1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_JMP32 0x06 /* jmp mode in word width */ 18 #define BPF_ALU64 0x07 /* alu mode in double word width */ 19 20 /* ld/ldx fields */ 21 #define BPF_DW 0x18 /* double word (64-bit) */ 22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */ 23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */ 24 25 /* alu/jmp fields */ 26 #define BPF_MOV 0xb0 /* mov reg to reg */ 27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 28 29 /* change endianness of a register */ 30 #define BPF_END 0xd0 /* flags for endianness conversion: */ 31 #define BPF_TO_LE 0x00 /* convert to little-endian */ 32 #define BPF_TO_BE 0x08 /* convert to big-endian */ 33 #define BPF_FROM_LE BPF_TO_LE 34 #define BPF_FROM_BE BPF_TO_BE 35 36 /* jmp encodings */ 37 #define BPF_JNE 0x50 /* jump != */ 38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 44 #define BPF_CALL 0x80 /* function call */ 45 #define BPF_EXIT 0x90 /* function return */ 46 47 /* atomic op type fields (stored in immediate) */ 48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */ 49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */ 50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */ 51 52 /* Register numbers */ 53 enum { 54 BPF_REG_0 = 0, 55 BPF_REG_1, 56 BPF_REG_2, 57 BPF_REG_3, 58 BPF_REG_4, 59 BPF_REG_5, 60 BPF_REG_6, 61 BPF_REG_7, 62 BPF_REG_8, 63 BPF_REG_9, 64 BPF_REG_10, 65 __MAX_BPF_REG, 66 }; 67 68 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 69 #define MAX_BPF_REG __MAX_BPF_REG 70 71 struct bpf_insn { 72 __u8 code; /* opcode */ 73 __u8 dst_reg:4; /* dest register */ 74 __u8 src_reg:4; /* source register */ 75 __s16 off; /* signed offset */ 76 __s32 imm; /* signed immediate constant */ 77 }; 78 79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 80 struct bpf_lpm_trie_key { 81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 82 __u8 data[0]; /* Arbitrary size */ 83 }; 84 85 struct bpf_cgroup_storage_key { 86 __u64 cgroup_inode_id; /* cgroup inode id */ 87 __u32 attach_type; /* program attach type */ 88 }; 89 90 union bpf_iter_link_info { 91 struct { 92 __u32 map_fd; 93 } map; 94 }; 95 96 /* BPF syscall commands, see bpf(2) man-page for more details. */ 97 /** 98 * DOC: eBPF Syscall Preamble 99 * 100 * The operation to be performed by the **bpf**\ () system call is determined 101 * by the *cmd* argument. Each operation takes an accompanying argument, 102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see 103 * below). The size argument is the size of the union pointed to by *attr*. 104 */ 105 /** 106 * DOC: eBPF Syscall Commands 107 * 108 * BPF_MAP_CREATE 109 * Description 110 * Create a map and return a file descriptor that refers to the 111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2)) 112 * is automatically enabled for the new file descriptor. 113 * 114 * Applying **close**\ (2) to the file descriptor returned by 115 * **BPF_MAP_CREATE** will delete the map (but see NOTES). 116 * 117 * Return 118 * A new file descriptor (a nonnegative integer), or -1 if an 119 * error occurred (in which case, *errno* is set appropriately). 120 * 121 * BPF_MAP_LOOKUP_ELEM 122 * Description 123 * Look up an element with a given *key* in the map referred to 124 * by the file descriptor *map_fd*. 125 * 126 * The *flags* argument may be specified as one of the 127 * following: 128 * 129 * **BPF_F_LOCK** 130 * Look up the value of a spin-locked map without 131 * returning the lock. This must be specified if the 132 * elements contain a spinlock. 133 * 134 * Return 135 * Returns zero on success. On error, -1 is returned and *errno* 136 * is set appropriately. 137 * 138 * BPF_MAP_UPDATE_ELEM 139 * Description 140 * Create or update an element (key/value pair) in a specified map. 141 * 142 * The *flags* argument should be specified as one of the 143 * following: 144 * 145 * **BPF_ANY** 146 * Create a new element or update an existing element. 147 * **BPF_NOEXIST** 148 * Create a new element only if it did not exist. 149 * **BPF_EXIST** 150 * Update an existing element. 151 * **BPF_F_LOCK** 152 * Update a spin_lock-ed map element. 153 * 154 * Return 155 * Returns zero on success. On error, -1 is returned and *errno* 156 * is set appropriately. 157 * 158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, 159 * **E2BIG**, **EEXIST**, or **ENOENT**. 160 * 161 * **E2BIG** 162 * The number of elements in the map reached the 163 * *max_entries* limit specified at map creation time. 164 * **EEXIST** 165 * If *flags* specifies **BPF_NOEXIST** and the element 166 * with *key* already exists in the map. 167 * **ENOENT** 168 * If *flags* specifies **BPF_EXIST** and the element with 169 * *key* does not exist in the map. 170 * 171 * BPF_MAP_DELETE_ELEM 172 * Description 173 * Look up and delete an element by key in a specified map. 174 * 175 * Return 176 * Returns zero on success. On error, -1 is returned and *errno* 177 * is set appropriately. 178 * 179 * BPF_MAP_GET_NEXT_KEY 180 * Description 181 * Look up an element by key in a specified map and return the key 182 * of the next element. Can be used to iterate over all elements 183 * in the map. 184 * 185 * Return 186 * Returns zero on success. On error, -1 is returned and *errno* 187 * is set appropriately. 188 * 189 * The following cases can be used to iterate over all elements of 190 * the map: 191 * 192 * * If *key* is not found, the operation returns zero and sets 193 * the *next_key* pointer to the key of the first element. 194 * * If *key* is found, the operation returns zero and sets the 195 * *next_key* pointer to the key of the next element. 196 * * If *key* is the last element, returns -1 and *errno* is set 197 * to **ENOENT**. 198 * 199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or 200 * **EINVAL** on error. 201 * 202 * BPF_PROG_LOAD 203 * Description 204 * Verify and load an eBPF program, returning a new file 205 * descriptor associated with the program. 206 * 207 * Applying **close**\ (2) to the file descriptor returned by 208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES). 209 * 210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is 211 * automatically enabled for the new file descriptor. 212 * 213 * Return 214 * A new file descriptor (a nonnegative integer), or -1 if an 215 * error occurred (in which case, *errno* is set appropriately). 216 * 217 * BPF_OBJ_PIN 218 * Description 219 * Pin an eBPF program or map referred by the specified *bpf_fd* 220 * to the provided *pathname* on the filesystem. 221 * 222 * The *pathname* argument must not contain a dot ("."). 223 * 224 * On success, *pathname* retains a reference to the eBPF object, 225 * preventing deallocation of the object when the original 226 * *bpf_fd* is closed. This allow the eBPF object to live beyond 227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent 228 * process. 229 * 230 * Applying **unlink**\ (2) or similar calls to the *pathname* 231 * unpins the object from the filesystem, removing the reference. 232 * If no other file descriptors or filesystem nodes refer to the 233 * same object, it will be deallocated (see NOTES). 234 * 235 * The filesystem type for the parent directory of *pathname* must 236 * be **BPF_FS_MAGIC**. 237 * 238 * Return 239 * Returns zero on success. On error, -1 is returned and *errno* 240 * is set appropriately. 241 * 242 * BPF_OBJ_GET 243 * Description 244 * Open a file descriptor for the eBPF object pinned to the 245 * specified *pathname*. 246 * 247 * Return 248 * A new file descriptor (a nonnegative integer), or -1 if an 249 * error occurred (in which case, *errno* is set appropriately). 250 * 251 * BPF_PROG_ATTACH 252 * Description 253 * Attach an eBPF program to a *target_fd* at the specified 254 * *attach_type* hook. 255 * 256 * The *attach_type* specifies the eBPF attachment point to 257 * attach the program to, and must be one of *bpf_attach_type* 258 * (see below). 259 * 260 * The *attach_bpf_fd* must be a valid file descriptor for a 261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap 262 * or sock_ops type corresponding to the specified *attach_type*. 263 * 264 * The *target_fd* must be a valid file descriptor for a kernel 265 * object which depends on the attach type of *attach_bpf_fd*: 266 * 267 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 268 * **BPF_PROG_TYPE_CGROUP_SKB**, 269 * **BPF_PROG_TYPE_CGROUP_SOCK**, 270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 273 * **BPF_PROG_TYPE_SOCK_OPS** 274 * 275 * Control Group v2 hierarchy with the eBPF controller 276 * enabled. Requires the kernel to be compiled with 277 * **CONFIG_CGROUP_BPF**. 278 * 279 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 280 * 281 * Network namespace (eg /proc/self/ns/net). 282 * 283 * **BPF_PROG_TYPE_LIRC_MODE2** 284 * 285 * LIRC device path (eg /dev/lircN). Requires the kernel 286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 287 * 288 * **BPF_PROG_TYPE_SK_SKB**, 289 * **BPF_PROG_TYPE_SK_MSG** 290 * 291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**). 292 * 293 * Return 294 * Returns zero on success. On error, -1 is returned and *errno* 295 * is set appropriately. 296 * 297 * BPF_PROG_DETACH 298 * Description 299 * Detach the eBPF program associated with the *target_fd* at the 300 * hook specified by *attach_type*. The program must have been 301 * previously attached using **BPF_PROG_ATTACH**. 302 * 303 * Return 304 * Returns zero on success. On error, -1 is returned and *errno* 305 * is set appropriately. 306 * 307 * BPF_PROG_TEST_RUN 308 * Description 309 * Run the eBPF program associated with the *prog_fd* a *repeat* 310 * number of times against a provided program context *ctx_in* and 311 * data *data_in*, and return the modified program context 312 * *ctx_out*, *data_out* (for example, packet data), result of the 313 * execution *retval*, and *duration* of the test run. 314 * 315 * The sizes of the buffers provided as input and output 316 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must 317 * be provided in the corresponding variables *ctx_size_in*, 318 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any 319 * of these parameters are not provided (ie set to NULL), the 320 * corresponding size field must be zero. 321 * 322 * Some program types have particular requirements: 323 * 324 * **BPF_PROG_TYPE_SK_LOOKUP** 325 * *data_in* and *data_out* must be NULL. 326 * 327 * **BPF_PROG_TYPE_XDP** 328 * *ctx_in* and *ctx_out* must be NULL. 329 * 330 * **BPF_PROG_TYPE_RAW_TRACEPOINT**, 331 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE** 332 * 333 * *ctx_out*, *data_in* and *data_out* must be NULL. 334 * *repeat* must be zero. 335 * 336 * Return 337 * Returns zero on success. On error, -1 is returned and *errno* 338 * is set appropriately. 339 * 340 * **ENOSPC** 341 * Either *data_size_out* or *ctx_size_out* is too small. 342 * **ENOTSUPP** 343 * This command is not supported by the program type of 344 * the program referred to by *prog_fd*. 345 * 346 * BPF_PROG_GET_NEXT_ID 347 * Description 348 * Fetch the next eBPF program currently loaded into the kernel. 349 * 350 * Looks for the eBPF program with an id greater than *start_id* 351 * and updates *next_id* on success. If no other eBPF programs 352 * remain with ids higher than *start_id*, returns -1 and sets 353 * *errno* to **ENOENT**. 354 * 355 * Return 356 * Returns zero on success. On error, or when no id remains, -1 357 * is returned and *errno* is set appropriately. 358 * 359 * BPF_MAP_GET_NEXT_ID 360 * Description 361 * Fetch the next eBPF map currently loaded into the kernel. 362 * 363 * Looks for the eBPF map with an id greater than *start_id* 364 * and updates *next_id* on success. If no other eBPF maps 365 * remain with ids higher than *start_id*, returns -1 and sets 366 * *errno* to **ENOENT**. 367 * 368 * Return 369 * Returns zero on success. On error, or when no id remains, -1 370 * is returned and *errno* is set appropriately. 371 * 372 * BPF_PROG_GET_FD_BY_ID 373 * Description 374 * Open a file descriptor for the eBPF program corresponding to 375 * *prog_id*. 376 * 377 * Return 378 * A new file descriptor (a nonnegative integer), or -1 if an 379 * error occurred (in which case, *errno* is set appropriately). 380 * 381 * BPF_MAP_GET_FD_BY_ID 382 * Description 383 * Open a file descriptor for the eBPF map corresponding to 384 * *map_id*. 385 * 386 * Return 387 * A new file descriptor (a nonnegative integer), or -1 if an 388 * error occurred (in which case, *errno* is set appropriately). 389 * 390 * BPF_OBJ_GET_INFO_BY_FD 391 * Description 392 * Obtain information about the eBPF object corresponding to 393 * *bpf_fd*. 394 * 395 * Populates up to *info_len* bytes of *info*, which will be in 396 * one of the following formats depending on the eBPF object type 397 * of *bpf_fd*: 398 * 399 * * **struct bpf_prog_info** 400 * * **struct bpf_map_info** 401 * * **struct bpf_btf_info** 402 * * **struct bpf_link_info** 403 * 404 * Return 405 * Returns zero on success. On error, -1 is returned and *errno* 406 * is set appropriately. 407 * 408 * BPF_PROG_QUERY 409 * Description 410 * Obtain information about eBPF programs associated with the 411 * specified *attach_type* hook. 412 * 413 * The *target_fd* must be a valid file descriptor for a kernel 414 * object which depends on the attach type of *attach_bpf_fd*: 415 * 416 * **BPF_PROG_TYPE_CGROUP_DEVICE**, 417 * **BPF_PROG_TYPE_CGROUP_SKB**, 418 * **BPF_PROG_TYPE_CGROUP_SOCK**, 419 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**, 420 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**, 421 * **BPF_PROG_TYPE_CGROUP_SYSCTL**, 422 * **BPF_PROG_TYPE_SOCK_OPS** 423 * 424 * Control Group v2 hierarchy with the eBPF controller 425 * enabled. Requires the kernel to be compiled with 426 * **CONFIG_CGROUP_BPF**. 427 * 428 * **BPF_PROG_TYPE_FLOW_DISSECTOR** 429 * 430 * Network namespace (eg /proc/self/ns/net). 431 * 432 * **BPF_PROG_TYPE_LIRC_MODE2** 433 * 434 * LIRC device path (eg /dev/lircN). Requires the kernel 435 * to be compiled with **CONFIG_BPF_LIRC_MODE2**. 436 * 437 * **BPF_PROG_QUERY** always fetches the number of programs 438 * attached and the *attach_flags* which were used to attach those 439 * programs. Additionally, if *prog_ids* is nonzero and the number 440 * of attached programs is less than *prog_cnt*, populates 441 * *prog_ids* with the eBPF program ids of the programs attached 442 * at *target_fd*. 443 * 444 * The following flags may alter the result: 445 * 446 * **BPF_F_QUERY_EFFECTIVE** 447 * Only return information regarding programs which are 448 * currently effective at the specified *target_fd*. 449 * 450 * Return 451 * Returns zero on success. On error, -1 is returned and *errno* 452 * is set appropriately. 453 * 454 * BPF_RAW_TRACEPOINT_OPEN 455 * Description 456 * Attach an eBPF program to a tracepoint *name* to access kernel 457 * internal arguments of the tracepoint in their raw form. 458 * 459 * The *prog_fd* must be a valid file descriptor associated with 460 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**. 461 * 462 * No ABI guarantees are made about the content of tracepoint 463 * arguments exposed to the corresponding eBPF program. 464 * 465 * Applying **close**\ (2) to the file descriptor returned by 466 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES). 467 * 468 * Return 469 * A new file descriptor (a nonnegative integer), or -1 if an 470 * error occurred (in which case, *errno* is set appropriately). 471 * 472 * BPF_BTF_LOAD 473 * Description 474 * Verify and load BPF Type Format (BTF) metadata into the kernel, 475 * returning a new file descriptor associated with the metadata. 476 * BTF is described in more detail at 477 * https://www.kernel.org/doc/html/latest/bpf/btf.html. 478 * 479 * The *btf* parameter must point to valid memory providing 480 * *btf_size* bytes of BTF binary metadata. 481 * 482 * The returned file descriptor can be passed to other **bpf**\ () 483 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to 484 * associate the BTF with those objects. 485 * 486 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional 487 * parameters to specify a *btf_log_buf*, *btf_log_size* and 488 * *btf_log_level* which allow the kernel to return freeform log 489 * output regarding the BTF verification process. 490 * 491 * Return 492 * A new file descriptor (a nonnegative integer), or -1 if an 493 * error occurred (in which case, *errno* is set appropriately). 494 * 495 * BPF_BTF_GET_FD_BY_ID 496 * Description 497 * Open a file descriptor for the BPF Type Format (BTF) 498 * corresponding to *btf_id*. 499 * 500 * Return 501 * A new file descriptor (a nonnegative integer), or -1 if an 502 * error occurred (in which case, *errno* is set appropriately). 503 * 504 * BPF_TASK_FD_QUERY 505 * Description 506 * Obtain information about eBPF programs associated with the 507 * target process identified by *pid* and *fd*. 508 * 509 * If the *pid* and *fd* are associated with a tracepoint, kprobe 510 * or uprobe perf event, then the *prog_id* and *fd_type* will 511 * be populated with the eBPF program id and file descriptor type 512 * of type **bpf_task_fd_type**. If associated with a kprobe or 513 * uprobe, the *probe_offset* and *probe_addr* will also be 514 * populated. Optionally, if *buf* is provided, then up to 515 * *buf_len* bytes of *buf* will be populated with the name of 516 * the tracepoint, kprobe or uprobe. 517 * 518 * The resulting *prog_id* may be introspected in deeper detail 519 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**. 520 * 521 * Return 522 * Returns zero on success. On error, -1 is returned and *errno* 523 * is set appropriately. 524 * 525 * BPF_MAP_LOOKUP_AND_DELETE_ELEM 526 * Description 527 * Look up an element with the given *key* in the map referred to 528 * by the file descriptor *fd*, and if found, delete the element. 529 * 530 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types 531 * implement this command as a "pop" operation, deleting the top 532 * element rather than one corresponding to *key*. 533 * The *key* and *key_len* parameters should be zeroed when 534 * issuing this operation for these map types. 535 * 536 * This command is only valid for the following map types: 537 * * **BPF_MAP_TYPE_QUEUE** 538 * * **BPF_MAP_TYPE_STACK** 539 * 540 * Return 541 * Returns zero on success. On error, -1 is returned and *errno* 542 * is set appropriately. 543 * 544 * BPF_MAP_FREEZE 545 * Description 546 * Freeze the permissions of the specified map. 547 * 548 * Write permissions may be frozen by passing zero *flags*. 549 * Upon success, no future syscall invocations may alter the 550 * map state of *map_fd*. Write operations from eBPF programs 551 * are still possible for a frozen map. 552 * 553 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**. 554 * 555 * Return 556 * Returns zero on success. On error, -1 is returned and *errno* 557 * is set appropriately. 558 * 559 * BPF_BTF_GET_NEXT_ID 560 * Description 561 * Fetch the next BPF Type Format (BTF) object currently loaded 562 * into the kernel. 563 * 564 * Looks for the BTF object with an id greater than *start_id* 565 * and updates *next_id* on success. If no other BTF objects 566 * remain with ids higher than *start_id*, returns -1 and sets 567 * *errno* to **ENOENT**. 568 * 569 * Return 570 * Returns zero on success. On error, or when no id remains, -1 571 * is returned and *errno* is set appropriately. 572 * 573 * BPF_MAP_LOOKUP_BATCH 574 * Description 575 * Iterate and fetch multiple elements in a map. 576 * 577 * Two opaque values are used to manage batch operations, 578 * *in_batch* and *out_batch*. Initially, *in_batch* must be set 579 * to NULL to begin the batched operation. After each subsequent 580 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant 581 * *out_batch* as the *in_batch* for the next operation to 582 * continue iteration from the current point. 583 * 584 * The *keys* and *values* are output parameters which must point 585 * to memory large enough to hold *count* items based on the key 586 * and value size of the map *map_fd*. The *keys* buffer must be 587 * of *key_size* * *count*. The *values* buffer must be of 588 * *value_size* * *count*. 589 * 590 * The *elem_flags* argument may be specified as one of the 591 * following: 592 * 593 * **BPF_F_LOCK** 594 * Look up the value of a spin-locked map without 595 * returning the lock. This must be specified if the 596 * elements contain a spinlock. 597 * 598 * On success, *count* elements from the map are copied into the 599 * user buffer, with the keys copied into *keys* and the values 600 * copied into the corresponding indices in *values*. 601 * 602 * If an error is returned and *errno* is not **EFAULT**, *count* 603 * is set to the number of successfully processed elements. 604 * 605 * Return 606 * Returns zero on success. On error, -1 is returned and *errno* 607 * is set appropriately. 608 * 609 * May set *errno* to **ENOSPC** to indicate that *keys* or 610 * *values* is too small to dump an entire bucket during 611 * iteration of a hash-based map type. 612 * 613 * BPF_MAP_LOOKUP_AND_DELETE_BATCH 614 * Description 615 * Iterate and delete all elements in a map. 616 * 617 * This operation has the same behavior as 618 * **BPF_MAP_LOOKUP_BATCH** with two exceptions: 619 * 620 * * Every element that is successfully returned is also deleted 621 * from the map. This is at least *count* elements. Note that 622 * *count* is both an input and an output parameter. 623 * * Upon returning with *errno* set to **EFAULT**, up to 624 * *count* elements may be deleted without returning the keys 625 * and values of the deleted elements. 626 * 627 * Return 628 * Returns zero on success. On error, -1 is returned and *errno* 629 * is set appropriately. 630 * 631 * BPF_MAP_UPDATE_BATCH 632 * Description 633 * Update multiple elements in a map by *key*. 634 * 635 * The *keys* and *values* are input parameters which must point 636 * to memory large enough to hold *count* items based on the key 637 * and value size of the map *map_fd*. The *keys* buffer must be 638 * of *key_size* * *count*. The *values* buffer must be of 639 * *value_size* * *count*. 640 * 641 * Each element specified in *keys* is sequentially updated to the 642 * value in the corresponding index in *values*. The *in_batch* 643 * and *out_batch* parameters are ignored and should be zeroed. 644 * 645 * The *elem_flags* argument should be specified as one of the 646 * following: 647 * 648 * **BPF_ANY** 649 * Create new elements or update a existing elements. 650 * **BPF_NOEXIST** 651 * Create new elements only if they do not exist. 652 * **BPF_EXIST** 653 * Update existing elements. 654 * **BPF_F_LOCK** 655 * Update spin_lock-ed map elements. This must be 656 * specified if the map value contains a spinlock. 657 * 658 * On success, *count* elements from the map are updated. 659 * 660 * If an error is returned and *errno* is not **EFAULT**, *count* 661 * is set to the number of successfully processed elements. 662 * 663 * Return 664 * Returns zero on success. On error, -1 is returned and *errno* 665 * is set appropriately. 666 * 667 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or 668 * **E2BIG**. **E2BIG** indicates that the number of elements in 669 * the map reached the *max_entries* limit specified at map 670 * creation time. 671 * 672 * May set *errno* to one of the following error codes under 673 * specific circumstances: 674 * 675 * **EEXIST** 676 * If *flags* specifies **BPF_NOEXIST** and the element 677 * with *key* already exists in the map. 678 * **ENOENT** 679 * If *flags* specifies **BPF_EXIST** and the element with 680 * *key* does not exist in the map. 681 * 682 * BPF_MAP_DELETE_BATCH 683 * Description 684 * Delete multiple elements in a map by *key*. 685 * 686 * The *keys* parameter is an input parameter which must point 687 * to memory large enough to hold *count* items based on the key 688 * size of the map *map_fd*, that is, *key_size* * *count*. 689 * 690 * Each element specified in *keys* is sequentially deleted. The 691 * *in_batch*, *out_batch*, and *values* parameters are ignored 692 * and should be zeroed. 693 * 694 * The *elem_flags* argument may be specified as one of the 695 * following: 696 * 697 * **BPF_F_LOCK** 698 * Look up the value of a spin-locked map without 699 * returning the lock. This must be specified if the 700 * elements contain a spinlock. 701 * 702 * On success, *count* elements from the map are updated. 703 * 704 * If an error is returned and *errno* is not **EFAULT**, *count* 705 * is set to the number of successfully processed elements. If 706 * *errno* is **EFAULT**, up to *count* elements may be been 707 * deleted. 708 * 709 * Return 710 * Returns zero on success. On error, -1 is returned and *errno* 711 * is set appropriately. 712 * 713 * BPF_LINK_CREATE 714 * Description 715 * Attach an eBPF program to a *target_fd* at the specified 716 * *attach_type* hook and return a file descriptor handle for 717 * managing the link. 718 * 719 * Return 720 * A new file descriptor (a nonnegative integer), or -1 if an 721 * error occurred (in which case, *errno* is set appropriately). 722 * 723 * BPF_LINK_UPDATE 724 * Description 725 * Update the eBPF program in the specified *link_fd* to 726 * *new_prog_fd*. 727 * 728 * Return 729 * Returns zero on success. On error, -1 is returned and *errno* 730 * is set appropriately. 731 * 732 * BPF_LINK_GET_FD_BY_ID 733 * Description 734 * Open a file descriptor for the eBPF Link corresponding to 735 * *link_id*. 736 * 737 * Return 738 * A new file descriptor (a nonnegative integer), or -1 if an 739 * error occurred (in which case, *errno* is set appropriately). 740 * 741 * BPF_LINK_GET_NEXT_ID 742 * Description 743 * Fetch the next eBPF link currently loaded into the kernel. 744 * 745 * Looks for the eBPF link with an id greater than *start_id* 746 * and updates *next_id* on success. If no other eBPF links 747 * remain with ids higher than *start_id*, returns -1 and sets 748 * *errno* to **ENOENT**. 749 * 750 * Return 751 * Returns zero on success. On error, or when no id remains, -1 752 * is returned and *errno* is set appropriately. 753 * 754 * BPF_ENABLE_STATS 755 * Description 756 * Enable eBPF runtime statistics gathering. 757 * 758 * Runtime statistics gathering for the eBPF runtime is disabled 759 * by default to minimize the corresponding performance overhead. 760 * This command enables statistics globally. 761 * 762 * Multiple programs may independently enable statistics. 763 * After gathering the desired statistics, eBPF runtime statistics 764 * may be disabled again by calling **close**\ (2) for the file 765 * descriptor returned by this function. Statistics will only be 766 * disabled system-wide when all outstanding file descriptors 767 * returned by prior calls for this subcommand are closed. 768 * 769 * Return 770 * A new file descriptor (a nonnegative integer), or -1 if an 771 * error occurred (in which case, *errno* is set appropriately). 772 * 773 * BPF_ITER_CREATE 774 * Description 775 * Create an iterator on top of the specified *link_fd* (as 776 * previously created using **BPF_LINK_CREATE**) and return a 777 * file descriptor that can be used to trigger the iteration. 778 * 779 * If the resulting file descriptor is pinned to the filesystem 780 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls 781 * for that path will trigger the iterator to read kernel state 782 * using the eBPF program attached to *link_fd*. 783 * 784 * Return 785 * A new file descriptor (a nonnegative integer), or -1 if an 786 * error occurred (in which case, *errno* is set appropriately). 787 * 788 * BPF_LINK_DETACH 789 * Description 790 * Forcefully detach the specified *link_fd* from its 791 * corresponding attachment point. 792 * 793 * Return 794 * Returns zero on success. On error, -1 is returned and *errno* 795 * is set appropriately. 796 * 797 * BPF_PROG_BIND_MAP 798 * Description 799 * Bind a map to the lifetime of an eBPF program. 800 * 801 * The map identified by *map_fd* is bound to the program 802 * identified by *prog_fd* and only released when *prog_fd* is 803 * released. This may be used in cases where metadata should be 804 * associated with a program which otherwise does not contain any 805 * references to the map (for example, embedded in the eBPF 806 * program instructions). 807 * 808 * Return 809 * Returns zero on success. On error, -1 is returned and *errno* 810 * is set appropriately. 811 * 812 * NOTES 813 * eBPF objects (maps and programs) can be shared between processes. 814 * 815 * * After **fork**\ (2), the child inherits file descriptors 816 * referring to the same eBPF objects. 817 * * File descriptors referring to eBPF objects can be transferred over 818 * **unix**\ (7) domain sockets. 819 * * File descriptors referring to eBPF objects can be duplicated in the 820 * usual way, using **dup**\ (2) and similar calls. 821 * * File descriptors referring to eBPF objects can be pinned to the 822 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2). 823 * 824 * An eBPF object is deallocated only after all file descriptors referring 825 * to the object have been closed and no references remain pinned to the 826 * filesystem or attached (for example, bound to a program or device). 