bpf.h 24 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808
  1. /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
  2. *
  3. * This program is free software; you can redistribute it and/or
  4. * modify it under the terms of version 2 of the GNU General Public
  5. * License as published by the Free Software Foundation.
  6. */
  7. #ifndef _UAPI__LINUX_BPF_H__
  8. #define _UAPI__LINUX_BPF_H__
  9. #include <linux/types.h>
  10. #include <linux/bpf_common.h>
  11. /* Extended instruction set based on top of classic BPF */
  12. /* instruction classes */
  13. #define BPF_ALU64 0x07 /* alu mode in double word width */
  14. /* ld/ldx fields */
  15. #define BPF_DW 0x18 /* double word */
  16. #define BPF_XADD 0xc0 /* exclusive add */
  17. /* alu/jmp fields */
  18. #define BPF_MOV 0xb0 /* mov reg to reg */
  19. #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
  20. /* change endianness of a register */
  21. #define BPF_END 0xd0 /* flags for endianness conversion: */
  22. #define BPF_TO_LE 0x00 /* convert to little-endian */
  23. #define BPF_TO_BE 0x08 /* convert to big-endian */
  24. #define BPF_FROM_LE BPF_TO_LE
  25. #define BPF_FROM_BE BPF_TO_BE
  26. /* jmp encodings */
  27. #define BPF_JNE 0x50 /* jump != */
  28. #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
  29. #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
  30. #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
  31. #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
  32. #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
  33. #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
  34. #define BPF_CALL 0x80 /* function call */
  35. #define BPF_EXIT 0x90 /* function return */
  36. /* Register numbers */
  37. enum {
  38. BPF_REG_0 = 0,
  39. BPF_REG_1,
  40. BPF_REG_2,
  41. BPF_REG_3,
  42. BPF_REG_4,
  43. BPF_REG_5,
  44. BPF_REG_6,
  45. BPF_REG_7,
  46. BPF_REG_8,
  47. BPF_REG_9,
  48. BPF_REG_10,
  49. __MAX_BPF_REG,
  50. };
  51. /* BPF has 10 general purpose 64-bit registers and stack frame. */
  52. #define MAX_BPF_REG __MAX_BPF_REG
  53. struct bpf_insn {
  54. __u8 code; /* opcode */
  55. __u8 dst_reg:4; /* dest register */
  56. __u8 src_reg:4; /* source register */
  57. __s16 off; /* signed offset */
  58. __s32 imm; /* signed immediate constant */
  59. };
  60. /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
  61. struct bpf_lpm_trie_key {
  62. __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
  63. __u8 data[0]; /* Arbitrary size */
  64. };
  65. /* BPF syscall commands, see bpf(2) man-page for details. */
  66. enum bpf_cmd {
  67. BPF_MAP_CREATE,
  68. BPF_MAP_LOOKUP_ELEM,
  69. BPF_MAP_UPDATE_ELEM,
  70. BPF_MAP_DELETE_ELEM,
  71. BPF_MAP_GET_NEXT_KEY,
  72. BPF_PROG_LOAD,
  73. BPF_OBJ_PIN,
  74. BPF_OBJ_GET,
  75. BPF_PROG_ATTACH,
  76. BPF_PROG_DETACH,
  77. BPF_PROG_TEST_RUN,
  78. BPF_PROG_GET_NEXT_ID,
  79. BPF_MAP_GET_NEXT_ID,
  80. BPF_PROG_GET_FD_BY_ID,
  81. BPF_MAP_GET_FD_BY_ID,
  82. BPF_OBJ_GET_INFO_BY_FD,
  83. };
  84. enum bpf_map_type {
  85. BPF_MAP_TYPE_UNSPEC,
  86. BPF_MAP_TYPE_HASH,
  87. BPF_MAP_TYPE_ARRAY,
  88. BPF_MAP_TYPE_PROG_ARRAY,
  89. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  90. BPF_MAP_TYPE_PERCPU_HASH,
  91. BPF_MAP_TYPE_PERCPU_ARRAY,
  92. BPF_MAP_TYPE_STACK_TRACE,
  93. BPF_MAP_TYPE_CGROUP_ARRAY,
  94. BPF_MAP_TYPE_LRU_HASH,
  95. BPF_MAP_TYPE_LRU_PERCPU_HASH,
  96. BPF_MAP_TYPE_LPM_TRIE,
  97. BPF_MAP_TYPE_ARRAY_OF_MAPS,
  98. BPF_MAP_TYPE_HASH_OF_MAPS,
  99. BPF_MAP_TYPE_DEVMAP,
  100. };
  101. enum bpf_prog_type {
  102. BPF_PROG_TYPE_UNSPEC,
