logo

oasis-root

Compiled tree of Oasis Linux based on own branch at <https://hacktivis.me/git/oasis/> git clone https://anongit.hacktivis.me/git/oasis-root.git

bpf.h (280212B)


  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 __LINUX_BPF_H__
  9. #define __LINUX_BPF_H__
  10. #include <linux/types.h>
  11. #include <linux/bpf_common.h>
  12. /* Extended instruction set based on top of classic BPF */
  13. /* instruction classes */
  14. #define BPF_JMP32 0x06 /* jmp mode in word width */
  15. #define BPF_ALU64 0x07 /* alu mode in double word width */
  16. /* ld/ldx fields */
  17. #define BPF_DW 0x18 /* double word (64-bit) */
  18. #define BPF_MEMSX 0x80 /* load with sign extension */
  19. #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
  20. #define BPF_XADD 0xc0 /* exclusive add - legacy name */
  21. /* alu/jmp fields */
  22. #define BPF_MOV 0xb0 /* mov reg to reg */
  23. #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
  24. /* change endianness of a register */
  25. #define BPF_END 0xd0 /* flags for endianness conversion: */
  26. #define BPF_TO_LE 0x00 /* convert to little-endian */
  27. #define BPF_TO_BE 0x08 /* convert to big-endian */
  28. #define BPF_FROM_LE BPF_TO_LE
  29. #define BPF_FROM_BE BPF_TO_BE
  30. /* jmp encodings */
  31. #define BPF_JNE 0x50 /* jump != */
  32. #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
  33. #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
  34. #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
  35. #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
  36. #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
  37. #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
  38. #define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */
  39. #define BPF_CALL 0x80 /* function call */
  40. #define BPF_EXIT 0x90 /* function return */
  41. /* atomic op type fields (stored in immediate) */
  42. #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
  43. #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
  44. #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
  45. enum bpf_cond_pseudo_jmp {
  46. BPF_MAY_GOTO = 0,
  47. };
  48. /* Register numbers */
  49. enum {
  50. BPF_REG_0 = 0,
  51. BPF_REG_1,
  52. BPF_REG_2,
  53. BPF_REG_3,
  54. BPF_REG_4,
  55. BPF_REG_5,
  56. BPF_REG_6,
  57. BPF_REG_7,
  58. BPF_REG_8,
  59. BPF_REG_9,
  60. BPF_REG_10,
  61. __MAX_BPF_REG,
  62. };
  63. /* BPF has 10 general purpose 64-bit registers and stack frame. */
  64. #define MAX_BPF_REG __MAX_BPF_REG
  65. struct bpf_insn {
  66. __u8 code; /* opcode */
  67. __u8 dst_reg:4; /* dest register */
  68. __u8 src_reg:4; /* source register */
  69. __s16 off; /* signed offset */
  70. __s32 imm; /* signed immediate constant */
  71. };
  72. /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
  73. * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
  74. * the trailing flexible array member) instead.
  75. */
  76. struct bpf_lpm_trie_key {
  77. __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
  78. __u8 data[0]; /* Arbitrary size */
  79. };
  80. /* Header for bpf_lpm_trie_key structs */
  81. struct bpf_lpm_trie_key_hdr {
  82. __u32 prefixlen;
  83. };
  84. /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
  85. struct bpf_lpm_trie_key_u8 {
  86. union {
  87. struct bpf_lpm_trie_key_hdr hdr;
  88. __u32 prefixlen;
  89. };
  90. __u8 data[]; /* Arbitrary size */
  91. };
  92. struct bpf_cgroup_storage_key {
  93. __u64 cgroup_inode_id; /* cgroup inode id */
  94. __u32 attach_type; /* program attach type (enum bpf_attach_type) */
  95. };
  96. enum bpf_cgroup_iter_order {
  97. BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
  98. BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */
  99. BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */
  100. BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */
  101. BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */
  102. };
  103. union bpf_iter_link_info {
  104. struct {
  105. __u32 map_fd;
  106. } map;
  107. struct {
  108. enum bpf_cgroup_iter_order order;
  109. /* At most one of cgroup_fd and cgroup_id can be non-zero. If
  110. * both are zero, the walk starts from the default cgroup v2
  111. * root. For walking v1 hierarchy, one should always explicitly
  112. * specify cgroup_fd.
  113. */
  114. __u32 cgroup_fd;
  115. __u64 cgroup_id;
  116. } cgroup;
  117. /* Parameters of task iterators. */
  118. struct {
  119. __u32 tid;
  120. __u32 pid;
  121. __u32 pid_fd;
  122. } task;
  123. };
  124. /* BPF syscall commands, see bpf(2) man-page for more details. */
  125. /**
  126. * DOC: eBPF Syscall Preamble
  127. *
  128. * The operation to be performed by the **bpf**\ () system call is determined
  129. * by the *cmd* argument. Each operation takes an accompanying argument,
  130. * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
  131. * below). The size argument is the size of the union pointed to by *attr*.
  132. */
  133. /**
  134. * DOC: eBPF Syscall Commands
  135. *
  136. * BPF_MAP_CREATE
  137. * Description
  138. * Create a map and return a file descriptor that refers to the
  139. * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
  140. * is automatically enabled for the new file descriptor.
  141. *
  142. * Applying **close**\ (2) to the file descriptor returned by
  143. * **BPF_MAP_CREATE** will delete the map (but see NOTES).
  144. *
  145. * Return
  146. * A new file descriptor (a nonnegative integer), or -1 if an
  147. * error occurred (in which case, *errno* is set appropriately).
  148. *
  149. * BPF_MAP_LOOKUP_ELEM
  150. * Description
  151. * Look up an element with a given *key* in the map referred to
  152. * by the file descriptor *map_fd*.
  153. *
  154. * The *flags* argument may be specified as one of the
  155. * following:
  156. *
  157. * **BPF_F_LOCK**
  158. * Look up the value of a spin-locked map without
  159. * returning the lock. This must be specified if the
  160. * elements contain a spinlock.
  161. *
  162. * Return
  163. * Returns zero on success. On error, -1 is returned and *errno*
  164. * is set appropriately.
  165. *
  166. * BPF_MAP_UPDATE_ELEM
  167. * Description
  168. * Create or update an element (key/value pair) in a specified map.
  169. *
  170. * The *flags* argument should be specified as one of the
  171. * following:
  172. *
  173. * **BPF_ANY**
  174. * Create a new element or update an existing element.
  175. * **BPF_NOEXIST**
  176. * Create a new element only if it did not exist.
  177. * **BPF_EXIST**
  178. * Update an existing element.
  179. * **BPF_F_LOCK**
  180. * Update a spin_lock-ed map element.
  181. *
  182. * Return
  183. * Returns zero on success. On error, -1 is returned and *errno*
  184. * is set appropriately.
  185. *
  186. * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
  187. * **E2BIG**, **EEXIST**, or **ENOENT**.
  188. *
  189. * **E2BIG**
  190. * The number of elements in the map reached the
  191. * *max_entries* limit specified at map creation time.
  192. * **EEXIST**
  193. * If *flags* specifies **BPF_NOEXIST** and the element
  194. * with *key* already exists in the map.
  195. * **ENOENT**
  196. * If *flags* specifies **BPF_EXIST** and the element with
  197. * *key* does not exist in the map.
  198. *
  199. * BPF_MAP_DELETE_ELEM
  200. * Description
  201. * Look up and delete an element by key in a specified map.
  202. *
  203. * Return
  204. * Returns zero on success. On error, -1 is returned and *errno*
  205. * is set appropriately.
  206. *
  207. * BPF_MAP_GET_NEXT_KEY
  208. * Description
  209. * Look up an element by key in a specified map and return the key
  210. * of the next element. Can be used to iterate over all elements
  211. * in the map.
  212. *
  213. * Return
  214. * Returns zero on success. On error, -1 is returned and *errno*
  215. * is set appropriately.
  216. *
  217. * The following cases can be used to iterate over all elements of
  218. * the map:
  219. *
  220. * * If *key* is not found, the operation returns zero and sets
  221. * the *next_key* pointer to the key of the first element.
  222. * * If *key* is found, the operation returns zero and sets the
  223. * *next_key* pointer to the key of the next element.
  224. * * If *key* is the last element, returns -1 and *errno* is set
  225. * to **ENOENT**.
  226. *
  227. * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
  228. * **EINVAL** on error.
  229. *
  230. * BPF_PROG_LOAD
  231. * Description
  232. * Verify and load an eBPF program, returning a new file
  233. * descriptor associated with the program.
  234. *
  235. * Applying **close**\ (2) to the file descriptor returned by
  236. * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
  237. *
  238. * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
  239. * automatically enabled for the new file descriptor.
  240. *
  241. * Return
  242. * A new file descriptor (a nonnegative integer), or -1 if an
  243. * error occurred (in which case, *errno* is set appropriately).
  244. *
  245. * BPF_OBJ_PIN
  246. * Description
  247. * Pin an eBPF program or map referred by the specified *bpf_fd*
  248. * to the provided *pathname* on the filesystem.
  249. *
  250. * The *pathname* argument must not contain a dot (".").
  251. *
  252. * On success, *pathname* retains a reference to the eBPF object,
  253. * preventing deallocation of the object when the original
  254. * *bpf_fd* is closed. This allow the eBPF object to live beyond
  255. * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
  256. * process.
  257. *
  258. * Applying **unlink**\ (2) or similar calls to the *pathname*
  259. * unpins the object from the filesystem, removing the reference.
  260. * If no other file descriptors or filesystem nodes refer to the
  261. * same object, it will be deallocated (see NOTES).
  262. *
  263. * The filesystem type for the parent directory of *pathname* must
  264. * be **BPF_FS_MAGIC**.
  265. *
  266. * Return
  267. * Returns zero on success. On error, -1 is returned and *errno*
  268. * is set appropriately.
  269. *
  270. * BPF_OBJ_GET
  271. * Description
  272. * Open a file descriptor for the eBPF object pinned to the
  273. * specified *pathname*.
  274. *
  275. * Return
  276. * A new file descriptor (a nonnegative integer), or -1 if an
  277. * error occurred (in which case, *errno* is set appropriately).
  278. *
  279. * BPF_PROG_ATTACH
  280. * Description
  281. * Attach an eBPF program to a *target_fd* at the specified
  282. * *attach_type* hook.
  283. *
  284. * The *attach_type* specifies the eBPF attachment point to
  285. * attach the program to, and must be one of *bpf_attach_type*
  286. * (see below).
  287. *
  288. * The *attach_bpf_fd* must be a valid file descriptor for a
  289. * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
  290. * or sock_ops type corresponding to the specified *attach_type*.
  291. *
  292. * The *target_fd* must be a valid file descriptor for a kernel
  293. * object which depends on the attach type of *attach_bpf_fd*:
  294. *
  295. * **BPF_PROG_TYPE_CGROUP_DEVICE**,
  296. * **BPF_PROG_TYPE_CGROUP_SKB**,
  297. * **BPF_PROG_TYPE_CGROUP_SOCK**,
  298. * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
  299. * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
  300. * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
  301. * **BPF_PROG_TYPE_SOCK_OPS**
  302. *
  303. * Control Group v2 hierarchy with the eBPF controller
  304. * enabled. Requires the kernel to be compiled with
  305. * **CONFIG_CGROUP_BPF**.
  306. *
  307. * **BPF_PROG_TYPE_FLOW_DISSECTOR**
  308. *
  309. * Network namespace (eg /proc/self/ns/net).
  310. *
  311. * **BPF_PROG_TYPE_LIRC_MODE2**
  312. *
  313. * LIRC device path (eg /dev/lircN). Requires the kernel
  314. * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
  315. *
  316. * **BPF_PROG_TYPE_SK_SKB**,
  317. * **BPF_PROG_TYPE_SK_MSG**
  318. *
  319. * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
  320. *
  321. * Return
  322. * Returns zero on success. On error, -1 is returned and *errno*
  323. * is set appropriately.
  324. *
  325. * BPF_PROG_DETACH
  326. * Description
  327. * Detach the eBPF program associated with the *target_fd* at the
  328. * hook specified by *attach_type*. The program must have been
  329. * previously attached using **BPF_PROG_ATTACH**.
  330. *
  331. * Return
  332. * Returns zero on success. On error, -1 is returned and *errno*
  333. * is set appropriately.
  334. *
  335. * BPF_PROG_TEST_RUN
  336. * Description
  337. * Run the eBPF program associated with the *prog_fd* a *repeat*
  338. * number of times against a provided program context *ctx_in* and
  339. * data *data_in*, and return the modified program context
  340. * *ctx_out*, *data_out* (for example, packet data), result of the
  341. * execution *retval*, and *duration* of the test run.
  342. *
  343. * The sizes of the buffers provided as input and output
  344. * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
  345. * be provided in the corresponding variables *ctx_size_in*,
  346. * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
  347. * of these parameters are not provided (ie set to NULL), the
  348. * corresponding size field must be zero.
  349. *
  350. * Some program types have particular requirements:
  351. *
  352. * **BPF_PROG_TYPE_SK_LOOKUP**
  353. * *data_in* and *data_out* must be NULL.
  354. *
  355. * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
  356. * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
  357. *
  358. * *ctx_out*, *data_in* and *data_out* must be NULL.
  359. * *repeat* must be zero.
  360. *
  361. * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
  362. *
  363. * Return
  364. * Returns zero on success. On error, -1 is returned and *errno*
  365. * is set appropriately.
  366. *
  367. * **ENOSPC**
  368. * Either *data_size_out* or *ctx_size_out* is too small.
  369. * **ENOTSUPP**
  370. * This command is not supported by the program type of
  371. * the program referred to by *prog_fd*.
  372. *
  373. * BPF_PROG_GET_NEXT_ID
  374. * Description
  375. * Fetch the next eBPF program currently loaded into the kernel.
  376. *
  377. * Looks for the eBPF program with an id greater than *start_id*
  378. * and updates *next_id* on success. If no other eBPF programs
  379. * remain with ids higher than *start_id*, returns -1 and sets
  380. * *errno* to **ENOENT**.
  381. *
  382. * Return
  383. * Returns zero on success. On error, or when no id remains, -1
  384. * is returned and *errno* is set appropriately.
  385. *
  386. * BPF_MAP_GET_NEXT_ID
  387. * Description
  388. * Fetch the next eBPF map currently loaded into the kernel.
  389. *
  390. * Looks for the eBPF map with an id greater than *start_id*
  391. * and updates *next_id* on success. If no other eBPF maps
  392. * remain with ids higher than *start_id*, returns -1 and sets
  393. * *errno* to **ENOENT**.
  394. *
  395. * Return
  396. * Returns zero on success. On error, or when no id remains, -1
  397. * is returned and *errno* is set appropriately.
  398. *
  399. * BPF_PROG_GET_FD_BY_ID
  400. * Description
  401. * Open a file descriptor for the eBPF program corresponding to
  402. * *prog_id*.
  403. *
  404. * Return
  405. * A new file descriptor (a nonnegative integer), or -1 if an
  406. * error occurred (in which case, *errno* is set appropriately).
  407. *
  408. * BPF_MAP_GET_FD_BY_ID
  409. * Description
  410. * Open a file descriptor for the eBPF map corresponding to
  411. * *map_id*.
  412. *
  413. * Return
  414. * A new file descriptor (a nonnegative integer), or -1 if an
  415. * error occurred (in which case, *errno* is set appropriately).
  416. *
  417. * BPF_OBJ_GET_INFO_BY_FD
  418. * Description
  419. * Obtain information about the eBPF object corresponding to
  420. * *bpf_fd*.
  421. *
  422. * Populates up to *info_len* bytes of *info*, which will be in
  423. * one of the following formats depending on the eBPF object type
  424. * of *bpf_fd*:
  425. *
  426. * * **struct bpf_prog_info**
  427. * * **struct bpf_map_info**
  428. * * **struct bpf_btf_info**
  429. * * **struct bpf_link_info**
  430. *
  431. * Return
  432. * Returns zero on success. On error, -1 is returned and *errno*
  433. * is set appropriately.
  434. *
  435. * BPF_PROG_QUERY
  436. * Description
  437. * Obtain information about eBPF programs associated with the
  438. * specified *attach_type* hook.
  439. *
  440. * The *target_fd* must be a valid file descriptor for a kernel
  441. * object which depends on the attach type of *attach_bpf_fd*:
  442. *
  443. * **BPF_PROG_TYPE_CGROUP_DEVICE**,
  444. * **BPF_PROG_TYPE_CGROUP_SKB**,
  445. * **BPF_PROG_TYPE_CGROUP_SOCK**,
  446. * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
  447. * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
  448. * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
  449. * **BPF_PROG_TYPE_SOCK_OPS**
  450. *
  451. * Control Group v2 hierarchy with the eBPF controller
  452. * enabled. Requires the kernel to be compiled with
  453. * **CONFIG_CGROUP_BPF**.
  454. *
  455. * **BPF_PROG_TYPE_FLOW_DISSECTOR**
  456. *
  457. * Network namespace (eg /proc/self/ns/net).
  458. *
  459. * **BPF_PROG_TYPE_LIRC_MODE2**
  460. *
  461. * LIRC device path (eg /dev/lircN). Requires the kernel
  462. * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
  463. *
  464. * **BPF_PROG_QUERY** always fetches the number of programs
  465. * attached and the *attach_flags* which were used to attach those
  466. * programs. Additionally, if *prog_ids* is nonzero and the number
  467. * of attached programs is less than *prog_cnt*, populates
  468. * *prog_ids* with the eBPF program ids of the programs attached
  469. * at *target_fd*.
  470. *
  471. * The following flags may alter the result:
  472. *
  473. * **BPF_F_QUERY_EFFECTIVE**
  474. * Only return information regarding programs which are
  475. * currently effective at the specified *target_fd*.
  476. *
  477. * Return
  478. * Returns zero on success. On error, -1 is returned and *errno*
  479. * is set appropriately.
  480. *
  481. * BPF_RAW_TRACEPOINT_OPEN
  482. * Description
  483. * Attach an eBPF program to a tracepoint *name* to access kernel
  484. * internal arguments of the tracepoint in their raw form.
  485. *
  486. * The *prog_fd* must be a valid file descriptor associated with
  487. * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
  488. *
  489. * No ABI guarantees are made about the content of tracepoint
  490. * arguments exposed to the corresponding eBPF program.
  491. *
  492. * Applying **close**\ (2) to the file descriptor returned by
  493. * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
  494. *
  495. * Return
  496. * A new file descriptor (a nonnegative integer), or -1 if an
  497. * error occurred (in which case, *errno* is set appropriately).
  498. *
  499. * BPF_BTF_LOAD
  500. * Description
  501. * Verify and load BPF Type Format (BTF) metadata into the kernel,
  502. * returning a new file descriptor associated with the metadata.
  503. * BTF is described in more detail at
  504. * https://www.kernel.org/doc/html/latest/bpf/btf.html.
  505. *
  506. * The *btf* parameter must point to valid memory providing
  507. * *btf_size* bytes of BTF binary metadata.
  508. *
  509. * The returned file descriptor can be passed to other **bpf**\ ()
  510. * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
  511. * associate the BTF with those objects.
  512. *
  513. * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
  514. * parameters to specify a *btf_log_buf*, *btf_log_size* and
  515. * *btf_log_level* which allow the kernel to return freeform log
  516. * output regarding the BTF verification process.
  517. *
  518. * Return
  519. * A new file descriptor (a nonnegative integer), or -1 if an
  520. * error occurred (in which case, *errno* is set appropriately).
  521. *
  522. * BPF_BTF_GET_FD_BY_ID
  523. * Description
  524. * Open a file descriptor for the BPF Type Format (BTF)
  525. * corresponding to *btf_id*.
  526. *
  527. * Return
  528. * A new file descriptor (a nonnegative integer), or -1 if an
  529. * error occurred (in which case, *errno* is set appropriately).
  530. *
  531. * BPF_TASK_FD_QUERY
  532. * Description
  533. * Obtain information about eBPF programs associated with the
  534. * target process identified by *pid* and *fd*.
  535. *
  536. * If the *pid* and *fd* are associated with a tracepoint, kprobe
  537. * or uprobe perf event, then the *prog_id* and *fd_type* will
  538. * be populated with the eBPF program id and file descriptor type
  539. * of type **bpf_task_fd_type**. If associated with a kprobe or
  540. * uprobe, the *probe_offset* and *probe_addr* will also be
  541. * populated. Optionally, if *buf* is provided, then up to
  542. * *buf_len* bytes of *buf* will be populated with the name of
  543. * the tracepoint, kprobe or uprobe.
  544. *
  545. * The resulting *prog_id* may be introspected in deeper detail
  546. * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
  547. *
  548. * Return
  549. * Returns zero on success. On error, -1 is returned and *errno*
  550. * is set appropriately.
  551. *
  552. * BPF_MAP_LOOKUP_AND_DELETE_ELEM
  553. * Description
  554. * Look up an element with the given *key* in the map referred to
  555. * by the file descriptor *fd*, and if found, delete the element.
  556. *
  557. * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
  558. * types, the *flags* argument needs to be set to 0, but for other
  559. * map types, it may be specified as:
  560. *
  561. * **BPF_F_LOCK**
  562. * Look up and delete the value of a spin-locked map
  563. * without returning the lock. This must be specified if
  564. * the elements contain a spinlock.
  565. *
  566. * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
  567. * implement this command as a "pop" operation, deleting the top
  568. * element rather than one corresponding to *key*.
  569. * The *key* and *key_len* parameters should be zeroed when
  570. * issuing this operation for these map types.
  571. *
  572. * This command is only valid for the following map types:
  573. * * **BPF_MAP_TYPE_QUEUE**
  574. * * **BPF_MAP_TYPE_STACK**
  575. * * **BPF_MAP_TYPE_HASH**
  576. * * **BPF_MAP_TYPE_PERCPU_HASH**
  577. * * **BPF_MAP_TYPE_LRU_HASH**
  578. * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
  579. *
  580. * Return
  581. * Returns zero on success. On error, -1 is returned and *errno*
  582. * is set appropriately.
  583. *
  584. * BPF_MAP_FREEZE
  585. * Description
  586. * Freeze the permissions of the specified map.
  587. *
  588. * Write permissions may be frozen by passing zero *flags*.
  589. * Upon success, no future syscall invocations may alter the
  590. * map state of *map_fd*. Write operations from eBPF programs
  591. * are still possible for a frozen map.
  592. *
  593. * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
  594. *
  595. * Return
  596. * Returns zero on success. On error, -1 is returned and *errno*
  597. * is set appropriately.
  598. *
  599. * BPF_BTF_GET_NEXT_ID
  600. * Description
  601. * Fetch the next BPF Type Format (BTF) object currently loaded
  602. * into the kernel.
  603. *
  604. * Looks for the BTF object with an id greater than *start_id*
  605. * and updates *next_id* on success. If no other BTF objects
  606. * remain with ids higher than *start_id*, returns -1 and sets
  607. * *errno* to **ENOENT**.
  608. *
  609. * Return
  610. * Returns zero on success. On error, or when no id remains, -1
  611. * is returned and *errno* is set appropriately.
  612. *
  613. * BPF_MAP_LOOKUP_BATCH
  614. * Description
  615. * Iterate and fetch multiple elements in a map.
  616. *
  617. * Two opaque values are used to manage batch operations,
  618. * *in_batch* and *out_batch*. Initially, *in_batch* must be set
  619. * to NULL to begin the batched operation. After each subsequent
  620. * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
  621. * *out_batch* as the *in_batch* for the next operation to
  622. * continue iteration from the current point. Both *in_batch* and
  623. * *out_batch* must point to memory large enough to hold a key,
  624. * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,
  625. * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters
  626. * must be at least 4 bytes wide regardless of key size.
  627. *
  628. * The *keys* and *values* are output parameters which must point
  629. * to memory large enough to hold *count* items based on the key
  630. * and value size of the map *map_fd*. The *keys* buffer must be
  631. * of *key_size* * *count*. The *values* buffer must be of
  632. * *value_size* * *count*.
  633. *
  634. * The *elem_flags* argument may be specified as one of the
  635. * following:
  636. *
  637. * **BPF_F_LOCK**
  638. * Look up the value of a spin-locked map without
  639. * returning the lock. This must be specified if the
  640. * elements contain a spinlock.
  641. *
  642. * On success, *count* elements from the map are copied into the
  643. * user buffer, with the keys copied into *keys* and the values
  644. * copied into the corresponding indices in *values*.
  645. *
  646. * If an error is returned and *errno* is not **EFAULT**, *count*
  647. * is set to the number of successfully processed elements.
  648. *
  649. * Return
  650. * Returns zero on success. On error, -1 is returned and *errno*
  651. * is set appropriately.
  652. *
  653. * May set *errno* to **ENOSPC** to indicate that *keys* or
  654. * *values* is too small to dump an entire bucket during
  655. * iteration of a hash-based map type.
  656. *
  657. * BPF_MAP_LOOKUP_AND_DELETE_BATCH
  658. * Description
  659. * Iterate and delete all elements in a map.
  660. *
  661. * This operation has the same behavior as
  662. * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
  663. *
  664. * * Every element that is successfully returned is also deleted
  665. * from the map. This is at least *count* elements. Note that
  666. * *count* is both an input and an output parameter.
  667. * * Upon returning with *errno* set to **EFAULT**, up to
  668. * *count* elements may be deleted without returning the keys
  669. * and values of the deleted elements.
  670. *
  671. * Return
  672. * Returns zero on success. On error, -1 is returned and *errno*
  673. * is set appropriately.
  674. *
  675. * BPF_MAP_UPDATE_BATCH
  676. * Description
  677. * Update multiple elements in a map by *key*.
  678. *
  679. * The *keys* and *values* are input parameters which must point
  680. * to memory large enough to hold *count* items based on the key
  681. * and value size of the map *map_fd*. The *keys* buffer must be
  682. * of *key_size* * *count*. The *values* buffer must be of
  683. * *value_size* * *count*.
  684. *
  685. * Each element specified in *keys* is sequentially updated to the
  686. * value in the corresponding index in *values*. The *in_batch*
  687. * and *out_batch* parameters are ignored and should be zeroed.
  688. *
  689. * The *elem_flags* argument should be specified as one of the
  690. * following:
  691. *
  692. * **BPF_ANY**
  693. * Create new elements or update a existing elements.
  694. * **BPF_NOEXIST**
  695. * Create new elements only if they do not exist.
  696. * **BPF_EXIST**
  697. * Update existing elements.
  698. * **BPF_F_LOCK**
  699. * Update spin_lock-ed map elements. This must be
  700. * specified if the map value contains a spinlock.
  701. *
  702. * On success, *count* elements from the map are updated.
  703. *
  704. * If an error is returned and *errno* is not **EFAULT**, *count*
  705. * is set to the number of successfully processed elements.
  706. *
  707. * Return
  708. * Returns zero on success. On error, -1 is returned and *errno*
  709. * is set appropriately.
  710. *
  711. * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
  712. * **E2BIG**. **E2BIG** indicates that the number of elements in
  713. * the map reached the *max_entries* limit specified at map
  714. * creation time.
  715. *
  716. * May set *errno* to one of the following error codes under
  717. * specific circumstances:
  718. *
  719. * **EEXIST**
  720. * If *flags* specifies **BPF_NOEXIST** and the element
  721. * with *key* already exists in the map.
  722. * **ENOENT**
  723. * If *flags* specifies **BPF_EXIST** and the element with
  724. * *key* does not exist in the map.
  725. *
  726. * BPF_MAP_DELETE_BATCH
  727. * Description
  728. * Delete multiple elements in a map by *key*.
  729. *
  730. * The *keys* parameter is an input parameter which must point
  731. * to memory large enough to hold *count* items based on the key
  732. * size of the map *map_fd*, that is, *key_size* * *count*.
  733. *
  734. * Each element specified in *keys* is sequentially deleted. The
  735. * *in_batch*, *out_batch*, and *values* parameters are ignored
  736. * and should be zeroed.
  737. *
  738. * The *elem_flags* argument may be specified as one of the
  739. * following:
  740. *
  741. * **BPF_F_LOCK**
  742. * Look up the value of a spin-locked map without
  743. * returning the lock. This must be specified if the
  744. * elements contain a spinlock.
  745. *
  746. * On success, *count* elements from the map are updated.
  747. *
  748. * If an error is returned and *errno* is not **EFAULT**, *count*
  749. * is set to the number of successfully processed elements. If
  750. * *errno* is **EFAULT**, up to *count* elements may be been
  751. * deleted.
  752. *
  753. * Return
  754. * Returns zero on success. On error, -1 is returned and *errno*
  755. * is set appropriately.
  756. *
  757. * BPF_LINK_CREATE
  758. * Description
  759. * Attach an eBPF program to a *target_fd* at the specified
  760. * *attach_type* hook and return a file descriptor handle for
  761. * managing the link.
  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_UPDATE
  768. * Description
  769. * Update the eBPF program in the specified *link_fd* to
  770. * *new_prog_fd*.
  771. *
  772. * Return
  773. * Returns zero on success. On error, -1 is returned and *errno*
  774. * is set appropriately.
  775. *
  776. * BPF_LINK_GET_FD_BY_ID
  777. * Description
  778. * Open a file descriptor for the eBPF Link corresponding to
  779. * *link_id*.
  780. *
  781. * Return
  782. * A new file descriptor (a nonnegative integer), or -1 if an
  783. * error occurred (in which case, *errno* is set appropriately).
  784. *
  785. * BPF_LINK_GET_NEXT_ID
  786. * Description
  787. * Fetch the next eBPF link currently loaded into the kernel.
  788. *
  789. * Looks for the eBPF link with an id greater than *start_id*
  790. * and updates *next_id* on success. If no other eBPF links
  791. * remain with ids higher than *start_id*, returns -1 and sets
  792. * *errno* to **ENOENT**.
  793. *
  794. * Return
  795. * Returns zero on success. On error, or when no id remains, -1
  796. * is returned and *errno* is set appropriately.
  797. *
  798. * BPF_ENABLE_STATS
  799. * Description
  800. * Enable eBPF runtime statistics gathering.
  801. *
  802. * Runtime statistics gathering for the eBPF runtime is disabled
  803. * by default to minimize the corresponding performance overhead.
  804. * This command enables statistics globally.
  805. *
  806. * Multiple programs may independently enable statistics.
  807. * After gathering the desired statistics, eBPF runtime statistics
  808. * may be disabled again by calling **close**\ (2) for the file
  809. * descriptor returned by this function. Statistics will only be
  810. * disabled system-wide when all outstanding file descriptors
  811. * returned by prior calls for this subcommand are closed.
  812. *
  813. * Return
  814. * A new file descriptor (a nonnegative integer), or -1 if an
  815. * error occurred (in which case, *errno* is set appropriately).
  816. *
  817. * BPF_ITER_CREATE
  818. * Description
  819. * Create an iterator on top of the specified *link_fd* (as
  820. * previously created using **BPF_LINK_CREATE**) and return a
  821. * file descriptor that can be used to trigger the iteration.
  822. *
  823. * If the resulting file descriptor is pinned to the filesystem
  824. * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
  825. * for that path will trigger the iterator to read kernel state
  826. * using the eBPF program attached to *link_fd*.
  827. *
  828. * Return
  829. * A new file descriptor (a nonnegative integer), or -1 if an
  830. * error occurred (in which case, *errno* is set appropriately).
  831. *
  832. * BPF_LINK_DETACH
  833. * Description
  834. * Forcefully detach the specified *link_fd* from its
  835. * corresponding attachment point.
  836. *
  837. * Return
  838. * Returns zero on success. On error, -1 is returned and *errno*
  839. * is set appropriately.
  840. *
  841. * BPF_PROG_BIND_MAP
  842. * Description
  843. * Bind a map to the lifetime of an eBPF program.
  844. *
  845. * The map identified by *map_fd* is bound to the program
  846. * identified by *prog_fd* and only released when *prog_fd* is
  847. * released. This may be used in cases where metadata should be
  848. * associated with a program which otherwise does not contain any
  849. * references to the map (for example, embedded in the eBPF
  850. * program instructions).
  851. *
  852. * Return
  853. * Returns zero on success. On error, -1 is returned and *errno*
  854. * is set appropriately.
  855. *
  856. * BPF_TOKEN_CREATE
  857. * Description
  858. * Create BPF token with embedded information about what
  859. * BPF-related functionality it allows:
  860. * - a set of allowed bpf() syscall commands;
  861. * - a set of allowed BPF map types to be created with
  862. * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;
  863. * - a set of allowed BPF program types and BPF program attach
  864. * types to be loaded with BPF_PROG_LOAD command, if
  865. * BPF_PROG_LOAD itself is allowed.
  866. *
  867. * BPF token is created (derived) from an instance of BPF FS,
  868. * assuming it has necessary delegation mount options specified.
  869. * This BPF token can be passed as an extra parameter to various
  870. * bpf() syscall commands to grant BPF subsystem functionality to
  871. * unprivileged processes.
  872. *
  873. * When created, BPF token is "associated" with the owning
  874. * user namespace of BPF FS instance (super block) that it was
  875. * derived from, and subsequent BPF operations performed with
  876. * BPF token would be performing capabilities checks (i.e.,
  877. * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within
  878. * that user namespace. Without BPF token, such capabilities
  879. * have to be granted in init user namespace, making bpf()
  880. * syscall incompatible with user namespace, for the most part.
  881. *
  882. * Return
  883. * A new file descriptor (a nonnegative integer), or -1 if an
  884. * error occurred (in which case, *errno* is set appropriately).
  885. *
  886. * NOTES
  887. * eBPF objects (maps and programs) can be shared between processes.
  888. *
  889. * * After **fork**\ (2), the child inherits file descriptors
  890. * referring to the same eBPF objects.
  891. * * File descriptors referring to eBPF objects can be transferred over
  892. * **unix**\ (7) domain sockets.
  893. * * File descriptors referring to eBPF objects can be duplicated in the
  894. * usual way, using **dup**\ (2) and similar calls.
  895. * * File descriptors referring to eBPF objects can be pinned to the
  896. * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
  897. *
  898. * An eBPF object is deallocated only after all file descriptors referring
  899. * to the object have been closed and no references remain pinned to the
  900. * filesystem or attached (for example, bound to a program or device).
  901. */
  902. enum bpf_cmd {
  903. BPF_MAP_CREATE,
  904. BPF_MAP_LOOKUP_ELEM,
  905. BPF_MAP_UPDATE_ELEM,
  906. BPF_MAP_DELETE_ELEM,
  907. BPF_MAP_GET_NEXT_KEY,
  908. BPF_PROG_LOAD,
  909. BPF_OBJ_PIN,
  910. BPF_OBJ_GET,
  911. BPF_PROG_ATTACH,
  912. BPF_PROG_DETACH,
  913. BPF_PROG_TEST_RUN,
  914. BPF_PROG_RUN = BPF_PROG_TEST_RUN,
  915. BPF_PROG_GET_NEXT_ID,
  916. BPF_MAP_GET_NEXT_ID,
  917. BPF_PROG_GET_FD_BY_ID,
  918. BPF_MAP_GET_FD_BY_ID,
  919. BPF_OBJ_GET_INFO_BY_FD,
  920. BPF_PROG_QUERY,
  921. BPF_RAW_TRACEPOINT_OPEN,
  922. BPF_BTF_LOAD,
  923. BPF_BTF_GET_FD_BY_ID,
  924. BPF_TASK_FD_QUERY,
  925. BPF_MAP_LOOKUP_AND_DELETE_ELEM,
  926. BPF_MAP_FREEZE,
  927. BPF_BTF_GET_NEXT_ID,
  928. BPF_MAP_LOOKUP_BATCH,
  929. BPF_MAP_LOOKUP_AND_DELETE_BATCH,
  930. BPF_MAP_UPDATE_BATCH,
  931. BPF_MAP_DELETE_BATCH,
  932. BPF_LINK_CREATE,
  933. BPF_LINK_UPDATE,
  934. BPF_LINK_GET_FD_BY_ID,
  935. BPF_LINK_GET_NEXT_ID,
  936. BPF_ENABLE_STATS,
  937. BPF_ITER_CREATE,
  938. BPF_LINK_DETACH,
  939. BPF_PROG_BIND_MAP,
  940. BPF_TOKEN_CREATE,
  941. __MAX_BPF_CMD,
  942. };
  943. enum bpf_map_type {
  944. BPF_MAP_TYPE_UNSPEC,
  945. BPF_MAP_TYPE_HASH,
  946. BPF_MAP_TYPE_ARRAY,
  947. BPF_MAP_TYPE_PROG_ARRAY,
  948. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  949. BPF_MAP_TYPE_PERCPU_HASH,
  950. BPF_MAP_TYPE_PERCPU_ARRAY,
  951. BPF_MAP_TYPE_STACK_TRACE,
  952. BPF_MAP_TYPE_CGROUP_ARRAY,
  953. BPF_MAP_TYPE_LRU_HASH,
  954. BPF_MAP_TYPE_LRU_PERCPU_HASH,
  955. BPF_MAP_TYPE_LPM_TRIE,
  956. BPF_MAP_TYPE_ARRAY_OF_MAPS,
  957. BPF_MAP_TYPE_HASH_OF_MAPS,
  958. BPF_MAP_TYPE_DEVMAP,
  959. BPF_MAP_TYPE_SOCKMAP,
  960. BPF_MAP_TYPE_CPUMAP,
  961. BPF_MAP_TYPE_XSKMAP,
  962. BPF_MAP_TYPE_SOCKHASH,
  963. BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
  964. /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
  965. * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
  966. * both cgroup-attached and other progs and supports all functionality
  967. * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
  968. * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
  969. */
  970. BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
  971. BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
  972. BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
  973. /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
  974. * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
  975. * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
  976. * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
  977. * deprecated.