827 */ 828 enum bpf_cmd { 829 BPF_MAP_CREATE, 830 BPF_MAP_LOOKUP_ELEM, 831 BPF_MAP_UPDATE_ELEM, 832 BPF_MAP_DELETE_ELEM, 833 BPF_MAP_GET_NEXT_KEY, 834 BPF_PROG_LOAD, 835 BPF_OBJ_PIN, 836 BPF_OBJ_GET, 837 BPF_PROG_ATTACH, 838 BPF_PROG_DETACH, 839 BPF_PROG_TEST_RUN, 840 BPF_PROG_RUN = BPF_PROG_TEST_RUN, 841 BPF_PROG_GET_NEXT_ID, 842 BPF_MAP_GET_NEXT_ID, 843 BPF_PROG_GET_FD_BY_ID, 844 BPF_MAP_GET_FD_BY_ID, 845 BPF_OBJ_GET_INFO_BY_FD, 846 BPF_PROG_QUERY, 847 BPF_RAW_TRACEPOINT_OPEN, 848 BPF_BTF_LOAD, 849 BPF_BTF_GET_FD_BY_ID, 850 BPF_TASK_FD_QUERY, 851 BPF_MAP_LOOKUP_AND_DELETE_ELEM, 852 BPF_MAP_FREEZE, 853 BPF_BTF_GET_NEXT_ID, 854 BPF_MAP_LOOKUP_BATCH, 855 BPF_MAP_LOOKUP_AND_DELETE_BATCH, 856 BPF_MAP_UPDATE_BATCH, 857 BPF_MAP_DELETE_BATCH, 858 BPF_LINK_CREATE, 859 BPF_LINK_UPDATE, 860 BPF_LINK_GET_FD_BY_ID, 861 BPF_LINK_GET_NEXT_ID, 862 BPF_ENABLE_STATS, 863 BPF_ITER_CREATE, 864 BPF_LINK_DETACH, 865 BPF_PROG_BIND_MAP, 866 }; 867 868 enum bpf_map_type { 869 BPF_MAP_TYPE_UNSPEC, 870 BPF_MAP_TYPE_HASH, 871 BPF_MAP_TYPE_ARRAY, 872 BPF_MAP_TYPE_PROG_ARRAY, 873 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 874 BPF_MAP_TYPE_PERCPU_HASH, 875 BPF_MAP_TYPE_PERCPU_ARRAY, 876 BPF_MAP_TYPE_STACK_TRACE, 877 BPF_MAP_TYPE_CGROUP_ARRAY, 878 BPF_MAP_TYPE_LRU_HASH, 879 BPF_MAP_TYPE_LRU_PERCPU_HASH, 880 BPF_MAP_TYPE_LPM_TRIE, 881 BPF_MAP_TYPE_ARRAY_OF_MAPS, 882 BPF_MAP_TYPE_HASH_OF_MAPS, 883 BPF_MAP_TYPE_DEVMAP, 884 BPF_MAP_TYPE_SOCKMAP, 885 BPF_MAP_TYPE_CPUMAP, 886 BPF_MAP_TYPE_XSKMAP, 887 BPF_MAP_TYPE_SOCKHASH, 888 BPF_MAP_TYPE_CGROUP_STORAGE, 889 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 890 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 891 BPF_MAP_TYPE_QUEUE, 892 BPF_MAP_TYPE_STACK, 893 BPF_MAP_TYPE_SK_STORAGE, 894 BPF_MAP_TYPE_DEVMAP_HASH, 895 BPF_MAP_TYPE_STRUCT_OPS, 896 BPF_MAP_TYPE_RINGBUF, 897 BPF_MAP_TYPE_INODE_STORAGE, 898 BPF_MAP_TYPE_TASK_STORAGE, 899 }; 900 901 /* Note that tracing related programs such as 902 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 903 * are not subject to a stable API since kernel internal data 904 * structures can change from release to release and may 905 * therefore break existing tracing BPF programs. Tracing BPF 906 * programs correspond to /a/ specific kernel which is to be 907 * analyzed, and not /a/ specific kernel /and/ all future ones. 908 */ 909 enum bpf_prog_type { 910 BPF_PROG_TYPE_UNSPEC, 911 BPF_PROG_TYPE_SOCKET_FILTER, 912 BPF_PROG_TYPE_KPROBE, 913 BPF_PROG_TYPE_SCHED_CLS, 914 BPF_PROG_TYPE_SCHED_ACT, 915 BPF_PROG_TYPE_TRACEPOINT, 916 BPF_PROG_TYPE_XDP, 917 BPF_PROG_TYPE_PERF_EVENT, 918 BPF_PROG_TYPE_CGROUP_SKB, 919 BPF_PROG_TYPE_CGROUP_SOCK, 920 BPF_PROG_TYPE_LWT_IN, 921 BPF_PROG_TYPE_LWT_OUT, 922 BPF_PROG_TYPE_LWT_XMIT, 923 BPF_PROG_TYPE_SOCK_OPS, 924 BPF_PROG_TYPE_SK_SKB, 925 BPF_PROG_TYPE_CGROUP_DEVICE, 926 BPF_PROG_TYPE_SK_MSG, 927 BPF_PROG_TYPE_RAW_TRACEPOINT, 928 BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 929 BPF_PROG_TYPE_LWT_SEG6LOCAL, 930 BPF_PROG_TYPE_LIRC_MODE2, 931 BPF_PROG_TYPE_SK_REUSEPORT, 932 BPF_PROG_TYPE_FLOW_DISSECTOR, 933 BPF_PROG_TYPE_CGROUP_SYSCTL, 934 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 935 BPF_PROG_TYPE_CGROUP_SOCKOPT, 936 BPF_PROG_TYPE_TRACING, 937 BPF_PROG_TYPE_STRUCT_OPS, 938 BPF_PROG_TYPE_EXT, 939 BPF_PROG_TYPE_LSM, 940 BPF_PROG_TYPE_SK_LOOKUP, 941 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ 942 }; 943 944 enum bpf_attach_type { 945 BPF_CGROUP_INET_INGRESS, 946 BPF_CGROUP_INET_EGRESS, 947 BPF_CGROUP_INET_SOCK_CREATE, 948 BPF_CGROUP_SOCK_OPS, 949 BPF_SK_SKB_STREAM_PARSER, 950 BPF_SK_SKB_STREAM_VERDICT, 951 BPF_CGROUP_DEVICE, 952 BPF_SK_MSG_VERDICT, 953 BPF_CGROUP_INET4_BIND, 954 BPF_CGROUP_INET6_BIND, 955 BPF_CGROUP_INET4_CONNECT, 956 BPF_CGROUP_INET6_CONNECT, 957 BPF_CGROUP_INET4_POST_BIND, 958 BPF_CGROUP_INET6_POST_BIND, 959 BPF_CGROUP_UDP4_SENDMSG, 960 BPF_CGROUP_UDP6_SENDMSG, 961 BPF_LIRC_MODE2, 962 BPF_FLOW_DISSECTOR, 963 BPF_CGROUP_SYSCTL, 964 BPF_CGROUP_UDP4_RECVMSG, 965 BPF_CGROUP_UDP6_RECVMSG, 966 BPF_CGROUP_GETSOCKOPT, 967 BPF_CGROUP_SETSOCKOPT, 968 BPF_TRACE_RAW_TP, 969 BPF_TRACE_FENTRY, 970 BPF_TRACE_FEXIT, 971 BPF_MODIFY_RETURN, 972 BPF_LSM_MAC, 973 BPF_TRACE_ITER, 974 BPF_CGROUP_INET4_GETPEERNAME, 975 BPF_CGROUP_INET6_GETPEERNAME, 976 BPF_CGROUP_INET4_GETSOCKNAME, 977 BPF_CGROUP_INET6_GETSOCKNAME, 978 BPF_XDP_DEVMAP, 979 BPF_CGROUP_INET_SOCK_RELEASE, 980 BPF_XDP_CPUMAP, 981 BPF_SK_LOOKUP, 982 BPF_XDP, 983 BPF_SK_SKB_VERDICT, 984 __MAX_BPF_ATTACH_TYPE 985 }; 986 987 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 988 989 enum bpf_link_type { 990 BPF_LINK_TYPE_UNSPEC = 0, 991 BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 992 BPF_LINK_TYPE_TRACING = 2, 993 BPF_LINK_TYPE_CGROUP = 3, 994 BPF_LINK_TYPE_ITER = 4, 995 BPF_LINK_TYPE_NETNS = 5, 996 BPF_LINK_TYPE_XDP = 6, 997 998 MAX_BPF_LINK_TYPE, 999 }; 1000 1001 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 1002 * 1003 * NONE(default): No further bpf programs allowed in the subtree. 1004 * 1005 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 1006 * the program in this cgroup yields to sub-cgroup program. 1007 * 1008 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 1009 * that cgroup program gets run in addition to the program in this cgroup. 1010 * 1011 * Only one program is allowed to be attached to a cgroup with 1012 * NONE or BPF_F_ALLOW_OVERRIDE flag. 1013 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 1014 * release old program and attach the new one. Attach flags has to match. 1015 * 1016 * Multiple programs are allowed to be attached to a cgroup with 1017 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 1018 * (those that were attached first, run first) 1019 * The programs of sub-cgroup are executed first, then programs of 1020 * this cgroup and then programs of parent cgroup. 1021 * When children program makes decision (like picking TCP CA or sock bind) 1022 * parent program has a chance to override it. 1023 * 1024 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 1025 * programs for a cgroup. Though it's possible to replace an old program at 1026 * any position by also specifying BPF_F_REPLACE flag and position itself in 1027 * replace_bpf_fd attribute. Old program at this position will be released. 1028 * 1029 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 1030 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 1031 * Ex1: 1032 * cgrp1 (MULTI progs A, B) -> 1033 * cgrp2 (OVERRIDE prog C) -> 1034 * cgrp3 (MULTI prog D) -> 1035 * cgrp4 (OVERRIDE prog E) -> 1036 * cgrp5 (NONE prog F) 1037 * the event in cgrp5 triggers execution of F,D,A,B in that order. 1038 * if prog F is detached, the execution is E,D,A,B 1039 * if prog F and D are detached, the execution is E,A,B 1040 * if prog F, E and D are detached, the execution is C,A,B 1041 * 1042 * All eligible programs are executed regardless of return code from 1043 * earlier programs. 1044 */ 1045 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 1046 #define BPF_F_ALLOW_MULTI (1U << 1) 1047 #define BPF_F_REPLACE (1U << 2) 1048 1049 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 1050 * verifier will perform strict alignment checking as if the kernel 1051 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 1052 * and NET_IP_ALIGN defined to 2. 1053 */ 1054 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 1055 1056 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 1057 * verifier will allow any alignment whatsoever. On platforms 1058 * with strict alignment requirements for loads ands stores (such 1059 * as sparc and mips) the verifier validates that all loads and 1060 * stores provably follow this requirement. This flag turns that 1061 * checking and enforcement off. 1062 * 1063 * It is mostly used for testing when we want to validate the 1064 * context and memory access aspects of the verifier, but because 1065 * of an unaligned access the alignment check would trigger before 1066 * the one we are interested in. 1067 */ 1068 #define BPF_F_ANY_ALIGNMENT (1U << 1) 1069 1070 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 1071 * Verifier does sub-register def/use analysis and identifies instructions whose 1072 * def only matters for low 32-bit, high 32-bit is never referenced later 1073 * through implicit zero extension. Therefore verifier notifies JIT back-ends 1074 * that it is safe to ignore clearing high 32-bit for these instructions. This 1075 * saves some back-ends a lot of code-gen. However such optimization is not 1076 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 1077 * hence hasn't used verifier's analysis result. But, we really want to have a 1078 * way to be able to verify the correctness of the described optimization on 1079 * x86_64 on which testsuites are frequently exercised. 1080 * 1081 * So, this flag is introduced. Once it is set, verifier will randomize high 1082 * 32-bit for those instructions who has been identified as safe to ignore them. 1083 * Then, if verifier is not doing correct analysis, such randomization will 1084 * regress tests to expose bugs. 1085 */ 1086 #define BPF_F_TEST_RND_HI32 (1U << 2) 1087 1088 /* The verifier internal test flag. Behavior is undefined */ 1089 #define BPF_F_TEST_STATE_FREQ (1U << 3) 1090 1091 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 1092 * restrict map and helper usage for such programs. Sleepable BPF programs can 1093 * only be attached to hooks where kernel execution context allows sleeping. 1094 * Such programs are allowed to use helpers that may sleep like 1095 * bpf_copy_from_user(). 1096 */ 1097 #define BPF_F_SLEEPABLE (1U << 4) 1098 1099 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 1100 * the following extensions: 1101 * 1102 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX] 1103 * insn[0].imm: map fd or fd_idx 1104 * insn[1].imm: 0 1105 * insn[0].off: 0 1106 * insn[1].off: 0 1107 * ldimm64 rewrite: address of map 1108 * verifier type: CONST_PTR_TO_MAP 1109 */ 1110 #define BPF_PSEUDO_MAP_FD 1 1111 #define BPF_PSEUDO_MAP_IDX 5 1112 1113 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE 1114 * insn[0].imm: map fd or fd_idx 1115 * insn[1].imm: offset into value 1116 * insn[0].off: 0 1117 * insn[1].off: 0 1118 * ldimm64 rewrite: address of map[0]+offset 1119 * verifier type: PTR_TO_MAP_VALUE 1120 */ 1121 #define BPF_PSEUDO_MAP_VALUE 2 1122 #define BPF_PSEUDO_MAP_IDX_VALUE 6 1123 1124 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID 1125 * insn[0].imm: kernel btd id of VAR 1126 * insn[1].imm: 0 1127 * insn[0].off: 0 1128 * insn[1].off: 0 1129 * ldimm64 rewrite: address of the kernel variable 1130 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var 1131 * is struct/union. 1132 */ 1133 #define BPF_PSEUDO_BTF_ID 3 1134 /* insn[0].src_reg: BPF_PSEUDO_FUNC 1135 * insn[0].imm: insn offset to the func 1136 * insn[1].imm: 0 1137 * insn[0].off: 0 1138 * insn[1].off: 0 1139 * ldimm64 rewrite: address of the function 1140 * verifier type: PTR_TO_FUNC. 1141 */ 1142 #define BPF_PSEUDO_FUNC 4 1143 1144 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 1145 * offset to another bpf function 1146 */ 1147 #define BPF_PSEUDO_CALL 1 1148 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL, 1149 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel 1150 */ 1151 #define BPF_PSEUDO_KFUNC_CALL 2 1152 1153 /* flags for BPF_MAP_UPDATE_ELEM command */ 1154 enum { 1155 BPF_ANY = 0, /* create new element or update existing */ 1156 BPF_NOEXIST = 1, /* create new element if it didn't exist */ 1157 BPF_EXIST = 2, /* update existing element */ 1158 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 1159 }; 1160 1161 /* flags for BPF_MAP_CREATE command */ 1162 enum { 1163 BPF_F_NO_PREALLOC = (1U << 0), 1164 /* Instead of having one common LRU list in the 1165 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 1166 * which can scale and perform better. 1167 * Note, the LRU nodes (including free nodes) cannot be moved 1168 * across different LRU lists. 1169 */ 1170 BPF_F_NO_COMMON_LRU = (1U << 1), 1171 /* Specify numa node during map creation */ 1172 BPF_F_NUMA_NODE = (1U << 2), 1173 1174 /* Flags for accessing BPF object from syscall side. */ 1175 BPF_F_RDONLY = (1U << 3), 1176 BPF_F_WRONLY = (1U << 4), 1177 1178 /* Flag for stack_map, store build_id+offset instead of pointer */ 1179 BPF_F_STACK_BUILD_ID = (1U << 5), 1180 1181 /* Zero-initialize hash function seed. This should only be used for testing. */ 1182 BPF_F_ZERO_SEED = (1U << 6), 1183 1184 /* Flags for accessing BPF object from program side. */ 1185 BPF_F_RDONLY_PROG = (1U << 7), 1186 BPF_F_WRONLY_PROG = (1U << 8), 1187 1188 /* Clone map from listener for newly accepted socket */ 1189 BPF_F_CLONE = (1U << 9), 1190 1191 /* Enable memory-mapping BPF map */ 1192 BPF_F_MMAPABLE = (1U << 10), 1193 1194 /* Share perf_event among processes */ 1195 BPF_F_PRESERVE_ELEMS = (1U << 11), 1196 1197 /* Create a map that is suitable to be an inner map with dynamic max entries */ 1198 BPF_F_INNER_MAP = (1U << 12), 1199 }; 1200 1201 /* Flags for BPF_PROG_QUERY. */ 1202 1203 /* Query effective (directly attached + inherited from ancestor cgroups) 1204 * programs that will be executed for events within a cgroup. 1205 * attach_flags with this flag are returned only for directly attached programs. 1206 */ 1207 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 1208 1209 /* Flags for BPF_PROG_TEST_RUN */ 1210 1211 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 1212 #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 1213 1214 /* type for BPF_ENABLE_STATS */ 1215 enum bpf_stats_type { 1216 /* enabled run_time_ns and run_cnt */ 1217 BPF_STATS_RUN_TIME = 0, 1218 }; 1219 1220 enum bpf_stack_build_id_status { 1221 /* user space need an empty entry to identify end of a trace */ 1222 BPF_STACK_BUILD_ID_EMPTY = 0, 1223 /* with valid build_id and offset */ 1224 BPF_STACK_BUILD_ID_VALID = 1, 1225 /* couldn't get build_id, fallback to ip */ 1226 BPF_STACK_BUILD_ID_IP = 2, 1227 }; 1228 1229 #define BPF_BUILD_ID_SIZE 20 1230 struct bpf_stack_build_id { 1231 __s32 status; 1232 unsigned char build_id[BPF_BUILD_ID_SIZE]; 1233 union { 1234 __u64 offset; 1235 __u64 ip; 1236 }; 1237 }; 1238 1239 #define BPF_OBJ_NAME_LEN 16U 1240 1241 union bpf_attr { 1242 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 1243 __u32 map_type; /* one of enum bpf_map_type */ 1244 __u32 key_size; /* size of key in bytes */ 1245 __u32 value_size; /* size of value in bytes */ 1246 __u32 max_entries; /* max number of entries in a map */ 1247 __u32 map_flags; /* BPF_MAP_CREATE related 1248 * flags defined above. 1249 */ 1250 __u32 inner_map_fd; /* fd pointing to the inner map */ 1251 __u32 numa_node; /* numa node (effective only if 1252 * BPF_F_NUMA_NODE is set). 1253 */ 1254 char map_name[BPF_OBJ_NAME_LEN]; 1255 __u32 map_ifindex; /* ifindex of netdev to create on */ 1256 __u32 btf_fd; /* fd pointing to a BTF type data */ 1257 __u32 btf_key_type_id; /* BTF type_id of the key */ 1258 __u32 btf_value_type_id; /* BTF type_id of the value */ 1259 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 1260 * struct stored as the 1261 * map value 1262 */ 1263 }; 1264 1265 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 1266 __u32 map_fd; 1267 __aligned_u64 key; 1268 union { 1269 __aligned_u64 value; 1270 __aligned_u64 next_key; 1271 }; 1272 __u64 flags; 1273 }; 1274 1275 struct { /* struct used by BPF_MAP_*_BATCH commands */ 1276 __aligned_u64 in_batch; /* start batch, 1277 * NULL to start from beginning 1278 */ 1279 __aligned_u64 out_batch; /* output: next start batch */ 1280 __aligned_u64 keys; 1281 __aligned_u64 values; 1282 __u32 count; /* input/output: 1283 * input: # of key/value 1284 * elements 1285 * output: # of filled elements 1286 */ 1287 __u32 map_fd; 1288 __u64 elem_flags; 1289 __u64 flags; 1290 } batch; 1291 1292 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 1293 __u32 prog_type; /* one of enum bpf_prog_type */ 1294 __u32 insn_cnt; 1295 __aligned_u64 insns; 1296 __aligned_u64 license; 1297 __u32 log_level; /* verbosity level of verifier */ 1298 __u32 log_size; /* size of user buffer */ 1299 __aligned_u64 log_buf; /* user supplied buffer */ 1300 __u32 kern_version; /* not used */ 1301 __u32 prog_flags; 1302 char prog_name[BPF_OBJ_NAME_LEN]; 1303 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 1304 /* For some prog types expected attach type must be known at 1305 * load time to verify attach type specific parts of prog 1306 * (context accesses, allowed helpers, etc). 1307 */ 1308 __u32 expected_attach_type; 1309 __u32 prog_btf_fd; /* fd pointing to BTF type data */ 1310 __u32 func_info_rec_size; /* userspace bpf_func_info size */ 1311 __aligned_u64 func_info; /* func info */ 1312 __u32 func_info_cnt; /* number of bpf_func_info records */ 1313 __u32 line_info_rec_size; /* userspace bpf_line_info size */ 1314 __aligned_u64 line_info; /* line info */ 1315 __u32 line_info_cnt; /* number of bpf_line_info records */ 1316 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1317 union { 1318 /* valid prog_fd to attach to bpf prog */ 1319 __u32 attach_prog_fd; 1320 /* or valid module BTF object fd or 0 to attach to vmlinux */ 1321 __u32 attach_btf_obj_fd; 1322 }; 1323 __u32 :32; /* pad */ 1324 __aligned_u64 fd_array; /* array of FDs */ 1325 }; 1326 1327 struct { /* anonymous struct used by BPF_OBJ_* commands */ 1328 __aligned_u64 pathname; 1329 __u32 bpf_fd; 1330 __u32 file_flags; 1331 }; 1332 1333 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 1334 __u32 target_fd; /* container object to attach to */ 1335 __u32 attach_bpf_fd; /* eBPF program to attach */ 1336 __u32 attach_type; 1337 __u32 attach_flags; 1338 __u32 replace_bpf_fd; /* previously attached eBPF 1339 * program to replace if 1340 * BPF_F_REPLACE is used 1341 */ 1342 }; 1343 1344 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 1345 __u32 prog_fd; 1346 __u32 retval; 1347 __u32 data_size_in; /* input: len of data_in */ 1348 __u32 data_size_out; /* input/output: len of data_out 1349 * returns ENOSPC if data_out 1350 * is too small. 1351 */ 1352 __aligned_u64 data_in; 1353 __aligned_u64 data_out; 1354 __u32 repeat; 1355 __u32 duration; 1356 __u32 ctx_size_in; /* input: len of ctx_in */ 1357 __u32 ctx_size_out; /* input/output: len of ctx_out 1358 * returns ENOSPC if ctx_out 1359 * is too small. 1360 */ 1361 __aligned_u64 ctx_in; 1362 __aligned_u64 ctx_out; 1363 __u32 flags; 1364 __u32 cpu; 1365 } test; 1366 1367 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 1368 union { 1369 __u32 start_id; 1370 __u32 prog_id; 1371 __u32 map_id; 1372 __u32 btf_id; 1373 __u32 link_id; 1374 }; 1375 __u32 next_id; 1376 __u32 open_flags; 1377 }; 1378 1379 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 1380 __u32 bpf_fd; 1381 __u32 info_len; 1382 __aligned_u64 info; 1383 } info; 1384 1385 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 1386 __u32 target_fd; /* container object to query */ 1387 __u32 attach_type; 1388 __u32 query_flags; 1389 __u32 attach_flags; 1390 __aligned_u64 prog_ids; 1391 __u32 prog_cnt; 1392 } query; 1393 1394 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 1395 __u64 name; 1396 __u32 prog_fd; 1397 } raw_tracepoint; 1398 1399 struct { /* anonymous struct for BPF_BTF_LOAD */ 1400 __aligned_u64 btf; 1401 __aligned_u64 btf_log_buf; 1402 __u32 btf_size; 1403 __u32 btf_log_size; 1404 __u32 btf_log_level; 1405 }; 1406 1407 struct { 1408 __u32 pid; /* input: pid */ 1409 __u32 fd; /* input: fd */ 1410 __u32 flags; /* input: flags */ 1411 __u32 buf_len; /* input/output: buf len */ 1412 __aligned_u64 buf; /* input/output: 1413 * tp_name for tracepoint 1414 * symbol for kprobe 1415 * filename for uprobe 1416 */ 1417 __u32 prog_id; /* output: prod_id */ 1418 __u32 fd_type; /* output: BPF_FD_TYPE_* */ 1419 __u64 probe_offset; /* output: probe_offset */ 1420 __u64 probe_addr; /* output: probe_addr */ 1421 } task_fd_query; 1422 1423 struct { /* struct used by BPF_LINK_CREATE command */ 1424 __u32 prog_fd; /* eBPF program to attach */ 1425 union { 1426 __u32 target_fd; /* object to attach to */ 1427 __u32 target_ifindex; /* target ifindex */ 1428 }; 1429 __u32 attach_type; /* attach type */ 1430 __u32 flags; /* extra flags */ 1431 union { 1432 __u32 target_btf_id; /* btf_id of target to attach to */ 1433 struct { 1434 __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 1435 __u32 iter_info_len; /* iter_info length */ 1436 }; 1437 }; 1438 } link_create; 1439 1440 struct { /* struct used by BPF_LINK_UPDATE command */ 1441 __u32 link_fd; /* link fd */ 1442 /* new program fd to update link with */ 1443 __u32 new_prog_fd; 1444 __u32 flags; /* extra flags */ 1445 /* expected link's program fd; is specified only if 1446 * BPF_F_REPLACE flag is set in flags */ 1447 __u32 old_prog_fd; 1448 } link_update; 1449 1450 struct { 1451 __u32 link_fd; 1452 } link_detach; 1453 1454 struct { /* struct used by BPF_ENABLE_STATS command */ 1455 __u32 type; 1456 } enable_stats; 1457 1458 struct { /* struct used by BPF_ITER_CREATE command */ 1459 __u32 link_fd; 1460 __u32 flags; 1461 } iter_create; 1462 1463 struct { /* struct used by BPF_PROG_BIND_MAP command */ 1464 __u32 prog_fd; 1465 __u32 map_fd; 1466 __u32 flags; /* extra flags */ 1467 } prog_bind_map; 1468 1469 } __attribute__((aligned(8))); 1470 1471 /* The description below is an attempt at providing documentation to eBPF 1472 * developers about the multiple available eBPF helper functions. It can be 1473 * parsed and used to produce a manual page. The workflow is the following, 1474 * and requires the rst2man utility: 1475 * 1476 * $ ./scripts/bpf_doc.py \ 1477 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 1478 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 1479 * $ man /tmp/bpf-helpers.7 1480 * 1481 * Note that in order to produce this external documentation, some RST 1482 * formatting is used in the descriptions to get "bold" and "italics" in 1483 * manual pages. Also note that the few trailing white spaces are 1484 * intentional, removing them would break paragraphs for rst2man. 1485 * 1486 * Start of BPF helper function descriptions: 1487 * 1488 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 1489 * Description 1490 * Perform a lookup in *map* for an entry associated to *key*. 1491 * Return 1492 * Map value associated to *key*, or **NULL** if no entry was 1493 * found. 1494 * 1495 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 1496 * Description 1497 * Add or update the value of the entry associated to *key* in 1498 * *map* with *value*. *flags* is one of: 1499 * 1500 * **BPF_NOEXIST** 1501 * The entry for *key* must not exist in the map. 1502 * **BPF_EXIST** 1503 * The entry for *key* must already exist in the map. 1504 * **BPF_ANY** 1505 * No condition on the existence of the entry for *key*. 1506 * 1507 * Flag value **BPF_NOEXIST** cannot be used for maps of types 1508 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 1509 * elements always exist), the helper would return an error. 1510 * Return 1511 * 0 on success, or a negative error in case of failure. 1512 * 1513 * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 1514 * Description 1515 * Delete entry with *key* from *map*. 1516 * Return 1517 * 0 on success, or a negative error in case of failure. 1518 * 1519 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 1520 * Description 1521 * For tracing programs, safely attempt to read *size* bytes from 1522 * kernel space address *unsafe_ptr* and store the data in *dst*. 1523 * 1524 * Generally, use **bpf_probe_read_user**\ () or 1525 * **bpf_probe_read_kernel**\ () instead. 1526 * Return 1527 * 0 on success, or a negative error in case of failure. 1528 * 1529 * u64 bpf_ktime_get_ns(void) 1530 * Description 1531 * Return the time elapsed since system boot, in nanoseconds. 1532 * Does not include time the system was suspended. 1533 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 1534 * Return 1535 * Current *ktime*. 1536 * 1537 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 1538 * Description 1539 * This helper is a "printk()-like" facility for debugging. It 1540 * prints a message defined by format *fmt* (of size *fmt_size*) 1541 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 1542 * available. It can take up to three additional **u64** 1543 * arguments (as an eBPF helpers, the total number of arguments is 1544 * limited to five). 1545 * 1546 * Each time the helper is called, it appends a line to the trace. 1547 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 1548 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 1549 * The format of the trace is customizable, and the exact output 1550 * one will get depends on the options set in 1551 * *\/sys/kernel/debug/tracing/trace_options* (see also the 1552 * *README* file under the same directory). However, it usually 1553 * defaults to something like: 1554 * 1555 * :: 1556 * 1557 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 1558 * 1559 * In the above: 1560 * 1561 * * ``telnet`` is the name of the current task. 1562 * * ``470`` is the PID of the current task. 1563 * * ``001`` is the CPU number on which the task is 1564 * running. 1565 * * In ``.N..``, each character refers to a set of 1566 * options (whether irqs are enabled, scheduling 1567 * options, whether hard/softirqs are running, level of 1568 * preempt_disabled respectively). **N** means that 1569 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 1570 * are set. 1571 * * ``419421.045894`` is a timestamp. 1572 * * ``0x00000001`` is a fake value used by BPF for the 1573 * instruction pointer register. 1574 * * ``<formatted msg>`` is the message formatted with 1575 * *fmt*. 1576 * 1577 * The conversion specifiers supported by *fmt* are similar, but 1578 * more limited than for printk(). They are **%d**, **%i**, 1579 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 1580 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 1581 * of field, padding with zeroes, etc.) is available, and the 1582 * helper will return **-EINVAL** (but print nothing) if it 1583 * encounters an unknown specifier. 1584 * 1585 * Also, note that **bpf_trace_printk**\ () is slow, and should 1586 * only be used for debugging purposes. For this reason, a notice 1587 * block (spanning several lines) is printed to kernel logs and 1588 * states that the helper should not be used "for production use" 1589 * the first time this helper is used (or more precisely, when 1590 * **trace_printk**\ () buffers are allocated). For passing values 1591 * to user space, perf events should be preferred. 1592 * Return 1593 * The number of bytes written to the buffer, or a negative error 1594 * in case of failure. 1595 * 1596 * u32 bpf_get_prandom_u32(void) 1597 * Description 1598 * Get a pseudo-random number. 1599 * 1600 * From a security point of view, this helper uses its own 1601 * pseudo-random internal state, and cannot be used to infer the 1602 * seed of other random functions in the kernel. However, it is 1603 * essential to note that the generator used by the helper is not 1604 * cryptographically secure. 1605 * Return 1606 * A random 32-bit unsigned value. 1607 * 1608 * u32 bpf_get_smp_processor_id(void) 1609 * Description 1610 * Get the SMP (symmetric multiprocessing) processor id. Note that 1611 * all programs run with preemption disabled, which means that the 1612 * SMP processor id is stable during all the execution of the 1613 * program. 1614 * Return 1615 * The SMP id of the processor running the program. 1616 * 1617 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 1618 * Description 1619 * Store *len* bytes from address *from* into the packet 1620 * associated to *skb*, at *offset*. *flags* are a combination of 1621 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 1622 * checksum for the packet after storing the bytes) and 1623 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 1624 * **->swhash** and *skb*\ **->l4hash** to 0). 1625 * 1626 * A call to this helper is susceptible to change the underlying 1627 * packet buffer. Therefore, at load time, all checks on pointers 1628 * previously done by the verifier are invalidated and must be 1629 * performed again, if the helper is used in combination with 1630 * direct packet access. 1631 * Return 1632 * 0 on success, or a negative error in case of failure. 1633 * 1634 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 1635 * Description 1636 * Recompute the layer 3 (e.g. IP) checksum for the packet 1637 * associated to *skb*. Computation is incremental, so the helper 1638 * must know the former value of the header field that was 1639 * modified (*from*), the new value of this field (*to*), and the 1640 * number of bytes (2 or 4) for this field, stored in *size*. 1641 * Alternatively, it is possible to store the difference between 1642 * the previous and the new values of the header field in *to*, by 1643 * setting *from* and *size* to 0. For both methods, *offset* 1644 * indicates the location of the IP checksum within the packet. 1645 * 1646 * This helper works in combination with **bpf_csum_diff**\ (), 1647 * which does not update the checksum in-place, but offers more 1648 * flexibility and can handle sizes larger than 2 or 4 for the 1649 * checksum to update. 1650 * 1651 * A call to this helper is susceptible to change the underlying 1652 * packet buffer. Therefore, at load time, all checks on pointers 1653 * previously done by the verifier are invalidated and must be 1654 * performed again, if the helper is used in combination with 1655 * direct packet access. 1656 * Return 1657 * 0 on success, or a negative error in case of failure. 