and 1190 * stores provably follow this requirement. This flag turns that 1191 * checking and enforcement off. 1192 * 1193 * It is mostly used for testing when we want to validate the 1194 * context and memory access aspects of the verifier, but because 1195 * of an unaligned access the alignment check would trigger before 1196 * the one we are interested in. 1197 */ 1198 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1199 1200 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1201 * Verifier does sub-register def/use analysis and identifies instructions whose 1202 * def only matters for low 32-bit, high 32-bit is never referenced later 1203 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1204 * that it is safe to ignore clearing high 32-bit for these instructions. This 1205 * saves some back-ends a lot of code-gen. However such optimization is not 1206 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1207 * hence hasn't used verifier's analysis result. But, we really want to have a 1208 * way to be able to verify the correctness of the described optimization on 1209 * x86_64 on which testsuites are frequently exercised. 1210 * 1211 * So, this flag is introduced. Once it is set, verifier will randomize high 1212 * 32-bit for those instructions who has been identified as safe to ignore them. 1213 * Then, if verifier is not doing correct analysis, such randomization will 1214 * regress tests to expose bugs. 1215 */ 1216 #define BPF_F_TEST_RND_HI32 (1U << 2) 1217 1218 /* The verifier internal test flag. Behavior is undefined */ 1219 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1220 1221 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1222 * restrict map and helper usage for such programs. Sleepable BPF programs can 1223 * only be attached to hooks where kernel execution context allows sleeping. 1224 * Such programs are allowed to use helpers that may sleep like 1225 * bpf_copy_from_user(). 1226 */ 1227 #define BPF_F_SLEEPABLE (1U << 4) 1228 1229 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program 1230 * fully support xdp frags. 1231 */ 1232 #define BPF_F_XDP_HAS_FRAGS (1U << 5) 1233 1234 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded 1235 * program becomes device-bound but can access XDP metadata. 1236 */ 1237 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6) 1238 1239 /* The verifier internal test flag. Behavior is undefined */ 1240 #define BPF_F_TEST_REG_INVARIANTS (1U << 7) 1241 1242 /* link_create.kprobe_multi.flags used in LINK_CREATE command for 1243 * BPF_TRACE_KPROBE_MULTI attach type to create return probe. 1244 */ 1245 enum { 1246 BPF_F_KPROBE_MULTI_RETURN = (1U << 0) 1247 }; 1248 1249 /* link_create.uprobe_multi.flags used in LINK_CREATE command for 1250 * BPF_TRACE_UPROBE_MULTI attach type to create return probe. 1251 */ 1252 enum { 1253 BPF_F_UPROBE_MULTI_RETURN = (1U << 0) 1254 }; 1255 1256 /* link_create.netfilter.flags used in LINK_CREATE command for 1257 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation. 1258 */ 1259 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0) 1260 1261 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1262 * the following extensions: 1263 * 1264 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1265 * insn[0].imm: map fd or fd_idx 1266 * insn[1].imm: 0 1267 * insn[0].off: 0 1268 * insn[1].off: 0 1269 * ldimm64 rewrite: address of map 1270 * verifier type: CONST_PTR_TO_MAP 1271 */ 1272 #define BPF_PSEUDO_MAP_FD 1 1273 #define BPF_PSEUDO_MAP_IDX 5 1274 1275 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1276 * insn[0].imm: map fd or fd_idx 1277 * insn[1].imm: offset into value 1278 * insn[0].off: 0 1279 * insn[1].off: 0 1280 * ldimm64 rewrite: address of map[0]+offset 1281 * verifier type: PTR_TO_MAP_VALUE 1282 */ 1283 #define BPF_PSEUDO_MAP_VALUE 2 1284 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1285 1286 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1287 * insn[0].imm: kernel btd id of VAR 1288 * insn[1].imm: 0 1289 * insn[0].off: 0 1290 * insn[1].off: 0 1291 * ldimm64 rewrite: address of the kernel variable 1292 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1293 * is struct/union. 1294 */ 1295 #define BPF_PSEUDO_BTF_ID 3 1296 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1297 * insn[0].imm: insn offset to the func 1298 * insn[1].imm: 0 1299 * insn[0].off: 0 1300 * insn[1].off: 0 1301 * ldimm64 rewrite: address of the function 1302 * verifier type: PTR_TO_FUNC. 1303 */ 1304 #define BPF_PSEUDO_FUNC 4 1305 1306 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1307 * offset to another bpf function 1308 */ 1309 #define BPF_PSEUDO_CALL 1 1310 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1311 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1312 */ 1313 #define BPF_PSEUDO_KFUNC_CALL 2 1314 1315 /* flags for BPF_MAP_UPDATE_ELEM command */ 1316 enum { 1317 BPF_ANY = 0, /* create new element or update existing */ 1318 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1319 BPF_EXIST = 2, /* update existing element */ 1320 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1321 }; 1322 1323 /* flags for BPF_MAP_CREATE command */ 1324 enum { 1325 BPF_F_NO_PREALLOC = (1U << 0), 1326 /* Instead of having one common LRU list in the 1327 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1328 * which can scale and perform better. 1329 * Note, the LRU nodes (including free nodes) cannot be moved 1330 * across different LRU lists. 1331 */ 1332 BPF_F_NO_COMMON_LRU = (1U << 1), 1333 /* Specify numa node during map creation */ 1334 BPF_F_NUMA_NODE = (1U << 2), 1335 1336 /* Flags for accessing BPF object from syscall side. */ 1337 BPF_F_RDONLY = (1U << 3), 1338 BPF_F_WRONLY = (1U << 4), 1339 1340 /* Flag for stack_map, store build_id+offset instead of pointer */ 1341 BPF_F_STACK_BUILD_ID = (1U << 5), 1342 1343 /* Zero-initialize hash function seed. This should only be used for testing. */ 1344 BPF_F_ZERO_SEED = (1U << 6), 1345 1346 /* Flags for accessing BPF object from program side. */ 1347 BPF_F_RDONLY_PROG = (1U << 7), 1348 BPF_F_WRONLY_PROG = (1U << 8), 1349 1350 /* Clone map from listener for newly accepted socket */ 1351 BPF_F_CLONE = (1U << 9), 1352 1353 /* Enable memory-mapping BPF map */ 1354 BPF_F_MMAPABLE = (1U << 10), 1355 1356 /* Share perf_event among processes */ 1357 BPF_F_PRESERVE_ELEMS = (1U << 11), 1358 1359 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1360 BPF_F_INNER_MAP = (1U << 12), 1361 1362 /* Create a map that will be registered/unregesitered by the backed bpf_link */ 1363 BPF_F_LINK = (1U << 13), 1364 1365 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */ 1366 BPF_F_PATH_FD = (1U << 14), 1367 1368 /* Flag for value_type_btf_obj_fd, the fd is available */ 1369 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15), 1370 1371 /* BPF token FD is passed in a corresponding command's token_fd field */ 1372 BPF_F_TOKEN_FD = (1U << 16), 1373 }; 1374 1375 /* Flags for BPF_PROG_QUERY. */ 1376 1377 /* Query effective (directly attached + inherited from ancestor cgroups) 1378 * programs that will be executed for events within a cgroup. 1379 * attach_flags with this flag are always returned 0. 1380 */ 1381 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1382 1383 /* Flags for BPF_PROG_TEST_RUN */ 1384 1385 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1386 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1387 /* If set, XDP frames will be transmitted after processing */ 1388 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1) 1389 1390 /* type for BPF_ENABLE_STATS */ 1391 enum bpf_stats_type { 1392 /* enabled run_time_ns and run_cnt */ 1393 BPF_STATS_RUN_TIME = 0, 1394 }; 1395 1396 enum bpf_stack_build_id_status { 1397 /* user space need an empty entry to identify end of a trace */ 1398 BPF_STACK_BUILD_ID_EMPTY = 0, 1399 /* with valid build_id and offset */ 1400 BPF_STACK_BUILD_ID_VALID = 1, 1401 /* couldn't get build_id, fallback to ip */ 1402 BPF_STACK_BUILD_ID_IP = 2, 1403 }; 1404 1405 #define BPF_BUILD_ID_SIZE 20 1406 struct bpf_stack_build_id { 1407 __s32 status; 1408 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1409 union { 1410 __u64 offset; 1411 __u64 ip; 1412 }; 1413 }; 1414 1415 #define BPF_OBJ_NAME_LEN 16U 1416 1417 union bpf_attr { 1418 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1419 __u32 map_type; /* one of enum bpf_map_type */ 1420 __u32 key_size; /* size of key in bytes */ 1421 __u32 value_size; /* size of value in bytes */ 1422 __u32 max_entries; /* max number of entries in a map */ 1423 __u32 map_flags; /* BPF_MAP_CREATE related 1424 * flags defined above. 1425 */ 1426 __u32 inner_map_fd; /* fd pointing to the inner map */ 1427 __u32 numa_node; /* numa node (effective only if 1428 * BPF_F_NUMA_NODE is set). 1429 */ 1430 char map_name[BPF_OBJ_NAME_LEN]; 1431 __u32 map_ifindex; /* ifindex of netdev to create on */ 1432 __u32 btf_fd; /* fd pointing to a BTF type data */ 1433 __u32 btf_key_type_id; /* BTF type_id of the key */ 1434 __u32 btf_value_type_id; /* BTF type_id of the value */ 1435 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1436 * struct stored as the 1437 * map value 1438 */ 1439 /* Any per-map-type extra fields 1440 * 1441 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the 1442 * number of hash functions (if 0, the bloom filter will default 1443 * to using 5 hash functions). 1444 */ 1445 __u64 map_extra; 1446 1447 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF 1448 * type data for 1449 * btf_vmlinux_value_type_id. 1450 */ 1451 /* BPF token FD to use with BPF_MAP_CREATE operation. 1452 * If provided, map_flags should have BPF_F_TOKEN_FD flag set. 1453 */ 1454 __s32 map_token_fd; 1455 }; 1456 1457 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1458 __u32 map_fd; 1459 __aligned_u64 key; 1460 union { 1461 __aligned_u64 value; 1462 __aligned_u64 next_key; 1463 }; 1464 __u64 flags; 1465 }; 1466 1467 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1468 __aligned_u64 in_batch; /* start batch, 1469 * NULL to start from beginning 1470 */ 1471 __aligned_u64 out_batch; /* output: next start batch */ 1472 __aligned_u64 keys; 1473 __aligned_u64 values; 1474 __u32 count; /* input/output: 1475 * input: # of key/value 1476 * elements 1477 * output: # of filled elements 1478 */ 1479 __u32 map_fd; 1480 __u64 elem_flags; 1481 __u64 flags; 1482 } batch; 1483 1484 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1485 __u32 prog_type; /* one of enum bpf_prog_type */ 1486 __u32 insn_cnt; 1487 __aligned_u64 insns; 1488 __aligned_u64 license; 1489 __u32 log_level; /* verbosity level of verifier */ 1490 __u32 log_size; /* size of user buffer */ 1491 __aligned_u64 log_buf; /* user supplied buffer */ 1492 __u32 kern_version; /* not used */ 1493 __u32 prog_flags; 1494 char prog_name[BPF_OBJ_NAME_LEN]; 1495 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1496 /* For some prog types expected attach type must be known at 1497 * load time to verify attach type specific parts of prog 1498 * (context accesses, allowed helpers, etc). 1499 */ 1500 __u32 expected_attach_type; 1501 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1502 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1503 __aligned_u64 func_info; /* func info */ 1504 __u32 func_info_cnt; /* number of bpf_func_info records */ 1505 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1506 __aligned_u64 line_info; /* line info */ 1507 __u32 line_info_cnt; /* number of bpf_line_info records */ 1508 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1509 union { 1510 /* valid prog_fd to attach to bpf prog */ 1511 __u32 attach_prog_fd; 1512 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1513 __u32 attach_btf_obj_fd; 1514 }; 1515 __u32 core_relo_cnt; /* number of bpf_core_relo */ 1516 __aligned_u64 fd_array; /* array of FDs */ 1517 __aligned_u64 core_relos; 1518 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */ 1519 /* output: actual total log contents size (including termintaing zero). 1520 * It could be both larger than original log_size (if log was 1521 * truncated), or smaller (if log buffer wasn't filled completely). 1522 */ 1523 __u32 log_true_size; 1524 /* BPF token FD to use with BPF_PROG_LOAD operation. 1525 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set. 1526 */ 1527 __s32 prog_token_fd; 1528 }; 1529 1530 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1531 __aligned_u64 pathname; 1532 __u32 bpf_fd; 1533 __u32 file_flags; 1534 /* Same as dirfd in openat() syscall; see openat(2) 1535 * manpage for details of path FD and pathname semantics; 1536 * path_fd should accompanied by BPF_F_PATH_FD flag set in 1537 * file_flags field, otherwise it should be set to zero; 1538 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed. 1539 */ 1540 __s32 path_fd; 1541 }; 1542 1543 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1544 union { 1545 __u32 target_fd; /* target object to attach to or ... */ 1546 __u32 target_ifindex; /* target ifindex */ 1547 }; 1548 __u32 attach_bpf_fd; 1549 __u32 attach_type; 1550 __u32 attach_flags; 1551 __u32 replace_bpf_fd; 1552 union { 1553 __u32 relative_fd; 1554 __u32 relative_id; 1555 }; 1556 __u64 expected_revision; 1557 }; 1558 1559 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1560 __u32 prog_fd; 1561 __u32 retval; 1562 __u32 data_size_in; /* input: len of data_in */ 1563 __u32 data_size_out; /* input/output: len of data_out 1564 * returns ENOSPC if data_out 1565 * is too small. 1566 */ 1567 __aligned_u64 data_in; 1568 __aligned_u64 data_out; 1569 __u32 repeat; 1570 __u32 duration; 1571 __u32 ctx_size_in; /* input: len of ctx_in */ 1572 __u32 ctx_size_out; /* input/output: len of ctx_out 1573 * returns ENOSPC if ctx_out 1574 * is too small. 1575 */ 1576 __aligned_u64 ctx_in; 1577 __aligned_u64 ctx_out; 1578 __u32 flags; 1579 __u32 cpu; 1580 __u32 batch_size; 1581 } test; 1582 1583 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1584 union { 1585 __u32 start_id; 1586 __u32 prog_id; 1587 __u32 map_id; 1588 __u32 btf_id; 1589 __u32 link_id; 1590 }; 1591 __u32 next_id; 1592 __u32 open_flags; 1593 }; 1594 1595 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1596 __u32 bpf_fd; 1597 __u32 info_len; 1598 __aligned_u64 info; 1599 } info; 1600 1601 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1602 union { 1603 __u32 target_fd; /* target object to query or ... */ 1604 __u32 target_ifindex; /* target ifindex */ 1605 }; 1606 __u32 attach_type; 1607 __u32 query_flags; 1608 __u32 attach_flags; 1609 __aligned_u64 prog_ids; 1610 union { 1611 __u32 prog_cnt; 1612 __u32 count; 1613 }; 1614 __u32 :32; 1615 /* output: per-program attach_flags. 1616 * not allowed to be set during effective query. 1617 */ 1618 __aligned_u64 prog_attach_flags; 1619 __aligned_u64 link_ids; 1620 __aligned_u64 link_attach_flags; 1621 __u64 revision; 1622 } query; 1623 1624 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1625 __u64 name; 1626 __u32 prog_fd; 1627 } raw_tracepoint; 1628 1629 struct { /* anonymous struct for BPF_BTF_LOAD */ 1630 __aligned_u64 btf; 1631 __aligned_u64 btf_log_buf; 1632 __u32 btf_size; 1633 __u32 btf_log_size; 1634 __u32 btf_log_level; 1635 /* output: actual total log contents size (including termintaing zero). 1636 * It could be both larger than original log_size (if log was 1637 * truncated), or smaller (if log buffer wasn't filled completely). 1638 */ 1639 __u32 btf_log_true_size; 1640 __u32 btf_flags; 1641 /* BPF token FD to use with BPF_BTF_LOAD operation. 1642 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set. 1643 */ 1644 __s32 btf_token_fd; 1645 }; 1646 1647 struct { 1648 __u32 pid; /* input: pid */ 1649 __u32 fd; /* input: fd */ 1650 __u32 flags; /* input: flags */ 1651 __u32 buf_len; /* input/output: buf len */ 1652 __aligned_u64 buf; /* input/output: 1653 * tp_name for tracepoint 1654 * symbol for kprobe 1655 * filename for uprobe 1656 */ 1657 __u32 prog_id; /* output: prod_id */ 1658 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1659 __u64 probe_offset; /* output: probe_offset */ 1660 __u64 probe_addr; /* output: probe_addr */ 1661 } task_fd_query; 1662 1663 struct { /* struct used by BPF_LINK_CREATE command */ 1664 union { 1665 __u32 prog_fd; /* eBPF program to attach */ 1666 __u32 map_fd; /* struct_ops to attach */ 1667 }; 1668 union { 1669 __u32 target_fd; /* target object to attach to or ... */ 1670 __u32 target_ifindex; /* target ifindex */ 1671 }; 1672 __u32 attach_type; /* attach type */ 1673 __u32 flags; /* extra flags */ 1674 union { 1675 __u32 target_btf_id; /* btf_id of target to attach to */ 1676 struct { 1677 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1678 __u32 iter_info_len; /* iter_info length */ 1679 }; 1680 struct { 1681 /* black box user-provided value passed through 1682 * to BPF program at the execution time and 1683 * accessible through bpf_get_attach_cookie() BPF helper 1684 */ 1685 __u64 bpf_cookie; 1686 } perf_event; 1687 struct { 1688 __u32 flags; 1689 __u32 cnt; 1690 __aligned_u64 syms; 1691 __aligned_u64 addrs; 1692 __aligned_u64 cookies; 1693 } kprobe_multi; 1694 struct { 1695 /* this is overlaid with the target_btf_id above. */ 1696 __u32 target_btf_id; 1697 /* black box user-provided value passed through 1698 * to BPF program at the execution time and 1699 * accessible through bpf_get_attach_cookie() BPF helper 1700 */ 1701 __u64 cookie; 1702 } tracing; 1703 struct { 1704 __u32 pf; 1705 __u32 hooknum; 1706 __s32 priority; 1707 __u32 flags; 1708 } netfilter; 1709 struct { 1710 union { 1711 __u32 relative_fd; 1712 __u32 relative_id; 1713 }; 1714 __u64 expected_revision; 1715 } tcx; 1716 struct { 1717 __aligned_u64 path; 1718 __aligned_u64 offsets; 1719 __aligned_u64 ref_ctr_offsets; 1720 __aligned_u64 cookies; 1721 __u32 cnt; 1722 __u32 flags; 1723 __u32 pid; 1724 } uprobe_multi; 1725 struct { 1726 union { 1727 __u32 relative_fd; 1728 __u32 relative_id; 1729 }; 1730 __u64 expected_revision; 1731 } netkit; 1732 }; 1733 } link_create; 1734 1735 struct { /* struct used by BPF_LINK_UPDATE command */ 1736 __u32 link_fd; /* link fd */ 1737 union { 1738 /* new program fd to update link with */ 1739 __u32 new_prog_fd; 1740 /* new struct_ops map fd to update link with */ 1741 __u32 new_map_fd; 1742 }; 1743 __u32 flags; /* extra flags */ 1744 union { 1745 /* expected link's program fd; is specified only if 1746 * BPF_F_REPLACE flag is set in flags. 1747 */ 1748 __u32 old_prog_fd; 1749 /* expected link's map fd; is specified only 1750 * if BPF_F_REPLACE flag is set. 1751 */ 1752 __u32 old_map_fd; 1753 }; 1754 } link_update; 1755 1756 struct { 1757 __u32 link_fd; 1758 } link_detach; 1759 1760 struct { /* struct used by BPF_ENABLE_STATS command */ 1761 __u32 type; 1762 } enable_stats; 1763 1764 struct { /* struct used by BPF_ITER_CREATE command */ 1765 __u32 link_fd; 1766 __u32 flags; 1767 } iter_create; 1768 1769 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1770 __u32 prog_fd; 1771 __u32 map_fd; 1772 __u32 flags; /* extra flags */ 1773 } prog_bind_map; 1774 1775 struct { /* struct used by BPF_TOKEN_CREATE command */ 1776 __u32 flags; 1777 __u32 bpffs_fd; 1778 } token_create; 1779 1780 } __attribute__((aligned(8))); 1781 1782 /* The description below is an attempt at providing documentation to eBPF 1783 * developers about the multiple available eBPF helper functions. It can be 1784 * parsed and used to produce a manual page. The workflow is the following, 1785 * and requires the rst2man utility: 1786 * 1787 * $ ./scripts/bpf_doc.py \ 1788 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1789 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1790 * $ man /tmp/bpf-helpers.7 1791 * 1792 * Note that in order to produce this external documentation, some RST 1793 * formatting is used in the descriptions to get "bold" and "italics" in 1794 * manual pages. Also note that the few trailing white spaces are 1795 * intentional, removing them would break paragraphs for rst2man. 1796 * 1797 * Start of BPF helper function descriptions: 1798 * 1799 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1800 * Description 1801 * Perform a lookup in *map* for an entry associated to *key*. 1802 * Return 1803 * Map value associated to *key*, or **NULL** if no entry was 1804 * found. 1805 * 1806 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1807 * Description 1808 * Add or update the value of the entry associated to *key* in 1809 * *map* with *value*. *flags* is one of: 1810 * 1811 * **BPF_NOEXIST** 1812 * The entry for *key* must not exist in the map. 1813 * **BPF_EXIST** 1814 * The entry for *key* must already exist in the map. 1815 * **BPF_ANY** 1816 * No condition on the existence of the entry for *key*. 1817 * 1818 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1819 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1820 * elements always exist), the helper would return an error. 1821 * Return 1822 * 0 on success, or a negative error in case of failure. 1823 * 1824 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1825 * Description 1826 * Delete entry with *key* from *map*. 1827 * Return 1828 * 0 on success, or a negative error in case of failure. 1829 * 1830 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1831 * Description 1832 * For tracing programs, safely attempt to read *size* bytes from 1833 * kernel space address *unsafe_ptr* and store the data in *dst*. 1834 * 1835 * Generally, use **bpf_probe_read_user**\ () or 1836 * **bpf_probe_read_kernel**\ () instead. 1837 * Return 1838 * 0 on success, or a negative error in case of failure. 1839 * 1840 * u64 bpf_ktime_get_ns(void) 1841 * Description 1842 * Return the time elapsed since system boot, in nanoseconds. 1843 * Does not include time the system was suspended. 1844 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1845 * Return 1846 * Current *ktime*. 1847 * 1848 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1849 * Description 1850 * This helper is a "printk()-like" facility for debugging. It 1851 * prints a message defined by format *fmt* (of size *fmt_size*) 1852 * to file *\/sys/kernel/tracing/trace* from TraceFS, if 1853 * available. It can take up to three additional **u64** 1854 * arguments (as an eBPF helpers, the total number of arguments is 1855 * limited to five). 1856 * 1857 * Each time the helper is called, it appends a line to the trace. 1858 * Lines are discarded while *\/sys/kernel/tracing/trace* is 1859 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this. 1860 * The format of the trace is customizable, and the exact output 1861 * one will get depends on the options set in 1862 * *\/sys/kernel/tracing/trace_options* (see also the 1863 * *README* file under the same directory). However, it usually 1864 * defaults to something like: 1865 * 1866 * :: 1867 * 1868 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1869 * 1870 * In the above: 1871 * 1872 * * ``telnet`` is the name of the current task. 1873 * * ``470`` is the PID of the current task. 1874 * * ``001`` is the CPU number on which the task is 1875 * running. 1876 * * In ``.N..``, each character refers to a set of 1877 * options (whether irqs are enabled, scheduling 1878 * options, whether hard/softirqs are running, level of 1879 * preempt_disabled respectively). **N** means that 1880 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1881 * are set. 1882 * * ``419421.045894`` is a timestamp. 1883 * * ``0x00000001`` is a fake value used by BPF for the 1884 * instruction pointer register. 1885 * * ``<formatted msg>`` is the message formatted with 1886 * *fmt*. 1887 * 1888 * The conversion specifiers supported by *fmt* are similar, but 1889 * more limited than for printk(). They are **%d**, **%i**, 1890 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1891 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1892 * of field, padding with zeroes, etc.) is available, and the 1893 * helper will return **-EINVAL** (but print nothing) if it 1894 * encounters an unknown specifier. 1895 * 1896 * Also, note that **bpf_trace_printk**\ () is slow, and should 1897 * only be used for debugging purposes. For this reason, a notice 1898 * block (spanning several lines) is printed to kernel logs and 1899 * states that the helper should not be used "for production use" 1900 * the first time this helper is used (or more precisely, when 1901 * **trace_printk**\ () buffers are allocated). For passing values 1902 * to user space, perf events should be preferred. 1903 * Return 1904 * The number of bytes written to the buffer, or a negative error 1905 * in case of failure. 1906 * 1907 * u32 bpf_get_prandom_u32(void) 1908 * Description 1909 * Get a pseudo-random number. 1910 * 1911 * From a security point of view, this helper uses its own 1912 * pseudo-random internal state, and cannot be used to infer the 1913 * seed of other random functions in the kernel. However, it is 1914 * essential to note that the generator used by the helper is not 1915 * cryptographically secure. 1916 * Return 1917 * A random 32-bit unsigned value. 1918 * 1919 * u32 bpf_get_smp_processor_id(void) 1920 * Description 1921 * Get the SMP (symmetric multiprocessing) processor id. Note that 1922 * all programs run with migration disabled, which means that the 1923 * SMP processor id is stable during all the execution of the 1924 * program. 1925 * Return 1926 * The SMP id of the processor running the program. 1927 * 1928 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1929 * Description 1930 * Store *len* bytes from address *from* into the packet 1931 * associated to *skb*, at *offset*. *flags* are a combination of 1932 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1933 * checksum for the packet after storing the bytes) and 1934 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1935 * **->swhash** and *skb*\ **->l4hash** to 0). 1936 * 1937 * A call to this helper is susceptible to change the underlying 1938 * packet buffer. Therefore, at load time, all checks on pointers 1939 * previously done by the verifier are invalidated and must be 1940 * performed again, if the helper is used in combination with 1941 * direct packet access. 1942 * Return 1943 * 0 on success, or a negative error in case of failure. 1944 * 1945 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1946 * Description 1947 * Recompute the layer 3 (e.g. IP) checksum for the packet 1948 * associated to *skb*. Computation is incremental, so the helper 1949 * must know the former value of the header field that was 1950 * modified (*from*), the new value of this field (*to*), and the 1951 * number of bytes (2 or 4) for this field, stored in *size*. 1952 * Alternatively, it is possible to store the difference between 1953 * the previous and the new values of the header field in *to*, by 1954 * setting *from* and *size* to 0. For both methods, *offset* 1955 * indicates the location of the IP checksum within the packet. 1956 * 1957 * This helper works in combination with **bpf_csum_diff**\ (), 1958 * which does not update the checksum in-place, but offers more 1959 * flexibility and can handle sizes larger than 2 or 4 for the 1960 * checksum to update. 1961 * 1962 * A call to this helper is susceptible to change the underlying 1963 * packet buffer. Therefore, at load time, all checks on pointers 1964 * previously done by the verifier are invalidated and must be 1965 * performed again, if the helper is used in combination with 1966 * direct packet access. 1967 * Return 1968 * 0 on success, or a negative error in case of failure. 1969 * 1970 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1971 * Description 1972 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1973 * packet associated to *skb*. Computation is incremental, so the 1974 * helper must know the former value of the header field that was 1975 * modified (*from*), the new value of this field (*to*), and the 1976 * number of bytes (2 or 4) for this field, stored on the lowest 1977 * four bits of *flags*. Alternatively, it is possible to store 1978 * the difference between the previous and the new values of the 1979 * header field in *to*, by setting *from* and the four lowest 1980 * bits of *flags* to 0. For both methods, *offset* indicates the 1981 * location of the IP checksum within the packet. In addition to 1982 * the size of the field, *flags* can be added (bitwise OR) actual 1983 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1984 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1985 * for updates resulting in a null checksum the value is set to 1986 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1987 * the checksum is to be computed against a pseudo-header. 1988 * 1989 * This helper works in combination with **bpf_csum_diff**\ (), 1990 * which does not update the checksum in-place, but offers more 1991 * flexibility and can handle sizes larger than 2 or 4 for the 1992 * checksum to update. 1993 * 1994 * A call to this helper is susceptible to change the underlying 1995 * packet buffer. Therefore, at load time, all checks on pointers 1996 * previously done by the verifier are invalidated and must be 1997 * performed again, if the helper is used in combination with 1998 * direct packet access. 1999 * Return 2000 * 0 on success, or a negative error in case of failure. 2001 * 2002 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 2003 * Description 2004 * This special helper is used to trigger a "tail call", or in 2005 * other words, to jump into another eBPF program. The same stack 2006 * frame is used (but values on stack and in registers for the 2007 * caller are not accessible to the callee). This mechanism allows 2008 * for program chaining, either for raising the maximum number of 2009 * available eBPF instructions, or to execute given programs in 2010 * conditional blocks. For security reasons, there is an upper 2011 * limit to the number of successive tail calls that can be 2012 * performed. 2013 * 2014 * Upon call of this helper, the program attempts to jump into a 2015 * program referenced at index *index* in *prog_array_map*, a 2016 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 2017 * *ctx*, a pointer to the context. 2018 * 2019 * If the call succeeds, the kernel immediately runs the first 2020 * instruction of the new program. This is not a function call, 2021 * and it never returns to the previous program. If the call 2022 * fails, then the helper has no effect, and the caller continues 2023 * to run its subsequent instructions. A call can fail if the 2024 * destination program for the jump does not exist (i.e. *index* 2025 * is superior to the number of entries in *prog_array_map*), or 2026 * if the maximum number of tail calls has been reached for this 2027 * chain of programs. This limit is defined in the kernel by the 2028 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 2029 * which is currently set to 33. 