  978. */
  979. BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
  980. BPF_MAP_TYPE_QUEUE,
  981. BPF_MAP_TYPE_STACK,
  982. BPF_MAP_TYPE_SK_STORAGE,
  983. BPF_MAP_TYPE_DEVMAP_HASH,
  984. BPF_MAP_TYPE_STRUCT_OPS,
  985. BPF_MAP_TYPE_RINGBUF,
  986. BPF_MAP_TYPE_INODE_STORAGE,
  987. BPF_MAP_TYPE_TASK_STORAGE,
  988. BPF_MAP_TYPE_BLOOM_FILTER,
  989. BPF_MAP_TYPE_USER_RINGBUF,
  990. BPF_MAP_TYPE_CGRP_STORAGE,
  991. BPF_MAP_TYPE_ARENA,
  992. __MAX_BPF_MAP_TYPE
  993. };
  994. /* Note that tracing related programs such as
  995. * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
  996. * are not subject to a stable API since kernel internal data
  997. * structures can change from release to release and may
  998. * therefore break existing tracing BPF programs. Tracing BPF
  999. * programs correspond to /a/ specific kernel which is to be
  1000. * analyzed, and not /a/ specific kernel /and/ all future ones.
  1001. */
  1002. enum bpf_prog_type {
  1003. BPF_PROG_TYPE_UNSPEC,
  1004. BPF_PROG_TYPE_SOCKET_FILTER,
  1005. BPF_PROG_TYPE_KPROBE,
  1006. BPF_PROG_TYPE_SCHED_CLS,
  1007. BPF_PROG_TYPE_SCHED_ACT,
  1008. BPF_PROG_TYPE_TRACEPOINT,
  1009. BPF_PROG_TYPE_XDP,
  1010. BPF_PROG_TYPE_PERF_EVENT,
  1011. BPF_PROG_TYPE_CGROUP_SKB,
  1012. BPF_PROG_TYPE_CGROUP_SOCK,
  1013. BPF_PROG_TYPE_LWT_IN,
  1014. BPF_PROG_TYPE_LWT_OUT,
  1015. BPF_PROG_TYPE_LWT_XMIT,
  1016. BPF_PROG_TYPE_SOCK_OPS,
  1017. BPF_PROG_TYPE_SK_SKB,
  1018. BPF_PROG_TYPE_CGROUP_DEVICE,
  1019. BPF_PROG_TYPE_SK_MSG,
  1020. BPF_PROG_TYPE_RAW_TRACEPOINT,
  1021. BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
  1022. BPF_PROG_TYPE_LWT_SEG6LOCAL,
  1023. BPF_PROG_TYPE_LIRC_MODE2,
  1024. BPF_PROG_TYPE_SK_REUSEPORT,
  1025. BPF_PROG_TYPE_FLOW_DISSECTOR,
  1026. BPF_PROG_TYPE_CGROUP_SYSCTL,
  1027. BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
  1028. BPF_PROG_TYPE_CGROUP_SOCKOPT,
  1029. BPF_PROG_TYPE_TRACING,
  1030. BPF_PROG_TYPE_STRUCT_OPS,
  1031. BPF_PROG_TYPE_EXT,
  1032. BPF_PROG_TYPE_LSM,
  1033. BPF_PROG_TYPE_SK_LOOKUP,
  1034. BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
  1035. BPF_PROG_TYPE_NETFILTER,
  1036. __MAX_BPF_PROG_TYPE
  1037. };
  1038. enum bpf_attach_type {
  1039. BPF_CGROUP_INET_INGRESS,
  1040. BPF_CGROUP_INET_EGRESS,
  1041. BPF_CGROUP_INET_SOCK_CREATE,
  1042. BPF_CGROUP_SOCK_OPS,
  1043. BPF_SK_SKB_STREAM_PARSER,
  1044. BPF_SK_SKB_STREAM_VERDICT,
  1045. BPF_CGROUP_DEVICE,
  1046. BPF_SK_MSG_VERDICT,
  1047. BPF_CGROUP_INET4_BIND,
  1048. BPF_CGROUP_INET6_BIND,
  1049. BPF_CGROUP_INET4_CONNECT,
  1050. BPF_CGROUP_INET6_CONNECT,
  1051. BPF_CGROUP_INET4_POST_BIND,
  1052. BPF_CGROUP_INET6_POST_BIND,
  1053. BPF_CGROUP_UDP4_SENDMSG,
  1054. BPF_CGROUP_UDP6_SENDMSG,
  1055. BPF_LIRC_MODE2,
  1056. BPF_FLOW_DISSECTOR,
  1057. BPF_CGROUP_SYSCTL,
  1058. BPF_CGROUP_UDP4_RECVMSG,
  1059. BPF_CGROUP_UDP6_RECVMSG,
  1060. BPF_CGROUP_GETSOCKOPT,
  1061. BPF_CGROUP_SETSOCKOPT,
  1062. BPF_TRACE_RAW_TP,
  1063. BPF_TRACE_FENTRY,
  1064. BPF_TRACE_FEXIT,
  1065. BPF_MODIFY_RETURN,
  1066. BPF_LSM_MAC,
  1067. BPF_TRACE_ITER,
  1068. BPF_CGROUP_INET4_GETPEERNAME,
  1069. BPF_CGROUP_INET6_GETPEERNAME,
  1070. BPF_CGROUP_INET4_GETSOCKNAME,
  1071. BPF_CGROUP_INET6_GETSOCKNAME,
  1072. BPF_XDP_DEVMAP,
  1073. BPF_CGROUP_INET_SOCK_RELEASE,
  1074. BPF_XDP_CPUMAP,
  1075. BPF_SK_LOOKUP,
  1076. BPF_XDP,
  1077. BPF_SK_SKB_VERDICT,
  1078. BPF_SK_REUSEPORT_SELECT,
  1079. BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
  1080. BPF_PERF_EVENT,
  1081. BPF_TRACE_KPROBE_MULTI,
  1082. BPF_LSM_CGROUP,
  1083. BPF_STRUCT_OPS,
  1084. BPF_NETFILTER,
  1085. BPF_TCX_INGRESS,
  1086. BPF_TCX_EGRESS,
  1087. BPF_TRACE_UPROBE_MULTI,
  1088. BPF_CGROUP_UNIX_CONNECT,
  1089. BPF_CGROUP_UNIX_SENDMSG,
  1090. BPF_CGROUP_UNIX_RECVMSG,
  1091. BPF_CGROUP_UNIX_GETPEERNAME,
  1092. BPF_CGROUP_UNIX_GETSOCKNAME,
  1093. BPF_NETKIT_PRIMARY,
  1094. BPF_NETKIT_PEER,
  1095. BPF_TRACE_KPROBE_SESSION,
  1096. __MAX_BPF_ATTACH_TYPE
  1097. };
  1098. #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  1099. /* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]
  1100. * in sync with the definitions below.
  1101. */
  1102. enum bpf_link_type {
  1103. BPF_LINK_TYPE_UNSPEC = 0,
  1104. BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
  1105. BPF_LINK_TYPE_TRACING = 2,
  1106. BPF_LINK_TYPE_CGROUP = 3,
  1107. BPF_LINK_TYPE_ITER = 4,
  1108. BPF_LINK_TYPE_NETNS = 5,
  1109. BPF_LINK_TYPE_XDP = 6,
  1110. BPF_LINK_TYPE_PERF_EVENT = 7,
  1111. BPF_LINK_TYPE_KPROBE_MULTI = 8,
  1112. BPF_LINK_TYPE_STRUCT_OPS = 9,
  1113. BPF_LINK_TYPE_NETFILTER = 10,
  1114. BPF_LINK_TYPE_TCX = 11,
  1115. BPF_LINK_TYPE_UPROBE_MULTI = 12,
  1116. BPF_LINK_TYPE_NETKIT = 13,
  1117. BPF_LINK_TYPE_SOCKMAP = 14,
  1118. __MAX_BPF_LINK_TYPE,
  1119. };
  1120. #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
  1121. enum bpf_perf_event_type {
  1122. BPF_PERF_EVENT_UNSPEC = 0,
  1123. BPF_PERF_EVENT_UPROBE = 1,
  1124. BPF_PERF_EVENT_URETPROBE = 2,
  1125. BPF_PERF_EVENT_KPROBE = 3,
  1126. BPF_PERF_EVENT_KRETPROBE = 4,
  1127. BPF_PERF_EVENT_TRACEPOINT = 5,
  1128. BPF_PERF_EVENT_EVENT = 6,
  1129. };
  1130. /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
  1131. *
  1132. * NONE(default): No further bpf programs allowed in the subtree.
  1133. *
  1134. * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
  1135. * the program in this cgroup yields to sub-cgroup program.
  1136. *
  1137. * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
  1138. * that cgroup program gets run in addition to the program in this cgroup.
  1139. *
  1140. * Only one program is allowed to be attached to a cgroup with
  1141. * NONE or BPF_F_ALLOW_OVERRIDE flag.
  1142. * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
  1143. * release old program and attach the new one. Attach flags has to match.
  1144. *
  1145. * Multiple programs are allowed to be attached to a cgroup with
  1146. * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
  1147. * (those that were attached first, run first)
  1148. * The programs of sub-cgroup are executed first, then programs of
  1149. * this cgroup and then programs of parent cgroup.
  1150. * When children program makes decision (like picking TCP CA or sock bind)
  1151. * parent program has a chance to override it.
  1152. *
  1153. * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
  1154. * programs for a cgroup. Though it's possible to replace an old program at
  1155. * any position by also specifying BPF_F_REPLACE flag and position itself in
  1156. * replace_bpf_fd attribute. Old program at this position will be released.
  1157. *
  1158. * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
  1159. * A cgroup with NONE doesn't allow any programs in sub-cgroups.
  1160. * Ex1:
  1161. * cgrp1 (MULTI progs A, B) ->
  1162. * cgrp2 (OVERRIDE prog C) ->
  1163. * cgrp3 (MULTI prog D) ->
  1164. * cgrp4 (OVERRIDE prog E) ->
  1165. * cgrp5 (NONE prog F)
  1166. * the event in cgrp5 triggers execution of F,D,A,B in that order.
  1167. * if prog F is detached, the execution is E,D,A,B
  1168. * if prog F and D are detached, the execution is E,A,B
  1169. * if prog F, E and D are detached, the execution is C,A,B
  1170. *
  1171. * All eligible programs are executed regardless of return code from
  1172. * earlier programs.
  1173. */
  1174. #define BPF_F_ALLOW_OVERRIDE (1U << 0)
  1175. #define BPF_F_ALLOW_MULTI (1U << 1)
  1176. /* Generic attachment flags. */
  1177. #define BPF_F_REPLACE (1U << 2)
  1178. #define BPF_F_BEFORE (1U << 3)
  1179. #define BPF_F_AFTER (1U << 4)
  1180. #define BPF_F_ID (1U << 5)
  1181. #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */
  1182. /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  1183. * verifier will perform strict alignment checking as if the kernel
  1184. * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  1185. * and NET_IP_ALIGN defined to 2.
  1186. */
  1187. #define BPF_F_STRICT_ALIGNMENT (1U << 0)
  1188. /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
  1189. * verifier will allow any alignment whatsoever. On platforms
  1190. * with strict alignment requirements for loads ands stores (such
  1191. * as sparc and mips) the verifier validates that all loads and
  1192. * stores provably follow this requirement. This flag turns that
  1193. * checking and enforcement off.
  1194. *
  1195. * It is mostly used for testing when we want to validate the
  1196. * context and memory access aspects of the verifier, but because
  1197. * of an unaligned access the alignment check would trigger before
  1198. * the one we are interested in.
  1199. */
  1200. #define BPF_F_ANY_ALIGNMENT (1U << 1)
  1201. /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
  1202. * Verifier does sub-register def/use analysis and identifies instructions whose
  1203. * def only matters for low 32-bit, high 32-bit is never referenced later
  1204. * through implicit zero extension. Therefore verifier notifies JIT back-ends
  1205. * that it is safe to ignore clearing high 32-bit for these instructions. This
  1206. * saves some back-ends a lot of code-gen. However such optimization is not
  1207. * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
  1208. * hence hasn't used verifier's analysis result. But, we really want to have a
  1209. * way to be able to verify the correctness of the described optimization on
  1210. * x86_64 on which testsuites are frequently exercised.
  1211. *
  1212. * So, this flag is introduced. Once it is set, verifier will randomize high
  1213. * 32-bit for those instructions who has been identified as safe to ignore them.
  1214. * Then, if verifier is not doing correct analysis, such randomization will
  1215. * regress tests to expose bugs.
  1216. */
  1217. #define BPF_F_TEST_RND_HI32 (1U << 2)
  1218. /* The verifier internal test flag. Behavior is undefined */
  1219. #define BPF_F_TEST_STATE_FREQ (1U << 3)
  1220. /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
  1221. * restrict map and helper usage for such programs. Sleepable BPF programs can
  1222. * only be attached to hooks where kernel execution context allows sleeping.
  1223. * Such programs are allowed to use helpers that may sleep like
  1224. * bpf_copy_from_user().
  1225. */
  1226. #define BPF_F_SLEEPABLE (1U << 4)
  1227. /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
  1228. * fully support xdp frags.
  1229. */
  1230. #define BPF_F_XDP_HAS_FRAGS (1U << 5)
  1231. /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
  1232. * program becomes device-bound but can access XDP metadata.
  1233. */
  1234. #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6)
  1235. /* The verifier internal test flag. Behavior is undefined */
  1236. #define BPF_F_TEST_REG_INVARIANTS (1U << 7)
  1237. /* link_create.kprobe_multi.flags used in LINK_CREATE command for
  1238. * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
  1239. */
  1240. enum {
  1241. BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
  1242. };
  1243. /* link_create.uprobe_multi.flags used in LINK_CREATE command for
  1244. * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
  1245. */
  1246. enum {
  1247. BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
  1248. };
  1249. /* link_create.netfilter.flags used in LINK_CREATE command for
  1250. * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
  1251. */
  1252. #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
  1253. /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
  1254. * the following extensions:
  1255. *
  1256. * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
  1257. * insn[0].imm: map fd or fd_idx
  1258. * insn[1].imm: 0
  1259. * insn[0].off: 0
  1260. * insn[1].off: 0
  1261. * ldimm64 rewrite: address of map
  1262. * verifier type: CONST_PTR_TO_MAP
  1263. */
  1264. #define BPF_PSEUDO_MAP_FD 1
  1265. #define BPF_PSEUDO_MAP_IDX 5
  1266. /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
  1267. * insn[0].imm: map fd or fd_idx
  1268. * insn[1].imm: offset into value
  1269. * insn[0].off: 0
  1270. * insn[1].off: 0
  1271. * ldimm64 rewrite: address of map[0]+offset
  1272. * verifier type: PTR_TO_MAP_VALUE
  1273. */
  1274. #define BPF_PSEUDO_MAP_VALUE 2
  1275. #define BPF_PSEUDO_MAP_IDX_VALUE 6
  1276. /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
  1277. * insn[0].imm: kernel btd id of VAR
  1278. * insn[1].imm: 0
  1279. * insn[0].off: 0
  1280. * insn[1].off: 0
  1281. * ldimm64 rewrite: address of the kernel variable
  1282. * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
  1283. * is struct/union.
  1284. */
  1285. #define BPF_PSEUDO_BTF_ID 3
  1286. /* insn[0].src_reg: BPF_PSEUDO_FUNC
  1287. * insn[0].imm: insn offset to the func
  1288. * insn[1].imm: 0
  1289. * insn[0].off: 0
  1290. * insn[1].off: 0
  1291. * ldimm64 rewrite: address of the function
  1292. * verifier type: PTR_TO_FUNC.
  1293. */
  1294. #define BPF_PSEUDO_FUNC 4
  1295. /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
  1296. * offset to another bpf function
  1297. */
  1298. #define BPF_PSEUDO_CALL 1
  1299. /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
  1300. * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
  1301. */
  1302. #define BPF_PSEUDO_KFUNC_CALL 2
  1303. enum bpf_addr_space_cast {
  1304. BPF_ADDR_SPACE_CAST = 1,
  1305. };
  1306. /* flags for BPF_MAP_UPDATE_ELEM command */
  1307. enum {
  1308. BPF_ANY = 0, /* create new element or update existing */
  1309. BPF_NOEXIST = 1, /* create new element if it didn't exist */
  1310. BPF_EXIST = 2, /* update existing element */
  1311. BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
  1312. };
  1313. /* flags for BPF_MAP_CREATE command */
  1314. enum {
  1315. BPF_F_NO_PREALLOC = (1U << 0),
  1316. /* Instead of having one common LRU list in the
  1317. * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  1318. * which can scale and perform better.
  1319. * Note, the LRU nodes (including free nodes) cannot be moved
  1320. * across different LRU lists.
  1321. */
  1322. BPF_F_NO_COMMON_LRU = (1U << 1),
  1323. /* Specify numa node during map creation */
  1324. BPF_F_NUMA_NODE = (1U << 2),
  1325. /* Flags for accessing BPF object from syscall side. */
  1326. BPF_F_RDONLY = (1U << 3),
  1327. BPF_F_WRONLY = (1U << 4),
  1328. /* Flag for stack_map, store build_id+offset instead of pointer */
  1329. BPF_F_STACK_BUILD_ID = (1U << 5),
  1330. /* Zero-initialize hash function seed. This should only be used for testing. */
  1331. BPF_F_ZERO_SEED = (1U << 6),
  1332. /* Flags for accessing BPF object from program side. */
  1333. BPF_F_RDONLY_PROG = (1U << 7),
  1334. BPF_F_WRONLY_PROG = (1U << 8),
  1335. /* Clone map from listener for newly accepted socket */
  1336. BPF_F_CLONE = (1U << 9),
  1337. /* Enable memory-mapping BPF map */
  1338. BPF_F_MMAPABLE = (1U << 10),
  1339. /* Share perf_event among processes */
  1340. BPF_F_PRESERVE_ELEMS = (1U << 11),
  1341. /* Create a map that is suitable to be an inner map with dynamic max entries */
  1342. BPF_F_INNER_MAP = (1U << 12),
  1343. /* Create a map that will be registered/unregesitered by the backed bpf_link */
  1344. BPF_F_LINK = (1U << 13),
  1345. /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
  1346. BPF_F_PATH_FD = (1U << 14),
  1347. /* Flag for value_type_btf_obj_fd, the fd is available */
  1348. BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15),
  1349. /* BPF token FD is passed in a corresponding command's token_fd field */
  1350. BPF_F_TOKEN_FD = (1U << 16),
  1351. /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */
  1352. BPF_F_SEGV_ON_FAULT = (1U << 17),
  1353. /* Do not translate kernel bpf_arena pointers to user pointers */
  1354. BPF_F_NO_USER_CONV = (1U << 18),
  1355. };
  1356. /* Flags for BPF_PROG_QUERY. */
  1357. /* Query effective (directly attached + inherited from ancestor cgroups)
  1358. * programs that will be executed for events within a cgroup.
  1359. * attach_flags with this flag are always returned 0.
  1360. */
  1361. #define BPF_F_QUERY_EFFECTIVE (1U << 0)
  1362. /* Flags for BPF_PROG_TEST_RUN */
  1363. /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
  1364. #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
  1365. /* If set, XDP frames will be transmitted after processing */
  1366. #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
  1367. /* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */
  1368. #define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2)
  1369. /* type for BPF_ENABLE_STATS */
  1370. enum bpf_stats_type {
  1371. /* enabled run_time_ns and run_cnt */
  1372. BPF_STATS_RUN_TIME = 0,
  1373. };
  1374. enum bpf_stack_build_id_status {
  1375. /* user space need an empty entry to identify end of a trace */
  1376. BPF_STACK_BUILD_ID_EMPTY = 0,
  1377. /* with valid build_id and offset */
  1378. BPF_STACK_BUILD_ID_VALID = 1,
  1379. /* couldn't get build_id, fallback to ip */
  1380. BPF_STACK_BUILD_ID_IP = 2,
  1381. };
  1382. #define BPF_BUILD_ID_SIZE 20
  1383. struct bpf_stack_build_id {
  1384. __s32 status;
  1385. unsigned char build_id[BPF_BUILD_ID_SIZE];
  1386. union {
  1387. __u64 offset;
  1388. __u64 ip;
  1389. };
  1390. };
  1391. #define BPF_OBJ_NAME_LEN 16U
  1392. union bpf_attr {
  1393. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  1394. __u32 map_type; /* one of enum bpf_map_type */
  1395. __u32 key_size; /* size of key in bytes */
  1396. __u32 value_size; /* size of value in bytes */
  1397. __u32 max_entries; /* max number of entries in a map */
  1398. __u32 map_flags; /* BPF_MAP_CREATE related
  1399. * flags defined above.
  1400. */
  1401. __u32 inner_map_fd; /* fd pointing to the inner map */
  1402. __u32 numa_node; /* numa node (effective only if
  1403. * BPF_F_NUMA_NODE is set).
  1404. */
  1405. char map_name[BPF_OBJ_NAME_LEN];
  1406. __u32 map_ifindex; /* ifindex of netdev to create on */
  1407. __u32 btf_fd; /* fd pointing to a BTF type data */
  1408. __u32 btf_key_type_id; /* BTF type_id of the key */
  1409. __u32 btf_value_type_id; /* BTF type_id of the value */
  1410. __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
  1411. * struct stored as the
  1412. * map value
  1413. */
  1414. /* Any per-map-type extra fields
  1415. *
  1416. * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
  1417. * number of hash functions (if 0, the bloom filter will default
  1418. * to using 5 hash functions).
  1419. *
  1420. * BPF_MAP_TYPE_ARENA - contains the address where user space
  1421. * is going to mmap() the arena. It has to be page aligned.
  1422. */
  1423. __u64 map_extra;
  1424. __s32 value_type_btf_obj_fd; /* fd pointing to a BTF
  1425. * type data for
  1426. * btf_vmlinux_value_type_id.
  1427. */
  1428. /* BPF token FD to use with BPF_MAP_CREATE operation.
  1429. * If provided, map_flags should have BPF_F_TOKEN_FD flag set.
  1430. */
  1431. __s32 map_token_fd;
  1432. };
  1433. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  1434. __u32 map_fd;
  1435. __aligned_u64 key;
  1436. union {
  1437. __aligned_u64 value;
  1438. __aligned_u64 next_key;
  1439. };
  1440. __u64 flags;
  1441. };
  1442. struct { /* struct used by BPF_MAP_*_BATCH commands */
  1443. __aligned_u64 in_batch; /* start batch,
  1444. * NULL to start from beginning
  1445. */
  1446. __aligned_u64 out_batch; /* output: next start batch */
  1447. __aligned_u64 keys;
  1448. __aligned_u64 values;
  1449. __u32 count; /* input/output:
  1450. * input: # of key/value
  1451. * elements
  1452. * output: # of filled elements
  1453. */
  1454. __u32 map_fd;
  1455. __u64 elem_flags;
  1456. __u64 flags;
  1457. } batch;
  1458. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  1459. __u32 prog_type; /* one of enum bpf_prog_type */
  1460. __u32 insn_cnt;
  1461. __aligned_u64 insns;
  1462. __aligned_u64 license;
  1463. __u32 log_level; /* verbosity level of verifier */
  1464. __u32 log_size; /* size of user buffer */
  1465. __aligned_u64 log_buf; /* user supplied buffer */
  1466. __u32 kern_version; /* not used */
  1467. __u32 prog_flags;
  1468. char prog_name[BPF_OBJ_NAME_LEN];
  1469. __u32 prog_ifindex; /* ifindex of netdev to prep for */
  1470. /* For some prog types expected attach type must be known at
  1471. * load time to verify attach type specific parts of prog
  1472. * (context accesses, allowed helpers, etc).
  1473. */
  1474. __u32 expected_attach_type;
  1475. __u32 prog_btf_fd; /* fd pointing to BTF type data */
  1476. __u32 func_info_rec_size; /* userspace bpf_func_info size */
  1477. __aligned_u64 func_info; /* func info */
  1478. __u32 func_info_cnt; /* number of bpf_func_info records */
  1479. __u32 line_info_rec_size; /* userspace bpf_line_info size */
  1480. __aligned_u64 line_info; /* line info */
  1481. __u32 line_info_cnt; /* number of bpf_line_info records */
  1482. __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
  1483. union {
  1484. /* valid prog_fd to attach to bpf prog */
  1485. __u32 attach_prog_fd;
  1486. /* or valid module BTF object fd or 0 to attach to vmlinux */
  1487. __u32 attach_btf_obj_fd;
  1488. };
  1489. __u32 core_relo_cnt; /* number of bpf_core_relo */
  1490. __aligned_u64 fd_array; /* array of FDs */
  1491. __aligned_u64 core_relos;
  1492. __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
  1493. /* output: actual total log contents size (including termintaing zero).
  1494. * It could be both larger than original log_size (if log was
  1495. * truncated), or smaller (if log buffer wasn't filled completely).
  1496. */
  1497. __u32 log_true_size;
  1498. /* BPF token FD to use with BPF_PROG_LOAD operation.
  1499. * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.
  1500. */
  1501. __s32 prog_token_fd;
  1502. };
  1503. struct { /* anonymous struct used by BPF_OBJ_* commands */
  1504. __aligned_u64 pathname;
  1505. __u32 bpf_fd;
  1506. __u32 file_flags;
  1507. /* Same as dirfd in openat() syscall; see openat(2)
  1508. * manpage for details of path FD and pathname semantics;
  1509. * path_fd should accompanied by BPF_F_PATH_FD flag set in
  1510. * file_flags field, otherwise it should be set to zero;
  1511. * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
  1512. */
  1513. __s32 path_fd;
  1514. };
  1515. struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  1516. union {
  1517. __u32 target_fd; /* target object to attach to or ... */
  1518. __u32 target_ifindex; /* target ifindex */
  1519. };
  1520. __u32 attach_bpf_fd;
  1521. __u32 attach_type;
  1522. __u32 attach_flags;
  1523. __u32 replace_bpf_fd;
  1524. union {
  1525. __u32 relative_fd;
  1526. __u32 relative_id;
  1527. };
  1528. __u64 expected_revision;
  1529. };
  1530. struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  1531. __u32 prog_fd;
  1532. __u32 retval;
  1533. __u32 data_size_in; /* input: len of data_in */
  1534. __u32 data_size_out; /* input/output: len of data_out
  1535. * returns ENOSPC if data_out
  1536. * is too small.
  1537. */
  1538. __aligned_u64 data_in;
  1539. __aligned_u64 data_out;
  1540. __u32 repeat;
  1541. __u32 duration;
  1542. __u32 ctx_size_in; /* input: len of ctx_in */
  1543. __u32 ctx_size_out; /* input/output: len of ctx_out
  1544. * returns ENOSPC if ctx_out
  1545. * is too small.
  1546. */
  1547. __aligned_u64 ctx_in;
  1548. __aligned_u64 ctx_out;
  1549. __u32 flags;
  1550. __u32 cpu;
  1551. __u32 batch_size;
  1552. } test;
  1553. struct { /* anonymous struct used by BPF_*_GET_*_ID */
  1554. union {
  1555. __u32 start_id;
  1556. __u32 prog_id;
  1557. __u32 map_id;
  1558. __u32 btf_id;
  1559. __u32 link_id;
  1560. };
  1561. __u32 next_id;
  1562. __u32 open_flags;
  1563. };
  1564. struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  1565. __u32 bpf_fd;
  1566. __u32 info_len;
  1567. __aligned_u64 info;
  1568. } info;
  1569. struct { /* anonymous struct used by BPF_PROG_QUERY command */
  1570. union {
  1571. __u32 target_fd; /* target object to query or ... */
  1572. __u32 target_ifindex; /* target ifindex */
  1573. };
  1574. __u32 attach_type;
  1575. __u32 query_flags;
  1576. __u32 attach_flags;
  1577. __aligned_u64 prog_ids;
  1578. union {
  1579. __u32 prog_cnt;
  1580. __u32 count;
  1581. };
  1582. __u32 :32;
  1583. /* output: per-program attach_flags.
  1584. * not allowed to be set during effective query.
  1585. */
  1586. __aligned_u64 prog_attach_flags;
  1587. __aligned_u64 link_ids;
  1588. __aligned_u64 link_attach_flags;
  1589. __u64 revision;
  1590. } query;
  1591. struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
  1592. __u64 name;
  1593. __u32 prog_fd;
  1594. __u32 :32;
  1595. __aligned_u64 cookie;
  1596. } raw_tracepoint;
  1597. struct { /* anonymous struct for BPF_BTF_LOAD */
  1598. __aligned_u64 btf;
  1599. __aligned_u64 btf_log_buf;
  1600. __u32 btf_size;
  1601. __u32 btf_log_size;
  1602. __u32 btf_log_level;
  1603. /* output: actual total log contents size (including termintaing zero).
  1604. * It could be both larger than original log_size (if log was
  1605. * truncated), or smaller (if log buffer wasn't filled completely).
  1606. */
  1607. __u32 btf_log_true_size;
  1608. __u32 btf_flags;
  1609. /* BPF token FD to use with BPF_BTF_LOAD operation.
  1610. * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.
  1611. */
  1612. __s32 btf_token_fd;
  1613. };
  1614. struct {
  1615. __u32 pid; /* input: pid */
  1616. __u32 fd; /* input: fd */
  1617. __u32 flags; /* input: flags */
  1618. __u32 buf_len; /* input/output: buf len */
  1619. __aligned_u64 buf; /* input/output:
  1620. * tp_name for tracepoint
  1621. * symbol for kprobe
  1622. * filename for uprobe
  1623. */
  1624. __u32 prog_id; /* output: prod_id */
  1625. __u32 fd_type; /* output: BPF_FD_TYPE_* */
  1626. __u64 probe_offset; /* output: probe_offset */
  1627. __u64 probe_addr; /* output: probe_addr */
  1628. } task_fd_query;
  1629. struct { /* struct used by BPF_LINK_CREATE command */
  1630. union {
  1631. __u32 prog_fd; /* eBPF program to attach */
  1632. __u32 map_fd; /* struct_ops to attach */
  1633. };
  1634. union {
  1635. __u32 target_fd; /* target object to attach to or ... */
  1636. __u32 target_ifindex; /* target ifindex */
  1637. };
  1638. __u32 attach_type; /* attach type */
  1639. __u32 flags; /* extra flags */
  1640. union {
  1641. __u32 target_btf_id; /* btf_id of target to attach to */
  1642. struct {
  1643. __aligned_u64 iter_info; /* extra bpf_iter_link_info */
  1644. __u32 iter_info_len; /* iter_info length */
  1645. };
  1646. struct {
  1647. /* black box user-provided value passed through
  1648. * to BPF program at the execution time and
  1649. * accessible through bpf_get_attach_cookie() BPF helper
  1650. */
  1651. __u64 bpf_cookie;
  1652. } perf_event;
  1653. struct {
  1654. __u32 flags;
  1655. __u32 cnt;
  1656. __aligned_u64 syms;
  1657. __aligned_u64 addrs;
  1658. __aligned_u64 cookies;
  1659. } kprobe_multi;
  1660. struct {
  1661. /* this is overlaid with the target_btf_id above. */
  1662. __u32 target_btf_id;
  1663. /* black box user-provided value passed through
  1664. * to BPF program at the execution time and
  1665. * accessible through bpf_get_attach_cookie() BPF helper
  1666. */
  1667. __u64 cookie;
  1668. } tracing;
  1669. struct {
  1670. __u32 pf;
  1671. __u32 hooknum;
  1672. __s32 priority;
  1673. __u32 flags;
  1674. } netfilter;
  1675. struct {
  1676. union {
  1677. __u32 relative_fd;
  1678. __u32 relative_id;
  1679. };
  1680. __u64 expected_revision;
  1681. } tcx;
  1682. struct {
  1683. __aligned_u64 path;
  1684. __aligned_u64 offsets;
  1685. __aligned_u64 ref_ctr_offsets;
  1686. __aligned_u64 cookies;
  1687. __u32 cnt;
  1688. __u32 flags;
  1689. __u32 pid;
  1690. } uprobe_multi;
  1691. struct {
  1692. union {
  1693. __u32 relative_fd;
  1694. __u32 relative_id;
  1695. };
  1696. __u64 expected_revision;
  1697. } netkit;
  1698. };
  1699. } link_create;
  1700. struct { /* struct used by BPF_LINK_UPDATE command */
  1701. __u32 link_fd; /* link fd */
  1702. union {
  1703. /* new program fd to update link with */
  1704. __u32 new_prog_fd;
  1705. /* new struct_ops map fd to update link with */
  1706. __u32 new_map_fd;
  1707. };
  1708. __u32 flags; /* extra flags */
  1709. union {
  1710. /* expected link's program fd; is specified only if
  1711. * BPF_F_REPLACE flag is set in flags.
  1712. */
  1713. __u32 old_prog_fd;
  1714. /* expected link's map fd; is specified only
  1715. * if BPF_F_REPLACE flag is set.
  1716. */
  1717. __u32 old_map_fd;
  1718. };
  1719. } link_update;
  1720. struct {
  1721. __u32 link_fd;
  1722. } link_detach;
  1723. struct { /* struct used by BPF_ENABLE_STATS command */
  1724. __u32 type;
  1725. } enable_stats;
  1726. struct { /* struct used by BPF_ITER_CREATE command */
  1727. __u32 link_fd;
  1728. __u32 flags;
  1729. } iter_create;
  1730. struct { /* struct used by BPF_PROG_BIND_MAP command */
  1731. __u32 prog_fd;
  1732. __u32 map_fd;
  1733. __u32 flags; /* extra flags */
  1734. } prog_bind_map;
  1735. struct { /* struct used by BPF_TOKEN_CREATE command */
  1736. __u32 flags;
  1737. __u32 bpffs_fd;
  1738. } token_create;
  1739. } __attribute__((aligned(8)));
  1740. /* The description below is an attempt at providing documentation to eBPF
  1741. * developers about the multiple available eBPF helper functions. It can be
  1742. * parsed and used to produce a manual page. The workflow is the following,
  1743. * and requires the rst2man utility:
  1744. *
  1745. * $ ./scripts/bpf_doc.py \
  1746. * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
  1747. * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
  1748. * $ man /tmp/bpf-helpers.7
  1749. *
  1750. * Note that in order to produce this external documentation, some RST
  1751. * formatting is used in the descriptions to get "bold" and "italics" in
  1752. * manual pages. Also note that the few trailing white spaces are
  1753. * intentional, removing them would break paragraphs for rst2man.
  1754. *
  1755. * Start of BPF helper function descriptions:
  1756. *
  1757. * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
  1758. * Description
  1759. * Perform a lookup in *map* for an entry associated to *key*.
  1760. * Return
  1761. * Map value associated to *key*, or **NULL** if no entry was
  1762. * found.
  1763. *
  1764. * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
  1765. * Description
  1766. * Add or update the value of the entry associated to *key* in
  1767. * *map* with *value*. *flags* is one of:
  1768. *
  1769. * **BPF_NOEXIST**
  1770. * The entry for *key* must not exist in the map.
  1771. * **BPF_EXIST**
  1772. * The entry for *key* must already exist in the map.
  1773. * **BPF_ANY**
  1774. * No condition on the existence of the entry for *key*.
  1775. *
  1776. * Flag value **BPF_NOEXIST** cannot be used for maps of types
  1777. * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
  1778. * elements always exist), the helper would return an error.
  1779. * Return
  1780. * 0 on success, or a negative error in case of failure.
  1781. *
  1782. * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
  1783. * Description
  1784. * Delete entry with *key* from *map*.
  1785. * Return
  1786. * 0 on success, or a negative error in case of failure.
  1787. *
  1788. * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
  1789. * Description
  1790. * For tracing programs, safely attempt to read *size* bytes from
  1791. * kernel space address *unsafe_ptr* and store the data in *dst*.
  1792. *
  1793. * Generally, use **bpf_probe_read_user**\ () or
  1794. * **bpf_probe_read_kernel**\ () instead.
  1795. * Return
  1796. * 0 on success, or a negative error in case of failure.
  1797. *
  1798. * u64 bpf_ktime_get_ns(void)
  1799. * Description
  1800. * Return the time elapsed since system boot, in nanoseconds.
  1801. * Does not include time the system was suspended.