1658 * 1659 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 1660 * Description 1661 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 1662 * packet associated to *skb*. Computation is incremental, so the 1663 * helper must know the former value of the header field that was 1664 * modified (*from*), the new value of this field (*to*), and the 1665 * number of bytes (2 or 4) for this field, stored on the lowest 1666 * four bits of *flags*. Alternatively, it is possible to store 1667 * the difference between the previous and the new values of the 1668 * header field in *to*, by setting *from* and the four lowest 1669 * bits of *flags* to 0. For both methods, *offset* indicates the 1670 * location of the IP checksum within the packet. In addition to 1671 * the size of the field, *flags* can be added (bitwise OR) actual 1672 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 1673 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 1674 * for updates resulting in a null checksum the value is set to 1675 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 1676 * the checksum is to be computed against a pseudo-header. 1677 * 1678 * This helper works in combination with **bpf_csum_diff**\ (), 1679 * which does not update the checksum in-place, but offers more 1680 * flexibility and can handle sizes larger than 2 or 4 for the 1681 * checksum to update. 1682 * 1683 * A call to this helper is susceptible to change the underlying 1684 * packet buffer. Therefore, at load time, all checks on pointers 1685 * previously done by the verifier are invalidated and must be 1686 * performed again, if the helper is used in combination with 1687 * direct packet access. 1688 * Return 1689 * 0 on success, or a negative error in case of failure. 1690 * 1691 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 1692 * Description 1693 * This special helper is used to trigger a "tail call", or in 1694 * other words, to jump into another eBPF program. The same stack 1695 * frame is used (but values on stack and in registers for the 1696 * caller are not accessible to the callee). This mechanism allows 1697 * for program chaining, either for raising the maximum number of 1698 * available eBPF instructions, or to execute given programs in 1699 * conditional blocks. For security reasons, there is an upper 1700 * limit to the number of successive tail calls that can be 1701 * performed. 1702 * 1703 * Upon call of this helper, the program attempts to jump into a 1704 * program referenced at index *index* in *prog_array_map*, a 1705 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 1706 * *ctx*, a pointer to the context. 1707 * 1708 * If the call succeeds, the kernel immediately runs the first 1709 * instruction of the new program. This is not a function call, 1710 * and it never returns to the previous program. If the call 1711 * fails, then the helper has no effect, and the caller continues 1712 * to run its subsequent instructions. A call can fail if the 1713 * destination program for the jump does not exist (i.e. *index* 1714 * is superior to the number of entries in *prog_array_map*), or 1715 * if the maximum number of tail calls has been reached for this 1716 * chain of programs. This limit is defined in the kernel by the 1717 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 1718 * which is currently set to 32. 1719 * Return 1720 * 0 on success, or a negative error in case of failure. 1721 * 1722 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 1723 * Description 1724 * Clone and redirect the packet associated to *skb* to another 1725 * net device of index *ifindex*. Both ingress and egress 1726 * interfaces can be used for redirection. The **BPF_F_INGRESS** 1727 * value in *flags* is used to make the distinction (ingress path 1728 * is selected if the flag is present, egress path otherwise). 1729 * This is the only flag supported for now. 1730 * 1731 * In comparison with **bpf_redirect**\ () helper, 1732 * **bpf_clone_redirect**\ () has the associated cost of 1733 * duplicating the packet buffer, but this can be executed out of 1734 * the eBPF program. Conversely, **bpf_redirect**\ () is more 1735 * efficient, but it is handled through an action code where the 1736 * redirection happens only after the eBPF program has returned. 1737 * 1738 * A call to this helper is susceptible to change the underlying 1739 * packet buffer. Therefore, at load time, all checks on pointers 1740 * previously done by the verifier are invalidated and must be 1741 * performed again, if the helper is used in combination with 1742 * direct packet access. 1743 * Return 1744 * 0 on success, or a negative error in case of failure. 1745 * 1746 * u64 bpf_get_current_pid_tgid(void) 1747 * Return 1748 * A 64-bit integer containing the current tgid and pid, and 1749 * created as such: 1750 * *current_task*\ **->tgid << 32 \|** 1751 * *current_task*\ **->pid**. 1752 * 1753 * u64 bpf_get_current_uid_gid(void) 1754 * Return 1755 * A 64-bit integer containing the current GID and UID, and 1756 * created as such: *current_gid* **<< 32 \|** *current_uid*. 1757 * 1758 * long bpf_get_current_comm(void *buf, u32 size_of_buf) 1759 * Description 1760 * Copy the **comm** attribute of the current task into *buf* of 1761 * *size_of_buf*. The **comm** attribute contains the name of 1762 * the executable (excluding the path) for the current task. The 1763 * *size_of_buf* must be strictly positive. On success, the 1764 * helper makes sure that the *buf* is NUL-terminated. On failure, 1765 * it is filled with zeroes. 1766 * Return 1767 * 0 on success, or a negative error in case of failure. 1768 * 1769 * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 1770 * Description 1771 * Retrieve the classid for the current task, i.e. for the net_cls 1772 * cgroup to which *skb* belongs. 1773 * 1774 * This helper can be used on TC egress path, but not on ingress. 1775 * 1776 * The net_cls cgroup provides an interface to tag network packets 1777 * based on a user-provided identifier for all traffic coming from 1778 * the tasks belonging to the related cgroup. See also the related 1779 * kernel documentation, available from the Linux sources in file 1780 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 1781 * 1782 * The Linux kernel has two versions for cgroups: there are 1783 * cgroups v1 and cgroups v2. Both are available to users, who can 1784 * use a mixture of them, but note that the net_cls cgroup is for 1785 * cgroup v1 only. This makes it incompatible with BPF programs 1786 * run on cgroups, which is a cgroup-v2-only feature (a socket can 1787 * only hold data for one version of cgroups at a time). 1788 * 1789 * This helper is only available is the kernel was compiled with 1790 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 1791 * "**y**" or to "**m**". 1792 * Return 1793 * The classid, or 0 for the default unconfigured classid. 1794 * 1795 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1796 * Description 1797 * Push a *vlan_tci* (VLAN tag control information) of protocol 1798 * *vlan_proto* to the packet associated to *skb*, then update 1799 * the checksum. Note that if *vlan_proto* is different from 1800 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1801 * be **ETH_P_8021Q**. 1802 * 1803 * A call to this helper is susceptible to change the underlying 1804 * packet buffer. Therefore, at load time, all checks on pointers 1805 * previously done by the verifier are invalidated and must be 1806 * performed again, if the helper is used in combination with 1807 * direct packet access. 1808 * Return 1809 * 0 on success, or a negative error in case of failure. 1810 * 1811 * long bpf_skb_vlan_pop(struct sk_buff *skb) 1812 * Description 1813 * Pop a VLAN header from the packet associated to *skb*. 1814 * 1815 * A call to this helper is susceptible to change the underlying 1816 * packet buffer. Therefore, at load time, all checks on pointers 1817 * previously done by the verifier are invalidated and must be 1818 * performed again, if the helper is used in combination with 1819 * direct packet access. 1820 * Return 1821 * 0 on success, or a negative error in case of failure. 1822 * 1823 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1824 * Description 1825 * Get tunnel metadata. This helper takes a pointer *key* to an 1826 * empty **struct bpf_tunnel_key** of **size**, that will be 1827 * filled with tunnel metadata for the packet associated to *skb*. 1828 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1829 * indicates that the tunnel is based on IPv6 protocol instead of 1830 * IPv4. 1831 * 1832 * The **struct bpf_tunnel_key** is an object that generalizes the 1833 * principal parameters used by various tunneling protocols into a 1834 * single struct. This way, it can be used to easily make a 1835 * decision based on the contents of the encapsulation header, 1836 * "summarized" in this struct. In particular, it holds the IP 1837 * address of the remote end (IPv4 or IPv6, depending on the case) 1838 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1839 * this struct exposes the *key*\ **->tunnel_id**, which is 1840 * generally mapped to a VNI (Virtual Network Identifier), making 1841 * it programmable together with the **bpf_skb_set_tunnel_key**\ 1842 * () helper. 1843 * 1844 * Let's imagine that the following code is part of a program 1845 * attached to the TC ingress interface, on one end of a GRE 1846 * tunnel, and is supposed to filter out all messages coming from 1847 * remote ends with IPv4 address other than 10.0.0.1: 1848 * 1849 * :: 1850 * 1851 * int ret; 1852 * struct bpf_tunnel_key key = {}; 1853 * 1854 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1855 * if (ret < 0) 1856 * return TC_ACT_SHOT; // drop packet 1857 * 1858 * if (key.remote_ipv4 != 0x0a000001) 1859 * return TC_ACT_SHOT; // drop packet 1860 * 1861 * return TC_ACT_OK; // accept packet 1862 * 1863 * This interface can also be used with all encapsulation devices 1864 * that can operate in "collect metadata" mode: instead of having 1865 * one network device per specific configuration, the "collect 1866 * metadata" mode only requires a single device where the 1867 * configuration can be extracted from this helper. 1868 * 1869 * This can be used together with various tunnels such as VXLan, 1870 * Geneve, GRE or IP in IP (IPIP). 1871 * Return 1872 * 0 on success, or a negative error in case of failure. 1873 * 1874 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1875 * Description 1876 * Populate tunnel metadata for packet associated to *skb.* The 1877 * tunnel metadata is set to the contents of *key*, of *size*. The 1878 * *flags* can be set to a combination of the following values: 1879 * 1880 * **BPF_F_TUNINFO_IPV6** 1881 * Indicate that the tunnel is based on IPv6 protocol 1882 * instead of IPv4. 1883 * **BPF_F_ZERO_CSUM_TX** 1884 * For IPv4 packets, add a flag to tunnel metadata 1885 * indicating that checksum computation should be skipped 1886 * and checksum set to zeroes. 1887 * **BPF_F_DONT_FRAGMENT** 1888 * Add a flag to tunnel metadata indicating that the 1889 * packet should not be fragmented. 1890 * **BPF_F_SEQ_NUMBER** 1891 * Add a flag to tunnel metadata indicating that a 1892 * sequence number should be added to tunnel header before 1893 * sending the packet. This flag was added for GRE 1894 * encapsulation, but might be used with other protocols 1895 * as well in the future. 1896 * 1897 * Here is a typical usage on the transmit path: 1898 * 1899 * :: 1900 * 1901 * struct bpf_tunnel_key key; 1902 * populate key ... 1903 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1904 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1905 * 1906 * See also the description of the **bpf_skb_get_tunnel_key**\ () 1907 * helper for additional information. 1908 * Return 1909 * 0 on success, or a negative error in case of failure. 1910 * 1911 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1912 * Description 1913 * Read the value of a perf event counter. This helper relies on a 1914 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1915 * the perf event counter is selected when *map* is updated with 1916 * perf event file descriptors. The *map* is an array whose size 1917 * is the number of available CPUs, and each cell contains a value 1918 * relative to one CPU. The value to retrieve is indicated by 1919 * *flags*, that contains the index of the CPU to look up, masked 1920 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1921 * **BPF_F_CURRENT_CPU** to indicate that the value for the 1922 * current CPU should be retrieved. 1923 * 1924 * Note that before Linux 4.13, only hardware perf event can be 1925 * retrieved. 1926 * 1927 * Also, be aware that the newer helper 1928 * **bpf_perf_event_read_value**\ () is recommended over 1929 * **bpf_perf_event_read**\ () in general. The latter has some ABI 1930 * quirks where error and counter value are used as a return code 1931 * (which is wrong to do since ranges may overlap). This issue is 1932 * fixed with **bpf_perf_event_read_value**\ (), which at the same 1933 * time provides more features over the **bpf_perf_event_read**\ 1934 * () interface. Please refer to the description of 1935 * **bpf_perf_event_read_value**\ () for details. 1936 * Return 1937 * The value of the perf event counter read from the map, or a 1938 * negative error code in case of failure. 1939 * 1940 * long bpf_redirect(u32 ifindex, u64 flags) 1941 * Description 1942 * Redirect the packet to another net device of index *ifindex*. 1943 * This helper is somewhat similar to **bpf_clone_redirect**\ 1944 * (), except that the packet is not cloned, which provides 1945 * increased performance. 1946 * 1947 * Except for XDP, both ingress and egress interfaces can be used 1948 * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1949 * to make the distinction (ingress path is selected if the flag 1950 * is present, egress path otherwise). Currently, XDP only 1951 * supports redirection to the egress interface, and accepts no 1952 * flag at all. 1953 * 1954 * The same effect can also be attained with the more generic 1955 * **bpf_redirect_map**\ (), which uses a BPF map to store the 1956 * redirect target instead of providing it directly to the helper. 1957 * Return 1958 * For XDP, the helper returns **XDP_REDIRECT** on success or 1959 * **XDP_ABORTED** on error. For other program types, the values 1960 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1961 * error. 1962 * 1963 * u32 bpf_get_route_realm(struct sk_buff *skb) 1964 * Description 1965 * Retrieve the realm or the route, that is to say the 1966 * **tclassid** field of the destination for the *skb*. The 1967 * identifier retrieved is a user-provided tag, similar to the 1968 * one used with the net_cls cgroup (see description for 1969 * **bpf_get_cgroup_classid**\ () helper), but here this tag is 1970 * held by a route (a destination entry), not by a task. 1971 * 1972 * Retrieving this identifier works with the clsact TC egress hook 1973 * (see also **tc-bpf(8)**), or alternatively on conventional 1974 * classful egress qdiscs, but not on TC ingress path. In case of 1975 * clsact TC egress hook, this has the advantage that, internally, 1976 * the destination entry has not been dropped yet in the transmit 1977 * path. Therefore, the destination entry does not need to be 1978 * artificially held via **netif_keep_dst**\ () for a classful 1979 * qdisc until the *skb* is freed. 1980 * 1981 * This helper is available only if the kernel was compiled with 1982 * **CONFIG_IP_ROUTE_CLASSID** configuration option. 1983 * Return 1984 * The realm of the route for the packet associated to *skb*, or 0 1985 * if none was found. 1986 * 1987 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1988 * Description 1989 * Write raw *data* blob into a special BPF perf event held by 1990 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1991 * event must have the following attributes: **PERF_SAMPLE_RAW** 1992 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1993 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1994 * 1995 * The *flags* are used to indicate the index in *map* for which 1996 * the value must be put, masked with **BPF_F_INDEX_MASK**. 1997 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1998 * to indicate that the index of the current CPU core should be 1999 * used. 2000 * 2001 * The value to write, of *size*, is passed through eBPF stack and 2002 * pointed by *data*. 2003 * 2004 * The context of the program *ctx* needs also be passed to the 2005 * helper. 2006 * 2007 * On user space, a program willing to read the values needs to 2008 * call **perf_event_open**\ () on the perf event (either for 2009 * one or for all CPUs) and to store the file descriptor into the 2010 * *map*. This must be done before the eBPF program can send data 2011 * into it. An example is available in file 2012 * *samples/bpf/trace_output_user.c* in the Linux kernel source 2013 * tree (the eBPF program counterpart is in 2014 * *samples/bpf/trace_output_kern.c*). 2015 * 2016 * **bpf_perf_event_output**\ () achieves better performance 2017 * than **bpf_trace_printk**\ () for sharing data with user 2018 * space, and is much better suitable for streaming data from eBPF 2019 * programs. 2020 * 2021 * Note that this helper is not restricted to tracing use cases 2022 * and can be used with programs attached to TC or XDP as well, 2023 * where it allows for passing data to user space listeners. Data 2024 * can be: 2025 * 2026 * * Only custom structs, 2027 * * Only the packet payload, or 2028 * * A combination of both. 2029 * Return 2030 * 0 on success, or a negative error in case of failure. 2031 * 2032 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 2033 * Description 2034 * This helper was provided as an easy way to load data from a 2035 * packet. It can be used to load *len* bytes from *offset* from 2036 * the packet associated to *skb*, into the buffer pointed by 2037 * *to*. 2038 * 2039 * Since Linux 4.7, usage of this helper has mostly been replaced 2040 * by "direct packet access", enabling packet data to be 2041 * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 2042 * pointing respectively to the first byte of packet data and to 2043 * the byte after the last byte of packet data. However, it 2044 * remains useful if one wishes to read large quantities of data 2045 * at once from a packet into the eBPF stack. 2046 * Return 2047 * 0 on success, or a negative error in case of failure. 2048 * 2049 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 2050 * Description 2051 * Walk a user or a kernel stack and return its id. To achieve 2052 * this, the helper needs *ctx*, which is a pointer to the context 2053 * on which the tracing program is executed, and a pointer to a 2054 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 2055 * 2056 * The last argument, *flags*, holds the number of stack frames to 2057 * skip (from 0 to 255), masked with 2058 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2059 * a combination of the following flags: 2060 * 2061 * **BPF_F_USER_STACK** 2062 * Collect a user space stack instead of a kernel stack. 2063 * **BPF_F_FAST_STACK_CMP** 2064 * Compare stacks by hash only. 2065 * **BPF_F_REUSE_STACKID** 2066 * If two different stacks hash into the same *stackid*, 2067 * discard the old one. 2068 * 2069 * The stack id retrieved is a 32 bit long integer handle which 2070 * can be further combined with other data (including other stack 2071 * ids) and used as a key into maps. This can be useful for 2072 * generating a variety of graphs (such as flame graphs or off-cpu 2073 * graphs). 2074 * 2075 * For walking a stack, this helper is an improvement over 2076 * **bpf_probe_read**\ (), which can be used with unrolled loops 2077 * but is not efficient and consumes a lot of eBPF instructions. 2078 * Instead, **bpf_get_stackid**\ () can collect up to 2079 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 2080 * this limit can be controlled with the **sysctl** program, and 2081 * that it should be manually increased in order to profile long 2082 * user stacks (such as stacks for Java programs). To do so, use: 2083 * 2084 * :: 2085 * 2086 * # sysctl kernel.perf_event_max_stack=<new value> 2087 * Return 2088 * The positive or null stack id on success, or a negative error 2089 * in case of failure. 2090 * 2091 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 2092 * Description 2093 * Compute a checksum difference, from the raw buffer pointed by 2094 * *from*, of length *from_size* (that must be a multiple of 4), 2095 * towards the raw buffer pointed by *to*, of size *to_size* 2096 * (same remark). An optional *seed* can be added to the value 2097 * (this can be cascaded, the seed may come from a previous call 2098 * to the helper). 2099 * 2100 * This is flexible enough to be used in several ways: 2101 * 2102 * * With *from_size* == 0, *to_size* > 0 and *seed* set to 2103 * checksum, it can be used when pushing new data. 2104 * * With *from_size* > 0, *to_size* == 0 and *seed* set to 2105 * checksum, it can be used when removing data from a packet. 2106 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 2107 * can be used to compute a diff. Note that *from_size* and 2108 * *to_size* do not need to be equal. 2109 * 2110 * This helper can be used in combination with 2111 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 2112 * which one can feed in the difference computed with 2113 * **bpf_csum_diff**\ (). 2114 * Return 2115 * The checksum result, or a negative error code in case of 2116 * failure. 2117 * 2118 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2119 * Description 2120 * Retrieve tunnel options metadata for the packet associated to 2121 * *skb*, and store the raw tunnel option data to the buffer *opt* 2122 * of *size*. 2123 * 2124 * This helper can be used with encapsulation devices that can 2125 * operate in "collect metadata" mode (please refer to the related 2126 * note in the description of **bpf_skb_get_tunnel_key**\ () for 2127 * more details). A particular example where this can be used is 2128 * in combination with the Geneve encapsulation protocol, where it 2129 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 2130 * and retrieving arbitrary TLVs (Type-Length-Value headers) from 2131 * the eBPF program. This allows for full customization of these 2132 * headers. 2133 * Return 2134 * The size of the option data retrieved. 2135 * 2136 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 2137 * Description 2138 * Set tunnel options metadata for the packet associated to *skb* 2139 * to the option data contained in the raw buffer *opt* of *size*. 2140 * 2141 * See also the description of the **bpf_skb_get_tunnel_opt**\ () 2142 * helper for additional information. 2143 * Return 2144 * 0 on success, or a negative error in case of failure. 2145 * 2146 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 2147 * Description 2148 * Change the protocol of the *skb* to *proto*. Currently 2149 * supported are transition from IPv4 to IPv6, and from IPv6 to 2150 * IPv4. The helper takes care of the groundwork for the 2151 * transition, including resizing the socket buffer. The eBPF 2152 * program is expected to fill the new headers, if any, via 2153 * **skb_store_bytes**\ () and to recompute the checksums with 2154 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 2155 * (). The main case for this helper is to perform NAT64 2156 * operations out of an eBPF program. 2157 * 2158 * Internally, the GSO type is marked as dodgy so that headers are 2159 * checked and segments are recalculated by the GSO/GRO engine. 2160 * The size for GSO target is adapted as well. 2161 * 2162 * All values for *flags* are reserved for future usage, and must 2163 * be left at zero. 2164 * 2165 * A call to this helper is susceptible to change the underlying 2166 * packet buffer. Therefore, at load time, all checks on pointers 2167 * previously done by the verifier are invalidated and must be 2168 * performed again, if the helper is used in combination with 2169 * direct packet access. 2170 * Return 2171 * 0 on success, or a negative error in case of failure. 2172 * 2173 * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 2174 * Description 2175 * Change the packet type for the packet associated to *skb*. This 2176 * comes down to setting *skb*\ **->pkt_type** to *type*, except 2177 * the eBPF program does not have a write access to *skb*\ 2178 * **->pkt_type** beside this helper. Using a helper here allows 2179 * for graceful handling of errors. 2180 * 2181 * The major use case is to change incoming *skb*s to 2182 * **PACKET_HOST** in a programmatic way instead of having to 2183 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 2184 * example. 2185 * 2186 * Note that *type* only allows certain values. At this time, they 2187 * are: 2188 * 2189 * **PACKET_HOST** 2190 * Packet is for us. 2191 * **PACKET_BROADCAST** 2192 * Send packet to all. 2193 * **PACKET_MULTICAST** 2194 * Send packet to group. 2195 * **PACKET_OTHERHOST** 2196 * Send packet to someone else. 2197 * Return 2198 * 0 on success, or a negative error in case of failure. 2199 * 2200 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 2201 * Description 2202 * Check whether *skb* is a descendant of the cgroup2 held by 2203 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2204 * Return 2205 * The return value depends on the result of the test, and can be: 2206 * 2207 * * 0, if the *skb* failed the cgroup2 descendant test. 2208 * * 1, if the *skb* succeeded the cgroup2 descendant test. 2209 * * A negative error code, if an error occurred. 2210 * 2211 * u32 bpf_get_hash_recalc(struct sk_buff *skb) 2212 * Description 2213 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 2214 * not set, in particular if the hash was cleared due to mangling, 2215 * recompute this hash. Later accesses to the hash can be done 2216 * directly with *skb*\ **->hash**. 2217 * 2218 * Calling **bpf_set_hash_invalid**\ (), changing a packet 2219 * prototype with **bpf_skb_change_proto**\ (), or calling 2220 * **bpf_skb_store_bytes**\ () with the 2221 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 2222 * the hash and to trigger a new computation for the next call to 2223 * **bpf_get_hash_recalc**\ (). 2224 * Return 2225 * The 32-bit hash. 2226 * 2227 * u64 bpf_get_current_task(void) 2228 * Return 2229 * A pointer to the current task struct. 2230 * 2231 * long bpf_probe_write_user(void *dst, const void *src, u32 len) 2232 * Description 2233 * Attempt in a safe way to write *len* bytes from the buffer 2234 * *src* to *dst* in memory. It only works for threads that are in 2235 * user context, and *dst* must be a valid user space address. 2236 * 2237 * This helper should not be used to implement any kind of 2238 * security mechanism because of TOC-TOU attacks, but rather to 2239 * debug, divert, and manipulate execution of semi-cooperative 2240 * processes. 2241 * 2242 * Keep in mind that this feature is meant for experiments, and it 2243 * has a risk of crashing the system and running programs. 2244 * Therefore, when an eBPF program using this helper is attached, 2245 * a warning including PID and process name is printed to kernel 2246 * logs. 2247 * Return 2248 * 0 on success, or a negative error in case of failure. 2249 * 2250 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 2251 * Description 2252 * Check whether the probe is being run is the context of a given 2253 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 2254 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 2255 * Return 2256 * The return value depends on the result of the test, and can be: 2257 * 2258 * * 0, if current task belongs to the cgroup2. 2259 * * 1, if current task does not belong to the cgroup2. 2260 * * A negative error code, if an error occurred. 2261 * 2262 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 2263 * Description 2264 * Resize (trim or grow) the packet associated to *skb* to the 2265 * new *len*. The *flags* are reserved for future usage, and must 2266 * be left at zero. 2267 * 2268 * The basic idea is that the helper performs the needed work to 2269 * change the size of the packet, then the eBPF program rewrites 2270 * the rest via helpers like **bpf_skb_store_bytes**\ (), 2271 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 2272 * and others. This helper is a slow path utility intended for 2273 * replies with control messages. And because it is targeted for 2274 * slow path, the helper itself can afford to be slow: it 2275 * implicitly linearizes, unclones and drops offloads from the 2276 * *skb*. 2277 * 2278 * A call to this helper is susceptible to change the underlying 2279 * packet buffer. Therefore, at load time, all checks on pointers 2280 * previously done by the verifier are invalidated and must be 2281 * performed again, if the helper is used in combination with 2282 * direct packet access. 2283 * Return 2284 * 0 on success, or a negative error in case of failure. 2285 * 2286 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 2287 * Description 2288 * Pull in non-linear data in case the *skb* is non-linear and not 2289 * all of *len* are part of the linear section. Make *len* bytes 2290 * from *skb* readable and writable. If a zero value is passed for 2291 * *len*, then the whole length of the *skb* is pulled. 2292 * 2293 * This helper is only needed for reading and writing with direct 2294 * packet access. 2295 * 2296 * For direct packet access, testing that offsets to access 2297 * are within packet boundaries (test on *skb*\ **->data_end**) is 2298 * susceptible to fail if offsets are invalid, or if the requested 2299 * data is in non-linear parts of the *skb*. On failure the 2300 * program can just bail out, or in the case of a non-linear 2301 * buffer, use a helper to make the data available. The 2302 * **bpf_skb_load_bytes**\ () helper is a first solution to access 2303 * the data. Another one consists in using **bpf_skb_pull_data** 2304 * to pull in once the non-linear parts, then retesting and 2305 * eventually access the data. 2306 * 2307 * At the same time, this also makes sure the *skb* is uncloned, 2308 * which is a necessary condition for direct write. As this needs 2309 * to be an invariant for the write part only, the verifier 2310 * detects writes and adds a prologue that is calling 2311 * **bpf_skb_pull_data()** to effectively unclone the *skb* from 2312 * the very beginning in case it is indeed cloned. 