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