  1802. * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
  1803. * Return
  1804. * Current *ktime*.
  1805. *
  1806. * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
  1807. * Description
  1808. * This helper is a "printk()-like" facility for debugging. It
  1809. * prints a message defined by format *fmt* (of size *fmt_size*)
  1810. * to file *\/sys/kernel/tracing/trace* from TraceFS, if
  1811. * available. It can take up to three additional **u64**
  1812. * arguments (as an eBPF helpers, the total number of arguments is
  1813. * limited to five).
  1814. *
  1815. * Each time the helper is called, it appends a line to the trace.
  1816. * Lines are discarded while *\/sys/kernel/tracing/trace* is
  1817. * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
  1818. * The format of the trace is customizable, and the exact output
  1819. * one will get depends on the options set in
  1820. * *\/sys/kernel/tracing/trace_options* (see also the
  1821. * *README* file under the same directory). However, it usually
  1822. * defaults to something like:
  1823. *
  1824. * ::
  1825. *
  1826. * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
  1827. *
  1828. * In the above:
  1829. *
  1830. * * ``telnet`` is the name of the current task.
  1831. * * ``470`` is the PID of the current task.
  1832. * * ``001`` is the CPU number on which the task is
  1833. * running.
  1834. * * In ``.N..``, each character refers to a set of
  1835. * options (whether irqs are enabled, scheduling
  1836. * options, whether hard/softirqs are running, level of
  1837. * preempt_disabled respectively). **N** means that
  1838. * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
  1839. * are set.
  1840. * * ``419421.045894`` is a timestamp.
  1841. * * ``0x00000001`` is a fake value used by BPF for the
  1842. * instruction pointer register.
  1843. * * ``<formatted msg>`` is the message formatted with
  1844. * *fmt*.
  1845. *
  1846. * The conversion specifiers supported by *fmt* are similar, but
  1847. * more limited than for printk(). They are **%d**, **%i**,
  1848. * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
  1849. * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
  1850. * of field, padding with zeroes, etc.) is available, and the
  1851. * helper will return **-EINVAL** (but print nothing) if it
  1852. * encounters an unknown specifier.
  1853. *
  1854. * Also, note that **bpf_trace_printk**\ () is slow, and should
  1855. * only be used for debugging purposes. For this reason, a notice
  1856. * block (spanning several lines) is printed to kernel logs and
  1857. * states that the helper should not be used "for production use"
  1858. * the first time this helper is used (or more precisely, when
  1859. * **trace_printk**\ () buffers are allocated). For passing values
  1860. * to user space, perf events should be preferred.
  1861. * Return
  1862. * The number of bytes written to the buffer, or a negative error
  1863. * in case of failure.
  1864. *
  1865. * u32 bpf_get_prandom_u32(void)
  1866. * Description
  1867. * Get a pseudo-random number.
  1868. *
  1869. * From a security point of view, this helper uses its own
  1870. * pseudo-random internal state, and cannot be used to infer the
  1871. * seed of other random functions in the kernel. However, it is
  1872. * essential to note that the generator used by the helper is not
  1873. * cryptographically secure.
  1874. * Return
  1875. * A random 32-bit unsigned value.
  1876. *
  1877. * u32 bpf_get_smp_processor_id(void)
  1878. * Description
  1879. * Get the SMP (symmetric multiprocessing) processor id. Note that
  1880. * all programs run with migration disabled, which means that the
  1881. * SMP processor id is stable during all the execution of the
  1882. * program.
  1883. * Return
  1884. * The SMP id of the processor running the program.
  1885. *
  1886. * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
  1887. * Description
  1888. * Store *len* bytes from address *from* into the packet
  1889. * associated to *skb*, at *offset*. *flags* are a combination of
  1890. * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
  1891. * checksum for the packet after storing the bytes) and
  1892. * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
  1893. * **->swhash** and *skb*\ **->l4hash** to 0).
  1894. *
  1895. * A call to this helper is susceptible to change the underlying
  1896. * packet buffer. Therefore, at load time, all checks on pointers
  1897. * previously done by the verifier are invalidated and must be
  1898. * performed again, if the helper is used in combination with
  1899. * direct packet access.
  1900. * Return
  1901. * 0 on success, or a negative error in case of failure.
  1902. *
  1903. * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
  1904. * Description
  1905. * Recompute the layer 3 (e.g. IP) checksum for the packet
  1906. * associated to *skb*. Computation is incremental, so the helper
  1907. * must know the former value of the header field that was
  1908. * modified (*from*), the new value of this field (*to*), and the
  1909. * number of bytes (2 or 4) for this field, stored in *size*.
  1910. * Alternatively, it is possible to store the difference between
  1911. * the previous and the new values of the header field in *to*, by
  1912. * setting *from* and *size* to 0. For both methods, *offset*
  1913. * indicates the location of the IP checksum within the packet.
  1914. *
  1915. * This helper works in combination with **bpf_csum_diff**\ (),
  1916. * which does not update the checksum in-place, but offers more
  1917. * flexibility and can handle sizes larger than 2 or 4 for the
  1918. * checksum to update.
  1919. *
  1920. * A call to this helper is susceptible to change the underlying
  1921. * packet buffer. Therefore, at load time, all checks on pointers
  1922. * previously done by the verifier are invalidated and must be
  1923. * performed again, if the helper is used in combination with
  1924. * direct packet access.
  1925. * Return
  1926. * 0 on success, or a negative error in case of failure.
  1927. *
  1928. * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
  1929. * Description
  1930. * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
  1931. * packet associated to *skb*. Computation is incremental, so the
  1932. * helper must know the former value of the header field that was
  1933. * modified (*from*), the new value of this field (*to*), and the
  1934. * number of bytes (2 or 4) for this field, stored on the lowest
  1935. * four bits of *flags*. Alternatively, it is possible to store
  1936. * the difference between the previous and the new values of the
  1937. * header field in *to*, by setting *from* and the four lowest
  1938. * bits of *flags* to 0. For both methods, *offset* indicates the
  1939. * location of the IP checksum within the packet. In addition to
  1940. * the size of the field, *flags* can be added (bitwise OR) actual
  1941. * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
  1942. * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
  1943. * for updates resulting in a null checksum the value is set to
  1944. * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
  1945. * the checksum is to be computed against a pseudo-header.
  1946. *
  1947. * This helper works in combination with **bpf_csum_diff**\ (),
  1948. * which does not update the checksum in-place, but offers more
  1949. * flexibility and can handle sizes larger than 2 or 4 for the
  1950. * checksum to update.
  1951. *
  1952. * A call to this helper is susceptible to change the underlying
  1953. * packet buffer. Therefore, at load time, all checks on pointers
  1954. * previously done by the verifier are invalidated and must be
  1955. * performed again, if the helper is used in combination with
  1956. * direct packet access.
  1957. * Return
  1958. * 0 on success, or a negative error in case of failure.
  1959. *
  1960. * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
  1961. * Description
  1962. * This special helper is used to trigger a "tail call", or in
  1963. * other words, to jump into another eBPF program. The same stack
  1964. * frame is used (but values on stack and in registers for the
  1965. * caller are not accessible to the callee). This mechanism allows
  1966. * for program chaining, either for raising the maximum number of
  1967. * available eBPF instructions, or to execute given programs in
  1968. * conditional blocks. For security reasons, there is an upper
  1969. * limit to the number of successive tail calls that can be
  1970. * performed.
  1971. *
  1972. * Upon call of this helper, the program attempts to jump into a
  1973. * program referenced at index *index* in *prog_array_map*, a
  1974. * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
  1975. * *ctx*, a pointer to the context.
  1976. *
  1977. * If the call succeeds, the kernel immediately runs the first
  1978. * instruction of the new program. This is not a function call,
  1979. * and it never returns to the previous program. If the call
  1980. * fails, then the helper has no effect, and the caller continues
  1981. * to run its subsequent instructions. A call can fail if the
  1982. * destination program for the jump does not exist (i.e. *index*
  1983. * is superior to the number of entries in *prog_array_map*), or
  1984. * if the maximum number of tail calls has been reached for this
  1985. * chain of programs. This limit is defined in the kernel by the
  1986. * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
  1987. * which is currently set to 33.
  1988. * Return
  1989. * 0 on success, or a negative error in case of failure.
  1990. *
  1991. * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
  1992. * Description
  1993. * Clone and redirect the packet associated to *skb* to another
  1994. * net device of index *ifindex*. Both ingress and egress
  1995. * interfaces can be used for redirection. The **BPF_F_INGRESS**
  1996. * value in *flags* is used to make the distinction (ingress path
  1997. * is selected if the flag is present, egress path otherwise).
  1998. * This is the only flag supported for now.
  1999. *
  2000. * In comparison with **bpf_redirect**\ () helper,
  2001. * **bpf_clone_redirect**\ () has the associated cost of
  2002. * duplicating the packet buffer, but this can be executed out of
  2003. * the eBPF program. Conversely, **bpf_redirect**\ () is more
  2004. * efficient, but it is handled through an action code where the
  2005. * redirection happens only after the eBPF program has returned.
  2006. *
  2007. * A call to this helper is susceptible to change the underlying
  2008. * packet buffer. Therefore, at load time, all checks on pointers
  2009. * previously done by the verifier are invalidated and must be
  2010. * performed again, if the helper is used in combination with
  2011. * direct packet access.
  2012. * Return
  2013. * 0 on success, or a negative error in case of failure. Positive
  2014. * error indicates a potential drop or congestion in the target
  2015. * device. The particular positive error codes are not defined.
  2016. *
  2017. * u64 bpf_get_current_pid_tgid(void)
  2018. * Description
  2019. * Get the current pid and tgid.
  2020. * Return
  2021. * A 64-bit integer containing the current tgid and pid, and
  2022. * created as such:
  2023. * *current_task*\ **->tgid << 32 \|**
  2024. * *current_task*\ **->pid**.
  2025. *
  2026. * u64 bpf_get_current_uid_gid(void)
  2027. * Description
  2028. * Get the current uid and gid.
  2029. * Return
  2030. * A 64-bit integer containing the current GID and UID, and
  2031. * created as such: *current_gid* **<< 32 \|** *current_uid*.
  2032. *
  2033. * long bpf_get_current_comm(void *buf, u32 size_of_buf)
  2034. * Description
  2035. * Copy the **comm** attribute of the current task into *buf* of
  2036. * *size_of_buf*. The **comm** attribute contains the name of
  2037. * the executable (excluding the path) for the current task. The
  2038. * *size_of_buf* must be strictly positive. On success, the
  2039. * helper makes sure that the *buf* is NUL-terminated. On failure,
  2040. * it is filled with zeroes.
  2041. * Return
  2042. * 0 on success, or a negative error in case of failure.
  2043. *
  2044. * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
  2045. * Description
  2046. * Retrieve the classid for the current task, i.e. for the net_cls
  2047. * cgroup to which *skb* belongs.
  2048. *
  2049. * This helper can be used on TC egress path, but not on ingress.
  2050. *
  2051. * The net_cls cgroup provides an interface to tag network packets
  2052. * based on a user-provided identifier for all traffic coming from
  2053. * the tasks belonging to the related cgroup. See also the related
  2054. * kernel documentation, available from the Linux sources in file
  2055. * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
  2056. *
  2057. * The Linux kernel has two versions for cgroups: there are
  2058. * cgroups v1 and cgroups v2. Both are available to users, who can
  2059. * use a mixture of them, but note that the net_cls cgroup is for
  2060. * cgroup v1 only. This makes it incompatible with BPF programs
  2061. * run on cgroups, which is a cgroup-v2-only feature (a socket can
  2062. * only hold data for one version of cgroups at a time).
  2063. *
  2064. * This helper is only available is the kernel was compiled with
  2065. * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
  2066. * "**y**" or to "**m**".
  2067. * Return
  2068. * The classid, or 0 for the default unconfigured classid.
  2069. *
  2070. * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
  2071. * Description
  2072. * Push a *vlan_tci* (VLAN tag control information) of protocol
  2073. * *vlan_proto* to the packet associated to *skb*, then update
  2074. * the checksum. Note that if *vlan_proto* is different from
  2075. * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
  2076. * be **ETH_P_8021Q**.
  2077. *
  2078. * A call to this helper is susceptible to change the underlying
  2079. * packet buffer. Therefore, at load time, all checks on pointers
  2080. * previously done by the verifier are invalidated and must be
  2081. * performed again, if the helper is used in combination with
  2082. * direct packet access.
  2083. * Return
  2084. * 0 on success, or a negative error in case of failure.
  2085. *
  2086. * long bpf_skb_vlan_pop(struct sk_buff *skb)
  2087. * Description
  2088. * Pop a VLAN header from the packet associated to *skb*.
  2089. *
  2090. * A call to this helper is susceptible to change the underlying
  2091. * packet buffer. Therefore, at load time, all checks on pointers
  2092. * previously done by the verifier are invalidated and must be
  2093. * performed again, if the helper is used in combination with
  2094. * direct packet access.
  2095. * Return
  2096. * 0 on success, or a negative error in case of failure.
  2097. *
  2098. * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  2099. * Description
  2100. * Get tunnel metadata. This helper takes a pointer *key* to an
  2101. * empty **struct bpf_tunnel_key** of **size**, that will be
  2102. * filled with tunnel metadata for the packet associated to *skb*.
  2103. * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
  2104. * indicates that the tunnel is based on IPv6 protocol instead of
  2105. * IPv4.
  2106. *
  2107. * The **struct bpf_tunnel_key** is an object that generalizes the
  2108. * principal parameters used by various tunneling protocols into a
  2109. * single struct. This way, it can be used to easily make a
  2110. * decision based on the contents of the encapsulation header,
  2111. * "summarized" in this struct. In particular, it holds the IP
  2112. * address of the remote end (IPv4 or IPv6, depending on the case)
  2113. * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
  2114. * this struct exposes the *key*\ **->tunnel_id**, which is
  2115. * generally mapped to a VNI (Virtual Network Identifier), making
  2116. * it programmable together with the **bpf_skb_set_tunnel_key**\
  2117. * () helper.
  2118. *
  2119. * Let's imagine that the following code is part of a program
  2120. * attached to the TC ingress interface, on one end of a GRE
  2121. * tunnel, and is supposed to filter out all messages coming from
  2122. * remote ends with IPv4 address other than 10.0.0.1:
  2123. *
  2124. * ::
  2125. *
  2126. * int ret;
  2127. * struct bpf_tunnel_key key = {};
  2128. *
  2129. * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
  2130. * if (ret < 0)
  2131. * return TC_ACT_SHOT; // drop packet
  2132. *
  2133. * if (key.remote_ipv4 != 0x0a000001)
  2134. * return TC_ACT_SHOT; // drop packet
  2135. *
  2136. * return TC_ACT_OK; // accept packet
  2137. *
  2138. * This interface can also be used with all encapsulation devices
  2139. * that can operate in "collect metadata" mode: instead of having
  2140. * one network device per specific configuration, the "collect
  2141. * metadata" mode only requires a single device where the
  2142. * configuration can be extracted from this helper.
  2143. *
  2144. * This can be used together with various tunnels such as VXLan,
  2145. * Geneve, GRE or IP in IP (IPIP).
  2146. * Return
  2147. * 0 on success, or a negative error in case of failure.
  2148. *
  2149. * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  2150. * Description
  2151. * Populate tunnel metadata for packet associated to *skb.* The
  2152. * tunnel metadata is set to the contents of *key*, of *size*. The
  2153. * *flags* can be set to a combination of the following values:
  2154. *
  2155. * **BPF_F_TUNINFO_IPV6**
  2156. * Indicate that the tunnel is based on IPv6 protocol
  2157. * instead of IPv4.
  2158. * **BPF_F_ZERO_CSUM_TX**
  2159. * For IPv4 packets, add a flag to tunnel metadata
  2160. * indicating that checksum computation should be skipped
  2161. * and checksum set to zeroes.
  2162. * **BPF_F_DONT_FRAGMENT**
  2163. * Add a flag to tunnel metadata indicating that the
  2164. * packet should not be fragmented.
  2165. * **BPF_F_SEQ_NUMBER**
  2166. * Add a flag to tunnel metadata indicating that a
  2167. * sequence number should be added to tunnel header before
  2168. * sending the packet. This flag was added for GRE
  2169. * encapsulation, but might be used with other protocols
  2170. * as well in the future.
  2171. * **BPF_F_NO_TUNNEL_KEY**
  2172. * Add a flag to tunnel metadata indicating that no tunnel
  2173. * key should be set in the resulting tunnel header.
  2174. *
  2175. * Here is a typical usage on the transmit path:
  2176. *
  2177. * ::
  2178. *
  2179. * struct bpf_tunnel_key key;
  2180. * populate key ...
  2181. * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
  2182. * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
  2183. *
  2184. * See also the description of the **bpf_skb_get_tunnel_key**\ ()
  2185. * helper for additional information.
  2186. * Return
  2187. * 0 on success, or a negative error in case of failure.
  2188. *
  2189. * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
  2190. * Description
  2191. * Read the value of a perf event counter. This helper relies on a
  2192. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
  2193. * the perf event counter is selected when *map* is updated with
  2194. * perf event file descriptors. The *map* is an array whose size
  2195. * is the number of available CPUs, and each cell contains a value
  2196. * relative to one CPU. The value to retrieve is indicated by
  2197. * *flags*, that contains the index of the CPU to look up, masked
  2198. * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  2199. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  2200. * current CPU should be retrieved.
  2201. *
  2202. * Note that before Linux 4.13, only hardware perf event can be
  2203. * retrieved.
  2204. *
  2205. * Also, be aware that the newer helper
  2206. * **bpf_perf_event_read_value**\ () is recommended over
  2207. * **bpf_perf_event_read**\ () in general. The latter has some ABI
  2208. * quirks where error and counter value are used as a return code
  2209. * (which is wrong to do since ranges may overlap). This issue is
  2210. * fixed with **bpf_perf_event_read_value**\ (), which at the same
  2211. * time provides more features over the **bpf_perf_event_read**\
  2212. * () interface. Please refer to the description of
  2213. * **bpf_perf_event_read_value**\ () for details.
  2214. * Return
  2215. * The value of the perf event counter read from the map, or a
  2216. * negative error code in case of failure.
  2217. *
  2218. * long bpf_redirect(u32 ifindex, u64 flags)
  2219. * Description
  2220. * Redirect the packet to another net device of index *ifindex*.
  2221. * This helper is somewhat similar to **bpf_clone_redirect**\
  2222. * (), except that the packet is not cloned, which provides
  2223. * increased performance.
  2224. *
  2225. * Except for XDP, both ingress and egress interfaces can be used
  2226. * for redirection. The **BPF_F_INGRESS** value in *flags* is used
  2227. * to make the distinction (ingress path is selected if the flag
  2228. * is present, egress path otherwise). Currently, XDP only
  2229. * supports redirection to the egress interface, and accepts no
  2230. * flag at all.
  2231. *
  2232. * The same effect can also be attained with the more generic
  2233. * **bpf_redirect_map**\ (), which uses a BPF map to store the
  2234. * redirect target instead of providing it directly to the helper.
  2235. * Return
  2236. * For XDP, the helper returns **XDP_REDIRECT** on success or
  2237. * **XDP_ABORTED** on error. For other program types, the values
  2238. * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
  2239. * error.
  2240. *
  2241. * u32 bpf_get_route_realm(struct sk_buff *skb)
  2242. * Description
  2243. * Retrieve the realm or the route, that is to say the
  2244. * **tclassid** field of the destination for the *skb*. The
  2245. * identifier retrieved is a user-provided tag, similar to the
  2246. * one used with the net_cls cgroup (see description for
  2247. * **bpf_get_cgroup_classid**\ () helper), but here this tag is
  2248. * held by a route (a destination entry), not by a task.
  2249. *
  2250. * Retrieving this identifier works with the clsact TC egress hook
  2251. * (see also **tc-bpf(8)**), or alternatively on conventional
  2252. * classful egress qdiscs, but not on TC ingress path. In case of
  2253. * clsact TC egress hook, this has the advantage that, internally,
  2254. * the destination entry has not been dropped yet in the transmit
  2255. * path. Therefore, the destination entry does not need to be
  2256. * artificially held via **netif_keep_dst**\ () for a classful
  2257. * qdisc until the *skb* is freed.
  2258. *
  2259. * This helper is available only if the kernel was compiled with
  2260. * **CONFIG_IP_ROUTE_CLASSID** configuration option.
  2261. * Return
  2262. * The realm of the route for the packet associated to *skb*, or 0
  2263. * if none was found.
  2264. *
  2265. * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  2266. * Description
  2267. * Write raw *data* blob into a special BPF perf event held by
  2268. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  2269. * event must have the following attributes: **PERF_SAMPLE_RAW**
  2270. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  2271. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  2272. *
  2273. * The *flags* are used to indicate the index in *map* for which
  2274. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  2275. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  2276. * to indicate that the index of the current CPU core should be
  2277. * used.
  2278. *
  2279. * The value to write, of *size*, is passed through eBPF stack and
  2280. * pointed by *data*.
  2281. *
  2282. * The context of the program *ctx* needs also be passed to the
  2283. * helper.
  2284. *
  2285. * On user space, a program willing to read the values needs to
  2286. * call **perf_event_open**\ () on the perf event (either for
  2287. * one or for all CPUs) and to store the file descriptor into the
  2288. * *map*. This must be done before the eBPF program can send data
  2289. * into it. An example is available in file
  2290. * *samples/bpf/trace_output_user.c* in the Linux kernel source
  2291. * tree (the eBPF program counterpart is in
  2292. * *samples/bpf/trace_output_kern.c*).
  2293. *
  2294. * **bpf_perf_event_output**\ () achieves better performance
  2295. * than **bpf_trace_printk**\ () for sharing data with user
  2296. * space, and is much better suitable for streaming data from eBPF
  2297. * programs.
  2298. *
  2299. * Note that this helper is not restricted to tracing use cases
  2300. * and can be used with programs attached to TC or XDP as well,
  2301. * where it allows for passing data to user space listeners. Data
  2302. * can be:
  2303. *
  2304. * * Only custom structs,
  2305. * * Only the packet payload, or
  2306. * * A combination of both.
  2307. * Return
  2308. * 0 on success, or a negative error in case of failure.
  2309. *
  2310. * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
  2311. * Description
  2312. * This helper was provided as an easy way to load data from a
  2313. * packet. It can be used to load *len* bytes from *offset* from
  2314. * the packet associated to *skb*, into the buffer pointed by
  2315. * *to*.
  2316. *
  2317. * Since Linux 4.7, usage of this helper has mostly been replaced
  2318. * by "direct packet access", enabling packet data to be
  2319. * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
  2320. * pointing respectively to the first byte of packet data and to
  2321. * the byte after the last byte of packet data. However, it
  2322. * remains useful if one wishes to read large quantities of data
  2323. * at once from a packet into the eBPF stack.
  2324. * Return
  2325. * 0 on success, or a negative error in case of failure.
  2326. *
  2327. * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
  2328. * Description
  2329. * Walk a user or a kernel stack and return its id. To achieve
  2330. * this, the helper needs *ctx*, which is a pointer to the context
  2331. * on which the tracing program is executed, and a pointer to a
  2332. * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
  2333. *
  2334. * The last argument, *flags*, holds the number of stack frames to
  2335. * skip (from 0 to 255), masked with
  2336. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  2337. * a combination of the following flags:
  2338. *
  2339. * **BPF_F_USER_STACK**
  2340. * Collect a user space stack instead of a kernel stack.
  2341. * **BPF_F_FAST_STACK_CMP**
  2342. * Compare stacks by hash only.
  2343. * **BPF_F_REUSE_STACKID**
  2344. * If two different stacks hash into the same *stackid*,
  2345. * discard the old one.
  2346. *
  2347. * The stack id retrieved is a 32 bit long integer handle which
  2348. * can be further combined with other data (including other stack
  2349. * ids) and used as a key into maps. This can be useful for
  2350. * generating a variety of graphs (such as flame graphs or off-cpu
  2351. * graphs).
  2352. *
  2353. * For walking a stack, this helper is an improvement over
  2354. * **bpf_probe_read**\ (), which can be used with unrolled loops
  2355. * but is not efficient and consumes a lot of eBPF instructions.
  2356. * Instead, **bpf_get_stackid**\ () can collect up to
  2357. * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
  2358. * this limit can be controlled with the **sysctl** program, and
  2359. * that it should be manually increased in order to profile long
  2360. * user stacks (such as stacks for Java programs). To do so, use:
  2361. *
  2362. * ::
  2363. *
  2364. * # sysctl kernel.perf_event_max_stack=<new value>
  2365. * Return
  2366. * The positive or null stack id on success, or a negative error
  2367. * in case of failure.
  2368. *
  2369. * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
  2370. * Description
  2371. * Compute a checksum difference, from the raw buffer pointed by
  2372. * *from*, of length *from_size* (that must be a multiple of 4),
  2373. * towards the raw buffer pointed by *to*, of size *to_size*
  2374. * (same remark). An optional *seed* can be added to the value
  2375. * (this can be cascaded, the seed may come from a previous call
  2376. * to the helper).
  2377. *
  2378. * This is flexible enough to be used in several ways:
  2379. *
  2380. * * With *from_size* == 0, *to_size* > 0 and *seed* set to
  2381. * checksum, it can be used when pushing new data.
  2382. * * With *from_size* > 0, *to_size* == 0 and *seed* set to
  2383. * checksum, it can be used when removing data from a packet.
  2384. * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
  2385. * can be used to compute a diff. Note that *from_size* and
  2386. * *to_size* do not need to be equal.
  2387. *
  2388. * This helper can be used in combination with
  2389. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
  2390. * which one can feed in the difference computed with
  2391. * **bpf_csum_diff**\ ().
  2392. * Return
  2393. * The checksum result, or a negative error code in case of
  2394. * failure.
  2395. *
  2396. * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
  2397. * Description
  2398. * Retrieve tunnel options metadata for the packet associated to
  2399. * *skb*, and store the raw tunnel option data to the buffer *opt*
  2400. * of *size*.
  2401. *
  2402. * This helper can be used with encapsulation devices that can
  2403. * operate in "collect metadata" mode (please refer to the related
  2404. * note in the description of **bpf_skb_get_tunnel_key**\ () for
  2405. * more details). A particular example where this can be used is
  2406. * in combination with the Geneve encapsulation protocol, where it
  2407. * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
  2408. * and retrieving arbitrary TLVs (Type-Length-Value headers) from
  2409. * the eBPF program. This allows for full customization of these
  2410. * headers.
  2411. * Return
  2412. * The size of the option data retrieved.
  2413. *
  2414. * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
  2415. * Description
  2416. * Set tunnel options metadata for the packet associated to *skb*
  2417. * to the option data contained in the raw buffer *opt* of *size*.
  2418. *
  2419. * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
  2420. * helper for additional information.
  2421. * Return
  2422. * 0 on success, or a negative error in case of failure.
  2423. *
  2424. * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
  2425. * Description
  2426. * Change the protocol of the *skb* to *proto*. Currently
  2427. * supported are transition from IPv4 to IPv6, and from IPv6 to
  2428. * IPv4. The helper takes care of the groundwork for the
  2429. * transition, including resizing the socket buffer. The eBPF
  2430. * program is expected to fill the new headers, if any, via
  2431. * **skb_store_bytes**\ () and to recompute the checksums with
  2432. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
  2433. * (). The main case for this helper is to perform NAT64
  2434. * operations out of an eBPF program.
  2435. *
  2436. * Internally, the GSO type is marked as dodgy so that headers are
  2437. * checked and segments are recalculated by the GSO/GRO engine.
  2438. * The size for GSO target is adapted as well.
  2439. *
  2440. * All values for *flags* are reserved for future usage, and must
  2441. * be left at zero.
  2442. *
  2443. * A call to this helper is susceptible to change the underlying
  2444. * packet buffer. Therefore, at load time, all checks on pointers
  2445. * previously done by the verifier are invalidated and must be
  2446. * performed again, if the helper is used in combination with
  2447. * direct packet access.
  2448. * Return
  2449. * 0 on success, or a negative error in case of failure.
  2450. *
  2451. * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
  2452. * Description
  2453. * Change the packet type for the packet associated to *skb*. This
  2454. * comes down to setting *skb*\ **->pkt_type** to *type*, except
  2455. * the eBPF program does not have a write access to *skb*\
  2456. * **->pkt_type** beside this helper. Using a helper here allows
  2457. * for graceful handling of errors.
  2458. *
  2459. * The major use case is to change incoming *skb*s to
  2460. * **PACKET_HOST** in a programmatic way instead of having to
  2461. * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
  2462. * example.
  2463. *
  2464. * Note that *type* only allows certain values. At this time, they
  2465. * are:
  2466. *
  2467. * **PACKET_HOST**
  2468. * Packet is for us.
  2469. * **PACKET_BROADCAST**
  2470. * Send packet to all.
  2471. * **PACKET_MULTICAST**
  2472. * Send packet to group.
  2473. * **PACKET_OTHERHOST**
  2474. * Send packet to someone else.
  2475. * Return
  2476. * 0 on success, or a negative error in case of failure.
  2477. *
  2478. * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
  2479. * Description
  2480. * Check whether *skb* is a descendant of the cgroup2 held by
  2481. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  2482. * Return
  2483. * The return value depends on the result of the test, and can be:
  2484. *
  2485. * * 0, if the *skb* failed the cgroup2 descendant test.
  2486. * * 1, if the *skb* succeeded the cgroup2 descendant test.
  2487. * * A negative error code, if an error occurred.
  2488. *
  2489. * u32 bpf_get_hash_recalc(struct sk_buff *skb)
  2490. * Description
  2491. * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
  2492. * not set, in particular if the hash was cleared due to mangling,
  2493. * recompute this hash. Later accesses to the hash can be done
  2494. * directly with *skb*\ **->hash**.
  2495. *
  2496. * Calling **bpf_set_hash_invalid**\ (), changing a packet
  2497. * prototype with **bpf_skb_change_proto**\ (), or calling
  2498. * **bpf_skb_store_bytes**\ () with the
  2499. * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
  2500. * the hash and to trigger a new computation for the next call to
  2501. * **bpf_get_hash_recalc**\ ().
  2502. * Return
  2503. * The 32-bit hash.
  2504. *
  2505. * u64 bpf_get_current_task(void)
  2506. * Description
  2507. * Get the current task.
  2508. * Return
  2509. * A pointer to the current task struct.
  2510. *
  2511. * long bpf_probe_write_user(void *dst, const void *src, u32 len)
  2512. * Description
  2513. * Attempt in a safe way to write *len* bytes from the buffer
  2514. * *src* to *dst* in memory. It only works for threads that are in
  2515. * user context, and *dst* must be a valid user space address.
  2516. *
  2517. * This helper should not be used to implement any kind of
  2518. * security mechanism because of TOC-TOU attacks, but rather to
  2519. * debug, divert, and manipulate execution of semi-cooperative
  2520. * processes.
  2521. *
  2522. * Keep in mind that this feature is meant for experiments, and it
  2523. * has a risk of crashing the system and running programs.
  2524. * Therefore, when an eBPF program using this helper is attached,
  2525. * a warning including PID and process name is printed to kernel
  2526. * logs.
  2527. * Return
  2528. * 0 on success, or a negative error in case of failure.
  2529. *
  2530. * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
  2531. * Description
  2532. * Check whether the probe is being run is the context of a given
  2533. * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
  2534. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  2535. * Return
  2536. * The return value depends on the result of the test, and can be:
  2537. *
  2538. * * 1, if current task belongs to the cgroup2.
  2539. * * 0, if current task does not belong to the cgroup2.
  2540. * * A negative error code, if an error occurred.
  2541. *
  2542. * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
  2543. * Description
  2544. * Resize (trim or grow) the packet associated to *skb* to the
  2545. * new *len*. The *flags* are reserved for future usage, and must
  2546. * be left at zero.
  2547. *
  2548. * The basic idea is that the helper performs the needed work to
  2549. * change the size of the packet, then the eBPF program rewrites
  2550. * the rest via helpers like **bpf_skb_store_bytes**\ (),
  2551. * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
  2552. * and others. This helper is a slow path utility intended for
  2553. * replies with control messages. And because it is targeted for
  2554. * slow path, the helper itself can afford to be slow: it
  2555. * implicitly linearizes, unclones and drops offloads from the
  2556. * *skb*.
  2557. *
  2558. * A call to this helper is susceptible to change the underlying
  2559. * packet buffer. Therefore, at load time, all checks on pointers
  2560. * previously done by the verifier are invalidated and must be
  2561. * performed again, if the helper is used in combination with
  2562. * direct packet access.
  2563. * Return
  2564. * 0 on success, or a negative error in case of failure.
  2565. *
  2566. * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
  2567. * Description
  2568. * Pull in non-linear data in case the *skb* is non-linear and not
  2569. * all of *len* are part of the linear section. Make *len* bytes
  2570. * from *skb* readable and writable. If a zero value is passed for
  2571. * *len*, then all bytes in the linear part of *skb* will be made
  2572. * readable and writable.
  2573. *
  2574. * This helper is only needed for reading and writing with direct
  2575. * packet access.
  2576. *
  2577. * For direct packet access, testing that offsets to access
  2578. * are within packet boundaries (test on *skb*\ **->data_end**) is
  2579. * susceptible to fail if offsets are invalid, or if the requested
  2580. * data is in non-linear parts of the *skb*. On failure the
  2581. * program can just bail out, or in the case of a non-linear
  2582. * buffer, use a helper to make the data available. The
  2583. * **bpf_skb_load_bytes**\ () helper is a first solution to access
  2584. * the data. Another one consists in using **bpf_skb_pull_data**
  2585. * to pull in once the non-linear parts, then retesting and
  2586. * eventually access the data.
  2587. *
  2588. * At the same time, this also makes sure the *skb* is uncloned,
  2589. * which is a necessary condition for direct write. As this needs
  2590. * to be an invariant for the write part only, the verifier
  2591. * detects writes and adds a prologue that is calling
  2592. * **bpf_skb_pull_data()** to effectively unclone the *skb* from
  2593. * the very beginning in case it is indeed cloned.
  2594. *
  2595. * A call to this helper is susceptible to change the underlying
  2596. * packet buffer. Therefore, at load time, all checks on pointers
  2597. * previously done by the verifier are invalidated and must be
  2598. * performed again, if the helper is used in combination with
  2599. * direct packet access.
  2600. * Return
  2601. * 0 on success, or a negative error in case of failure.
  2602. *
  2603. * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
  2604. * Description
  2605. * Add the checksum *csum* into *skb*\ **->csum** in case the
  2606. * driver has supplied a checksum for the entire packet into that
  2607. * field. Return an error otherwise. This helper is intended to be
  2608. * used in combination with **bpf_csum_diff**\ (), in particular
  2609. * when the checksum needs to be updated after data has been
  2610. * written into the packet through direct packet access.
  2611. * Return
  2612. * The checksum on success, or a negative error code in case of
  2613. * failure.
  2614. *
  2615. * void bpf_set_hash_invalid(struct sk_buff *skb)
  2616. * Description
  2617. * Invalidate the current *skb*\ **->hash**. It can be used after
  2618. * mangling on headers through direct packet access, in order to
  2619. * indicate that the hash is outdated and to trigger a
  2620. * recalculation the next time the kernel tries to access this
  2621. * hash or when the **bpf_get_hash_recalc**\ () helper is called.
  2622. * Return
  2623. * void.
  2624. *
  2625. * long bpf_get_numa_node_id(void)
  2626. * Description
  2627. * Return the id of the current NUMA node. The primary use case
  2628. * for this helper is the selection of sockets for the local NUMA
  2629. * node, when the program is attached to sockets using the
  2630. * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
  2631. * but the helper is also available to other eBPF program types,
  2632. * similarly to **bpf_get_smp_processor_id**\ ().
  2633. * Return
  2634. * The id of current NUMA node.
  2635. *
  2636. * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
  2637. * Description
  2638. * Grows headroom of packet associated to *skb* and adjusts the
  2639. * offset of the MAC header accordingly, adding *len* bytes of
  2640. * space. It automatically extends and reallocates memory as
  2641. * required.
  2642. *
  2643. * This helper can be used on a layer 3 *skb* to push a MAC header
  2644. * for redirection into a layer 2 device.
  2645. *
  2646. * All values for *flags* are reserved for future usage, and must
  2647. * be left at zero.
  2648. *
  2649. * A call to this helper is susceptible to change the underlying
  2650. * packet buffer. Therefore, at load time, all checks on pointers
  2651. * previously done by the verifier are invalidated and must be
  2652. * performed again, if the helper is used in combination with
  2653. * direct packet access.
  2654. * Return
  2655. * 0 on success, or a negative error in case of failure.
  2656. *
  2657. * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
  2658. * Description
  2659. * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
  2660. * it is possible to use a negative value for *delta*. This helper
  2661. * can be used to prepare the packet for pushing or popping
  2662. * headers.