2313 * 2314 * A call to this helper is susceptible to change the underlying 2315 * packet buffer. Therefore, at load time, all checks on pointers 2316 * previously done by the verifier are invalidated and must be 2317 * performed again, if the helper is used in combination with 2318 * direct packet access. 2319 * Return 2320 * 0 on success, or a negative error in case of failure. 2321 * 2322 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 2323 * Description 2324 * Add the checksum *csum* into *skb*\ **->csum** in case the 2325 * driver has supplied a checksum for the entire packet into that 2326 * field. Return an error otherwise. This helper is intended to be 2327 * used in combination with **bpf_csum_diff**\ (), in particular 2328 * when the checksum needs to be updated after data has been 2329 * written into the packet through direct packet access. 2330 * Return 2331 * The checksum on success, or a negative error code in case of 2332 * failure. 2333 * 2334 * void bpf_set_hash_invalid(struct sk_buff *skb) 2335 * Description 2336 * Invalidate the current *skb*\ **->hash**. It can be used after 2337 * mangling on headers through direct packet access, in order to 2338 * indicate that the hash is outdated and to trigger a 2339 * recalculation the next time the kernel tries to access this 2340 * hash or when the **bpf_get_hash_recalc**\ () helper is called. 2341 * 2342 * long bpf_get_numa_node_id(void) 2343 * Description 2344 * Return the id of the current NUMA node. The primary use case 2345 * for this helper is the selection of sockets for the local NUMA 2346 * node, when the program is attached to sockets using the 2347 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 2348 * but the helper is also available to other eBPF program types, 2349 * similarly to **bpf_get_smp_processor_id**\ (). 2350 * Return 2351 * The id of current NUMA node. 2352 * 2353 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 2354 * Description 2355 * Grows headroom of packet associated to *skb* and adjusts the 2356 * offset of the MAC header accordingly, adding *len* bytes of 2357 * space. It automatically extends and reallocates memory as 2358 * required. 2359 * 2360 * This helper can be used on a layer 3 *skb* to push a MAC header 2361 * for redirection into a layer 2 device. 2362 * 2363 * All values for *flags* are reserved for future usage, and must 2364 * be left at zero. 2365 * 2366 * A call to this helper is susceptible to change the underlying 2367 * packet buffer. Therefore, at load time, all checks on pointers 2368 * previously done by the verifier are invalidated and must be 2369 * performed again, if the helper is used in combination with 2370 * direct packet access. 2371 * Return 2372 * 0 on success, or a negative error in case of failure. 2373 * 2374 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 2375 * Description 2376 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 2377 * it is possible to use a negative value for *delta*. This helper 2378 * can be used to prepare the packet for pushing or popping 2379 * headers. 2380 * 2381 * A call to this helper is susceptible to change the underlying 2382 * packet buffer. Therefore, at load time, all checks on pointers 2383 * previously done by the verifier are invalidated and must be 2384 * performed again, if the helper is used in combination with 2385 * direct packet access. 2386 * Return 2387 * 0 on success, or a negative error in case of failure. 2388 * 2389 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 2390 * Description 2391 * Copy a NUL terminated string from an unsafe kernel address 2392 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 2393 * more details. 2394 * 2395 * Generally, use **bpf_probe_read_user_str**\ () or 2396 * **bpf_probe_read_kernel_str**\ () instead. 2397 * Return 2398 * On success, the strictly positive length of the string, 2399 * including the trailing NUL character. On error, a negative 2400 * value. 2401 * 2402 * u64 bpf_get_socket_cookie(struct sk_buff *skb) 2403 * Description 2404 * If the **struct sk_buff** pointed by *skb* has a known socket, 2405 * retrieve the cookie (generated by the kernel) of this socket. 2406 * If no cookie has been set yet, generate a new cookie. Once 2407 * generated, the socket cookie remains stable for the life of the 2408 * socket. This helper can be useful for monitoring per socket 2409 * networking traffic statistics as it provides a global socket 2410 * identifier that can be assumed unique. 2411 * Return 2412 * A 8-byte long unique number on success, or 0 if the socket 2413 * field is missing inside *skb*. 2414 * 2415 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 2416 * Description 2417 * Equivalent to bpf_get_socket_cookie() helper that accepts 2418 * *skb*, but gets socket from **struct bpf_sock_addr** context. 2419 * Return 2420 * A 8-byte long unique number. 2421 * 2422 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 2423 * Description 2424 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2425 * *skb*, but gets socket from **struct bpf_sock_ops** context. 2426 * Return 2427 * A 8-byte long unique number. 2428 * 2429 * u64 bpf_get_socket_cookie(struct sock *sk) 2430 * Description 2431 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 2432 * *sk*, but gets socket from a BTF **struct sock**. This helper 2433 * also works for sleepable programs. 2434 * Return 2435 * A 8-byte long unique number or 0 if *sk* is NULL. 2436 * 2437 * u32 bpf_get_socket_uid(struct sk_buff *skb) 2438 * Return 2439 * The owner UID of the socket associated to *skb*. If the socket 2440 * is **NULL**, or if it is not a full socket (i.e. if it is a 2441 * time-wait or a request socket instead), **overflowuid** value 2442 * is returned (note that **overflowuid** might also be the actual 2443 * UID value for the socket). 2444 * 2445 * long bpf_set_hash(struct sk_buff *skb, u32 hash) 2446 * Description 2447 * Set the full hash for *skb* (set the field *skb*\ **->hash**) 2448 * to value *hash*. 2449 * Return 2450 * 0 2451 * 2452 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2453 * Description 2454 * Emulate a call to **setsockopt()** on the socket associated to 2455 * *bpf_socket*, which must be a full socket. The *level* at 2456 * which the option resides and the name *optname* of the option 2457 * must be specified, see **setsockopt(2)** for more information. 2458 * The option value of length *optlen* is pointed by *optval*. 2459 * 2460 * *bpf_socket* should be one of the following: 2461 * 2462 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2463 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2464 * and **BPF_CGROUP_INET6_CONNECT**. 2465 * 2466 * This helper actually implements a subset of **setsockopt()**. 2467 * It supports the following *level*\ s: 2468 * 2469 * * **SOL_SOCKET**, which supports the following *optname*\ s: 2470 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 2471 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 2472 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 2473 * * **IPPROTO_TCP**, which supports the following *optname*\ s: 2474 * **TCP_CONGESTION**, **TCP_BPF_IW**, 2475 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 2476 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 2477 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**. 2478 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2479 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2480 * Return 2481 * 0 on success, or a negative error in case of failure. 2482 * 2483 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 2484 * Description 2485 * Grow or shrink the room for data in the packet associated to 2486 * *skb* by *len_diff*, and according to the selected *mode*. 2487 * 2488 * By default, the helper will reset any offloaded checksum 2489 * indicator of the skb to CHECKSUM_NONE. This can be avoided 2490 * by the following flag: 2491 * 2492 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 2493 * checksum data of the skb to CHECKSUM_NONE. 2494 * 2495 * There are two supported modes at this time: 2496 * 2497 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 2498 * (room space is added or removed below the layer 2 header). 2499 * 2500 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 2501 * (room space is added or removed below the layer 3 header). 2502 * 2503 * The following flags are supported at this time: 2504 * 2505 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 2506 * Adjusting mss in this way is not allowed for datagrams. 2507 * 2508 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 2509 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 2510 * Any new space is reserved to hold a tunnel header. 2511 * Configure skb offsets and other fields accordingly. 2512 * 2513 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 2514 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 2515 * Use with ENCAP_L3 flags to further specify the tunnel type. 2516 * 2517 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 2518 * Use with ENCAP_L3/L4 flags to further specify the tunnel 2519 * type; *len* is the length of the inner MAC header. 2520 * 2521 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**: 2522 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the 2523 * L2 type as Ethernet. 2524 * 2525 * A call to this helper is susceptible to change the underlying 2526 * packet buffer. Therefore, at load time, all checks on pointers 2527 * previously done by the verifier are invalidated and must be 2528 * performed again, if the helper is used in combination with 2529 * direct packet access. 2530 * Return 2531 * 0 on success, or a negative error in case of failure. 2532 * 2533 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 2534 * Description 2535 * Redirect the packet to the endpoint referenced by *map* at 2536 * index *key*. Depending on its type, this *map* can contain 2537 * references to net devices (for forwarding packets through other 2538 * ports), or to CPUs (for redirecting XDP frames to another CPU; 2539 * but this is only implemented for native XDP (with driver 2540 * support) as of this writing). 2541 * 2542 * The lower two bits of *flags* are used as the return code if 2543 * the map lookup fails. This is so that the return value can be 2544 * one of the XDP program return codes up to **XDP_TX**, as chosen 2545 * by the caller. Any higher bits in the *flags* argument must be 2546 * unset. 2547 * 2548 * See also **bpf_redirect**\ (), which only supports redirecting 2549 * to an ifindex, but doesn't require a map to do so. 2550 * Return 2551 * **XDP_REDIRECT** on success, or the value of the two lower bits 2552 * of the *flags* argument on error. 2553 * 2554 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 2555 * Description 2556 * Redirect the packet to the socket referenced by *map* (of type 2557 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2558 * egress interfaces can be used for redirection. The 2559 * **BPF_F_INGRESS** value in *flags* is used to make the 2560 * distinction (ingress path is selected if the flag is present, 2561 * egress path otherwise). This is the only flag supported for now. 2562 * Return 2563 * **SK_PASS** on success, or **SK_DROP** on error. 2564 * 2565 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 2566 * Description 2567 * Add an entry to, or update a *map* referencing sockets. The 2568 * *skops* is used as a new value for the entry associated to 2569 * *key*. *flags* is one of: 2570 * 2571 * **BPF_NOEXIST** 2572 * The entry for *key* must not exist in the map. 2573 * **BPF_EXIST** 2574 * The entry for *key* must already exist in the map. 2575 * **BPF_ANY** 2576 * No condition on the existence of the entry for *key*. 2577 * 2578 * If the *map* has eBPF programs (parser and verdict), those will 2579 * be inherited by the socket being added. If the socket is 2580 * already attached to eBPF programs, this results in an error. 2581 * Return 2582 * 0 on success, or a negative error in case of failure. 2583 * 2584 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 2585 * Description 2586 * Adjust the address pointed by *xdp_md*\ **->data_meta** by 2587 * *delta* (which can be positive or negative). Note that this 2588 * operation modifies the address stored in *xdp_md*\ **->data**, 2589 * so the latter must be loaded only after the helper has been 2590 * called. 2591 * 2592 * The use of *xdp_md*\ **->data_meta** is optional and programs 2593 * are not required to use it. The rationale is that when the 2594 * packet is processed with XDP (e.g. as DoS filter), it is 2595 * possible to push further meta data along with it before passing 2596 * to the stack, and to give the guarantee that an ingress eBPF 2597 * program attached as a TC classifier on the same device can pick 2598 * this up for further post-processing. Since TC works with socket 2599 * buffers, it remains possible to set from XDP the **mark** or 2600 * **priority** pointers, or other pointers for the socket buffer. 2601 * Having this scratch space generic and programmable allows for 2602 * more flexibility as the user is free to store whatever meta 2603 * data they need. 2604 * 2605 * A call to this helper is susceptible to change the underlying 2606 * packet buffer. Therefore, at load time, all checks on pointers 2607 * previously done by the verifier are invalidated and must be 2608 * performed again, if the helper is used in combination with 2609 * direct packet access. 2610 * Return 2611 * 0 on success, or a negative error in case of failure. 2612 * 2613 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 2614 * Description 2615 * Read the value of a perf event counter, and store it into *buf* 2616 * of size *buf_size*. This helper relies on a *map* of type 2617 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 2618 * counter is selected when *map* is updated with perf event file 2619 * descriptors. The *map* is an array whose size is the number of 2620 * available CPUs, and each cell contains a value relative to one 2621 * CPU. The value to retrieve is indicated by *flags*, that 2622 * contains the index of the CPU to look up, masked with 2623 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 2624 * **BPF_F_CURRENT_CPU** to indicate that the value for the 2625 * current CPU should be retrieved. 2626 * 2627 * This helper behaves in a way close to 2628 * **bpf_perf_event_read**\ () helper, save that instead of 2629 * just returning the value observed, it fills the *buf* 2630 * structure. This allows for additional data to be retrieved: in 2631 * particular, the enabled and running times (in *buf*\ 2632 * **->enabled** and *buf*\ **->running**, respectively) are 2633 * copied. In general, **bpf_perf_event_read_value**\ () is 2634 * recommended over **bpf_perf_event_read**\ (), which has some 2635 * ABI issues and provides fewer functionalities. 2636 * 2637 * These values are interesting, because hardware PMU (Performance 2638 * Monitoring Unit) counters are limited resources. When there are 2639 * more PMU based perf events opened than available counters, 2640 * kernel will multiplex these events so each event gets certain 2641 * percentage (but not all) of the PMU time. In case that 2642 * multiplexing happens, the number of samples or counter value 2643 * will not reflect the case compared to when no multiplexing 2644 * occurs. This makes comparison between different runs difficult. 2645 * Typically, the counter value should be normalized before 2646 * comparing to other experiments. The usual normalization is done 2647 * as follows. 2648 * 2649 * :: 2650 * 2651 * normalized_counter = counter * t_enabled / t_running 2652 * 2653 * Where t_enabled is the time enabled for event and t_running is 2654 * the time running for event since last normalization. The 2655 * enabled and running times are accumulated since the perf event 2656 * open. To achieve scaling factor between two invocations of an 2657 * eBPF program, users can use CPU id as the key (which is 2658 * typical for perf array usage model) to remember the previous 2659 * value and do the calculation inside the eBPF program. 2660 * Return 2661 * 0 on success, or a negative error in case of failure. 2662 * 2663 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 2664 * Description 2665 * For en eBPF program attached to a perf event, retrieve the 2666 * value of the event counter associated to *ctx* and store it in 2667 * the structure pointed by *buf* and of size *buf_size*. Enabled 2668 * and running times are also stored in the structure (see 2669 * description of helper **bpf_perf_event_read_value**\ () for 2670 * more details). 2671 * Return 2672 * 0 on success, or a negative error in case of failure. 2673 * 2674 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 2675 * Description 2676 * Emulate a call to **getsockopt()** on the socket associated to 2677 * *bpf_socket*, which must be a full socket. The *level* at 2678 * which the option resides and the name *optname* of the option 2679 * must be specified, see **getsockopt(2)** for more information. 2680 * The retrieved value is stored in the structure pointed by 2681 * *opval* and of length *optlen*. 2682 * 2683 * *bpf_socket* should be one of the following: 2684 * 2685 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 2686 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 2687 * and **BPF_CGROUP_INET6_CONNECT**. 2688 * 2689 * This helper actually implements a subset of **getsockopt()**. 2690 * It supports the following *level*\ s: 2691 * 2692 * * **IPPROTO_TCP**, which supports *optname* 2693 * **TCP_CONGESTION**. 2694 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 2695 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 2696 * Return 2697 * 0 on success, or a negative error in case of failure. 2698 * 2699 * long bpf_override_return(struct pt_regs *regs, u64 rc) 2700 * Description 2701 * Used for error injection, this helper uses kprobes to override 2702 * the return value of the probed function, and to set it to *rc*. 2703 * The first argument is the context *regs* on which the kprobe 2704 * works. 2705 * 2706 * This helper works by setting the PC (program counter) 2707 * to an override function which is run in place of the original 2708 * probed function. This means the probed function is not run at 2709 * all. The replacement function just returns with the required 2710 * value. 2711 * 2712 * This helper has security implications, and thus is subject to 2713 * restrictions. It is only available if the kernel was compiled 2714 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 2715 * option, and in this case it only works on functions tagged with 2716 * **ALLOW_ERROR_INJECTION** in the kernel code. 2717 * 2718 * Also, the helper is only available for the architectures having 2719 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 2720 * x86 architecture is the only one to support this feature. 2721 * Return 2722 * 0 2723 * 2724 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 2725 * Description 2726 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 2727 * for the full TCP socket associated to *bpf_sock_ops* to 2728 * *argval*. 2729 * 2730 * The primary use of this field is to determine if there should 2731 * be calls to eBPF programs of type 2732 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 2733 * code. A program of the same type can change its value, per 2734 * connection and as necessary, when the connection is 2735 * established. This field is directly accessible for reading, but 2736 * this helper must be used for updates in order to return an 2737 * error if an eBPF program tries to set a callback that is not 2738 * supported in the current kernel. 2739 * 2740 * *argval* is a flag array which can combine these flags: 2741 * 2742 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 2743 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 2744 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 2745 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 2746 * 2747 * Therefore, this function can be used to clear a callback flag by 2748 * setting the appropriate bit to zero. e.g. to disable the RTO 2749 * callback: 2750 * 2751 * **bpf_sock_ops_cb_flags_set(bpf_sock,** 2752 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 2753 * 2754 * Here are some examples of where one could call such eBPF 2755 * program: 2756 * 2757 * * When RTO fires. 2758 * * When a packet is retransmitted. 2759 * * When the connection terminates. 2760 * * When a packet is sent. 2761 * * When a packet is received. 2762 * Return 2763 * Code **-EINVAL** if the socket is not a full TCP socket; 2764 * otherwise, a positive number containing the bits that could not 2765 * be set is returned (which comes down to 0 if all bits were set 2766 * as required). 2767 * 2768 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 2769 * Description 2770 * This helper is used in programs implementing policies at the 2771 * socket level. If the message *msg* is allowed to pass (i.e. if 2772 * the verdict eBPF program returns **SK_PASS**), redirect it to 2773 * the socket referenced by *map* (of type 2774 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 2775 * egress interfaces can be used for redirection. The 2776 * **BPF_F_INGRESS** value in *flags* is used to make the 2777 * distinction (ingress path is selected if the flag is present, 2778 * egress path otherwise). This is the only flag supported for now. 2779 * Return 2780 * **SK_PASS** on success, or **SK_DROP** on error. 2781 * 2782 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 2783 * Description 2784 * For socket policies, apply the verdict of the eBPF program to 2785 * the next *bytes* (number of bytes) of message *msg*. 2786 * 2787 * For example, this helper can be used in the following cases: 2788 * 2789 * * A single **sendmsg**\ () or **sendfile**\ () system call 2790 * contains multiple logical messages that the eBPF program is 2791 * supposed to read and for which it should apply a verdict. 2792 * * An eBPF program only cares to read the first *bytes* of a 2793 * *msg*. If the message has a large payload, then setting up 2794 * and calling the eBPF program repeatedly for all bytes, even 2795 * though the verdict is already known, would create unnecessary 2796 * overhead. 2797 * 2798 * When called from within an eBPF program, the helper sets a 2799 * counter internal to the BPF infrastructure, that is used to 2800 * apply the last verdict to the next *bytes*. If *bytes* is 2801 * smaller than the current data being processed from a 2802 * **sendmsg**\ () or **sendfile**\ () system call, the first 2803 * *bytes* will be sent and the eBPF program will be re-run with 2804 * the pointer for start of data pointing to byte number *bytes* 2805 * **+ 1**. If *bytes* is larger than the current data being 2806 * processed, then the eBPF verdict will be applied to multiple 2807 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2808 * consumed. 2809 * 2810 * Note that if a socket closes with the internal counter holding 2811 * a non-zero value, this is not a problem because data is not 2812 * being buffered for *bytes* and is sent as it is received. 2813 * Return 2814 * 0 2815 * 2816 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2817 * Description 2818 * For socket policies, prevent the execution of the verdict eBPF 2819 * program for message *msg* until *bytes* (byte number) have been 2820 * accumulated. 2821 * 2822 * This can be used when one needs a specific number of bytes 2823 * before a verdict can be assigned, even if the data spans 2824 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2825 * case would be a user calling **sendmsg**\ () repeatedly with 2826 * 1-byte long message segments. Obviously, this is bad for 2827 * performance, but it is still valid. If the eBPF program needs 2828 * *bytes* bytes to validate a header, this helper can be used to 2829 * prevent the eBPF program to be called again until *bytes* have 2830 * been accumulated. 2831 * Return 2832 * 0 2833 * 2834 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2835 * Description 2836 * For socket policies, pull in non-linear data from user space 2837 * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2838 * **->data_end** to *start* and *end* bytes offsets into *msg*, 2839 * respectively. 2840 * 2841 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2842 * *msg* it can only parse data that the (**data**, **data_end**) 2843 * pointers have already consumed. For **sendmsg**\ () hooks this 2844 * is likely the first scatterlist element. But for calls relying 2845 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2846 * be the range (**0**, **0**) because the data is shared with 2847 * user space and by default the objective is to avoid allowing 2848 * user space to modify data while (or after) eBPF verdict is 2849 * being decided. This helper can be used to pull in data and to 2850 * set the start and end pointer to given values. Data will be 2851 * copied if necessary (i.e. if data was not linear and if start 2852 * and end pointers do not point to the same chunk). 2853 * 2854 * A call to this helper is susceptible to change the underlying 2855 * packet buffer. Therefore, at load time, all checks on pointers 2856 * previously done by the verifier are invalidated and must be 2857 * performed again, if the helper is used in combination with 2858 * direct packet access. 2859 * 2860 * All values for *flags* are reserved for future usage, and must 2861 * be left at zero. 2862 * Return 2863 * 0 on success, or a negative error in case of failure. 2864 * 2865 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 2866 * Description 2867 * Bind the socket associated to *ctx* to the address pointed by 2868 * *addr*, of length *addr_len*. This allows for making outgoing 2869 * connection from the desired IP address, which can be useful for 2870 * example when all processes inside a cgroup should use one 2871 * single IP address on a host that has multiple IP configured. 2872 * 2873 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 2874 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 2875 * **AF_INET6**). It's advised to pass zero port (**sin_port** 2876 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 2877 * behavior and lets the kernel efficiently pick up an unused 2878 * port as long as 4-tuple is unique. Passing non-zero port might 2879 * lead to degraded performance. 2880 * Return 2881 * 0 on success, or a negative error in case of failure. 2882 * 2883 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 2884 * Description 2885 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2886 * possible to both shrink and grow the packet tail. 2887 * Shrink done via *delta* being a negative integer. 2888 * 2889 * A call to this helper is susceptible to change the underlying 2890 * packet buffer. Therefore, at load time, all checks on pointers 2891 * previously done by the verifier are invalidated and must be 2892 * performed again, if the helper is used in combination with 2893 * direct packet access. 2894 * Return 2895 * 0 on success, or a negative error in case of failure. 2896 * 2897 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 2898 * Description 2899 * Retrieve the XFRM state (IP transform framework, see also 2900 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 2901 * 2902 * The retrieved value is stored in the **struct bpf_xfrm_state** 2903 * pointed by *xfrm_state* and of length *size*. 2904 * 2905 * All values for *flags* are reserved for future usage, and must 2906 * be left at zero. 2907 * 2908 * This helper is available only if the kernel was compiled with 2909 * **CONFIG_XFRM** configuration option. 2910 * Return 2911 * 0 on success, or a negative error in case of failure. 2912 * 2913 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2914 * Description 2915 * Return a user or a kernel stack in bpf program provided buffer. 2916 * To achieve this, the helper needs *ctx*, which is a pointer 2917 * to the context on which the tracing program is executed. 2918 * To store the stacktrace, the bpf program provides *buf* with 2919 * a nonnegative *size*. 2920 * 2921 * The last argument, *flags*, holds the number of stack frames to 2922 * skip (from 0 to 255), masked with 2923 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2924 * the following flags: 2925 * 2926 * **BPF_F_USER_STACK** 2927 * Collect a user space stack instead of a kernel stack. 2928 * **BPF_F_USER_BUILD_ID** 2929 * Collect buildid+offset instead of ips for user stack, 2930 * only valid if **BPF_F_USER_STACK** is also specified. 2931 * 2932 * **bpf_get_stack**\ () can collect up to 2933 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2934 * to sufficient large buffer size. Note that 2935 * this limit can be controlled with the **sysctl** program, and 2936 * that it should be manually increased in order to profile long 2937 * user stacks (such as stacks for Java programs). To do so, use: 2938 * 2939 * :: 2940 * 2941 * # sysctl kernel.perf_event_max_stack=<new value> 2942 * Return 2943 * A non-negative value equal to or less than *size* on success, 2944 * or a negative error in case of failure. 