2030 * Return 2031 * 0 on success, or a negative error in case of failure. 2032 * 2033 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 2034 * Description 2035 * Clone and redirect the packet associated to *skb* to another 2036 * net device of index *ifindex*. Both ingress and egress 2037 * interfaces can be used for redirection. The **BPF_F_INGRESS** 2038 * value in *flags* is used to make the distinction (ingress path 2039 * is selected if the flag is present, egress path otherwise). 2040 * This is the only flag supported for now. 2041 * 2042 * In comparison with **bpf_redirect**\ () helper, 2043 * **bpf_clone_redirect**\ () has the associated cost of 2044 * duplicating the packet buffer, but this can be executed out of 2045 * the eBPF program. Conversely, **bpf_redirect**\ () is more 2046 * efficient, but it is handled through an action code where the 2047 * redirection happens only after the eBPF program has returned. 2048 * 2049 * A call to this helper is susceptible to change the underlying 2050 * packet buffer. Therefore, at load time, all checks on pointers 2051 * previously done by the verifier are invalidated and must be 2052 * performed again, if the helper is used in combination with 2053 * direct packet access. 2054 * Return 2055 * 0 on success, or a negative error in case of failure. Positive 2056 * error indicates a potential drop or congestion in the target 2057 * device. The particular positive error codes are not defined. 2058 * 2059 * u64 bpf_get_current_pid_tgid(void) 2060 * Description 2061 * Get the current pid and tgid. 2062 * Return 2063 * A 64-bit integer containing the current tgid and pid, and 2064 * created as such: 2065 * *current_task*\ **->tgid << 32 \|** 2066 * *current_task*\ **->pid**. 2067 * 2068 * u64 bpf_get_current_uid_gid(void) 2069 * Description 2070 * Get the current uid and gid. 2071 * Return 2072 * A 64-bit integer containing the current GID and UID, and 2073 * created as such: *current_gid* **<< 32 \|** *current_uid*. 2074 * 2075 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 2076 * Description 2077 * Copy the **comm** attribute of the current task into *buf* of 2078 * *size_of_buf*. The **comm** attribute contains the name of 2079 * the executable (excluding the path) for the current task. The 2080 * *size_of_buf* must be strictly positive. On success, the 2081 * helper makes sure that the *buf* is NUL-terminated. On failure, 2082 * it is filled with zeroes. 2083 * Return 2084 * 0 on success, or a negative error in case of failure. 2085 * 2086 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 2087 * Description 2088 * Retrieve the classid for the current task, i.e. for the net_cls 2089 * cgroup to which *skb* belongs. 2090 * 2091 * This helper can be used on TC egress path, but not on ingress. 2092 * 2093 * The net_cls cgroup provides an interface to tag network packets 2094 * based on a user-provided identifier for all traffic coming from 2095 * the tasks belonging to the related cgroup. See also the related 2096 * kernel documentation, available from the Linux sources in file 2097 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 2098 * 2099 * The Linux kernel has two versions for cgroups: there are 2100 * cgroups v1 and cgroups v2. Both are available to users, who can 2101 * use a mixture of them, but note that the net_cls cgroup is for 2102 * cgroup v1 only. This makes it incompatible with BPF programs 2103 * run on cgroups, which is a cgroup-v2-only feature (a socket can 2104 * only hold data for one version of cgroups at a time). 2105 * 2106 * This helper is only available is the kernel was compiled with 2107 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 2108 * "**y**" or to "**m**". 2109 * Return 2110 * The classid, or 0 for the default unconfigured classid. 2111 * 2112 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 2113 * Description 2114 * Push a *vlan_tci* (VLAN tag control information) of protocol 2115 * *vlan_proto* to the packet associated to *skb*, then update 2116 * the checksum. Note that if *vlan_proto* is different from 2117 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 2118 * be **ETH_P_8021Q**. 2119 * 2120 * A call to this helper is susceptible to change the underlying 2121 * packet buffer. Therefore, at load time, all checks on pointers 2122 * previously done by the verifier are invalidated and must be 2123 * performed again, if the helper is used in combination with 2124 * direct packet access. 2125 * Return 2126 * 0 on success, or a negative error in case of failure. 2127 * 2128 * long bpf_skb_vlan_pop(struct sk_buff *skb) 2129 * Description 2130 * Pop a VLAN header from the packet associated to *skb*. 2131 * 2132 * A call to this helper is susceptible to change the underlying 2133 * packet buffer. Therefore, at load time, all checks on pointers 2134 * previously done by the verifier are invalidated and must be 2135 * performed again, if the helper is used in combination with 2136 * direct packet access. 2137 * Return 2138 * 0 on success, or a negative error in case of failure. 2139 * 2140 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2141 * Description 2142 * Get tunnel metadata. This helper takes a pointer *key* to an 2143 * empty **struct bpf_tunnel_key** of **size**, that will be 2144 * filled with tunnel metadata for the packet associated to *skb*. 2145 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 2146 * indicates that the tunnel is based on IPv6 protocol instead of 2147 * IPv4. 2148 * 2149 * The **struct bpf_tunnel_key** is an object that generalizes the 2150 * principal parameters used by various tunneling protocols into a 2151 * single struct. This way, it can be used to easily make a 2152 * decision based on the contents of the encapsulation header, 2153 * "summarized" in this struct. In particular, it holds the IP 2154 * address of the remote end (IPv4 or IPv6, depending on the case) 2155 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 2156 * this struct exposes the *key*\ **->tunnel_id**, which is 2157 * generally mapped to a VNI (Virtual Network Identifier), making 2158 * it programmable together with the **bpf_skb_set_tunnel_key**\ 2159 * () helper. 2160 * 2161 * Let's imagine that the following code is part of a program 2162 * attached to the TC ingress interface, on one end of a GRE 2163 * tunnel, and is supposed to filter out all messages coming from 2164 * remote ends with IPv4 address other than 10.0.0.1: 2165 * 2166 * :: 2167 * 2168 * int ret; 2169 * struct bpf_tunnel_key key = {}; 2170 * 2171 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 2172 * if (ret < 0) 2173 * return TC_ACT_SHOT; // drop packet 2174 * 2175 * if (key.remote_ipv4 != 0x0a000001) 2176 * return TC_ACT_SHOT; // drop packet 2177 * 2178 * return TC_ACT_OK; // accept packet 2179 * 2180 * This interface can also be used with all encapsulation devices 2181 * that can operate in "collect metadata" mode: instead of having 2182 * one network device per specific configuration, the "collect 2183 * metadata" mode only requires a single device where the 2184 * configuration can be extracted from this helper. 2185 * 2186 * This can be used together with various tunnels such as VXLan, 2187 * Geneve, GRE or IP in IP (IPIP). 2188 * Return 2189 * 0 on success, or a negative error in case of failure. 2190 * 2191 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 2192 * Description 2193 * Populate tunnel metadata for packet associated to *skb.* The 2194 * tunnel metadata is set to the contents of *key*, of *size*. The 2195 * *flags* can be set to a combination of the following values: 2196 * 2197 * **BPF_F_TUNINFO_IPV6** 2198 * Indicate that the tunnel is based on IPv6 protocol 2199 * instead of IPv4. 2200 * **BPF_F_ZERO_CSUM_TX** 2201 * For IPv4 packets, add a flag to tunnel metadata 2202 * indicating that checksum computation should be skipped 2203 * and checksum set to zeroes. 2204 * **BPF_F_DONT_FRAGMENT** 2205 * Add a flag to tunnel metadata indicating that the 2206 * packet should not be fragmented. 2207 * **BPF_F_SEQ_NUMBER** 2208 * Add a flag to tunnel metadata indicating that a 2209 * sequence number should be added to tunnel header before 2210 * sending the packet. This flag was added for GRE 2211 * encapsulation, but might be used with other protocols 2212 * as well in the future. 2213 * **BPF_F_NO_TUNNEL_KEY** 2214 * Add a flag to tunnel metadata indicating that no tunnel 2215 * key should be set in the resulting tunnel header. 2216 * 2217 * Here is a typical usage on the transmit path: 2218 * 2219 * :: 2220 * 2221 * struct bpf_tunnel_key key; 2222 * populate key ... 2223 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 2224 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 2225 * 2226 * See also the description of the **bpf_skb_get_tunnel_key**\ () 2227 * helper for additional information. 2228 * Return 2229 * 0 on success, or a negative error in case of failure. 2230 * 2231 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 2232 * Description 2233 * Read the value of a perf event counter. This helper relies on a 2234 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 2235 * the perf event counter is selected when *map* is updated with 2236 * perf event file descriptors. The *map* is an array whose size 2237 * is the number of available CPUs, and each cell contains a value 2238 * relative to one CPU. The value to retrieve is indicated by 2239 * *flags*, that contains the index of the CPU to look up, masked 2240 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2241 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2242 * current CPU should be retrieved. 2243 * 2244 * Note that before Linux 4.13, only hardware perf event can be 2245 * retrieved. 2246 * 2247 * Also, be aware that the newer helper 2248 * **bpf_perf_event_read_value**\ () is recommended over 2249 * **bpf_perf_event_read**\ () in general. The latter has some ABI 2250 * quirks where error and counter value are used as a return code 2251 * (which is wrong to do since ranges may overlap). This issue is 2252 * fixed with **bpf_perf_event_read_value**\ (), which at the same 2253 * time provides more features over the **bpf_perf_event_read**\ 2254 * () interface. Please refer to the description of 2255 * **bpf_perf_event_read_value**\ () for details. 2256 * Return 2257 * The value of the perf event counter read from the map, or a 2258 * negative error code in case of failure. 2259 * 2260 * long bpf_redirect(u32 ifindex, u64 flags) 2261 * Description 2262 * Redirect the packet to another net device of index *ifindex*. 2263 * This helper is somewhat similar to **bpf_clone_redirect**\ 2264 * (), except that the packet is not cloned, which provides 2265 * increased performance. 2266 * 2267 * Except for XDP, both ingress and egress interfaces can be used 2268 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 2269 * to make the distinction (ingress path is selected if the flag 2270 * is present, egress path otherwise). Currently, XDP only 2271 * supports redirection to the egress interface, and accepts no 2272 * flag at all. 2273 * 2274 * The same effect can also be attained with the more generic 2275 * **bpf_redirect_map**\ (), which uses a BPF map to store the 2276 * redirect target instead of providing it directly to the helper. 2277 * Return 2278 * For XDP, the helper returns **XDP_REDIRECT** on success or 2279 * **XDP_ABORTED** on error. For other program types, the values 2280 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 2281 * error. 2282 * 2283 * u32 bpf_get_route_realm(struct sk_buff *skb) 2284 * Description 2285 * Retrieve the realm or the route, that is to say the 2286 * **tclassid** field of the destination for the *skb*. The 2287 * identifier retrieved is a user-provided tag, similar to the 2288 * one used with the net_cls cgroup (see description for 2289 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 2290 * held by a route (a destination entry), not by a task. 2291 * 2292 * Retrieving this identifier works with the clsact TC egress hook 2293 * (see also **tc-bpf(8)**), or alternatively on conventional 2294 * classful egress qdiscs, but not on TC ingress path. In case of 2295 * clsact TC egress hook, this has the advantage that, internally, 2296 * the destination entry has not been dropped yet in the transmit 2297 * path. Therefore, the destination entry does not need to be 2298 * artificially held via **netif_keep_dst**\ () for a classful 2299 * qdisc until the *skb* is freed. 2300 * 2301 * This helper is available only if the kernel was compiled with 2302 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 2303 * Return 2304 * The realm of the route for the packet associated to *skb*, or 0 2305 * if none was found. 2306 * 2307 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2308 * Description 2309 * Write raw *data* blob into a special BPF perf event held by 2310 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2311 * event must have the following attributes: **PERF_SAMPLE_RAW** 2312 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2313 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2314 * 2315 * The *flags* are used to indicate the index in *map* for which 2316 * the value must be put, masked with **BPF_F_INDEX_MASK**. 2317 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2318 * to indicate that the index of the current CPU core should be 2319 * used. 2320 * 2321 * The value to write, of *size*, is passed through eBPF stack and 2322 * pointed by *data*. 2323 * 2324 * The context of the program *ctx* needs also be passed to the 2325 * helper. 2326 * 2327 * On user space, a program willing to read the values needs to 2328 * call **perf_event_open**\ () on the perf event (either for 2329 * one or for all CPUs) and to store the file descriptor into the 2330 * *map*. This must be done before the eBPF program can send data 2331 * into it. An example is available in file 2332 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2333 * tree (the eBPF program counterpart is in 2334 * *samples/bpf/trace_output_kern.c*). 2335 * 2336 * **bpf_perf_event_output**\ () achieves better performance 2337 * than **bpf_trace_printk**\ () for sharing data with user 2338 * space, and is much better suitable for streaming data from eBPF 2339 * programs. 2340 * 2341 * Note that this helper is not restricted to tracing use cases 2342 * and can be used with programs attached to TC or XDP as well, 2343 * where it allows for passing data to user space listeners. Data 2344 * can be: 2345 * 2346 * * Only custom structs, 2347 * * Only the packet payload, or 2348 * * A combination of both. 2349 * Return 2350 * 0 on success, or a negative error in case of failure. 2351 * 2352 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2353 * Description 2354 * This helper was provided as an easy way to load data from a 2355 * packet. It can be used to load *len* bytes from *offset* from 2356 * the packet associated to *skb*, into the buffer pointed by 2357 * *to*. 2358 * 2359 * Since Linux 4.7, usage of this helper has mostly been replaced 2360 * by "direct packet access", enabling packet data to be 2361 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2362 * pointing respectively to the first byte of packet data and to 2363 * the byte after the last byte of packet data. However, it 2364 * remains useful if one wishes to read large quantities of data 2365 * at once from a packet into the eBPF stack. 2366 * Return 2367 * 0 on success, or a negative error in case of failure. 2368 * 2369 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2370 * Description 2371 * Walk a user or a kernel stack and return its id. To achieve 2372 * this, the helper needs *ctx*, which is a pointer to the context 2373 * on which the tracing program is executed, and a pointer to a 2374 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2375 * 2376 * The last argument, *flags*, holds the number of stack frames to 2377 * skip (from 0 to 255), masked with 2378 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2379 * a combination of the following flags: 2380 * 2381 * **BPF_F_USER_STACK** 2382 * Collect a user space stack instead of a kernel stack. 2383 * **BPF_F_FAST_STACK_CMP** 2384 * Compare stacks by hash only. 2385 * **BPF_F_REUSE_STACKID** 2386 * If two different stacks hash into the same *stackid*, 2387 * discard the old one. 2388 * 2389 * The stack id retrieved is a 32 bit long integer handle which 2390 * can be further combined with other data (including other stack 2391 * ids) and used as a key into maps. This can be useful for 2392 * generating a variety of graphs (such as flame graphs or off-cpu 2393 * graphs). 2394 * 2395 * For walking a stack, this helper is an improvement over 2396 * **bpf_probe_read**\ (), which can be used with unrolled loops 2397 * but is not efficient and consumes a lot of eBPF instructions. 2398 * Instead, **bpf_get_stackid**\ () can collect up to 2399 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2400 * this limit can be controlled with the **sysctl** program, and 2401 * that it should be manually increased in order to profile long 2402 * user stacks (such as stacks for Java programs). To do so, use: 2403 * 2404 * :: 2405 * 2406 * # sysctl kernel.perf_event_max_stack=<new value> 2407 * Return 2408 * The positive or null stack id on success, or a negative error 2409 * in case of failure. 2410 * 2411 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2412 * Description 2413 * Compute a checksum difference, from the raw buffer pointed by 2414 * *from*, of length *from_size* (that must be a multiple of 4), 2415 * towards the raw buffer pointed by *to*, of size *to_size* 2416 * (same remark). An optional *seed* can be added to the value 2417 * (this can be cascaded, the seed may come from a previous call 2418 * to the helper). 2419 * 2420 * This is flexible enough to be used in several ways: 2421 * 2422 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2423 * checksum, it can be used when pushing new data. 2424 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2425 * checksum, it can be used when removing data from a packet. 2426 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2427 * can be used to compute a diff. Note that *from_size* and 2428 * *to_size* do not need to be equal. 2429 * 2430 * This helper can be used in combination with 2431 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2432 * which one can feed in the difference computed with 2433 * **bpf_csum_diff**\ (). 2434 * Return 2435 * The checksum result, or a negative error code in case of 2436 * failure. 2437 * 2438 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2439 * Description 2440 * Retrieve tunnel options metadata for the packet associated to 2441 * *skb*, and store the raw tunnel option data to the buffer *opt* 2442 * of *size*. 2443 * 2444 * This helper can be used with encapsulation devices that can 2445 * operate in "collect metadata" mode (please refer to the related 2446 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2447 * more details). A particular example where this can be used is 2448 * in combination with the Geneve encapsulation protocol, where it 2449 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2450 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2451 * the eBPF program. This allows for full customization of these 2452 * headers. 2453 * Return 2454 * The size of the option data retrieved. 2455 * 2456 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2457 * Description 2458 * Set tunnel options metadata for the packet associated to *skb* 2459 * to the option data contained in the raw buffer *opt* of *size*. 2460 * 2461 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2462 * helper for additional information. 2463 * Return 2464 * 0 on success, or a negative error in case of failure. 2465 * 2466 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2467 * Description 2468 * Change the protocol of the *skb* to *proto*. Currently 2469 * supported are transition from IPv4 to IPv6, and from IPv6 to 2470 * IPv4. The helper takes care of the groundwork for the 2471 * transition, including resizing the socket buffer. The eBPF 2472 * program is expected to fill the new headers, if any, via 2473 * **skb_store_bytes**\ () and to recompute the checksums with 2474 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2475 * (). The main case for this helper is to perform NAT64 2476 * operations out of an eBPF program. 2477 * 2478 * Internally, the GSO type is marked as dodgy so that headers are 2479 * checked and segments are recalculated by the GSO/GRO engine. 2480 * The size for GSO target is adapted as well. 2481 * 2482 * All values for *flags* are reserved for future usage, and must 2483 * be left at zero. 2484 * 2485 * A call to this helper is susceptible to change the underlying 2486 * packet buffer. Therefore, at load time, all checks on pointers 2487 * previously done by the verifier are invalidated and must be 2488 * performed again, if the helper is used in combination with 2489 * direct packet access. 2490 * Return 2491 * 0 on success, or a negative error in case of failure. 2492 * 2493 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2494 * Description 2495 * Change the packet type for the packet associated to *skb*. This 2496 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2497 * the eBPF program does not have a write access to *skb*\ 2498 * **->pkt_type** beside this helper. Using a helper here allows 2499 * for graceful handling of errors. 2500 * 2501 * The major use case is to change incoming *skb*s to 2502 * **PACKET_HOST** in a programmatic way instead of having to 2503 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2504 * example. 2505 * 2506 * Note that *type* only allows certain values. At this time, they 2507 * are: 2508 * 2509 * **PACKET_HOST** 2510 * Packet is for us. 2511 * **PACKET_BROADCAST** 2512 * Send packet to all. 2513 * **PACKET_MULTICAST** 2514 * Send packet to group. 2515 * **PACKET_OTHERHOST** 2516 * Send packet to someone else. 2517 * Return 2518 * 0 on success, or a negative error in case of failure. 2519 * 2520 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2521 * Description 2522 * Check whether *skb* is a descendant of the cgroup2 held by 2523 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2524 * Return 2525 * The return value depends on the result of the test, and can be: 2526 * 2527 * * 0, if the *skb* failed the cgroup2 descendant test. 2528 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2529 * * A negative error code, if an error occurred. 2530 * 2531 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2532 * Description 2533 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2534 * not set, in particular if the hash was cleared due to mangling, 2535 * recompute this hash. Later accesses to the hash can be done 2536 * directly with *skb*\ **->hash**. 2537 * 2538 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2539 * prototype with **bpf_skb_change_proto**\ (), or calling 2540 * **bpf_skb_store_bytes**\ () with the 2541 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2542 * the hash and to trigger a new computation for the next call to 2543 * **bpf_get_hash_recalc**\ (). 2544 * Return 2545 * The 32-bit hash. 2546 * 2547 * u64 bpf_get_current_task(void) 2548 * Description 2549 * Get the current task. 2550 * Return 2551 * A pointer to the current task struct. 2552 * 2553 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2554 * Description 2555 * Attempt in a safe way to write *len* bytes from the buffer 2556 * *src* to *dst* in memory. It only works for threads that are in 2557 * user context, and *dst* must be a valid user space address. 2558 * 2559 * This helper should not be used to implement any kind of 2560 * security mechanism because of TOC-TOU attacks, but rather to 2561 * debug, divert, and manipulate execution of semi-cooperative 2562 * processes. 2563 * 2564 * Keep in mind that this feature is meant for experiments, and it 2565 * has a risk of crashing the system and running programs. 2566 * Therefore, when an eBPF program using this helper is attached, 2567 * a warning including PID and process name is printed to kernel 2568 * logs. 2569 * Return 2570 * 0 on success, or a negative error in case of failure. 2571 * 2572 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2573 * Description 2574 * Check whether the probe is being run is the context of a given 2575 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2576 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2577 * Return 2578 * The return value depends on the result of the test, and can be: 2579 * 2580 * * 1, if current task belongs to the cgroup2. 2581 * * 0, if current task does not belong to the cgroup2. 2582 * * A negative error code, if an error occurred. 2583 * 2584 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2585 * Description 2586 * Resize (trim or grow) the packet associated to *skb* to the 2587 * new *len*. The *flags* are reserved for future usage, and must 2588 * be left at zero. 2589 * 2590 * The basic idea is that the helper performs the needed work to 2591 * change the size of the packet, then the eBPF program rewrites 2592 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2593 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2594 * and others. This helper is a slow path utility intended for 2595 * replies with control messages. And because it is targeted for 2596 * slow path, the helper itself can afford to be slow: it 2597 * implicitly linearizes, unclones and drops offloads from the 2598 * *skb*. 2599 * 2600 * A call to this helper is susceptible to change the underlying 2601 * packet buffer. Therefore, at load time, all checks on pointers 2602 * previously done by the verifier are invalidated and must be 2603 * performed again, if the helper is used in combination with 2604 * direct packet access. 2605 * Return 2606 * 0 on success, or a negative error in case of failure. 2607 * 2608 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2609 * Description 2610 * Pull in non-linear data in case the *skb* is non-linear and not 2611 * all of *len* are part of the linear section. Make *len* bytes 2612 * from *skb* readable and writable. If a zero value is passed for 2613 * *len*, then all bytes in the linear part of *skb* will be made 2614 * readable and writable. 2615 * 2616 * This helper is only needed for reading and writing with direct 2617 * packet access. 2618 * 2619 * For direct packet access, testing that offsets to access 2620 * are within packet boundaries (test on *skb*\ **->data_end**) is 2621 * susceptible to fail if offsets are invalid, or if the requested 2622 * data is in non-linear parts of the *skb*. On failure the 2623 * program can just bail out, or in the case of a non-linear 2624 * buffer, use a helper to make the data available. The 2625 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2626 * the data. Another one consists in using **bpf_skb_pull_data** 2627 * to pull in once the non-linear parts, then retesting and 2628 * eventually access the data. 2629 * 2630 * At the same time, this also makes sure the *skb* is uncloned, 2631 * which is a necessary condition for direct write. As this needs 2632 * to be an invariant for the write part only, the verifier 2633 * detects writes and adds a prologue that is calling 2634 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2635 * the very beginning in case it is indeed cloned. 2636 * 2637 * A call to this helper is susceptible to change the underlying 2638 * packet buffer. Therefore, at load time, all checks on pointers 2639 * previously done by the verifier are invalidated and must be 2640 * performed again, if the helper is used in combination with 2641 * direct packet access. 2642 * Return 2643 * 0 on success, or a negative error in case of failure. 2644 * 2645 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2646 * Description 2647 * Add the checksum *csum* into *skb*\ **->csum** in case the 2648 * driver has supplied a checksum for the entire packet into that 2649 * field. Return an error otherwise. This helper is intended to be 2650 * used in combination with **bpf_csum_diff**\ (), in particular 2651 * when the checksum needs to be updated after data has been 2652 * written into the packet through direct packet access. 2653 * Return 2654 * The checksum on success, or a negative error code in case of 2655 * failure. 2656 * 2657 * void bpf_set_hash_invalid(struct sk_buff *skb) 2658 * Description 2659 * Invalidate the current *skb*\ **->hash**. It can be used after 2660 * mangling on headers through direct packet access, in order to 2661 * indicate that the hash is outdated and to trigger a 2662 * recalculation the next time the kernel tries to access this 2663 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2664 * Return 2665 * void. 2666 * 2667 * long bpf_get_numa_node_id(void) 2668 * Description 2669 * Return the id of the current NUMA node. The primary use case 2670 * for this helper is the selection of sockets for the local NUMA 2671 * node, when the program is attached to sockets using the 2672 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2673 * but the helper is also available to other eBPF program types, 2674 * similarly to **bpf_get_smp_processor_id**\ (). 2675 * Return 2676 * The id of current NUMA node. 2677 * 2678 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2679 * Description 2680 * Grows headroom of packet associated to *skb* and adjusts the 2681 * offset of the MAC header accordingly, adding *len* bytes of 2682 * space. It automatically extends and reallocates memory as 2683 * required. 2684 * 2685 * This helper can be used on a layer 3 *skb* to push a MAC header 2686 * for redirection into a layer 2 device. 2687 * 2688 * All values for *flags* are reserved for future usage, and must 2689 * be left at zero. 2690 * 2691 * A call to this helper is susceptible to change the underlying 2692 * packet buffer. Therefore, at load time, all checks on pointers 2693 * previously done by the verifier are invalidated and must be 2694 * performed again, if the helper is used in combination with 2695 * direct packet access. 2696 * Return 2697 * 0 on success, or a negative error in case of failure. 2698 * 2699 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2700 * Description 2701 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2702 * it is possible to use a negative value for *delta*. This helper 2703 * can be used to prepare the packet for pushing or popping 2704 * headers. 2705 * 2706 * A call to this helper is susceptible to change the underlying 2707 * packet buffer. Therefore, at load time, all checks on pointers 2708 * previously done by the verifier are invalidated and must be 2709 * performed again, if the helper is used in combination with 2710 * direct packet access. 