  2663. *
  2664. * A call to this helper is susceptible to change the underlying
  2665. * packet buffer. Therefore, at load time, all checks on pointers
  2666. * previously done by the verifier are invalidated and must be
  2667. * performed again, if the helper is used in combination with
  2668. * direct packet access.
  2669. * Return
  2670. * 0 on success, or a negative error in case of failure.
  2671. *
  2672. * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
  2673. * Description
  2674. * Copy a NUL terminated string from an unsafe kernel address
  2675. * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
  2676. * more details.
  2677. *
  2678. * Generally, use **bpf_probe_read_user_str**\ () or
  2679. * **bpf_probe_read_kernel_str**\ () instead.
  2680. * Return
  2681. * On success, the strictly positive length of the string,
  2682. * including the trailing NUL character. On error, a negative
  2683. * value.
  2684. *
  2685. * u64 bpf_get_socket_cookie(struct sk_buff *skb)
  2686. * Description
  2687. * If the **struct sk_buff** pointed by *skb* has a known socket,
  2688. * retrieve the cookie (generated by the kernel) of this socket.
  2689. * If no cookie has been set yet, generate a new cookie. Once
  2690. * generated, the socket cookie remains stable for the life of the
  2691. * socket. This helper can be useful for monitoring per socket
  2692. * networking traffic statistics as it provides a global socket
  2693. * identifier that can be assumed unique.
  2694. * Return
  2695. * A 8-byte long unique number on success, or 0 if the socket
  2696. * field is missing inside *skb*.
  2697. *
  2698. * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
  2699. * Description
  2700. * Equivalent to bpf_get_socket_cookie() helper that accepts
  2701. * *skb*, but gets socket from **struct bpf_sock_addr** context.
  2702. * Return
  2703. * A 8-byte long unique number.
  2704. *
  2705. * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
  2706. * Description
  2707. * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
  2708. * *skb*, but gets socket from **struct bpf_sock_ops** context.
  2709. * Return
  2710. * A 8-byte long unique number.
  2711. *
  2712. * u64 bpf_get_socket_cookie(struct sock *sk)
  2713. * Description
  2714. * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
  2715. * *sk*, but gets socket from a BTF **struct sock**. This helper
  2716. * also works for sleepable programs.
  2717. * Return
  2718. * A 8-byte long unique number or 0 if *sk* is NULL.
  2719. *
  2720. * u32 bpf_get_socket_uid(struct sk_buff *skb)
  2721. * Description
  2722. * Get the owner UID of the socked associated to *skb*.
  2723. * Return
  2724. * The owner UID of the socket associated to *skb*. If the socket
  2725. * is **NULL**, or if it is not a full socket (i.e. if it is a
  2726. * time-wait or a request socket instead), **overflowuid** value
  2727. * is returned (note that **overflowuid** might also be the actual
  2728. * UID value for the socket).
  2729. *
  2730. * long bpf_set_hash(struct sk_buff *skb, u32 hash)
  2731. * Description
  2732. * Set the full hash for *skb* (set the field *skb*\ **->hash**)
  2733. * to value *hash*.
  2734. * Return
  2735. * 0
  2736. *
  2737. * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
  2738. * Description
  2739. * Emulate a call to **setsockopt()** on the socket associated to
  2740. * *bpf_socket*, which must be a full socket. The *level* at
  2741. * which the option resides and the name *optname* of the option
  2742. * must be specified, see **setsockopt(2)** for more information.
  2743. * The option value of length *optlen* is pointed by *optval*.
  2744. *
  2745. * *bpf_socket* should be one of the following:
  2746. *
  2747. * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
  2748. * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
  2749. * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
  2750. *
  2751. * This helper actually implements a subset of **setsockopt()**.
  2752. * It supports the following *level*\ s:
  2753. *
  2754. * * **SOL_SOCKET**, which supports the following *optname*\ s:
  2755. * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
  2756. * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
  2757. * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
  2758. * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
  2759. * * **IPPROTO_TCP**, which supports the following *optname*\ s:
  2760. * **TCP_CONGESTION**, **TCP_BPF_IW**,
  2761. * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
  2762. * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
  2763. * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
  2764. * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
  2765. * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
  2766. * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.
  2767. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  2768. * * **IPPROTO_IPV6**, which supports the following *optname*\ s:
  2769. * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
  2770. * Return
  2771. * 0 on success, or a negative error in case of failure.
  2772. *
  2773. * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
  2774. * Description
  2775. * Grow or shrink the room for data in the packet associated to
  2776. * *skb* by *len_diff*, and according to the selected *mode*.
  2777. *
  2778. * By default, the helper will reset any offloaded checksum
  2779. * indicator of the skb to CHECKSUM_NONE. This can be avoided
  2780. * by the following flag:
  2781. *
  2782. * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
  2783. * checksum data of the skb to CHECKSUM_NONE.
  2784. *
  2785. * There are two supported modes at this time:
  2786. *
  2787. * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
  2788. * (room space is added or removed between the layer 2 and
  2789. * layer 3 headers).
  2790. *
  2791. * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  2792. * (room space is added or removed between the layer 3 and
  2793. * layer 4 headers).
  2794. *
  2795. * The following flags are supported at this time:
  2796. *
  2797. * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
  2798. * Adjusting mss in this way is not allowed for datagrams.
  2799. *
  2800. * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
  2801. * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
  2802. * Any new space is reserved to hold a tunnel header.
  2803. * Configure skb offsets and other fields accordingly.
  2804. *
  2805. * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
  2806. * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
  2807. * Use with ENCAP_L3 flags to further specify the tunnel type.
  2808. *
  2809. * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
  2810. * Use with ENCAP_L3/L4 flags to further specify the tunnel
  2811. * type; *len* is the length of the inner MAC header.
  2812. *
  2813. * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
  2814. * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
  2815. * L2 type as Ethernet.
  2816. *
  2817. * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
  2818. * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
  2819. * Indicate the new IP header version after decapsulating the outer
  2820. * IP header. Used when the inner and outer IP versions are different.
  2821. *
  2822. * A call to this helper is susceptible to change the underlying
  2823. * packet buffer. Therefore, at load time, all checks on pointers
  2824. * previously done by the verifier are invalidated and must be
  2825. * performed again, if the helper is used in combination with
  2826. * direct packet access.
  2827. * Return
  2828. * 0 on success, or a negative error in case of failure.
  2829. *
  2830. * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
  2831. * Description
  2832. * Redirect the packet to the endpoint referenced by *map* at
  2833. * index *key*. Depending on its type, this *map* can contain
  2834. * references to net devices (for forwarding packets through other
  2835. * ports), or to CPUs (for redirecting XDP frames to another CPU;
  2836. * but this is only implemented for native XDP (with driver
  2837. * support) as of this writing).
  2838. *
  2839. * The lower two bits of *flags* are used as the return code if
  2840. * the map lookup fails. This is so that the return value can be
  2841. * one of the XDP program return codes up to **XDP_TX**, as chosen
  2842. * by the caller. The higher bits of *flags* can be set to
  2843. * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
  2844. *
  2845. * With BPF_F_BROADCAST the packet will be broadcasted to all the
  2846. * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
  2847. * interface will be excluded when do broadcasting.
  2848. *
  2849. * See also **bpf_redirect**\ (), which only supports redirecting
  2850. * to an ifindex, but doesn't require a map to do so.
  2851. * Return
  2852. * **XDP_REDIRECT** on success, or the value of the two lower bits
  2853. * of the *flags* argument on error.
  2854. *
  2855. * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
  2856. * Description
  2857. * Redirect the packet to the socket referenced by *map* (of type
  2858. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  2859. * egress interfaces can be used for redirection. The
  2860. * **BPF_F_INGRESS** value in *flags* is used to make the
  2861. * distinction (ingress path is selected if the flag is present,
  2862. * egress path otherwise). This is the only flag supported for now.
  2863. * Return
  2864. * **SK_PASS** on success, or **SK_DROP** on error.
  2865. *
  2866. * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  2867. * Description
  2868. * Add an entry to, or update a *map* referencing sockets. The
  2869. * *skops* is used as a new value for the entry associated to
  2870. * *key*. *flags* is one of:
  2871. *
  2872. * **BPF_NOEXIST**
  2873. * The entry for *key* must not exist in the map.
  2874. * **BPF_EXIST**
  2875. * The entry for *key* must already exist in the map.
  2876. * **BPF_ANY**
  2877. * No condition on the existence of the entry for *key*.
  2878. *
  2879. * If the *map* has eBPF programs (parser and verdict), those will
  2880. * be inherited by the socket being added. If the socket is
  2881. * already attached to eBPF programs, this results in an error.
  2882. * Return
  2883. * 0 on success, or a negative error in case of failure.
  2884. *
  2885. * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
  2886. * Description
  2887. * Adjust the address pointed by *xdp_md*\ **->data_meta** by
  2888. * *delta* (which can be positive or negative). Note that this
  2889. * operation modifies the address stored in *xdp_md*\ **->data**,
  2890. * so the latter must be loaded only after the helper has been
  2891. * called.
  2892. *
  2893. * The use of *xdp_md*\ **->data_meta** is optional and programs
  2894. * are not required to use it. The rationale is that when the
  2895. * packet is processed with XDP (e.g. as DoS filter), it is
  2896. * possible to push further meta data along with it before passing
  2897. * to the stack, and to give the guarantee that an ingress eBPF
  2898. * program attached as a TC classifier on the same device can pick
  2899. * this up for further post-processing. Since TC works with socket
  2900. * buffers, it remains possible to set from XDP the **mark** or
  2901. * **priority** pointers, or other pointers for the socket buffer.
  2902. * Having this scratch space generic and programmable allows for
  2903. * more flexibility as the user is free to store whatever meta
  2904. * data they need.
  2905. *
  2906. * A call to this helper is susceptible to change the underlying
  2907. * packet buffer. Therefore, at load time, all checks on pointers
  2908. * previously done by the verifier are invalidated and must be
  2909. * performed again, if the helper is used in combination with
  2910. * direct packet access.
  2911. * Return
  2912. * 0 on success, or a negative error in case of failure.
  2913. *
  2914. * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
  2915. * Description
  2916. * Read the value of a perf event counter, and store it into *buf*
  2917. * of size *buf_size*. This helper relies on a *map* of type
  2918. * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
  2919. * counter is selected when *map* is updated with perf event file
  2920. * descriptors. The *map* is an array whose size is the number of
  2921. * available CPUs, and each cell contains a value relative to one
  2922. * CPU. The value to retrieve is indicated by *flags*, that
  2923. * contains the index of the CPU to look up, masked with
  2924. * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  2925. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  2926. * current CPU should be retrieved.
  2927. *
  2928. * This helper behaves in a way close to
  2929. * **bpf_perf_event_read**\ () helper, save that instead of
  2930. * just returning the value observed, it fills the *buf*
  2931. * structure. This allows for additional data to be retrieved: in
  2932. * particular, the enabled and running times (in *buf*\
  2933. * **->enabled** and *buf*\ **->running**, respectively) are
  2934. * copied. In general, **bpf_perf_event_read_value**\ () is
  2935. * recommended over **bpf_perf_event_read**\ (), which has some
  2936. * ABI issues and provides fewer functionalities.
  2937. *
  2938. * These values are interesting, because hardware PMU (Performance
  2939. * Monitoring Unit) counters are limited resources. When there are
  2940. * more PMU based perf events opened than available counters,
  2941. * kernel will multiplex these events so each event gets certain
  2942. * percentage (but not all) of the PMU time. In case that
  2943. * multiplexing happens, the number of samples or counter value
  2944. * will not reflect the case compared to when no multiplexing
  2945. * occurs. This makes comparison between different runs difficult.
  2946. * Typically, the counter value should be normalized before
  2947. * comparing to other experiments. The usual normalization is done
  2948. * as follows.
  2949. *
  2950. * ::
  2951. *
  2952. * normalized_counter = counter * t_enabled / t_running
  2953. *
  2954. * Where t_enabled is the time enabled for event and t_running is
  2955. * the time running for event since last normalization. The
  2956. * enabled and running times are accumulated since the perf event
  2957. * open. To achieve scaling factor between two invocations of an
  2958. * eBPF program, users can use CPU id as the key (which is
  2959. * typical for perf array usage model) to remember the previous
  2960. * value and do the calculation inside the eBPF program.
  2961. * Return
  2962. * 0 on success, or a negative error in case of failure.
  2963. *
  2964. * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
  2965. * Description
  2966. * For an eBPF program attached to a perf event, retrieve the
  2967. * value of the event counter associated to *ctx* and store it in
  2968. * the structure pointed by *buf* and of size *buf_size*. Enabled
  2969. * and running times are also stored in the structure (see
  2970. * description of helper **bpf_perf_event_read_value**\ () for
  2971. * more details).
  2972. * Return
  2973. * 0 on success, or a negative error in case of failure.
  2974. *
  2975. * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
  2976. * Description
  2977. * Emulate a call to **getsockopt()** on the socket associated to
  2978. * *bpf_socket*, which must be a full socket. The *level* at
  2979. * which the option resides and the name *optname* of the option
  2980. * must be specified, see **getsockopt(2)** for more information.
  2981. * The retrieved value is stored in the structure pointed by
  2982. * *opval* and of length *optlen*.
  2983. *
  2984. * *bpf_socket* should be one of the following:
  2985. *
  2986. * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
  2987. * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
  2988. * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
  2989. *
  2990. * This helper actually implements a subset of **getsockopt()**.
  2991. * It supports the same set of *optname*\ s that is supported by
  2992. * the **bpf_setsockopt**\ () helper. The exceptions are
  2993. * **TCP_BPF_*** is **bpf_setsockopt**\ () only and
  2994. * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
  2995. * Return
  2996. * 0 on success, or a negative error in case of failure.
  2997. *
  2998. * long bpf_override_return(struct pt_regs *regs, u64 rc)
  2999. * Description
  3000. * Used for error injection, this helper uses kprobes to override
  3001. * the return value of the probed function, and to set it to *rc*.
  3002. * The first argument is the context *regs* on which the kprobe
  3003. * works.
  3004. *
  3005. * This helper works by setting the PC (program counter)
  3006. * to an override function which is run in place of the original
  3007. * probed function. This means the probed function is not run at
  3008. * all. The replacement function just returns with the required
  3009. * value.
  3010. *
  3011. * This helper has security implications, and thus is subject to
  3012. * restrictions. It is only available if the kernel was compiled
  3013. * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
  3014. * option, and in this case it only works on functions tagged with
  3015. * **ALLOW_ERROR_INJECTION** in the kernel code.
  3016. *
  3017. * Also, the helper is only available for the architectures having
  3018. * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
  3019. * x86 architecture is the only one to support this feature.
  3020. * Return
  3021. * 0
  3022. *
  3023. * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
  3024. * Description
  3025. * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
  3026. * for the full TCP socket associated to *bpf_sock_ops* to
  3027. * *argval*.
  3028. *
  3029. * The primary use of this field is to determine if there should
  3030. * be calls to eBPF programs of type
  3031. * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
  3032. * code. A program of the same type can change its value, per
  3033. * connection and as necessary, when the connection is
  3034. * established. This field is directly accessible for reading, but
  3035. * this helper must be used for updates in order to return an
  3036. * error if an eBPF program tries to set a callback that is not
  3037. * supported in the current kernel.
  3038. *
  3039. * *argval* is a flag array which can combine these flags:
  3040. *
  3041. * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
  3042. * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
  3043. * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
  3044. * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
  3045. *
  3046. * Therefore, this function can be used to clear a callback flag by
  3047. * setting the appropriate bit to zero. e.g. to disable the RTO
  3048. * callback:
  3049. *
  3050. * **bpf_sock_ops_cb_flags_set(bpf_sock,**
  3051. * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
  3052. *
  3053. * Here are some examples of where one could call such eBPF
  3054. * program:
  3055. *
  3056. * * When RTO fires.
  3057. * * When a packet is retransmitted.
  3058. * * When the connection terminates.
  3059. * * When a packet is sent.
  3060. * * When a packet is received.
  3061. * Return
  3062. * Code **-EINVAL** if the socket is not a full TCP socket;
  3063. * otherwise, a positive number containing the bits that could not
  3064. * be set is returned (which comes down to 0 if all bits were set
  3065. * as required).
  3066. *
  3067. * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
  3068. * Description
  3069. * This helper is used in programs implementing policies at the
  3070. * socket level. If the message *msg* is allowed to pass (i.e. if
  3071. * the verdict eBPF program returns **SK_PASS**), redirect it to
  3072. * the socket referenced by *map* (of type
  3073. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  3074. * egress interfaces can be used for redirection. The
  3075. * **BPF_F_INGRESS** value in *flags* is used to make the
  3076. * distinction (ingress path is selected if the flag is present,
  3077. * egress path otherwise). This is the only flag supported for now.
  3078. * Return
  3079. * **SK_PASS** on success, or **SK_DROP** on error.
  3080. *
  3081. * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
  3082. * Description
  3083. * For socket policies, apply the verdict of the eBPF program to
  3084. * the next *bytes* (number of bytes) of message *msg*.
  3085. *
  3086. * For example, this helper can be used in the following cases:
  3087. *
  3088. * * A single **sendmsg**\ () or **sendfile**\ () system call
  3089. * contains multiple logical messages that the eBPF program is
  3090. * supposed to read and for which it should apply a verdict.
  3091. * * An eBPF program only cares to read the first *bytes* of a
  3092. * *msg*. If the message has a large payload, then setting up
  3093. * and calling the eBPF program repeatedly for all bytes, even
  3094. * though the verdict is already known, would create unnecessary
  3095. * overhead.
  3096. *
  3097. * When called from within an eBPF program, the helper sets a
  3098. * counter internal to the BPF infrastructure, that is used to
  3099. * apply the last verdict to the next *bytes*. If *bytes* is
  3100. * smaller than the current data being processed from a
  3101. * **sendmsg**\ () or **sendfile**\ () system call, the first
  3102. * *bytes* will be sent and the eBPF program will be re-run with
  3103. * the pointer for start of data pointing to byte number *bytes*
  3104. * **+ 1**. If *bytes* is larger than the current data being
  3105. * processed, then the eBPF verdict will be applied to multiple
  3106. * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
  3107. * consumed.
  3108. *
  3109. * Note that if a socket closes with the internal counter holding
  3110. * a non-zero value, this is not a problem because data is not
  3111. * being buffered for *bytes* and is sent as it is received.
  3112. * Return
  3113. * 0
  3114. *
  3115. * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
  3116. * Description
  3117. * For socket policies, prevent the execution of the verdict eBPF
  3118. * program for message *msg* until *bytes* (byte number) have been
  3119. * accumulated.
  3120. *
  3121. * This can be used when one needs a specific number of bytes
  3122. * before a verdict can be assigned, even if the data spans
  3123. * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
  3124. * case would be a user calling **sendmsg**\ () repeatedly with
  3125. * 1-byte long message segments. Obviously, this is bad for
  3126. * performance, but it is still valid. If the eBPF program needs
  3127. * *bytes* bytes to validate a header, this helper can be used to
  3128. * prevent the eBPF program to be called again until *bytes* have
  3129. * been accumulated.
  3130. * Return
  3131. * 0
  3132. *
  3133. * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
  3134. * Description
  3135. * For socket policies, pull in non-linear data from user space
  3136. * for *msg* and set pointers *msg*\ **->data** and *msg*\
  3137. * **->data_end** to *start* and *end* bytes offsets into *msg*,
  3138. * respectively.
  3139. *
  3140. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  3141. * *msg* it can only parse data that the (**data**, **data_end**)
  3142. * pointers have already consumed. For **sendmsg**\ () hooks this
  3143. * is likely the first scatterlist element. But for calls relying
  3144. * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
  3145. * be the range (**0**, **0**) because the data is shared with
  3146. * user space and by default the objective is to avoid allowing
  3147. * user space to modify data while (or after) eBPF verdict is
  3148. * being decided. This helper can be used to pull in data and to
  3149. * set the start and end pointer to given values. Data will be
  3150. * copied if necessary (i.e. if data was not linear and if start
  3151. * and end pointers do not point to the same chunk).
  3152. *
  3153. * A call to this helper is susceptible to change the underlying
  3154. * packet buffer. Therefore, at load time, all checks on pointers
  3155. * previously done by the verifier are invalidated and must be
  3156. * performed again, if the helper is used in combination with
  3157. * direct packet access.
  3158. *
  3159. * All values for *flags* are reserved for future usage, and must
  3160. * be left at zero.
  3161. * Return
  3162. * 0 on success, or a negative error in case of failure.
  3163. *
  3164. * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
  3165. * Description
  3166. * Bind the socket associated to *ctx* to the address pointed by
  3167. * *addr*, of length *addr_len*. This allows for making outgoing
  3168. * connection from the desired IP address, which can be useful for
  3169. * example when all processes inside a cgroup should use one
  3170. * single IP address on a host that has multiple IP configured.
  3171. *
  3172. * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
  3173. * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
  3174. * **AF_INET6**). It's advised to pass zero port (**sin_port**
  3175. * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
  3176. * behavior and lets the kernel efficiently pick up an unused
  3177. * port as long as 4-tuple is unique. Passing non-zero port might
  3178. * lead to degraded performance.
  3179. * Return
  3180. * 0 on success, or a negative error in case of failure.
  3181. *
  3182. * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
  3183. * Description
  3184. * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
  3185. * possible to both shrink and grow the packet tail.
  3186. * Shrink done via *delta* being a negative integer.
  3187. *
  3188. * A call to this helper is susceptible to change the underlying
  3189. * packet buffer. Therefore, at load time, all checks on pointers
  3190. * previously done by the verifier are invalidated and must be
  3191. * performed again, if the helper is used in combination with
  3192. * direct packet access.
  3193. * Return
  3194. * 0 on success, or a negative error in case of failure.
  3195. *
  3196. * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
  3197. * Description
  3198. * Retrieve the XFRM state (IP transform framework, see also
  3199. * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
  3200. *
  3201. * The retrieved value is stored in the **struct bpf_xfrm_state**
  3202. * pointed by *xfrm_state* and of length *size*.
  3203. *
  3204. * All values for *flags* are reserved for future usage, and must
  3205. * be left at zero.
  3206. *
  3207. * This helper is available only if the kernel was compiled with
  3208. * **CONFIG_XFRM** configuration option.
  3209. * Return
  3210. * 0 on success, or a negative error in case of failure.
  3211. *
  3212. * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
  3213. * Description
  3214. * Return a user or a kernel stack in bpf program provided buffer.
  3215. * To achieve this, the helper needs *ctx*, which is a pointer
  3216. * to the context on which the tracing program is executed.
  3217. * To store the stacktrace, the bpf program provides *buf* with
  3218. * a nonnegative *size*.
  3219. *
  3220. * The last argument, *flags*, holds the number of stack frames to
  3221. * skip (from 0 to 255), masked with
  3222. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  3223. * the following flags:
  3224. *
  3225. * **BPF_F_USER_STACK**
  3226. * Collect a user space stack instead of a kernel stack.
  3227. * **BPF_F_USER_BUILD_ID**
  3228. * Collect (build_id, file_offset) instead of ips for user
  3229. * stack, only valid if **BPF_F_USER_STACK** is also
  3230. * specified.
  3231. *
  3232. * *file_offset* is an offset relative to the beginning
  3233. * of the executable or shared object file backing the vma
  3234. * which the *ip* falls in. It is *not* an offset relative
  3235. * to that object's base address. Accordingly, it must be
  3236. * adjusted by adding (sh_addr - sh_offset), where
  3237. * sh_{addr,offset} correspond to the executable section
  3238. * containing *file_offset* in the object, for comparisons
  3239. * to symbols' st_value to be valid.
  3240. *
  3241. * **bpf_get_stack**\ () can collect up to
  3242. * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  3243. * to sufficient large buffer size. Note that
  3244. * this limit can be controlled with the **sysctl** program, and
  3245. * that it should be manually increased in order to profile long
  3246. * user stacks (such as stacks for Java programs). To do so, use:
  3247. *
  3248. * ::
  3249. *
  3250. * # sysctl kernel.perf_event_max_stack=<new value>
  3251. * Return
  3252. * The non-negative copied *buf* length equal to or less than
  3253. * *size* on success, or a negative error in case of failure.
  3254. *
  3255. * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
  3256. * Description
  3257. * This helper is similar to **bpf_skb_load_bytes**\ () in that
  3258. * it provides an easy way to load *len* bytes from *offset*
  3259. * from the packet associated to *skb*, into the buffer pointed
  3260. * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
  3261. * a fifth argument *start_header* exists in order to select a
  3262. * base offset to start from. *start_header* can be one of:
  3263. *
  3264. * **BPF_HDR_START_MAC**
  3265. * Base offset to load data from is *skb*'s mac header.
  3266. * **BPF_HDR_START_NET**
  3267. * Base offset to load data from is *skb*'s network header.
  3268. *
  3269. * In general, "direct packet access" is the preferred method to
  3270. * access packet data, however, this helper is in particular useful
  3271. * in socket filters where *skb*\ **->data** does not always point
  3272. * to the start of the mac header and where "direct packet access"
  3273. * is not available.
  3274. * Return
  3275. * 0 on success, or a negative error in case of failure.
  3276. *
  3277. * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
  3278. * Description
  3279. * Do FIB lookup in kernel tables using parameters in *params*.
  3280. * If lookup is successful and result shows packet is to be
  3281. * forwarded, the neighbor tables are searched for the nexthop.
  3282. * If successful (ie., FIB lookup shows forwarding and nexthop
  3283. * is resolved), the nexthop address is returned in ipv4_dst
  3284. * or ipv6_dst based on family, smac is set to mac address of
  3285. * egress device, dmac is set to nexthop mac address, rt_metric
  3286. * is set to metric from route (IPv4/IPv6 only), and ifindex
  3287. * is set to the device index of the nexthop from the FIB lookup.
  3288. *
  3289. * *plen* argument is the size of the passed in struct.
  3290. * *flags* argument can be a combination of one or more of the
  3291. * following values:
  3292. *
  3293. * **BPF_FIB_LOOKUP_DIRECT**
  3294. * Do a direct table lookup vs full lookup using FIB
  3295. * rules.
  3296. * **BPF_FIB_LOOKUP_TBID**
  3297. * Used with BPF_FIB_LOOKUP_DIRECT.
  3298. * Use the routing table ID present in *params*->tbid
  3299. * for the fib lookup.
  3300. * **BPF_FIB_LOOKUP_OUTPUT**
  3301. * Perform lookup from an egress perspective (default is
  3302. * ingress).
  3303. * **BPF_FIB_LOOKUP_SKIP_NEIGH**
  3304. * Skip the neighbour table lookup. *params*->dmac
  3305. * and *params*->smac will not be set as output. A common
  3306. * use case is to call **bpf_redirect_neigh**\ () after
  3307. * doing **bpf_fib_lookup**\ ().
  3308. * **BPF_FIB_LOOKUP_SRC**
  3309. * Derive and set source IP addr in *params*->ipv{4,6}_src
  3310. * for the nexthop. If the src addr cannot be derived,
  3311. * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
  3312. * case, *params*->dmac and *params*->smac are not set either.
  3313. * **BPF_FIB_LOOKUP_MARK**
  3314. * Use the mark present in *params*->mark for the fib lookup.
  3315. * This option should not be used with BPF_FIB_LOOKUP_DIRECT,
  3316. * as it only has meaning for full lookups.
  3317. *
  3318. * *ctx* is either **struct xdp_md** for XDP programs or
  3319. * **struct sk_buff** tc cls_act programs.
  3320. * Return
  3321. * * < 0 if any input argument is invalid
  3322. * * 0 on success (packet is forwarded, nexthop neighbor exists)
  3323. * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
  3324. * packet is not forwarded or needs assist from full stack
  3325. *
  3326. * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
  3327. * was exceeded and output params->mtu_result contains the MTU.
  3328. *
  3329. * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  3330. * Description
  3331. * Add an entry to, or update a sockhash *map* referencing sockets.
  3332. * The *skops* is used as a new value for the entry associated to
  3333. * *key*. *flags* is one of:
  3334. *
  3335. * **BPF_NOEXIST**
  3336. * The entry for *key* must not exist in the map.
  3337. * **BPF_EXIST**
  3338. * The entry for *key* must already exist in the map.
  3339. * **BPF_ANY**
  3340. * No condition on the existence of the entry for *key*.
  3341. *
  3342. * If the *map* has eBPF programs (parser and verdict), those will
  3343. * be inherited by the socket being added. If the socket is
  3344. * already attached to eBPF programs, this results in an error.
  3345. * Return
  3346. * 0 on success, or a negative error in case of failure.
  3347. *
  3348. * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
  3349. * Description
  3350. * This helper is used in programs implementing policies at the
  3351. * socket level. If the message *msg* is allowed to pass (i.e. if
  3352. * the verdict eBPF program returns **SK_PASS**), redirect it to
  3353. * the socket referenced by *map* (of type
  3354. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  3355. * egress interfaces can be used for redirection. The
  3356. * **BPF_F_INGRESS** value in *flags* is used to make the
  3357. * distinction (ingress path is selected if the flag is present,
  3358. * egress path otherwise). This is the only flag supported for now.
  3359. * Return
  3360. * **SK_PASS** on success, or **SK_DROP** on error.
  3361. *
  3362. * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
  3363. * Description
  3364. * This helper is used in programs implementing policies at the
  3365. * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
  3366. * if the verdict eBPF program returns **SK_PASS**), redirect it
  3367. * to the socket referenced by *map* (of type
  3368. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  3369. * egress interfaces can be used for redirection. The
  3370. * **BPF_F_INGRESS** value in *flags* is used to make the
  3371. * distinction (ingress path is selected if the flag is present,
  3372. * egress otherwise). This is the only flag supported for now.
  3373. * Return
  3374. * **SK_PASS** on success, or **SK_DROP** on error.
  3375. *
  3376. * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
  3377. * Description
  3378. * Encapsulate the packet associated to *skb* within a Layer 3
  3379. * protocol header. This header is provided in the buffer at
  3380. * address *hdr*, with *len* its size in bytes. *type* indicates
  3381. * the protocol of the header and can be one of:
  3382. *
  3383. * **BPF_LWT_ENCAP_SEG6**
  3384. * IPv6 encapsulation with Segment Routing Header
  3385. * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
  3386. * the IPv6 header is computed by the kernel.
  3387. * **BPF_LWT_ENCAP_SEG6_INLINE**
  3388. * Only works if *skb* contains an IPv6 packet. Insert a
  3389. * Segment Routing Header (**struct ipv6_sr_hdr**) inside
  3390. * the IPv6 header.
  3391. * **BPF_LWT_ENCAP_IP**
  3392. * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
  3393. * must be IPv4 or IPv6, followed by zero or more
  3394. * additional headers, up to **LWT_BPF_MAX_HEADROOM**
  3395. * total bytes in all prepended headers. Please note that
  3396. * if **skb_is_gso**\ (*skb*) is true, no more than two
  3397. * headers can be prepended, and the inner header, if
  3398. * present, should be either GRE or UDP/GUE.
  3399. *
  3400. * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
  3401. * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
  3402. * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
  3403. * **BPF_PROG_TYPE_LWT_XMIT**.
  3404. *
  3405. * A call to this helper is susceptible to change the underlying
  3406. * packet buffer. Therefore, at load time, all checks on pointers
  3407. * previously done by the verifier are invalidated and must be
  3408. * performed again, if the helper is used in combination with
  3409. * direct packet access.
  3410. * Return
  3411. * 0 on success, or a negative error in case of failure.
  3412. *
  3413. * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
  3414. * Description
  3415. * Store *len* bytes from address *from* into the packet
  3416. * associated to *skb*, at *offset*. Only the flags, tag and TLVs
  3417. * inside the outermost IPv6 Segment Routing Header can be
  3418. * modified through this helper.
  3419. *
  3420. * A call to this helper is susceptible to change the underlying
  3421. * packet buffer. Therefore, at load time, all checks on pointers
  3422. * previously done by the verifier are invalidated and must be
  3423. * performed again, if the helper is used in combination with
  3424. * direct packet access.
  3425. * Return
  3426. * 0 on success, or a negative error in case of failure.
  3427. *
  3428. * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
  3429. * Description
  3430. * Adjust the size allocated to TLVs in the outermost IPv6
  3431. * Segment Routing Header contained in the packet associated to
  3432. * *skb*, at position *offset* by *delta* bytes. Only offsets
  3433. * after the segments are accepted. *delta* can be as well
  3434. * positive (growing) as negative (shrinking).
  3435. *
  3436. * A call to this helper is susceptible to change the underlying
  3437. * packet buffer. Therefore, at load time, all checks on pointers
  3438. * previously done by the verifier are invalidated and must be
  3439. * performed again, if the helper is used in combination with
  3440. * direct packet access.
  3441. * Return
  3442. * 0 on success, or a negative error in case of failure.
  3443. *
  3444. * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
  3445. * Description
  3446. * Apply an IPv6 Segment Routing action of type *action* to the
  3447. * packet associated to *skb*. Each action takes a parameter
  3448. * contained at address *param*, and of length *param_len* bytes.
  3449. * *action* can be one of:
  3450. *
  3451. * **SEG6_LOCAL_ACTION_END_X**
  3452. * End.X action: Endpoint with Layer-3 cross-connect.
  3453. * Type of *param*: **struct in6_addr**.
  3454. * **SEG6_LOCAL_ACTION_END_T**
  3455. * End.T action: Endpoint with specific IPv6 table lookup.
  3456. * Type of *param*: **int**.
  3457. * **SEG6_LOCAL_ACTION_END_B6**
  3458. * End.B6 action: Endpoint bound to an SRv6 policy.
  3459. * Type of *param*: **struct ipv6_sr_hdr**.
  3460. * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
  3461. * End.B6.Encap action: Endpoint bound to an SRv6
  3462. * encapsulation policy.
  3463. * Type of *param*: **struct ipv6_sr_hdr**.
  3464. *
  3465. * A call to this helper is susceptible to change the underlying
  3466. * packet buffer. Therefore, at load time, all checks on pointers
  3467. * previously done by the verifier are invalidated and must be
  3468. * performed again, if the helper is used in combination with
  3469. * direct packet access.
  3470. * Return
  3471. * 0 on success, or a negative error in case of failure.
  3472. *
  3473. * long bpf_rc_repeat(void *ctx)
  3474. * Description
  3475. * This helper is used in programs implementing IR decoding, to
  3476. * report a successfully decoded repeat key message. This delays
  3477. * the generation of a key up event for previously generated
  3478. * key down event.
  3479. *
  3480. * Some IR protocols like NEC have a special IR message for
  3481. * repeating last button, for when a button is held down.
  3482. *
  3483. * The *ctx* should point to the lirc sample as passed into
  3484. * the program.
  3485. *
  3486. * This helper is only available is the kernel was compiled with
  3487. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3488. * "**y**".
  3489. * Return
  3490. * 0
  3491. *
  3492. * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
  3493. * Description
  3494. * This helper is used in programs implementing IR decoding, to
  3495. * report a successfully decoded key press with *scancode*,
  3496. * *toggle* value in the given *protocol*. The scancode will be
  3497. * translated to a keycode using the rc keymap, and reported as
  3498. * an input key down event. After a period a key up event is
  3499. * generated. This period can be extended by calling either
  3500. * **bpf_rc_keydown**\ () again with the same values, or calling
  3501. * **bpf_rc_repeat**\ ().
  3502. *
  3503. * Some protocols include a toggle bit, in case the button was
  3504. * released and pressed again between consecutive scancodes.
  3505. *
  3506. * The *ctx* should point to the lirc sample as passed into
  3507. * the program.
  3508. *
  3509. * The *protocol* is the decoded protocol number (see
  3510. * **enum rc_proto** for some predefined values).
  3511. *
  3512. * This helper is only available is the kernel was compiled with
  3513. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3514. * "**y**".
  3515. * Return
  3516. * 0
  3517. *
  3518. * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
  3519. * Description
  3520. * Return the cgroup v2 id of the socket associated with the *skb*.
  3521. * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
  3522. * helper for cgroup v1 by providing a tag resp. identifier that
  3523. * can be matched on or used for map lookups e.g. to implement
  3524. * policy. The cgroup v2 id of a given path in the hierarchy is
  3525. * exposed in user space through the f_handle API in order to get
  3526. * to the same 64-bit id.
  3527. *
  3528. * This helper can be used on TC egress path, but not on ingress,
  3529. * and is available only if the kernel was compiled with the
  3530. * **CONFIG_SOCK_CGROUP_DATA** configuration option.
  3531. * Return
  3532. * The id is returned or 0 in case the id could not be retrieved.
  3533. *
  3534. * u64 bpf_get_current_cgroup_id(void)
  3535. * Description
  3536. * Get the current cgroup id based on the cgroup within which
  3537. * the current task is running.
  3538. * Return
  3539. * A 64-bit integer containing the current cgroup id based
  3540. * on the cgroup within which the current task is running.
  3541. *
  3542. * void *bpf_get_local_storage(void *map, u64 flags)
  3543. * Description
  3544. * Get the pointer to the local storage area.
  3545. * The type and the size of the local storage is defined
  3546. * by the *map* argument.