2945 * 2946 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 2947 * Description 2948 * This helper is similar to **bpf_skb_load_bytes**\ () in that 2949 * it provides an easy way to load *len* bytes from *offset* 2950 * from the packet associated to *skb*, into the buffer pointed 2951 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 2952 * a fifth argument *start_header* exists in order to select a 2953 * base offset to start from. *start_header* can be one of: 2954 * 2955 * **BPF_HDR_START_MAC** 2956 * Base offset to load data from is *skb*'s mac header. 2957 * **BPF_HDR_START_NET** 2958 * Base offset to load data from is *skb*'s network header. 2959 * 2960 * In general, "direct packet access" is the preferred method to 2961 * access packet data, however, this helper is in particular useful 2962 * in socket filters where *skb*\ **->data** does not always point 2963 * to the start of the mac header and where "direct packet access" 2964 * is not available. 2965 * Return 2966 * 0 on success, or a negative error in case of failure. 2967 * 2968 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 2969 * Description 2970 * Do FIB lookup in kernel tables using parameters in *params*. 2971 * If lookup is successful and result shows packet is to be 2972 * forwarded, the neighbor tables are searched for the nexthop. 2973 * If successful (ie., FIB lookup shows forwarding and nexthop 2974 * is resolved), the nexthop address is returned in ipv4_dst 2975 * or ipv6_dst based on family, smac is set to mac address of 2976 * egress device, dmac is set to nexthop mac address, rt_metric 2977 * is set to metric from route (IPv4/IPv6 only), and ifindex 2978 * is set to the device index of the nexthop from the FIB lookup. 2979 * 2980 * *plen* argument is the size of the passed in struct. 2981 * *flags* argument can be a combination of one or more of the 2982 * following values: 2983 * 2984 * **BPF_FIB_LOOKUP_DIRECT** 2985 * Do a direct table lookup vs full lookup using FIB 2986 * rules. 2987 * **BPF_FIB_LOOKUP_OUTPUT** 2988 * Perform lookup from an egress perspective (default is 2989 * ingress). 2990 * 2991 * *ctx* is either **struct xdp_md** for XDP programs or 2992 * **struct sk_buff** tc cls_act programs. 2993 * Return 2994 * * < 0 if any input argument is invalid 2995 * * 0 on success (packet is forwarded, nexthop neighbor exists) 2996 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 2997 * packet is not forwarded or needs assist from full stack 2998 * 2999 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU 3000 * was exceeded and output params->mtu_result contains the MTU. 3001 * 3002 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 3003 * Description 3004 * Add an entry to, or update a sockhash *map* referencing sockets. 3005 * The *skops* is used as a new value for the entry associated to 3006 * *key*. *flags* is one of: 3007 * 3008 * **BPF_NOEXIST** 3009 * The entry for *key* must not exist in the map. 3010 * **BPF_EXIST** 3011 * The entry for *key* must already exist in the map. 3012 * **BPF_ANY** 3013 * No condition on the existence of the entry for *key*. 3014 * 3015 * If the *map* has eBPF programs (parser and verdict), those will 3016 * be inherited by the socket being added. If the socket is 3017 * already attached to eBPF programs, this results in an error. 3018 * Return 3019 * 0 on success, or a negative error in case of failure. 3020 * 3021 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 3022 * Description 3023 * This helper is used in programs implementing policies at the 3024 * socket level. If the message *msg* is allowed to pass (i.e. if 3025 * the verdict eBPF program returns **SK_PASS**), redirect it to 3026 * the socket referenced by *map* (of type 3027 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3028 * egress interfaces can be used for redirection. The 3029 * **BPF_F_INGRESS** value in *flags* is used to make the 3030 * distinction (ingress path is selected if the flag is present, 3031 * egress path otherwise). This is the only flag supported for now. 3032 * Return 3033 * **SK_PASS** on success, or **SK_DROP** on error. 3034 * 3035 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 3036 * Description 3037 * This helper is used in programs implementing policies at the 3038 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 3039 * if the verdict eBPF program returns **SK_PASS**), redirect it 3040 * to the socket referenced by *map* (of type 3041 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 3042 * egress interfaces can be used for redirection. The 3043 * **BPF_F_INGRESS** value in *flags* is used to make the 3044 * distinction (ingress path is selected if the flag is present, 3045 * egress otherwise). This is the only flag supported for now. 3046 * Return 3047 * **SK_PASS** on success, or **SK_DROP** on error. 3048 * 3049 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 3050 * Description 3051 * Encapsulate the packet associated to *skb* within a Layer 3 3052 * protocol header. This header is provided in the buffer at 3053 * address *hdr*, with *len* its size in bytes. *type* indicates 3054 * the protocol of the header and can be one of: 3055 * 3056 * **BPF_LWT_ENCAP_SEG6** 3057 * IPv6 encapsulation with Segment Routing Header 3058 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 3059 * the IPv6 header is computed by the kernel. 3060 * **BPF_LWT_ENCAP_SEG6_INLINE** 3061 * Only works if *skb* contains an IPv6 packet. Insert a 3062 * Segment Routing Header (**struct ipv6_sr_hdr**) inside 3063 * the IPv6 header. 3064 * **BPF_LWT_ENCAP_IP** 3065 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 3066 * must be IPv4 or IPv6, followed by zero or more 3067 * additional headers, up to **LWT_BPF_MAX_HEADROOM** 3068 * total bytes in all prepended headers. Please note that 3069 * if **skb_is_gso**\ (*skb*) is true, no more than two 3070 * headers can be prepended, and the inner header, if 3071 * present, should be either GRE or UDP/GUE. 3072 * 3073 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 3074 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 3075 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 3076 * **BPF_PROG_TYPE_LWT_XMIT**. 3077 * 3078 * A call to this helper is susceptible to change the underlying 3079 * packet buffer. Therefore, at load time, all checks on pointers 3080 * previously done by the verifier are invalidated and must be 3081 * performed again, if the helper is used in combination with 3082 * direct packet access. 3083 * Return 3084 * 0 on success, or a negative error in case of failure. 3085 * 3086 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 3087 * Description 3088 * Store *len* bytes from address *from* into the packet 3089 * associated to *skb*, at *offset*. Only the flags, tag and TLVs 3090 * inside the outermost IPv6 Segment Routing Header can be 3091 * modified through this helper. 3092 * 3093 * A call to this helper is susceptible to change the underlying 3094 * packet buffer. Therefore, at load time, all checks on pointers 3095 * previously done by the verifier are invalidated and must be 3096 * performed again, if the helper is used in combination with 3097 * direct packet access. 3098 * Return 3099 * 0 on success, or a negative error in case of failure. 3100 * 3101 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 3102 * Description 3103 * Adjust the size allocated to TLVs in the outermost IPv6 3104 * Segment Routing Header contained in the packet associated to 3105 * *skb*, at position *offset* by *delta* bytes. Only offsets 3106 * after the segments are accepted. *delta* can be as well 3107 * positive (growing) as negative (shrinking). 3108 * 3109 * A call to this helper is susceptible to change the underlying 3110 * packet buffer. Therefore, at load time, all checks on pointers 3111 * previously done by the verifier are invalidated and must be 3112 * performed again, if the helper is used in combination with 3113 * direct packet access. 3114 * Return 3115 * 0 on success, or a negative error in case of failure. 3116 * 3117 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 3118 * Description 3119 * Apply an IPv6 Segment Routing action of type *action* to the 3120 * packet associated to *skb*. Each action takes a parameter 3121 * contained at address *param*, and of length *param_len* bytes. 3122 * *action* can be one of: 3123 * 3124 * **SEG6_LOCAL_ACTION_END_X** 3125 * End.X action: Endpoint with Layer-3 cross-connect. 3126 * Type of *param*: **struct in6_addr**. 3127 * **SEG6_LOCAL_ACTION_END_T** 3128 * End.T action: Endpoint with specific IPv6 table lookup. 3129 * Type of *param*: **int**. 3130 * **SEG6_LOCAL_ACTION_END_B6** 3131 * End.B6 action: Endpoint bound to an SRv6 policy. 3132 * Type of *param*: **struct ipv6_sr_hdr**. 3133 * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 3134 * End.B6.Encap action: Endpoint bound to an SRv6 3135 * encapsulation policy. 3136 * Type of *param*: **struct ipv6_sr_hdr**. 3137 * 3138 * A call to this helper is susceptible to change the underlying 3139 * packet buffer. Therefore, at load time, all checks on pointers 3140 * previously done by the verifier are invalidated and must be 3141 * performed again, if the helper is used in combination with 3142 * direct packet access. 3143 * Return 3144 * 0 on success, or a negative error in case of failure. 3145 * 3146 * long bpf_rc_repeat(void *ctx) 3147 * Description 3148 * This helper is used in programs implementing IR decoding, to 3149 * report a successfully decoded repeat key message. This delays 3150 * the generation of a key up event for previously generated 3151 * key down event. 3152 * 3153 * Some IR protocols like NEC have a special IR message for 3154 * repeating last button, for when a button is held down. 3155 * 3156 * The *ctx* should point to the lirc sample as passed into 3157 * the program. 3158 * 3159 * This helper is only available is the kernel was compiled with 3160 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3161 * "**y**". 3162 * Return 3163 * 0 3164 * 3165 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 3166 * Description 3167 * This helper is used in programs implementing IR decoding, to 3168 * report a successfully decoded key press with *scancode*, 3169 * *toggle* value in the given *protocol*. The scancode will be 3170 * translated to a keycode using the rc keymap, and reported as 3171 * an input key down event. After a period a key up event is 3172 * generated. This period can be extended by calling either 3173 * **bpf_rc_keydown**\ () again with the same values, or calling 3174 * **bpf_rc_repeat**\ (). 3175 * 3176 * Some protocols include a toggle bit, in case the button was 3177 * released and pressed again between consecutive scancodes. 3178 * 3179 * The *ctx* should point to the lirc sample as passed into 3180 * the program. 3181 * 3182 * The *protocol* is the decoded protocol number (see 3183 * **enum rc_proto** for some predefined values). 3184 * 3185 * This helper is only available is the kernel was compiled with 3186 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3187 * "**y**". 3188 * Return 3189 * 0 3190 * 3191 * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 3192 * Description 3193 * Return the cgroup v2 id of the socket associated with the *skb*. 3194 * This is roughly similar to the **bpf_get_cgroup_classid**\ () 3195 * helper for cgroup v1 by providing a tag resp. identifier that 3196 * can be matched on or used for map lookups e.g. to implement 3197 * policy. The cgroup v2 id of a given path in the hierarchy is 3198 * exposed in user space through the f_handle API in order to get 3199 * to the same 64-bit id. 3200 * 3201 * This helper can be used on TC egress path, but not on ingress, 3202 * and is available only if the kernel was compiled with the 3203 * **CONFIG_SOCK_CGROUP_DATA** configuration option. 3204 * Return 3205 * The id is returned or 0 in case the id could not be retrieved. 3206 * 3207 * u64 bpf_get_current_cgroup_id(void) 3208 * Return 3209 * A 64-bit integer containing the current cgroup id based 3210 * on the cgroup within which the current task is running. 3211 * 3212 * void *bpf_get_local_storage(void *map, u64 flags) 3213 * Description 3214 * Get the pointer to the local storage area. 3215 * The type and the size of the local storage is defined 3216 * by the *map* argument. 3217 * The *flags* meaning is specific for each map type, 3218 * and has to be 0 for cgroup local storage. 3219 * 3220 * Depending on the BPF program type, a local storage area 3221 * can be shared between multiple instances of the BPF program, 3222 * running simultaneously. 3223 * 3224 * A user should care about the synchronization by himself. 3225 * For example, by using the **BPF_ATOMIC** instructions to alter 3226 * the shared data. 3227 * Return 3228 * A pointer to the local storage area. 3229 * 3230 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 3231 * Description 3232 * Select a **SO_REUSEPORT** socket from a 3233 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 3234 * It checks the selected socket is matching the incoming 3235 * request in the socket buffer. 3236 * Return 3237 * 0 on success, or a negative error in case of failure. 3238 * 3239 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 3240 * Description 3241 * Return id of cgroup v2 that is ancestor of cgroup associated 3242 * with the *skb* at the *ancestor_level*. The root cgroup is at 3243 * *ancestor_level* zero and each step down the hierarchy 3244 * increments the level. If *ancestor_level* == level of cgroup 3245 * associated with *skb*, then return value will be same as that 3246 * of **bpf_skb_cgroup_id**\ (). 3247 * 3248 * The helper is useful to implement policies based on cgroups 3249 * that are upper in hierarchy than immediate cgroup associated 3250 * with *skb*. 3251 * 3252 * The format of returned id and helper limitations are same as in 3253 * **bpf_skb_cgroup_id**\ (). 3254 * Return 3255 * The id is returned or 0 in case the id could not be retrieved. 3256 * 3257 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3258 * Description 3259 * Look for TCP socket matching *tuple*, optionally in a child 3260 * network namespace *netns*. The return value must be checked, 3261 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3262 * 3263 * The *ctx* should point to the context of the program, such as 3264 * the skb or socket (depending on the hook in use). This is used 3265 * to determine the base network namespace for the lookup. 3266 * 3267 * *tuple_size* must be one of: 3268 * 3269 * **sizeof**\ (*tuple*\ **->ipv4**) 3270 * Look for an IPv4 socket. 3271 * **sizeof**\ (*tuple*\ **->ipv6**) 3272 * Look for an IPv6 socket. 3273 * 3274 * If the *netns* is a negative signed 32-bit integer, then the 3275 * socket lookup table in the netns associated with the *ctx* 3276 * will be used. For the TC hooks, this is the netns of the device 3277 * in the skb. For socket hooks, this is the netns of the socket. 3278 * If *netns* is any other signed 32-bit value greater than or 3279 * equal to zero then it specifies the ID of the netns relative to 3280 * the netns associated with the *ctx*. *netns* values beyond the 3281 * range of 32-bit integers are reserved for future use. 3282 * 3283 * All values for *flags* are reserved for future usage, and must 3284 * be left at zero. 3285 * 3286 * This helper is available only if the kernel was compiled with 3287 * **CONFIG_NET** configuration option. 3288 * Return 3289 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3290 * For sockets with reuseport option, the **struct bpf_sock** 3291 * result is from *reuse*\ **->socks**\ [] using the hash of the 3292 * tuple. 3293 * 3294 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3295 * Description 3296 * Look for UDP socket matching *tuple*, optionally in a child 3297 * network namespace *netns*. The return value must be checked, 3298 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3299 * 3300 * The *ctx* should point to the context of the program, such as 3301 * the skb or socket (depending on the hook in use). This is used 3302 * to determine the base network namespace for the lookup. 3303 * 3304 * *tuple_size* must be one of: 3305 * 3306 * **sizeof**\ (*tuple*\ **->ipv4**) 3307 * Look for an IPv4 socket. 3308 * **sizeof**\ (*tuple*\ **->ipv6**) 3309 * Look for an IPv6 socket. 3310 * 3311 * If the *netns* is a negative signed 32-bit integer, then the 3312 * socket lookup table in the netns associated with the *ctx* 3313 * will be used. For the TC hooks, this is the netns of the device 3314 * in the skb. For socket hooks, this is the netns of the socket. 3315 * If *netns* is any other signed 32-bit value greater than or 3316 * equal to zero then it specifies the ID of the netns relative to 3317 * the netns associated with the *ctx*. *netns* values beyond the 3318 * range of 32-bit integers are reserved for future use. 3319 * 3320 * All values for *flags* are reserved for future usage, and must 3321 * be left at zero. 3322 * 3323 * This helper is available only if the kernel was compiled with 3324 * **CONFIG_NET** configuration option. 3325 * Return 3326 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3327 * For sockets with reuseport option, the **struct bpf_sock** 3328 * result is from *reuse*\ **->socks**\ [] using the hash of the 3329 * tuple. 3330 * 3331 * long bpf_sk_release(void *sock) 3332 * Description 3333 * Release the reference held by *sock*. *sock* must be a 3334 * non-**NULL** pointer that was returned from 3335 * **bpf_sk_lookup_xxx**\ (). 3336 * Return 3337 * 0 on success, or a negative error in case of failure. 3338 * 3339 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 3340 * Description 3341 * Push an element *value* in *map*. *flags* is one of: 3342 * 3343 * **BPF_EXIST** 3344 * If the queue/stack is full, the oldest element is 3345 * removed to make room for this. 3346 * Return 3347 * 0 on success, or a negative error in case of failure. 3348 * 3349 * long bpf_map_pop_elem(struct bpf_map *map, void *value) 3350 * Description 3351 * Pop an element from *map*. 3352 * Return 3353 * 0 on success, or a negative error in case of failure. 3354 * 3355 * long bpf_map_peek_elem(struct bpf_map *map, void *value) 3356 * Description 3357 * Get an element from *map* without removing it. 3358 * Return 3359 * 0 on success, or a negative error in case of failure. 3360 * 3361 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3362 * Description 3363 * For socket policies, insert *len* bytes into *msg* at offset 3364 * *start*. 3365 * 3366 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 3367 * *msg* it may want to insert metadata or options into the *msg*. 3368 * This can later be read and used by any of the lower layer BPF 3369 * hooks. 3370 * 3371 * This helper may fail if under memory pressure (a malloc 3372 * fails) in these cases BPF programs will get an appropriate 3373 * error and BPF programs will need to handle them. 3374 * Return 3375 * 0 on success, or a negative error in case of failure. 3376 * 3377 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 3378 * Description 3379 * Will remove *len* bytes from a *msg* starting at byte *start*. 3380 * This may result in **ENOMEM** errors under certain situations if 3381 * an allocation and copy are required due to a full ring buffer. 3382 * However, the helper will try to avoid doing the allocation 3383 * if possible. Other errors can occur if input parameters are 3384 * invalid either due to *start* byte not being valid part of *msg* 3385 * payload and/or *pop* value being to large. 3386 * Return 3387 * 0 on success, or a negative error in case of failure. 3388 * 3389 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 3390 * Description 3391 * This helper is used in programs implementing IR decoding, to 3392 * report a successfully decoded pointer movement. 3393 * 3394 * The *ctx* should point to the lirc sample as passed into 3395 * the program. 3396 * 3397 * This helper is only available is the kernel was compiled with 3398 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 3399 * "**y**". 3400 * Return 3401 * 0 3402 * 3403 * long bpf_spin_lock(struct bpf_spin_lock *lock) 3404 * Description 3405 * Acquire a spinlock represented by the pointer *lock*, which is 3406 * stored as part of a value of a map. Taking the lock allows to 3407 * safely update the rest of the fields in that value. The 3408 * spinlock can (and must) later be released with a call to 3409 * **bpf_spin_unlock**\ (\ *lock*\ ). 3410 * 3411 * Spinlocks in BPF programs come with a number of restrictions 3412 * and constraints: 3413 * 3414 * * **bpf_spin_lock** objects are only allowed inside maps of 3415 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 3416 * list could be extended in the future). 3417 * * BTF description of the map is mandatory. 3418 * * The BPF program can take ONE lock at a time, since taking two 3419 * or more could cause dead locks. 3420 * * Only one **struct bpf_spin_lock** is allowed per map element. 3421 * * When the lock is taken, calls (either BPF to BPF or helpers) 3422 * are not allowed. 3423 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 3424 * allowed inside a spinlock-ed region. 3425 * * The BPF program MUST call **bpf_spin_unlock**\ () to release 3426 * the lock, on all execution paths, before it returns. 3427 * * The BPF program can access **struct bpf_spin_lock** only via 3428 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 3429 * helpers. Loading or storing data into the **struct 3430 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 3431 * * To use the **bpf_spin_lock**\ () helper, the BTF description 3432 * of the map value must be a struct and have **struct 3433 * bpf_spin_lock** *anyname*\ **;** field at the top level. 3434 * Nested lock inside another struct is not allowed. 3435 * * The **struct bpf_spin_lock** *lock* field in a map value must 3436 * be aligned on a multiple of 4 bytes in that value. 3437 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 3438 * the **bpf_spin_lock** field to user space. 3439 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 3440 * a BPF program, do not update the **bpf_spin_lock** field. 3441 * * **bpf_spin_lock** cannot be on the stack or inside a 3442 * networking packet (it can only be inside of a map values). 3443 * * **bpf_spin_lock** is available to root only. 3444 * * Tracing programs and socket filter programs cannot use 3445 * **bpf_spin_lock**\ () due to insufficient preemption checks 3446 * (but this may change in the future). 3447 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 3448 * Return 3449 * 0 3450 * 3451 * long bpf_spin_unlock(struct bpf_spin_lock *lock) 3452 * Description 3453 * Release the *lock* previously locked by a call to 3454 * **bpf_spin_lock**\ (\ *lock*\ ). 3455 * Return 3456 * 0 3457 * 3458 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 3459 * Description 3460 * This helper gets a **struct bpf_sock** pointer such 3461 * that all the fields in this **bpf_sock** can be accessed. 3462 * Return 3463 * A **struct bpf_sock** pointer on success, or **NULL** in 3464 * case of failure. 3465 * 3466 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 3467 * Description 3468 * This helper gets a **struct bpf_tcp_sock** pointer from a 3469 * **struct bpf_sock** pointer. 3470 * Return 3471 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 3472 * case of failure. 3473 * 3474 * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 3475 * Description 3476 * Set ECN (Explicit Congestion Notification) field of IP header 3477 * to **CE** (Congestion Encountered) if current value is **ECT** 3478 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 3479 * and IPv4. 3480 * Return 3481 * 1 if the **CE** flag is set (either by the current helper call 3482 * or because it was already present), 0 if it is not set. 3483 * 3484 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 3485 * Description 3486 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 3487 * **bpf_sk_release**\ () is unnecessary and not allowed. 3488 * Return 3489 * A **struct bpf_sock** pointer on success, or **NULL** in 3490 * case of failure. 3491 * 3492 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 3493 * Description 3494 * Look for TCP socket matching *tuple*, optionally in a child 3495 * network namespace *netns*. The return value must be checked, 3496 * and if non-**NULL**, released via **bpf_sk_release**\ (). 3497 * 3498 * This function is identical to **bpf_sk_lookup_tcp**\ (), except 3499 * that it also returns timewait or request sockets. Use 3500 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 3501 * full structure. 3502 * 3503 * This helper is available only if the kernel was compiled with 3504 * **CONFIG_NET** configuration option. 3505 * Return 3506 * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 3507 * For sockets with reuseport option, the **struct bpf_sock** 3508 * result is from *reuse*\ **->socks**\ [] using the hash of the 3509 * tuple. 3510 * 3511 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3512 * Description 3513 * Check whether *iph* and *th* contain a valid SYN cookie ACK for 3514 * the listening socket in *sk*. 3515 * 3516 * *iph* points to the start of the IPv4 or IPv6 header, while 3517 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3518 * **sizeof**\ (**struct ip6hdr**). 3519 * 3520 * *th* points to the start of the TCP header, while *th_len* 3521 * contains **sizeof**\ (**struct tcphdr**). 3522 * Return 3523 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 3524 * error otherwise. 3525 * 3526 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 3527 * Description 3528 * Get name of sysctl in /proc/sys/ and copy it into provided by 3529 * program buffer *buf* of size *buf_len*. 3530 * 3531 * The buffer is always NUL terminated, unless it's zero-sized. 3532 * 3533 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 3534 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 3535 * only (e.g. "tcp_mem"). 3536 * Return 3537 * Number of character copied (not including the trailing NUL). 3538 * 3539 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3540 * truncated name in this case). 3541 * 3542 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3543 * Description 3544 * Get current value of sysctl as it is presented in /proc/sys 3545 * (incl. newline, etc), and copy it as a string into provided 3546 * by program buffer *buf* of size *buf_len*. 3547 * 3548 * The whole value is copied, no matter what file position user 3549 * space issued e.g. sys_read at. 3550 * 3551 * The buffer is always NUL terminated, unless it's zero-sized. 3552 * Return 3553 * Number of character copied (not including the trailing NUL). 3554 * 3555 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3556 * truncated name in this case). 3557 * 3558 * **-EINVAL** if current value was unavailable, e.g. because 3559 * sysctl is uninitialized and read returns -EIO for it. 3560 * 3561 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 3562 * Description 3563 * Get new value being written by user space to sysctl (before 3564 * the actual write happens) and copy it as a string into 3565 * provided by program buffer *buf* of size *buf_len*. 3566 * 3567 * User space may write new value at file position > 0. 3568 * 3569 * The buffer is always NUL terminated, unless it's zero-sized. 3570 * Return 3571 * Number of character copied (not including the trailing NUL). 3572 * 3573 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 3574 * truncated name in this case). 3575 * 3576 * **-EINVAL** if sysctl is being read. 3577 * 3578 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 3579 * Description 3580 * Override new value being written by user space to sysctl with 3581 * value provided by program in buffer *buf* of size *buf_len*. 3582 * 3583 * *buf* should contain a string in same form as provided by user 3584 * space on sysctl write. 3585 * 3586 * User space may write new value at file position > 0. To override 3587 * the whole sysctl value file position should be set to zero. 3588 * Return 3589 * 0 on success. 3590 * 3591 * **-E2BIG** if the *buf_len* is too big. 3592 * 3593 * **-EINVAL** if sysctl is being read. 3594 * 3595 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 3596 * Description 3597 * Convert the initial part of the string from buffer *buf* of 3598 * size *buf_len* to a long integer according to the given base 3599 * and save the result in *res*. 3600 * 3601 * The string may begin with an arbitrary amount of white space 3602 * (as determined by **isspace**\ (3)) followed by a single 3603 * optional '**-**' sign. 3604 * 3605 * Five least significant bits of *flags* encode base, other bits 3606 * are currently unused. 3607 * 3608 * Base must be either 8, 10, 16 or 0 to detect it automatically 3609 * similar to user space **strtol**\ (3). 3610 * Return 3611 * Number of characters consumed on success. Must be positive but 3612 * no more than *buf_len*. 3613 * 3614 * **-EINVAL** if no valid digits were found or unsupported base 3615 * was provided. 3616 * 3617 * **-ERANGE** if resulting value was out of range. 3618 * 3619 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 3620 * Description 3621 * Convert the initial part of the string from buffer *buf* of 3622 * size *buf_len* to an unsigned long integer according to the 3623 * given base and save the result in *res*. 