2711 * Return 2712 * 0 on success, or a negative error in case of failure. 2713 * 2714 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2715 * Description 2716 * Copy a NUL terminated string from an unsafe kernel address 2717 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2718 * more details. 2719 * 2720 * Generally, use **bpf_probe_read_user_str**\ () or 2721 * **bpf_probe_read_kernel_str**\ () instead. 2722 * Return 2723 * On success, the strictly positive length of the string, 2724 * including the trailing NUL character. On error, a negative 2725 * value. 2726 * 2727 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2728 * Description 2729 * If the **struct sk_buff** pointed by *skb* has a known socket, 2730 * retrieve the cookie (generated by the kernel) of this socket. 2731 * If no cookie has been set yet, generate a new cookie. Once 2732 * generated, the socket cookie remains stable for the life of the 2733 * socket. This helper can be useful for monitoring per socket 2734 * networking traffic statistics as it provides a global socket 2735 * identifier that can be assumed unique. 2736 * Return 2737 * A 8-byte long unique number on success, or 0 if the socket 2738 * field is missing inside *skb*. 2739 * 2740 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2741 * Description 2742 * Equivalent to bpf_get_socket_cookie() helper that accepts 2743 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2744 * Return 2745 * A 8-byte long unique number. 2746 * 2747 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2748 * Description 2749 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2750 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2751 * Return 2752 * A 8-byte long unique number. 2753 * 2754 * u64 bpf_get_socket_cookie(struct sock *sk) 2755 * Description 2756 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2757 * *sk*, but gets socket from a BTF **struct sock**. This helper 2758 * also works for sleepable programs. 2759 * Return 2760 * A 8-byte long unique number or 0 if *sk* is NULL. 2761 * 2762 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2763 * Description 2764 * Get the owner UID of the socked associated to *skb*. 2765 * Return 2766 * The owner UID of the socket associated to *skb*. If the socket 2767 * is **NULL**, or if it is not a full socket (i.e. if it is a 2768 * time-wait or a request socket instead), **overflowuid** value 2769 * is returned (note that **overflowuid** might also be the actual 2770 * UID value for the socket). 2771 * 2772 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2773 * Description 2774 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2775 * to value *hash*. 2776 * Return 2777 * 0 2778 * 2779 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2780 * Description 2781 * Emulate a call to **setsockopt()** on the socket associated to 2782 * *bpf_socket*, which must be a full socket. The *level* at 2783 * which the option resides and the name *optname* of the option 2784 * must be specified, see **setsockopt(2)** for more information. 2785 * The option value of length *optlen* is pointed by *optval*. 2786 * 2787 * *bpf_socket* should be one of the following: 2788 * 2789 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2790 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 2791 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 2792 * 2793 * This helper actually implements a subset of **setsockopt()**. 2794 * It supports the following *level*\ s: 2795 * 2796 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2797 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2798 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2799 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**, 2800 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**. 2801 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2802 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2803 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2804 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2805 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**, 2806 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**, 2807 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**, 2808 * **TCP_BPF_RTO_MIN**. 2809 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2810 * * **IPPROTO_IPV6**, which supports the following *optname*\ s: 2811 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**. 2812 * Return 2813 * 0 on success, or a negative error in case of failure. 2814 * 2815 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2816 * Description 2817 * Grow or shrink the room for data in the packet associated to 2818 * *skb* by *len_diff*, and according to the selected *mode*. 2819 * 2820 * By default, the helper will reset any offloaded checksum 2821 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2822 * by the following flag: 2823 * 2824 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2825 * checksum data of the skb to CHECKSUM_NONE. 2826 * 2827 * There are two supported modes at this time: 2828 * 2829 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2830 * (room space is added or removed between the layer 2 and 2831 * layer 3 headers). 2832 * 2833 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2834 * (room space is added or removed between the layer 3 and 2835 * layer 4 headers). 2836 * 2837 * The following flags are supported at this time: 2838 * 2839 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2840 * Adjusting mss in this way is not allowed for datagrams. 2841 * 2842 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2843 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2844 * Any new space is reserved to hold a tunnel header. 2845 * Configure skb offsets and other fields accordingly. 2846 * 2847 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2848 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2849 * Use with ENCAP_L3 flags to further specify the tunnel type. 2850 * 2851 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2852 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2853 * type; *len* is the length of the inner MAC header. 2854 * 2855 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2856 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2857 * L2 type as Ethernet. 2858 * 2859 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**, 2860 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**: 2861 * Indicate the new IP header version after decapsulating the outer 2862 * IP header. Used when the inner and outer IP versions are different. 2863 * 2864 * A call to this helper is susceptible to change the underlying 2865 * packet buffer. Therefore, at load time, all checks on pointers 2866 * previously done by the verifier are invalidated and must be 2867 * performed again, if the helper is used in combination with 2868 * direct packet access. 2869 * Return 2870 * 0 on success, or a negative error in case of failure. 2871 * 2872 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags) 2873 * Description 2874 * Redirect the packet to the endpoint referenced by *map* at 2875 * index *key*. Depending on its type, this *map* can contain 2876 * references to net devices (for forwarding packets through other 2877 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2878 * but this is only implemented for native XDP (with driver 2879 * support) as of this writing). 2880 * 2881 * The lower two bits of *flags* are used as the return code if 2882 * the map lookup fails. This is so that the return value can be 2883 * one of the XDP program return codes up to **XDP_TX**, as chosen 2884 * by the caller. The higher bits of *flags* can be set to 2885 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below. 2886 * 2887 * With BPF_F_BROADCAST the packet will be broadcasted to all the 2888 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress 2889 * interface will be excluded when do broadcasting. 2890 * 2891 * See also **bpf_redirect**\ (), which only supports redirecting 2892 * to an ifindex, but doesn't require a map to do so. 2893 * Return 2894 * **XDP_REDIRECT** on success, or the value of the two lower bits 2895 * of the *flags* argument on error. 2896 * 2897 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2898 * Description 2899 * Redirect the packet to the socket referenced by *map* (of type 2900 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2901 * egress interfaces can be used for redirection. The 2902 * **BPF_F_INGRESS** value in *flags* is used to make the 2903 * distinction (ingress path is selected if the flag is present, 2904 * egress path otherwise). This is the only flag supported for now. 2905 * Return 2906 * **SK_PASS** on success, or **SK_DROP** on error. 2907 * 2908 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2909 * Description 2910 * Add an entry to, or update a *map* referencing sockets. The 2911 * *skops* is used as a new value for the entry associated to 2912 * *key*. *flags* is one of: 2913 * 2914 * **BPF_NOEXIST** 2915 * The entry for *key* must not exist in the map. 2916 * **BPF_EXIST** 2917 * The entry for *key* must already exist in the map. 2918 * **BPF_ANY** 2919 * No condition on the existence of the entry for *key*. 2920 * 2921 * If the *map* has eBPF programs (parser and verdict), those will 2922 * be inherited by the socket being added. If the socket is 2923 * already attached to eBPF programs, this results in an error. 2924 * Return 2925 * 0 on success, or a negative error in case of failure. 2926 * 2927 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2928 * Description 2929 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2930 * *delta* (which can be positive or negative). Note that this 2931 * operation modifies the address stored in *xdp_md*\ **->data**, 2932 * so the latter must be loaded only after the helper has been 2933 * called. 2934 * 2935 * The use of *xdp_md*\ **->data_meta** is optional and programs 2936 * are not required to use it. The rationale is that when the 2937 * packet is processed with XDP (e.g. as DoS filter), it is 2938 * possible to push further meta data along with it before passing 2939 * to the stack, and to give the guarantee that an ingress eBPF 2940 * program attached as a TC classifier on the same device can pick 2941 * this up for further post-processing. Since TC works with socket 2942 * buffers, it remains possible to set from XDP the **mark** or 2943 * **priority** pointers, or other pointers for the socket buffer. 2944 * Having this scratch space generic and programmable allows for 2945 * more flexibility as the user is free to store whatever meta 2946 * data they need. 2947 * 2948 * A call to this helper is susceptible to change the underlying 2949 * packet buffer. Therefore, at load time, all checks on pointers 2950 * previously done by the verifier are invalidated and must be 2951 * performed again, if the helper is used in combination with 2952 * direct packet access. 2953 * Return 2954 * 0 on success, or a negative error in case of failure. 2955 * 2956 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2957 * Description 2958 * Read the value of a perf event counter, and store it into *buf* 2959 * of size *buf_size*. This helper relies on a *map* of type 2960 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2961 * counter is selected when *map* is updated with perf event file 2962 * descriptors. The *map* is an array whose size is the number of 2963 * available CPUs, and each cell contains a value relative to one 2964 * CPU. The value to retrieve is indicated by *flags*, that 2965 * contains the index of the CPU to look up, masked with 2966 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2967 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2968 * current CPU should be retrieved. 2969 * 2970 * This helper behaves in a way close to 2971 * **bpf_perf_event_read**\ () helper, save that instead of 2972 * just returning the value observed, it fills the *buf* 2973 * structure. This allows for additional data to be retrieved: in 2974 * particular, the enabled and running times (in *buf*\ 2975 * **->enabled** and *buf*\ **->running**, respectively) are 2976 * copied. In general, **bpf_perf_event_read_value**\ () is 2977 * recommended over **bpf_perf_event_read**\ (), which has some 2978 * ABI issues and provides fewer functionalities. 2979 * 2980 * These values are interesting, because hardware PMU (Performance 2981 * Monitoring Unit) counters are limited resources. When there are 2982 * more PMU based perf events opened than available counters, 2983 * kernel will multiplex these events so each event gets certain 2984 * percentage (but not all) of the PMU time. In case that 2985 * multiplexing happens, the number of samples or counter value 2986 * will not reflect the case compared to when no multiplexing 2987 * occurs. This makes comparison between different runs difficult. 2988 * Typically, the counter value should be normalized before 2989 * comparing to other experiments. The usual normalization is done 2990 * as follows. 2991 * 2992 * :: 2993 * 2994 * normalized_counter = counter * t_enabled / t_running 2995 * 2996 * Where t_enabled is the time enabled for event and t_running is 2997 * the time running for event since last normalization. The 2998 * enabled and running times are accumulated since the perf event 2999 * open. To achieve scaling factor between two invocations of an 3000 * eBPF program, users can use CPU id as the key (which is 3001 * typical for perf array usage model) to remember the previous 3002 * value and do the calculation inside the eBPF program. 3003 * Return 3004 * 0 on success, or a negative error in case of failure. 3005 * 3006 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 3007 * Description 3008 * For an eBPF program attached to a perf event, retrieve the 3009 * value of the event counter associated to *ctx* and store it in 3010 * the structure pointed by *buf* and of size *buf_size*. Enabled 3011 * and running times are also stored in the structure (see 3012 * description of helper **bpf_perf_event_read_value**\ () for 3013 * more details). 3014 * Return 3015 * 0 on success, or a negative error in case of failure. 3016 * 3017 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 3018 * Description 3019 * Emulate a call to **getsockopt()** on the socket associated to 3020 * *bpf_socket*, which must be a full socket. The *level* at 3021 * which the option resides and the name *optname* of the option 3022 * must be specified, see **getsockopt(2)** for more information. 3023 * The retrieved value is stored in the structure pointed by 3024 * *opval* and of length *optlen*. 3025 * 3026 * *bpf_socket* should be one of the following: 3027 * 3028 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 3029 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**, 3030 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**. 3031 * 3032 * This helper actually implements a subset of **getsockopt()**. 3033 * It supports the same set of *optname*\ s that is supported by 3034 * the **bpf_setsockopt**\ () helper. The exceptions are 3035 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and 3036 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only. 3037 * Return 3038 * 0 on success, or a negative error in case of failure. 3039 * 3040 * long bpf_override_return(struct pt_regs *regs, u64 rc) 3041 * Description 3042 * Used for error injection, this helper uses kprobes to override 3043 * the return value of the probed function, and to set it to *rc*. 3044 * The first argument is the context *regs* on which the kprobe 3045 * works. 3046 * 3047 * This helper works by setting the PC (program counter) 3048 * to an override function which is run in place of the original 3049 * probed function. This means the probed function is not run at 3050 * all. The replacement function just returns with the required 3051 * value. 3052 * 3053 * This helper has security implications, and thus is subject to 3054 * restrictions. It is only available if the kernel was compiled 3055 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 3056 * option, and in this case it only works on functions tagged with 3057 * **ALLOW_ERROR_INJECTION** in the kernel code. 3058 * 3059 * Also, the helper is only available for the architectures having 3060 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 3061 * x86 architecture is the only one to support this feature. 3062 * Return 3063 * 0 3064 * 3065 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 3066 * Description 3067 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 3068 * for the full TCP socket associated to *bpf_sock_ops* to 3069 * *argval*. 3070 * 3071 * The primary use of this field is to determine if there should 3072 * be calls to eBPF programs of type 3073 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 3074 * code. A program of the same type can change its value, per 3075 * connection and as necessary, when the connection is 3076 * established. This field is directly accessible for reading, but 3077 * this helper must be used for updates in order to return an 3078 * error if an eBPF program tries to set a callback that is not 3079 * supported in the current kernel. 3080 * 3081 * *argval* is a flag array which can combine these flags: 3082 * 3083 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 3084 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 3085 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 3086 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 3087 * 3088 * Therefore, this function can be used to clear a callback flag by 3089 * setting the appropriate bit to zero. e.g. to disable the RTO 3090 * callback: 3091 * 3092 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 3093 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 3094 * 3095 * Here are some examples of where one could call such eBPF 3096 * program: 3097 * 3098 * * When RTO fires. 3099 * * When a packet is retransmitted. 3100 * * When the connection terminates. 3101 * * When a packet is sent. 3102 * * When a packet is received. 3103 * Return 3104 * Code **-EINVAL** if the socket is not a full TCP socket; 3105 * otherwise, a positive number containing the bits that could not 3106 * be set is returned (which comes down to 0 if all bits were set 3107 * as required). 3108 * 3109 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 3110 * Description 3111 * This helper is used in programs implementing policies at the 3112 * socket level. If the message *msg* is allowed to pass (i.e. if 3113 * the verdict eBPF program returns **SK_PASS**), redirect it to 3114 * the socket referenced by *map* (of type 3115 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 3116 * egress interfaces can be used for redirection. The 3117 * **BPF_F_INGRESS** value in *flags* is used to make the 3118 * distinction (ingress path is selected if the flag is present, 3119 * egress path otherwise). This is the only flag supported for now. 3120 * Return 3121 * **SK_PASS** on success, or **SK_DROP** on error. 3122 * 3123 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 3124 * Description 3125 * For socket policies, apply the verdict of the eBPF program to 3126 * the next *bytes* (number of bytes) of message *msg*. 3127 * 3128 * For example, this helper can be used in the following cases: 3129 * 3130 * * A single **sendmsg**\ () or **sendfile**\ () system call 3131 * contains multiple logical messages that the eBPF program is 3132 * supposed to read and for which it should apply a verdict. 3133 * * An eBPF program only cares to read the first *bytes* of a 3134 * *msg*. If the message has a large payload, then setting up 3135 * and calling the eBPF program repeatedly for all bytes, even 3136 * though the verdict is already known, would create unnecessary 3137 * overhead. 3138 * 3139 * When called from within an eBPF program, the helper sets a 3140 * counter internal to the BPF infrastructure, that is used to 3141 * apply the last verdict to the next *bytes*. If *bytes* is 3142 * smaller than the current data being processed from a 3143 * **sendmsg**\ () or **sendfile**\ () system call, the first 3144 * *bytes* will be sent and the eBPF program will be re-run with 3145 * the pointer for start of data pointing to byte number *bytes* 3146 * **+ 1**. If *bytes* is larger than the current data being 3147 * processed, then the eBPF verdict will be applied to multiple 3148 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 3149 * consumed. 3150 * 3151 * Note that if a socket closes with the internal counter holding 3152 * a non-zero value, this is not a problem because data is not 3153 * being buffered for *bytes* and is sent as it is received. 3154 * Return 3155 * 0 3156 * 3157 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 3158 * Description 3159 * For socket policies, prevent the execution of the verdict eBPF 3160 * program for message *msg* until *bytes* (byte number) have been 3161 * accumulated. 3162 * 3163 * This can be used when one needs a specific number of bytes 3164 * before a verdict can be assigned, even if the data spans 3165 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 3166 * case would be a user calling **sendmsg**\ () repeatedly with 3167 * 1-byte long message segments. Obviously, this is bad for 3168 * performance, but it is still valid. If the eBPF program needs 3169 * *bytes* bytes to validate a header, this helper can be used to 3170 * prevent the eBPF program to be called again until *bytes* have 3171 * been accumulated. 3172 * Return 3173 * 0 3174 * 3175 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 3176 * Description 3177 * For socket policies, pull in non-linear data from user space 3178 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 3179 * **->data_end** to *start* and *end* bytes offsets into *msg*, 3180 * respectively. 3181 * 3182 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3183 * *msg* it can only parse data that the (**data**, **data_end**) 3184 * pointers have already consumed. For **sendmsg**\ () hooks this 3185 * is likely the first scatterlist element. But for calls relying 3186 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 3187 * be the range (**0**, **0**) because the data is shared with 3188 * user space and by default the objective is to avoid allowing 3189 * user space to modify data while (or after) eBPF verdict is 3190 * being decided. This helper can be used to pull in data and to 3191 * set the start and end pointer to given values. Data will be 3192 * copied if necessary (i.e. if data was not linear and if start 3193 * and end pointers do not point to the same chunk). 3194 * 3195 * A call to this helper is susceptible to change the underlying 3196 * packet buffer. Therefore, at load time, all checks on pointers 3197 * previously done by the verifier are invalidated and must be 3198 * performed again, if the helper is used in combination with 3199 * direct packet access. 3200 * 3201 * All values for *flags* are reserved for future usage, and must 3202 * be left at zero. 3203 * Return 3204 * 0 on success, or a negative error in case of failure. 3205 * 3206 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 3207 * Description 3208 * Bind the socket associated to *ctx* to the address pointed by 3209 * *addr*, of length *addr_len*. This allows for making outgoing 3210 * connection from the desired IP address, which can be useful for 3211 * example when all processes inside a cgroup should use one 3212 * single IP address on a host that has multiple IP configured. 3213 * 3214 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 3215 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 3216 * **AF_INET6**). It's advised to pass zero port (**sin_port** 3217 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 3218 * behavior and lets the kernel efficiently pick up an unused 3219 * port as long as 4-tuple is unique. Passing non-zero port might 3220 * lead to degraded performance. 3221 * Return 3222 * 0 on success, or a negative error in case of failure. 3223 * 3224 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 3225 * Description 3226 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 3227 * possible to both shrink and grow the packet tail. 3228 * Shrink done via *delta* being a negative integer. 3229 * 3230 * A call to this helper is susceptible to change the underlying 3231 * packet buffer. Therefore, at load time, all checks on pointers 3232 * previously done by the verifier are invalidated and must be 3233 * performed again, if the helper is used in combination with 3234 * direct packet access. 3235 * Return 3236 * 0 on success, or a negative error in case of failure. 3237 * 3238 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 3239 * Description 3240 * Retrieve the XFRM state (IP transform framework, see also 3241 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 3242 * 3243 * The retrieved value is stored in the **struct bpf_xfrm_state** 3244 * pointed by *xfrm_state* and of length *size*. 3245 * 3246 * All values for *flags* are reserved for future usage, and must 3247 * be left at zero. 3248 * 3249 * This helper is available only if the kernel was compiled with 3250 * **CONFIG_XFRM** configuration option. 3251 * Return 3252 * 0 on success, or a negative error in case of failure. 3253 * 3254 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 3255 * Description 3256 * Return a user or a kernel stack in bpf program provided buffer. 3257 * To achieve this, the helper needs *ctx*, which is a pointer 3258 * to the context on which the tracing program is executed. 3259 * To store the stacktrace, the bpf program provides *buf* with 3260 * a nonnegative *size*. 3261 * 3262 * The last argument, *flags*, holds the number of stack frames to 3263 * skip (from 0 to 255), masked with 3264 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3265 * the following flags: 3266 * 3267 * **BPF_F_USER_STACK** 3268 * Collect a user space stack instead of a kernel stack. 3269 * **BPF_F_USER_BUILD_ID** 3270 * Collect (build_id, file_offset) instead of ips for user 3271 * stack, only valid if **BPF_F_USER_STACK** is also 3272 * specified. 3273 * 3274 * *file_offset* is an offset relative to the beginning 3275 * of the executable or shared object file backing the vma 3276 * which the *ip* falls in. It is *not* an offset relative 3277 * to that object's base address. Accordingly, it must be 3278 * adjusted by adding (sh_addr - sh_offset), where 3279 * sh_{addr,offset} correspond to the executable section 3280 * containing *file_offset* in the object, for comparisons 3281 * to symbols' st_value to be valid. 3282 * 3283 * **bpf_get_stack**\ () can collect up to 3284 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3285 * to sufficient large buffer size. Note that 3286 * this limit can be controlled with the **sysctl** program, and 3287 * that it should be manually increased in order to profile long 3288 * user stacks (such as stacks for Java programs). To do so, use: 3289 * 3290 * :: 3291 * 3292 * # sysctl kernel.perf_event_max_stack=<new value> 3293 * Return 3294 * The non-negative copied *buf* length equal to or less than 3295 * *size* on success, or a negative error in case of failure. 3296 * 3297 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 3298 * Description 3299 * This helper is similar to **bpf_skb_load_bytes**\ () in that 3300 * it provides an easy way to load *len* bytes from *offset* 3301 * from the packet associated to *skb*, into the buffer pointed 3302 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 3303 * a fifth argument *start_header* exists in order to select a 3304 * base offset to start from. *start_header* can be one of: 3305 * 3306 * **BPF_HDR_START_MAC** 3307 * Base offset to load data from is *skb*'s mac header. 3308 * **BPF_HDR_START_NET** 3309 * Base offset to load data from is *skb*'s network header. 3310 * 3311 * In general, "direct packet access" is the preferred method to 3312 * access packet data, however, this helper is in particular useful 3313 * in socket filters where *skb*\ **->data** does not always point 3314 * to the start of the mac header and where "direct packet access" 3315 * is not available. 3316 * Return 3317 * 0 on success, or a negative error in case of failure. 3318 * 3319 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 3320 * Description 3321 * Do FIB lookup in kernel tables using parameters in *params*. 3322 * If lookup is successful and result shows packet is to be 3323 * forwarded, the neighbor tables are searched for the nexthop. 3324 * If successful (ie., FIB lookup shows forwarding and nexthop 3325 * is resolved), the nexthop address is returned in ipv4_dst 3326 * or ipv6_dst based on family, smac is set to mac address of 3327 * egress device, dmac is set to nexthop mac address, rt_metric 3328 * is set to metric from route (IPv4/IPv6 only), and ifindex 3329 * is set to the device index of the nexthop from the FIB lookup. 3330 * 3331 * *plen* argument is the size of the passed in struct. 3332 * *flags* argument can be a combination of one or more of the 3333 * following values: 3334 * 3335 * **BPF_FIB_LOOKUP_DIRECT** 3336 * Do a direct table lookup vs full lookup using FIB 3337 * rules. 3338 * **BPF_FIB_LOOKUP_TBID** 3339 * Used with BPF_FIB_LOOKUP_DIRECT. 3340 * Use the routing table ID present in *params*->tbid 3341 * for the fib lookup. 3342 * **BPF_FIB_LOOKUP_OUTPUT** 3343 * Perform lookup from an egress perspective (default is 3344 * ingress). 3345 * **BPF_FIB_LOOKUP_SKIP_NEIGH** 3346 * Skip the neighbour table lookup. *params*->dmac 3347 * and *params*->smac will not be set as output. A common 3348 * use case is to call **bpf_redirect_neigh**\ () after 3349 * doing **bpf_fib_lookup**\ (). 3350 * **BPF_FIB_LOOKUP_SRC** 3351 * Derive and set source IP addr in *params*->ipv{4,6}_src 3352 * for the nexthop. If the src addr cannot be derived, 3353 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this 3354 * case, *params*->dmac and *params*->smac are not set either. 3355 * 3356 * *ctx* is either **struct xdp_md** for XDP programs or 3357 * **struct sk_buff** tc cls_act programs. 3358 * Return 3359 * * < 0 if any input argument is invalid 3360 * * 0 on success (packet is forwarded, nexthop neighbor exists) 3361 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 3362 * packet is not forwarded or needs assist from full stack 3363 * 3364 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3365 * was exceeded and output params->mtu_result contains the MTU. 3366 * 3367 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3368 * Description 3369 * Add an entry to, or update a sockhash *map* referencing sockets. 