  3547. * The *flags* meaning is specific for each map type,
  3548. * and has to be 0 for cgroup local storage.
  3549. *
  3550. * Depending on the BPF program type, a local storage area
  3551. * can be shared between multiple instances of the BPF program,
  3552. * running simultaneously.
  3553. *
  3554. * A user should care about the synchronization by himself.
  3555. * For example, by using the **BPF_ATOMIC** instructions to alter
  3556. * the shared data.
  3557. * Return
  3558. * A pointer to the local storage area.
  3559. *
  3560. * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
  3561. * Description
  3562. * Select a **SO_REUSEPORT** socket from a
  3563. * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
  3564. * It checks the selected socket is matching the incoming
  3565. * request in the socket buffer.
  3566. * Return
  3567. * 0 on success, or a negative error in case of failure.
  3568. *
  3569. * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
  3570. * Description
  3571. * Return id of cgroup v2 that is ancestor of cgroup associated
  3572. * with the *skb* at the *ancestor_level*. The root cgroup is at
  3573. * *ancestor_level* zero and each step down the hierarchy
  3574. * increments the level. If *ancestor_level* == level of cgroup
  3575. * associated with *skb*, then return value will be same as that
  3576. * of **bpf_skb_cgroup_id**\ ().
  3577. *
  3578. * The helper is useful to implement policies based on cgroups
  3579. * that are upper in hierarchy than immediate cgroup associated
  3580. * with *skb*.
  3581. *
  3582. * The format of returned id and helper limitations are same as in
  3583. * **bpf_skb_cgroup_id**\ ().
  3584. * Return
  3585. * The id is returned or 0 in case the id could not be retrieved.
  3586. *
  3587. * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3588. * Description
  3589. * Look for TCP socket matching *tuple*, optionally in a child
  3590. * network namespace *netns*. The return value must be checked,
  3591. * and if non-**NULL**, released via **bpf_sk_release**\ ().
  3592. *
  3593. * The *ctx* should point to the context of the program, such as
  3594. * the skb or socket (depending on the hook in use). This is used
  3595. * to determine the base network namespace for the lookup.
  3596. *
  3597. * *tuple_size* must be one of:
  3598. *
  3599. * **sizeof**\ (*tuple*\ **->ipv4**)
  3600. * Look for an IPv4 socket.
  3601. * **sizeof**\ (*tuple*\ **->ipv6**)
  3602. * Look for an IPv6 socket.
  3603. *
  3604. * If the *netns* is a negative signed 32-bit integer, then the
  3605. * socket lookup table in the netns associated with the *ctx*
  3606. * will be used. For the TC hooks, this is the netns of the device
  3607. * in the skb. For socket hooks, this is the netns of the socket.
  3608. * If *netns* is any other signed 32-bit value greater than or
  3609. * equal to zero then it specifies the ID of the netns relative to
  3610. * the netns associated with the *ctx*. *netns* values beyond the
  3611. * range of 32-bit integers are reserved for future use.
  3612. *
  3613. * All values for *flags* are reserved for future usage, and must
  3614. * be left at zero.
  3615. *
  3616. * This helper is available only if the kernel was compiled with
  3617. * **CONFIG_NET** configuration option.
  3618. * Return
  3619. * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3620. * For sockets with reuseport option, the **struct bpf_sock**
  3621. * result is from *reuse*\ **->socks**\ [] using the hash of the
  3622. * tuple.
  3623. *
  3624. * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3625. * Description
  3626. * Look for UDP socket matching *tuple*, optionally in a child
  3627. * network namespace *netns*. The return value must be checked,
  3628. * and if non-**NULL**, released via **bpf_sk_release**\ ().
  3629. *
  3630. * The *ctx* should point to the context of the program, such as
  3631. * the skb or socket (depending on the hook in use). This is used
  3632. * to determine the base network namespace for the lookup.
  3633. *
  3634. * *tuple_size* must be one of:
  3635. *
  3636. * **sizeof**\ (*tuple*\ **->ipv4**)
  3637. * Look for an IPv4 socket.
  3638. * **sizeof**\ (*tuple*\ **->ipv6**)
  3639. * Look for an IPv6 socket.
  3640. *
  3641. * If the *netns* is a negative signed 32-bit integer, then the
  3642. * socket lookup table in the netns associated with the *ctx*
  3643. * will be used. For the TC hooks, this is the netns of the device
  3644. * in the skb. For socket hooks, this is the netns of the socket.
  3645. * If *netns* is any other signed 32-bit value greater than or
  3646. * equal to zero then it specifies the ID of the netns relative to
  3647. * the netns associated with the *ctx*. *netns* values beyond the
  3648. * range of 32-bit integers are reserved for future use.
  3649. *
  3650. * All values for *flags* are reserved for future usage, and must
  3651. * be left at zero.
  3652. *
  3653. * This helper is available only if the kernel was compiled with
  3654. * **CONFIG_NET** configuration option.
  3655. * Return
  3656. * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3657. * For sockets with reuseport option, the **struct bpf_sock**
  3658. * result is from *reuse*\ **->socks**\ [] using the hash of the
  3659. * tuple.
  3660. *
  3661. * long bpf_sk_release(void *sock)
  3662. * Description
  3663. * Release the reference held by *sock*. *sock* must be a
  3664. * non-**NULL** pointer that was returned from
  3665. * **bpf_sk_lookup_xxx**\ ().
  3666. * Return
  3667. * 0 on success, or a negative error in case of failure.
  3668. *
  3669. * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
  3670. * Description
  3671. * Push an element *value* in *map*. *flags* is one of:
  3672. *
  3673. * **BPF_EXIST**
  3674. * If the queue/stack is full, the oldest element is
  3675. * removed to make room for this.
  3676. * Return
  3677. * 0 on success, or a negative error in case of failure.
  3678. *
  3679. * long bpf_map_pop_elem(struct bpf_map *map, void *value)
  3680. * Description
  3681. * Pop an element from *map*.
  3682. * Return
  3683. * 0 on success, or a negative error in case of failure.
  3684. *
  3685. * long bpf_map_peek_elem(struct bpf_map *map, void *value)
  3686. * Description
  3687. * Get an element from *map* without removing it.
  3688. * Return
  3689. * 0 on success, or a negative error in case of failure.
  3690. *
  3691. * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
  3692. * Description
  3693. * For socket policies, insert *len* bytes into *msg* at offset
  3694. * *start*.
  3695. *
  3696. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  3697. * *msg* it may want to insert metadata or options into the *msg*.
  3698. * This can later be read and used by any of the lower layer BPF
  3699. * hooks.
  3700. *
  3701. * This helper may fail if under memory pressure (a malloc
  3702. * fails) in these cases BPF programs will get an appropriate
  3703. * error and BPF programs will need to handle them.
  3704. * Return
  3705. * 0 on success, or a negative error in case of failure.
  3706. *
  3707. * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
  3708. * Description
  3709. * Will remove *len* bytes from a *msg* starting at byte *start*.
  3710. * This may result in **ENOMEM** errors under certain situations if
  3711. * an allocation and copy are required due to a full ring buffer.
  3712. * However, the helper will try to avoid doing the allocation
  3713. * if possible. Other errors can occur if input parameters are
  3714. * invalid either due to *start* byte not being valid part of *msg*
  3715. * payload and/or *pop* value being to large.
  3716. * Return
  3717. * 0 on success, or a negative error in case of failure.
  3718. *
  3719. * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
  3720. * Description
  3721. * This helper is used in programs implementing IR decoding, to
  3722. * report a successfully decoded pointer movement.
  3723. *
  3724. * The *ctx* should point to the lirc sample as passed into
  3725. * the program.
  3726. *
  3727. * This helper is only available is the kernel was compiled with
  3728. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3729. * "**y**".
  3730. * Return
  3731. * 0
  3732. *
  3733. * long bpf_spin_lock(struct bpf_spin_lock *lock)
  3734. * Description
  3735. * Acquire a spinlock represented by the pointer *lock*, which is
  3736. * stored as part of a value of a map. Taking the lock allows to
  3737. * safely update the rest of the fields in that value. The
  3738. * spinlock can (and must) later be released with a call to
  3739. * **bpf_spin_unlock**\ (\ *lock*\ ).
  3740. *
  3741. * Spinlocks in BPF programs come with a number of restrictions
  3742. * and constraints:
  3743. *
  3744. * * **bpf_spin_lock** objects are only allowed inside maps of
  3745. * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
  3746. * list could be extended in the future).
  3747. * * BTF description of the map is mandatory.
  3748. * * The BPF program can take ONE lock at a time, since taking two
  3749. * or more could cause dead locks.
  3750. * * Only one **struct bpf_spin_lock** is allowed per map element.
  3751. * * When the lock is taken, calls (either BPF to BPF or helpers)
  3752. * are not allowed.
  3753. * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
  3754. * allowed inside a spinlock-ed region.
  3755. * * The BPF program MUST call **bpf_spin_unlock**\ () to release
  3756. * the lock, on all execution paths, before it returns.
  3757. * * The BPF program can access **struct bpf_spin_lock** only via
  3758. * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
  3759. * helpers. Loading or storing data into the **struct
  3760. * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
  3761. * * To use the **bpf_spin_lock**\ () helper, the BTF description
  3762. * of the map value must be a struct and have **struct
  3763. * bpf_spin_lock** *anyname*\ **;** field at the top level.
  3764. * Nested lock inside another struct is not allowed.
  3765. * * The **struct bpf_spin_lock** *lock* field in a map value must
  3766. * be aligned on a multiple of 4 bytes in that value.
  3767. * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
  3768. * the **bpf_spin_lock** field to user space.
  3769. * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
  3770. * a BPF program, do not update the **bpf_spin_lock** field.
  3771. * * **bpf_spin_lock** cannot be on the stack or inside a
  3772. * networking packet (it can only be inside of a map values).
  3773. * * **bpf_spin_lock** is available to root only.
  3774. * * Tracing programs and socket filter programs cannot use
  3775. * **bpf_spin_lock**\ () due to insufficient preemption checks
  3776. * (but this may change in the future).
  3777. * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
  3778. * Return
  3779. * 0
  3780. *
  3781. * long bpf_spin_unlock(struct bpf_spin_lock *lock)
  3782. * Description
  3783. * Release the *lock* previously locked by a call to
  3784. * **bpf_spin_lock**\ (\ *lock*\ ).
  3785. * Return
  3786. * 0
  3787. *
  3788. * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
  3789. * Description
  3790. * This helper gets a **struct bpf_sock** pointer such
  3791. * that all the fields in this **bpf_sock** can be accessed.
  3792. * Return
  3793. * A **struct bpf_sock** pointer on success, or **NULL** in
  3794. * case of failure.
  3795. *
  3796. * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
  3797. * Description
  3798. * This helper gets a **struct bpf_tcp_sock** pointer from a
  3799. * **struct bpf_sock** pointer.
  3800. * Return
  3801. * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
  3802. * case of failure.
  3803. *
  3804. * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
  3805. * Description
  3806. * Set ECN (Explicit Congestion Notification) field of IP header
  3807. * to **CE** (Congestion Encountered) if current value is **ECT**
  3808. * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
  3809. * and IPv4.
  3810. * Return
  3811. * 1 if the **CE** flag is set (either by the current helper call
  3812. * or because it was already present), 0 if it is not set.
  3813. *
  3814. * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
  3815. * Description
  3816. * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
  3817. * **bpf_sk_release**\ () is unnecessary and not allowed.
  3818. * Return
  3819. * A **struct bpf_sock** pointer on success, or **NULL** in
  3820. * case of failure.
  3821. *
  3822. * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3823. * Description
  3824. * Look for TCP socket matching *tuple*, optionally in a child
  3825. * network namespace *netns*. The return value must be checked,
  3826. * and if non-**NULL**, released via **bpf_sk_release**\ ().
  3827. *
  3828. * This function is identical to **bpf_sk_lookup_tcp**\ (), except
  3829. * that it also returns timewait or request sockets. Use
  3830. * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
  3831. * full structure.
  3832. *
  3833. * This helper is available only if the kernel was compiled with
  3834. * **CONFIG_NET** configuration option.
  3835. * Return
  3836. * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3837. * For sockets with reuseport option, the **struct bpf_sock**
  3838. * result is from *reuse*\ **->socks**\ [] using the hash of the
  3839. * tuple.
  3840. *
  3841. * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
  3842. * Description
  3843. * Check whether *iph* and *th* contain a valid SYN cookie ACK for
  3844. * the listening socket in *sk*.
  3845. *
  3846. * *iph* points to the start of the IPv4 or IPv6 header, while
  3847. * *iph_len* contains **sizeof**\ (**struct iphdr**) or
  3848. * **sizeof**\ (**struct ipv6hdr**).
  3849. *
  3850. * *th* points to the start of the TCP header, while *th_len*
  3851. * contains the length of the TCP header (at least
  3852. * **sizeof**\ (**struct tcphdr**)).
  3853. * Return
  3854. * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
  3855. * error otherwise.
  3856. *
  3857. * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
  3858. * Description
  3859. * Get name of sysctl in /proc/sys/ and copy it into provided by
  3860. * program buffer *buf* of size *buf_len*.
  3861. *
  3862. * The buffer is always NUL terminated, unless it's zero-sized.
  3863. *
  3864. * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
  3865. * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
  3866. * only (e.g. "tcp_mem").
  3867. * Return
  3868. * Number of character copied (not including the trailing NUL).
  3869. *
  3870. * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3871. * truncated name in this case).
  3872. *
  3873. * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
  3874. * Description
  3875. * Get current value of sysctl as it is presented in /proc/sys
  3876. * (incl. newline, etc), and copy it as a string into provided
  3877. * by program buffer *buf* of size *buf_len*.
  3878. *
  3879. * The whole value is copied, no matter what file position user
  3880. * space issued e.g. sys_read at.
  3881. *
  3882. * The buffer is always NUL terminated, unless it's zero-sized.
  3883. * Return
  3884. * Number of character copied (not including the trailing NUL).
  3885. *
  3886. * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3887. * truncated name in this case).
  3888. *
  3889. * **-EINVAL** if current value was unavailable, e.g. because
  3890. * sysctl is uninitialized and read returns -EIO for it.
  3891. *
  3892. * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
  3893. * Description
  3894. * Get new value being written by user space to sysctl (before
  3895. * the actual write happens) and copy it as a string into
  3896. * provided by program buffer *buf* of size *buf_len*.
  3897. *
  3898. * User space may write new value at file position > 0.
  3899. *
  3900. * The buffer is always NUL terminated, unless it's zero-sized.
  3901. * Return
  3902. * Number of character copied (not including the trailing NUL).
  3903. *
  3904. * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3905. * truncated name in this case).
  3906. *
  3907. * **-EINVAL** if sysctl is being read.
  3908. *
  3909. * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
  3910. * Description
  3911. * Override new value being written by user space to sysctl with
  3912. * value provided by program in buffer *buf* of size *buf_len*.
  3913. *
  3914. * *buf* should contain a string in same form as provided by user
  3915. * space on sysctl write.
  3916. *
  3917. * User space may write new value at file position > 0. To override
  3918. * the whole sysctl value file position should be set to zero.
  3919. * Return
  3920. * 0 on success.
  3921. *
  3922. * **-E2BIG** if the *buf_len* is too big.
  3923. *
  3924. * **-EINVAL** if sysctl is being read.
  3925. *
  3926. * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
  3927. * Description
  3928. * Convert the initial part of the string from buffer *buf* of
  3929. * size *buf_len* to a long integer according to the given base
  3930. * and save the result in *res*.
  3931. *
  3932. * The string may begin with an arbitrary amount of white space
  3933. * (as determined by **isspace**\ (3)) followed by a single
  3934. * optional '**-**' sign.
  3935. *
  3936. * Five least significant bits of *flags* encode base, other bits
  3937. * are currently unused.
  3938. *
  3939. * Base must be either 8, 10, 16 or 0 to detect it automatically
  3940. * similar to user space **strtol**\ (3).
  3941. * Return
  3942. * Number of characters consumed on success. Must be positive but
  3943. * no more than *buf_len*.
  3944. *
  3945. * **-EINVAL** if no valid digits were found or unsupported base
  3946. * was provided.
  3947. *
  3948. * **-ERANGE** if resulting value was out of range.
  3949. *
  3950. * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
  3951. * Description
  3952. * Convert the initial part of the string from buffer *buf* of
  3953. * size *buf_len* to an unsigned long integer according to the
  3954. * given base and save the result in *res*.
  3955. *
  3956. * The string may begin with an arbitrary amount of white space
  3957. * (as determined by **isspace**\ (3)).
  3958. *
  3959. * Five least significant bits of *flags* encode base, other bits
  3960. * are currently unused.
  3961. *
  3962. * Base must be either 8, 10, 16 or 0 to detect it automatically
  3963. * similar to user space **strtoul**\ (3).
  3964. * Return
  3965. * Number of characters consumed on success. Must be positive but
  3966. * no more than *buf_len*.
  3967. *
  3968. * **-EINVAL** if no valid digits were found or unsupported base
  3969. * was provided.
  3970. *
  3971. * **-ERANGE** if resulting value was out of range.
  3972. *
  3973. * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
  3974. * Description
  3975. * Get a bpf-local-storage from a *sk*.
  3976. *
  3977. * Logically, it could be thought of getting the value from
  3978. * a *map* with *sk* as the **key**. From this
  3979. * perspective, the usage is not much different from
  3980. * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
  3981. * helper enforces the key must be a full socket and the map must
  3982. * be a **BPF_MAP_TYPE_SK_STORAGE** also.
  3983. *
  3984. * Underneath, the value is stored locally at *sk* instead of
  3985. * the *map*. The *map* is used as the bpf-local-storage
  3986. * "type". The bpf-local-storage "type" (i.e. the *map*) is
  3987. * searched against all bpf-local-storages residing at *sk*.
  3988. *
  3989. * *sk* is a kernel **struct sock** pointer for LSM program.
  3990. * *sk* is a **struct bpf_sock** pointer for other program types.
  3991. *
  3992. * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
  3993. * used such that a new bpf-local-storage will be
  3994. * created if one does not exist. *value* can be used
  3995. * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
  3996. * the initial value of a bpf-local-storage. If *value* is
  3997. * **NULL**, the new bpf-local-storage will be zero initialized.
  3998. * Return
  3999. * A bpf-local-storage pointer is returned on success.
  4000. *
  4001. * **NULL** if not found or there was an error in adding
  4002. * a new bpf-local-storage.
  4003. *
  4004. * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
  4005. * Description
  4006. * Delete a bpf-local-storage from a *sk*.
  4007. * Return
  4008. * 0 on success.
  4009. *
  4010. * **-ENOENT** if the bpf-local-storage cannot be found.
  4011. * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
  4012. *
  4013. * long bpf_send_signal(u32 sig)
  4014. * Description
  4015. * Send signal *sig* to the process of the current task.
  4016. * The signal may be delivered to any of this process's threads.
  4017. * Return
  4018. * 0 on success or successfully queued.
  4019. *
  4020. * **-EBUSY** if work queue under nmi is full.
  4021. *
  4022. * **-EINVAL** if *sig* is invalid.
  4023. *
  4024. * **-EPERM** if no permission to send the *sig*.
  4025. *
  4026. * **-EAGAIN** if bpf program can try again.
  4027. *
  4028. * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
  4029. * Description
  4030. * Try to issue a SYN cookie for the packet with corresponding
  4031. * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
  4032. *
  4033. * *iph* points to the start of the IPv4 or IPv6 header, while
  4034. * *iph_len* contains **sizeof**\ (**struct iphdr**) or
  4035. * **sizeof**\ (**struct ipv6hdr**).
  4036. *
  4037. * *th* points to the start of the TCP header, while *th_len*
  4038. * contains the length of the TCP header with options (at least
  4039. * **sizeof**\ (**struct tcphdr**)).
  4040. * Return
  4041. * On success, lower 32 bits hold the generated SYN cookie in
  4042. * followed by 16 bits which hold the MSS value for that cookie,
  4043. * and the top 16 bits are unused.
  4044. *
  4045. * On failure, the returned value is one of the following:
  4046. *
  4047. * **-EINVAL** SYN cookie cannot be issued due to error
  4048. *
  4049. * **-ENOENT** SYN cookie should not be issued (no SYN flood)
  4050. *
  4051. * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
  4052. *
  4053. * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
  4054. *
  4055. * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  4056. * Description
  4057. * Write raw *data* blob into a special BPF perf event held by
  4058. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  4059. * event must have the following attributes: **PERF_SAMPLE_RAW**
  4060. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  4061. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  4062. *
  4063. * The *flags* are used to indicate the index in *map* for which
  4064. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  4065. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  4066. * to indicate that the index of the current CPU core should be
  4067. * used.
  4068. *
  4069. * The value to write, of *size*, is passed through eBPF stack and
  4070. * pointed by *data*.
  4071. *
  4072. * *ctx* is a pointer to in-kernel struct sk_buff.
  4073. *
  4074. * This helper is similar to **bpf_perf_event_output**\ () but
  4075. * restricted to raw_tracepoint bpf programs.
  4076. * Return
  4077. * 0 on success, or a negative error in case of failure.
  4078. *
  4079. * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
  4080. * Description
  4081. * Safely attempt to read *size* bytes from user space address
  4082. * *unsafe_ptr* and store the data in *dst*.
  4083. * Return
  4084. * 0 on success, or a negative error in case of failure.
  4085. *
  4086. * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
  4087. * Description
  4088. * Safely attempt to read *size* bytes from kernel space address
  4089. * *unsafe_ptr* and store the data in *dst*.
  4090. * Return
  4091. * 0 on success, or a negative error in case of failure.
  4092. *
  4093. * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
  4094. * Description
  4095. * Copy a NUL terminated string from an unsafe user address
  4096. * *unsafe_ptr* to *dst*. The *size* should include the
  4097. * terminating NUL byte. In case the string length is smaller than
  4098. * *size*, the target is not padded with further NUL bytes. If the
  4099. * string length is larger than *size*, just *size*-1 bytes are
  4100. * copied and the last byte is set to NUL.
  4101. *
  4102. * On success, returns the number of bytes that were written,
  4103. * including the terminal NUL. This makes this helper useful in
  4104. * tracing programs for reading strings, and more importantly to
  4105. * get its length at runtime. See the following snippet:
  4106. *
  4107. * ::
  4108. *
  4109. * SEC("kprobe/sys_open")
  4110. * void bpf_sys_open(struct pt_regs *ctx)
  4111. * {
  4112. * char buf[PATHLEN]; // PATHLEN is defined to 256
  4113. * int res = bpf_probe_read_user_str(buf, sizeof(buf),
  4114. * ctx->di);
  4115. *
  4116. * // Consume buf, for example push it to
  4117. * // userspace via bpf_perf_event_output(); we
  4118. * // can use res (the string length) as event
  4119. * // size, after checking its boundaries.
  4120. * }
  4121. *
  4122. * In comparison, using **bpf_probe_read_user**\ () helper here
  4123. * instead to read the string would require to estimate the length
  4124. * at compile time, and would often result in copying more memory
  4125. * than necessary.
  4126. *
  4127. * Another useful use case is when parsing individual process
  4128. * arguments or individual environment variables navigating
  4129. * *current*\ **->mm->arg_start** and *current*\
  4130. * **->mm->env_start**: using this helper and the return value,
  4131. * one can quickly iterate at the right offset of the memory area.
  4132. * Return
  4133. * On success, the strictly positive length of the output string,
  4134. * including the trailing NUL character. On error, a negative
  4135. * value.
  4136. *
  4137. * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
  4138. * Description
  4139. * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
  4140. * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
  4141. * Return
  4142. * On success, the strictly positive length of the string, including
  4143. * the trailing NUL character. On error, a negative value.
  4144. *
  4145. * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
  4146. * Description
  4147. * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
  4148. * *rcv_nxt* is the ack_seq to be sent out.
  4149. * Return
  4150. * 0 on success, or a negative error in case of failure.
  4151. *
  4152. * long bpf_send_signal_thread(u32 sig)
  4153. * Description
  4154. * Send signal *sig* to the thread corresponding to the current task.
  4155. * Return
  4156. * 0 on success or successfully queued.
  4157. *
  4158. * **-EBUSY** if work queue under nmi is full.
  4159. *
  4160. * **-EINVAL** if *sig* is invalid.
  4161. *
  4162. * **-EPERM** if no permission to send the *sig*.
  4163. *
  4164. * **-EAGAIN** if bpf program can try again.
  4165. *
  4166. * u64 bpf_jiffies64(void)
  4167. * Description
  4168. * Obtain the 64bit jiffies
  4169. * Return
  4170. * The 64 bit jiffies
  4171. *
  4172. * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
  4173. * Description
  4174. * For an eBPF program attached to a perf event, retrieve the
  4175. * branch records (**struct perf_branch_entry**) associated to *ctx*
  4176. * and store it in the buffer pointed by *buf* up to size
  4177. * *size* bytes.
  4178. * Return
  4179. * On success, number of bytes written to *buf*. On error, a
  4180. * negative value.
  4181. *
  4182. * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
  4183. * instead return the number of bytes required to store all the
  4184. * branch entries. If this flag is set, *buf* may be NULL.
  4185. *
  4186. * **-EINVAL** if arguments invalid or **size** not a multiple
  4187. * of **sizeof**\ (**struct perf_branch_entry**\ ).
  4188. *
  4189. * **-ENOENT** if architecture does not support branch records.
  4190. *
  4191. * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
  4192. * Description
  4193. * Returns 0 on success, values for *pid* and *tgid* as seen from the current
  4194. * *namespace* will be returned in *nsdata*.
  4195. * Return
  4196. * 0 on success, or one of the following in case of failure:
  4197. *
  4198. * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
  4199. * with nsfs of current task, or if dev conversion to dev_t lost high bits.
  4200. *
  4201. * **-ENOENT** if pidns does not exists for the current task.
  4202. *
  4203. * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  4204. * Description
  4205. * Write raw *data* blob into a special BPF perf event held by
  4206. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  4207. * event must have the following attributes: **PERF_SAMPLE_RAW**
  4208. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  4209. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  4210. *
  4211. * The *flags* are used to indicate the index in *map* for which
  4212. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  4213. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  4214. * to indicate that the index of the current CPU core should be
  4215. * used.
  4216. *
  4217. * The value to write, of *size*, is passed through eBPF stack and
  4218. * pointed by *data*.
  4219. *
  4220. * *ctx* is a pointer to in-kernel struct xdp_buff.
  4221. *
  4222. * This helper is similar to **bpf_perf_eventoutput**\ () but
  4223. * restricted to raw_tracepoint bpf programs.
  4224. * Return
  4225. * 0 on success, or a negative error in case of failure.
  4226. *
  4227. * u64 bpf_get_netns_cookie(void *ctx)
  4228. * Description
  4229. * Retrieve the cookie (generated by the kernel) of the network
  4230. * namespace the input *ctx* is associated with. The network
  4231. * namespace cookie remains stable for its lifetime and provides
  4232. * a global identifier that can be assumed unique. If *ctx* is
  4233. * NULL, then the helper returns the cookie for the initial
  4234. * network namespace. The cookie itself is very similar to that
  4235. * of **bpf_get_socket_cookie**\ () helper, but for network
  4236. * namespaces instead of sockets.
  4237. * Return
  4238. * A 8-byte long opaque number.
  4239. *
  4240. * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
  4241. * Description
  4242. * Return id of cgroup v2 that is ancestor of the cgroup associated
  4243. * with the current task at the *ancestor_level*. The root cgroup
  4244. * is at *ancestor_level* zero and each step down the hierarchy
  4245. * increments the level. If *ancestor_level* == level of cgroup
  4246. * associated with the current task, then return value will be the
  4247. * same as that of **bpf_get_current_cgroup_id**\ ().
  4248. *
  4249. * The helper is useful to implement policies based on cgroups
  4250. * that are upper in hierarchy than immediate cgroup associated
  4251. * with the current task.
  4252. *
  4253. * The format of returned id and helper limitations are same as in
  4254. * **bpf_get_current_cgroup_id**\ ().
  4255. * Return
  4256. * The id is returned or 0 in case the id could not be retrieved.
  4257. *
  4258. * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
  4259. * Description
  4260. * Helper is overloaded depending on BPF program type. This
  4261. * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
  4262. * **BPF_PROG_TYPE_SCHED_ACT** programs.
  4263. *
  4264. * Assign the *sk* to the *skb*. When combined with appropriate
  4265. * routing configuration to receive the packet towards the socket,
  4266. * will cause *skb* to be delivered to the specified socket.
  4267. * Subsequent redirection of *skb* via **bpf_redirect**\ (),
  4268. * **bpf_clone_redirect**\ () or other methods outside of BPF may
  4269. * interfere with successful delivery to the socket.
  4270. *
  4271. * This operation is only valid from TC ingress path.
  4272. *
  4273. * The *flags* argument must be zero.
  4274. * Return
  4275. * 0 on success, or a negative error in case of failure:
  4276. *
  4277. * **-EINVAL** if specified *flags* are not supported.
  4278. *
  4279. * **-ENOENT** if the socket is unavailable for assignment.
  4280. *
  4281. * **-ENETUNREACH** if the socket is unreachable (wrong netns).
  4282. *
  4283. * **-EOPNOTSUPP** if the operation is not supported, for example
  4284. * a call from outside of TC ingress.
  4285. *
  4286. * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
  4287. * Description
  4288. * Helper is overloaded depending on BPF program type. This
  4289. * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
  4290. *
  4291. * Select the *sk* as a result of a socket lookup.
  4292. *
  4293. * For the operation to succeed passed socket must be compatible
  4294. * with the packet description provided by the *ctx* object.
  4295. *
  4296. * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
  4297. * be an exact match. While IP family (**AF_INET** or
  4298. * **AF_INET6**) must be compatible, that is IPv6 sockets
  4299. * that are not v6-only can be selected for IPv4 packets.
  4300. *
  4301. * Only TCP listeners and UDP unconnected sockets can be
  4302. * selected. *sk* can also be NULL to reset any previous
  4303. * selection.
  4304. *
  4305. * *flags* argument can combination of following values:
  4306. *
  4307. * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
  4308. * socket selection, potentially done by a BPF program
  4309. * that ran before us.
  4310. *
  4311. * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
  4312. * load-balancing within reuseport group for the socket
  4313. * being selected.
  4314. *
  4315. * On success *ctx->sk* will point to the selected socket.
  4316. *
  4317. * Return
  4318. * 0 on success, or a negative errno in case of failure.
  4319. *
  4320. * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
  4321. * not compatible with packet family (*ctx->family*).
  4322. *
  4323. * * **-EEXIST** if socket has been already selected,
  4324. * potentially by another program, and
  4325. * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
  4326. *
  4327. * * **-EINVAL** if unsupported flags were specified.
  4328. *
  4329. * * **-EPROTOTYPE** if socket L4 protocol
  4330. * (*sk->protocol*) doesn't match packet protocol
  4331. * (*ctx->protocol*).
  4332. *
  4333. * * **-ESOCKTNOSUPPORT** if socket is not in allowed
  4334. * state (TCP listening or UDP unconnected).
  4335. *
  4336. * u64 bpf_ktime_get_boot_ns(void)
  4337. * Description
  4338. * Return the time elapsed since system boot, in nanoseconds.
  4339. * Does include the time the system was suspended.
  4340. * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
  4341. * Return
  4342. * Current *ktime*.
  4343. *
  4344. * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
  4345. * Description
  4346. * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
  4347. * out the format string.
  4348. * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
  4349. * the format string itself. The *data* and *data_len* are format string
  4350. * arguments. The *data* are a **u64** array and corresponding format string
  4351. * values are stored in the array. For strings and pointers where pointees
  4352. * are accessed, only the pointer values are stored in the *data* array.
  4353. * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
  4354. *
  4355. * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
  4356. * Reading kernel memory may fail due to either invalid address or
  4357. * valid address but requiring a major memory fault. If reading kernel memory
  4358. * fails, the string for **%s** will be an empty string, and the ip
  4359. * address for **%p{i,I}{4,6}** will be 0. Not returning error to
  4360. * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
  4361. * Return
  4362. * 0 on success, or a negative error in case of failure:
  4363. *
  4364. * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
  4365. * by returning 1 from bpf program.
  4366. *
  4367. * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
  4368. *
  4369. * **-E2BIG** if *fmt* contains too many format specifiers.
  4370. *
  4371. * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
  4372. *
  4373. * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
  4374. * Description
  4375. * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
  4376. * The *m* represents the seq_file. The *data* and *len* represent the
  4377. * data to write in bytes.
  4378. * Return
  4379. * 0 on success, or a negative error in case of failure:
  4380. *
  4381. * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
  4382. *
  4383. * u64 bpf_sk_cgroup_id(void *sk)
  4384. * Description
  4385. * Return the cgroup v2 id of the socket *sk*.
  4386. *
  4387. * *sk* must be a non-**NULL** pointer to a socket, e.g. one
  4388. * returned from **bpf_sk_lookup_xxx**\ (),
  4389. * **bpf_sk_fullsock**\ (), etc. The format of returned id is
  4390. * same as in **bpf_skb_cgroup_id**\ ().
  4391. *
  4392. * This helper is available only if the kernel was compiled with
  4393. * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
  4394. * Return
  4395. * The id is returned or 0 in case the id could not be retrieved.
  4396. *
  4397. * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
  4398. * Description
  4399. * Return id of cgroup v2 that is ancestor of cgroup associated
  4400. * with the *sk* at the *ancestor_level*. The root cgroup is at
  4401. * *ancestor_level* zero and each step down the hierarchy
  4402. * increments the level. If *ancestor_level* == level of cgroup
  4403. * associated with *sk*, then return value will be same as that
  4404. * of **bpf_sk_cgroup_id**\ ().
  4405. *
  4406. * The helper is useful to implement policies based on cgroups
  4407. * that are upper in hierarchy than immediate cgroup associated
  4408. * with *sk*.
  4409. *
  4410. * The format of returned id and helper limitations are same as in
  4411. * **bpf_sk_cgroup_id**\ ().
  4412. * Return
  4413. * The id is returned or 0 in case the id could not be retrieved.
  4414. *
  4415. * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
  4416. * Description
  4417. * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
  4418. * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4419. * of new data availability is sent.
  4420. * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4421. * of new data availability is sent unconditionally.
  4422. * If **0** is specified in *flags*, an adaptive notification
  4423. * of new data availability is sent.
  4424. *
  4425. * An adaptive notification is a notification sent whenever the user-space
  4426. * process has caught up and consumed all available payloads. In case the user-space
  4427. * process is still processing a previous payload, then no notification is needed
  4428. * as it will process the newly added payload automatically.
  4429. * Return
  4430. * 0 on success, or a negative error in case of failure.
  4431. *
  4432. * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
  4433. * Description
  4434. * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
  4435. * *flags* must be 0.
  4436. * Return
  4437. * Valid pointer with *size* bytes of memory available; NULL,
  4438. * otherwise.
  4439. *
  4440. * void bpf_ringbuf_submit(void *data, u64 flags)
  4441. * Description
  4442. * Submit reserved ring buffer sample, pointed to by *data*.
  4443. * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4444. * of new data availability is sent.
  4445. * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4446. * of new data availability is sent unconditionally.
  4447. * If **0** is specified in *flags*, an adaptive notification
  4448. * of new data availability is sent.
  4449. *
  4450. * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
  4451. * Return
  4452. * Nothing. Always succeeds.
  4453. *
  4454. * void bpf_ringbuf_discard(void *data, u64 flags)
  4455. * Description
  4456. * Discard reserved ring buffer sample, pointed to by *data*.
  4457. * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4458. * of new data availability is sent.
  4459. * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4460. * of new data availability is sent unconditionally.
  4461. * If **0** is specified in *flags*, an adaptive notification
  4462. * of new data availability is sent.
  4463. *
  4464. * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
  4465. * Return
  4466. * Nothing. Always succeeds.
  4467. *
  4468. * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
  4469. * Description
  4470. * Query various characteristics of provided ring buffer. What
  4471. * exactly is queries is determined by *flags*:
  4472. *
  4473. * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
  4474. * * **BPF_RB_RING_SIZE**: The size of ring buffer.
  4475. * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
  4476. * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
  4477. *
  4478. * Data returned is just a momentary snapshot of actual values
  4479. * and could be inaccurate, so this facility should be used to
  4480. * power heuristics and for reporting, not to make 100% correct
  4481. * calculation.
  4482. * Return
  4483. * Requested value, or 0, if *flags* are not recognized.
  4484. *
  4485. * long bpf_csum_level(struct sk_buff *skb, u64 level)
  4486. * Description
  4487. * Change the skbs checksum level by one layer up or down, or
  4488. * reset it entirely to none in order to have the stack perform
  4489. * checksum validation. The level is applicable to the following
  4490. * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
  4491. * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
  4492. * through **bpf_skb_adjust_room**\ () helper with passing in
  4493. * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
  4494. * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
  4495. * the UDP header is removed. Similarly, an encap of the latter
  4496. * into the former could be accompanied by a helper call to
  4497. * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
  4498. * skb is still intended to be processed in higher layers of the
  4499. * stack instead of just egressing at tc.