3624 * 3625 * The string may begin with an arbitrary amount of white space 3626 * (as determined by **isspace**\ (3)). 3627 * 3628 * Five least significant bits of *flags* encode base, other bits 3629 * are currently unused. 3630 * 3631 * Base must be either 8, 10, 16 or 0 to detect it automatically 3632 * similar to user space **strtoul**\ (3). 3633 * Return 3634 * Number of characters consumed on success. Must be positive but 3635 * no more than *buf_len*. 3636 * 3637 * **-EINVAL** if no valid digits were found or unsupported base 3638 * was provided. 3639 * 3640 * **-ERANGE** if resulting value was out of range. 3641 * 3642 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 3643 * Description 3644 * Get a bpf-local-storage from a *sk*. 3645 * 3646 * Logically, it could be thought of getting the value from 3647 * a *map* with *sk* as the **key**. From this 3648 * perspective, the usage is not much different from 3649 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 3650 * helper enforces the key must be a full socket and the map must 3651 * be a **BPF_MAP_TYPE_SK_STORAGE** also. 3652 * 3653 * Underneath, the value is stored locally at *sk* instead of 3654 * the *map*. The *map* is used as the bpf-local-storage 3655 * "type". The bpf-local-storage "type" (i.e. the *map*) is 3656 * searched against all bpf-local-storages residing at *sk*. 3657 * 3658 * *sk* is a kernel **struct sock** pointer for LSM program. 3659 * *sk* is a **struct bpf_sock** pointer for other program types. 3660 * 3661 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 3662 * used such that a new bpf-local-storage will be 3663 * created if one does not exist. *value* can be used 3664 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 3665 * the initial value of a bpf-local-storage. If *value* is 3666 * **NULL**, the new bpf-local-storage will be zero initialized. 3667 * Return 3668 * A bpf-local-storage pointer is returned on success. 3669 * 3670 * **NULL** if not found or there was an error in adding 3671 * a new bpf-local-storage. 3672 * 3673 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 3674 * Description 3675 * Delete a bpf-local-storage from a *sk*. 3676 * Return 3677 * 0 on success. 3678 * 3679 * **-ENOENT** if the bpf-local-storage cannot be found. 3680 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 3681 * 3682 * long bpf_send_signal(u32 sig) 3683 * Description 3684 * Send signal *sig* to the process of the current task. 3685 * The signal may be delivered to any of this process's threads. 3686 * Return 3687 * 0 on success or successfully queued. 3688 * 3689 * **-EBUSY** if work queue under nmi is full. 3690 * 3691 * **-EINVAL** if *sig* is invalid. 3692 * 3693 * **-EPERM** if no permission to send the *sig*. 3694 * 3695 * **-EAGAIN** if bpf program can try again. 3696 * 3697 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 3698 * Description 3699 * Try to issue a SYN cookie for the packet with corresponding 3700 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 3701 * 3702 * *iph* points to the start of the IPv4 or IPv6 header, while 3703 * *iph_len* contains **sizeof**\ (**struct iphdr**) or 3704 * **sizeof**\ (**struct ip6hdr**). 3705 * 3706 * *th* points to the start of the TCP header, while *th_len* 3707 * contains the length of the TCP header. 3708 * Return 3709 * On success, lower 32 bits hold the generated SYN cookie in 3710 * followed by 16 bits which hold the MSS value for that cookie, 3711 * and the top 16 bits are unused. 3712 * 3713 * On failure, the returned value is one of the following: 3714 * 3715 * **-EINVAL** SYN cookie cannot be issued due to error 3716 * 3717 * **-ENOENT** SYN cookie should not be issued (no SYN flood) 3718 * 3719 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 3720 * 3721 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 3722 * 3723 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3724 * Description 3725 * Write raw *data* blob into a special BPF perf event held by 3726 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3727 * event must have the following attributes: **PERF_SAMPLE_RAW** 3728 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3729 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3730 * 3731 * The *flags* are used to indicate the index in *map* for which 3732 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3733 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3734 * to indicate that the index of the current CPU core should be 3735 * used. 3736 * 3737 * The value to write, of *size*, is passed through eBPF stack and 3738 * pointed by *data*. 3739 * 3740 * *ctx* is a pointer to in-kernel struct sk_buff. 3741 * 3742 * This helper is similar to **bpf_perf_event_output**\ () but 3743 * restricted to raw_tracepoint bpf programs. 3744 * Return 3745 * 0 on success, or a negative error in case of failure. 3746 * 3747 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 3748 * Description 3749 * Safely attempt to read *size* bytes from user space address 3750 * *unsafe_ptr* and store the data in *dst*. 3751 * Return 3752 * 0 on success, or a negative error in case of failure. 3753 * 3754 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 3755 * Description 3756 * Safely attempt to read *size* bytes from kernel space address 3757 * *unsafe_ptr* and store the data in *dst*. 3758 * Return 3759 * 0 on success, or a negative error in case of failure. 3760 * 3761 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 3762 * Description 3763 * Copy a NUL terminated string from an unsafe user address 3764 * *unsafe_ptr* to *dst*. The *size* should include the 3765 * terminating NUL byte. In case the string length is smaller than 3766 * *size*, the target is not padded with further NUL bytes. If the 3767 * string length is larger than *size*, just *size*-1 bytes are 3768 * copied and the last byte is set to NUL. 3769 * 3770 * On success, returns the number of bytes that were written, 3771 * including the terminal NUL. This makes this helper useful in 3772 * tracing programs for reading strings, and more importantly to 3773 * get its length at runtime. See the following snippet: 3774 * 3775 * :: 3776 * 3777 * SEC("kprobe/sys_open") 3778 * void bpf_sys_open(struct pt_regs *ctx) 3779 * { 3780 * char buf[PATHLEN]; // PATHLEN is defined to 256 3781 * int res = bpf_probe_read_user_str(buf, sizeof(buf), 3782 * ctx->di); 3783 * 3784 * // Consume buf, for example push it to 3785 * // userspace via bpf_perf_event_output(); we 3786 * // can use res (the string length) as event 3787 * // size, after checking its boundaries. 3788 * } 3789 * 3790 * In comparison, using **bpf_probe_read_user**\ () helper here 3791 * instead to read the string would require to estimate the length 3792 * at compile time, and would often result in copying more memory 3793 * than necessary. 3794 * 3795 * Another useful use case is when parsing individual process 3796 * arguments or individual environment variables navigating 3797 * *current*\ **->mm->arg_start** and *current*\ 3798 * **->mm->env_start**: using this helper and the return value, 3799 * one can quickly iterate at the right offset of the memory area. 3800 * Return 3801 * On success, the strictly positive length of the output string, 3802 * including the trailing NUL character. On error, a negative 3803 * value. 3804 * 3805 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 3806 * Description 3807 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3808 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 3809 * Return 3810 * On success, the strictly positive length of the string, including 3811 * the trailing NUL character. On error, a negative value. 3812 * 3813 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3814 * Description 3815 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3816 * *rcv_nxt* is the ack_seq to be sent out. 3817 * Return 3818 * 0 on success, or a negative error in case of failure. 3819 * 3820 * long bpf_send_signal_thread(u32 sig) 3821 * Description 3822 * Send signal *sig* to the thread corresponding to the current task. 3823 * Return 3824 * 0 on success or successfully queued. 3825 * 3826 * **-EBUSY** if work queue under nmi is full. 3827 * 3828 * **-EINVAL** if *sig* is invalid. 3829 * 3830 * **-EPERM** if no permission to send the *sig*. 3831 * 3832 * **-EAGAIN** if bpf program can try again. 3833 * 3834 * u64 bpf_jiffies64(void) 3835 * Description 3836 * Obtain the 64bit jiffies 3837 * Return 3838 * The 64 bit jiffies 3839 * 3840 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3841 * Description 3842 * For an eBPF program attached to a perf event, retrieve the 3843 * branch records (**struct perf_branch_entry**) associated to *ctx* 3844 * and store it in the buffer pointed by *buf* up to size 3845 * *size* bytes. 3846 * Return 3847 * On success, number of bytes written to *buf*. On error, a 3848 * negative value. 3849 * 3850 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3851 * instead return the number of bytes required to store all the 3852 * branch entries. If this flag is set, *buf* may be NULL. 3853 * 3854 * **-EINVAL** if arguments invalid or **size** not a multiple 3855 * of **sizeof**\ (**struct perf_branch_entry**\ ). 3856 * 3857 * **-ENOENT** if architecture does not support branch records. 3858 * 3859 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3860 * Description 3861 * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3862 * *namespace* will be returned in *nsdata*. 3863 * Return 3864 * 0 on success, or one of the following in case of failure: 3865 * 3866 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3867 * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3868 * 3869 * **-ENOENT** if pidns does not exists for the current task. 3870 * 3871 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3872 * Description 3873 * Write raw *data* blob into a special BPF perf event held by 3874 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3875 * event must have the following attributes: **PERF_SAMPLE_RAW** 3876 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3877 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3878 * 3879 * The *flags* are used to indicate the index in *map* for which 3880 * the value must be put, masked with **BPF_F_INDEX_MASK**. 3881 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3882 * to indicate that the index of the current CPU core should be 3883 * used. 3884 * 3885 * The value to write, of *size*, is passed through eBPF stack and 3886 * pointed by *data*. 3887 * 3888 * *ctx* is a pointer to in-kernel struct xdp_buff. 3889 * 3890 * This helper is similar to **bpf_perf_eventoutput**\ () but 3891 * restricted to raw_tracepoint bpf programs. 3892 * Return 3893 * 0 on success, or a negative error in case of failure. 3894 * 3895 * u64 bpf_get_netns_cookie(void *ctx) 3896 * Description 3897 * Retrieve the cookie (generated by the kernel) of the network 3898 * namespace the input *ctx* is associated with. The network 3899 * namespace cookie remains stable for its lifetime and provides 3900 * a global identifier that can be assumed unique. If *ctx* is 3901 * NULL, then the helper returns the cookie for the initial 3902 * network namespace. The cookie itself is very similar to that 3903 * of **bpf_get_socket_cookie**\ () helper, but for network 3904 * namespaces instead of sockets. 3905 * Return 3906 * A 8-byte long opaque number. 3907 * 3908 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 3909 * Description 3910 * Return id of cgroup v2 that is ancestor of the cgroup associated 3911 * with the current task at the *ancestor_level*. The root cgroup 3912 * is at *ancestor_level* zero and each step down the hierarchy 3913 * increments the level. If *ancestor_level* == level of cgroup 3914 * associated with the current task, then return value will be the 3915 * same as that of **bpf_get_current_cgroup_id**\ (). 3916 * 3917 * The helper is useful to implement policies based on cgroups 3918 * that are upper in hierarchy than immediate cgroup associated 3919 * with the current task. 3920 * 3921 * The format of returned id and helper limitations are same as in 3922 * **bpf_get_current_cgroup_id**\ (). 3923 * Return 3924 * The id is returned or 0 in case the id could not be retrieved. 3925 * 3926 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3927 * Description 3928 * Helper is overloaded depending on BPF program type. This 3929 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3930 * **BPF_PROG_TYPE_SCHED_ACT** programs. 3931 * 3932 * Assign the *sk* to the *skb*. When combined with appropriate 3933 * routing configuration to receive the packet towards the socket, 3934 * will cause *skb* to be delivered to the specified socket. 3935 * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3936 * **bpf_clone_redirect**\ () or other methods outside of BPF may 3937 * interfere with successful delivery to the socket. 3938 * 3939 * This operation is only valid from TC ingress path. 3940 * 3941 * The *flags* argument must be zero. 3942 * Return 3943 * 0 on success, or a negative error in case of failure: 3944 * 3945 * **-EINVAL** if specified *flags* are not supported. 3946 * 3947 * **-ENOENT** if the socket is unavailable for assignment. 3948 * 3949 * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3950 * 3951 * **-EOPNOTSUPP** if the operation is not supported, for example 3952 * a call from outside of TC ingress. 3953 * 3954 * **-ESOCKTNOSUPPORT** if the socket type is not supported 3955 * (reuseport). 3956 * 3957 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3958 * Description 3959 * Helper is overloaded depending on BPF program type. This 3960 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3961 * 3962 * Select the *sk* as a result of a socket lookup. 3963 * 3964 * For the operation to succeed passed socket must be compatible 3965 * with the packet description provided by the *ctx* object. 3966 * 3967 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3968 * be an exact match. While IP family (**AF_INET** or 3969 * **AF_INET6**) must be compatible, that is IPv6 sockets 3970 * that are not v6-only can be selected for IPv4 packets. 3971 * 3972 * Only TCP listeners and UDP unconnected sockets can be 3973 * selected. *sk* can also be NULL to reset any previous 3974 * selection. 3975 * 3976 * *flags* argument can combination of following values: 3977 * 3978 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3979 * socket selection, potentially done by a BPF program 3980 * that ran before us. 3981 * 3982 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3983 * load-balancing within reuseport group for the socket 3984 * being selected. 3985 * 3986 * On success *ctx->sk* will point to the selected socket. 3987 * 3988 * Return 3989 * 0 on success, or a negative errno in case of failure. 3990 * 3991 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3992 * not compatible with packet family (*ctx->family*). 3993 * 3994 * * **-EEXIST** if socket has been already selected, 3995 * potentially by another program, and 3996 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3997 * 3998 * * **-EINVAL** if unsupported flags were specified. 3999 * 4000 * * **-EPROTOTYPE** if socket L4 protocol 4001 * (*sk->protocol*) doesn't match packet protocol 4002 * (*ctx->protocol*). 4003 * 4004 * * **-ESOCKTNOSUPPORT** if socket is not in allowed 4005 * state (TCP listening or UDP unconnected). 4006 * 4007 * u64 bpf_ktime_get_boot_ns(void) 4008 * Description 4009 * Return the time elapsed since system boot, in nanoseconds. 4010 * Does include the time the system was suspended. 4011 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 4012 * Return 4013 * Current *ktime*. 4014 * 4015 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 4016 * Description 4017 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 4018 * out the format string. 4019 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 4020 * the format string itself. The *data* and *data_len* are format string 4021 * arguments. The *data* are a **u64** array and corresponding format string 4022 * values are stored in the array. For strings and pointers where pointees 4023 * are accessed, only the pointer values are stored in the *data* array. 4024 * The *data_len* is the size of *data* in bytes. 4025 * 4026 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 4027 * Reading kernel memory may fail due to either invalid address or 4028 * valid address but requiring a major memory fault. If reading kernel memory 4029 * fails, the string for **%s** will be an empty string, and the ip 4030 * address for **%p{i,I}{4,6}** will be 0. Not returning error to 4031 * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 4032 * Return 4033 * 0 on success, or a negative error in case of failure: 4034 * 4035 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 4036 * by returning 1 from bpf program. 4037 * 4038 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 4039 * 4040 * **-E2BIG** if *fmt* contains too many format specifiers. 4041 * 4042 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4043 * 4044 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 4045 * Description 4046 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 4047 * The *m* represents the seq_file. The *data* and *len* represent the 4048 * data to write in bytes. 4049 * Return 4050 * 0 on success, or a negative error in case of failure: 4051 * 4052 * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 4053 * 4054 * u64 bpf_sk_cgroup_id(void *sk) 4055 * Description 4056 * Return the cgroup v2 id of the socket *sk*. 4057 * 4058 * *sk* must be a non-**NULL** pointer to a socket, e.g. one 4059 * returned from **bpf_sk_lookup_xxx**\ (), 4060 * **bpf_sk_fullsock**\ (), etc. The format of returned id is 4061 * same as in **bpf_skb_cgroup_id**\ (). 4062 * 4063 * This helper is available only if the kernel was compiled with 4064 * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 4065 * Return 4066 * The id is returned or 0 in case the id could not be retrieved. 4067 * 4068 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 4069 * Description 4070 * Return id of cgroup v2 that is ancestor of cgroup associated 4071 * with the *sk* at the *ancestor_level*. The root cgroup is at 4072 * *ancestor_level* zero and each step down the hierarchy 4073 * increments the level. If *ancestor_level* == level of cgroup 4074 * associated with *sk*, then return value will be same as that 4075 * of **bpf_sk_cgroup_id**\ (). 4076 * 4077 * The helper is useful to implement policies based on cgroups 4078 * that are upper in hierarchy than immediate cgroup associated 4079 * with *sk*. 4080 * 4081 * The format of returned id and helper limitations are same as in 4082 * **bpf_sk_cgroup_id**\ (). 4083 * Return 4084 * The id is returned or 0 in case the id could not be retrieved. 4085 * 4086 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 4087 * Description 4088 * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 4089 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4090 * of new data availability is sent. 4091 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4092 * of new data availability is sent unconditionally. 4093 * If **0** is specified in *flags*, an adaptive notification 4094 * of new data availability is sent. 4095 * 4096 * An adaptive notification is a notification sent whenever the user-space 4097 * process has caught up and consumed all available payloads. In case the user-space 4098 * process is still processing a previous payload, then no notification is needed 4099 * as it will process the newly added payload automatically. 4100 * Return 4101 * 0 on success, or a negative error in case of failure. 4102 * 4103 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 4104 * Description 4105 * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 4106 * *flags* must be 0. 4107 * Return 4108 * Valid pointer with *size* bytes of memory available; NULL, 4109 * otherwise. 4110 * 4111 * void bpf_ringbuf_submit(void *data, u64 flags) 4112 * Description 4113 * Submit reserved ring buffer sample, pointed to by *data*. 4114 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4115 * of new data availability is sent. 4116 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4117 * of new data availability is sent unconditionally. 4118 * If **0** is specified in *flags*, an adaptive notification 4119 * of new data availability is sent. 4120 * 4121 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4122 * Return 4123 * Nothing. Always succeeds. 4124 * 4125 * void bpf_ringbuf_discard(void *data, u64 flags) 4126 * Description 4127 * Discard reserved ring buffer sample, pointed to by *data*. 4128 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 4129 * of new data availability is sent. 4130 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 4131 * of new data availability is sent unconditionally. 4132 * If **0** is specified in *flags*, an adaptive notification 4133 * of new data availability is sent. 4134 * 4135 * See 'bpf_ringbuf_output()' for the definition of adaptive notification. 4136 * Return 4137 * Nothing. Always succeeds. 4138 * 4139 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 4140 * Description 4141 * Query various characteristics of provided ring buffer. What 4142 * exactly is queries is determined by *flags*: 4143 * 4144 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 4145 * * **BPF_RB_RING_SIZE**: The size of ring buffer. 4146 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 4147 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 4148 * 4149 * Data returned is just a momentary snapshot of actual values 4150 * and could be inaccurate, so this facility should be used to 4151 * power heuristics and for reporting, not to make 100% correct 4152 * calculation. 4153 * Return 4154 * Requested value, or 0, if *flags* are not recognized. 4155 * 4156 * long bpf_csum_level(struct sk_buff *skb, u64 level) 4157 * Description 4158 * Change the skbs checksum level by one layer up or down, or 4159 * reset it entirely to none in order to have the stack perform 4160 * checksum validation. The level is applicable to the following 4161 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 4162 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 4163 * through **bpf_skb_adjust_room**\ () helper with passing in 4164 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 4165 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 4166 * the UDP header is removed. Similarly, an encap of the latter 4167 * into the former could be accompanied by a helper call to 4168 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 4169 * skb is still intended to be processed in higher layers of the 4170 * stack instead of just egressing at tc. 4171 * 4172 * There are three supported level settings at this time: 4173 * 4174 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 4175 * with CHECKSUM_UNNECESSARY. 4176 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 4177 * with CHECKSUM_UNNECESSARY. 4178 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 4179 * sets CHECKSUM_NONE to force checksum validation by the stack. 4180 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 4181 * skb->csum_level. 4182 * Return 4183 * 0 on success, or a negative error in case of failure. In the 4184 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 4185 * is returned or the error code -EACCES in case the skb is not 4186 * subject to CHECKSUM_UNNECESSARY. 4187 * 4188 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 4189 * Description 4190 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 4191 * Return 4192 * *sk* if casting is valid, or **NULL** otherwise. 4193 * 4194 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 4195 * Description 4196 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 4197 * Return 4198 * *sk* if casting is valid, or **NULL** otherwise. 4199 * 4200 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 4201 * Description 4202 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 4203 * Return 4204 * *sk* if casting is valid, or **NULL** otherwise. 4205 * 4206 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 4207 * Description 4208 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 4209 * Return 4210 * *sk* if casting is valid, or **NULL** otherwise. 4211 * 4212 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 4213 * Description 4214 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 4215 * Return 4216 * *sk* if casting is valid, or **NULL** otherwise. 4217 * 4218 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 4219 * Description 4220 * Return a user or a kernel stack in bpf program provided buffer. 4221 * To achieve this, the helper needs *task*, which is a valid 4222 * pointer to **struct task_struct**. To store the stacktrace, the 4223 * bpf program provides *buf* with a nonnegative *size*. 4224 * 4225 * The last argument, *flags*, holds the number of stack frames to 4226 * skip (from 0 to 255), masked with 4227 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 4228 * the following flags: 4229 * 4230 * **BPF_F_USER_STACK** 4231 * Collect a user space stack instead of a kernel stack. 4232 * **BPF_F_USER_BUILD_ID** 4233 * Collect buildid+offset instead of ips for user stack, 4234 * only valid if **BPF_F_USER_STACK** is also specified. 4235 * 4236 * **bpf_get_task_stack**\ () can collect up to 4237 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 4238 * to sufficient large buffer size. Note that 4239 * this limit can be controlled with the **sysctl** program, and 4240 * that it should be manually increased in order to profile long 4241 * user stacks (such as stacks for Java programs). To do so, use: 4242 * 4243 * :: 4244 * 4245 * # sysctl kernel.perf_event_max_stack=<new value> 4246 * Return 4247 * A non-negative value equal to or less than *size* on success, 4248 * or a negative error in case of failure. 4249 * 4250 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 4251 * Description 4252 * Load header option. Support reading a particular TCP header 4253 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 4254 * 4255 * If *flags* is 0, it will search the option from the 4256 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 4257 * has details on what skb_data contains under different 4258 * *skops*\ **->op**. 4259 * 4260 * The first byte of the *searchby_res* specifies the 4261 * kind that it wants to search. 4262 * 4263 * If the searching kind is an experimental kind 4264 * (i.e. 253 or 254 according to RFC6994). It also 4265 * needs to specify the "magic" which is either 4266 * 2 bytes or 4 bytes. It then also needs to 4267 * specify the size of the magic by using 4268 * the 2nd byte which is "kind-length" of a TCP 4269 * header option and the "kind-length" also 4270 * includes the first 2 bytes "kind" and "kind-length" 4271 * itself as a normal TCP header option also does. 4272 * 4273 * For example, to search experimental kind 254 with 4274 * 2 byte magic 0xeB9F, the searchby_res should be 4275 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 4276 * 4277 * To search for the standard window scale option (3), 4278 * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 4279 * Note, kind-length must be 0 for regular option. 4280 * 4281 * Searching for No-Op (0) and End-of-Option-List (1) are 4282 * not supported. 4283 * 4284 * *len* must be at least 2 bytes which is the minimal size 4285 * of a header option. 4286 * 4287 * Supported flags: 4288 * 4289 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 4290 * saved_syn packet or the just-received syn packet. 4291 * 4292 * Return 4293 * > 0 when found, the header option is copied to *searchby_res*. 4294 * The return value is the total length copied. On failure, a 4295 * negative error code is returned: 4296 * 4297 * **-EINVAL** if a parameter is invalid. 4298 * 4299 * **-ENOMSG** if the option is not found. 4300 * 4301 * **-ENOENT** if no syn packet is available when 4302 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 4303 * 4304 * **-ENOSPC** if there is not enough space. Only *len* number of 4305 * bytes are copied. 4306 * 4307 * **-EFAULT** on failure to parse the header options in the 4308 * packet. 4309 * 4310 * **-EPERM** if the helper cannot be used under the current 4311 * *skops*\ **->op**. 4312 * 4313 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 4314 * Description 4315 * Store header option. The data will be copied 4316 * from buffer *from* with length *len* to the TCP header. 4317 * 4318 * The buffer *from* should have the whole option that 4319 * includes the kind, kind-length, and the actual 4320 * option data. The *len* must be at least kind-length 4321 * long. The kind-length does not have to be 4 byte 4322 * aligned. The kernel will take care of the padding 4323 * and setting the 4 bytes aligned value to th->doff. 4324 * 4325 * This helper will check for duplicated option 4326 * by searching the same option in the outgoing skb. 4327 * 4328 * This helper can only be called during 4329 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4330 * 4331 * Return 4332 * 0 on success, or negative error in case of failure: 4333 * 4334 * **-EINVAL** If param is invalid. 4335 * 4336 * **-ENOSPC** if there is not enough space in the header. 4337 * Nothing has been written 4338 * 4339 * **-EEXIST** if the option already exists. 4340 * 4341 * **-EFAULT** on failrue to parse the existing header options. 4342 * 4343 * **-EPERM** if the helper cannot be used under the current 4344 * *skops*\ **->op**. 