3370 * The *skops* is used as a new value for the entry associated to 3371 * *key*. *flags* is one of: 3372 * 3373 * **BPF_NOEXIST** 3374 * The entry for *key* must not exist in the map. 3375 * **BPF_EXIST** 3376 * The entry for *key* must already exist in the map. 3377 * **BPF_ANY** 3378 * No condition on the existence of the entry for *key*. 3379 * 3380 * If the *map* has eBPF programs (parser and verdict), those will 3381 * be inherited by the socket being added. If the socket is 3382 * already attached to eBPF programs, this results in an error. 3383 * Return 3384 * 0 on success, or a negative error in case of failure. 3385 * 3386 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3387 * Description 3388 * This helper is used in programs implementing policies at the 3389 * socket level. If the message *msg* is allowed to pass (i.e. if 3390 * the verdict eBPF program returns **SK_PASS**), redirect it to 3391 * the socket referenced by *map* (of type 3392 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3393 * egress interfaces can be used for redirection. The 3394 * **BPF_F_INGRESS** value in *flags* is used to make the 3395 * distinction (ingress path is selected if the flag is present, 3396 * egress path otherwise). This is the only flag supported for now. 3397 * Return 3398 * **SK_PASS** on success, or **SK_DROP** on error. 3399 * 3400 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3401 * Description 3402 * This helper is used in programs implementing policies at the 3403 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3404 * if the verdict eBPF program returns **SK_PASS**), redirect it 3405 * to the socket referenced by *map* (of type 3406 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3407 * egress interfaces can be used for redirection. The 3408 * **BPF_F_INGRESS** value in *flags* is used to make the 3409 * distinction (ingress path is selected if the flag is present, 3410 * egress otherwise). This is the only flag supported for now. 3411 * Return 3412 * **SK_PASS** on success, or **SK_DROP** on error. 3413 * 3414 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3415 * Description 3416 * Encapsulate the packet associated to *skb* within a Layer 3 3417 * protocol header. This header is provided in the buffer at 3418 * address *hdr*, with *len* its size in bytes. *type* indicates 3419 * the protocol of the header and can be one of: 3420 * 3421 * **BPF_LWT_ENCAP_SEG6** 3422 * IPv6 encapsulation with Segment Routing Header 3423 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3424 * the IPv6 header is computed by the kernel. 3425 * **BPF_LWT_ENCAP_SEG6_INLINE** 3426 * Only works if *skb* contains an IPv6 packet. Insert a 3427 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3428 * the IPv6 header. 3429 * **BPF_LWT_ENCAP_IP** 3430 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3431 * must be IPv4 or IPv6, followed by zero or more 3432 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3433 * total bytes in all prepended headers. Please note that 3434 * if **skb_is_gso**\ (*skb*) is true, no more than two 3435 * headers can be prepended, and the inner header, if 3436 * present, should be either GRE or UDP/GUE. 3437 * 3438 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3439 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3440 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3441 * **BPF_PROG_TYPE_LWT_XMIT**. 3442 * 3443 * A call to this helper is susceptible to change the underlying 3444 * packet buffer. Therefore, at load time, all checks on pointers 3445 * previously done by the verifier are invalidated and must be 3446 * performed again, if the helper is used in combination with 3447 * direct packet access. 3448 * Return 3449 * 0 on success, or a negative error in case of failure. 3450 * 3451 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3452 * Description 3453 * Store *len* bytes from address *from* into the packet 3454 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3455 * inside the outermost IPv6 Segment Routing Header can be 3456 * modified through this helper. 3457 * 3458 * A call to this helper is susceptible to change the underlying 3459 * packet buffer. Therefore, at load time, all checks on pointers 3460 * previously done by the verifier are invalidated and must be 3461 * performed again, if the helper is used in combination with 3462 * direct packet access. 3463 * Return 3464 * 0 on success, or a negative error in case of failure. 3465 * 3466 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3467 * Description 3468 * Adjust the size allocated to TLVs in the outermost IPv6 3469 * Segment Routing Header contained in the packet associated to 3470 * *skb*, at position *offset* by *delta* bytes. Only offsets 3471 * after the segments are accepted. *delta* can be as well 3472 * positive (growing) as negative (shrinking). 3473 * 3474 * A call to this helper is susceptible to change the underlying 3475 * packet buffer. Therefore, at load time, all checks on pointers 3476 * previously done by the verifier are invalidated and must be 3477 * performed again, if the helper is used in combination with 3478 * direct packet access. 3479 * Return 3480 * 0 on success, or a negative error in case of failure. 3481 * 3482 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3483 * Description 3484 * Apply an IPv6 Segment Routing action of type *action* to the 3485 * packet associated to *skb*. Each action takes a parameter 3486 * contained at address *param*, and of length *param_len* bytes. 3487 * *action* can be one of: 3488 * 3489 * **SEG6_LOCAL_ACTION_END_X** 3490 * End.X action: Endpoint with Layer-3 cross-connect. 3491 * Type of *param*: **struct in6_addr**. 3492 * **SEG6_LOCAL_ACTION_END_T** 3493 * End.T action: Endpoint with specific IPv6 table lookup. 3494 * Type of *param*: **int**. 3495 * **SEG6_LOCAL_ACTION_END_B6** 3496 * End.B6 action: Endpoint bound to an SRv6 policy. 3497 * Type of *param*: **struct ipv6_sr_hdr**. 3498 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3499 * End.B6.Encap action: Endpoint bound to an SRv6 3500 * encapsulation policy. 3501 * Type of *param*: **struct ipv6_sr_hdr**. 3502 * 3503 * A call to this helper is susceptible to change the underlying 3504 * packet buffer. Therefore, at load time, all checks on pointers 3505 * previously done by the verifier are invalidated and must be 3506 * performed again, if the helper is used in combination with 3507 * direct packet access. 3508 * Return 3509 * 0 on success, or a negative error in case of failure. 3510 * 3511 * long bpf_rc_repeat(void *ctx) 3512 * Description 3513 * This helper is used in programs implementing IR decoding, to 3514 * report a successfully decoded repeat key message. This delays 3515 * the generation of a key up event for previously generated 3516 * key down event. 3517 * 3518 * Some IR protocols like NEC have a special IR message for 3519 * repeating last button, for when a button is held down. 3520 * 3521 * The *ctx* should point to the lirc sample as passed into 3522 * the program. 3523 * 3524 * This helper is only available is the kernel was compiled with 3525 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3526 * "**y**". 3527 * Return 3528 * 0 3529 * 3530 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3531 * Description 3532 * This helper is used in programs implementing IR decoding, to 3533 * report a successfully decoded key press with *scancode*, 3534 * *toggle* value in the given *protocol*. The scancode will be 3535 * translated to a keycode using the rc keymap, and reported as 3536 * an input key down event. After a period a key up event is 3537 * generated. This period can be extended by calling either 3538 * **bpf_rc_keydown**\ () again with the same values, or calling 3539 * **bpf_rc_repeat**\ (). 3540 * 3541 * Some protocols include a toggle bit, in case the button was 3542 * released and pressed again between consecutive scancodes. 3543 * 3544 * The *ctx* should point to the lirc sample as passed into 3545 * the program. 3546 * 3547 * The *protocol* is the decoded protocol number (see 3548 * **enum rc_proto** for some predefined values). 3549 * 3550 * This helper is only available is the kernel was compiled with 3551 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3552 * "**y**". 3553 * Return 3554 * 0 3555 * 3556 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3557 * Description 3558 * Return the cgroup v2 id of the socket associated with the *skb*. 3559 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3560 * helper for cgroup v1 by providing a tag resp. identifier that 3561 * can be matched on or used for map lookups e.g. to implement 3562 * policy. The cgroup v2 id of a given path in the hierarchy is 3563 * exposed in user space through the f_handle API in order to get 3564 * to the same 64-bit id. 3565 * 3566 * This helper can be used on TC egress path, but not on ingress, 3567 * and is available only if the kernel was compiled with the 3568 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3569 * Return 3570 * The id is returned or 0 in case the id could not be retrieved. 3571 * 3572 * u64 bpf_get_current_cgroup_id(void) 3573 * Description 3574 * Get the current cgroup id based on the cgroup within which 3575 * the current task is running. 3576 * Return 3577 * A 64-bit integer containing the current cgroup id based 3578 * on the cgroup within which the current task is running. 3579 * 3580 * void *bpf_get_local_storage(void *map, u64 flags) 3581 * Description 3582 * Get the pointer to the local storage area. 3583 * The type and the size of the local storage is defined 3584 * by the *map* argument. 3585 * The *flags* meaning is specific for each map type, 3586 * and has to be 0 for cgroup local storage. 3587 * 3588 * Depending on the BPF program type, a local storage area 3589 * can be shared between multiple instances of the BPF program, 3590 * running simultaneously. 3591 * 3592 * A user should care about the synchronization by himself. 3593 * For example, by using the **BPF_ATOMIC** instructions to alter 3594 * the shared data. 3595 * Return 3596 * A pointer to the local storage area. 3597 * 3598 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3599 * Description 3600 * Select a **SO_REUSEPORT** socket from a 3601 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*. 3602 * It checks the selected socket is matching the incoming 3603 * request in the socket buffer. 3604 * Return 3605 * 0 on success, or a negative error in case of failure. 3606 * 3607 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3608 * Description 3609 * Return id of cgroup v2 that is ancestor of cgroup associated 3610 * with the *skb* at the *ancestor_level*. The root cgroup is at 3611 * *ancestor_level* zero and each step down the hierarchy 3612 * increments the level. If *ancestor_level* == level of cgroup 3613 * associated with *skb*, then return value will be same as that 3614 * of **bpf_skb_cgroup_id**\ (). 3615 * 3616 * The helper is useful to implement policies based on cgroups 3617 * that are upper in hierarchy than immediate cgroup associated 3618 * with *skb*. 3619 * 3620 * The format of returned id and helper limitations are same as in 3621 * **bpf_skb_cgroup_id**\ (). 3622 * Return 3623 * The id is returned or 0 in case the id could not be retrieved. 3624 * 3625 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3626 * Description 3627 * Look for TCP socket matching *tuple*, optionally in a child 3628 * network namespace *netns*. The return value must be checked, 3629 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3630 * 3631 * The *ctx* should point to the context of the program, such as 3632 * the skb or socket (depending on the hook in use). This is used 3633 * to determine the base network namespace for the lookup. 3634 * 3635 * *tuple_size* must be one of: 3636 * 3637 * **sizeof**\ (*tuple*\ **->ipv4**) 3638 * Look for an IPv4 socket. 3639 * **sizeof**\ (*tuple*\ **->ipv6**) 3640 * Look for an IPv6 socket. 3641 * 3642 * If the *netns* is a negative signed 32-bit integer, then the 3643 * socket lookup table in the netns associated with the *ctx* 3644 * will be used. For the TC hooks, this is the netns of the device 3645 * in the skb. For socket hooks, this is the netns of the socket. 3646 * If *netns* is any other signed 32-bit value greater than or 3647 * equal to zero then it specifies the ID of the netns relative to 3648 * the netns associated with the *ctx*. *netns* values beyond the 3649 * range of 32-bit integers are reserved for future use. 3650 * 3651 * All values for *flags* are reserved for future usage, and must 3652 * be left at zero. 3653 * 3654 * This helper is available only if the kernel was compiled with 3655 * **CONFIG_NET** configuration option. 3656 * Return 3657 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3658 * For sockets with reuseport option, the **struct bpf_sock** 3659 * result is from *reuse*\ **->socks**\ [] using the hash of the 3660 * tuple. 3661 * 3662 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3663 * Description 3664 * Look for UDP socket matching *tuple*, optionally in a child 3665 * network namespace *netns*. The return value must be checked, 3666 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3667 * 3668 * The *ctx* should point to the context of the program, such as 3669 * the skb or socket (depending on the hook in use). This is used 3670 * to determine the base network namespace for the lookup. 3671 * 3672 * *tuple_size* must be one of: 3673 * 3674 * **sizeof**\ (*tuple*\ **->ipv4**) 3675 * Look for an IPv4 socket. 3676 * **sizeof**\ (*tuple*\ **->ipv6**) 3677 * Look for an IPv6 socket. 3678 * 3679 * If the *netns* is a negative signed 32-bit integer, then the 3680 * socket lookup table in the netns associated with the *ctx* 3681 * will be used. For the TC hooks, this is the netns of the device 3682 * in the skb. For socket hooks, this is the netns of the socket. 3683 * If *netns* is any other signed 32-bit value greater than or 3684 * equal to zero then it specifies the ID of the netns relative to 3685 * the netns associated with the *ctx*. *netns* values beyond the 3686 * range of 32-bit integers are reserved for future use. 3687 * 3688 * All values for *flags* are reserved for future usage, and must 3689 * be left at zero. 3690 * 3691 * This helper is available only if the kernel was compiled with 3692 * **CONFIG_NET** configuration option. 3693 * Return 3694 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3695 * For sockets with reuseport option, the **struct bpf_sock** 3696 * result is from *reuse*\ **->socks**\ [] using the hash of the 3697 * tuple. 3698 * 3699 * long bpf_sk_release(void *sock) 3700 * Description 3701 * Release the reference held by *sock*. *sock* must be a 3702 * non-**NULL** pointer that was returned from 3703 * **bpf_sk_lookup_xxx**\ (). 3704 * Return 3705 * 0 on success, or a negative error in case of failure. 3706 * 3707 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3708 * Description 3709 * Push an element *value* in *map*. *flags* is one of: 3710 * 3711 * **BPF_EXIST** 3712 * If the queue/stack is full, the oldest element is 3713 * removed to make room for this. 3714 * Return 3715 * 0 on success, or a negative error in case of failure. 3716 * 3717 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3718 * Description 3719 * Pop an element from *map*. 3720 * Return 3721 * 0 on success, or a negative error in case of failure. 3722 * 3723 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3724 * Description 3725 * Get an element from *map* without removing it. 3726 * Return 3727 * 0 on success, or a negative error in case of failure. 3728 * 3729 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3730 * Description 3731 * For socket policies, insert *len* bytes into *msg* at offset 3732 * *start*. 3733 * 3734 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3735 * *msg* it may want to insert metadata or options into the *msg*. 3736 * This can later be read and used by any of the lower layer BPF 3737 * hooks. 3738 * 3739 * This helper may fail if under memory pressure (a malloc 3740 * fails) in these cases BPF programs will get an appropriate 3741 * error and BPF programs will need to handle them. 3742 * Return 3743 * 0 on success, or a negative error in case of failure. 3744 * 3745 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3746 * Description 3747 * Will remove *len* bytes from a *msg* starting at byte *start*. 3748 * This may result in **ENOMEM** errors under certain situations if 3749 * an allocation and copy are required due to a full ring buffer. 3750 * However, the helper will try to avoid doing the allocation 3751 * if possible. Other errors can occur if input parameters are 3752 * invalid either due to *start* byte not being valid part of *msg* 3753 * payload and/or *pop* value being to large. 3754 * Return 3755 * 0 on success, or a negative error in case of failure. 3756 * 3757 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3758 * Description 3759 * This helper is used in programs implementing IR decoding, to 3760 * report a successfully decoded pointer movement. 3761 * 3762 * The *ctx* should point to the lirc sample as passed into 3763 * the program. 3764 * 3765 * This helper is only available is the kernel was compiled with 3766 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3767 * "**y**". 3768 * Return 3769 * 0 3770 * 3771 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3772 * Description 3773 * Acquire a spinlock represented by the pointer *lock*, which is 3774 * stored as part of a value of a map. Taking the lock allows to 3775 * safely update the rest of the fields in that value. The 3776 * spinlock can (and must) later be released with a call to 3777 * **bpf_spin_unlock**\ (\ *lock*\ ). 3778 * 3779 * Spinlocks in BPF programs come with a number of restrictions 3780 * and constraints: 3781 * 3782 * * **bpf_spin_lock** objects are only allowed inside maps of 3783 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3784 * list could be extended in the future). 3785 * * BTF description of the map is mandatory. 3786 * * The BPF program can take ONE lock at a time, since taking two 3787 * or more could cause dead locks. 3788 * * Only one **struct bpf_spin_lock** is allowed per map element. 3789 * * When the lock is taken, calls (either BPF to BPF or helpers) 3790 * are not allowed. 3791 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3792 * allowed inside a spinlock-ed region. 3793 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3794 * the lock, on all execution paths, before it returns. 3795 * * The BPF program can access **struct bpf_spin_lock** only via 3796 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3797 * helpers. Loading or storing data into the **struct 3798 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3799 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3800 * of the map value must be a struct and have **struct 3801 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3802 * Nested lock inside another struct is not allowed. 3803 * * The **struct bpf_spin_lock** *lock* field in a map value must 3804 * be aligned on a multiple of 4 bytes in that value. 3805 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3806 * the **bpf_spin_lock** field to user space. 3807 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3808 * a BPF program, do not update the **bpf_spin_lock** field. 3809 * * **bpf_spin_lock** cannot be on the stack or inside a 3810 * networking packet (it can only be inside of a map values). 3811 * * **bpf_spin_lock** is available to root only. 3812 * * Tracing programs and socket filter programs cannot use 3813 * **bpf_spin_lock**\ () due to insufficient preemption checks 3814 * (but this may change in the future). 3815 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3816 * Return 3817 * 0 3818 * 3819 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3820 * Description 3821 * Release the *lock* previously locked by a call to 3822 * **bpf_spin_lock**\ (\ *lock*\ ). 3823 * Return 3824 * 0 3825 * 3826 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3827 * Description 3828 * This helper gets a **struct bpf_sock** pointer such 3829 * that all the fields in this **bpf_sock** can be accessed. 3830 * Return 3831 * A **struct bpf_sock** pointer on success, or **NULL** in 3832 * case of failure. 3833 * 3834 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3835 * Description 3836 * This helper gets a **struct bpf_tcp_sock** pointer from a 3837 * **struct bpf_sock** pointer. 3838 * Return 3839 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3840 * case of failure. 3841 * 3842 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3843 * Description 3844 * Set ECN (Explicit Congestion Notification) field of IP header 3845 * to **CE** (Congestion Encountered) if current value is **ECT** 3846 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3847 * and IPv4. 3848 * Return 3849 * 1 if the **CE** flag is set (either by the current helper call 3850 * or because it was already present), 0 if it is not set. 3851 * 3852 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3853 * Description 3854 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3855 * **bpf_sk_release**\ () is unnecessary and not allowed. 3856 * Return 3857 * A **struct bpf_sock** pointer on success, or **NULL** in 3858 * case of failure. 3859 * 3860 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3861 * Description 3862 * Look for TCP socket matching *tuple*, optionally in a child 3863 * network namespace *netns*. The return value must be checked, 3864 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3865 * 3866 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3867 * that it also returns timewait or request sockets. Use 3868 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3869 * full structure. 3870 * 3871 * This helper is available only if the kernel was compiled with 3872 * **CONFIG_NET** configuration option. 3873 * Return 3874 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3875 * For sockets with reuseport option, the **struct bpf_sock** 3876 * result is from *reuse*\ **->socks**\ [] using the hash of the 3877 * tuple. 3878 * 3879 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3880 * Description 3881 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3882 * the listening socket in *sk*. 3883 * 3884 * *iph* points to the start of the IPv4 or IPv6 header, while 3885 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3886 * **sizeof**\ (**struct ipv6hdr**). 3887 * 3888 * *th* points to the start of the TCP header, while *th_len* 3889 * contains the length of the TCP header (at least 3890 * **sizeof**\ (**struct tcphdr**)). 3891 * Return 3892 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3893 * error otherwise. 3894 * 3895 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3896 * Description 3897 * Get name of sysctl in /proc/sys/ and copy it into provided by 3898 * program buffer *buf* of size *buf_len*. 3899 * 3900 * The buffer is always NUL terminated, unless it's zero-sized. 3901 * 3902 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3903 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3904 * only (e.g. "tcp_mem"). 3905 * Return 3906 * Number of character copied (not including the trailing NUL). 3907 * 3908 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3909 * truncated name in this case). 3910 * 3911 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3912 * Description 3913 * Get current value of sysctl as it is presented in /proc/sys 3914 * (incl. newline, etc), and copy it as a string into provided 3915 * by program buffer *buf* of size *buf_len*. 3916 * 3917 * The whole value is copied, no matter what file position user 3918 * space issued e.g. sys_read at. 3919 * 3920 * The buffer is always NUL terminated, unless it's zero-sized. 3921 * Return 3922 * Number of character copied (not including the trailing NUL). 3923 * 3924 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3925 * truncated name in this case). 3926 * 3927 * **-EINVAL** if current value was unavailable, e.g. because 3928 * sysctl is uninitialized and read returns -EIO for it. 3929 * 3930 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3931 * Description 3932 * Get new value being written by user space to sysctl (before 3933 * the actual write happens) and copy it as a string into 3934 * provided by program buffer *buf* of size *buf_len*. 3935 * 3936 * User space may write new value at file position > 0. 3937 * 3938 * The buffer is always NUL terminated, unless it's zero-sized. 3939 * Return 3940 * Number of character copied (not including the trailing NUL). 3941 * 3942 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3943 * truncated name in this case). 3944 * 3945 * **-EINVAL** if sysctl is being read. 3946 * 3947 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3948 * Description 3949 * Override new value being written by user space to sysctl with 3950 * value provided by program in buffer *buf* of size *buf_len*. 3951 * 3952 * *buf* should contain a string in same form as provided by user 3953 * space on sysctl write. 3954 * 3955 * User space may write new value at file position > 0. To override 3956 * the whole sysctl value file position should be set to zero. 3957 * Return 3958 * 0 on success. 3959 * 3960 * **-E2BIG** if the *buf_len* is too big. 3961 * 3962 * **-EINVAL** if sysctl is being read. 3963 * 3964 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3965 * Description 3966 * Convert the initial part of the string from buffer *buf* of 3967 * size *buf_len* to a long integer according to the given base 3968 * and save the result in *res*. 3969 * 3970 * The string may begin with an arbitrary amount of white space 3971 * (as determined by **isspace**\ (3)) followed by a single 3972 * optional '**-**' sign. 3973 * 3974 * Five least significant bits of *flags* encode base, other bits 3975 * are currently unused. 3976 * 3977 * Base must be either 8, 10, 16 or 0 to detect it automatically 3978 * similar to user space **strtol**\ (3). 3979 * Return 3980 * Number of characters consumed on success. Must be positive but 3981 * no more than *buf_len*. 3982 * 3983 * **-EINVAL** if no valid digits were found or unsupported base 3984 * was provided. 3985 * 3986 * **-ERANGE** if resulting value was out of range. 3987 * 3988 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3989 * Description 3990 * Convert the initial part of the string from buffer *buf* of 3991 * size *buf_len* to an unsigned long integer according to the 3992 * given base and save the result in *res*. 3993 * 3994 * The string may begin with an arbitrary amount of white space 3995 * (as determined by **isspace**\ (3)). 3996 * 3997 * Five least significant bits of *flags* encode base, other bits 3998 * are currently unused. 3999 * 4000 * Base must be either 8, 10, 16 or 0 to detect it automatically 4001 * similar to user space **strtoul**\ (3). 4002 * Return 4003 * Number of characters consumed on success. Must be positive but 4004 * no more than *buf_len*. 4005 * 4006 * **-EINVAL** if no valid digits were found or unsupported base 4007 * was provided. 4008 * 4009 * **-ERANGE** if resulting value was out of range. 4010 * 4011 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 4012 * Description 4013 * Get a bpf-local-storage from a *sk*. 4014 * 4015 * Logically, it could be thought of getting the value from 4016 * a *map* with *sk* as the **key**. From this 4017 * perspective, the usage is not much different from 4018 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 4019 * helper enforces the key must be a full socket and the map must 4020 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 4021 * 4022 * Underneath, the value is stored locally at *sk* instead of 4023 * the *map*. The *map* is used as the bpf-local-storage 4024 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4025 * searched against all bpf-local-storages residing at *sk*. 4026 * 4027 * *sk* is a kernel **struct sock** pointer for LSM program. 4028 * *sk* is a **struct bpf_sock** pointer for other program types. 4029 * 4030 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 4031 * used such that a new bpf-local-storage will be 4032 * created if one does not exist. *value* can be used 4033 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 4034 * the initial value of a bpf-local-storage. If *value* is 4035 * **NULL**, the new bpf-local-storage will be zero initialized. 4036 * Return 4037 * A bpf-local-storage pointer is returned on success. 4038 * 4039 * **NULL** if not found or there was an error in adding 4040 * a new bpf-local-storage. 4041 * 4042 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 4043 * Description 4044 * Delete a bpf-local-storage from a *sk*. 4045 * Return 4046 * 0 on success. 4047 * 4048 * **-ENOENT** if the bpf-local-storage cannot be found. 4049 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 4050 * 4051 * long bpf_send_signal(u32 sig) 4052 * Description 4053 * Send signal *sig* to the process of the current task. 4054 * The signal may be delivered to any of this process's threads. 4055 * Return 4056 * 0 on success or successfully queued. 4057 * 4058 * **-EBUSY** if work queue under nmi is full. 4059 * 4060 * **-EINVAL** if *sig* is invalid. 4061 * 4062 * **-EPERM** if no permission to send the *sig*. 4063 * 4064 * **-EAGAIN** if bpf program can try again. 4065 * 4066 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 4067 * Description 4068 * Try to issue a SYN cookie for the packet with corresponding 4069 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 4070 * 4071 * *iph* points to the start of the IPv4 or IPv6 header, while 4072 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 4073 * **sizeof**\ (**struct ipv6hdr**). 4074 * 4075 * *th* points to the start of the TCP header, while *th_len* 4076 * contains the length of the TCP header with options (at least 4077 * **sizeof**\ (**struct tcphdr**)). 4078 * Return 4079 * On success, lower 32 bits hold the generated SYN cookie in 4080 * followed by 16 bits which hold the MSS value for that cookie, 4081 * and the top 16 bits are unused. 4082 * 4083 * On failure, the returned value is one of the following: 4084 * 4085 * **-EINVAL** SYN cookie cannot be issued due to error 4086 * 4087 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 4088 * 4089 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 4090 * 4091 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 4092 * 4093 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4094 * Description 4095 * Write raw *data* blob into a special BPF perf event held by 4096 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4097 * event must have the following attributes: **PERF_SAMPLE_RAW** 4098 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4099 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4100 * 4101 * The *flags* are used to indicate the index in *map* for which 4102 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4103 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4104 * to indicate that the index of the current CPU core should be 4105 * used. 4106 * 4107 * The value to write, of *size*, is passed through eBPF stack and 4108 * pointed by *data*. 