  4500. *
  4501. * There are three supported level settings at this time:
  4502. *
  4503. * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
  4504. * with CHECKSUM_UNNECESSARY.
  4505. * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
  4506. * with CHECKSUM_UNNECESSARY.
  4507. * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
  4508. * sets CHECKSUM_NONE to force checksum validation by the stack.
  4509. * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
  4510. * skb->csum_level.
  4511. * Return
  4512. * 0 on success, or a negative error in case of failure. In the
  4513. * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
  4514. * is returned or the error code -EACCES in case the skb is not
  4515. * subject to CHECKSUM_UNNECESSARY.
  4516. *
  4517. * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
  4518. * Description
  4519. * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
  4520. * Return
  4521. * *sk* if casting is valid, or **NULL** otherwise.
  4522. *
  4523. * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
  4524. * Description
  4525. * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
  4526. * Return
  4527. * *sk* if casting is valid, or **NULL** otherwise.
  4528. *
  4529. * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
  4530. * Description
  4531. * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
  4532. * Return
  4533. * *sk* if casting is valid, or **NULL** otherwise.
  4534. *
  4535. * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
  4536. * Description
  4537. * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
  4538. * Return
  4539. * *sk* if casting is valid, or **NULL** otherwise.
  4540. *
  4541. * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
  4542. * Description
  4543. * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
  4544. * Return
  4545. * *sk* if casting is valid, or **NULL** otherwise.
  4546. *
  4547. * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
  4548. * Description
  4549. * Return a user or a kernel stack in bpf program provided buffer.
  4550. * Note: the user stack will only be populated if the *task* is
  4551. * the current task; all other tasks will return -EOPNOTSUPP.
  4552. * To achieve this, the helper needs *task*, which is a valid
  4553. * pointer to **struct task_struct**. To store the stacktrace, the
  4554. * bpf program provides *buf* with a nonnegative *size*.
  4555. *
  4556. * The last argument, *flags*, holds the number of stack frames to
  4557. * skip (from 0 to 255), masked with
  4558. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  4559. * the following flags:
  4560. *
  4561. * **BPF_F_USER_STACK**
  4562. * Collect a user space stack instead of a kernel stack.
  4563. * The *task* must be the current task.
  4564. * **BPF_F_USER_BUILD_ID**
  4565. * Collect buildid+offset instead of ips for user stack,
  4566. * only valid if **BPF_F_USER_STACK** is also specified.
  4567. *
  4568. * **bpf_get_task_stack**\ () can collect up to
  4569. * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  4570. * to sufficient large buffer size. Note that
  4571. * this limit can be controlled with the **sysctl** program, and
  4572. * that it should be manually increased in order to profile long
  4573. * user stacks (such as stacks for Java programs). To do so, use:
  4574. *
  4575. * ::
  4576. *
  4577. * # sysctl kernel.perf_event_max_stack=<new value>
  4578. * Return
  4579. * The non-negative copied *buf* length equal to or less than
  4580. * *size* on success, or a negative error in case of failure.
  4581. *
  4582. * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
  4583. * Description
  4584. * Load header option. Support reading a particular TCP header
  4585. * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
  4586. *
  4587. * If *flags* is 0, it will search the option from the
  4588. * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
  4589. * has details on what skb_data contains under different
  4590. * *skops*\ **->op**.
  4591. *
  4592. * The first byte of the *searchby_res* specifies the
  4593. * kind that it wants to search.
  4594. *
  4595. * If the searching kind is an experimental kind
  4596. * (i.e. 253 or 254 according to RFC6994). It also
  4597. * needs to specify the "magic" which is either
  4598. * 2 bytes or 4 bytes. It then also needs to
  4599. * specify the size of the magic by using
  4600. * the 2nd byte which is "kind-length" of a TCP
  4601. * header option and the "kind-length" also
  4602. * includes the first 2 bytes "kind" and "kind-length"
  4603. * itself as a normal TCP header option also does.
  4604. *
  4605. * For example, to search experimental kind 254 with
  4606. * 2 byte magic 0xeB9F, the searchby_res should be
  4607. * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
  4608. *
  4609. * To search for the standard window scale option (3),
  4610. * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
  4611. * Note, kind-length must be 0 for regular option.
  4612. *
  4613. * Searching for No-Op (0) and End-of-Option-List (1) are
  4614. * not supported.
  4615. *
  4616. * *len* must be at least 2 bytes which is the minimal size
  4617. * of a header option.
  4618. *
  4619. * Supported flags:
  4620. *
  4621. * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
  4622. * saved_syn packet or the just-received syn packet.
  4623. *
  4624. * Return
  4625. * > 0 when found, the header option is copied to *searchby_res*.
  4626. * The return value is the total length copied. On failure, a
  4627. * negative error code is returned:
  4628. *
  4629. * **-EINVAL** if a parameter is invalid.
  4630. *
  4631. * **-ENOMSG** if the option is not found.
  4632. *
  4633. * **-ENOENT** if no syn packet is available when
  4634. * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
  4635. *
  4636. * **-ENOSPC** if there is not enough space. Only *len* number of
  4637. * bytes are copied.
  4638. *
  4639. * **-EFAULT** on failure to parse the header options in the
  4640. * packet.
  4641. *
  4642. * **-EPERM** if the helper cannot be used under the current
  4643. * *skops*\ **->op**.
  4644. *
  4645. * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
  4646. * Description
  4647. * Store header option. The data will be copied
  4648. * from buffer *from* with length *len* to the TCP header.
  4649. *
  4650. * The buffer *from* should have the whole option that
  4651. * includes the kind, kind-length, and the actual
  4652. * option data. The *len* must be at least kind-length
  4653. * long. The kind-length does not have to be 4 byte
  4654. * aligned. The kernel will take care of the padding
  4655. * and setting the 4 bytes aligned value to th->doff.
  4656. *
  4657. * This helper will check for duplicated option
  4658. * by searching the same option in the outgoing skb.
  4659. *
  4660. * This helper can only be called during
  4661. * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
  4662. *
  4663. * Return
  4664. * 0 on success, or negative error in case of failure:
  4665. *
  4666. * **-EINVAL** If param is invalid.
  4667. *
  4668. * **-ENOSPC** if there is not enough space in the header.
  4669. * Nothing has been written
  4670. *
  4671. * **-EEXIST** if the option already exists.
  4672. *
  4673. * **-EFAULT** on failure to parse the existing header options.
  4674. *
  4675. * **-EPERM** if the helper cannot be used under the current
  4676. * *skops*\ **->op**.
  4677. *
  4678. * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
  4679. * Description
  4680. * Reserve *len* bytes for the bpf header option. The
  4681. * space will be used by **bpf_store_hdr_opt**\ () later in
  4682. * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
  4683. *
  4684. * If **bpf_reserve_hdr_opt**\ () is called multiple times,
  4685. * the total number of bytes will be reserved.
  4686. *
  4687. * This helper can only be called during
  4688. * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
  4689. *
  4690. * Return
  4691. * 0 on success, or negative error in case of failure:
  4692. *
  4693. * **-EINVAL** if a parameter is invalid.
  4694. *
  4695. * **-ENOSPC** if there is not enough space in the header.
  4696. *
  4697. * **-EPERM** if the helper cannot be used under the current
  4698. * *skops*\ **->op**.
  4699. *
  4700. * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
  4701. * Description
  4702. * Get a bpf_local_storage from an *inode*.
  4703. *
  4704. * Logically, it could be thought of as getting the value from
  4705. * a *map* with *inode* as the **key**. From this
  4706. * perspective, the usage is not much different from
  4707. * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
  4708. * helper enforces the key must be an inode and the map must also
  4709. * be a **BPF_MAP_TYPE_INODE_STORAGE**.
  4710. *
  4711. * Underneath, the value is stored locally at *inode* instead of
  4712. * the *map*. The *map* is used as the bpf-local-storage
  4713. * "type". The bpf-local-storage "type" (i.e. the *map*) is
  4714. * searched against all bpf_local_storage residing at *inode*.
  4715. *
  4716. * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  4717. * used such that a new bpf_local_storage will be
  4718. * created if one does not exist. *value* can be used
  4719. * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  4720. * the initial value of a bpf_local_storage. If *value* is
  4721. * **NULL**, the new bpf_local_storage will be zero initialized.
  4722. * Return
  4723. * A bpf_local_storage pointer is returned on success.
  4724. *
  4725. * **NULL** if not found or there was an error in adding
  4726. * a new bpf_local_storage.
  4727. *
  4728. * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
  4729. * Description
  4730. * Delete a bpf_local_storage from an *inode*.
  4731. * Return
  4732. * 0 on success.
  4733. *
  4734. * **-ENOENT** if the bpf_local_storage cannot be found.
  4735. *
  4736. * long bpf_d_path(struct path *path, char *buf, u32 sz)
  4737. * Description
  4738. * Return full path for given **struct path** object, which
  4739. * needs to be the kernel BTF *path* object. The path is
  4740. * returned in the provided buffer *buf* of size *sz* and
  4741. * is zero terminated.
  4742. *
  4743. * Return
  4744. * On success, the strictly positive length of the string,
  4745. * including the trailing NUL character. On error, a negative
  4746. * value.
  4747. *
  4748. * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
  4749. * Description
  4750. * Read *size* bytes from user space address *user_ptr* and store
  4751. * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
  4752. * Return
  4753. * 0 on success, or a negative error in case of failure.
  4754. *
  4755. * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
  4756. * Description
  4757. * Use BTF to store a string representation of *ptr*->ptr in *str*,
  4758. * using *ptr*->type_id. This value should specify the type
  4759. * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
  4760. * can be used to look up vmlinux BTF type ids. Traversing the
  4761. * data structure using BTF, the type information and values are
  4762. * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
  4763. * the pointer data is carried out to avoid kernel crashes during
  4764. * operation. Smaller types can use string space on the stack;
  4765. * larger programs can use map data to store the string
  4766. * representation.
  4767. *
  4768. * The string can be subsequently shared with userspace via
  4769. * bpf_perf_event_output() or ring buffer interfaces.
  4770. * bpf_trace_printk() is to be avoided as it places too small
  4771. * a limit on string size to be useful.
  4772. *
  4773. * *flags* is a combination of
  4774. *
  4775. * **BTF_F_COMPACT**
  4776. * no formatting around type information
  4777. * **BTF_F_NONAME**
  4778. * no struct/union member names/types
  4779. * **BTF_F_PTR_RAW**
  4780. * show raw (unobfuscated) pointer values;
  4781. * equivalent to printk specifier %px.
  4782. * **BTF_F_ZERO**
  4783. * show zero-valued struct/union members; they
  4784. * are not displayed by default
  4785. *
  4786. * Return
  4787. * The number of bytes that were written (or would have been
  4788. * written if output had to be truncated due to string size),
  4789. * or a negative error in cases of failure.
  4790. *
  4791. * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
  4792. * Description
  4793. * Use BTF to write to seq_write a string representation of
  4794. * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
  4795. * *flags* are identical to those used for bpf_snprintf_btf.
  4796. * Return
  4797. * 0 on success or a negative error in case of failure.
  4798. *
  4799. * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
  4800. * Description
  4801. * See **bpf_get_cgroup_classid**\ () for the main description.
  4802. * This helper differs from **bpf_get_cgroup_classid**\ () in that
  4803. * the cgroup v1 net_cls class is retrieved only from the *skb*'s
  4804. * associated socket instead of the current process.
  4805. * Return
  4806. * The id is returned or 0 in case the id could not be retrieved.
  4807. *
  4808. * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
  4809. * Description
  4810. * Redirect the packet to another net device of index *ifindex*
  4811. * and fill in L2 addresses from neighboring subsystem. This helper
  4812. * is somewhat similar to **bpf_redirect**\ (), except that it
  4813. * populates L2 addresses as well, meaning, internally, the helper
  4814. * relies on the neighbor lookup for the L2 address of the nexthop.
  4815. *
  4816. * The helper will perform a FIB lookup based on the skb's
  4817. * networking header to get the address of the next hop, unless
  4818. * this is supplied by the caller in the *params* argument. The
  4819. * *plen* argument indicates the len of *params* and should be set
  4820. * to 0 if *params* is NULL.
  4821. *
  4822. * The *flags* argument is reserved and must be 0. The helper is
  4823. * currently only supported for tc BPF program types, and enabled
  4824. * for IPv4 and IPv6 protocols.
  4825. * Return
  4826. * The helper returns **TC_ACT_REDIRECT** on success or
  4827. * **TC_ACT_SHOT** on error.
  4828. *
  4829. * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
  4830. * Description
  4831. * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
  4832. * pointer to the percpu kernel variable on *cpu*. A ksym is an
  4833. * extern variable decorated with '__ksym'. For ksym, there is a
  4834. * global var (either static or global) defined of the same name
  4835. * in the kernel. The ksym is percpu if the global var is percpu.
  4836. * The returned pointer points to the global percpu var on *cpu*.
  4837. *
  4838. * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
  4839. * kernel, except that bpf_per_cpu_ptr() may return NULL. This
  4840. * happens if *cpu* is larger than nr_cpu_ids. The caller of
  4841. * bpf_per_cpu_ptr() must check the returned value.
  4842. * Return
  4843. * A pointer pointing to the kernel percpu variable on *cpu*, or
  4844. * NULL, if *cpu* is invalid.
  4845. *
  4846. * void *bpf_this_cpu_ptr(const void *percpu_ptr)
  4847. * Description
  4848. * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
  4849. * pointer to the percpu kernel variable on this cpu. See the
  4850. * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
  4851. *
  4852. * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
  4853. * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
  4854. * never return NULL.
  4855. * Return
  4856. * A pointer pointing to the kernel percpu variable on this cpu.
  4857. *
  4858. * long bpf_redirect_peer(u32 ifindex, u64 flags)
  4859. * Description
  4860. * Redirect the packet to another net device of index *ifindex*.
  4861. * This helper is somewhat similar to **bpf_redirect**\ (), except
  4862. * that the redirection happens to the *ifindex*' peer device and
  4863. * the netns switch takes place from ingress to ingress without
  4864. * going through the CPU's backlog queue.
  4865. *
  4866. * The *flags* argument is reserved and must be 0. The helper is
  4867. * currently only supported for tc BPF program types at the
  4868. * ingress hook and for veth and netkit target device types. The
  4869. * peer device must reside in a different network namespace.
  4870. * Return
  4871. * The helper returns **TC_ACT_REDIRECT** on success or
  4872. * **TC_ACT_SHOT** on error.
  4873. *
  4874. * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
  4875. * Description
  4876. * Get a bpf_local_storage from the *task*.
  4877. *
  4878. * Logically, it could be thought of as getting the value from
  4879. * a *map* with *task* as the **key**. From this
  4880. * perspective, the usage is not much different from
  4881. * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
  4882. * helper enforces the key must be a task_struct and the map must also
  4883. * be a **BPF_MAP_TYPE_TASK_STORAGE**.
  4884. *
  4885. * Underneath, the value is stored locally at *task* instead of
  4886. * the *map*. The *map* is used as the bpf-local-storage
  4887. * "type". The bpf-local-storage "type" (i.e. the *map*) is
  4888. * searched against all bpf_local_storage residing at *task*.
  4889. *
  4890. * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  4891. * used such that a new bpf_local_storage will be
  4892. * created if one does not exist. *value* can be used
  4893. * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  4894. * the initial value of a bpf_local_storage. If *value* is
  4895. * **NULL**, the new bpf_local_storage will be zero initialized.
  4896. * Return
  4897. * A bpf_local_storage pointer is returned on success.
  4898. *
  4899. * **NULL** if not found or there was an error in adding
  4900. * a new bpf_local_storage.
  4901. *
  4902. * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
  4903. * Description
  4904. * Delete a bpf_local_storage from a *task*.
  4905. * Return
  4906. * 0 on success.
  4907. *
  4908. * **-ENOENT** if the bpf_local_storage cannot be found.
  4909. *
  4910. * struct task_struct *bpf_get_current_task_btf(void)
  4911. * Description
  4912. * Return a BTF pointer to the "current" task.
  4913. * This pointer can also be used in helpers that accept an
  4914. * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
  4915. * Return
  4916. * Pointer to the current task.
  4917. *
  4918. * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
  4919. * Description
  4920. * Set or clear certain options on *bprm*:
  4921. *
  4922. * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
  4923. * which sets the **AT_SECURE** auxv for glibc. The bit
  4924. * is cleared if the flag is not specified.
  4925. * Return
  4926. * **-EINVAL** if invalid *flags* are passed, zero otherwise.
  4927. *
  4928. * u64 bpf_ktime_get_coarse_ns(void)
  4929. * Description
  4930. * Return a coarse-grained version of the time elapsed since
  4931. * system boot, in nanoseconds. Does not include time the system
  4932. * was suspended.
  4933. *
  4934. * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
  4935. * Return
  4936. * Current *ktime*.
  4937. *
  4938. * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
  4939. * Description
  4940. * Returns the stored IMA hash of the *inode* (if it's available).
  4941. * If the hash is larger than *size*, then only *size*
  4942. * bytes will be copied to *dst*
  4943. * Return
  4944. * The **hash_algo** is returned on success,
  4945. * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if
  4946. * invalid arguments are passed.
  4947. *
  4948. * struct socket *bpf_sock_from_file(struct file *file)
  4949. * Description
  4950. * If the given file represents a socket, returns the associated
  4951. * socket.
  4952. * Return
  4953. * A pointer to a struct socket on success or NULL if the file is
  4954. * not a socket.
  4955. *
  4956. * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
  4957. * Description
  4958. * Check packet size against exceeding MTU of net device (based
  4959. * on *ifindex*). This helper will likely be used in combination
  4960. * with helpers that adjust/change the packet size.
  4961. *
  4962. * The argument *len_diff* can be used for querying with a planned
  4963. * size change. This allows to check MTU prior to changing packet
  4964. * ctx. Providing a *len_diff* adjustment that is larger than the
  4965. * actual packet size (resulting in negative packet size) will in
  4966. * principle not exceed the MTU, which is why it is not considered
  4967. * a failure. Other BPF helpers are needed for performing the
  4968. * planned size change; therefore the responsibility for catching
  4969. * a negative packet size belongs in those helpers.
  4970. *
  4971. * Specifying *ifindex* zero means the MTU check is performed
  4972. * against the current net device. This is practical if this isn't
  4973. * used prior to redirect.
  4974. *
  4975. * On input *mtu_len* must be a valid pointer, else verifier will
  4976. * reject BPF program. If the value *mtu_len* is initialized to
  4977. * zero then the ctx packet size is use. When value *mtu_len* is
  4978. * provided as input this specify the L3 length that the MTU check
  4979. * is done against. Remember XDP and TC length operate at L2, but
  4980. * this value is L3 as this correlate to MTU and IP-header tot_len
  4981. * values which are L3 (similar behavior as bpf_fib_lookup).
  4982. *
  4983. * The Linux kernel route table can configure MTUs on a more
  4984. * specific per route level, which is not provided by this helper.
  4985. * For route level MTU checks use the **bpf_fib_lookup**\ ()
  4986. * helper.
  4987. *
  4988. * *ctx* is either **struct xdp_md** for XDP programs or
  4989. * **struct sk_buff** for tc cls_act programs.
  4990. *
  4991. * The *flags* argument can be a combination of one or more of the
  4992. * following values:
  4993. *
  4994. * **BPF_MTU_CHK_SEGS**
  4995. * This flag will only works for *ctx* **struct sk_buff**.
  4996. * If packet context contains extra packet segment buffers
  4997. * (often knows as GSO skb), then MTU check is harder to
  4998. * check at this point, because in transmit path it is
  4999. * possible for the skb packet to get re-segmented
  5000. * (depending on net device features). This could still be
  5001. * a MTU violation, so this flag enables performing MTU
  5002. * check against segments, with a different violation
  5003. * return code to tell it apart. Check cannot use len_diff.
  5004. *
  5005. * On return *mtu_len* pointer contains the MTU value of the net
  5006. * device. Remember the net device configured MTU is the L3 size,
  5007. * which is returned here and XDP and TC length operate at L2.
  5008. * Helper take this into account for you, but remember when using
  5009. * MTU value in your BPF-code.
  5010. *
  5011. * Return
  5012. * * 0 on success, and populate MTU value in *mtu_len* pointer.
  5013. *
  5014. * * < 0 if any input argument is invalid (*mtu_len* not updated)
  5015. *
  5016. * MTU violations return positive values, but also populate MTU
  5017. * value in *mtu_len* pointer, as this can be needed for
  5018. * implementing PMTU handing:
  5019. *
  5020. * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
  5021. * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
  5022. *
  5023. * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
  5024. * Description
  5025. * For each element in **map**, call **callback_fn** function with
  5026. * **map**, **callback_ctx** and other map-specific parameters.
  5027. * The **callback_fn** should be a static function and
  5028. * the **callback_ctx** should be a pointer to the stack.
  5029. * The **flags** is used to control certain aspects of the helper.
  5030. * Currently, the **flags** must be 0.
  5031. *
  5032. * The following are a list of supported map types and their
  5033. * respective expected callback signatures:
  5034. *
  5035. * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
  5036. * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
  5037. * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
  5038. *
  5039. * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
  5040. *
  5041. * For per_cpu maps, the map_value is the value on the cpu where the
  5042. * bpf_prog is running.
  5043. *
  5044. * If **callback_fn** return 0, the helper will continue to the next
  5045. * element. If return value is 1, the helper will skip the rest of
  5046. * elements and return. Other return values are not used now.
  5047. *
  5048. * Return
  5049. * The number of traversed map elements for success, **-EINVAL** for
  5050. * invalid **flags**.
  5051. *
  5052. * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
  5053. * Description
  5054. * Outputs a string into the **str** buffer of size **str_size**
  5055. * based on a format string stored in a read-only map pointed by
  5056. * **fmt**.
  5057. *
  5058. * Each format specifier in **fmt** corresponds to one u64 element
  5059. * in the **data** array. For strings and pointers where pointees
  5060. * are accessed, only the pointer values are stored in the *data*
  5061. * array. The *data_len* is the size of *data* in bytes - must be
  5062. * a multiple of 8.
  5063. *
  5064. * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
  5065. * memory. Reading kernel memory may fail due to either invalid
  5066. * address or valid address but requiring a major memory fault. If
  5067. * reading kernel memory fails, the string for **%s** will be an
  5068. * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
  5069. * Not returning error to bpf program is consistent with what
  5070. * **bpf_trace_printk**\ () does for now.
  5071. *
  5072. * Return
  5073. * The strictly positive length of the formatted string, including
  5074. * the trailing zero character. If the return value is greater than
  5075. * **str_size**, **str** contains a truncated string, guaranteed to
  5076. * be zero-terminated except when **str_size** is 0.
  5077. *
  5078. * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
  5079. *
  5080. * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
  5081. * Description
  5082. * Execute bpf syscall with given arguments.
  5083. * Return
  5084. * A syscall result.
  5085. *
  5086. * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
  5087. * Description
  5088. * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
  5089. * Return
  5090. * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
  5091. *
  5092. * long bpf_sys_close(u32 fd)
  5093. * Description
  5094. * Execute close syscall for given FD.
  5095. * Return
  5096. * A syscall result.
  5097. *
  5098. * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
  5099. * Description
  5100. * Initialize the timer.
  5101. * First 4 bits of *flags* specify clockid.
  5102. * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
  5103. * All other bits of *flags* are reserved.
  5104. * The verifier will reject the program if *timer* is not from
  5105. * the same *map*.
  5106. * Return
  5107. * 0 on success.
  5108. * **-EBUSY** if *timer* is already initialized.
  5109. * **-EINVAL** if invalid *flags* are passed.
  5110. * **-EPERM** if *timer* is in a map that doesn't have any user references.
  5111. * The user space should either hold a file descriptor to a map with timers
  5112. * or pin such map in bpffs. When map is unpinned or file descriptor is
  5113. * closed all timers in the map will be cancelled and freed.
  5114. *
  5115. * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
  5116. * Description
  5117. * Configure the timer to call *callback_fn* static function.
  5118. * Return
  5119. * 0 on success.
  5120. * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
  5121. * **-EPERM** if *timer* is in a map that doesn't have any user references.
  5122. * The user space should either hold a file descriptor to a map with timers
  5123. * or pin such map in bpffs. When map is unpinned or file descriptor is
  5124. * closed all timers in the map will be cancelled and freed.
  5125. *
  5126. * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
  5127. * Description
  5128. * Set timer expiration N nanoseconds from the current time. The
  5129. * configured callback will be invoked in soft irq context on some cpu
  5130. * and will not repeat unless another bpf_timer_start() is made.
  5131. * In such case the next invocation can migrate to a different cpu.
  5132. * Since struct bpf_timer is a field inside map element the map
  5133. * owns the timer. The bpf_timer_set_callback() will increment refcnt
  5134. * of BPF program to make sure that callback_fn code stays valid.
  5135. * When user space reference to a map reaches zero all timers
  5136. * in a map are cancelled and corresponding program's refcnts are
  5137. * decremented. This is done to make sure that Ctrl-C of a user
  5138. * process doesn't leave any timers running. If map is pinned in
  5139. * bpffs the callback_fn can re-arm itself indefinitely.
  5140. * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
  5141. * cancel and free the timer in the given map element.
  5142. * The map can contain timers that invoke callback_fn-s from different
  5143. * programs. The same callback_fn can serve different timers from
  5144. * different maps if key/value layout matches across maps.
  5145. * Every bpf_timer_set_callback() can have different callback_fn.
  5146. *
  5147. * *flags* can be one of:
  5148. *
  5149. * **BPF_F_TIMER_ABS**
  5150. * Start the timer in absolute expire value instead of the
  5151. * default relative one.
  5152. * **BPF_F_TIMER_CPU_PIN**
  5153. * Timer will be pinned to the CPU of the caller.
  5154. *
  5155. * Return
  5156. * 0 on success.
  5157. * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
  5158. * or invalid *flags* are passed.
  5159. *
  5160. * long bpf_timer_cancel(struct bpf_timer *timer)
  5161. * Description
  5162. * Cancel the timer and wait for callback_fn to finish if it was running.
  5163. * Return
  5164. * 0 if the timer was not active.
  5165. * 1 if the timer was active.
  5166. * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
  5167. * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
  5168. * own timer which would have led to a deadlock otherwise.
  5169. *
  5170. * u64 bpf_get_func_ip(void *ctx)
  5171. * Description
  5172. * Get address of the traced function (for tracing and kprobe programs).
  5173. *
  5174. * When called for kprobe program attached as uprobe it returns
  5175. * probe address for both entry and return uprobe.
  5176. *
  5177. * Return
  5178. * Address of the traced function for kprobe.
  5179. * 0 for kprobes placed within the function (not at the entry).
  5180. * Address of the probe for uprobe and return uprobe.
  5181. *
  5182. * u64 bpf_get_attach_cookie(void *ctx)
  5183. * Description
  5184. * Get bpf_cookie value provided (optionally) during the program
  5185. * attachment. It might be different for each individual
  5186. * attachment, even if BPF program itself is the same.
  5187. * Expects BPF program context *ctx* as a first argument.
  5188. *
  5189. * Supported for the following program types:
  5190. * - kprobe/uprobe;
  5191. * - tracepoint;
  5192. * - perf_event.
  5193. * Return
  5194. * Value specified by user at BPF link creation/attachment time
  5195. * or 0, if it was not specified.
  5196. *
  5197. * long bpf_task_pt_regs(struct task_struct *task)
  5198. * Description
  5199. * Get the struct pt_regs associated with **task**.
  5200. * Return
  5201. * A pointer to struct pt_regs.
  5202. *
  5203. * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
  5204. * Description
  5205. * Get branch trace from hardware engines like Intel LBR. The
  5206. * hardware engine is stopped shortly after the helper is
  5207. * called. Therefore, the user need to filter branch entries
  5208. * based on the actual use case. To capture branch trace
  5209. * before the trigger point of the BPF program, the helper
  5210. * should be called at the beginning of the BPF program.
  5211. *
  5212. * The data is stored as struct perf_branch_entry into output
  5213. * buffer *entries*. *size* is the size of *entries* in bytes.
  5214. * *flags* is reserved for now and must be zero.
  5215. *
  5216. * Return
  5217. * On success, number of bytes written to *buf*. On error, a
  5218. * negative value.
  5219. *
  5220. * **-EINVAL** if *flags* is not zero.
  5221. *
  5222. * **-ENOENT** if architecture does not support branch records.
  5223. *
  5224. * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
  5225. * Description
  5226. * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
  5227. * to format and can handle more format args as a result.
  5228. *
  5229. * Arguments are to be used as in **bpf_seq_printf**\ () helper.
  5230. * Return
  5231. * The number of bytes written to the buffer, or a negative error
  5232. * in case of failure.
  5233. *
  5234. * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
  5235. * Description
  5236. * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
  5237. * Return
  5238. * *sk* if casting is valid, or **NULL** otherwise.
  5239. *
  5240. * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
  5241. * Description
  5242. * Get the address of a kernel symbol, returned in *res*. *res* is
  5243. * set to 0 if the symbol is not found.
  5244. * Return
  5245. * On success, zero. On error, a negative value.
  5246. *
  5247. * **-EINVAL** if *flags* is not zero.
  5248. *
  5249. * **-EINVAL** if string *name* is not the same size as *name_sz*.
  5250. *
  5251. * **-ENOENT** if symbol is not found.
  5252. *
  5253. * **-EPERM** if caller does not have permission to obtain kernel address.
  5254. *
  5255. * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
  5256. * Description
  5257. * Find vma of *task* that contains *addr*, call *callback_fn*
  5258. * function with *task*, *vma*, and *callback_ctx*.
  5259. * The *callback_fn* should be a static function and
  5260. * the *callback_ctx* should be a pointer to the stack.
  5261. * The *flags* is used to control certain aspects of the helper.
  5262. * Currently, the *flags* must be 0.
  5263. *
  5264. * The expected callback signature is
  5265. *
  5266. * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
  5267. *
  5268. * Return
  5269. * 0 on success.
  5270. * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
  5271. * **-EBUSY** if failed to try lock mmap_lock.
  5272. * **-EINVAL** for invalid **flags**.
  5273. *
  5274. * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
  5275. * Description
  5276. * For **nr_loops**, call **callback_fn** function
  5277. * with **callback_ctx** as the context parameter.
  5278. * The **callback_fn** should be a static function and
  5279. * the **callback_ctx** should be a pointer to the stack.
  5280. * The **flags** is used to control certain aspects of the helper.
  5281. * Currently, the **flags** must be 0. Currently, nr_loops is
  5282. * limited to 1 << 23 (~8 million) loops.
  5283. *
  5284. * long (\*callback_fn)(u32 index, void \*ctx);
  5285. *
  5286. * where **index** is the current index in the loop. The index
  5287. * is zero-indexed.
  5288. *
  5289. * If **callback_fn** returns 0, the helper will continue to the next
  5290. * loop. If return value is 1, the helper will skip the rest of
  5291. * the loops and return. Other return values are not used now,
  5292. * and will be rejected by the verifier.
  5293. *
  5294. * Return
  5295. * The number of loops performed, **-EINVAL** for invalid **flags**,
  5296. * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
  5297. *
  5298. * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
  5299. * Description
  5300. * Do strncmp() between **s1** and **s2**. **s1** doesn't need
  5301. * to be null-terminated and **s1_sz** is the maximum storage
  5302. * size of **s1**. **s2** must be a read-only string.
  5303. * Return
  5304. * An integer less than, equal to, or greater than zero
  5305. * if the first **s1_sz** bytes of **s1** is found to be
  5306. * less than, to match, or be greater than **s2**.
  5307. *
  5308. * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
  5309. * Description
  5310. * Get **n**-th argument register (zero based) of the traced function (for tracing programs)
  5311. * returned in **value**.
  5312. *
  5313. * Return
  5314. * 0 on success.
  5315. * **-EINVAL** if n >= argument register count of traced function.
  5316. *
  5317. * long bpf_get_func_ret(void *ctx, u64 *value)
  5318. * Description
  5319. * Get return value of the traced function (for tracing programs)
  5320. * in **value**.
  5321. *
  5322. * Return
  5323. * 0 on success.
  5324. * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
  5325. *
  5326. * long bpf_get_func_arg_cnt(void *ctx)
  5327. * Description
  5328. * Get number of registers of the traced function (for tracing programs) where
  5329. * function arguments are stored in these registers.
  5330. *
  5331. * Return
  5332. * The number of argument registers of the traced function.
  5333. *
  5334. * int bpf_get_retval(void)
  5335. * Description
  5336. * Get the BPF program's return value that will be returned to the upper layers.
  5337. *
  5338. * This helper is currently supported by cgroup programs and only by the hooks
  5339. * where BPF program's return value is returned to the userspace via errno.
  5340. * Return
  5341. * The BPF program's return value.
  5342. *
  5343. * int bpf_set_retval(int retval)
  5344. * Description
  5345. * Set the BPF program's return value that will be returned to the upper layers.
  5346. *
  5347. * This helper is currently supported by cgroup programs and only by the hooks
  5348. * where BPF program's return value is returned to the userspace via errno.
  5349. *
  5350. * Note that there is the following corner case where the program exports an error
  5351. * via bpf_set_retval but signals success via 'return 1':
  5352. *
  5353. * bpf_set_retval(-EPERM);
  5354. * return 1;
  5355. *
  5356. * In this case, the BPF program's return value will use helper's -EPERM. This
  5357. * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
  5358. *
  5359. * Return
  5360. * 0 on success, or a negative error in case of failure.
  5361. *
  5362. * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
  5363. * Description
  5364. * Get the total size of a given xdp buff (linear and paged area)
  5365. * Return
  5366. * The total size of a given xdp buffer.
  5367. *
  5368. * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
  5369. * Description
  5370. * This helper is provided as an easy way to load data from a
  5371. * xdp buffer. It can be used to load *len* bytes from *offset* from
  5372. * the frame associated to *xdp_md*, into the buffer pointed by
  5373. * *buf*.
  5374. * Return
  5375. * 0 on success, or a negative error in case of failure.
  5376. *
  5377. * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
  5378. * Description
  5379. * Store *len* bytes from buffer *buf* into the frame
  5380. * associated to *xdp_md*, at *offset*.
  5381. * Return
  5382. * 0 on success, or a negative error in case of failure.
  5383. *
  5384. * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
  5385. * Description
  5386. * Read *size* bytes from user space address *user_ptr* in *tsk*'s
  5387. * address space, and stores the data in *dst*. *flags* is not
  5388. * used yet and is provided for future extensibility. This helper
  5389. * can only be used by sleepable programs.
  5390. * Return
  5391. * 0 on success, or a negative error in case of failure. On error
  5392. * *dst* buffer is zeroed out.
  5393. *
  5394. * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
  5395. * Description
  5396. * Change the __sk_buff->tstamp_type to *tstamp_type*
  5397. * and set *tstamp* to the __sk_buff->tstamp together.
  5398. *
  5399. * If there is no need to change the __sk_buff->tstamp_type,
  5400. * the tstamp value can be directly written to __sk_buff->tstamp
  5401. * instead.
  5402. *
  5403. * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
  5404. * will be kept during bpf_redirect_*(). A non zero
  5405. * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
  5406. * *tstamp_type*.
  5407. *
  5408. * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
  5409. * with a zero *tstamp*.
  5410. *
  5411. * Only IPv4 and IPv6 skb->protocol are supported.
  5412. *
  5413. * This function is most useful when it needs to set a
  5414. * mono delivery time to __sk_buff->tstamp and then
  5415. * bpf_redirect_*() to the egress of an iface. For example,
  5416. * changing the (rcv) timestamp in __sk_buff->tstamp at
  5417. * ingress to a mono delivery time and then bpf_redirect_*()
  5418. * to sch_fq@phy-dev.
  5419. * Return
  5420. * 0 on success.
  5421. * **-EINVAL** for invalid input
  5422. * **-EOPNOTSUPP** for unsupported protocol
  5423. *
  5424. * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
  5425. * Description
  5426. * Returns a calculated IMA hash of the *file*.
  5427. * If the hash is larger than *size*, then only *size*
  5428. * bytes will be copied to *dst*
  5429. * Return
  5430. * The **hash_algo** is returned on success,
  5431. * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if
  5432. * invalid arguments are passed.
  5433. *
  5434. * void *bpf_kptr_xchg(void *dst, void *ptr)
  5435. * Description
  5436. * Exchange kptr at pointer *dst* with *ptr*, and return the old value.
  5437. * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise
  5438. * it must be a referenced pointer which will be released when this helper
  5439. * is called.
  5440. * Return
  5441. * The old value of kptr (which can be NULL). The returned pointer
  5442. * if not NULL, is a reference which must be released using its
  5443. * corresponding release function, or moved into a BPF map before
  5444. * program exit.
  5445. *
  5446. * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
  5447. * Description
  5448. * Perform a lookup in *percpu map* for an entry associated to
  5449. * *key* on *cpu*.