4345 * 4346 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 4347 * Description 4348 * Reserve *len* bytes for the bpf header option. The 4349 * space will be used by **bpf_store_hdr_opt**\ () later in 4350 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 4351 * 4352 * If **bpf_reserve_hdr_opt**\ () is called multiple times, 4353 * the total number of bytes will be reserved. 4354 * 4355 * This helper can only be called during 4356 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 4357 * 4358 * Return 4359 * 0 on success, or negative error in case of failure: 4360 * 4361 * **-EINVAL** if a parameter is invalid. 4362 * 4363 * **-ENOSPC** if there is not enough space in the header. 4364 * 4365 * **-EPERM** if the helper cannot be used under the current 4366 * *skops*\ **->op**. 4367 * 4368 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 4369 * Description 4370 * Get a bpf_local_storage from an *inode*. 4371 * 4372 * Logically, it could be thought of as getting the value from 4373 * a *map* with *inode* as the **key**. From this 4374 * perspective, the usage is not much different from 4375 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 4376 * helper enforces the key must be an inode and the map must also 4377 * be a **BPF_MAP_TYPE_INODE_STORAGE**. 4378 * 4379 * Underneath, the value is stored locally at *inode* instead of 4380 * the *map*. The *map* is used as the bpf-local-storage 4381 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4382 * searched against all bpf_local_storage residing at *inode*. 4383 * 4384 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4385 * used such that a new bpf_local_storage will be 4386 * created if one does not exist. *value* can be used 4387 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4388 * the initial value of a bpf_local_storage. If *value* is 4389 * **NULL**, the new bpf_local_storage will be zero initialized. 4390 * Return 4391 * A bpf_local_storage pointer is returned on success. 4392 * 4393 * **NULL** if not found or there was an error in adding 4394 * a new bpf_local_storage. 4395 * 4396 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 4397 * Description 4398 * Delete a bpf_local_storage from an *inode*. 4399 * Return 4400 * 0 on success. 4401 * 4402 * **-ENOENT** if the bpf_local_storage cannot be found. 4403 * 4404 * long bpf_d_path(struct path *path, char *buf, u32 sz) 4405 * Description 4406 * Return full path for given **struct path** object, which 4407 * needs to be the kernel BTF *path* object. The path is 4408 * returned in the provided buffer *buf* of size *sz* and 4409 * is zero terminated. 4410 * 4411 * Return 4412 * On success, the strictly positive length of the string, 4413 * including the trailing NUL character. On error, a negative 4414 * value. 4415 * 4416 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 4417 * Description 4418 * Read *size* bytes from user space address *user_ptr* and store 4419 * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 4420 * Return 4421 * 0 on success, or a negative error in case of failure. 4422 * 4423 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 4424 * Description 4425 * Use BTF to store a string representation of *ptr*->ptr in *str*, 4426 * using *ptr*->type_id. This value should specify the type 4427 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 4428 * can be used to look up vmlinux BTF type ids. Traversing the 4429 * data structure using BTF, the type information and values are 4430 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 4431 * the pointer data is carried out to avoid kernel crashes during 4432 * operation. Smaller types can use string space on the stack; 4433 * larger programs can use map data to store the string 4434 * representation. 4435 * 4436 * The string can be subsequently shared with userspace via 4437 * bpf_perf_event_output() or ring buffer interfaces. 4438 * bpf_trace_printk() is to be avoided as it places too small 4439 * a limit on string size to be useful. 4440 * 4441 * *flags* is a combination of 4442 * 4443 * **BTF_F_COMPACT** 4444 * no formatting around type information 4445 * **BTF_F_NONAME** 4446 * no struct/union member names/types 4447 * **BTF_F_PTR_RAW** 4448 * show raw (unobfuscated) pointer values; 4449 * equivalent to printk specifier %px. 4450 * **BTF_F_ZERO** 4451 * show zero-valued struct/union members; they 4452 * are not displayed by default 4453 * 4454 * Return 4455 * The number of bytes that were written (or would have been 4456 * written if output had to be truncated due to string size), 4457 * or a negative error in cases of failure. 4458 * 4459 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 4460 * Description 4461 * Use BTF to write to seq_write a string representation of 4462 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 4463 * *flags* are identical to those used for bpf_snprintf_btf. 4464 * Return 4465 * 0 on success or a negative error in case of failure. 4466 * 4467 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 4468 * Description 4469 * See **bpf_get_cgroup_classid**\ () for the main description. 4470 * This helper differs from **bpf_get_cgroup_classid**\ () in that 4471 * the cgroup v1 net_cls class is retrieved only from the *skb*'s 4472 * associated socket instead of the current process. 4473 * Return 4474 * The id is returned or 0 in case the id could not be retrieved. 4475 * 4476 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags) 4477 * Description 4478 * Redirect the packet to another net device of index *ifindex* 4479 * and fill in L2 addresses from neighboring subsystem. This helper 4480 * is somewhat similar to **bpf_redirect**\ (), except that it 4481 * populates L2 addresses as well, meaning, internally, the helper 4482 * relies on the neighbor lookup for the L2 address of the nexthop. 4483 * 4484 * The helper will perform a FIB lookup based on the skb's 4485 * networking header to get the address of the next hop, unless 4486 * this is supplied by the caller in the *params* argument. The 4487 * *plen* argument indicates the len of *params* and should be set 4488 * to 0 if *params* is NULL. 4489 * 4490 * The *flags* argument is reserved and must be 0. The helper is 4491 * currently only supported for tc BPF program types, and enabled 4492 * for IPv4 and IPv6 protocols. 4493 * Return 4494 * The helper returns **TC_ACT_REDIRECT** on success or 4495 * **TC_ACT_SHOT** on error. 4496 * 4497 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu) 4498 * Description 4499 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4500 * pointer to the percpu kernel variable on *cpu*. A ksym is an 4501 * extern variable decorated with '__ksym'. For ksym, there is a 4502 * global var (either static or global) defined of the same name 4503 * in the kernel. The ksym is percpu if the global var is percpu. 4504 * The returned pointer points to the global percpu var on *cpu*. 4505 * 4506 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the 4507 * kernel, except that bpf_per_cpu_ptr() may return NULL. This 4508 * happens if *cpu* is larger than nr_cpu_ids. The caller of 4509 * bpf_per_cpu_ptr() must check the returned value. 4510 * Return 4511 * A pointer pointing to the kernel percpu variable on *cpu*, or 4512 * NULL, if *cpu* is invalid. 4513 * 4514 * void *bpf_this_cpu_ptr(const void *percpu_ptr) 4515 * Description 4516 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a 4517 * pointer to the percpu kernel variable on this cpu. See the 4518 * description of 'ksym' in **bpf_per_cpu_ptr**\ (). 4519 * 4520 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in 4521 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would 4522 * never return NULL. 4523 * Return 4524 * A pointer pointing to the kernel percpu variable on this cpu. 4525 * 4526 * long bpf_redirect_peer(u32 ifindex, u64 flags) 4527 * Description 4528 * Redirect the packet to another net device of index *ifindex*. 4529 * This helper is somewhat similar to **bpf_redirect**\ (), except 4530 * that the redirection happens to the *ifindex*' peer device and 4531 * the netns switch takes place from ingress to ingress without 4532 * going through the CPU's backlog queue. 4533 * 4534 * The *flags* argument is reserved and must be 0. The helper is 4535 * currently only supported for tc BPF program types at the ingress 4536 * hook and for veth device types. The peer device must reside in a 4537 * different network namespace. 4538 * Return 4539 * The helper returns **TC_ACT_REDIRECT** on success or 4540 * **TC_ACT_SHOT** on error. 4541 * 4542 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags) 4543 * Description 4544 * Get a bpf_local_storage from the *task*. 4545 * 4546 * Logically, it could be thought of as getting the value from 4547 * a *map* with *task* as the **key**. From this 4548 * perspective, the usage is not much different from 4549 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this 4550 * helper enforces the key must be an task_struct and the map must also 4551 * be a **BPF_MAP_TYPE_TASK_STORAGE**. 4552 * 4553 * Underneath, the value is stored locally at *task* instead of 4554 * the *map*. The *map* is used as the bpf-local-storage 4555 * "type". The bpf-local-storage "type" (i.e. the *map*) is 4556 * searched against all bpf_local_storage residing at *task*. 4557 * 4558 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 4559 * used such that a new bpf_local_storage will be 4560 * created if one does not exist. *value* can be used 4561 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 4562 * the initial value of a bpf_local_storage. If *value* is 4563 * **NULL**, the new bpf_local_storage will be zero initialized. 4564 * Return 4565 * A bpf_local_storage pointer is returned on success. 4566 * 4567 * **NULL** if not found or there was an error in adding 4568 * a new bpf_local_storage. 4569 * 4570 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task) 4571 * Description 4572 * Delete a bpf_local_storage from a *task*. 4573 * Return 4574 * 0 on success. 4575 * 4576 * **-ENOENT** if the bpf_local_storage cannot be found. 4577 * 4578 * struct task_struct *bpf_get_current_task_btf(void) 4579 * Description 4580 * Return a BTF pointer to the "current" task. 4581 * This pointer can also be used in helpers that accept an 4582 * *ARG_PTR_TO_BTF_ID* of type *task_struct*. 4583 * Return 4584 * Pointer to the current task. 4585 * 4586 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags) 4587 * Description 4588 * Set or clear certain options on *bprm*: 4589 * 4590 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit 4591 * which sets the **AT_SECURE** auxv for glibc. The bit 4592 * is cleared if the flag is not specified. 4593 * Return 4594 * **-EINVAL** if invalid *flags* are passed, zero otherwise. 4595 * 4596 * u64 bpf_ktime_get_coarse_ns(void) 4597 * Description 4598 * Return a coarse-grained version of the time elapsed since 4599 * system boot, in nanoseconds. Does not include time the system 4600 * was suspended. 4601 * 4602 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**) 4603 * Return 4604 * Current *ktime*. 4605 * 4606 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size) 4607 * Description 4608 * Returns the stored IMA hash of the *inode* (if it's avaialable). 4609 * If the hash is larger than *size*, then only *size* 4610 * bytes will be copied to *dst* 4611 * Return 4612 * The **hash_algo** is returned on success, 4613 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if 4614 * invalid arguments are passed. 4615 * 4616 * struct socket *bpf_sock_from_file(struct file *file) 4617 * Description 4618 * If the given file represents a socket, returns the associated 4619 * socket. 4620 * Return 4621 * A pointer to a struct socket on success or NULL if the file is 4622 * not a socket. 4623 * 4624 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags) 4625 * Description 4626 * Check packet size against exceeding MTU of net device (based 4627 * on *ifindex*). This helper will likely be used in combination 4628 * with helpers that adjust/change the packet size. 4629 * 4630 * The argument *len_diff* can be used for querying with a planned 4631 * size change. This allows to check MTU prior to changing packet 4632 * ctx. Providing an *len_diff* adjustment that is larger than the 4633 * actual packet size (resulting in negative packet size) will in 4634 * principle not exceed the MTU, why it is not considered a 4635 * failure. Other BPF-helpers are needed for performing the 4636 * planned size change, why the responsability for catch a negative 4637 * packet size belong in those helpers. 4638 * 4639 * Specifying *ifindex* zero means the MTU check is performed 4640 * against the current net device. This is practical if this isn't 4641 * used prior to redirect. 4642 * 4643 * On input *mtu_len* must be a valid pointer, else verifier will 4644 * reject BPF program. If the value *mtu_len* is initialized to 4645 * zero then the ctx packet size is use. When value *mtu_len* is 4646 * provided as input this specify the L3 length that the MTU check 4647 * is done against. Remember XDP and TC length operate at L2, but 4648 * this value is L3 as this correlate to MTU and IP-header tot_len 4649 * values which are L3 (similar behavior as bpf_fib_lookup). 4650 * 4651 * The Linux kernel route table can configure MTUs on a more 4652 * specific per route level, which is not provided by this helper. 4653 * For route level MTU checks use the **bpf_fib_lookup**\ () 4654 * helper. 4655 * 4656 * *ctx* is either **struct xdp_md** for XDP programs or 4657 * **struct sk_buff** for tc cls_act programs. 4658 * 4659 * The *flags* argument can be a combination of one or more of the 4660 * following values: 4661 * 4662 * **BPF_MTU_CHK_SEGS** 4663 * This flag will only works for *ctx* **struct sk_buff**. 4664 * If packet context contains extra packet segment buffers 4665 * (often knows as GSO skb), then MTU check is harder to 4666 * check at this point, because in transmit path it is 4667 * possible for the skb packet to get re-segmented 4668 * (depending on net device features). This could still be 4669 * a MTU violation, so this flag enables performing MTU 4670 * check against segments, with a different violation 4671 * return code to tell it apart. Check cannot use len_diff. 4672 * 4673 * On return *mtu_len* pointer contains the MTU value of the net 4674 * device. Remember the net device configured MTU is the L3 size, 4675 * which is returned here and XDP and TC length operate at L2. 4676 * Helper take this into account for you, but remember when using 4677 * MTU value in your BPF-code. 4678 * 4679 * Return 4680 * * 0 on success, and populate MTU value in *mtu_len* pointer. 4681 * 4682 * * < 0 if any input argument is invalid (*mtu_len* not updated) 4683 * 4684 * MTU violations return positive values, but also populate MTU 4685 * value in *mtu_len* pointer, as this can be needed for 4686 * implementing PMTU handing: 4687 * 4688 * * **BPF_MTU_CHK_RET_FRAG_NEEDED** 4689 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG** 4690 * 4691 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags) 4692 * Description 4693 * For each element in **map**, call **callback_fn** function with 4694 * **map**, **callback_ctx** and other map-specific parameters. 4695 * The **callback_fn** should be a static function and 4696 * the **callback_ctx** should be a pointer to the stack. 4697 * The **flags** is used to control certain aspects of the helper. 4698 * Currently, the **flags** must be 0. 4699 * 4700 * The following are a list of supported map types and their 4701 * respective expected callback signatures: 4702 * 4703 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH, 4704 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH, 4705 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY 4706 * 4707 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx); 4708 * 4709 * For per_cpu maps, the map_value is the value on the cpu where the 4710 * bpf_prog is running. 4711 * 4712 * If **callback_fn** return 0, the helper will continue to the next 4713 * element. If return value is 1, the helper will skip the rest of 4714 * elements and return. Other return values are not used now. 4715 * 4716 * Return 4717 * The number of traversed map elements for success, **-EINVAL** for 4718 * invalid **flags**. 4719 * 4720 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len) 4721 * Description 4722 * Outputs a string into the **str** buffer of size **str_size** 4723 * based on a format string stored in a read-only map pointed by 4724 * **fmt**. 4725 * 4726 * Each format specifier in **fmt** corresponds to one u64 element 4727 * in the **data** array. For strings and pointers where pointees 4728 * are accessed, only the pointer values are stored in the *data* 4729 * array. The *data_len* is the size of *data* in bytes. 4730 * 4731 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel 4732 * memory. Reading kernel memory may fail due to either invalid 4733 * address or valid address but requiring a major memory fault. If 4734 * reading kernel memory fails, the string for **%s** will be an 4735 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0. 4736 * Not returning error to bpf program is consistent with what 4737 * **bpf_trace_printk**\ () does for now. 4738 * 4739 * Return 4740 * The strictly positive length of the formatted string, including 4741 * the trailing zero character. If the return value is greater than 4742 * **str_size**, **str** contains a truncated string, guaranteed to 4743 * be zero-terminated except when **str_size** is 0. 4744 * 4745 * Or **-EBUSY** if the per-CPU memory copy buffer is busy. 4746 * 4747 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size) 4748 * Description 4749 * Execute bpf syscall with given arguments. 4750 * Return 4751 * A syscall result. 4752 * 4753 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags) 4754 * Description 4755 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs. 4756 * Return 4757 * Returns btf_id and btf_obj_fd in lower and upper 32 bits. 4758 * 4759 * long bpf_sys_close(u32 fd) 4760 * Description 4761 * Execute close syscall for given FD. 4762 * Return 4763 * A syscall result. 4764 */ 4765 #define __BPF_FUNC_MAPPER(FN) \ 4766 FN(unspec), \ 4767 FN(map_lookup_elem), \ 4768 FN(map_update_elem), \ 4769 FN(map_delete_elem), \ 4770 FN(probe_read), \ 4771 FN(ktime_get_ns), \ 4772 FN(trace_printk), \ 4773 FN(get_prandom_u32), \ 4774 FN(get_smp_processor_id), \ 4775 FN(skb_store_bytes), \ 4776 FN(l3_csum_replace), \ 4777 FN(l4_csum_replace), \ 4778 FN(tail_call), \ 4779 FN(clone_redirect), \ 4780 FN(get_current_pid_tgid), \ 4781 FN(get_current_uid_gid), \ 4782 FN(get_current_comm), \ 4783 FN(get_cgroup_classid), \ 4784 FN(skb_vlan_push), \ 4785 FN(skb_vlan_pop), \ 4786 FN(skb_get_tunnel_key), \ 4787 FN(skb_set_tunnel_key), \ 4788 FN(perf_event_read), \ 4789 FN(redirect), \ 4790 FN(get_route_realm), \ 4791 FN(perf_event_output), \ 4792 FN(skb_load_bytes), \ 4793 FN(get_stackid), \ 4794 FN(csum_diff), \ 4795 FN(skb_get_tunnel_opt), \ 4796 FN(skb_set_tunnel_opt), \ 4797 FN(skb_change_proto), \ 4798 FN(skb_change_type), \ 4799 FN(skb_under_cgroup), \ 4800 FN(get_hash_recalc), \ 4801 FN(get_current_task), \ 4802 FN(probe_write_user), \ 4803 FN(current_task_under_cgroup), \ 4804 FN(skb_change_tail), \ 4805 FN(skb_pull_data), \ 4806 FN(csum_update), \ 4807 FN(set_hash_invalid), \ 4808 FN(get_numa_node_id), \ 4809 FN(skb_change_head), \ 4810 FN(xdp_adjust_head), \ 4811 FN(probe_read_str), \ 4812 FN(get_socket_cookie), \ 4813 FN(get_socket_uid), \ 4814 FN(set_hash), \ 4815 FN(setsockopt), \ 4816 FN(skb_adjust_room), \ 4817 FN(redirect_map), \ 4818 FN(sk_redirect_map), \ 4819 FN(sock_map_update), \ 4820 FN(xdp_adjust_meta), \ 4821 FN(perf_event_read_value), \ 4822 FN(perf_prog_read_value), \ 4823 FN(getsockopt), \ 4824 FN(override_return), \ 4825 FN(sock_ops_cb_flags_set), \ 4826 FN(msg_redirect_map), \ 4827 FN(msg_apply_bytes), \ 4828 FN(msg_cork_bytes), \ 4829 FN(msg_pull_data), \ 4830 FN(bind), \ 4831 FN(xdp_adjust_tail), \ 4832 FN(skb_get_xfrm_state), \ 4833 FN(get_stack), \ 4834 FN(skb_load_bytes_relative), \ 4835 FN(fib_lookup), \ 4836 FN(sock_hash_update), \ 4837 FN(msg_redirect_hash), \ 4838 FN(sk_redirect_hash), \ 4839 FN(lwt_push_encap), \ 4840 FN(lwt_seg6_store_bytes), \ 4841 FN(lwt_seg6_adjust_srh), \ 4842 FN(lwt_seg6_action), \ 4843 FN(rc_repeat), \ 4844 FN(rc_keydown), \ 4845 FN(skb_cgroup_id), \ 4846 FN(get_current_cgroup_id), \ 4847 FN(get_local_storage), \ 4848 FN(sk_select_reuseport), \ 4849 FN(skb_ancestor_cgroup_id), \ 4850 FN(sk_lookup_tcp), \ 4851 FN(sk_lookup_udp), \ 4852 FN(sk_release), \ 4853 FN(map_push_elem), \ 4854 FN(map_pop_elem), \ 4855 FN(map_peek_elem), \ 4856 FN(msg_push_data), \ 4857 FN(msg_pop_data), \ 4858 FN(rc_pointer_rel), \ 4859 FN(spin_lock), \ 4860 FN(spin_unlock), \ 4861 FN(sk_fullsock), \ 4862 FN(tcp_sock), \ 4863 FN(skb_ecn_set_ce), \ 4864 FN(get_listener_sock), \ 4865 FN(skc_lookup_tcp), \ 4866 FN(tcp_check_syncookie), \ 4867 FN(sysctl_get_name), \ 4868 FN(sysctl_get_current_value), \ 4869 FN(sysctl_get_new_value), \ 4870 FN(sysctl_set_new_value), \ 4871 FN(strtol), \ 4872 FN(strtoul), \ 4873 FN(sk_storage_get), \ 4874 FN(sk_storage_delete), \ 4875 FN(send_signal), \ 4876 FN(tcp_gen_syncookie), \ 4877 FN(skb_output), \ 4878 FN(probe_read_user), \ 4879 FN(probe_read_kernel), \ 4880 FN(probe_read_user_str), \ 4881 FN(probe_read_kernel_str), \ 4882 FN(tcp_send_ack), \ 4883 FN(send_signal_thread), \ 4884 FN(jiffies64), \ 4885 FN(read_branch_records), \ 4886 FN(get_ns_current_pid_tgid), \ 4887 FN(xdp_output), \ 4888 FN(get_netns_cookie), \ 4889 FN(get_current_ancestor_cgroup_id), \ 4890 FN(sk_assign), \ 4891 FN(ktime_get_boot_ns), \ 4892 FN(seq_printf), \ 4893 FN(seq_write), \ 4894 FN(sk_cgroup_id), \ 4895 FN(sk_ancestor_cgroup_id), \ 4896 FN(ringbuf_output), \ 4897 FN(ringbuf_reserve), \ 4898 FN(ringbuf_submit), \ 4899 FN(ringbuf_discard), \ 4900 FN(ringbuf_query), \ 4901 FN(csum_level), \ 4902 FN(skc_to_tcp6_sock), \ 4903 FN(skc_to_tcp_sock), \ 4904 FN(skc_to_tcp_timewait_sock), \ 4905 FN(skc_to_tcp_request_sock), \ 4906 FN(skc_to_udp6_sock), \ 4907 FN(get_task_stack), \ 4908 FN(load_hdr_opt), \ 4909 FN(store_hdr_opt), \ 4910 FN(reserve_hdr_opt), \ 4911 FN(inode_storage_get), \ 4912 FN(inode_storage_delete), \ 4913 FN(d_path), \ 4914 FN(copy_from_user), \ 4915 FN(snprintf_btf), \ 4916 FN(seq_printf_btf), \ 4917 FN(skb_cgroup_classid), \ 4918 FN(redirect_neigh), \ 4919 FN(per_cpu_ptr), \ 4920 FN(this_cpu_ptr), \ 4921 FN(redirect_peer), \ 4922 FN(task_storage_get), \ 4923 FN(task_storage_delete), \ 4924 FN(get_current_task_btf), \ 4925 FN(bprm_opts_set), \ 4926 FN(ktime_get_coarse_ns), \ 4927 FN(ima_inode_hash), \ 4928 FN(sock_from_file), \ 4929 FN(check_mtu), \ 4930 FN(for_each_map_elem), \ 4931 FN(snprintf), \ 4932 FN(sys_bpf), \ 4933 FN(btf_find_by_name_kind), \ 4934 FN(sys_close), \ 4935 /* */ 4936 4937 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 4938 * function eBPF program intends to call 4939 */ 4940 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 4941 enum bpf_func_id { 4942 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 4943 __BPF_FUNC_MAX_ID, 4944 }; 4945 #undef __BPF_ENUM_FN 4946 4947 /* All flags used by eBPF helper functions, placed here. */ 4948 4949 /* BPF_FUNC_skb_store_bytes flags. */ 4950 enum { 4951 BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 4952 BPF_F_INVALIDATE_HASH = (1ULL << 1), 4953 }; 4954 4955 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 4956 * First 4 bits are for passing the header field size. 4957 */ 4958 enum { 4959 BPF_F_HDR_FIELD_MASK = 0xfULL, 4960 }; 4961 4962 /* BPF_FUNC_l4_csum_replace flags. */ 4963 enum { 4964 BPF_F_PSEUDO_HDR = (1ULL << 4), 4965 BPF_F_MARK_MANGLED_0 = (1ULL << 5), 4966 BPF_F_MARK_ENFORCE = (1ULL << 6), 4967 }; 4968 4969 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 4970 enum { 4971 BPF_F_INGRESS = (1ULL << 0), 4972 }; 4973 4974 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 4975 enum { 4976 BPF_F_TUNINFO_IPV6 = (1ULL << 0), 4977 }; 4978 4979 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 4980 enum { 4981 BPF_F_SKIP_FIELD_MASK = 0xffULL, 4982 BPF_F_USER_STACK = (1ULL << 8), 4983 /* flags used by BPF_FUNC_get_stackid only. */ 4984 BPF_F_FAST_STACK_CMP = (1ULL << 9), 4985 BPF_F_REUSE_STACKID = (1ULL << 10), 4986 /* flags used by BPF_FUNC_get_stack only. */ 4987 BPF_F_USER_BUILD_ID = (1ULL << 11), 4988 }; 4989 4990 /* BPF_FUNC_skb_set_tunnel_key flags. */ 4991 enum { 4992 BPF_F_ZERO_CSUM_TX = (1ULL << 1), 4993 BPF_F_DONT_FRAGMENT = (1ULL << 2), 4994 BPF_F_SEQ_NUMBER = (1ULL << 3), 4995 }; 4996 4997 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 4998 * BPF_FUNC_perf_event_read_value flags. 4999 */ 5000 enum { 5001 BPF_F_INDEX_MASK = 0xffffffffULL, 5002 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 5003 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 5004 BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 5005 }; 5006 5007 /* Current network namespace */ 5008 enum { 5009 BPF_F_CURRENT_NETNS = (-1L), 5010 }; 5011 5012 /* BPF_FUNC_csum_level level values. */ 5013 enum { 5014 BPF_CSUM_LEVEL_QUERY, 5015 BPF_CSUM_LEVEL_INC, 5016 BPF_CSUM_LEVEL_DEC, 5017 BPF_CSUM_LEVEL_RESET, 5018 }; 5019 5020 /* BPF_FUNC_skb_adjust_room flags. */ 5021 enum { 5022 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 5023 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 5024 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 5025 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 5026 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 5027 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 5028 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6), 5029 }; 5030 5031 enum { 5032 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 5033 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 5034 }; 5035 5036 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 5037 BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 5038 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 5039 5040 /* BPF_FUNC_sysctl_get_name flags. */ 5041 enum { 5042 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 5043 }; 5044 5045 /* BPF_FUNC_<kernel_obj>_storage_get flags */ 5046 enum { 5047 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 5048 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 5049 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 5050 */ 5051 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 5052 }; 5053 5054 /* BPF_FUNC_read_branch_records flags. */ 5055 enum { 5056 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 5057 }; 5058 5059 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 5060 * BPF_FUNC_bpf_ringbuf_output flags. 5061 */ 5062 enum { 5063 BPF_RB_NO_WAKEUP = (1ULL << 0), 5064 BPF_RB_FORCE_WAKEUP = (1ULL << 1), 5065 }; 5066 5067 /* BPF_FUNC_bpf_ringbuf_query flags */ 5068 enum { 5069 BPF_RB_AVAIL_DATA = 0, 5070 BPF_RB_RING_SIZE = 1, 5071 BPF_RB_CONS_POS = 2, 5072 BPF_RB_PROD_POS = 3, 5073 }; 5074 5075 /* BPF ring buffer constants */ 5076 enum { 5077 BPF_RINGBUF_BUSY_BIT = (1U << 31), 5078 BPF_RINGBUF_DISCARD_BIT = (1U << 30), 5079 BPF_RINGBUF_HDR_SZ = 8, 5080 }; 5081 5082 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 5083 enum { 5084 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 5085 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 5086 }; 5087 5088 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 5089 enum bpf_adj_room_mode { 5090 BPF_ADJ_ROOM_NET, 5091 BPF_ADJ_ROOM_MAC, 5092 }; 5093 5094 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 5095 enum bpf_hdr_start_off { 5096 BPF_HDR_START_MAC, 5097 BPF_HDR_START_NET, 5098 }; 5099 5100 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 5101 enum bpf_lwt_encap_mode { 5102 BPF_LWT_ENCAP_SEG6, 5103 BPF_LWT_ENCAP_SEG6_INLINE, 5104 BPF_LWT_ENCAP_IP, 5105 }; 5106 5107 /* Flags for bpf_bprm_opts_set helper */ 5108 enum { 5109 BPF_F_BPRM_SECUREEXEC = (1ULL << 0), 5110 }; 5111 5112 #define __bpf_md_ptr(type, name) \ 5113 union { \ 5114 type name; \ 5115 __u64 :64; \ 5116 } __attribute__((aligned(8))) 5117 5118 /* user accessible mirror of in-kernel sk_buff. 5119 * new fields can only be added to the end of this structure 5120 */ 5121 struct __sk_buff { 5122 __u32 len; 5123 __u32 pkt_type; 5124 __u32 mark; 5125 __u32 queue_mapping; 5126 __u32 protocol; 5127 __u32 vlan_present; 5128 __u32 vlan_tci; 5129 __u32 vlan_proto; 5130 __u32 priority; 5131 __u32 ingress_ifindex; 5132 __u32 ifindex; 5133 __u32 tc_index; 5134 __u32 cb[5]; 5135 __u32 hash; 5136 __u32 tc_classid; 5137 __u32 data; 5138 __u32 data_end; 5139 __u32 napi_id; 5140 5141 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 5142 __u32 family; 5143 __u32 remote_ip4; /* Stored in network byte order */ 5144 __u32 local_ip4; /* Stored in network byte order */ 5145 __u32 remote_ip6[4]; /* Stored in network byte order */ 5146 __u32 local_ip6[4]; /* Stored in network byte order */ 5147 __u32 remote_port; /* Stored in network byte order */ 5148 __u32 local_port; /* stored in host byte order */ 5149 /* ... here. */ 5150 5151 __u32 data_meta; 5152 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 5153 __u64 tstamp; 5154 __u32 wire_len; 5155 __u32 gso_segs; 5156 __bpf_md_ptr(struct bpf_sock *, sk); 5157 __u32 gso_size; 5158 }; 5159 5160 struct bpf_tunnel_key { 5161 __u32 tunnel_id; 5162 union { 5163 __u32 remote_ipv4; 5164 __u32 remote_ipv6[4]; 5165 }; 5166 __u8 tunnel_tos; 5167 __u8 tunnel_ttl; 5168 __u16 tunnel_ext; /* Padding, future use. */ 5169 __u32 tunnel_label; 5170 }; 5171 5172 /* user accessible mirror of in-kernel xfrm_state. 