4109 * 4110 * *ctx* is a pointer to in-kernel struct sk_buff. 4111 * 4112 * This helper is similar to **bpf_perf_event_output**\ () but 4113 * restricted to raw_tracepoint bpf programs. 4114 * Return 4115 * 0 on success, or a negative error in case of failure. 4116 * 4117 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 4118 * Description 4119 * Safely attempt to read *size* bytes from user space address 4120 * *unsafe_ptr* and store the data in *dst*. 4121 * Return 4122 * 0 on success, or a negative error in case of failure. 4123 * 4124 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 4125 * Description 4126 * Safely attempt to read *size* bytes from kernel space address 4127 * *unsafe_ptr* and store the data in *dst*. 4128 * Return 4129 * 0 on success, or a negative error in case of failure. 4130 * 4131 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 4132 * Description 4133 * Copy a NUL terminated string from an unsafe user address 4134 * *unsafe_ptr* to *dst*. The *size* should include the 4135 * terminating NUL byte. In case the string length is smaller than 4136 * *size*, the target is not padded with further NUL bytes. If the 4137 * string length is larger than *size*, just *size*-1 bytes are 4138 * copied and the last byte is set to NUL. 4139 * 4140 * On success, returns the number of bytes that were written, 4141 * including the terminal NUL. This makes this helper useful in 4142 * tracing programs for reading strings, and more importantly to 4143 * get its length at runtime. See the following snippet: 4144 * 4145 * :: 4146 * 4147 * SEC("kprobe/sys_open") 4148 * void bpf_sys_open(struct pt_regs *ctx) 4149 * { 4150 * char buf[PATHLEN]; // PATHLEN is defined to 256 4151 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 4152 * ctx->di); 4153 * 4154 * // Consume buf, for example push it to 4155 * // userspace via bpf_perf_event_output(); we 4156 * // can use res (the string length) as event 4157 * // size, after checking its boundaries. 4158 * } 4159 * 4160 * In comparison, using **bpf_probe_read_user**\ () helper here 4161 * instead to read the string would require to estimate the length 4162 * at compile time, and would often result in copying more memory 4163 * than necessary. 4164 * 4165 * Another useful use case is when parsing individual process 4166 * arguments or individual environment variables navigating 4167 * *current*\ **->mm->arg_start** and *current*\ 4168 * **->mm->env_start**: using this helper and the return value, 4169 * one can quickly iterate at the right offset of the memory area. 4170 * Return 4171 * On success, the strictly positive length of the output string, 4172 * including the trailing NUL character. On error, a negative 4173 * value. 4174 * 4175 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 4176 * Description 4177 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 4178 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 4179 * Return 4180 * On success, the strictly positive length of the string, including 4181 * the trailing NUL character. On error, a negative value. 4182 * 4183 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 4184 * Description 4185 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 4186 * *rcv_nxt* is the ack_seq to be sent out. 4187 * Return 4188 * 0 on success, or a negative error in case of failure. 4189 * 4190 * long bpf_send_signal_thread(u32 sig) 4191 * Description 4192 * Send signal *sig* to the thread corresponding to the current task. 4193 * Return 4194 * 0 on success or successfully queued. 4195 * 4196 * **-EBUSY** if work queue under nmi is full. 4197 * 4198 * **-EINVAL** if *sig* is invalid. 4199 * 4200 * **-EPERM** if no permission to send the *sig*. 4201 * 4202 * **-EAGAIN** if bpf program can try again. 4203 * 4204 * u64 bpf_jiffies64(void) 4205 * Description 4206 * Obtain the 64bit jiffies 4207 * Return 4208 * The 64 bit jiffies 4209 * 4210 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 4211 * Description 4212 * For an eBPF program attached to a perf event, retrieve the 4213 * branch records (**struct perf_branch_entry**) associated to *ctx* 4214 * and store it in the buffer pointed by *buf* up to size 4215 * *size* bytes. 4216 * Return 4217 * On success, number of bytes written to *buf*. On error, a 4218 * negative value. 4219 * 4220 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 4221 * instead return the number of bytes required to store all the 4222 * branch entries. If this flag is set, *buf* may be NULL. 4223 * 4224 * **-EINVAL** if arguments invalid or **size** not a multiple 4225 * of **sizeof**\ (**struct perf_branch_entry**\ ). 4226 * 4227 * **-ENOENT** if architecture does not support branch records. 4228 * 4229 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 4230 * Description 4231 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 4232 * *namespace* will be returned in *nsdata*. 4233 * Return 4234 * 0 on success, or one of the following in case of failure: 4235 * 4236 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 4237 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 4238 * 4239 * **-ENOENT** if pidns does not exists for the current task. 4240 * 4241 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 4242 * Description 4243 * Write raw *data* blob into a special BPF perf event held by 4244 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 4245 * event must have the following attributes: **PERF_SAMPLE_RAW** 4246 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 4247 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 4248 * 4249 * The *flags* are used to indicate the index in *map* for which 4250 * the value must be put, masked with **BPF_F_INDEX_MASK**. 4251 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 4252 * to indicate that the index of the current CPU core should be 4253 * used. 4254 * 4255 * The value to write, of *size*, is passed through eBPF stack and 4256 * pointed by *data*. 4257 * 4258 * *ctx* is a pointer to in-kernel struct xdp_buff. 4259 * 4260 * This helper is similar to **bpf_perf_eventoutput**\ () but 4261 * restricted to raw_tracepoint bpf programs. 4262 * Return 4263 * 0 on success, or a negative error in case of failure. 4264 * 4265 * u64 bpf_get_netns_cookie(void *ctx) 4266 * Description 4267 * Retrieve the cookie (generated by the kernel) of the network 4268 * namespace the input *ctx* is associated with. The network 4269 * namespace cookie remains stable for its lifetime and provides 4270 * a global identifier that can be assumed unique. If *ctx* is 4271 * NULL, then the helper returns the cookie for the initial 4272 * network namespace. The cookie itself is very similar to that 4273 * of **bpf_get_socket_cookie**\ () helper, but for network 4274 * namespaces instead of sockets. 4275 * Return 4276 * A 8-byte long opaque number. 4277 * 4278 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 4279 * Description 4280 * Return id of cgroup v2 that is ancestor of the cgroup associated 4281 * with the current task at the *ancestor_level*. The root cgroup 4282 * is at *ancestor_level* zero and each step down the hierarchy 4283 * increments the level. If *ancestor_level* == level of cgroup 4284 * associated with the current task, then return value will be the 4285 * same as that of **bpf_get_current_cgroup_id**\ (). 4286 * 4287 * The helper is useful to implement policies based on cgroups 4288 * that are upper in hierarchy than immediate cgroup associated 4289 * with the current task. 4290 * 4291 * The format of returned id and helper limitations are same as in 4292 * **bpf_get_current_cgroup_id**\ (). 4293 * Return 4294 * The id is returned or 0 in case the id could not be retrieved. 4295 * 4296 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 4297 * Description 4298 * Helper is overloaded depending on BPF program type. This 4299 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 4300 * **BPF_PROG_TYPE_SCHED_ACT** programs. 4301 * 4302 * Assign the *sk* to the *skb*. When combined with appropriate 4303 * routing configuration to receive the packet towards the socket, 4304 * will cause *skb* to be delivered to the specified socket. 4305 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 4306 * **bpf_clone_redirect**\ () or other methods outside of BPF may 4307 * interfere with successful delivery to the socket. 4308 * 4309 * This operation is only valid from TC ingress path. 4310 * 4311 * The *flags* argument must be zero. 4312 * Return 4313 * 0 on success, or a negative error in case of failure: 4314 * 4315 * **-EINVAL** if specified *flags* are not supported. 4316 * 4317 * **-ENOENT** if the socket is unavailable for assignment. 4318 * 4319 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 4320 * 4321 * **-EOPNOTSUPP** if the operation is not supported, for example 4322 * a call from outside of TC ingress. 4323 * 4324 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 4325 * Description 4326 * Helper is overloaded depending on BPF program type. This 4327 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 4328 * 4329 * Select the *sk* as a result of a socket lookup. 4330 * 4331 * For the operation to succeed passed socket must be compatible 4332 * with the packet description provided by the *ctx* object. 4333 * 4334 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 4335 * be an exact match. While IP family (**AF_INET** or 4336 * **AF_INET6**) must be compatible, that is IPv6 sockets 4337 * that are not v6-only can be selected for IPv4 packets. 4338 * 4339 * Only TCP listeners and UDP unconnected sockets can be 4340 * selected. *sk* can also be NULL to reset any previous 4341 * selection. 4342 * 4343 * *flags* argument can combination of following values: 4344 * 4345 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 4346 * socket selection, potentially done by a BPF program 4347 * that ran before us. 4348 * 4349 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 4350 * load-balancing within reuseport group for the socket 4351 * being selected. 4352 * 4353 * On success *ctx->sk* will point to the selected socket. 4354 * 4355 * Return 4356 * 0 on success, or a negative errno in case of failure. 4357 * 4358 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 4359 * not compatible with packet family (*ctx->family*). 4360 * 4361 * * **-EEXIST** if socket has been already selected, 4362 * potentially by another program, and 4363 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 4364 * 4365 * * **-EINVAL** if unsupported flags were specified. 4366 * 4367 * * **-EPROTOTYPE** if socket L4 protocol 4368 * (*sk->protocol*) doesn't match packet protocol 4369 * (*ctx->protocol*). 4370 * 4371 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4372 * state (TCP listening or UDP unconnected). 4373 * 4374 * u64 bpf_ktime_get_boot_ns(void) 4375 * Description 4376 * Return the time elapsed since system boot, in nanoseconds. 4377 * Does include the time the system was suspended. 4378 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4379 * Return 4380 * Current *ktime*. 4381 * 4382 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4383 * Description 4384 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4385 * out the format string. 4386 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4387 * the format string itself. The *data* and *data_len* are format string 4388 * arguments. The *data* are a **u64** array and corresponding format string 4389 * values are stored in the array. For strings and pointers where pointees 4390 * are accessed, only the pointer values are stored in the *data* array. 4391 * The *data_len* is the size of *data* in bytes - must be a multiple of 8. 4392 * 4393 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4394 * Reading kernel memory may fail due to either invalid address or 4395 * valid address but requiring a major memory fault. If reading kernel memory 4396 * fails, the string for **%s** will be an empty string, and the ip 4397 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4398 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4399 * Return 4400 * 0 on success, or a negative error in case of failure: 4401 * 4402 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4403 * by returning 1 from bpf program. 4404 * 4405 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4406 * 4407 * **-E2BIG** if *fmt* contains too many format specifiers. 4408 * 4409 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4410 * 4411 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4412 * Description 4413 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4414 * The *m* represents the seq_file. The *data* and *len* represent the 4415 * data to write in bytes. 4416 * Return 4417 * 0 on success, or a negative error in case of failure: 4418 * 4419 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4420 * 4421 * u64 bpf_sk_cgroup_id(void *sk) 4422 * Description 4423 * Return the cgroup v2 id of the socket *sk*. 4424 * 4425 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4426 * returned from **bpf_sk_lookup_xxx**\ (), 4427 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4428 * same as in **bpf_skb_cgroup_id**\ (). 4429 * 4430 * This helper is available only if the kernel was compiled with 4431 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4432 * Return 4433 * The id is returned or 0 in case the id could not be retrieved. 4434 * 4435 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4436 * Description 4437 * Return id of cgroup v2 that is ancestor of cgroup associated 4438 * with the *sk* at the *ancestor_level*. The root cgroup is at 4439 * *ancestor_level* zero and each step down the hierarchy 4440 * increments the level. If *ancestor_level* == level of cgroup 4441 * associated with *sk*, then return value will be same as that 4442 * of **bpf_sk_cgroup_id**\ (). 4443 * 4444 * The helper is useful to implement policies based on cgroups 4445 * that are upper in hierarchy than immediate cgroup associated 4446 * with *sk*. 4447 * 4448 * The format of returned id and helper limitations are same as in 4449 * **bpf_sk_cgroup_id**\ (). 4450 * Return 4451 * The id is returned or 0 in case the id could not be retrieved. 4452 * 4453 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4454 * Description 4455 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4456 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4457 * of new data availability is sent. 4458 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4459 * of new data availability is sent unconditionally. 4460 * If **0** is specified in *flags*, an adaptive notification 4461 * of new data availability is sent. 4462 * 4463 * An adaptive notification is a notification sent whenever the user-space 4464 * process has caught up and consumed all available payloads. In case the user-space 4465 * process is still processing a previous payload, then no notification is needed 4466 * as it will process the newly added payload automatically. 4467 * Return 4468 * 0 on success, or a negative error in case of failure. 4469 * 4470 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4471 * Description 4472 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4473 * *flags* must be 0. 4474 * Return 4475 * Valid pointer with *size* bytes of memory available; NULL, 4476 * otherwise. 4477 * 4478 * void bpf_ringbuf_submit(void *data, u64 flags) 4479 * Description 4480 * Submit reserved ring buffer sample, pointed to by *data*. 4481 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4482 * of new data availability is sent. 4483 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4484 * of new data availability is sent unconditionally. 4485 * If **0** is specified in *flags*, an adaptive notification 4486 * of new data availability is sent. 4487 * 4488 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4489 * Return 4490 * Nothing. Always succeeds. 4491 * 4492 * void bpf_ringbuf_discard(void *data, u64 flags) 4493 * Description 4494 * Discard reserved ring buffer sample, pointed to by *data*. 4495 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4496 * of new data availability is sent. 4497 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4498 * of new data availability is sent unconditionally. 4499 * If **0** is specified in *flags*, an adaptive notification 4500 * of new data availability is sent. 4501 * 4502 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4503 * Return 4504 * Nothing. Always succeeds. 4505 * 4506 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4507 * Description 4508 * Query various characteristics of provided ring buffer. What 4509 * exactly is queries is determined by *flags*: 4510 * 4511 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4512 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4513 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4514 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4515 * 4516 * Data returned is just a momentary snapshot of actual values 4517 * and could be inaccurate, so this facility should be used to 4518 * power heuristics and for reporting, not to make 100% correct 4519 * calculation. 4520 * Return 4521 * Requested value, or 0, if *flags* are not recognized. 4522 * 4523 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4524 * Description 4525 * Change the skbs checksum level by one layer up or down, or 4526 * reset it entirely to none in order to have the stack perform 4527 * checksum validation. The level is applicable to the following 4528 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4529 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4530 * through **bpf_skb_adjust_room**\ () helper with passing in 4531 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4532 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4533 * the UDP header is removed. Similarly, an encap of the latter 4534 * into the former could be accompanied by a helper call to 4535 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4536 * skb is still intended to be processed in higher layers of the 4537 * stack instead of just egressing at tc. 4538 * 4539 * There are three supported level settings at this time: 4540 * 4541 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4542 * with CHECKSUM_UNNECESSARY. 4543 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4544 * with CHECKSUM_UNNECESSARY. 4545 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4546 * sets CHECKSUM_NONE to force checksum validation by the stack. 4547 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4548 * skb->csum_level. 4549 * Return 4550 * 0 on success, or a negative error in case of failure. In the 4551 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4552 * is returned or the error code -EACCES in case the skb is not 4553 * subject to CHECKSUM_UNNECESSARY. 4554 * 4555 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4556 * Description 4557 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4558 * Return 4559 * *sk* if casting is valid, or **NULL** otherwise. 4560 * 4561 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4562 * Description 4563 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4564 * Return 4565 * *sk* if casting is valid, or **NULL** otherwise. 4566 * 4567 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4568 * Description 4569 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4570 * Return 4571 * *sk* if casting is valid, or **NULL** otherwise. 4572 * 4573 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4574 * Description 4575 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4576 * Return 4577 * *sk* if casting is valid, or **NULL** otherwise. 4578 * 4579 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4580 * Description 4581 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4582 * Return 4583 * *sk* if casting is valid, or **NULL** otherwise. 4584 * 4585 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4586 * Description 4587 * Return a user or a kernel stack in bpf program provided buffer. 4588 * Note: the user stack will only be populated if the *task* is 4589 * the current task; all other tasks will return -EOPNOTSUPP. 4590 * To achieve this, the helper needs *task*, which is a valid 4591 * pointer to **struct task_struct**. To store the stacktrace, the 4592 * bpf program provides *buf* with a nonnegative *size*. 4593 * 4594 * The last argument, *flags*, holds the number of stack frames to 4595 * skip (from 0 to 255), masked with 4596 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4597 * the following flags: 4598 * 4599 * **BPF_F_USER_STACK** 4600 * Collect a user space stack instead of a kernel stack. 4601 * The *task* must be the current task. 4602 * **BPF_F_USER_BUILD_ID** 4603 * Collect buildid+offset instead of ips for user stack, 4604 * only valid if **BPF_F_USER_STACK** is also specified. 4605 * 4606 * **bpf_get_task_stack**\ () can collect up to 4607 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4608 * to sufficient large buffer size. Note that 4609 * this limit can be controlled with the **sysctl** program, and 4610 * that it should be manually increased in order to profile long 4611 * user stacks (such as stacks for Java programs). To do so, use: 4612 * 4613 * :: 4614 * 4615 * # sysctl kernel.perf_event_max_stack=<new value> 4616 * Return 4617 * The non-negative copied *buf* length equal to or less than 4618 * *size* on success, or a negative error in case of failure. 4619 * 4620 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4621 * Description 4622 * Load header option. Support reading a particular TCP header 4623 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4624 * 4625 * If *flags* is 0, it will search the option from the 4626 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4627 * has details on what skb_data contains under different 4628 * *skops*\ **->op**. 4629 * 4630 * The first byte of the *searchby_res* specifies the 4631 * kind that it wants to search. 4632 * 4633 * If the searching kind is an experimental kind 4634 * (i.e. 253 or 254 according to RFC6994). It also 4635 * needs to specify the "magic" which is either 4636 * 2 bytes or 4 bytes. It then also needs to 4637 * specify the size of the magic by using 4638 * the 2nd byte which is "kind-length" of a TCP 4639 * header option and the "kind-length" also 4640 * includes the first 2 bytes "kind" and "kind-length" 4641 * itself as a normal TCP header option also does. 4642 * 4643 * For example, to search experimental kind 254 with 4644 * 2 byte magic 0xeB9F, the searchby_res should be 4645 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4646 * 4647 * To search for the standard window scale option (3), 4648 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4649 * Note, kind-length must be 0 for regular option. 4650 * 4651 * Searching for No-Op (0) and End-of-Option-List (1) are 4652 * not supported. 4653 * 4654 * *len* must be at least 2 bytes which is the minimal size 4655 * of a header option. 4656 * 4657 * Supported flags: 4658 * 4659 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4660 * saved_syn packet or the just-received syn packet. 4661 * 4662 * Return 4663 * > 0 when found, the header option is copied to *searchby_res*. 4664 * The return value is the total length copied. On failure, a 4665 * negative error code is returned: 4666 * 4667 * **-EINVAL** if a parameter is invalid. 4668 * 4669 * **-ENOMSG** if the option is not found. 4670 * 4671 * **-ENOENT** if no syn packet is available when 4672 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4673 * 4674 * **-ENOSPC** if there is not enough space. Only *len* number of 4675 * bytes are copied. 4676 * 4677 * **-EFAULT** on failure to parse the header options in the 4678 * packet. 4679 * 4680 * **-EPERM** if the helper cannot be used under the current 4681 * *skops*\ **->op**. 4682 * 4683 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4684 * Description 4685 * Store header option. The data will be copied 4686 * from buffer *from* with length *len* to the TCP header. 4687 * 4688 * The buffer *from* should have the whole option that 4689 * includes the kind, kind-length, and the actual 4690 * option data. The *len* must be at least kind-length 4691 * long. The kind-length does not have to be 4 byte 4692 * aligned. The kernel will take care of the padding 4693 * and setting the 4 bytes aligned value to th->doff. 4694 * 4695 * This helper will check for duplicated option 4696 * by searching the same option in the outgoing skb. 4697 * 4698 * This helper can only be called during 4699 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4700 * 4701 * Return 4702 * 0 on success, or negative error in case of failure: 4703 * 4704 * **-EINVAL** If param is invalid. 4705 * 4706 * **-ENOSPC** if there is not enough space in the header. 4707 * Nothing has been written 4708 * 4709 * **-EEXIST** if the option already exists. 4710 * 4711 * **-EFAULT** on failure to parse the existing header options. 4712 * 4713 * **-EPERM** if the helper cannot be used under the current 4714 * *skops*\ **->op**. 4715 * 4716 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4717 * Description 4718 * Reserve *len* bytes for the bpf header option. The 4719 * space will be used by **bpf_store_hdr_opt**\ () later in 4720 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4721 * 4722 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4723 * the total number of bytes will be reserved. 4724 * 4725 * This helper can only be called during 4726 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4727 * 4728 * Return 4729 * 0 on success, or negative error in case of failure: 4730 * 4731 * **-EINVAL** if a parameter is invalid. 4732 * 4733 * **-ENOSPC** if there is not enough space in the header. 4734 * 4735 * **-EPERM** if the helper cannot be used under the current 4736 * *skops*\ **->op**. 4737 * 4738 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4739 * Description 4740 * Get a bpf_local_storage from an *inode*. 4741 * 4742 * Logically, it could be thought of as getting the value from 4743 * a *map* with *inode* as the **key**. From this 4744 * perspective, the usage is not much different from 4745 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4746 * helper enforces the key must be an inode and the map must also 4747 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4748 * 4749 * Underneath, the value is stored locally at *inode* instead of 4750 * the *map*. The *map* is used as the bpf-local-storage 4751 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4752 * searched against all bpf_local_storage residing at *inode*. 4753 * 4754 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4755 * used such that a new bpf_local_storage will be 4756 * created if one does not exist. *value* can be used 4757 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4758 * the initial value of a bpf_local_storage. If *value* is 4759 * **NULL**, the new bpf_local_storage will be zero initialized. 4760 * Return 4761 * A bpf_local_storage pointer is returned on success. 4762 * 4763 * **NULL** if not found or there was an error in adding 4764 * a new bpf_local_storage. 4765 * 4766 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4767 * Description 4768 * Delete a bpf_local_storage from an *inode*. 4769 * Return 4770 * 0 on success. 4771 * 4772 * **-ENOENT** if the bpf_local_storage cannot be found. 4773 * 4774 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4775 * Description 4776 * Return full path for given **struct path** object, which 4777 * needs to be the kernel BTF *path* object. The path is 4778 * returned in the provided buffer *buf* of size *sz* and 4779 * is zero terminated. 4780 * 4781 * Return 4782 * On success, the strictly positive length of the string, 4783 * including the trailing NUL character. On error, a negative 4784 * value. 4785 * 4786 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4787 * Description 4788 * Read *size* bytes from user space address *user_ptr* and store 4789 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4790 * Return 4791 * 0 on success, or a negative error in case of failure. 4792 * 4793 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4794 * Description 4795 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4796 * using *ptr*->type_id. This value should specify the type 4797 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4798 * can be used to look up vmlinux BTF type ids. Traversing the 4799 * data structure using BTF, the type information and values are 4800 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4801 * the pointer data is carried out to avoid kernel crashes during 4802 * operation. Smaller types can use string space on the stack; 4803 * larger programs can use map data to store the string 4804 * representation. 4805 * 4806 * The string can be subsequently shared with userspace via 4807 * bpf_perf_event_output() or ring buffer interfaces. 4808 * bpf_trace_printk() is to be avoided as it places too small 4809 * a limit on string size to be useful. 4810 * 4811 * *flags* is a combination of 4812 * 4813 * **BTF_F_COMPACT** 4814 * no formatting around type information 4815 * **BTF_F_NONAME** 4816 * no struct/union member names/types 4817 * **BTF_F_PTR_RAW** 4818 * show raw (unobfuscated) pointer values; 4819 * equivalent to printk specifier %px. 4820 * **BTF_F_ZERO** 4821 * show zero-valued struct/union members; they 4822 * are not displayed by default 4823 * 4824 * Return 4825 * The number of bytes that were written (or would have been 4826 * written if output had to be truncated due to string size), 4827 * or a negative error in cases of failure. 4828 * 4829 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4830 * Description 4831 * Use BTF to write to seq_write a string representation of 4832 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4833 * *flags* are identical to those used for bpf_snprintf_btf. 4834 * Return 4835 * 0 on success or a negative error in case of failure. 4836 * 4837 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4838 * Description 4839 * See **bpf_get_cgroup_classid**\ () for the main description. 4840 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4841 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4842 * associated socket instead of the current process. 