  5450. * Return
  5451. * Map value associated to *key* on *cpu*, or **NULL** if no entry
  5452. * was found or *cpu* is invalid.
  5453. *
  5454. * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
  5455. * Description
  5456. * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
  5457. * Return
  5458. * *sk* if casting is valid, or **NULL** otherwise.
  5459. *
  5460. * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
  5461. * Description
  5462. * Get a dynptr to local memory *data*.
  5463. *
  5464. * *data* must be a ptr to a map value.
  5465. * The maximum *size* supported is DYNPTR_MAX_SIZE.
  5466. * *flags* is currently unused.
  5467. * Return
  5468. * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
  5469. * -EINVAL if flags is not 0.
  5470. *
  5471. * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
  5472. * Description
  5473. * Reserve *size* bytes of payload in a ring buffer *ringbuf*
  5474. * through the dynptr interface. *flags* must be 0.
  5475. *
  5476. * Please note that a corresponding bpf_ringbuf_submit_dynptr or
  5477. * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
  5478. * reservation fails. This is enforced by the verifier.
  5479. * Return
  5480. * 0 on success, or a negative error in case of failure.
  5481. *
  5482. * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
  5483. * Description
  5484. * Submit reserved ring buffer sample, pointed to by *data*,
  5485. * through the dynptr interface. This is a no-op if the dynptr is
  5486. * invalid/null.
  5487. *
  5488. * For more information on *flags*, please see
  5489. * 'bpf_ringbuf_submit'.
  5490. * Return
  5491. * Nothing. Always succeeds.
  5492. *
  5493. * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
  5494. * Description
  5495. * Discard reserved ring buffer sample through the dynptr
  5496. * interface. This is a no-op if the dynptr is invalid/null.
  5497. *
  5498. * For more information on *flags*, please see
  5499. * 'bpf_ringbuf_discard'.
  5500. * Return
  5501. * Nothing. Always succeeds.
  5502. *
  5503. * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
  5504. * Description
  5505. * Read *len* bytes from *src* into *dst*, starting from *offset*
  5506. * into *src*.
  5507. * *flags* is currently unused.
  5508. * Return
  5509. * 0 on success, -E2BIG if *offset* + *len* exceeds the length
  5510. * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
  5511. * *flags* is not 0.
  5512. *
  5513. * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
  5514. * Description
  5515. * Write *len* bytes from *src* into *dst*, starting from *offset*
  5516. * into *dst*.
  5517. *
  5518. * *flags* must be 0 except for skb-type dynptrs.
  5519. *
  5520. * For skb-type dynptrs:
  5521. * * All data slices of the dynptr are automatically
  5522. * invalidated after **bpf_dynptr_write**\ (). This is
  5523. * because writing may pull the skb and change the
  5524. * underlying packet buffer.
  5525. *
  5526. * * For *flags*, please see the flags accepted by
  5527. * **bpf_skb_store_bytes**\ ().
  5528. * Return
  5529. * 0 on success, -E2BIG if *offset* + *len* exceeds the length
  5530. * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
  5531. * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
  5532. * other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
  5533. *
  5534. * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
  5535. * Description
  5536. * Get a pointer to the underlying dynptr data.
  5537. *
  5538. * *len* must be a statically known value. The returned data slice
  5539. * is invalidated whenever the dynptr is invalidated.
  5540. *
  5541. * skb and xdp type dynptrs may not use bpf_dynptr_data. They should
  5542. * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
  5543. * Return
  5544. * Pointer to the underlying dynptr data, NULL if the dynptr is
  5545. * read-only, if the dynptr is invalid, or if the offset and length
  5546. * is out of bounds.
  5547. *
  5548. * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
  5549. * Description
  5550. * Try to issue a SYN cookie for the packet with corresponding
  5551. * IPv4/TCP headers, *iph* and *th*, without depending on a
  5552. * listening socket.
  5553. *
  5554. * *iph* points to the IPv4 header.
  5555. *
  5556. * *th* points to the start of the TCP header, while *th_len*
  5557. * contains the length of the TCP header (at least
  5558. * **sizeof**\ (**struct tcphdr**)).
  5559. * Return
  5560. * On success, lower 32 bits hold the generated SYN cookie in
  5561. * followed by 16 bits which hold the MSS value for that cookie,
  5562. * and the top 16 bits are unused.
  5563. *
  5564. * On failure, the returned value is one of the following:
  5565. *
  5566. * **-EINVAL** if *th_len* is invalid.
  5567. *
  5568. * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
  5569. * Description
  5570. * Try to issue a SYN cookie for the packet with corresponding
  5571. * IPv6/TCP headers, *iph* and *th*, without depending on a
  5572. * listening socket.
  5573. *
  5574. * *iph* points to the IPv6 header.
  5575. *
  5576. * *th* points to the start of the TCP header, while *th_len*
  5577. * contains the length of the TCP header (at least
  5578. * **sizeof**\ (**struct tcphdr**)).
  5579. * Return
  5580. * On success, lower 32 bits hold the generated SYN cookie in
  5581. * followed by 16 bits which hold the MSS value for that cookie,
  5582. * and the top 16 bits are unused.
  5583. *
  5584. * On failure, the returned value is one of the following:
  5585. *
  5586. * **-EINVAL** if *th_len* is invalid.
  5587. *
  5588. * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
  5589. *
  5590. * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
  5591. * Description
  5592. * Check whether *iph* and *th* contain a valid SYN cookie ACK
  5593. * without depending on a listening socket.
  5594. *
  5595. * *iph* points to the IPv4 header.
  5596. *
  5597. * *th* points to the TCP header.
  5598. * Return
  5599. * 0 if *iph* and *th* are a valid SYN cookie ACK.
  5600. *
  5601. * On failure, the returned value is one of the following:
  5602. *
  5603. * **-EACCES** if the SYN cookie is not valid.
  5604. *
  5605. * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
  5606. * Description
  5607. * Check whether *iph* and *th* contain a valid SYN cookie ACK
  5608. * without depending on a listening socket.
  5609. *
  5610. * *iph* points to the IPv6 header.
  5611. *
  5612. * *th* points to the TCP header.
  5613. * Return
  5614. * 0 if *iph* and *th* are a valid SYN cookie ACK.
  5615. *
  5616. * On failure, the returned value is one of the following:
  5617. *
  5618. * **-EACCES** if the SYN cookie is not valid.
  5619. *
  5620. * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
  5621. *
  5622. * u64 bpf_ktime_get_tai_ns(void)
  5623. * Description
  5624. * A nonsettable system-wide clock derived from wall-clock time but
  5625. * ignoring leap seconds. This clock does not experience
  5626. * discontinuities and backwards jumps caused by NTP inserting leap
  5627. * seconds as CLOCK_REALTIME does.
  5628. *
  5629. * See: **clock_gettime**\ (**CLOCK_TAI**)
  5630. * Return
  5631. * Current *ktime*.
  5632. *
  5633. * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
  5634. * Description
  5635. * Drain samples from the specified user ring buffer, and invoke
  5636. * the provided callback for each such sample:
  5637. *
  5638. * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
  5639. *
  5640. * If **callback_fn** returns 0, the helper will continue to try
  5641. * and drain the next sample, up to a maximum of
  5642. * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
  5643. * the helper will skip the rest of the samples and return. Other
  5644. * return values are not used now, and will be rejected by the
  5645. * verifier.
  5646. * Return
  5647. * The number of drained samples if no error was encountered while
  5648. * draining samples, or 0 if no samples were present in the ring
  5649. * buffer. If a user-space producer was epoll-waiting on this map,
  5650. * and at least one sample was drained, they will receive an event
  5651. * notification notifying them of available space in the ring
  5652. * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
  5653. * function, no wakeup notification will be sent. If the
  5654. * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
  5655. * be sent even if no sample was drained.
  5656. *
  5657. * On failure, the returned value is one of the following:
  5658. *
  5659. * **-EBUSY** if the ring buffer is contended, and another calling
  5660. * context was concurrently draining the ring buffer.
  5661. *
  5662. * **-EINVAL** if user-space is not properly tracking the ring
  5663. * buffer due to the producer position not being aligned to 8
  5664. * bytes, a sample not being aligned to 8 bytes, or the producer
  5665. * position not matching the advertised length of a sample.
  5666. *
  5667. * **-E2BIG** if user-space has tried to publish a sample which is
  5668. * larger than the size of the ring buffer, or which cannot fit
  5669. * within a struct bpf_dynptr.
  5670. *
  5671. * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
  5672. * Description
  5673. * Get a bpf_local_storage from the *cgroup*.
  5674. *
  5675. * Logically, it could be thought of as getting the value from
  5676. * a *map* with *cgroup* as the **key**. From this
  5677. * perspective, the usage is not much different from
  5678. * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
  5679. * helper enforces the key must be a cgroup struct and the map must also
  5680. * be a **BPF_MAP_TYPE_CGRP_STORAGE**.
  5681. *
  5682. * In reality, the local-storage value is embedded directly inside of the
  5683. * *cgroup* object itself, rather than being located in the
  5684. * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
  5685. * queried for some *map* on a *cgroup* object, the kernel will perform an
  5686. * O(n) iteration over all of the live local-storage values for that
  5687. * *cgroup* object until the local-storage value for the *map* is found.
  5688. *
  5689. * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  5690. * used such that a new bpf_local_storage will be
  5691. * created if one does not exist. *value* can be used
  5692. * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  5693. * the initial value of a bpf_local_storage. If *value* is
  5694. * **NULL**, the new bpf_local_storage will be zero initialized.
  5695. * Return
  5696. * A bpf_local_storage pointer is returned on success.
  5697. *
  5698. * **NULL** if not found or there was an error in adding
  5699. * a new bpf_local_storage.
  5700. *
  5701. * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
  5702. * Description
  5703. * Delete a bpf_local_storage from a *cgroup*.
  5704. * Return
  5705. * 0 on success.
  5706. *
  5707. * **-ENOENT** if the bpf_local_storage cannot be found.
  5708. */
  5709. #define ___BPF_FUNC_MAPPER(FN, ctx...) \
  5710. FN(unspec, 0, ##ctx) \
  5711. FN(map_lookup_elem, 1, ##ctx) \
  5712. FN(map_update_elem, 2, ##ctx) \
  5713. FN(map_delete_elem, 3, ##ctx) \
  5714. FN(probe_read, 4, ##ctx) \
  5715. FN(ktime_get_ns, 5, ##ctx) \
  5716. FN(trace_printk, 6, ##ctx) \
  5717. FN(get_prandom_u32, 7, ##ctx) \
  5718. FN(get_smp_processor_id, 8, ##ctx) \
  5719. FN(skb_store_bytes, 9, ##ctx) \
  5720. FN(l3_csum_replace, 10, ##ctx) \
  5721. FN(l4_csum_replace, 11, ##ctx) \
  5722. FN(tail_call, 12, ##ctx) \
  5723. FN(clone_redirect, 13, ##ctx) \
  5724. FN(get_current_pid_tgid, 14, ##ctx) \
  5725. FN(get_current_uid_gid, 15, ##ctx) \
  5726. FN(get_current_comm, 16, ##ctx) \
  5727. FN(get_cgroup_classid, 17, ##ctx) \
  5728. FN(skb_vlan_push, 18, ##ctx) \
  5729. FN(skb_vlan_pop, 19, ##ctx) \
  5730. FN(skb_get_tunnel_key, 20, ##ctx) \
  5731. FN(skb_set_tunnel_key, 21, ##ctx) \
  5732. FN(perf_event_read, 22, ##ctx) \
  5733. FN(redirect, 23, ##ctx) \
  5734. FN(get_route_realm, 24, ##ctx) \
  5735. FN(perf_event_output, 25, ##ctx) \
  5736. FN(skb_load_bytes, 26, ##ctx) \
  5737. FN(get_stackid, 27, ##ctx) \
  5738. FN(csum_diff, 28, ##ctx) \
  5739. FN(skb_get_tunnel_opt, 29, ##ctx) \
  5740. FN(skb_set_tunnel_opt, 30, ##ctx) \
  5741. FN(skb_change_proto, 31, ##ctx) \
  5742. FN(skb_change_type, 32, ##ctx) \
  5743. FN(skb_under_cgroup, 33, ##ctx) \
  5744. FN(get_hash_recalc, 34, ##ctx) \
  5745. FN(get_current_task, 35, ##ctx) \
  5746. FN(probe_write_user, 36, ##ctx) \
  5747. FN(current_task_under_cgroup, 37, ##ctx) \
  5748. FN(skb_change_tail, 38, ##ctx) \
  5749. FN(skb_pull_data, 39, ##ctx) \
  5750. FN(csum_update, 40, ##ctx) \
  5751. FN(set_hash_invalid, 41, ##ctx) \
  5752. FN(get_numa_node_id, 42, ##ctx) \
  5753. FN(skb_change_head, 43, ##ctx) \
  5754. FN(xdp_adjust_head, 44, ##ctx) \
  5755. FN(probe_read_str, 45, ##ctx) \
  5756. FN(get_socket_cookie, 46, ##ctx) \
  5757. FN(get_socket_uid, 47, ##ctx) \
  5758. FN(set_hash, 48, ##ctx) \
  5759. FN(setsockopt, 49, ##ctx) \
  5760. FN(skb_adjust_room, 50, ##ctx) \
  5761. FN(redirect_map, 51, ##ctx) \
  5762. FN(sk_redirect_map, 52, ##ctx) \
  5763. FN(sock_map_update, 53, ##ctx) \
  5764. FN(xdp_adjust_meta, 54, ##ctx) \
  5765. FN(perf_event_read_value, 55, ##ctx) \
  5766. FN(perf_prog_read_value, 56, ##ctx) \
  5767. FN(getsockopt, 57, ##ctx) \
  5768. FN(override_return, 58, ##ctx) \
  5769. FN(sock_ops_cb_flags_set, 59, ##ctx) \
  5770. FN(msg_redirect_map, 60, ##ctx) \
  5771. FN(msg_apply_bytes, 61, ##ctx) \
  5772. FN(msg_cork_bytes, 62, ##ctx) \
  5773. FN(msg_pull_data, 63, ##ctx) \
  5774. FN(bind, 64, ##ctx) \
  5775. FN(xdp_adjust_tail, 65, ##ctx) \
  5776. FN(skb_get_xfrm_state, 66, ##ctx) \
  5777. FN(get_stack, 67, ##ctx) \
  5778. FN(skb_load_bytes_relative, 68, ##ctx) \
  5779. FN(fib_lookup, 69, ##ctx) \
  5780. FN(sock_hash_update, 70, ##ctx) \
  5781. FN(msg_redirect_hash, 71, ##ctx) \
  5782. FN(sk_redirect_hash, 72, ##ctx) \
  5783. FN(lwt_push_encap, 73, ##ctx) \
  5784. FN(lwt_seg6_store_bytes, 74, ##ctx) \
  5785. FN(lwt_seg6_adjust_srh, 75, ##ctx) \
  5786. FN(lwt_seg6_action, 76, ##ctx) \
  5787. FN(rc_repeat, 77, ##ctx) \
  5788. FN(rc_keydown, 78, ##ctx) \
  5789. FN(skb_cgroup_id, 79, ##ctx) \
  5790. FN(get_current_cgroup_id, 80, ##ctx) \
  5791. FN(get_local_storage, 81, ##ctx) \
  5792. FN(sk_select_reuseport, 82, ##ctx) \
  5793. FN(skb_ancestor_cgroup_id, 83, ##ctx) \
  5794. FN(sk_lookup_tcp, 84, ##ctx) \
  5795. FN(sk_lookup_udp, 85, ##ctx) \
  5796. FN(sk_release, 86, ##ctx) \
  5797. FN(map_push_elem, 87, ##ctx) \
  5798. FN(map_pop_elem, 88, ##ctx) \
  5799. FN(map_peek_elem, 89, ##ctx) \
  5800. FN(msg_push_data, 90, ##ctx) \
  5801. FN(msg_pop_data, 91, ##ctx) \
  5802. FN(rc_pointer_rel, 92, ##ctx) \
  5803. FN(spin_lock, 93, ##ctx) \
  5804. FN(spin_unlock, 94, ##ctx) \
  5805. FN(sk_fullsock, 95, ##ctx) \
  5806. FN(tcp_sock, 96, ##ctx) \
  5807. FN(skb_ecn_set_ce, 97, ##ctx) \
  5808. FN(get_listener_sock, 98, ##ctx) \
  5809. FN(skc_lookup_tcp, 99, ##ctx) \
  5810. FN(tcp_check_syncookie, 100, ##ctx) \
  5811. FN(sysctl_get_name, 101, ##ctx) \
  5812. FN(sysctl_get_current_value, 102, ##ctx) \
  5813. FN(sysctl_get_new_value, 103, ##ctx) \
  5814. FN(sysctl_set_new_value, 104, ##ctx) \
  5815. FN(strtol, 105, ##ctx) \
  5816. FN(strtoul, 106, ##ctx) \
  5817. FN(sk_storage_get, 107, ##ctx) \
  5818. FN(sk_storage_delete, 108, ##ctx) \
  5819. FN(send_signal, 109, ##ctx) \
  5820. FN(tcp_gen_syncookie, 110, ##ctx) \
  5821. FN(skb_output, 111, ##ctx) \
  5822. FN(probe_read_user, 112, ##ctx) \
  5823. FN(probe_read_kernel, 113, ##ctx) \
  5824. FN(probe_read_user_str, 114, ##ctx) \
  5825. FN(probe_read_kernel_str, 115, ##ctx) \
  5826. FN(tcp_send_ack, 116, ##ctx) \
  5827. FN(send_signal_thread, 117, ##ctx) \
  5828. FN(jiffies64, 118, ##ctx) \
  5829. FN(read_branch_records, 119, ##ctx) \
  5830. FN(get_ns_current_pid_tgid, 120, ##ctx) \
  5831. FN(xdp_output, 121, ##ctx) \
  5832. FN(get_netns_cookie, 122, ##ctx) \
  5833. FN(get_current_ancestor_cgroup_id, 123, ##ctx) \
  5834. FN(sk_assign, 124, ##ctx) \
  5835. FN(ktime_get_boot_ns, 125, ##ctx) \
  5836. FN(seq_printf, 126, ##ctx) \
  5837. FN(seq_write, 127, ##ctx) \
  5838. FN(sk_cgroup_id, 128, ##ctx) \
  5839. FN(sk_ancestor_cgroup_id, 129, ##ctx) \
  5840. FN(ringbuf_output, 130, ##ctx) \
  5841. FN(ringbuf_reserve, 131, ##ctx) \
  5842. FN(ringbuf_submit, 132, ##ctx) \
  5843. FN(ringbuf_discard, 133, ##ctx) \
  5844. FN(ringbuf_query, 134, ##ctx) \
  5845. FN(csum_level, 135, ##ctx) \
  5846. FN(skc_to_tcp6_sock, 136, ##ctx) \
  5847. FN(skc_to_tcp_sock, 137, ##ctx) \
  5848. FN(skc_to_tcp_timewait_sock, 138, ##ctx) \
  5849. FN(skc_to_tcp_request_sock, 139, ##ctx) \
  5850. FN(skc_to_udp6_sock, 140, ##ctx) \
  5851. FN(get_task_stack, 141, ##ctx) \
  5852. FN(load_hdr_opt, 142, ##ctx) \
  5853. FN(store_hdr_opt, 143, ##ctx) \
  5854. FN(reserve_hdr_opt, 144, ##ctx) \
  5855. FN(inode_storage_get, 145, ##ctx) \
  5856. FN(inode_storage_delete, 146, ##ctx) \
  5857. FN(d_path, 147, ##ctx) \
  5858. FN(copy_from_user, 148, ##ctx) \
  5859. FN(snprintf_btf, 149, ##ctx) \
  5860. FN(seq_printf_btf, 150, ##ctx) \
  5861. FN(skb_cgroup_classid, 151, ##ctx) \
  5862. FN(redirect_neigh, 152, ##ctx) \
  5863. FN(per_cpu_ptr, 153, ##ctx) \
  5864. FN(this_cpu_ptr, 154, ##ctx) \
  5865. FN(redirect_peer, 155, ##ctx) \
  5866. FN(task_storage_get, 156, ##ctx) \
  5867. FN(task_storage_delete, 157, ##ctx) \
  5868. FN(get_current_task_btf, 158, ##ctx) \
  5869. FN(bprm_opts_set, 159, ##ctx) \
  5870. FN(ktime_get_coarse_ns, 160, ##ctx) \
  5871. FN(ima_inode_hash, 161, ##ctx) \
  5872. FN(sock_from_file, 162, ##ctx) \
  5873. FN(check_mtu, 163, ##ctx) \
  5874. FN(for_each_map_elem, 164, ##ctx) \
  5875. FN(snprintf, 165, ##ctx) \
  5876. FN(sys_bpf, 166, ##ctx) \
  5877. FN(btf_find_by_name_kind, 167, ##ctx) \
  5878. FN(sys_close, 168, ##ctx) \
  5879. FN(timer_init, 169, ##ctx) \
  5880. FN(timer_set_callback, 170, ##ctx) \
  5881. FN(timer_start, 171, ##ctx) \
  5882. FN(timer_cancel, 172, ##ctx) \
  5883. FN(get_func_ip, 173, ##ctx) \
  5884. FN(get_attach_cookie, 174, ##ctx) \
  5885. FN(task_pt_regs, 175, ##ctx) \
  5886. FN(get_branch_snapshot, 176, ##ctx) \
  5887. FN(trace_vprintk, 177, ##ctx) \
  5888. FN(skc_to_unix_sock, 178, ##ctx) \
  5889. FN(kallsyms_lookup_name, 179, ##ctx) \
  5890. FN(find_vma, 180, ##ctx) \
  5891. FN(loop, 181, ##ctx) \
  5892. FN(strncmp, 182, ##ctx) \
  5893. FN(get_func_arg, 183, ##ctx) \
  5894. FN(get_func_ret, 184, ##ctx) \
  5895. FN(get_func_arg_cnt, 185, ##ctx) \
  5896. FN(get_retval, 186, ##ctx) \
  5897. FN(set_retval, 187, ##ctx) \
  5898. FN(xdp_get_buff_len, 188, ##ctx) \
  5899. FN(xdp_load_bytes, 189, ##ctx) \
  5900. FN(xdp_store_bytes, 190, ##ctx) \
  5901. FN(copy_from_user_task, 191, ##ctx) \
  5902. FN(skb_set_tstamp, 192, ##ctx) \
  5903. FN(ima_file_hash, 193, ##ctx) \
  5904. FN(kptr_xchg, 194, ##ctx) \
  5905. FN(map_lookup_percpu_elem, 195, ##ctx) \
  5906. FN(skc_to_mptcp_sock, 196, ##ctx) \
  5907. FN(dynptr_from_mem, 197, ##ctx) \
  5908. FN(ringbuf_reserve_dynptr, 198, ##ctx) \
  5909. FN(ringbuf_submit_dynptr, 199, ##ctx) \
  5910. FN(ringbuf_discard_dynptr, 200, ##ctx) \
  5911. FN(dynptr_read, 201, ##ctx) \
  5912. FN(dynptr_write, 202, ##ctx) \
  5913. FN(dynptr_data, 203, ##ctx) \
  5914. FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \
  5915. FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \
  5916. FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \
  5917. FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \
  5918. FN(ktime_get_tai_ns, 208, ##ctx) \
  5919. FN(user_ringbuf_drain, 209, ##ctx) \
  5920. FN(cgrp_storage_get, 210, ##ctx) \
  5921. FN(cgrp_storage_delete, 211, ##ctx) \
  5922. /* */
  5923. /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
  5924. * know or care about integer value that is now passed as second argument
  5925. */
  5926. #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
  5927. #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
  5928. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  5929. * function eBPF program intends to call
  5930. */
  5931. #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
  5932. enum bpf_func_id {
  5933. ___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  5934. __BPF_FUNC_MAX_ID,
  5935. };
  5936. #undef __BPF_ENUM_FN
  5937. /* All flags used by eBPF helper functions, placed here. */
  5938. /* BPF_FUNC_skb_store_bytes flags. */
  5939. enum {
  5940. BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
  5941. BPF_F_INVALIDATE_HASH = (1ULL << 1),
  5942. };
  5943. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  5944. * First 4 bits are for passing the header field size.
  5945. */
  5946. enum {
  5947. BPF_F_HDR_FIELD_MASK = 0xfULL,
  5948. };
  5949. /* BPF_FUNC_l4_csum_replace flags. */
  5950. enum {
  5951. BPF_F_PSEUDO_HDR = (1ULL << 4),
  5952. BPF_F_MARK_MANGLED_0 = (1ULL << 5),
  5953. BPF_F_MARK_ENFORCE = (1ULL << 6),
  5954. };
  5955. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  5956. enum {
  5957. BPF_F_TUNINFO_IPV6 = (1ULL << 0),
  5958. };
  5959. /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
  5960. enum {
  5961. BPF_F_SKIP_FIELD_MASK = 0xffULL,
  5962. BPF_F_USER_STACK = (1ULL << 8),
  5963. /* flags used by BPF_FUNC_get_stackid only. */
  5964. BPF_F_FAST_STACK_CMP = (1ULL << 9),
  5965. BPF_F_REUSE_STACKID = (1ULL << 10),
  5966. /* flags used by BPF_FUNC_get_stack only. */
  5967. BPF_F_USER_BUILD_ID = (1ULL << 11),
  5968. };
  5969. /* BPF_FUNC_skb_set_tunnel_key flags. */
  5970. enum {
  5971. BPF_F_ZERO_CSUM_TX = (1ULL << 1),
  5972. BPF_F_DONT_FRAGMENT = (1ULL << 2),
  5973. BPF_F_SEQ_NUMBER = (1ULL << 3),
  5974. BPF_F_NO_TUNNEL_KEY = (1ULL << 4),
  5975. };
  5976. /* BPF_FUNC_skb_get_tunnel_key flags. */
  5977. enum {
  5978. BPF_F_TUNINFO_FLAGS = (1ULL << 4),
  5979. };
  5980. /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
  5981. * BPF_FUNC_perf_event_read_value flags.
  5982. */
  5983. enum {
  5984. BPF_F_INDEX_MASK = 0xffffffffULL,
  5985. BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
  5986. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  5987. BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
  5988. };
  5989. /* Current network namespace */
  5990. enum {
  5991. BPF_F_CURRENT_NETNS = (-1L),
  5992. };
  5993. /* BPF_FUNC_csum_level level values. */
  5994. enum {
  5995. BPF_CSUM_LEVEL_QUERY,
  5996. BPF_CSUM_LEVEL_INC,
  5997. BPF_CSUM_LEVEL_DEC,
  5998. BPF_CSUM_LEVEL_RESET,
  5999. };
  6000. /* BPF_FUNC_skb_adjust_room flags. */
  6001. enum {
  6002. BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
  6003. BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
  6004. BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
  6005. BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
  6006. BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
  6007. BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
  6008. BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
  6009. BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7),
  6010. BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8),
  6011. };
  6012. enum {
  6013. BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
  6014. BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
  6015. };
  6016. #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
  6017. BPF_ADJ_ROOM_ENCAP_L2_MASK) \
  6018. << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
  6019. /* BPF_FUNC_sysctl_get_name flags. */
  6020. enum {
  6021. BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
  6022. };
  6023. /* BPF_FUNC_<kernel_obj>_storage_get flags */
  6024. enum {
  6025. BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
  6026. /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
  6027. * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
  6028. */
  6029. BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
  6030. };
  6031. /* BPF_FUNC_read_branch_records flags. */
  6032. enum {
  6033. BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
  6034. };
  6035. /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
  6036. * BPF_FUNC_bpf_ringbuf_output flags.
  6037. */
  6038. enum {
  6039. BPF_RB_NO_WAKEUP = (1ULL << 0),
  6040. BPF_RB_FORCE_WAKEUP = (1ULL << 1),
  6041. };
  6042. /* BPF_FUNC_bpf_ringbuf_query flags */
  6043. enum {
  6044. BPF_RB_AVAIL_DATA = 0,
  6045. BPF_RB_RING_SIZE = 1,
  6046. BPF_RB_CONS_POS = 2,
  6047. BPF_RB_PROD_POS = 3,
  6048. };
  6049. /* BPF ring buffer constants */
  6050. enum {
  6051. BPF_RINGBUF_BUSY_BIT = (1U << 31),
  6052. BPF_RINGBUF_DISCARD_BIT = (1U << 30),
  6053. BPF_RINGBUF_HDR_SZ = 8,
  6054. };
  6055. /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
  6056. enum {
  6057. BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
  6058. BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
  6059. };
  6060. /* Mode for BPF_FUNC_skb_adjust_room helper. */
  6061. enum bpf_adj_room_mode {
  6062. BPF_ADJ_ROOM_NET,
  6063. BPF_ADJ_ROOM_MAC,
  6064. };
  6065. /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
  6066. enum bpf_hdr_start_off {
  6067. BPF_HDR_START_MAC,
  6068. BPF_HDR_START_NET,
  6069. };
  6070. /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
  6071. enum bpf_lwt_encap_mode {
  6072. BPF_LWT_ENCAP_SEG6,
  6073. BPF_LWT_ENCAP_SEG6_INLINE,
  6074. BPF_LWT_ENCAP_IP,
  6075. };
  6076. /* Flags for bpf_bprm_opts_set helper */
  6077. enum {
  6078. BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
  6079. };
  6080. /* Flags for bpf_redirect and bpf_redirect_map helpers */
  6081. enum {
  6082. BPF_F_INGRESS = (1ULL << 0), /* used for skb path */
  6083. BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */
  6084. BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */
  6085. #define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
  6086. };
  6087. #define __bpf_md_ptr(type, name) \
  6088. union { \
  6089. type name; \
  6090. __u64 :64; \
  6091. } __attribute__((aligned(8)))
  6092. /* The enum used in skb->tstamp_type. It specifies the clock type
  6093. * of the time stored in the skb->tstamp.
  6094. */
  6095. enum {
  6096. BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */
  6097. BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */
  6098. BPF_SKB_CLOCK_REALTIME = 0,
  6099. BPF_SKB_CLOCK_MONOTONIC = 1,
  6100. BPF_SKB_CLOCK_TAI = 2,
  6101. /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,
  6102. * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.
  6103. */
  6104. };
  6105. /* user accessible mirror of in-kernel sk_buff.
  6106. * new fields can only be added to the end of this structure
  6107. */
  6108. struct __sk_buff {
  6109. __u32 len;
  6110. __u32 pkt_type;
  6111. __u32 mark;
  6112. __u32 queue_mapping;
  6113. __u32 protocol;
  6114. __u32 vlan_present;
  6115. __u32 vlan_tci;
  6116. __u32 vlan_proto;
  6117. __u32 priority;
  6118. __u32 ingress_ifindex;
  6119. __u32 ifindex;
  6120. __u32 tc_index;
  6121. __u32 cb[5];
  6122. __u32 hash;
  6123. __u32 tc_classid;
  6124. __u32 data;
  6125. __u32 data_end;
  6126. __u32 napi_id;
  6127. /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
  6128. __u32 family;
  6129. __u32 remote_ip4; /* Stored in network byte order */
  6130. __u32 local_ip4; /* Stored in network byte order */
  6131. __u32 remote_ip6[4]; /* Stored in network byte order */
  6132. __u32 local_ip6[4]; /* Stored in network byte order */
  6133. __u32 remote_port; /* Stored in network byte order */
  6134. __u32 local_port; /* stored in host byte order */
  6135. /* ... here. */
  6136. __u32 data_meta;
  6137. __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
  6138. __u64 tstamp;
  6139. __u32 wire_len;
  6140. __u32 gso_segs;
  6141. __bpf_md_ptr(struct bpf_sock *, sk);
  6142. __u32 gso_size;
  6143. __u8 tstamp_type;
  6144. __u32 :24; /* Padding, future use. */
  6145. __u64 hwtstamp;
  6146. };
  6147. struct bpf_tunnel_key {
  6148. __u32 tunnel_id;
  6149. union {
  6150. __u32 remote_ipv4;
  6151. __u32 remote_ipv6[4];
  6152. };
  6153. __u8 tunnel_tos;
  6154. __u8 tunnel_ttl;
  6155. union {
  6156. __u16 tunnel_ext; /* compat */
  6157. __be16 tunnel_flags;
  6158. };
  6159. __u32 tunnel_label;
  6160. union {
  6161. __u32 local_ipv4;
  6162. __u32 local_ipv6[4];
  6163. };
  6164. };
  6165. /* user accessible mirror of in-kernel xfrm_state.
  6166. * new fields can only be added to the end of this structure
  6167. */
  6168. struct bpf_xfrm_state {
  6169. __u32 reqid;
  6170. __u32 spi; /* Stored in network byte order */
  6171. __u16 family;
  6172. __u16 ext; /* Padding, future use. */
  6173. union {
  6174. __u32 remote_ipv4; /* Stored in network byte order */
  6175. __u32 remote_ipv6[4]; /* Stored in network byte order */
  6176. };
  6177. };
  6178. /* Generic BPF return codes which all BPF program types may support.
  6179. * The values are binary compatible with their TC_ACT_* counter-part to
  6180. * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  6181. * programs.
  6182. *
  6183. * XDP is handled seprately, see XDP_*.
  6184. */
  6185. enum bpf_ret_code {
  6186. BPF_OK = 0,
  6187. /* 1 reserved */
  6188. BPF_DROP = 2,
  6189. /* 3-6 reserved */
  6190. BPF_REDIRECT = 7,
  6191. /* >127 are reserved for prog type specific return codes.
  6192. *
  6193. * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
  6194. * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
  6195. * changed and should be routed based on its new L3 header.
  6196. * (This is an L3 redirect, as opposed to L2 redirect
  6197. * represented by BPF_REDIRECT above).
  6198. */
  6199. BPF_LWT_REROUTE = 128,
  6200. /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
  6201. * to indicate that no custom dissection was performed, and
  6202. * fallback to standard dissector is requested.
  6203. */
  6204. BPF_FLOW_DISSECTOR_CONTINUE = 129,
  6205. };
  6206. struct bpf_sock {
  6207. __u32 bound_dev_if;
  6208. __u32 family;
  6209. __u32 type;
  6210. __u32 protocol;
  6211. __u32 mark;
  6212. __u32 priority;
  6213. /* IP address also allows 1 and 2 bytes access */
  6214. __u32 src_ip4;
  6215. __u32 src_ip6[4];
  6216. __u32 src_port; /* host byte order */
  6217. __be16 dst_port; /* network byte order */
  6218. __u16 :16; /* zero padding */
  6219. __u32 dst_ip4;
  6220. __u32 dst_ip6[4];
  6221. __u32 state;
  6222. __s32 rx_queue_mapping;
  6223. };
  6224. struct bpf_tcp_sock {
  6225. __u32 snd_cwnd; /* Sending congestion window */
  6226. __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
  6227. __u32 rtt_min;
  6228. __u32 snd_ssthresh; /* Slow start size threshold */
  6229. __u32 rcv_nxt; /* What we want to receive next */
  6230. __u32 snd_nxt; /* Next sequence we send */
  6231. __u32 snd_una; /* First byte we want an ack for */
  6232. __u32 mss_cache; /* Cached effective mss, not including SACKS */
  6233. __u32 ecn_flags; /* ECN status bits. */
  6234. __u32 rate_delivered; /* saved rate sample: packets delivered */
  6235. __u32 rate_interval_us; /* saved rate sample: time elapsed */
  6236. __u32 packets_out; /* Packets which are "in flight" */
  6237. __u32 retrans_out; /* Retransmitted packets out */
  6238. __u32 total_retrans; /* Total retransmits for entire connection */
  6239. __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
  6240. * total number of segments in.
  6241. */
  6242. __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
  6243. * total number of data segments in.
  6244. */
  6245. __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
  6246. * The total number of segments sent.
  6247. */
  6248. __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
  6249. * total number of data segments sent.
  6250. */
  6251. __u32 lost_out; /* Lost packets */
  6252. __u32 sacked_out; /* SACK'd packets */
  6253. __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
  6254. * sum(delta(rcv_nxt)), or how many bytes
  6255. * were acked.
  6256. */
  6257. __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
  6258. * sum(delta(snd_una)), or how many bytes
  6259. * were acked.
  6260. */
  6261. __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
  6262. * total number of DSACK blocks received
  6263. */
  6264. __u32 delivered; /* Total data packets delivered incl. rexmits */
  6265. __u32 delivered_ce; /* Like the above but only ECE marked packets */
  6266. __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
  6267. };
  6268. struct bpf_sock_tuple {
  6269. union {
  6270. struct {
  6271. __be32 saddr;
  6272. __be32 daddr;
  6273. __be16 sport;
  6274. __be16 dport;
  6275. } ipv4;
  6276. struct {
  6277. __be32 saddr[4];
  6278. __be32 daddr[4];
  6279. __be16 sport;
  6280. __be16 dport;
  6281. } ipv6;
  6282. };
  6283. };
  6284. /* (Simplified) user return codes for tcx prog type.