5173 * new fields can only be added to the end of this structure 5174 */ 5175 struct bpf_xfrm_state { 5176 __u32 reqid; 5177 __u32 spi; /* Stored in network byte order */ 5178 __u16 family; 5179 __u16 ext; /* Padding, future use. */ 5180 union { 5181 __u32 remote_ipv4; /* Stored in network byte order */ 5182 __u32 remote_ipv6[4]; /* Stored in network byte order */ 5183 }; 5184 }; 5185 5186 /* Generic BPF return codes which all BPF program types may support. 5187 * The values are binary compatible with their TC_ACT_* counter-part to 5188 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 5189 * programs. 5190 * 5191 * XDP is handled seprately, see XDP_*. 5192 */ 5193 enum bpf_ret_code { 5194 BPF_OK = 0, 5195 /* 1 reserved */ 5196 BPF_DROP = 2, 5197 /* 3-6 reserved */ 5198 BPF_REDIRECT = 7, 5199 /* >127 are reserved for prog type specific return codes. 5200 * 5201 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 5202 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 5203 * changed and should be routed based on its new L3 header. 5204 * (This is an L3 redirect, as opposed to L2 redirect 5205 * represented by BPF_REDIRECT above). 5206 */ 5207 BPF_LWT_REROUTE = 128, 5208 }; 5209 5210 struct bpf_sock { 5211 __u32 bound_dev_if; 5212 __u32 family; 5213 __u32 type; 5214 __u32 protocol; 5215 __u32 mark; 5216 __u32 priority; 5217 /* IP address also allows 1 and 2 bytes access */ 5218 __u32 src_ip4; 5219 __u32 src_ip6[4]; 5220 __u32 src_port; /* host byte order */ 5221 __u32 dst_port; /* network byte order */ 5222 __u32 dst_ip4; 5223 __u32 dst_ip6[4]; 5224 __u32 state; 5225 __s32 rx_queue_mapping; 5226 }; 5227 5228 struct bpf_tcp_sock { 5229 __u32 snd_cwnd; /* Sending congestion window */ 5230 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 5231 __u32 rtt_min; 5232 __u32 snd_ssthresh; /* Slow start size threshold */ 5233 __u32 rcv_nxt; /* What we want to receive next */ 5234 __u32 snd_nxt; /* Next sequence we send */ 5235 __u32 snd_una; /* First byte we want an ack for */ 5236 __u32 mss_cache; /* Cached effective mss, not including SACKS */ 5237 __u32 ecn_flags; /* ECN status bits. */ 5238 __u32 rate_delivered; /* saved rate sample: packets delivered */ 5239 __u32 rate_interval_us; /* saved rate sample: time elapsed */ 5240 __u32 packets_out; /* Packets which are "in flight" */ 5241 __u32 retrans_out; /* Retransmitted packets out */ 5242 __u32 total_retrans; /* Total retransmits for entire connection */ 5243 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 5244 * total number of segments in. 5245 */ 5246 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 5247 * total number of data segments in. 5248 */ 5249 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 5250 * The total number of segments sent. 5251 */ 5252 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 5253 * total number of data segments sent. 5254 */ 5255 __u32 lost_out; /* Lost packets */ 5256 __u32 sacked_out; /* SACK'd packets */ 5257 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 5258 * sum(delta(rcv_nxt)), or how many bytes 5259 * were acked. 5260 */ 5261 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 5262 * sum(delta(snd_una)), or how many bytes 5263 * were acked. 5264 */ 5265 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 5266 * total number of DSACK blocks received 5267 */ 5268 __u32 delivered; /* Total data packets delivered incl. rexmits */ 5269 __u32 delivered_ce; /* Like the above but only ECE marked packets */ 5270 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 5271 }; 5272 5273 struct bpf_sock_tuple { 5274 union { 5275 struct { 5276 __be32 saddr; 5277 __be32 daddr; 5278 __be16 sport; 5279 __be16 dport; 5280 } ipv4; 5281 struct { 5282 __be32 saddr[4]; 5283 __be32 daddr[4]; 5284 __be16 sport; 5285 __be16 dport; 5286 } ipv6; 5287 }; 5288 }; 5289 5290 struct bpf_xdp_sock { 5291 __u32 queue_id; 5292 }; 5293 5294 #define XDP_PACKET_HEADROOM 256 5295 5296 /* User return codes for XDP prog type. 5297 * A valid XDP program must return one of these defined values. All other 5298 * return codes are reserved for future use. Unknown return codes will 5299 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 5300 */ 5301 enum xdp_action { 5302 XDP_ABORTED = 0, 5303 XDP_DROP, 5304 XDP_PASS, 5305 XDP_TX, 5306 XDP_REDIRECT, 5307 }; 5308 5309 /* user accessible metadata for XDP packet hook 5310 * new fields must be added to the end of this structure 5311 */ 5312 struct xdp_md { 5313 __u32 data; 5314 __u32 data_end; 5315 __u32 data_meta; 5316 /* Below access go through struct xdp_rxq_info */ 5317 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 5318 __u32 rx_queue_index; /* rxq->queue_index */ 5319 5320 __u32 egress_ifindex; /* txq->dev->ifindex */ 5321 }; 5322 5323 /* DEVMAP map-value layout 5324 * 5325 * The struct data-layout of map-value is a configuration interface. 5326 * New members can only be added to the end of this structure. 5327 */ 5328 struct bpf_devmap_val { 5329 __u32 ifindex; /* device index */ 5330 union { 5331 int fd; /* prog fd on map write */ 5332 __u32 id; /* prog id on map read */ 5333 } bpf_prog; 5334 }; 5335 5336 /* CPUMAP map-value layout 5337 * 5338 * The struct data-layout of map-value is a configuration interface. 5339 * New members can only be added to the end of this structure. 5340 */ 5341 struct bpf_cpumap_val { 5342 __u32 qsize; /* queue size to remote target CPU */ 5343 union { 5344 int fd; /* prog fd on map write */ 5345 __u32 id; /* prog id on map read */ 5346 } bpf_prog; 5347 }; 5348 5349 enum sk_action { 5350 SK_DROP = 0, 5351 SK_PASS, 5352 }; 5353 5354 /* user accessible metadata for SK_MSG packet hook, new fields must 5355 * be added to the end of this structure 5356 */ 5357 struct sk_msg_md { 5358 __bpf_md_ptr(void *, data); 5359 __bpf_md_ptr(void *, data_end); 5360 5361 __u32 family; 5362 __u32 remote_ip4; /* Stored in network byte order */ 5363 __u32 local_ip4; /* Stored in network byte order */ 5364 __u32 remote_ip6[4]; /* Stored in network byte order */ 5365 __u32 local_ip6[4]; /* Stored in network byte order */ 5366 __u32 remote_port; /* Stored in network byte order */ 5367 __u32 local_port; /* stored in host byte order */ 5368 __u32 size; /* Total size of sk_msg */ 5369 5370 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 5371 }; 5372 5373 struct sk_reuseport_md { 5374 /* 5375 * Start of directly accessible data. It begins from 5376 * the tcp/udp header. 5377 */ 5378 __bpf_md_ptr(void *, data); 5379 /* End of directly accessible data */ 5380 __bpf_md_ptr(void *, data_end); 5381 /* 5382 * Total length of packet (starting from the tcp/udp header). 5383 * Note that the directly accessible bytes (data_end - data) 5384 * could be less than this "len". Those bytes could be 5385 * indirectly read by a helper "bpf_skb_load_bytes()". 5386 */ 5387 __u32 len; 5388 /* 5389 * Eth protocol in the mac header (network byte order). e.g. 5390 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 5391 */ 5392 __u32 eth_protocol; 5393 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 5394 __u32 bind_inany; /* Is sock bound to an INANY address? */ 5395 __u32 hash; /* A hash of the packet 4 tuples */ 5396 }; 5397 5398 #define BPF_TAG_SIZE 8 5399 5400 struct bpf_prog_info { 5401 __u32 type; 5402 __u32 id; 5403 __u8 tag[BPF_TAG_SIZE]; 5404 __u32 jited_prog_len; 5405 __u32 xlated_prog_len; 5406 __aligned_u64 jited_prog_insns; 5407 __aligned_u64 xlated_prog_insns; 5408 __u64 load_time; /* ns since boottime */ 5409 __u32 created_by_uid; 5410 __u32 nr_map_ids; 5411 __aligned_u64 map_ids; 5412 char name[BPF_OBJ_NAME_LEN]; 5413 __u32 ifindex; 5414 __u32 gpl_compatible:1; 5415 __u32 :31; /* alignment pad */ 5416 __u64 netns_dev; 5417 __u64 netns_ino; 5418 __u32 nr_jited_ksyms; 5419 __u32 nr_jited_func_lens; 5420 __aligned_u64 jited_ksyms; 5421 __aligned_u64 jited_func_lens; 5422 __u32 btf_id; 5423 __u32 func_info_rec_size; 5424 __aligned_u64 func_info; 5425 __u32 nr_func_info; 5426 __u32 nr_line_info; 5427 __aligned_u64 line_info; 5428 __aligned_u64 jited_line_info; 5429 __u32 nr_jited_line_info; 5430 __u32 line_info_rec_size; 5431 __u32 jited_line_info_rec_size; 5432 __u32 nr_prog_tags; 5433 __aligned_u64 prog_tags; 5434 __u64 run_time_ns; 5435 __u64 run_cnt; 5436 __u64 recursion_misses; 5437 } __attribute__((aligned(8))); 5438 5439 struct bpf_map_info { 5440 __u32 type; 5441 __u32 id; 5442 __u32 key_size; 5443 __u32 value_size; 5444 __u32 max_entries; 5445 __u32 map_flags; 5446 char name[BPF_OBJ_NAME_LEN]; 5447 __u32 ifindex; 5448 __u32 btf_vmlinux_value_type_id; 5449 __u64 netns_dev; 5450 __u64 netns_ino; 5451 __u32 btf_id; 5452 __u32 btf_key_type_id; 5453 __u32 btf_value_type_id; 5454 } __attribute__((aligned(8))); 5455 5456 struct bpf_btf_info { 5457 __aligned_u64 btf; 5458 __u32 btf_size; 5459 __u32 id; 5460 __aligned_u64 name; 5461 __u32 name_len; 5462 __u32 kernel_btf; 5463 } __attribute__((aligned(8))); 5464 5465 struct bpf_link_info { 5466 __u32 type; 5467 __u32 id; 5468 __u32 prog_id; 5469 union { 5470 struct { 5471 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 5472 __u32 tp_name_len; /* in/out: tp_name buffer len */ 5473 } raw_tracepoint; 5474 struct { 5475 __u32 attach_type; 5476 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */ 5477 __u32 target_btf_id; /* BTF type id inside the object */ 5478 } tracing; 5479 struct { 5480 __u64 cgroup_id; 5481 __u32 attach_type; 5482 } cgroup; 5483 struct { 5484 __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 5485 __u32 target_name_len; /* in/out: target_name buffer len */ 5486 union { 5487 struct { 5488 __u32 map_id; 5489 } map; 5490 }; 5491 } iter; 5492 struct { 5493 __u32 netns_ino; 5494 __u32 attach_type; 5495 } netns; 5496 struct { 5497 __u32 ifindex; 5498 } xdp; 5499 }; 5500 } __attribute__((aligned(8))); 5501 5502 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 5503 * by user and intended to be used by socket (e.g. to bind to, depends on 5504 * attach type). 5505 */ 5506 struct bpf_sock_addr { 5507 __u32 user_family; /* Allows 4-byte read, but no write. */ 5508 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5509 * Stored in network byte order. 5510 */ 5511 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5512 * Stored in network byte order. 5513 */ 5514 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 5515 * Stored in network byte order 5516 */ 5517 __u32 family; /* Allows 4-byte read, but no write */ 5518 __u32 type; /* Allows 4-byte read, but no write */ 5519 __u32 protocol; /* Allows 4-byte read, but no write */ 5520 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 5521 * Stored in network byte order. 5522 */ 5523 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 5524 * Stored in network byte order. 5525 */ 5526 __bpf_md_ptr(struct bpf_sock *, sk); 5527 }; 5528 5529 /* User bpf_sock_ops struct to access socket values and specify request ops 5530 * and their replies. 5531 * Some of this fields are in network (bigendian) byte order and may need 5532 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 5533 * New fields can only be added at the end of this structure 5534 */ 5535 struct bpf_sock_ops { 5536 __u32 op; 5537 union { 5538 __u32 args[4]; /* Optionally passed to bpf program */ 5539 __u32 reply; /* Returned by bpf program */ 5540 __u32 replylong[4]; /* Optionally returned by bpf prog */ 5541 }; 5542 __u32 family; 5543 __u32 remote_ip4; /* Stored in network byte order */ 5544 __u32 local_ip4; /* Stored in network byte order */ 5545 __u32 remote_ip6[4]; /* Stored in network byte order */ 5546 __u32 local_ip6[4]; /* Stored in network byte order */ 5547 __u32 remote_port; /* Stored in network byte order */ 5548 __u32 local_port; /* stored in host byte order */ 5549 __u32 is_fullsock; /* Some TCP fields are only valid if 5550 * there is a full socket. If not, the 5551 * fields read as zero. 5552 */ 5553 __u32 snd_cwnd; 5554 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 5555 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 5556 __u32 state; 5557 __u32 rtt_min; 5558 __u32 snd_ssthresh; 5559 __u32 rcv_nxt; 5560 __u32 snd_nxt; 5561 __u32 snd_una; 5562 __u32 mss_cache; 5563 __u32 ecn_flags; 5564 __u32 rate_delivered; 5565 __u32 rate_interval_us; 5566 __u32 packets_out; 5567 __u32 retrans_out; 5568 __u32 total_retrans; 5569 __u32 segs_in; 5570 __u32 data_segs_in; 5571 __u32 segs_out; 5572 __u32 data_segs_out; 5573 __u32 lost_out; 5574 __u32 sacked_out; 5575 __u32 sk_txhash; 5576 __u64 bytes_received; 5577 __u64 bytes_acked; 5578 __bpf_md_ptr(struct bpf_sock *, sk); 5579 /* [skb_data, skb_data_end) covers the whole TCP header. 5580 * 5581 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 5582 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 5583 * header has not been written. 5584 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 5585 * been written so far. 5586 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 5587 * the 3WHS. 5588 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 5589 * the 3WHS. 5590 * 5591 * bpf_load_hdr_opt() can also be used to read a particular option. 5592 */ 5593 __bpf_md_ptr(void *, skb_data); 5594 __bpf_md_ptr(void *, skb_data_end); 5595 __u32 skb_len; /* The total length of a packet. 5596 * It includes the header, options, 5597 * and payload. 5598 */ 5599 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 5600 * an easy way to check for tcp_flags 5601 * without parsing skb_data. 5602 * 5603 * In particular, the skb_tcp_flags 5604 * will still be available in 5605 * BPF_SOCK_OPS_HDR_OPT_LEN even though 5606 * the outgoing header has not 5607 * been written yet. 5608 */ 5609 }; 5610 5611 /* Definitions for bpf_sock_ops_cb_flags */ 5612 enum { 5613 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 5614 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 5615 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 5616 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 5617 /* Call bpf for all received TCP headers. The bpf prog will be 5618 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5619 * 5620 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5621 * for the header option related helpers that will be useful 5622 * to the bpf programs. 5623 * 5624 * It could be used at the client/active side (i.e. connect() side) 5625 * when the server told it that the server was in syncookie 5626 * mode and required the active side to resend the bpf-written 5627 * options. The active side can keep writing the bpf-options until 5628 * it received a valid packet from the server side to confirm 5629 * the earlier packet (and options) has been received. The later 5630 * example patch is using it like this at the active side when the 5631 * server is in syncookie mode. 5632 * 5633 * The bpf prog will usually turn this off in the common cases. 5634 */ 5635 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 5636 /* Call bpf when kernel has received a header option that 5637 * the kernel cannot handle. The bpf prog will be called under 5638 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 5639 * 5640 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 5641 * for the header option related helpers that will be useful 5642 * to the bpf programs. 5643 */ 5644 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 5645 /* Call bpf when the kernel is writing header options for the 5646 * outgoing packet. The bpf prog will first be called 5647 * to reserve space in a skb under 5648 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 5649 * the bpf prog will be called to write the header option(s) 5650 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5651 * 5652 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 5653 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 5654 * related helpers that will be useful to the bpf programs. 5655 * 5656 * The kernel gets its chance to reserve space and write 5657 * options first before the BPF program does. 5658 */ 5659 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 5660 /* Mask of all currently supported cb flags */ 5661 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 5662 }; 5663 5664 /* List of known BPF sock_ops operators. 5665 * New entries can only be added at the end 5666 */ 5667 enum { 5668 BPF_SOCK_OPS_VOID, 5669 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 5670 * -1 if default value should be used 5671 */ 5672 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 5673 * window (in packets) or -1 if default 5674 * value should be used 5675 */ 5676 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 5677 * active connection is initialized 5678 */ 5679 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 5680 * active connection is 5681 * established 5682 */ 5683 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 5684 * passive connection is 5685 * established 5686 */ 5687 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 5688 * needs ECN 5689 */ 5690 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 5691 * based on the path and may be 5692 * dependent on the congestion control 5693 * algorithm. In general it indicates 5694 * a congestion threshold. RTTs above 5695 * this indicate congestion 5696 */ 5697 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 5698 * Arg1: value of icsk_retransmits 5699 * Arg2: value of icsk_rto 5700 * Arg3: whether RTO has expired 5701 */ 5702 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 5703 * Arg1: sequence number of 1st byte 5704 * Arg2: # segments 5705 * Arg3: return value of 5706 * tcp_transmit_skb (0 => success) 5707 */ 5708 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 5709 * Arg1: old_state 5710 * Arg2: new_state 5711 */ 5712 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 5713 * socket transition to LISTEN state. 5714 */ 5715 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 5716 */ 5717 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 5718 * It will be called to handle 5719 * the packets received at 5720 * an already established 5721 * connection. 5722 * 5723 * sock_ops->skb_data: 5724 * Referring to the received skb. 5725 * It covers the TCP header only. 5726 * 5727 * bpf_load_hdr_opt() can also 5728 * be used to search for a 5729 * particular option. 5730 */ 5731 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 5732 * header option later in 5733 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5734 * Arg1: bool want_cookie. (in 5735 * writing SYNACK only) 5736 * 5737 * sock_ops->skb_data: 5738 * Not available because no header has 5739 * been written yet. 5740 * 5741 * sock_ops->skb_tcp_flags: 5742 * The tcp_flags of the 5743 * outgoing skb. (e.g. SYN, ACK, FIN). 5744 * 5745 * bpf_reserve_hdr_opt() should 5746 * be used to reserve space. 5747 */ 5748 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 5749 * Arg1: bool want_cookie. (in 5750 * writing SYNACK only) 5751 * 5752 * sock_ops->skb_data: 5753 * Referring to the outgoing skb. 5754 * It covers the TCP header 5755 * that has already been written 5756 * by the kernel and the 5757 * earlier bpf-progs. 5758 * 5759 * sock_ops->skb_tcp_flags: 5760 * The tcp_flags of the outgoing 5761 * skb. (e.g. SYN, ACK, FIN). 5762 * 5763 * bpf_store_hdr_opt() should 5764 * be used to write the 5765 * option. 5766 * 5767 * bpf_load_hdr_opt() can also 5768 * be used to search for a 5769 * particular option that 5770 * has already been written 5771 * by the kernel or the 5772 * earlier bpf-progs. 5773 */ 5774 }; 5775 5776 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 5777 * changes between the TCP and BPF versions. Ideally this should never happen. 5778 * If it does, we need to add code to convert them before calling 5779 * the BPF sock_ops function. 5780 */ 5781 enum { 5782 BPF_TCP_ESTABLISHED = 1, 5783 BPF_TCP_SYN_SENT, 5784 BPF_TCP_SYN_RECV, 5785 BPF_TCP_FIN_WAIT1, 5786 BPF_TCP_FIN_WAIT2, 5787 BPF_TCP_TIME_WAIT, 5788 BPF_TCP_CLOSE, 5789 BPF_TCP_CLOSE_WAIT, 5790 BPF_TCP_LAST_ACK, 5791 BPF_TCP_LISTEN, 5792 BPF_TCP_CLOSING, /* Now a valid state */ 5793 BPF_TCP_NEW_SYN_RECV, 5794 5795 BPF_TCP_MAX_STATES /* Leave at the end! */ 5796 }; 5797 5798 enum { 5799 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 5800 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 5801 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 5802 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 5803 /* Copy the SYN pkt to optval 5804 * 5805 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 5806 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 5807 * to only getting from the saved_syn. It can either get the 5808 * syn packet from: 5809 * 5810 * 1. the just-received SYN packet (only available when writing the 5811 * SYNACK). It will be useful when it is not necessary to 5812 * save the SYN packet for latter use. It is also the only way 5813 * to get the SYN during syncookie mode because the syn 5814 * packet cannot be saved during syncookie. 5815 * 5816 * OR 5817 * 5818 * 2. the earlier saved syn which was done by 5819 * bpf_setsockopt(TCP_SAVE_SYN). 5820 * 5821 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 5822 * SYN packet is obtained. 5823 * 5824 * If the bpf-prog does not need the IP[46] header, the 5825 * bpf-prog can avoid parsing the IP header by using 5826 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 5827 * IP[46] and TCP header by using TCP_BPF_SYN_IP. 5828 * 5829 * >0: Total number of bytes copied 5830 * -ENOSPC: Not enough space in optval. Only optlen number of 5831 * bytes is copied. 5832 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 5833 * is not saved by setsockopt(TCP_SAVE_SYN). 5834 */ 5835 TCP_BPF_SYN = 1005, /* Copy the TCP header */ 5836 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 5837 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 5838 }; 5839 5840 enum { 5841 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 5842 }; 5843 5844 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 5845 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 5846 */ 5847 enum { 5848 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 5849 * total option spaces 5850 * required for an established 5851 * sk in order to calculate the 5852 * MSS. No skb is actually 5853 * sent. 5854 */ 5855 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 5856 * when sending a SYN. 5857 */ 5858 }; 5859 5860 struct bpf_perf_event_value { 5861 __u64 counter; 5862 __u64 enabled; 5863 __u64 running; 5864 }; 5865 5866 enum { 5867 BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 5868 BPF_DEVCG_ACC_READ = (1ULL << 1), 5869 BPF_DEVCG_ACC_WRITE = (1ULL << 2), 5870 }; 5871 5872 enum { 5873 BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 5874 BPF_DEVCG_DEV_CHAR = (1ULL << 1), 5875 }; 5876 5877 struct bpf_cgroup_dev_ctx { 5878 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 5879 __u32 access_type; 5880 __u32 major; 5881 __u32 minor; 5882 }; 5883 5884 struct bpf_raw_tracepoint_args { 5885 __u64 args[0]; 5886 }; 5887 5888 /* DIRECT: Skip the FIB rules and go to FIB table associated with device 5889 * OUTPUT: Do lookup from egress perspective; default is ingress 5890 */ 5891 enum { 5892 BPF_FIB_LOOKUP_DIRECT = (1U << 0), 5893 BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 5894 }; 5895 5896 enum { 5897 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 5898 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 5899 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 5900 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 5901 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 5902 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 5903 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 5904 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 5905 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 5906 }; 5907 5908 struct bpf_fib_lookup { 5909 /* input: network family for lookup (AF_INET, AF_INET6) 5910 * output: network family of egress nexthop 5911 */ 5912 __u8 family; 5913 5914 /* set if lookup is to consider L4 data - e.g., FIB rules */ 5915 __u8 l4_protocol; 5916 __be16 sport; 5917 __be16 dport; 5918 5919 union { /* used for MTU check */ 5920 /* input to lookup */ 5921 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */ 5922 5923 /* output: MTU value */ 5924 __u16 mtu_result; 5925 }; 5926 /* input: L3 device index for lookup 5927 * output: device index from FIB lookup 5928 */ 5929 __u32 ifindex; 5930 5931 union { 5932 /* inputs to lookup */ 5933 __u8 tos; /* AF_INET */ 5934 __be32 flowinfo; /* AF_INET6, flow_label + priority */ 5935 5936 /* output: metric of fib result (IPv4/IPv6 only) */ 5937 __u32 rt_metric; 5938 }; 5939 5940 union { 5941 __be32 ipv4_src; 5942 __u32 ipv6_src[4]; /* in6_addr; network order */ 5943 }; 5944 5945 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 5946 * network header. output: bpf_fib_lookup sets to gateway address 5947 * if FIB lookup returns gateway route 5948 */ 5949 union { 5950 __be32 ipv4_dst; 5951 __u32 ipv6_dst[4]; /* in6_addr; network order */ 5952 }; 5953 5954 /* output */ 5955 __be16 h_vlan_proto; 5956 __be16 h_vlan_TCI; 5957 __u8 smac[6]; /* ETH_ALEN */ 5958 __u8 dmac[6]; /* ETH_ALEN */ 5959 }; 5960 5961 struct bpf_redir_neigh { 5962 /* network family for lookup (AF_INET, AF_INET6) */ 5963 __u32 nh_family; 5964 /* network address of nexthop; skips fib lookup to find gateway */ 5965 union { 5966 __be32 ipv4_nh; 5967 __u32 ipv6_nh[4]; /* in6_addr; network order */ 5968 }; 5969 }; 5970 5971 /* bpf_check_mtu flags*/ 5972 enum bpf_check_mtu_flags { 5973 BPF_MTU_CHK_SEGS = (1U << 0), 5974 }; 5975 5976 enum bpf_check_mtu_ret { 5977 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */ 5978 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 5979 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */ 5980 }; 5981 5982 enum bpf_task_fd_type { 5983 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 5984 BPF_FD_TYPE_TRACEPOINT, /* tp name */ 5985 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 5986 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 5987 BPF_FD_TYPE_UPROBE, /* filename + offset */ 5988 BPF_FD_TYPE_URETPROBE, /* filename + offset */ 5989 }; 5990 5991 enum { 5992 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 5993 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 5994 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 5995 }; 5996 5997 struct bpf_flow_keys { 5998 __u16 nhoff; 5999 __u16 thoff; 6000 __u16 addr_proto; /* ETH_P_* of valid addrs */ 6001 __u8 is_frag; 6002 __u8 is_first_frag; 6003 __u8 is_encap; 6004 __u8 ip_proto; 6005 __be16 n_proto; 6006 __be16 sport; 6007 __be16 dport; 6008 union { 6009 struct { 6010 __be32 ipv4_src; 6011 __be32 ipv4_dst; 6012 }; 6013 struct { 6014 __u32 ipv6_src[4]; /* in6_addr; network order */ 6015 __u32 ipv6_dst[4]; /* in6_addr; network order */ 6016 }; 6017 }; 6018 __u32 flags; 6019 __be32 flow_label; 6020 }; 6021 6022 struct bpf_func_info { 6023 __u32 insn_off; 6024 __u32 type_id; 6025 }; 6026 6027 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 6028 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 6029 6030 struct bpf_line_info { 6031 __u32 insn_off; 6032 __u32 file_name_off; 6033 __u32 line_off; 6034 __u32 line_col; 6035 }; 6036 6037 struct bpf_spin_lock { 6038 __u32 val; 6039 }; 6040 6041 struct bpf_sysctl { 6042 __u32 write; /* Sysctl is being read (= 0) or written (= 1). 6043 * Allows 1,2,4-byte read, but no write. 6044 */ 6045 __u32 file_pos; /* Sysctl file position to read from, write to. 6046 * Allows 1,2,4-byte read an 4-byte write. 6047 */ 6048 }; 6049 6050 struct bpf_sockopt { 6051 __bpf_md_ptr(struct bpf_sock *, sk); 6052 __bpf_md_ptr(void *, optval); 6053 __bpf_md_ptr(void *, optval_end); 6054 6055 __s32 level; 6056 __s32 optname; 6057 __s32 optlen; 6058 __s32 retval; 6059 }; 6060 6061 struct bpf_pidns_info { 6062 __u32 pid; 6063 __u32 tgid; 6064 }; 6065 6066 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 6067 struct bpf_sk_lookup { 6068 union { 6069 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 6070 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */ 6071 }; 6072 6073 __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 6074 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 6075 __u32 remote_ip4; /* Network byte order */ 6076 __u32 remote_ip6[4]; /* Network byte order */ 6077 __u32 remote_port; /* Network byte order */ 6078 __u32 local_ip4; /* Network byte order */ 6079 __u32 local_ip6[4]; /* Network byte order */ 6080 __u32 local_port; /* Host byte order */ 6081 }; 6082 6083 /* 6084 * struct btf_ptr is used for typed pointer representation; the 6085 * type id is used to render the pointer data as the appropriate type 6086 * via the bpf_snprintf_btf() helper described above. A flags field - 6087 * potentially to specify additional details about the BTF pointer 6088 * (rather than its mode of display) - is included for future use. 6089 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 6090 */ 6091 struct btf_ptr { 6092 void *ptr; 6093 __u32 type_id; 6094 __u32 flags; /* BTF ptr flags; unused at present. */ 6095 }; 6096 6097 /* 6098 * Flags to control bpf_snprintf_btf() behaviour. 6099 * - BTF_F_COMPACT: no formatting around type information 6100 * - BTF_F_NONAME: no struct/union member names/types 6101 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 6102 * equivalent to %px. 6103 * - BTF_F_ZERO: show zero-valued struct/union members; they 6104 * are not displayed by default 6105 */ 6106 enum { 6107 BTF_F_COMPACT = (1ULL << 0), 6108 BTF_F_NONAME = (1ULL << 1), 6109 BTF_F_PTR_RAW = (1ULL << 2), 6110 BTF_F_ZERO = (1ULL << 3), 6111 }; 6112 6113 #endif /* _UAPI__LINUX_BPF_H__ */ 6114