4843 * Return 4844 * The id is returned or 0 in case the id could not be retrieved. 4845 * 4846 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4847 * Description 4848 * Redirect the packet to another net device of index *ifindex* 4849 * and fill in L2 addresses from neighboring subsystem. This helper 4850 * is somewhat similar to **bpf_redirect**\ (), except that it 4851 * populates L2 addresses as well, meaning, internally, the helper 4852 * relies on the neighbor lookup for the L2 address of the nexthop. 4853 * 4854 * The helper will perform a FIB lookup based on the skb's 4855 * networking header to get the address of the next hop, unless 4856 * this is supplied by the caller in the *params* argument. The 4857 * *plen* argument indicates the len of *params* and should be set 4858 * to 0 if *params* is NULL. 4859 * 4860 * The *flags* argument is reserved and must be 0. The helper is 4861 * currently only supported for tc BPF program types, and enabled 4862 * for IPv4 and IPv6 protocols. 4863 * Return 4864 * The helper returns **TC_ACT_REDIRECT** on success or 4865 * **TC_ACT_SHOT** on error. 4866 * 4867 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4868 * Description 4869 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4870 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4871 * extern variable decorated with '__ksym'. For ksym, there is a 4872 * global var (either static or global) defined of the same name 4873 * in the kernel. The ksym is percpu if the global var is percpu. 4874 * The returned pointer points to the global percpu var on *cpu*. 4875 * 4876 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4877 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4878 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4879 * bpf_per_cpu_ptr() must check the returned value. 4880 * Return 4881 * A pointer pointing to the kernel percpu variable on *cpu*, or 4882 * NULL, if *cpu* is invalid. 4883 * 4884 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4885 * Description 4886 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4887 * pointer to the percpu kernel variable on this cpu. See the 4888 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4889 * 4890 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4891 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4892 * never return NULL. 4893 * Return 4894 * A pointer pointing to the kernel percpu variable on this cpu. 4895 * 4896 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4897 * Description 4898 * Redirect the packet to another net device of index *ifindex*. 4899 * This helper is somewhat similar to **bpf_redirect**\ (), except 4900 * that the redirection happens to the *ifindex*' peer device and 4901 * the netns switch takes place from ingress to ingress without 4902 * going through the CPU's backlog queue. 4903 * 4904 * The *flags* argument is reserved and must be 0. The helper is 4905 * currently only supported for tc BPF program types at the 4906 * ingress hook and for veth and netkit target device types. The 4907 * peer device must reside in a different network namespace. 4908 * Return 4909 * The helper returns **TC_ACT_REDIRECT** on success or 4910 * **TC_ACT_SHOT** on error. 4911 * 4912 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4913 * Description 4914 * Get a bpf_local_storage from the *task*. 4915 * 4916 * Logically, it could be thought of as getting the value from 4917 * a *map* with *task* as the **key**. From this 4918 * perspective, the usage is not much different from 4919 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4920 * helper enforces the key must be a task_struct and the map must also 4921 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4922 * 4923 * Underneath, the value is stored locally at *task* instead of 4924 * the *map*. The *map* is used as the bpf-local-storage 4925 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4926 * searched against all bpf_local_storage residing at *task*. 4927 * 4928 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4929 * used such that a new bpf_local_storage will be 4930 * created if one does not exist. *value* can be used 4931 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4932 * the initial value of a bpf_local_storage. If *value* is 4933 * **NULL**, the new bpf_local_storage will be zero initialized. 4934 * Return 4935 * A bpf_local_storage pointer is returned on success. 4936 * 4937 * **NULL** if not found or there was an error in adding 4938 * a new bpf_local_storage. 4939 * 4940 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4941 * Description 4942 * Delete a bpf_local_storage from a *task*. 4943 * Return 4944 * 0 on success. 4945 * 4946 * **-ENOENT** if the bpf_local_storage cannot be found. 4947 * 4948 * struct task_struct *bpf_get_current_task_btf(void) 4949 * Description 4950 * Return a BTF pointer to the "current" task. 4951 * This pointer can also be used in helpers that accept an 4952 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4953 * Return 4954 * Pointer to the current task. 4955 * 4956 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4957 * Description 4958 * Set or clear certain options on *bprm*: 4959 * 4960 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4961 * which sets the **AT_SECURE** auxv for glibc. The bit 4962 * is cleared if the flag is not specified. 4963 * Return 4964 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4965 * 4966 * u64 bpf_ktime_get_coarse_ns(void) 4967 * Description 4968 * Return a coarse-grained version of the time elapsed since 4969 * system boot, in nanoseconds. Does not include time the system 4970 * was suspended. 4971 * 4972 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4973 * Return 4974 * Current *ktime*. 4975 * 4976 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4977 * Description 4978 * Returns the stored IMA hash of the *inode* (if it's available). 4979 * If the hash is larger than *size*, then only *size* 4980 * bytes will be copied to *dst* 4981 * Return 4982 * The **hash_algo** is returned on success, 4983 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4984 * invalid arguments are passed. 4985 * 4986 * struct socket *bpf_sock_from_file(struct file *file) 4987 * Description 4988 * If the given file represents a socket, returns the associated 4989 * socket. 4990 * Return 4991 * A pointer to a struct socket on success or NULL if the file is 4992 * not a socket. 4993 * 4994 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4995 * Description 4996 * Check packet size against exceeding MTU of net device (based 4997 * on *ifindex*). This helper will likely be used in combination 4998 * with helpers that adjust/change the packet size. 4999 * 5000 * The argument *len_diff* can be used for querying with a planned 5001 * size change. This allows to check MTU prior to changing packet 5002 * ctx. Providing a *len_diff* adjustment that is larger than the 5003 * actual packet size (resulting in negative packet size) will in 5004 * principle not exceed the MTU, which is why it is not considered 5005 * a failure. Other BPF helpers are needed for performing the 5006 * planned size change; therefore the responsibility for catching 5007 * a negative packet size belongs in those helpers. 5008 * 5009 * Specifying *ifindex* zero means the MTU check is performed 5010 * against the current net device. This is practical if this isn't 5011 * used prior to redirect. 5012 * 5013 * On input *mtu_len* must be a valid pointer, else verifier will 5014 * reject BPF program. If the value *mtu_len* is initialized to 5015 * zero then the ctx packet size is use. When value *mtu_len* is 5016 * provided as input this specify the L3 length that the MTU check 5017 * is done against. Remember XDP and TC length operate at L2, but 5018 * this value is L3 as this correlate to MTU and IP-header tot_len 5019 * values which are L3 (similar behavior as bpf_fib_lookup). 5020 * 5021 * The Linux kernel route table can configure MTUs on a more 5022 * specific per route level, which is not provided by this helper. 5023 * For route level MTU checks use the **bpf_fib_lookup**\ () 5024 * helper. 5025 * 5026 * *ctx* is either **struct xdp_md** for XDP programs or 5027 * **struct sk_buff** for tc cls_act programs. 5028 * 5029 * The *flags* argument can be a combination of one or more of the 5030 * following values: 5031 * 5032 * **BPF_MTU_CHK_SEGS** 5033 * This flag will only works for *ctx* **struct sk_buff**. 5034 * If packet context contains extra packet segment buffers 5035 * (often knows as GSO skb), then MTU check is harder to 5036 * check at this point, because in transmit path it is 5037 * possible for the skb packet to get re-segmented 5038 * (depending on net device features). This could still be 5039 * a MTU violation, so this flag enables performing MTU 5040 * check against segments, with a different violation 5041 * return code to tell it apart. Check cannot use len_diff. 5042 * 5043 * On return *mtu_len* pointer contains the MTU value of the net 5044 * device. Remember the net device configured MTU is the L3 size, 5045 * which is returned here and XDP and TC length operate at L2. 5046 * Helper take this into account for you, but remember when using 5047 * MTU value in your BPF-code. 5048 * 5049 * Return 5050 * * 0 on success, and populate MTU value in *mtu_len* pointer. 5051 * 5052 * * < 0 if any input argument is invalid (*mtu_len* not updated) 5053 * 5054 * MTU violations return positive values, but also populate MTU 5055 * value in *mtu_len* pointer, as this can be needed for 5056 * implementing PMTU handing: 5057 * 5058 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 5059 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 5060 * 5061 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 5062 * Description 5063 * For each element in **map**, call **callback_fn** function with 5064 * **map**, **callback_ctx** and other map-specific parameters. 5065 * The **callback_fn** should be a static function and 5066 * the **callback_ctx** should be a pointer to the stack. 5067 * The **flags** is used to control certain aspects of the helper. 5068 * Currently, the **flags** must be 0. 5069 * 5070 * The following are a list of supported map types and their 5071 * respective expected callback signatures: 5072 * 5073 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 5074 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 5075 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 5076 * 5077 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 5078 * 5079 * For per_cpu maps, the map_value is the value on the cpu where the 5080 * bpf_prog is running. 5081 * 5082 * If **callback_fn** return 0, the helper will continue to the next 5083 * element. If return value is 1, the helper will skip the rest of 5084 * elements and return. Other return values are not used now. 5085 * 5086 * Return 5087 * The number of traversed map elements for success, **-EINVAL** for 5088 * invalid **flags**. 5089 * 5090 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 5091 * Description 5092 * Outputs a string into the **str** buffer of size **str_size** 5093 * based on a format string stored in a read-only map pointed by 5094 * **fmt**. 5095 * 5096 * Each format specifier in **fmt** corresponds to one u64 element 5097 * in the **data** array. For strings and pointers where pointees 5098 * are accessed, only the pointer values are stored in the *data* 5099 * array. The *data_len* is the size of *data* in bytes - must be 5100 * a multiple of 8. 5101 * 5102 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 5103 * memory. Reading kernel memory may fail due to either invalid 5104 * address or valid address but requiring a major memory fault. If 5105 * reading kernel memory fails, the string for **%s** will be an 5106 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 5107 * Not returning error to bpf program is consistent with what 5108 * **bpf_trace_printk**\ () does for now. 5109 * 5110 * Return 5111 * The strictly positive length of the formatted string, including 5112 * the trailing zero character. If the return value is greater than 5113 * **str_size**, **str** contains a truncated string, guaranteed to 5114 * be zero-terminated except when **str_size** is 0. 5115 * 5116 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 5117 * 5118 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 5119 * Description 5120 * Execute bpf syscall with given arguments. 5121 * Return 5122 * A syscall result. 5123 * 5124 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 5125 * Description 5126 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 5127 * Return 5128 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 5129 * 5130 * long bpf_sys_close(u32 fd) 5131 * Description 5132 * Execute close syscall for given FD. 5133 * Return 5134 * A syscall result. 5135 * 5136 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags) 5137 * Description 5138 * Initialize the timer. 5139 * First 4 bits of *flags* specify clockid. 5140 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed. 5141 * All other bits of *flags* are reserved. 5142 * The verifier will reject the program if *timer* is not from 5143 * the same *map*. 5144 * Return 5145 * 0 on success. 5146 * **-EBUSY** if *timer* is already initialized. 5147 * **-EINVAL** if invalid *flags* are passed. 5148 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5149 * The user space should either hold a file descriptor to a map with timers 5150 * or pin such map in bpffs. When map is unpinned or file descriptor is 5151 * closed all timers in the map will be cancelled and freed. 5152 * 5153 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn) 5154 * Description 5155 * Configure the timer to call *callback_fn* static function. 5156 * Return 5157 * 0 on success. 5158 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5159 * **-EPERM** if *timer* is in a map that doesn't have any user references. 5160 * The user space should either hold a file descriptor to a map with timers 5161 * or pin such map in bpffs. When map is unpinned or file descriptor is 5162 * closed all timers in the map will be cancelled and freed. 5163 * 5164 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags) 5165 * Description 5166 * Set timer expiration N nanoseconds from the current time. The 5167 * configured callback will be invoked in soft irq context on some cpu 5168 * and will not repeat unless another bpf_timer_start() is made. 5169 * In such case the next invocation can migrate to a different cpu. 5170 * Since struct bpf_timer is a field inside map element the map 5171 * owns the timer. The bpf_timer_set_callback() will increment refcnt 5172 * of BPF program to make sure that callback_fn code stays valid. 5173 * When user space reference to a map reaches zero all timers 5174 * in a map are cancelled and corresponding program's refcnts are 5175 * decremented. This is done to make sure that Ctrl-C of a user 5176 * process doesn't leave any timers running. If map is pinned in 5177 * bpffs the callback_fn can re-arm itself indefinitely. 5178 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands 5179 * cancel and free the timer in the given map element. 5180 * The map can contain timers that invoke callback_fn-s from different 5181 * programs. The same callback_fn can serve different timers from 5182 * different maps if key/value layout matches across maps. 5183 * Every bpf_timer_set_callback() can have different callback_fn. 5184 * 5185 * *flags* can be one of: 5186 * 5187 * **BPF_F_TIMER_ABS** 5188 * Start the timer in absolute expire value instead of the 5189 * default relative one. 5190 * **BPF_F_TIMER_CPU_PIN** 5191 * Timer will be pinned to the CPU of the caller. 5192 * 5193 * Return 5194 * 0 on success. 5195 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier 5196 * or invalid *flags* are passed. 5197 * 5198 * long bpf_timer_cancel(struct bpf_timer *timer) 5199 * Description 5200 * Cancel the timer and wait for callback_fn to finish if it was running. 5201 * Return 5202 * 0 if the timer was not active. 5203 * 1 if the timer was active. 5204 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier. 5205 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its 5206 * own timer which would have led to a deadlock otherwise. 5207 * 5208 * u64 bpf_get_func_ip(void *ctx) 5209 * Description 5210 * Get address of the traced function (for tracing and kprobe programs). 5211 * 5212 * When called for kprobe program attached as uprobe it returns 5213 * probe address for both entry and return uprobe. 5214 * 5215 * Return 5216 * Address of the traced function for kprobe. 5217 * 0 for kprobes placed within the function (not at the entry). 5218 * Address of the probe for uprobe and return uprobe. 5219 * 5220 * u64 bpf_get_attach_cookie(void *ctx) 5221 * Description 5222 * Get bpf_cookie value provided (optionally) during the program 5223 * attachment. It might be different for each individual 5224 * attachment, even if BPF program itself is the same. 5225 * Expects BPF program context *ctx* as a first argument. 5226 * 5227 * Supported for the following program types: 5228 * - kprobe/uprobe; 5229 * - tracepoint; 5230 * - perf_event. 5231 * Return 5232 * Value specified by user at BPF link creation/attachment time 5233 * or 0, if it was not specified. 5234 * 5235 * long bpf_task_pt_regs(struct task_struct *task) 5236 * Description 5237 * Get the struct pt_regs associated with **task**. 5238 * Return 5239 * A pointer to struct pt_regs. 5240 * 5241 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags) 5242 * Description 5243 * Get branch trace from hardware engines like Intel LBR. The 5244 * hardware engine is stopped shortly after the helper is 5245 * called. Therefore, the user need to filter branch entries 5246 * based on the actual use case. To capture branch trace 5247 * before the trigger point of the BPF program, the helper 5248 * should be called at the beginning of the BPF program. 5249 * 5250 * The data is stored as struct perf_branch_entry into output 5251 * buffer *entries*. *size* is the size of *entries* in bytes. 5252 * *flags* is reserved for now and must be zero. 5253 * 5254 * Return 5255 * On success, number of bytes written to *buf*. On error, a 5256 * negative value. 5257 * 5258 * **-EINVAL** if *flags* is not zero. 5259 * 5260 * **-ENOENT** if architecture does not support branch records. 5261 * 5262 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len) 5263 * Description 5264 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64 5265 * to format and can handle more format args as a result. 5266 * 5267 * Arguments are to be used as in **bpf_seq_printf**\ () helper. 5268 * Return 5269 * The number of bytes written to the buffer, or a negative error 5270 * in case of failure. 5271 * 5272 * struct unix_sock *bpf_skc_to_unix_sock(void *sk) 5273 * Description 5274 * Dynamically cast a *sk* pointer to a *unix_sock* pointer. 5275 * Return 5276 * *sk* if casting is valid, or **NULL** otherwise. 5277 * 5278 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res) 5279 * Description 5280 * Get the address of a kernel symbol, returned in *res*. *res* is 5281 * set to 0 if the symbol is not found. 5282 * Return 5283 * On success, zero. On error, a negative value. 5284 * 5285 * **-EINVAL** if *flags* is not zero. 5286 * 5287 * **-EINVAL** if string *name* is not the same size as *name_sz*. 5288 * 5289 * **-ENOENT** if symbol is not found. 5290 * 5291 * **-EPERM** if caller does not have permission to obtain kernel address. 5292 * 5293 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags) 5294 * Description 5295 * Find vma of *task* that contains *addr*, call *callback_fn* 5296 * function with *task*, *vma*, and *callback_ctx*. 5297 * The *callback_fn* should be a static function and 5298 * the *callback_ctx* should be a pointer to the stack. 5299 * The *flags* is used to control certain aspects of the helper. 5300 * Currently, the *flags* must be 0. 5301 * 5302 * The expected callback signature is 5303 * 5304 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx); 5305 * 5306 * Return 5307 * 0 on success. 5308 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*. 5309 * **-EBUSY** if failed to try lock mmap_lock. 5310 * **-EINVAL** for invalid **flags**. 5311 * 5312 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags) 5313 * Description 5314 * For **nr_loops**, call **callback_fn** function 5315 * with **callback_ctx** as the context parameter. 5316 * The **callback_fn** should be a static function and 5317 * the **callback_ctx** should be a pointer to the stack. 5318 * The **flags** is used to control certain aspects of the helper. 5319 * Currently, the **flags** must be 0. Currently, nr_loops is 5320 * limited to 1 << 23 (~8 million) loops. 5321 * 5322 * long (\*callback_fn)(u32 index, void \*ctx); 5323 * 5324 * where **index** is the current index in the loop. The index 5325 * is zero-indexed. 5326 * 5327 * If **callback_fn** returns 0, the helper will continue to the next 5328 * loop. If return value is 1, the helper will skip the rest of 5329 * the loops and return. Other return values are not used now, 5330 * and will be rejected by the verifier. 5331 * 5332 * Return 5333 * The number of loops performed, **-EINVAL** for invalid **flags**, 5334 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops. 5335 * 5336 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2) 5337 * Description 5338 * Do strncmp() between **s1** and **s2**. **s1** doesn't need 5339 * to be null-terminated and **s1_sz** is the maximum storage 5340 * size of **s1**. **s2** must be a read-only string. 5341 * Return 5342 * An integer less than, equal to, or greater than zero 5343 * if the first **s1_sz** bytes of **s1** is found to be 5344 * less than, to match, or be greater than **s2**. 5345 * 5346 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value) 5347 * Description 5348 * Get **n**-th argument register (zero based) of the traced function (for tracing programs) 5349 * returned in **value**. 5350 * 5351 * Return 5352 * 0 on success. 5353 * **-EINVAL** if n >= argument register count of traced function. 5354 * 5355 * long bpf_get_func_ret(void *ctx, u64 *value) 5356 * Description 5357 * Get return value of the traced function (for tracing programs) 5358 * in **value**. 5359 * 5360 * Return 5361 * 0 on success. 5362 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN. 5363 * 5364 * long bpf_get_func_arg_cnt(void *ctx) 5365 * Description 5366 * Get number of registers of the traced function (for tracing programs) where 5367 * function arguments are stored in these registers. 5368 * 5369 * Return 5370 * The number of argument registers of the traced function. 5371 * 5372 * int bpf_get_retval(void) 5373 * Description 5374 * Get the BPF program's return value that will be returned to the upper layers. 5375 * 5376 * This helper is currently supported by cgroup programs and only by the hooks 5377 * where BPF program's return value is returned to the userspace via errno. 5378 * Return 5379 * The BPF program's return value. 5380 * 5381 * int bpf_set_retval(int retval) 5382 * Description 5383 * Set the BPF program's return value that will be returned to the upper layers. 5384 * 5385 * This helper is currently supported by cgroup programs and only by the hooks 5386 * where BPF program's return value is returned to the userspace via errno. 5387 * 5388 * Note that there is the following corner case where the program exports an error 5389 * via bpf_set_retval but signals success via 'return 1': 5390 * 5391 * bpf_set_retval(-EPERM); 5392 * return 1; 5393 * 5394 * In this case, the BPF program's return value will use helper's -EPERM. This 5395 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case. 5396 * 5397 * Return 5398 * 0 on success, or a negative error in case of failure. 5399 * 5400 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md) 5401 * Description 5402 * Get the total size of a given xdp buff (linear and paged area) 5403 * Return 5404 * The total size of a given xdp buffer. 5405 * 5406 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5407 * Description 5408 * This helper is provided as an easy way to load data from a 5409 * xdp buffer. It can be used to load *len* bytes from *offset* from 5410 * the frame associated to *xdp_md*, into the buffer pointed by 5411 * *buf*. 5412 * Return 5413 * 0 on success, or a negative error in case of failure. 5414 * 5415 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len) 5416 * Description 5417 * Store *len* bytes from buffer *buf* into the frame 5418 * associated to *xdp_md*, at *offset*. 5419 * Return 5420 * 0 on success, or a negative error in case of failure. 5421 * 5422 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags) 5423 * Description 5424 * Read *size* bytes from user space address *user_ptr* in *tsk*'s 5425 * address space, and stores the data in *dst*. *flags* is not 5426 * used yet and is provided for future extensibility. This helper 5427 * can only be used by sleepable programs. 5428 * Return 5429 * 0 on success, or a negative error in case of failure. On error 5430 * *dst* buffer is zeroed out. 5431 * 5432 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type) 5433 * Description 5434 * Change the __sk_buff->tstamp_type to *tstamp_type* 5435 * and set *tstamp* to the __sk_buff->tstamp together. 5436 * 5437 * If there is no need to change the __sk_buff->tstamp_type, 5438 * the tstamp value can be directly written to __sk_buff->tstamp 5439 * instead. 5440 * 5441 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that 5442 * will be kept during bpf_redirect_*(). A non zero 5443 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO 5444 * *tstamp_type*. 5445 * 5446 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used 5447 * with a zero *tstamp*. 5448 * 5449 * Only IPv4 and IPv6 skb->protocol are supported. 5450 * 5451 * This function is most useful when it needs to set a 5452 * mono delivery time to __sk_buff->tstamp and then 5453 * bpf_redirect_*() to the egress of an iface. For example, 5454 * changing the (rcv) timestamp in __sk_buff->tstamp at 5455 * ingress to a mono delivery time and then bpf_redirect_*() 5456 * to sch_fq@phy-dev. 5457 * Return 5458 * 0 on success. 5459 * **-EINVAL** for invalid input 5460 * **-EOPNOTSUPP** for unsupported protocol 5461 * 5462 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size) 5463 * Description 5464 * Returns a calculated IMA hash of the *file*. 5465 * If the hash is larger than *size*, then only *size* 5466 * bytes will be copied to *dst* 5467 * Return 5468 * The **hash_algo** is returned on success, 5469 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if 5470 * invalid arguments are passed. 5471 * 5472 * void *bpf_kptr_xchg(void *map_value, void *ptr) 5473 * Description 5474 * Exchange kptr at pointer *map_value* with *ptr*, and return the 5475 * old value. *ptr* can be NULL, otherwise it must be a referenced 5476 * pointer which will be released when this helper is called. 5477 * Return 5478 * The old value of kptr (which can be NULL). The returned pointer 5479 * if not NULL, is a reference which must be released using its 5480 * corresponding release function, or moved into a BPF map before 5481 * program exit. 5482 * 5483 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu) 5484 * Description 5485 * Perform a lookup in *percpu map* for an entry associated to 5486 * *key* on *cpu*. 5487 * Return 5488 * Map value associated to *key* on *cpu*, or **NULL** if no entry 5489 * was found or *cpu* is invalid. 5490 * 5491 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk) 5492 * Description 5493 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer. 5494 * Return 5495 * *sk* if casting is valid, or **NULL** otherwise. 5496 * 5497 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr) 5498 * Description 5499 * Get a dynptr to local memory *data*. 5500 * 5501 * *data* must be a ptr to a map value. 5502 * The maximum *size* supported is DYNPTR_MAX_SIZE. 5503 * *flags* is currently unused. 5504 * Return 5505 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE, 5506 * -EINVAL if flags is not 0. 5507 * 5508 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr) 5509 * Description 5510 * Reserve *size* bytes of payload in a ring buffer *ringbuf* 5511 * through the dynptr interface. *flags* must be 0. 5512 * 5513 * Please note that a corresponding bpf_ringbuf_submit_dynptr or 5514 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the 5515 * reservation fails. This is enforced by the verifier. 5516 * Return 5517 * 0 on success, or a negative error in case of failure. 5518 * 5519 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags) 5520 * Description 5521 * Submit reserved ring buffer sample, pointed to by *data*, 5522 * through the dynptr interface. This is a no-op if the dynptr is 5523 * invalid/null. 5524 * 5525 * For more information on *flags*, please see 5526 * 'bpf_ringbuf_submit'. 5527 * Return 5528 * Nothing. Always succeeds. 5529 * 5530 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags) 5531 * Description 5532 * Discard reserved ring buffer sample through the dynptr 5533 * interface. This is a no-op if the dynptr is invalid/null. 5534 * 5535 * For more information on *flags*, please see 5536 * 'bpf_ringbuf_discard'. 5537 * Return 5538 * Nothing. Always succeeds. 5539 * 5540 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags) 5541 * Description 5542 * Read *len* bytes from *src* into *dst*, starting from *offset* 5543 * into *src*. 5544 * *flags* is currently unused. 5545 * Return 5546 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5547 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if 5548 * *flags* is not 0. 5549 * 5550 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags) 5551 * Description 5552 * Write *len* bytes from *src* into *dst*, starting from *offset* 5553 * into *dst*. 5554 * 5555 * *flags* must be 0 except for skb-type dynptrs. 5556 * 5557 * For skb-type dynptrs: 5558 * * All data slices of the dynptr are automatically 5559 * invalidated after **bpf_dynptr_write**\ (). This is 5560 * because writing may pull the skb and change the 5561 * underlying packet buffer. 5562 * 5563 * * For *flags*, please see the flags accepted by 5564 * **bpf_skb_store_bytes**\ (). 5565 * Return 5566 * 0 on success, -E2BIG if *offset* + *len* exceeds the length 5567 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst* 5568 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs, 5569 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ (). 5570 * 5571 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len) 5572 * Description 5573 * Get a pointer to the underlying dynptr data. 5574 * 5575 * *len* must be a statically known value. The returned data slice 5576 * is invalidated whenever the dynptr is invalidated. 