  6285. * A valid tcx program must return one of these defined values. All other
  6286. * return codes are reserved for future use. Must remain compatible with
  6287. * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
  6288. * return codes are mapped to TCX_NEXT.
  6289. */
  6290. enum tcx_action_base {
  6291. TCX_NEXT = -1,
  6292. TCX_PASS = 0,
  6293. TCX_DROP = 2,
  6294. TCX_REDIRECT = 7,
  6295. };
  6296. struct bpf_xdp_sock {
  6297. __u32 queue_id;
  6298. };
  6299. #define XDP_PACKET_HEADROOM 256
  6300. /* User return codes for XDP prog type.
  6301. * A valid XDP program must return one of these defined values. All other
  6302. * return codes are reserved for future use. Unknown return codes will
  6303. * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
  6304. */
  6305. enum xdp_action {
  6306. XDP_ABORTED = 0,
  6307. XDP_DROP,
  6308. XDP_PASS,
  6309. XDP_TX,
  6310. XDP_REDIRECT,
  6311. };
  6312. /* user accessible metadata for XDP packet hook
  6313. * new fields must be added to the end of this structure
  6314. */
  6315. struct xdp_md {
  6316. __u32 data;
  6317. __u32 data_end;
  6318. __u32 data_meta;
  6319. /* Below access go through struct xdp_rxq_info */
  6320. __u32 ingress_ifindex; /* rxq->dev->ifindex */
  6321. __u32 rx_queue_index; /* rxq->queue_index */
  6322. __u32 egress_ifindex; /* txq->dev->ifindex */
  6323. };
  6324. /* DEVMAP map-value layout
  6325. *
  6326. * The struct data-layout of map-value is a configuration interface.
  6327. * New members can only be added to the end of this structure.
  6328. */
  6329. struct bpf_devmap_val {
  6330. __u32 ifindex; /* device index */
  6331. union {
  6332. int fd; /* prog fd on map write */
  6333. __u32 id; /* prog id on map read */
  6334. } bpf_prog;
  6335. };
  6336. /* CPUMAP map-value layout
  6337. *
  6338. * The struct data-layout of map-value is a configuration interface.
  6339. * New members can only be added to the end of this structure.
  6340. */
  6341. struct bpf_cpumap_val {
  6342. __u32 qsize; /* queue size to remote target CPU */
  6343. union {
  6344. int fd; /* prog fd on map write */
  6345. __u32 id; /* prog id on map read */
  6346. } bpf_prog;
  6347. };
  6348. enum sk_action {
  6349. SK_DROP = 0,
  6350. SK_PASS,
  6351. };
  6352. /* user accessible metadata for SK_MSG packet hook, new fields must
  6353. * be added to the end of this structure
  6354. */
  6355. struct sk_msg_md {
  6356. __bpf_md_ptr(void *, data);
  6357. __bpf_md_ptr(void *, data_end);
  6358. __u32 family;
  6359. __u32 remote_ip4; /* Stored in network byte order */
  6360. __u32 local_ip4; /* Stored in network byte order */
  6361. __u32 remote_ip6[4]; /* Stored in network byte order */
  6362. __u32 local_ip6[4]; /* Stored in network byte order */
  6363. __u32 remote_port; /* Stored in network byte order */
  6364. __u32 local_port; /* stored in host byte order */
  6365. __u32 size; /* Total size of sk_msg */
  6366. __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
  6367. };
  6368. struct sk_reuseport_md {
  6369. /*
  6370. * Start of directly accessible data. It begins from
  6371. * the tcp/udp header.
  6372. */
  6373. __bpf_md_ptr(void *, data);
  6374. /* End of directly accessible data */
  6375. __bpf_md_ptr(void *, data_end);
  6376. /*
  6377. * Total length of packet (starting from the tcp/udp header).
  6378. * Note that the directly accessible bytes (data_end - data)
  6379. * could be less than this "len". Those bytes could be
  6380. * indirectly read by a helper "bpf_skb_load_bytes()".
  6381. */
  6382. __u32 len;
  6383. /*
  6384. * Eth protocol in the mac header (network byte order). e.g.
  6385. * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
  6386. */
  6387. __u32 eth_protocol;
  6388. __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
  6389. __u32 bind_inany; /* Is sock bound to an INANY address? */
  6390. __u32 hash; /* A hash of the packet 4 tuples */
  6391. /* When reuse->migrating_sk is NULL, it is selecting a sk for the
  6392. * new incoming connection request (e.g. selecting a listen sk for
  6393. * the received SYN in the TCP case). reuse->sk is one of the sk
  6394. * in the reuseport group. The bpf prog can use reuse->sk to learn
  6395. * the local listening ip/port without looking into the skb.
  6396. *
  6397. * When reuse->migrating_sk is not NULL, reuse->sk is closed and
  6398. * reuse->migrating_sk is the socket that needs to be migrated
  6399. * to another listening socket. migrating_sk could be a fullsock
  6400. * sk that is fully established or a reqsk that is in-the-middle
  6401. * of 3-way handshake.
  6402. */
  6403. __bpf_md_ptr(struct bpf_sock *, sk);
  6404. __bpf_md_ptr(struct bpf_sock *, migrating_sk);
  6405. };
  6406. #define BPF_TAG_SIZE 8
  6407. struct bpf_prog_info {
  6408. __u32 type;
  6409. __u32 id;
  6410. __u8 tag[BPF_TAG_SIZE];
  6411. __u32 jited_prog_len;
  6412. __u32 xlated_prog_len;
  6413. __aligned_u64 jited_prog_insns;
  6414. __aligned_u64 xlated_prog_insns;
  6415. __u64 load_time; /* ns since boottime */
  6416. __u32 created_by_uid;
  6417. __u32 nr_map_ids;
  6418. __aligned_u64 map_ids;
  6419. char name[BPF_OBJ_NAME_LEN];
  6420. __u32 ifindex;
  6421. __u32 gpl_compatible:1;
  6422. __u32 :31; /* alignment pad */
  6423. __u64 netns_dev;
  6424. __u64 netns_ino;
  6425. __u32 nr_jited_ksyms;
  6426. __u32 nr_jited_func_lens;
  6427. __aligned_u64 jited_ksyms;
  6428. __aligned_u64 jited_func_lens;
  6429. __u32 btf_id;
  6430. __u32 func_info_rec_size;
  6431. __aligned_u64 func_info;
  6432. __u32 nr_func_info;
  6433. __u32 nr_line_info;
  6434. __aligned_u64 line_info;
  6435. __aligned_u64 jited_line_info;
  6436. __u32 nr_jited_line_info;
  6437. __u32 line_info_rec_size;
  6438. __u32 jited_line_info_rec_size;
  6439. __u32 nr_prog_tags;
  6440. __aligned_u64 prog_tags;
  6441. __u64 run_time_ns;
  6442. __u64 run_cnt;
  6443. __u64 recursion_misses;
  6444. __u32 verified_insns;
  6445. __u32 attach_btf_obj_id;
  6446. __u32 attach_btf_id;
  6447. } __attribute__((aligned(8)));
  6448. struct bpf_map_info {
  6449. __u32 type;
  6450. __u32 id;
  6451. __u32 key_size;
  6452. __u32 value_size;
  6453. __u32 max_entries;
  6454. __u32 map_flags;
  6455. char name[BPF_OBJ_NAME_LEN];
  6456. __u32 ifindex;
  6457. __u32 btf_vmlinux_value_type_id;
  6458. __u64 netns_dev;
  6459. __u64 netns_ino;
  6460. __u32 btf_id;
  6461. __u32 btf_key_type_id;
  6462. __u32 btf_value_type_id;
  6463. __u32 btf_vmlinux_id;
  6464. __u64 map_extra;
  6465. } __attribute__((aligned(8)));
  6466. struct bpf_btf_info {
  6467. __aligned_u64 btf;
  6468. __u32 btf_size;
  6469. __u32 id;
  6470. __aligned_u64 name;
  6471. __u32 name_len;
  6472. __u32 kernel_btf;
  6473. } __attribute__((aligned(8)));
  6474. struct bpf_link_info {
  6475. __u32 type;
  6476. __u32 id;
  6477. __u32 prog_id;
  6478. union {
  6479. struct {
  6480. __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
  6481. __u32 tp_name_len; /* in/out: tp_name buffer len */
  6482. } raw_tracepoint;
  6483. struct {
  6484. __u32 attach_type;
  6485. __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
  6486. __u32 target_btf_id; /* BTF type id inside the object */
  6487. } tracing;
  6488. struct {
  6489. __u64 cgroup_id;
  6490. __u32 attach_type;
  6491. } cgroup;
  6492. struct {
  6493. __aligned_u64 target_name; /* in/out: target_name buffer ptr */
  6494. __u32 target_name_len; /* in/out: target_name buffer len */
  6495. /* If the iter specific field is 32 bits, it can be put
  6496. * in the first or second union. Otherwise it should be
  6497. * put in the second union.
  6498. */
  6499. union {
  6500. struct {
  6501. __u32 map_id;
  6502. } map;
  6503. };
  6504. union {
  6505. struct {
  6506. __u64 cgroup_id;
  6507. __u32 order;
  6508. } cgroup;
  6509. struct {
  6510. __u32 tid;
  6511. __u32 pid;
  6512. } task;
  6513. };
  6514. } iter;
  6515. struct {
  6516. __u32 netns_ino;
  6517. __u32 attach_type;
  6518. } netns;
  6519. struct {
  6520. __u32 ifindex;
  6521. } xdp;
  6522. struct {
  6523. __u32 map_id;
  6524. } struct_ops;
  6525. struct {
  6526. __u32 pf;
  6527. __u32 hooknum;
  6528. __s32 priority;
  6529. __u32 flags;
  6530. } netfilter;
  6531. struct {
  6532. __aligned_u64 addrs;
  6533. __u32 count; /* in/out: kprobe_multi function count */
  6534. __u32 flags;
  6535. __u64 missed;
  6536. __aligned_u64 cookies;
  6537. } kprobe_multi;
  6538. struct {
  6539. __aligned_u64 path;
  6540. __aligned_u64 offsets;
  6541. __aligned_u64 ref_ctr_offsets;
  6542. __aligned_u64 cookies;
  6543. __u32 path_size; /* in/out: real path size on success, including zero byte */
  6544. __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
  6545. __u32 flags;
  6546. __u32 pid;
  6547. } uprobe_multi;
  6548. struct {
  6549. __u32 type; /* enum bpf_perf_event_type */
  6550. __u32 :32;
  6551. union {
  6552. struct {
  6553. __aligned_u64 file_name; /* in/out */
  6554. __u32 name_len;
  6555. __u32 offset; /* offset from file_name */
  6556. __u64 cookie;
  6557. } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
  6558. struct {
  6559. __aligned_u64 func_name; /* in/out */
  6560. __u32 name_len;
  6561. __u32 offset; /* offset from func_name */
  6562. __u64 addr;
  6563. __u64 missed;
  6564. __u64 cookie;
  6565. } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
  6566. struct {
  6567. __aligned_u64 tp_name; /* in/out */
  6568. __u32 name_len;
  6569. __u32 :32;
  6570. __u64 cookie;
  6571. } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
  6572. struct {
  6573. __u64 config;
  6574. __u32 type;
  6575. __u32 :32;
  6576. __u64 cookie;
  6577. } event; /* BPF_PERF_EVENT_EVENT */
  6578. };
  6579. } perf_event;
  6580. struct {
  6581. __u32 ifindex;
  6582. __u32 attach_type;
  6583. } tcx;
  6584. struct {
  6585. __u32 ifindex;
  6586. __u32 attach_type;
  6587. } netkit;
  6588. struct {
  6589. __u32 map_id;
  6590. __u32 attach_type;
  6591. } sockmap;
  6592. };
  6593. } __attribute__((aligned(8)));
  6594. /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
  6595. * by user and intended to be used by socket (e.g. to bind to, depends on
  6596. * attach type).
  6597. */
  6598. struct bpf_sock_addr {
  6599. __u32 user_family; /* Allows 4-byte read, but no write. */
  6600. __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
  6601. * Stored in network byte order.
  6602. */
  6603. __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
  6604. * Stored in network byte order.
  6605. */
  6606. __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
  6607. * Stored in network byte order
  6608. */
  6609. __u32 family; /* Allows 4-byte read, but no write */
  6610. __u32 type; /* Allows 4-byte read, but no write */
  6611. __u32 protocol; /* Allows 4-byte read, but no write */
  6612. __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
  6613. * Stored in network byte order.
  6614. */
  6615. __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
  6616. * Stored in network byte order.
  6617. */
  6618. __bpf_md_ptr(struct bpf_sock *, sk);
  6619. };
  6620. /* User bpf_sock_ops struct to access socket values and specify request ops
  6621. * and their replies.
  6622. * Some of this fields are in network (bigendian) byte order and may need
  6623. * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
  6624. * New fields can only be added at the end of this structure
  6625. */
  6626. struct bpf_sock_ops {
  6627. __u32 op;
  6628. union {
  6629. __u32 args[4]; /* Optionally passed to bpf program */
  6630. __u32 reply; /* Returned by bpf program */
  6631. __u32 replylong[4]; /* Optionally returned by bpf prog */
  6632. };
  6633. __u32 family;
  6634. __u32 remote_ip4; /* Stored in network byte order */
  6635. __u32 local_ip4; /* Stored in network byte order */
  6636. __u32 remote_ip6[4]; /* Stored in network byte order */
  6637. __u32 local_ip6[4]; /* Stored in network byte order */
  6638. __u32 remote_port; /* Stored in network byte order */
  6639. __u32 local_port; /* stored in host byte order */
  6640. __u32 is_fullsock; /* Some TCP fields are only valid if
  6641. * there is a full socket. If not, the
  6642. * fields read as zero.
  6643. */
  6644. __u32 snd_cwnd;
  6645. __u32 srtt_us; /* Averaged RTT << 3 in usecs */
  6646. __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
  6647. __u32 state;
  6648. __u32 rtt_min;
  6649. __u32 snd_ssthresh;
  6650. __u32 rcv_nxt;
  6651. __u32 snd_nxt;
  6652. __u32 snd_una;
  6653. __u32 mss_cache;
  6654. __u32 ecn_flags;
  6655. __u32 rate_delivered;
  6656. __u32 rate_interval_us;
  6657. __u32 packets_out;
  6658. __u32 retrans_out;
  6659. __u32 total_retrans;
  6660. __u32 segs_in;
  6661. __u32 data_segs_in;
  6662. __u32 segs_out;
  6663. __u32 data_segs_out;
  6664. __u32 lost_out;
  6665. __u32 sacked_out;
  6666. __u32 sk_txhash;
  6667. __u64 bytes_received;
  6668. __u64 bytes_acked;
  6669. __bpf_md_ptr(struct bpf_sock *, sk);
  6670. /* [skb_data, skb_data_end) covers the whole TCP header.
  6671. *
  6672. * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
  6673. * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
  6674. * header has not been written.
  6675. * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
  6676. * been written so far.
  6677. * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
  6678. * the 3WHS.
  6679. * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
  6680. * the 3WHS.
  6681. *
  6682. * bpf_load_hdr_opt() can also be used to read a particular option.
  6683. */
  6684. __bpf_md_ptr(void *, skb_data);
  6685. __bpf_md_ptr(void *, skb_data_end);
  6686. __u32 skb_len; /* The total length of a packet.
  6687. * It includes the header, options,
  6688. * and payload.
  6689. */
  6690. __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
  6691. * an easy way to check for tcp_flags
  6692. * without parsing skb_data.
  6693. *
  6694. * In particular, the skb_tcp_flags
  6695. * will still be available in
  6696. * BPF_SOCK_OPS_HDR_OPT_LEN even though
  6697. * the outgoing header has not
  6698. * been written yet.
  6699. */
  6700. __u64 skb_hwtstamp;
  6701. };
  6702. /* Definitions for bpf_sock_ops_cb_flags */
  6703. enum {
  6704. BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
  6705. BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
  6706. BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
  6707. BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
  6708. /* Call bpf for all received TCP headers. The bpf prog will be
  6709. * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6710. *
  6711. * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6712. * for the header option related helpers that will be useful
  6713. * to the bpf programs.
  6714. *
  6715. * It could be used at the client/active side (i.e. connect() side)
  6716. * when the server told it that the server was in syncookie
  6717. * mode and required the active side to resend the bpf-written
  6718. * options. The active side can keep writing the bpf-options until
  6719. * it received a valid packet from the server side to confirm
  6720. * the earlier packet (and options) has been received. The later
  6721. * example patch is using it like this at the active side when the
  6722. * server is in syncookie mode.
  6723. *
  6724. * The bpf prog will usually turn this off in the common cases.
  6725. */
  6726. BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
  6727. /* Call bpf when kernel has received a header option that
  6728. * the kernel cannot handle. The bpf prog will be called under
  6729. * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
  6730. *
  6731. * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6732. * for the header option related helpers that will be useful
  6733. * to the bpf programs.
  6734. */
  6735. BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
  6736. /* Call bpf when the kernel is writing header options for the
  6737. * outgoing packet. The bpf prog will first be called
  6738. * to reserve space in a skb under
  6739. * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
  6740. * the bpf prog will be called to write the header option(s)
  6741. * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6742. *
  6743. * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
  6744. * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
  6745. * related helpers that will be useful to the bpf programs.
  6746. *
  6747. * The kernel gets its chance to reserve space and write
  6748. * options first before the BPF program does.
  6749. */
  6750. BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
  6751. /* Mask of all currently supported cb flags */
  6752. BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
  6753. };
  6754. /* List of known BPF sock_ops operators.
  6755. * New entries can only be added at the end
  6756. */
  6757. enum {
  6758. BPF_SOCK_OPS_VOID,
  6759. BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
  6760. * -1 if default value should be used
  6761. */
  6762. BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
  6763. * window (in packets) or -1 if default
  6764. * value should be used
  6765. */
  6766. BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
  6767. * active connection is initialized
  6768. */
  6769. BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
  6770. * active connection is
  6771. * established
  6772. */
  6773. BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
  6774. * passive connection is
  6775. * established
  6776. */
  6777. BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
  6778. * needs ECN
  6779. */
  6780. BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
  6781. * based on the path and may be
  6782. * dependent on the congestion control
  6783. * algorithm. In general it indicates
  6784. * a congestion threshold. RTTs above
  6785. * this indicate congestion
  6786. */
  6787. BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
  6788. * Arg1: value of icsk_retransmits
  6789. * Arg2: value of icsk_rto
  6790. * Arg3: whether RTO has expired
  6791. */
  6792. BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
  6793. * Arg1: sequence number of 1st byte
  6794. * Arg2: # segments
  6795. * Arg3: return value of
  6796. * tcp_transmit_skb (0 => success)
  6797. */
  6798. BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
  6799. * Arg1: old_state
  6800. * Arg2: new_state
  6801. */
  6802. BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
  6803. * socket transition to LISTEN state.
  6804. */
  6805. BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
  6806. * Arg1: measured RTT input (mrtt)
  6807. * Arg2: updated srtt
  6808. */
  6809. BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
  6810. * It will be called to handle
  6811. * the packets received at
  6812. * an already established
  6813. * connection.
  6814. *
  6815. * sock_ops->skb_data:
  6816. * Referring to the received skb.
  6817. * It covers the TCP header only.
  6818. *
  6819. * bpf_load_hdr_opt() can also
  6820. * be used to search for a
  6821. * particular option.
  6822. */
  6823. BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
  6824. * header option later in
  6825. * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6826. * Arg1: bool want_cookie. (in
  6827. * writing SYNACK only)
  6828. *
  6829. * sock_ops->skb_data:
  6830. * Not available because no header has
  6831. * been written yet.
  6832. *
  6833. * sock_ops->skb_tcp_flags:
  6834. * The tcp_flags of the
  6835. * outgoing skb. (e.g. SYN, ACK, FIN).
  6836. *
  6837. * bpf_reserve_hdr_opt() should
  6838. * be used to reserve space.
  6839. */
  6840. BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
  6841. * Arg1: bool want_cookie. (in
  6842. * writing SYNACK only)
  6843. *
  6844. * sock_ops->skb_data:
  6845. * Referring to the outgoing skb.
  6846. * It covers the TCP header
  6847. * that has already been written
  6848. * by the kernel and the
  6849. * earlier bpf-progs.
  6850. *
  6851. * sock_ops->skb_tcp_flags:
  6852. * The tcp_flags of the outgoing
  6853. * skb. (e.g. SYN, ACK, FIN).
  6854. *
  6855. * bpf_store_hdr_opt() should
  6856. * be used to write the
  6857. * option.
  6858. *
  6859. * bpf_load_hdr_opt() can also
  6860. * be used to search for a
  6861. * particular option that
  6862. * has already been written
  6863. * by the kernel or the
  6864. * earlier bpf-progs.
  6865. */
  6866. };
  6867. /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
  6868. * changes between the TCP and BPF versions. Ideally this should never happen.
  6869. * If it does, we need to add code to convert them before calling
  6870. * the BPF sock_ops function.
  6871. */
  6872. enum {
  6873. BPF_TCP_ESTABLISHED = 1,
  6874. BPF_TCP_SYN_SENT,
  6875. BPF_TCP_SYN_RECV,
  6876. BPF_TCP_FIN_WAIT1,
  6877. BPF_TCP_FIN_WAIT2,
  6878. BPF_TCP_TIME_WAIT,
  6879. BPF_TCP_CLOSE,
  6880. BPF_TCP_CLOSE_WAIT,
  6881. BPF_TCP_LAST_ACK,
  6882. BPF_TCP_LISTEN,
  6883. BPF_TCP_CLOSING, /* Now a valid state */
  6884. BPF_TCP_NEW_SYN_RECV,
  6885. BPF_TCP_BOUND_INACTIVE,
  6886. BPF_TCP_MAX_STATES /* Leave at the end! */
  6887. };
  6888. enum {
  6889. TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
  6890. TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
  6891. TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
  6892. TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
  6893. /* Copy the SYN pkt to optval
  6894. *
  6895. * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
  6896. * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
  6897. * to only getting from the saved_syn. It can either get the
  6898. * syn packet from:
  6899. *
  6900. * 1. the just-received SYN packet (only available when writing the
  6901. * SYNACK). It will be useful when it is not necessary to
  6902. * save the SYN packet for latter use. It is also the only way
  6903. * to get the SYN during syncookie mode because the syn
  6904. * packet cannot be saved during syncookie.
  6905. *
  6906. * OR
  6907. *
  6908. * 2. the earlier saved syn which was done by
  6909. * bpf_setsockopt(TCP_SAVE_SYN).
  6910. *
  6911. * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
  6912. * SYN packet is obtained.
  6913. *
  6914. * If the bpf-prog does not need the IP[46] header, the
  6915. * bpf-prog can avoid parsing the IP header by using
  6916. * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
  6917. * IP[46] and TCP header by using TCP_BPF_SYN_IP.
  6918. *
  6919. * >0: Total number of bytes copied
  6920. * -ENOSPC: Not enough space in optval. Only optlen number of
  6921. * bytes is copied.
  6922. * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
  6923. * is not saved by setsockopt(TCP_SAVE_SYN).
  6924. */
  6925. TCP_BPF_SYN = 1005, /* Copy the TCP header */
  6926. TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
  6927. TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
  6928. TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */
  6929. };
  6930. enum {
  6931. BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
  6932. };
  6933. /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
  6934. * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6935. */
  6936. enum {
  6937. BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
  6938. * total option spaces
  6939. * required for an established
  6940. * sk in order to calculate the
  6941. * MSS. No skb is actually
  6942. * sent.
  6943. */
  6944. BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
  6945. * when sending a SYN.
  6946. */
  6947. };
  6948. struct bpf_perf_event_value {
  6949. __u64 counter;
  6950. __u64 enabled;
  6951. __u64 running;
  6952. };
  6953. enum {
  6954. BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
  6955. BPF_DEVCG_ACC_READ = (1ULL << 1),
  6956. BPF_DEVCG_ACC_WRITE = (1ULL << 2),
  6957. };
  6958. enum {
  6959. BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
  6960. BPF_DEVCG_DEV_CHAR = (1ULL << 1),
  6961. };
  6962. struct bpf_cgroup_dev_ctx {
  6963. /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
  6964. __u32 access_type;
  6965. __u32 major;
  6966. __u32 minor;
  6967. };
  6968. struct bpf_raw_tracepoint_args {
  6969. __u64 args[0];
  6970. };
  6971. /* DIRECT: Skip the FIB rules and go to FIB table associated with device
  6972. * OUTPUT: Do lookup from egress perspective; default is ingress
  6973. */
  6974. enum {
  6975. BPF_FIB_LOOKUP_DIRECT = (1U << 0),
  6976. BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
  6977. BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
  6978. BPF_FIB_LOOKUP_TBID = (1U << 3),
  6979. BPF_FIB_LOOKUP_SRC = (1U << 4),
  6980. BPF_FIB_LOOKUP_MARK = (1U << 5),
  6981. };
  6982. enum {
  6983. BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
  6984. BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
  6985. BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
  6986. BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
  6987. BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
  6988. BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
  6989. BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
  6990. BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
  6991. BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
  6992. BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */
  6993. };
  6994. struct bpf_fib_lookup {
  6995. /* input: network family for lookup (AF_INET, AF_INET6)
  6996. * output: network family of egress nexthop
  6997. */
  6998. __u8 family;
  6999. /* set if lookup is to consider L4 data - e.g., FIB rules */
  7000. __u8 l4_protocol;
  7001. __be16 sport;
  7002. __be16 dport;
  7003. union { /* used for MTU check */
  7004. /* input to lookup */
  7005. __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
  7006. /* output: MTU value */
  7007. __u16 mtu_result;
  7008. } __attribute__((packed, aligned(2)));
  7009. /* input: L3 device index for lookup
  7010. * output: device index from FIB lookup
  7011. */
  7012. __u32 ifindex;
  7013. union {
  7014. /* inputs to lookup */
  7015. __u8 tos; /* AF_INET */
  7016. __be32 flowinfo; /* AF_INET6, flow_label + priority */
  7017. /* output: metric of fib result (IPv4/IPv6 only) */
  7018. __u32 rt_metric;
  7019. };
  7020. /* input: source address to consider for lookup
  7021. * output: source address result from lookup
  7022. */
  7023. union {
  7024. __be32 ipv4_src;
  7025. __u32 ipv6_src[4]; /* in6_addr; network order */
  7026. };
  7027. /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
  7028. * network header. output: bpf_fib_lookup sets to gateway address
  7029. * if FIB lookup returns gateway route
  7030. */
  7031. union {
  7032. __be32 ipv4_dst;
  7033. __u32 ipv6_dst[4]; /* in6_addr; network order */
  7034. };
  7035. union {
  7036. struct {
  7037. /* output */
  7038. __be16 h_vlan_proto;
  7039. __be16 h_vlan_TCI;
  7040. };
  7041. /* input: when accompanied with the
  7042. * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
  7043. * specific routing table to use for the fib lookup.
  7044. */
  7045. __u32 tbid;
  7046. };
  7047. union {
  7048. /* input */
  7049. struct {
  7050. __u32 mark; /* policy routing */
  7051. /* 2 4-byte holes for input */
  7052. };
  7053. /* output: source and dest mac */
  7054. struct {
  7055. __u8 smac[6]; /* ETH_ALEN */
  7056. __u8 dmac[6]; /* ETH_ALEN */
  7057. };
  7058. };
  7059. };
  7060. struct bpf_redir_neigh {
  7061. /* network family for lookup (AF_INET, AF_INET6) */
  7062. __u32 nh_family;
  7063. /* network address of nexthop; skips fib lookup to find gateway */
  7064. union {
  7065. __be32 ipv4_nh;
  7066. __u32 ipv6_nh[4]; /* in6_addr; network order */
  7067. };
  7068. };
  7069. /* bpf_check_mtu flags*/
  7070. enum bpf_check_mtu_flags {
  7071. BPF_MTU_CHK_SEGS = (1U << 0),
  7072. };
  7073. enum bpf_check_mtu_ret {
  7074. BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
  7075. BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
  7076. BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
  7077. };
  7078. enum bpf_task_fd_type {
  7079. BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
  7080. BPF_FD_TYPE_TRACEPOINT, /* tp name */
  7081. BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
  7082. BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
  7083. BPF_FD_TYPE_UPROBE, /* filename + offset */
  7084. BPF_FD_TYPE_URETPROBE, /* filename + offset */
  7085. };
  7086. enum {
  7087. BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
  7088. BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
  7089. BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
  7090. };
  7091. struct bpf_flow_keys {
  7092. __u16 nhoff;
  7093. __u16 thoff;
  7094. __u16 addr_proto; /* ETH_P_* of valid addrs */
  7095. __u8 is_frag;
  7096. __u8 is_first_frag;
  7097. __u8 is_encap;
  7098. __u8 ip_proto;
  7099. __be16 n_proto;
  7100. __be16 sport;
  7101. __be16 dport;
  7102. union {
  7103. struct {
  7104. __be32 ipv4_src;
  7105. __be32 ipv4_dst;
  7106. };
  7107. struct {
  7108. __u32 ipv6_src[4]; /* in6_addr; network order */
  7109. __u32 ipv6_dst[4]; /* in6_addr; network order */
  7110. };
  7111. };
  7112. __u32 flags;
  7113. __be32 flow_label;
  7114. };
  7115. struct bpf_func_info {
  7116. __u32 insn_off;
  7117. __u32 type_id;
  7118. };
  7119. #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
  7120. #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
  7121. struct bpf_line_info {
  7122. __u32 insn_off;
  7123. __u32 file_name_off;
  7124. __u32 line_off;
  7125. __u32 line_col;
  7126. };
  7127. struct bpf_spin_lock {
  7128. __u32 val;
  7129. };
  7130. struct bpf_timer {
  7131. __u64 __opaque[2];
  7132. } __attribute__((aligned(8)));
  7133. struct bpf_wq {
  7134. __u64 __opaque[2];
  7135. } __attribute__((aligned(8)));
  7136. struct bpf_dynptr {
  7137. __u64 __opaque[2];
  7138. } __attribute__((aligned(8)));
  7139. struct bpf_list_head {
  7140. __u64 __opaque[2];
  7141. } __attribute__((aligned(8)));
  7142. struct bpf_list_node {
  7143. __u64 __opaque[3];
  7144. } __attribute__((aligned(8)));
  7145. struct bpf_rb_root {
  7146. __u64 __opaque[2];
  7147. } __attribute__((aligned(8)));
  7148. struct bpf_rb_node {
  7149. __u64 __opaque[4];
  7150. } __attribute__((aligned(8)));
  7151. struct bpf_refcount {
  7152. __u32 __opaque[1];
  7153. } __attribute__((aligned(4)));
  7154. struct bpf_sysctl {
  7155. __u32 write; /* Sysctl is being read (= 0) or written (= 1).
  7156. * Allows 1,2,4-byte read, but no write.
  7157. */
  7158. __u32 file_pos; /* Sysctl file position to read from, write to.
  7159. * Allows 1,2,4-byte read an 4-byte write.
  7160. */
  7161. };
  7162. struct bpf_sockopt {
  7163. __bpf_md_ptr(struct bpf_sock *, sk);
  7164. __bpf_md_ptr(void *, optval);
  7165. __bpf_md_ptr(void *, optval_end);
  7166. __s32 level;
  7167. __s32 optname;
  7168. __s32 optlen;
  7169. __s32 retval;
  7170. };
  7171. struct bpf_pidns_info {
  7172. __u32 pid;
  7173. __u32 tgid;
  7174. };
  7175. /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
  7176. struct bpf_sk_lookup {
  7177. union {
  7178. __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
  7179. __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
  7180. };
  7181. __u32 family; /* Protocol family (AF_INET, AF_INET6) */
  7182. __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
  7183. __u32 remote_ip4; /* Network byte order */
  7184. __u32 remote_ip6[4]; /* Network byte order */
  7185. __be16 remote_port; /* Network byte order */
  7186. __u16 :16; /* Zero padding */
  7187. __u32 local_ip4; /* Network byte order */
  7188. __u32 local_ip6[4]; /* Network byte order */
  7189. __u32 local_port; /* Host byte order */
  7190. __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */
  7191. };
  7192. /*
  7193. * struct btf_ptr is used for typed pointer representation; the
  7194. * type id is used to render the pointer data as the appropriate type
  7195. * via the bpf_snprintf_btf() helper described above. A flags field -
  7196. * potentially to specify additional details about the BTF pointer
  7197. * (rather than its mode of display) - is included for future use.
  7198. * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
  7199. */
  7200. struct btf_ptr {
  7201. void *ptr;
  7202. __u32 type_id;
  7203. __u32 flags; /* BTF ptr flags; unused at present. */
  7204. };
  7205. /*
  7206. * Flags to control bpf_snprintf_btf() behaviour.
  7207. * - BTF_F_COMPACT: no formatting around type information
  7208. * - BTF_F_NONAME: no struct/union member names/types
  7209. * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
  7210. * equivalent to %px.
  7211. * - BTF_F_ZERO: show zero-valued struct/union members; they
  7212. * are not displayed by default
  7213. */
  7214. enum {
  7215. BTF_F_COMPACT = (1ULL << 0),
  7216. BTF_F_NONAME = (1ULL << 1),
  7217. BTF_F_PTR_RAW = (1ULL << 2),
  7218. BTF_F_ZERO = (1ULL << 3),
  7219. };
  7220. /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
  7221. * has to be adjusted by relocations. It is emitted by llvm and passed to
  7222. * libbpf and later to the kernel.
  7223. */
  7224. enum bpf_core_relo_kind {
  7225. BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */
  7226. BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */
  7227. BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */
  7228. BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */
  7229. BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */
  7230. BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */
  7231. BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */
  7232. BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */
  7233. BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */
  7234. BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */
  7235. BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */
  7236. BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */
  7237. BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */
  7238. };
  7239. /*
  7240. * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
  7241. * and from libbpf to the kernel.
  7242. *
  7243. * CO-RE relocation captures the following data:
  7244. * - insn_off - instruction offset (in bytes) within a BPF program that needs
  7245. * its insn->imm field to be relocated with actual field info;
  7246. * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
  7247. * type or field;
  7248. * - access_str_off - offset into corresponding .BTF string section. String
  7249. * interpretation depends on specific relocation kind:
  7250. * - for field-based relocations, string encodes an accessed field using
  7251. * a sequence of field and array indices, separated by colon (:). It's
  7252. * conceptually very close to LLVM's getelementptr ([0]) instruction's
  7253. * arguments for identifying offset to a field.
  7254. * - for type-based relocations, strings is expected to be just "0";
  7255. * - for enum value-based relocations, string contains an index of enum
  7256. * value within its enum type;
  7257. * - kind - one of enum bpf_core_relo_kind;
  7258. *
  7259. * Example:
  7260. * struct sample {
  7261. * int a;
  7262. * struct {
  7263. * int b[10];
  7264. * };
  7265. * };
  7266. *
  7267. * struct sample *s = ...;
  7268. * int *x = &s->a; // encoded as "0:0" (a is field #0)
  7269. * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
  7270. * // b is field #0 inside anon struct, accessing elem #5)
  7271. * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
  7272. *
  7273. * type_id for all relocs in this example will capture BTF type id of
  7274. * `struct sample`.
  7275. *
  7276. * Such relocation is emitted when using __builtin_preserve_access_index()
  7277. * Clang built-in, passing expression that captures field address, e.g.:
  7278. *
  7279. * bpf_probe_read(&dst, sizeof(dst),
  7280. * __builtin_preserve_access_index(&src->a.b.c));
  7281. *
  7282. * In this case Clang will emit field relocation recording necessary data to
  7283. * be able to find offset of embedded `a.b.c` field within `src` struct.
  7284. *
  7285. * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
  7286. */
  7287. struct bpf_core_relo {
  7288. __u32 insn_off;
  7289. __u32 type_id;
  7290. __u32 access_str_off;
  7291. enum bpf_core_relo_kind kind;
  7292. };
  7293. /*
  7294. * Flags to control bpf_timer_start() behaviour.
  7295. * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
  7296. * relative to current time.
  7297. * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
  7298. */
  7299. enum {
  7300. BPF_F_TIMER_ABS = (1ULL << 0),
  7301. BPF_F_TIMER_CPU_PIN = (1ULL << 1),
  7302. };
  7303. /* BPF numbers iterator state */
  7304. struct bpf_iter_num {
  7305. /* opaque iterator state; having __u64 here allows to preserve correct
  7306. * alignment requirements in vmlinux.h, generated from BTF
  7307. */
  7308. __u64 __opaque[1];
  7309. } __attribute__((aligned(8)));
  7310. /*
  7311. * Flags to control BPF kfunc behaviour.
  7312. * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective
  7313. * helper documentation for details.)
  7314. */
  7315. enum bpf_kfunc_flags {
  7316. BPF_F_PAD_ZEROS = (1ULL << 0),
  7317. };
  7318. #endif /* __LINUX_BPF_H__ */