5577 * 5578 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should 5579 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr. 5580 * Return 5581 * Pointer to the underlying dynptr data, NULL if the dynptr is 5582 * read-only, if the dynptr is invalid, or if the offset and length 5583 * is out of bounds. 5584 * 5585 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len) 5586 * Description 5587 * Try to issue a SYN cookie for the packet with corresponding 5588 * IPv4/TCP headers, *iph* and *th*, without depending on a 5589 * listening socket. 5590 * 5591 * *iph* points to the IPv4 header. 5592 * 5593 * *th* points to the start of the TCP header, while *th_len* 5594 * contains the length of the TCP header (at least 5595 * **sizeof**\ (**struct tcphdr**)). 5596 * Return 5597 * On success, lower 32 bits hold the generated SYN cookie in 5598 * followed by 16 bits which hold the MSS value for that cookie, 5599 * and the top 16 bits are unused. 5600 * 5601 * On failure, the returned value is one of the following: 5602 * 5603 * **-EINVAL** if *th_len* is invalid. 5604 * 5605 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len) 5606 * Description 5607 * Try to issue a SYN cookie for the packet with corresponding 5608 * IPv6/TCP headers, *iph* and *th*, without depending on a 5609 * listening socket. 5610 * 5611 * *iph* points to the IPv6 header. 5612 * 5613 * *th* points to the start of the TCP header, while *th_len* 5614 * contains the length of the TCP header (at least 5615 * **sizeof**\ (**struct tcphdr**)). 5616 * Return 5617 * On success, lower 32 bits hold the generated SYN cookie in 5618 * followed by 16 bits which hold the MSS value for that cookie, 5619 * and the top 16 bits are unused. 5620 * 5621 * On failure, the returned value is one of the following: 5622 * 5623 * **-EINVAL** if *th_len* is invalid. 5624 * 5625 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5626 * 5627 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th) 5628 * Description 5629 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5630 * without depending on a listening socket. 5631 * 5632 * *iph* points to the IPv4 header. 5633 * 5634 * *th* points to the TCP header. 5635 * Return 5636 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5637 * 5638 * On failure, the returned value is one of the following: 5639 * 5640 * **-EACCES** if the SYN cookie is not valid. 5641 * 5642 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th) 5643 * Description 5644 * Check whether *iph* and *th* contain a valid SYN cookie ACK 5645 * without depending on a listening socket. 5646 * 5647 * *iph* points to the IPv6 header. 5648 * 5649 * *th* points to the TCP header. 5650 * Return 5651 * 0 if *iph* and *th* are a valid SYN cookie ACK. 5652 * 5653 * On failure, the returned value is one of the following: 5654 * 5655 * **-EACCES** if the SYN cookie is not valid. 5656 * 5657 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin. 5658 * 5659 * u64 bpf_ktime_get_tai_ns(void) 5660 * Description 5661 * A nonsettable system-wide clock derived from wall-clock time but 5662 * ignoring leap seconds. This clock does not experience 5663 * discontinuities and backwards jumps caused by NTP inserting leap 5664 * seconds as CLOCK_REALTIME does. 5665 * 5666 * See: **clock_gettime**\ (**CLOCK_TAI**) 5667 * Return 5668 * Current *ktime*. 5669 * 5670 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags) 5671 * Description 5672 * Drain samples from the specified user ring buffer, and invoke 5673 * the provided callback for each such sample: 5674 * 5675 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx); 5676 * 5677 * If **callback_fn** returns 0, the helper will continue to try 5678 * and drain the next sample, up to a maximum of 5679 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1, 5680 * the helper will skip the rest of the samples and return. Other 5681 * return values are not used now, and will be rejected by the 5682 * verifier. 5683 * Return 5684 * The number of drained samples if no error was encountered while 5685 * draining samples, or 0 if no samples were present in the ring 5686 * buffer. If a user-space producer was epoll-waiting on this map, 5687 * and at least one sample was drained, they will receive an event 5688 * notification notifying them of available space in the ring 5689 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this 5690 * function, no wakeup notification will be sent. If the 5691 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will 5692 * be sent even if no sample was drained. 5693 * 5694 * On failure, the returned value is one of the following: 5695 * 5696 * **-EBUSY** if the ring buffer is contended, and another calling 5697 * context was concurrently draining the ring buffer. 5698 * 5699 * **-EINVAL** if user-space is not properly tracking the ring 5700 * buffer due to the producer position not being aligned to 8 5701 * bytes, a sample not being aligned to 8 bytes, or the producer 5702 * position not matching the advertised length of a sample. 5703 * 5704 * **-E2BIG** if user-space has tried to publish a sample which is 5705 * larger than the size of the ring buffer, or which cannot fit 5706 * within a struct bpf_dynptr. 5707 * 5708 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags) 5709 * Description 5710 * Get a bpf_local_storage from the *cgroup*. 5711 * 5712 * Logically, it could be thought of as getting the value from 5713 * a *map* with *cgroup* as the **key**. From this 5714 * perspective, the usage is not much different from 5715 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this 5716 * helper enforces the key must be a cgroup struct and the map must also 5717 * be a **BPF_MAP_TYPE_CGRP_STORAGE**. 5718 * 5719 * In reality, the local-storage value is embedded directly inside of the 5720 * *cgroup* object itself, rather than being located in the 5721 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is 5722 * queried for some *map* on a *cgroup* object, the kernel will perform an 5723 * O(n) iteration over all of the live local-storage values for that 5724 * *cgroup* object until the local-storage value for the *map* is found. 5725 * 5726 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 5727 * used such that a new bpf_local_storage will be 5728 * created if one does not exist. *value* can be used 5729 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 5730 * the initial value of a bpf_local_storage. If *value* is 5731 * **NULL**, the new bpf_local_storage will be zero initialized. 5732 * Return 5733 * A bpf_local_storage pointer is returned on success. 5734 * 5735 * **NULL** if not found or there was an error in adding 5736 * a new bpf_local_storage. 5737 * 5738 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup) 5739 * Description 5740 * Delete a bpf_local_storage from a *cgroup*. 5741 * Return 5742 * 0 on success. 5743 * 5744 * **-ENOENT** if the bpf_local_storage cannot be found. 5745 */ 5746 #define ___BPF_FUNC_MAPPER(FN, ctx...) \ 5747 FN(unspec, 0, ##ctx) \ 5748 FN(map_lookup_elem, 1, ##ctx) \ 5749 FN(map_update_elem, 2, ##ctx) \ 5750 FN(map_delete_elem, 3, ##ctx) \ 5751 FN(probe_read, 4, ##ctx) \ 5752 FN(ktime_get_ns, 5, ##ctx) \ 5753 FN(trace_printk, 6, ##ctx) \ 5754 FN(get_prandom_u32, 7, ##ctx) \ 5755 FN(get_smp_processor_id, 8, ##ctx) \ 5756 FN(skb_store_bytes, 9, ##ctx) \ 5757 FN(l3_csum_replace, 10, ##ctx) \ 5758 FN(l4_csum_replace, 11, ##ctx) \ 5759 FN(tail_call, 12, ##ctx) \ 5760 FN(clone_redirect, 13, ##ctx) \ 5761 FN(get_current_pid_tgid, 14, ##ctx) \ 5762 FN(get_current_uid_gid, 15, ##ctx) \ 5763 FN(get_current_comm, 16, ##ctx) \ 5764 FN(get_cgroup_classid, 17, ##ctx) \ 5765 FN(skb_vlan_push, 18, ##ctx) \ 5766 FN(skb_vlan_pop, 19, ##ctx) \ 5767 FN(skb_get_tunnel_key, 20, ##ctx) \ 5768 FN(skb_set_tunnel_key, 21, ##ctx) \ 5769 FN(perf_event_read, 22, ##ctx) \ 5770 FN(redirect, 23, ##ctx) \ 5771 FN(get_route_realm, 24, ##ctx) \ 5772 FN(perf_event_output, 25, ##ctx) \ 5773 FN(skb_load_bytes, 26, ##ctx) \ 5774 FN(get_stackid, 27, ##ctx) \ 5775 FN(csum_diff, 28, ##ctx) \ 5776 FN(skb_get_tunnel_opt, 29, ##ctx) \ 5777 FN(skb_set_tunnel_opt, 30, ##ctx) \ 5778 FN(skb_change_proto, 31, ##ctx) \ 5779 FN(skb_change_type, 32, ##ctx) \ 5780 FN(skb_under_cgroup, 33, ##ctx) \ 5781 FN(get_hash_recalc, 34, ##ctx) \ 5782 FN(get_current_task, 35, ##ctx) \ 5783 FN(probe_write_user, 36, ##ctx) \ 5784 FN(current_task_under_cgroup, 37, ##ctx) \ 5785 FN(skb_change_tail, 38, ##ctx) \ 5786 FN(skb_pull_data, 39, ##ctx) \ 5787 FN(csum_update, 40, ##ctx) \ 5788 FN(set_hash_invalid, 41, ##ctx) \ 5789 FN(get_numa_node_id, 42, ##ctx) \ 5790 FN(skb_change_head, 43, ##ctx) \ 5791 FN(xdp_adjust_head, 44, ##ctx) \ 5792 FN(probe_read_str, 45, ##ctx) \ 5793 FN(get_socket_cookie, 46, ##ctx) \ 5794 FN(get_socket_uid, 47, ##ctx) \ 5795 FN(set_hash, 48, ##ctx) \ 5796 FN(setsockopt, 49, ##ctx) \ 5797 FN(skb_adjust_room, 50, ##ctx) \ 5798 FN(redirect_map, 51, ##ctx) \ 5799 FN(sk_redirect_map, 52, ##ctx) \ 5800 FN(sock_map_update, 53, ##ctx) \ 5801 FN(xdp_adjust_meta, 54, ##ctx) \ 5802 FN(perf_event_read_value, 55, ##ctx) \ 5803 FN(perf_prog_read_value, 56, ##ctx) \ 5804 FN(getsockopt, 57, ##ctx) \ 5805 FN(override_return, 58, ##ctx) \ 5806 FN(sock_ops_cb_flags_set, 59, ##ctx) \ 5807 FN(msg_redirect_map, 60, ##ctx) \ 5808 FN(msg_apply_bytes, 61, ##ctx) \ 5809 FN(msg_cork_bytes, 62, ##ctx) \ 5810 FN(msg_pull_data, 63, ##ctx) \ 5811 FN(bind, 64, ##ctx) \ 5812 FN(xdp_adjust_tail, 65, ##ctx) \ 5813 FN(skb_get_xfrm_state, 66, ##ctx) \ 5814 FN(get_stack, 67, ##ctx) \ 5815 FN(skb_load_bytes_relative, 68, ##ctx) \ 5816 FN(fib_lookup, 69, ##ctx) \ 5817 FN(sock_hash_update, 70, ##ctx) \ 5818 FN(msg_redirect_hash, 71, ##ctx) \ 5819 FN(sk_redirect_hash, 72, ##ctx) \ 5820 FN(lwt_push_encap, 73, ##ctx) \ 5821 FN(lwt_seg6_store_bytes, 74, ##ctx) \ 5822 FN(lwt_seg6_adjust_srh, 75, ##ctx) \ 5823 FN(lwt_seg6_action, 76, ##ctx) \ 5824 FN(rc_repeat, 77, ##ctx) \ 5825 FN(rc_keydown, 78, ##ctx) \ 5826 FN(skb_cgroup_id, 79, ##ctx) \ 5827 FN(get_current_cgroup_id, 80, ##ctx) \ 5828 FN(get_local_storage, 81, ##ctx) \ 5829 FN(sk_select_reuseport, 82, ##ctx) \ 5830 FN(skb_ancestor_cgroup_id, 83, ##ctx) \ 5831 FN(sk_lookup_tcp, 84, ##ctx) \ 5832 FN(sk_lookup_udp, 85, ##ctx) \ 5833 FN(sk_release, 86, ##ctx) \ 5834 FN(map_push_elem, 87, ##ctx) \ 5835 FN(map_pop_elem, 88, ##ctx) \ 5836 FN(map_peek_elem, 89, ##ctx) \ 5837 FN(msg_push_data, 90, ##ctx) \ 5838 FN(msg_pop_data, 91, ##ctx) \ 5839 FN(rc_pointer_rel, 92, ##ctx) \ 5840 FN(spin_lock, 93, ##ctx) \ 5841 FN(spin_unlock, 94, ##ctx) \ 5842 FN(sk_fullsock, 95, ##ctx) \ 5843 FN(tcp_sock, 96, ##ctx) \ 5844 FN(skb_ecn_set_ce, 97, ##ctx) \ 5845 FN(get_listener_sock, 98, ##ctx) \ 5846 FN(skc_lookup_tcp, 99, ##ctx) \ 5847 FN(tcp_check_syncookie, 100, ##ctx) \ 5848 FN(sysctl_get_name, 101, ##ctx) \ 5849 FN(sysctl_get_current_value, 102, ##ctx) \ 5850 FN(sysctl_get_new_value, 103, ##ctx) \ 5851 FN(sysctl_set_new_value, 104, ##ctx) \ 5852 FN(strtol, 105, ##ctx) \ 5853 FN(strtoul, 106, ##ctx) \ 5854 FN(sk_storage_get, 107, ##ctx) \ 5855 FN(sk_storage_delete, 108, ##ctx) \ 5856 FN(send_signal, 109, ##ctx) \ 5857 FN(tcp_gen_syncookie, 110, ##ctx) \ 5858 FN(skb_output, 111, ##ctx) \ 5859 FN(probe_read_user, 112, ##ctx) \ 5860 FN(probe_read_kernel, 113, ##ctx) \ 5861 FN(probe_read_user_str, 114, ##ctx) \ 5862 FN(probe_read_kernel_str, 115, ##ctx) \ 5863 FN(tcp_send_ack, 116, ##ctx) \ 5864 FN(send_signal_thread, 117, ##ctx) \ 5865 FN(jiffies64, 118, ##ctx) \ 5866 FN(read_branch_records, 119, ##ctx) \ 5867 FN(get_ns_current_pid_tgid, 120, ##ctx) \ 5868 FN(xdp_output, 121, ##ctx) \ 5869 FN(get_netns_cookie, 122, ##ctx) \ 5870 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \ 5871 FN(sk_assign, 124, ##ctx) \ 5872 FN(ktime_get_boot_ns, 125, ##ctx) \ 5873 FN(seq_printf, 126, ##ctx) \ 5874 FN(seq_write, 127, ##ctx) \ 5875 FN(sk_cgroup_id, 128, ##ctx) \ 5876 FN(sk_ancestor_cgroup_id, 129, ##ctx) \ 5877 FN(ringbuf_output, 130, ##ctx) \ 5878 FN(ringbuf_reserve, 131, ##ctx) \ 5879 FN(ringbuf_submit, 132, ##ctx) \ 5880 FN(ringbuf_discard, 133, ##ctx) \ 5881 FN(ringbuf_query, 134, ##ctx) \ 5882 FN(csum_level, 135, ##ctx) \ 5883 FN(skc_to_tcp6_sock, 136, ##ctx) \ 5884 FN(skc_to_tcp_sock, 137, ##ctx) \ 5885 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \ 5886 FN(skc_to_tcp_request_sock, 139, ##ctx) \ 5887 FN(skc_to_udp6_sock, 140, ##ctx) \ 5888 FN(get_task_stack, 141, ##ctx) \ 5889 FN(load_hdr_opt, 142, ##ctx) \ 5890 FN(store_hdr_opt, 143, ##ctx) \ 5891 FN(reserve_hdr_opt, 144, ##ctx) \ 5892 FN(inode_storage_get, 145, ##ctx) \ 5893 FN(inode_storage_delete, 146, ##ctx) \ 5894 FN(d_path, 147, ##ctx) \ 5895 FN(copy_from_user, 148, ##ctx) \ 5896 FN(snprintf_btf, 149, ##ctx) \ 5897 FN(seq_printf_btf, 150, ##ctx) \ 5898 FN(skb_cgroup_classid, 151, ##ctx) \ 5899 FN(redirect_neigh, 152, ##ctx) \ 5900 FN(per_cpu_ptr, 153, ##ctx) \ 5901 FN(this_cpu_ptr, 154, ##ctx) \ 5902 FN(redirect_peer, 155, ##ctx) \ 5903 FN(task_storage_get, 156, ##ctx) \ 5904 FN(task_storage_delete, 157, ##ctx) \ 5905 FN(get_current_task_btf, 158, ##ctx) \ 5906 FN(bprm_opts_set, 159, ##ctx) \ 5907 FN(ktime_get_coarse_ns, 160, ##ctx) \ 5908 FN(ima_inode_hash, 161, ##ctx) \ 5909 FN(sock_from_file, 162, ##ctx) \ 5910 FN(check_mtu, 163, ##ctx) \ 5911 FN(for_each_map_elem, 164, ##ctx) \ 5912 FN(snprintf, 165, ##ctx) \ 5913 FN(sys_bpf, 166, ##ctx) \ 5914 FN(btf_find_by_name_kind, 167, ##ctx) \ 5915 FN(sys_close, 168, ##ctx) \ 5916 FN(timer_init, 169, ##ctx) \ 5917 FN(timer_set_callback, 170, ##ctx) \ 5918 FN(timer_start, 171, ##ctx) \ 5919 FN(timer_cancel, 172, ##ctx) \ 5920 FN(get_func_ip, 173, ##ctx) \ 5921 FN(get_attach_cookie, 174, ##ctx) \ 5922 FN(task_pt_regs, 175, ##ctx) \ 5923 FN(get_branch_snapshot, 176, ##ctx) \ 5924 FN(trace_vprintk, 177, ##ctx) \ 5925 FN(skc_to_unix_sock, 178, ##ctx) \ 5926 FN(kallsyms_lookup_name, 179, ##ctx) \ 5927 FN(find_vma, 180, ##ctx) \ 5928 FN(loop, 181, ##ctx) \ 5929 FN(strncmp, 182, ##ctx) \ 5930 FN(get_func_arg, 183, ##ctx) \ 5931 FN(get_func_ret, 184, ##ctx) \ 5932 FN(get_func_arg_cnt, 185, ##ctx) \ 5933 FN(get_retval, 186, ##ctx) \ 5934 FN(set_retval, 187, ##ctx) \ 5935 FN(xdp_get_buff_len, 188, ##ctx) \ 5936 FN(xdp_load_bytes, 189, ##ctx) \ 5937 FN(xdp_store_bytes, 190, ##ctx) \ 5938 FN(copy_from_user_task, 191, ##ctx) \ 5939 FN(skb_set_tstamp, 192, ##ctx) \ 5940 FN(ima_file_hash, 193, ##ctx) \ 5941 FN(kptr_xchg, 194, ##ctx) \ 5942 FN(map_lookup_percpu_elem, 195, ##ctx) \ 5943 FN(skc_to_mptcp_sock, 196, ##ctx) \ 5944 FN(dynptr_from_mem, 197, ##ctx) \ 5945 FN(ringbuf_reserve_dynptr, 198, ##ctx) \ 5946 FN(ringbuf_submit_dynptr, 199, ##ctx) \ 5947 FN(ringbuf_discard_dynptr, 200, ##ctx) \ 5948 FN(dynptr_read, 201, ##ctx) \ 5949 FN(dynptr_write, 202, ##ctx) \ 5950 FN(dynptr_data, 203, ##ctx) \ 5951 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \ 5952 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \ 5953 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \ 5954 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \ 5955 FN(ktime_get_tai_ns, 208, ##ctx) \ 5956 FN(user_ringbuf_drain, 209, ##ctx) \ 5957 FN(cgrp_storage_get, 210, ##ctx) \ 5958 FN(cgrp_storage_delete, 211, ##ctx) \ 5959 /* */ 5960 5961 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't 5962 * know or care about integer value that is now passed as second argument 5963 */ 5964 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name), 5965 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN) 5966 5967 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 5968 * function eBPF program intends to call 5969 */ 5970 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y, 5971 enum bpf_func_id { 5972 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN) 5973 __BPF_FUNC_MAX_ID, 5974 }; 5975 #undef __BPF_ENUM_FN 5976 5977 /* All flags used by eBPF helper functions, placed here. */ 5978 5979 /* BPF_FUNC_skb_store_bytes flags. */ 5980 enum { 5981 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 5982 BPF_F_INVALIDATE_HASH = (1ULL << 1), 5983 }; 5984 5985 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 5986 * First 4 bits are for passing the header field size. 5987 */ 5988 enum { 5989 BPF_F_HDR_FIELD_MASK = 0xfULL, 5990 }; 5991 5992 /* BPF_FUNC_l4_csum_replace flags. */ 5993 enum { 5994 BPF_F_PSEUDO_HDR = (1ULL << 4), 5995 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 5996 BPF_F_MARK_ENFORCE = (1ULL << 6), 5997 }; 5998 5999 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 6000 enum { 6001 BPF_F_INGRESS = (1ULL << 0), 6002 }; 6003 6004 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 6005 enum { 6006 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 6007 }; 6008 6009 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 6010 enum { 6011 BPF_F_SKIP_FIELD_MASK = 0xffULL, 6012 BPF_F_USER_STACK = (1ULL << 8), 6013 /* flags used by BPF_FUNC_get_stackid only. */ 6014 BPF_F_FAST_STACK_CMP = (1ULL << 9), 6015 BPF_F_REUSE_STACKID = (1ULL << 10), 6016 /* flags used by BPF_FUNC_get_stack only. */ 6017 BPF_F_USER_BUILD_ID = (1ULL << 11), 6018 }; 6019 6020 /* BPF_FUNC_skb_set_tunnel_key flags. */ 6021 enum { 6022 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 6023 BPF_F_DONT_FRAGMENT = (1ULL << 2), 6024 BPF_F_SEQ_NUMBER = (1ULL << 3), 6025 BPF_F_NO_TUNNEL_KEY = (1ULL << 4), 6026 }; 6027 6028 /* BPF_FUNC_skb_get_tunnel_key flags. */ 6029 enum { 6030 BPF_F_TUNINFO_FLAGS = (1ULL << 4), 6031 }; 6032 6033 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 6034 * BPF_FUNC_perf_event_read_value flags. 6035 */ 6036 enum { 6037 BPF_F_INDEX_MASK = 0xffffffffULL, 6038 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 6039 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 6040 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 6041 }; 6042 6043 /* Current network namespace */ 6044 enum { 6045 BPF_F_CURRENT_NETNS = (-1L), 6046 }; 6047 6048 /* BPF_FUNC_csum_level level values. */ 6049 enum { 6050 BPF_CSUM_LEVEL_QUERY, 6051 BPF_CSUM_LEVEL_INC, 6052 BPF_CSUM_LEVEL_DEC, 6053 BPF_CSUM_LEVEL_RESET, 6054 }; 6055 6056 /* BPF_FUNC_skb_adjust_room flags. */ 6057 enum { 6058 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 6059 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 6060 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 6061 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 6062 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 6063 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 6064 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 6065 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7), 6066 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8), 6067 }; 6068 6069 enum { 6070 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 6071 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 6072 }; 6073 6074 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 6075 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 6076 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 6077 6078 /* BPF_FUNC_sysctl_get_name flags. */ 6079 enum { 6080 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 6081 }; 6082 6083 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 6084 enum { 6085 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 6086 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 6087 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 6088 */ 6089 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 6090 }; 6091 6092 /* BPF_FUNC_read_branch_records flags. */ 6093 enum { 6094 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 6095 }; 6096 6097 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 6098 * BPF_FUNC_bpf_ringbuf_output flags. 6099 */ 6100 enum { 6101 BPF_RB_NO_WAKEUP = (1ULL << 0), 6102 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 6103 }; 6104 6105 /* BPF_FUNC_bpf_ringbuf_query flags */ 6106 enum { 6107 BPF_RB_AVAIL_DATA = 0, 6108 BPF_RB_RING_SIZE = 1, 6109 BPF_RB_CONS_POS = 2, 6110 BPF_RB_PROD_POS = 3, 6111 }; 6112 6113 /* BPF ring buffer constants */ 6114 enum { 6115 BPF_RINGBUF_BUSY_BIT = (1U << 31), 6116 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 6117 BPF_RINGBUF_HDR_SZ = 8, 6118 }; 6119 6120 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 6121 enum { 6122 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 6123 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 6124 }; 6125 6126 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 6127 enum bpf_adj_room_mode { 6128 BPF_ADJ_ROOM_NET, 6129 BPF_ADJ_ROOM_MAC, 6130 }; 6131 6132 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 6133 enum bpf_hdr_start_off { 6134 BPF_HDR_START_MAC, 6135 BPF_HDR_START_NET, 6136 }; 6137 6138 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 6139 enum bpf_lwt_encap_mode { 6140 BPF_LWT_ENCAP_SEG6, 6141 BPF_LWT_ENCAP_SEG6_INLINE, 6142 BPF_LWT_ENCAP_IP, 6143 }; 6144 6145 /* Flags for bpf_bprm_opts_set helper */ 6146 enum { 6147 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 6148 }; 6149 6150 /* Flags for bpf_redirect_map helper */ 6151 enum { 6152 BPF_F_BROADCAST = (1ULL << 3), 6153 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), 6154 }; 6155 6156 #define __bpf_md_ptr(type, name) \ 6157 union { \ 6158 type name; \ 6159 __u64 :64; \ 6160 } __attribute__((aligned(8))) 6161 6162 enum { 6163 BPF_SKB_TSTAMP_UNSPEC, 6164 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */ 6165 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle, 6166 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC 6167 * and try to deduce it by ingress, egress or skb->sk->sk_clockid. 6168 */ 6169 }; 6170 6171 /* user accessible mirror of in-kernel sk_buff. 6172 * new fields can only be added to the end of this structure 6173 */ 6174 struct __sk_buff { 6175 __u32 len; 6176 __u32 pkt_type; 6177 __u32 mark; 6178 __u32 queue_mapping; 6179 __u32 protocol; 6180 __u32 vlan_present; 6181 __u32 vlan_tci; 6182 __u32 vlan_proto; 6183 __u32 priority; 6184 __u32 ingress_ifindex; 6185 __u32 ifindex; 6186 __u32 tc_index; 6187 __u32 cb[5]; 6188 __u32 hash; 6189 __u32 tc_classid; 6190 __u32 data; 6191 __u32 data_end; 6192 __u32 napi_id; 6193 6194 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 6195 __u32 family; 6196 __u32 remote_ip4; /* Stored in network byte order */ 6197 __u32 local_ip4; /* Stored in network byte order */ 6198 __u32 remote_ip6[4]; /* Stored in network byte order */ 6199 __u32 local_ip6[4]; /* Stored in network byte order */ 6200 __u32 remote_port; /* Stored in network byte order */ 6201 __u32 local_port; /* stored in host byte order */ 6202 /* ... here. */ 6203 6204 __u32 data_meta; 6205 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 6206 __u64 tstamp; 6207 __u32 wire_len; 6208 __u32 gso_segs; 6209 __bpf_md_ptr(struct bpf_sock *, sk); 6210 __u32 gso_size; 6211 __u8 tstamp_type; 6212 __u32 :24; /* Padding, future use. */ 6213 __u64 hwtstamp; 6214 }; 6215 6216 struct bpf_tunnel_key { 6217 __u32 tunnel_id; 6218 union { 6219 __u32 remote_ipv4; 6220 __u32 remote_ipv6[4]; 6221 }; 6222 __u8 tunnel_tos; 6223 __u8 tunnel_ttl; 6224 union { 6225 __u16 tunnel_ext; /* compat */ 6226 __be16 tunnel_flags; 6227 }; 6228 __u32 tunnel_label; 6229 union { 6230 __u32 local_ipv4; 6231 __u32 local_ipv6[4]; 6232 }; 6233 }; 6234 6235 /* user accessible mirror of in-kernel xfrm_state. 6236 * new fields can only be added to the end of this structure 6237 */ 6238 struct bpf_xfrm_state { 6239 __u32 reqid; 6240 __u32 spi; /* Stored in network byte order */ 6241 __u16 family; 6242 __u16 ext; /* Padding, future use. */ 6243 union { 6244 __u32 remote_ipv4; /* Stored in network byte order */ 6245 __u32 remote_ipv6[4]; /* Stored in network byte order */ 6246 }; 6247 }; 6248 6249 /* Generic BPF return codes which all BPF program types may support. 6250 * The values are binary compatible with their TC_ACT_* counter-part to 6251 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 6252 * programs. 6253 * 6254 * XDP is handled seprately, see XDP_*. 6255 */ 6256 enum bpf_ret_code { 6257 BPF_OK = 0, 6258 /* 1 reserved */ 6259 BPF_DROP = 2, 6260 /* 3-6 reserved */ 6261 BPF_REDIRECT = 7, 6262 /* >127 are reserved for prog type specific return codes. 6263 * 6264 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 6265 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 6266 * changed and should be routed based on its new L3 header. 6267 * (This is an L3 redirect, as opposed to L2 redirect 6268 * represented by BPF_REDIRECT above). 6269 */ 6270 BPF_LWT_REROUTE = 128, 6271 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR 6272 * to indicate that no custom dissection was performed, and 6273 * fallback to standard dissector is requested. 6274 */ 6275 BPF_FLOW_DISSECTOR_CONTINUE = 129, 6276 }; 6277 6278 struct bpf_sock { 6279 __u32 bound_dev_if; 6280 __u32 family; 6281 __u32 type; 6282 __u32 protocol; 6283 __u32 mark; 6284 __u32 priority; 6285 /* IP address also allows 1 and 2 bytes access */ 6286 __u32 src_ip4; 6287 __u32 src_ip6[4]; 6288 __u32 src_port; /* host byte order */ 6289 __be16 dst_port; /* network byte order */ 6290 __u16 :16; /* zero padding */ 6291 __u32 dst_ip4; 6292 __u32 dst_ip6[4]; 6293 __u32 state; 6294 __s32 rx_queue_mapping; 6295 }; 6296 6297 struct bpf_tcp_sock { 6298 __u32 snd_cwnd; /* Sending congestion window */ 6299 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 6300 __u32 rtt_min; 6301 __u32 snd_ssthresh; /* Slow start size threshold */ 6302 __u32 rcv_nxt; /* What we want to receive next */ 6303 __u32 snd_nxt; /* Next sequence we send */ 6304 __u32 snd_una; /* First byte we want an ack for */ 6305 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 6306 __u32 ecn_flags; /* ECN status bits. */ 6307 __u32 rate_delivered; /* saved rate sample: packets delivered */ 6308 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 6309 __u32 packets_out; /* Packets which are "in flight" */ 6310 __u32 retrans_out; /* Retransmitted packets out */ 6311 __u32 total_retrans; /* Total retransmits for entire connection */ 6312 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 6313 * total number of segments in. 6314 */ 6315 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 6316 * total number of data segments in. 6317 */ 6318 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 6319 * The total number of segments sent. 6320 */ 6321 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 6322 * total number of data segments sent. 6323 */ 6324 __u32 lost_out; /* Lost packets */ 6325 __u32 sacked_out; /* SACK'd packets */ 6326 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 6327 * sum(delta(rcv_nxt)), or how many bytes 6328 * were acked. 6329 */ 6330 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 6331 * sum(delta(snd_una)), or how many bytes 6332 * were acked. 6333 */ 6334 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 6335 * total number of DSACK blocks received 6336 */ 6337 __u32 delivered; /* Total data packets delivered incl. rexmits */ 6338 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 6339 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 6340 }; 6341 6342 struct bpf_sock_tuple { 6343 union { 6344 struct { 6345 __be32 saddr; 6346 __be32 daddr; 6347 __be16 sport; 6348 __be16 dport; 6349 } ipv4; 6350 struct { 6351 __be32 saddr[4]; 6352 __be32 daddr[4]; 6353 __be16 sport; 6354 __be16 dport; 6355 } ipv6; 6356 }; 6357 }; 6358 6359 /* (Simplified) user return codes for tcx prog type. 6360 * A valid tcx program must return one of these defined values. All other 6361 * return codes are reserved for future use. Must remain compatible with 6362 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown 6363 * return codes are mapped to TCX_NEXT. 6364 */ 6365 enum tcx_action_base { 6366 TCX_NEXT = -1, 6367 TCX_PASS = 0, 6368 TCX_DROP = 2, 6369 TCX_REDIRECT = 7, 6370 }; 6371 6372 struct bpf_xdp_sock { 6373 __u32 queue_id; 6374 }; 6375 6376 #define XDP_PACKET_HEADROOM 256 6377 6378 /* User return codes for XDP prog type. 6379 * A valid XDP program must return one of these defined values. All other 6380 * return codes are reserved for future use. Unknown return codes will 6381 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 6382 */ 6383 enum xdp_action { 6384 XDP_ABORTED = 0, 6385 XDP_DROP, 6386 XDP_PASS, 6387 XDP_TX, 6388 XDP_REDIRECT, 6389 }; 6390 6391 /* user accessible metadata for XDP packet hook 6392 * new fields must be added to the end of this structure 6393 */ 6394 struct xdp_md { 6395 __u32 data; 6396 __u32 data_end; 6397 __u32 data_meta; 6398 /* Below access go through struct xdp_rxq_info */ 6399 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 6400 __u32 rx_queue_index; /* rxq->queue_index */ 6401 6402 __u32 egress_ifindex; /* txq->dev->ifindex */ 6403 }; 6404 6405 /* DEVMAP map-value layout 6406 * 6407 * The struct data-layout of map-value is a configuration interface. 6408 * New members can only be added to the end of this structure. 6409 */ 6410 struct bpf_devmap_val { 6411 __u32 ifindex; /* device index */ 6412 union { 6413 int fd; /* prog fd on map write */ 6414 __u32 id; /* prog id on map read */ 6415 } bpf_prog; 6416 }; 6417 6418 /* CPUMAP map-value layout 6419 * 6420 * The struct data-layout of map-value is a configuration interface. 6421 * New members can only be added to the end of this structure. 6422 */ 6423 struct bpf_cpumap_val { 6424 __u32 qsize; /* queue size to remote target CPU */ 6425 union { 6426 int fd; /* prog fd on map write */ 6427 __u32 id; /* prog id on map read */ 6428 } bpf_prog; 6429 }; 6430 6431 enum sk_action { 6432 SK_DROP = 0, 6433 SK_PASS, 6434 }; 6435 6436 /* user accessible metadata for SK_MSG packet hook, new fields must 6437 * be added to the end of this structure 6438 */ 6439 struct sk_msg_md { 6440 __bpf_md_ptr(void *, data); 6441 __bpf_md_ptr(void *, data_end); 6442 6443 __u32 family; 6444 __u32 remote_ip4; /* Stored in network byte order */ 6445 __u32 local_ip4; /* Stored in network byte order */ 6446 __u32 remote_ip6[4]; /* Stored in network byte order */ 6447 __u32 local_ip6[4]; /* Stored in network byte order */ 6448 __u32 remote_port; /* Stored in network byte order */ 6449 __u32 local_port; /* stored in host byte order */ 6450 __u32 size; /* Total size of sk_msg */ 6451 6452 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 6453 }; 6454 6455 struct sk_reuseport_md { 6456 /* 6457 * Start of directly accessible data. It begins from 6458 * the tcp/udp header. 6459 */ 6460 __bpf_md_ptr(void *, data); 6461 /* End of directly accessible data */ 6462 __bpf_md_ptr(void *, data_end); 6463 /* 6464 * Total length of packet (starting from the tcp/udp header). 6465 * Note that the directly accessible bytes (data_end - data) 6466 * could be less than this "len". Those bytes could be 6467 * indirectly read by a helper "bpf_skb_load_bytes()". 6468 */ 6469 __u32 len; 6470 /* 6471 * Eth protocol in the mac header (network byte order). e.g. 6472 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 6473 */ 6474 __u32 eth_protocol; 6475 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 6476 __u32 bind_inany; /* Is sock bound to an INANY address? */ 6477 __u32 hash; /* A hash of the packet 4 tuples */ 6478 /* When reuse->migrating_sk is NULL, it is selecting a sk for the 6479 * new incoming connection request (e.g. selecting a listen sk for 6480 * the received SYN in the TCP case). reuse->sk is one of the sk 6481 * in the reuseport group. The bpf prog can use reuse->sk to learn 6482 * the local listening ip/port without looking into the skb. 6483 * 6484 * When reuse->migrating_sk is not NULL, reuse->sk is closed and 6485 * reuse->migrating_sk is the socket that needs to be migrated 6486 * to another listening socket. migrating_sk could be a fullsock 6487 * sk that is fully established or a reqsk that is in-the-middle 6488 * of 3-way handshake. 6489 */ 6490 __bpf_md_ptr(struct bpf_sock *, sk); 6491 __bpf_md_ptr(struct bpf_sock *, migrating_sk); 6492 }; 6493 6494 #define BPF_TAG_SIZE 8 6495 6496 struct bpf_prog_info { 6497 __u32 type; 6498 __u32 id; 6499 __u8 tag[BPF_TAG_SIZE]; 6500 __u32 jited_prog_len; 6501 __u32 xlated_prog_len; 6502 __aligned_u64 jited_prog_insns; 6503 __aligned_u64 xlated_prog_insns; 6504 __u64 load_time; /* ns since boottime */ 6505 __u32 created_by_uid; 6506 __u32 nr_map_ids; 6507 __aligned_u64 map_ids; 6508 char name[BPF_OBJ_NAME_LEN]; 6509 __u32 ifindex; 6510 __u32 gpl_compatible:1; 6511 __u32 :31; /* alignment pad */ 6512 __u64 netns_dev; 6513 __u64 netns_ino; 6514 __u32 nr_jited_ksyms; 6515 __u32 nr_jited_func_lens; 6516 __aligned_u64 jited_ksyms; 6517 __aligned_u64 jited_func_lens; 6518 __u32 btf_id; 6519 __u32 func_info_rec_size; 6520 __aligned_u64 func_info; 6521 __u32 nr_func_info; 6522 __u32 nr_line_info; 6523 __aligned_u64 line_info; 6524 __aligned_u64 jited_line_info; 6525 __u32 nr_jited_line_info; 6526 __u32 line_info_rec_size; 6527 __u32 jited_line_info_rec_size; 6528 __u32 nr_prog_tags; 6529 __aligned_u64 prog_tags; 6530 __u64 run_time_ns; 6531 __u64 run_cnt; 6532 __u64 recursion_misses; 6533 __u32 verified_insns; 6534 __u32 attach_btf_obj_id; 6535 __u32 attach_btf_id; 6536 } __attribute__((aligned(8))); 6537 6538 struct bpf_map_info { 6539 __u32 type; 6540 __u32 id; 6541 __u32 key_size; 6542 __u32 value_size; 6543 __u32 max_entries; 6544 __u32 map_flags; 6545 char name[BPF_OBJ_NAME_LEN]; 6546 __u32 ifindex; 6547 __u32 btf_vmlinux_value_type_id; 6548 __u64 netns_dev; 6549 __u64 netns_ino; 6550 __u32 btf_id; 6551 __u32 btf_key_type_id; 6552 __u32 btf_value_type_id; 6553 __u32 btf_vmlinux_id; 6554 __u64 map_extra; 6555 } __attribute__((aligned(8))); 6556 6557 struct bpf_btf_info { 6558 __aligned_u64 btf; 6559 __u32 btf_size; 6560 __u32 id; 6561 __aligned_u64 name; 6562 __u32 name_len; 6563 __u32 kernel_btf; 6564 } __attribute__((aligned(8))); 6565 6566 struct bpf_link_info { 6567 __u32 type; 6568 __u32 id; 6569 __u32 prog_id; 6570 union { 6571 struct { 6572 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 6573 __u32 tp_name_len; /* in/out: tp_name buffer len */ 6574 } raw_tracepoint; 6575 struct { 6576 __u32 attach_type; 6577 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 6578 __u32 target_btf_id; /* BTF type id inside the object */ 6579 } tracing; 6580 struct { 6581 __u64 cgroup_id; 6582 __u32 attach_type; 6583 } cgroup; 6584 struct { 6585 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 6586 __u32 target_name_len; /* in/out: target_name buffer len */ 6587 6588 /* If the iter specific field is 32 bits, it can be put 6589 * in the first or second union. Otherwise it should be 6590 * put in the second union. 6591 */ 6592 union { 6593 struct { 6594 __u32 map_id; 6595 } map; 6596 }; 6597 union { 6598 struct { 6599 __u64 cgroup_id; 6600 __u32 order; 6601 } cgroup; 6602 struct { 6603 __u32 tid; 6604 __u32 pid; 6605 } task; 6606 }; 6607 } iter; 6608 struct { 6609 __u32 netns_ino; 6610 __u32 attach_type; 6611 } netns; 6612 struct { 6613 __u32 ifindex; 6614 } xdp; 6615 struct { 6616 __u32 map_id; 6617 } struct_ops; 6618 struct { 6619 __u32 pf; 6620 __u32 hooknum; 6621 __s32 priority; 6622 __u32 flags; 6623 } netfilter; 6624 struct { 6625 __aligned_u64 addrs; 6626 __u32 count; /* in/out: kprobe_multi function count */ 6627 __u32 flags; 6628 __u64 missed; 6629 __aligned_u64 cookies; 6630 } kprobe_multi; 6631 struct { 6632 __aligned_u64 path; 6633 __aligned_u64 offsets; 6634 __aligned_u64 ref_ctr_offsets; 6635 __aligned_u64 cookies; 6636 __u32 path_size; /* in/out: real path size on success, including zero byte */ 6637 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */ 6638 __u32 flags; 6639 __u32 pid; 6640 } uprobe_multi; 6641 struct { 6642 __u32 type; /* enum bpf_perf_event_type */ 6643 __u32 :32; 6644 union { 6645 struct { 6646 __aligned_u64 file_name; /* in/out */ 6647 __u32 name_len; 6648 __u32 offset; /* offset from file_name */ 6649 __u64 cookie; 6650 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */ 6651 struct { 6652 __aligned_u64 func_name; /* in/out */ 6653 __u32 name_len; 6654 __u32 offset; /* offset from func_name */ 6655 __u64 addr; 6656 __u64 missed; 6657 __u64 cookie; 6658 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */ 6659 struct { 6660 __aligned_u64 tp_name; /* in/out */ 6661 __u32 name_len; 6662 __u32 :32; 6663 __u64 cookie; 6664 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */ 6665 struct { 6666 __u64 config; 6667 __u32 type; 6668 __u32 :32; 6669 __u64 cookie; 6670 } event; /* BPF_PERF_EVENT_EVENT */ 6671 }; 6672 } perf_event; 6673 struct { 6674 __u32 ifindex; 6675 __u32 attach_type; 6676 } tcx; 6677 struct { 6678 __u32 ifindex; 6679 __u32 attach_type; 6680 } netkit; 6681 }; 6682 } __attribute__((aligned(8))); 6683 6684 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 6685 * by user and intended to be used by socket (e.g. to bind to, depends on 6686 * attach type). 6687 */ 6688 struct bpf_sock_addr { 6689 __u32 user_family; /* Allows 4-byte read, but no write. */ 6690 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6691 * Stored in network byte order. 6692 */ 6693 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6694 * Stored in network byte order. 6695 */ 6696 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 6697 * Stored in network byte order 6698 */ 6699 __u32 family; /* Allows 4-byte read, but no write */ 6700 __u32 type; /* Allows 4-byte read, but no write */ 6701 __u32 protocol; /* Allows 4-byte read, but no write */ 6702 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 6703 * Stored in network byte order. 6704 */ 6705 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 6706 * Stored in network byte order. 6707 */ 6708 __bpf_md_ptr(struct bpf_sock *, sk); 6709 }; 6710 6711 /* User bpf_sock_ops struct to access socket values and specify request ops 6712 * and their replies. 6713 * Some of this fields are in network (bigendian) byte order and may need 6714 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 6715 * New fields can only be added at the end of this structure 6716 */ 6717 struct bpf_sock_ops { 6718 __u32 op; 6719 union { 6720 __u32 args[4]; /* Optionally passed to bpf program */ 6721 __u32 reply; /* Returned by bpf program */ 6722 __u32 replylong[4]; /* Optionally returned by bpf prog */ 6723 }; 6724 __u32 family; 6725 __u32 remote_ip4; /* Stored in network byte order */ 6726 __u32 local_ip4; /* Stored in network byte order */ 6727 __u32 remote_ip6[4]; /* Stored in network byte order */ 6728 __u32 local_ip6[4]; /* Stored in network byte order */ 6729 __u32 remote_port; /* Stored in network byte order */ 6730 __u32 local_port; /* stored in host byte order */ 6731 __u32 is_fullsock; /* Some TCP fields are only valid if 6732 * there is a full socket. If not, the 6733 * fields read as zero. 6734 */ 6735 __u32 snd_cwnd; 6736 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 6737 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 6738 __u32 state; 6739 __u32 rtt_min; 6740 __u32 snd_ssthresh; 6741 __u32 rcv_nxt; 6742 __u32 snd_nxt; 6743 __u32 snd_una; 6744 __u32 mss_cache; 6745 __u32 ecn_flags; 6746 __u32 rate_delivered; 6747 __u32 rate_interval_us; 6748 __u32 packets_out; 6749 __u32 retrans_out; 6750 __u32 total_retrans; 6751 __u32 segs_in; 6752 __u32 data_segs_in; 6753 __u32 segs_out; 6754 __u32 data_segs_out; 6755 __u32 lost_out; 6756 __u32 sacked_out; 6757 __u32 sk_txhash; 6758 __u64 bytes_received; 6759 __u64 bytes_acked; 6760 __bpf_md_ptr(struct bpf_sock *, sk); 6761 /* [skb_data, skb_data_end) covers the whole TCP header. 6762 * 6763 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 6764 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 6765 * header has not been written. 6766 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 6767 * been written so far. 6768 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 6769 * the 3WHS. 6770 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 6771 * the 3WHS. 6772 * 6773 * bpf_load_hdr_opt() can also be used to read a particular option. 6774 */ 6775 __bpf_md_ptr(void *, skb_data); 6776 __bpf_md_ptr(void *, skb_data_end); 6777 __u32 skb_len; /* The total length of a packet. 6778 * It includes the header, options, 6779 * and payload. 6780 */ 6781 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 6782 * an easy way to check for tcp_flags 6783 * without parsing skb_data. 6784 * 6785 * In particular, the skb_tcp_flags 6786 * will still be available in 6787 * BPF_SOCK_OPS_HDR_OPT_LEN even though 6788 * the outgoing header has not 6789 * been written yet. 6790 */ 6791 __u64 skb_hwtstamp; 6792 }; 6793 6794 /* Definitions for bpf_sock_ops_cb_flags */ 6795 enum { 6796 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 6797 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 6798 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 6799 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 6800 /* Call bpf for all received TCP headers. The bpf prog will be 6801 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6802 * 6803 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6804 * for the header option related helpers that will be useful 6805 * to the bpf programs. 6806 * 6807 * It could be used at the client/active side (i.e. connect() side) 6808 * when the server told it that the server was in syncookie 6809 * mode and required the active side to resend the bpf-written 6810 * options. The active side can keep writing the bpf-options until 6811 * it received a valid packet from the server side to confirm 6812 * the earlier packet (and options) has been received. The later 6813 * example patch is using it like this at the active side when the 6814 * server is in syncookie mode. 6815 * 6816 * The bpf prog will usually turn this off in the common cases. 6817 */ 6818 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 6819 /* Call bpf when kernel has received a header option that 6820 * the kernel cannot handle. The bpf prog will be called under 6821 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 6822 * 6823 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 6824 * for the header option related helpers that will be useful 6825 * to the bpf programs. 6826 */ 6827 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 6828 /* Call bpf when the kernel is writing header options for the 6829 * outgoing packet. The bpf prog will first be called 6830 * to reserve space in a skb under 6831 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 6832 * the bpf prog will be called to write the header option(s) 6833 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6834 * 6835 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 6836 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 6837 * related helpers that will be useful to the bpf programs. 6838 * 6839 * The kernel gets its chance to reserve space and write 6840 * options first before the BPF program does. 6841 */ 6842 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 6843 /* Mask of all currently supported cb flags */ 6844 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 6845 }; 6846 6847 /* List of known BPF sock_ops operators. 6848 * New entries can only be added at the end 6849 */ 6850 enum { 6851 BPF_SOCK_OPS_VOID, 6852 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 6853 * -1 if default value should be used 6854 */ 6855 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 6856 * window (in packets) or -1 if default 6857 * value should be used 6858 */ 6859 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 6860 * active connection is initialized 6861 */ 6862 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 6863 * active connection is 6864 * established 6865 */ 6866 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 6867 * passive connection is 6868 * established 6869 */ 6870 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 6871 * needs ECN 6872 */ 6873 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 6874 * based on the path and may be 6875 * dependent on the congestion control 6876 * algorithm. In general it indicates 6877 * a congestion threshold. RTTs above 6878 * this indicate congestion 6879 */ 6880 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 6881 * Arg1: value of icsk_retransmits 6882 * Arg2: value of icsk_rto 6883 * Arg3: whether RTO has expired 6884 */ 6885 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 6886 * Arg1: sequence number of 1st byte 6887 * Arg2: # segments 6888 * Arg3: return value of 6889 * tcp_transmit_skb (0 => success) 6890 */ 6891 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 6892 * Arg1: old_state 6893 * Arg2: new_state 6894 */ 6895 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 6896 * socket transition to LISTEN state. 6897 */ 6898 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 6899 */ 6900 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 6901 * It will be called to handle 6902 * the packets received at 6903 * an already established 6904 * connection. 6905 * 6906 * sock_ops->skb_data: 6907 * Referring to the received skb. 6908 * It covers the TCP header only. 6909 * 6910 * bpf_load_hdr_opt() can also 6911 * be used to search for a 6912 * particular option. 6913 */ 6914 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 6915 * header option later in 6916 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 6917 * Arg1: bool want_cookie. (in 6918 * writing SYNACK only) 6919 * 6920 * sock_ops->skb_data: 6921 * Not available because no header has 6922 * been written yet. 6923 * 6924 * sock_ops->skb_tcp_flags: 6925 * The tcp_flags of the 6926 * outgoing skb. (e.g. SYN, ACK, FIN). 6927 * 6928 * bpf_reserve_hdr_opt() should 6929 * be used to reserve space. 6930 */ 6931 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 6932 * Arg1: bool want_cookie. (in 6933 * writing SYNACK only) 6934 * 6935 * sock_ops->skb_data: 6936 * Referring to the outgoing skb. 6937 * It covers the TCP header 6938 * that has already been written 6939 * by the kernel and the 6940 * earlier bpf-progs. 6941 * 6942 * sock_ops->skb_tcp_flags: 6943 * The tcp_flags of the outgoing 6944 * skb. (e.g. SYN, ACK, FIN). 6945 * 6946 * bpf_store_hdr_opt() should 6947 * be used to write the 6948 * option. 6949 * 6950 * bpf_load_hdr_opt() can also 6951 * be used to search for a 6952 * particular option that 6953 * has already been written 6954 * by the kernel or the 6955 * earlier bpf-progs. 6956 */ 6957 }; 6958 6959 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 6960 * changes between the TCP and BPF versions. Ideally this should never happen. 6961 * If it does, we need to add code to convert them before calling 6962 * the BPF sock_ops function. 6963 */ 6964 enum { 6965 BPF_TCP_ESTABLISHED = 1, 6966 BPF_TCP_SYN_SENT, 6967 BPF_TCP_SYN_RECV, 6968 BPF_TCP_FIN_WAIT1, 6969 BPF_TCP_FIN_WAIT2, 6970 BPF_TCP_TIME_WAIT, 6971 BPF_TCP_CLOSE, 6972 BPF_TCP_CLOSE_WAIT, 6973 BPF_TCP_LAST_ACK, 6974 BPF_TCP_LISTEN, 6975 BPF_TCP_CLOSING, /* Now a valid state */ 6976 BPF_TCP_NEW_SYN_RECV, 6977 BPF_TCP_BOUND_INACTIVE, 6978 6979 BPF_TCP_MAX_STATES /* Leave at the end! */ 6980 }; 6981 6982 enum { 6983 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 6984 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 6985 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 6986 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 6987 /* Copy the SYN pkt to optval 6988 * 6989 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 6990 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 6991 * to only getting from the saved_syn. It can either get the 6992 * syn packet from: 6993 * 6994 * 1. the just-received SYN packet (only available when writing the 6995 * SYNACK). It will be useful when it is not necessary to 6996 * save the SYN packet for latter use. It is also the only way 6997 * to get the SYN during syncookie mode because the syn 6998 * packet cannot be saved during syncookie. 6999 * 7000 * OR 7001 * 7002 * 2. the earlier saved syn which was done by 7003 * bpf_setsockopt(TCP_SAVE_SYN). 7004 * 7005 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 7006 * SYN packet is obtained. 7007 * 7008 * If the bpf-prog does not need the IP[46] header, the 7009 * bpf-prog can avoid parsing the IP header by using 7010 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 7011 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 7012 * 7013 * >0: Total number of bytes copied 7014 * -ENOSPC: Not enough space in optval. Only optlen number of 7015 * bytes is copied. 7016 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 7017 * is not saved by setsockopt(TCP_SAVE_SYN). 7018 */ 7019 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 7020 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 7021 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 7022 }; 7023 7024 enum { 7025 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 7026 }; 7027 7028 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 7029 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 7030 */ 7031 enum { 7032 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 7033 * total option spaces 7034 * required for an established 7035 * sk in order to calculate the 7036 * MSS. No skb is actually 7037 * sent. 7038 */ 7039 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 7040 * when sending a SYN. 7041 */ 7042 }; 7043 7044 struct bpf_perf_event_value { 7045 __u64 counter; 7046 __u64 enabled; 7047 __u64 running; 7048 }; 7049 7050 enum { 7051 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 7052 BPF_DEVCG_ACC_READ = (1ULL << 1), 7053 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 7054 }; 7055 7056 enum { 7057 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 7058 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 7059 }; 7060 7061 struct bpf_cgroup_dev_ctx { 7062 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 7063 __u32 access_type; 7064 __u32 major; 7065 __u32 minor; 7066 }; 7067 7068 struct bpf_raw_tracepoint_args { 7069 __u64 args[0]; 7070 }; 7071 7072 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 7073 * OUTPUT: Do lookup from egress perspective; default is ingress 7074 */ 7075 enum { 7076 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 7077 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 7078 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2), 7079 BPF_FIB_LOOKUP_TBID = (1U << 3), 7080 BPF_FIB_LOOKUP_SRC = (1U << 4), 7081 }; 7082 7083 enum { 7084 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 7085 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 7086 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 7087 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 7088 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 7089 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 7090 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 7091 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 7092 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7093 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */ 7094 }; 7095 7096 struct bpf_fib_lookup { 7097 /* input: network family for lookup (AF_INET, AF_INET6) 7098 * output: network family of egress nexthop 7099 */ 7100 __u8 family; 7101 7102 /* set if lookup is to consider L4 data - e.g., FIB rules */ 7103 __u8 l4_protocol; 7104 __be16 sport; 7105 __be16 dport; 7106 7107 union { /* used for MTU check */ 7108 /* input to lookup */ 7109 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 7110 7111 /* output: MTU value */ 7112 __u16 mtu_result; 7113 }; 7114 /* input: L3 device index for lookup 7115 * output: device index from FIB lookup 7116 */ 7117 __u32 ifindex; 7118 7119 union { 7120 /* inputs to lookup */ 7121 __u8 tos; /* AF_INET */ 7122 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 7123 7124 /* output: metric of fib result (IPv4/IPv6 only) */ 7125 __u32 rt_metric; 7126 }; 7127 7128 /* input: source address to consider for lookup 7129 * output: source address result from lookup 7130 */ 7131 union { 7132 __be32 ipv4_src; 7133 __u32 ipv6_src[4]; /* in6_addr; network order */ 7134 }; 7135 7136 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 7137 * network header. output: bpf_fib_lookup sets to gateway address 7138 * if FIB lookup returns gateway route 7139 */ 7140 union { 7141 __be32 ipv4_dst; 7142 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7143 }; 7144 7145 union { 7146 struct { 7147 /* output */ 7148 __be16 h_vlan_proto; 7149 __be16 h_vlan_TCI; 7150 }; 7151 /* input: when accompanied with the 7152 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a 7153 * specific routing table to use for the fib lookup. 7154 */ 7155 __u32 tbid; 7156 }; 7157 7158 __u8 smac[6]; /* ETH_ALEN */ 7159 __u8 dmac[6]; /* ETH_ALEN */ 7160 }; 7161 7162 struct bpf_redir_neigh { 7163 /* network family for lookup (AF_INET, AF_INET6) */ 7164 __u32 nh_family; 7165 /* network address of nexthop; skips fib lookup to find gateway */ 7166 union { 7167 __be32 ipv4_nh; 7168 __u32 ipv6_nh[4]; /* in6_addr; network order */ 7169 }; 7170 }; 7171 7172 /* bpf_check_mtu flags*/ 7173 enum bpf_check_mtu_flags { 7174 BPF_MTU_CHK_SEGS = (1U << 0), 7175 }; 7176 7177 enum bpf_check_mtu_ret { 7178 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 7179 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 7180 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 7181 }; 7182 7183 enum bpf_task_fd_type { 7184 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 7185 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 7186 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 7187 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 7188 BPF_FD_TYPE_UPROBE, /* filename + offset */ 7189 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 7190 }; 7191 7192 enum { 7193 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 7194 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 7195 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 7196 }; 7197 7198 struct bpf_flow_keys { 7199 __u16 nhoff; 7200 __u16 thoff; 7201 __u16 addr_proto; /* ETH_P_* of valid addrs */ 7202 __u8 is_frag; 7203 __u8 is_first_frag; 7204 __u8 is_encap; 7205 __u8 ip_proto; 7206 __be16 n_proto; 7207 __be16 sport; 7208 __be16 dport; 7209 union { 7210 struct { 7211 __be32 ipv4_src; 7212 __be32 ipv4_dst; 7213 }; 7214 struct { 7215 __u32 ipv6_src[4]; /* in6_addr; network order */ 7216 __u32 ipv6_dst[4]; /* in6_addr; network order */ 7217 }; 7218 }; 7219 __u32 flags; 7220 __be32 flow_label; 7221 }; 7222 7223 struct bpf_func_info { 7224 __u32 insn_off; 7225 __u32 type_id; 7226 }; 7227 7228 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 7229 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 7230 7231 struct bpf_line_info { 7232 __u32 insn_off; 7233 __u32 file_name_off; 7234 __u32 line_off; 7235 __u32 line_col; 7236 }; 7237 7238 struct bpf_spin_lock { 7239 __u32 val; 7240 }; 7241 7242 struct bpf_timer { 7243 __u64 __opaque[2]; 7244 } __attribute__((aligned(8))); 7245 7246 struct bpf_dynptr { 7247 __u64 __opaque[2]; 7248 } __attribute__((aligned(8))); 7249 7250 struct bpf_list_head { 7251 __u64 __opaque[2]; 7252 } __attribute__((aligned(8))); 7253 7254 struct bpf_list_node { 7255 __u64 __opaque[3]; 7256 } __attribute__((aligned(8))); 7257 7258 struct bpf_rb_root { 7259 __u64 __opaque[2]; 7260 } __attribute__((aligned(8))); 7261 7262 struct bpf_rb_node { 7263 __u64 __opaque[4]; 7264 } __attribute__((aligned(8))); 7265 7266 struct bpf_refcount { 7267 __u32 __opaque[1]; 7268 } __attribute__((aligned(4))); 7269 7270 struct bpf_sysctl { 7271 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 7272 * Allows 1,2,4-byte read, but no write. 7273 */ 7274 __u32 file_pos; /* Sysctl file position to read from, write to. 7275 * Allows 1,2,4-byte read an 4-byte write. 7276 */ 7277 }; 7278 7279 struct bpf_sockopt { 7280 __bpf_md_ptr(struct bpf_sock *, sk); 7281 __bpf_md_ptr(void *, optval); 7282 __bpf_md_ptr(void *, optval_end); 7283 7284 __s32 level; 7285 __s32 optname; 7286 __s32 optlen; 7287 __s32 retval; 7288 }; 7289 7290 struct bpf_pidns_info { 7291 __u32 pid; 7292 __u32 tgid; 7293 }; 7294 7295 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 7296 struct bpf_sk_lookup { 7297 union { 7298 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 7299 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 7300 }; 7301 7302 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 7303 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 7304 __u32 remote_ip4; /* Network byte order */ 7305 __u32 remote_ip6[4]; /* Network byte order */ 7306 __be16 remote_port; /* Network byte order */ 7307 __u16 :16; /* Zero padding */ 7308 __u32 local_ip4; /* Network byte order */ 7309 __u32 local_ip6[4]; /* Network byte order */ 7310 __u32 local_port; /* Host byte order */ 7311 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */ 7312 }; 7313 7314 /* 7315 * struct btf_ptr is used for typed pointer representation; the 7316 * type id is used to render the pointer data as the appropriate type 7317 * via the bpf_snprintf_btf() helper described above. A flags field - 7318 * potentially to specify additional details about the BTF pointer 7319 * (rather than its mode of display) - is included for future use. 7320 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 7321 */ 7322 struct btf_ptr { 7323 void *ptr; 7324 __u32 type_id; 7325 __u32 flags; /* BTF ptr flags; unused at present. */ 7326 }; 7327 7328 /* 7329 * Flags to control bpf_snprintf_btf() behaviour. 7330 * - BTF_F_COMPACT: no formatting around type information 7331 * - BTF_F_NONAME: no struct/union member names/types 7332 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 7333 * equivalent to %px. 7334 * - BTF_F_ZERO: show zero-valued struct/union members; they 7335 * are not displayed by default 7336 */ 7337 enum { 7338 BTF_F_COMPACT = (1ULL << 0), 7339 BTF_F_NONAME = (1ULL << 1), 7340 BTF_F_PTR_RAW = (1ULL << 2), 7341 BTF_F_ZERO = (1ULL << 3), 7342 }; 7343 7344 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value 7345 * has to be adjusted by relocations. It is emitted by llvm and passed to 7346 * libbpf and later to the kernel. 7347 */ 7348 enum bpf_core_relo_kind { 7349 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */ 7350 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */ 7351 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */ 7352 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */ 7353 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */ 7354 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */ 7355 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */ 7356 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */ 7357 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */ 7358 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */ 7359 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */ 7360 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */ 7361 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */ 7362 }; 7363 7364 /* 7365 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf 7366 * and from libbpf to the kernel. 7367 * 7368 * CO-RE relocation captures the following data: 7369 * - insn_off - instruction offset (in bytes) within a BPF program that needs 7370 * its insn->imm field to be relocated with actual field info; 7371 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable 7372 * type or field; 7373 * - access_str_off - offset into corresponding .BTF string section. String 7374 * interpretation depends on specific relocation kind: 7375 * - for field-based relocations, string encodes an accessed field using 7376 * a sequence of field and array indices, separated by colon (:). It's 7377 * conceptually very close to LLVM's getelementptr ([0]) instruction's 7378 * arguments for identifying offset to a field. 7379 * - for type-based relocations, strings is expected to be just "0"; 7380 * - for enum value-based relocations, string contains an index of enum 7381 * value within its enum type; 7382 * - kind - one of enum bpf_core_relo_kind; 7383 * 7384 * Example: 7385 * struct sample { 7386 * int a; 7387 * struct { 7388 * int b[10]; 7389 * }; 7390 * }; 7391 * 7392 * struct sample *s = ...; 7393 * int *x = &s->a; // encoded as "0:0" (a is field #0) 7394 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1, 7395 * // b is field #0 inside anon struct, accessing elem #5) 7396 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array) 7397 * 7398 * type_id for all relocs in this example will capture BTF type id of 7399 * `struct sample`. 7400 * 7401 * Such relocation is emitted when using __builtin_preserve_access_index() 7402 * Clang built-in, passing expression that captures field address, e.g.: 7403 * 7404 * bpf_probe_read(&dst, sizeof(dst), 7405 * __builtin_preserve_access_index(&src->a.b.c)); 7406 * 7407 * In this case Clang will emit field relocation recording necessary data to 7408 * be able to find offset of embedded `a.b.c` field within `src` struct. 7409 * 7410 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction 7411 */ 7412 struct bpf_core_relo { 7413 __u32 insn_off; 7414 __u32 type_id; 7415 __u32 access_str_off; 7416 enum bpf_core_relo_kind kind; 7417 }; 7418 7419 /* 7420 * Flags to control bpf_timer_start() behaviour. 7421 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is 7422 * relative to current time. 7423 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller. 7424 */ 7425 enum { 7426 BPF_F_TIMER_ABS = (1ULL << 0), 7427 BPF_F_TIMER_CPU_PIN = (1ULL << 1), 7428 }; 7429 7430 /* BPF numbers iterator state */ 7431 struct bpf_iter_num { 7432 /* opaque iterator state; having __u64 here allows to preserve correct 7433 * alignment requirements in vmlinux.h, generated from BTF 7434 */ 7435 __u64 __opaque[1]; 7436 } __attribute__((aligned(8))); 7437 7438 #endif /* _UAPI__LINUX_BPF_H__ */ 7439