  103. BPF_PROG_TYPE_SOCKET_FILTER,
  104. BPF_PROG_TYPE_KPROBE,
  105. BPF_PROG_TYPE_SCHED_CLS,
  106. BPF_PROG_TYPE_SCHED_ACT,
  107. BPF_PROG_TYPE_TRACEPOINT,
  108. BPF_PROG_TYPE_XDP,
  109. BPF_PROG_TYPE_PERF_EVENT,
  110. BPF_PROG_TYPE_CGROUP_SKB,
  111. BPF_PROG_TYPE_CGROUP_SOCK,
  112. BPF_PROG_TYPE_LWT_IN,
  113. BPF_PROG_TYPE_LWT_OUT,
  114. BPF_PROG_TYPE_LWT_XMIT,
  115. BPF_PROG_TYPE_SOCK_OPS,
  116. };
  117. enum bpf_attach_type {
  118. BPF_CGROUP_INET_INGRESS,
  119. BPF_CGROUP_INET_EGRESS,
  120. BPF_CGROUP_INET_SOCK_CREATE,
  121. BPF_CGROUP_SOCK_OPS,
  122. __MAX_BPF_ATTACH_TYPE
  123. };
  124. #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  125. /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
  126. * to the given target_fd cgroup the descendent cgroup will be able to
  127. * override effective bpf program that was inherited from this cgroup
  128. */
  129. #define BPF_F_ALLOW_OVERRIDE (1U << 0)
  130. /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  131. * verifier will perform strict alignment checking as if the kernel
  132. * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  133. * and NET_IP_ALIGN defined to 2.
  134. */
  135. #define BPF_F_STRICT_ALIGNMENT (1U << 0)
  136. #define BPF_PSEUDO_MAP_FD 1
  137. /* flags for BPF_MAP_UPDATE_ELEM command */
  138. #define BPF_ANY 0 /* create new element or update existing */
  139. #define BPF_NOEXIST 1 /* create new element if it didn't exist */
  140. #define BPF_EXIST 2 /* update existing element */
  141. #define BPF_F_NO_PREALLOC (1U << 0)
  142. /* Instead of having one common LRU list in the
  143. * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  144. * which can scale and perform better.
  145. * Note, the LRU nodes (including free nodes) cannot be moved
  146. * across different LRU lists.
  147. */
  148. #define BPF_F_NO_COMMON_LRU (1U << 1)
  149. union bpf_attr {
  150. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  151. __u32 map_type; /* one of enum bpf_map_type */
  152. __u32 key_size; /* size of key in bytes */
  153. __u32 value_size; /* size of value in bytes */
  154. __u32 max_entries; /* max number of entries in a map */
  155. __u32 map_flags; /* prealloc or not */
  156. __u32 inner_map_fd; /* fd pointing to the inner map */
  157. };
  158. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  159. __u32 map_fd;
  160. __aligned_u64 key;
  161. union {
  162. __aligned_u64 value;
  163. __aligned_u64 next_key;
  164. };
  165. __u64 flags;
  166. };
  167. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  168. __u32 prog_type; /* one of enum bpf_prog_type */
  169. __u32 insn_cnt;
  170. __aligned_u64 insns;
  171. __aligned_u64 license;
  172. __u32 log_level; /* verbosity level of verifier */
  173. __u32 log_size; /* size of user buffer */
  174. __aligned_u64 log_buf; /* user supplied buffer */
  175. __u32 kern_version; /* checked when prog_type=kprobe */
  176. __u32 prog_flags;
  177. };
  178. struct { /* anonymous struct used by BPF_OBJ_* commands */
  179. __aligned_u64 pathname;
  180. __u32 bpf_fd;
  181. };
  182. struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  183. __u32 target_fd; /* container object to attach to */
  184. __u32 attach_bpf_fd; /* eBPF program to attach */
  185. __u32 attach_type;
  186. __u32 attach_flags;
  187. };
  188. struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  189. __u32 prog_fd;
  190. __u32 retval;
  191. __u32 data_size_in;
  192. __u32 data_size_out;
  193. __aligned_u64 data_in;
  194. __aligned_u64 data_out;
  195. __u32 repeat;
  196. __u32 duration;
  197. } test;
  198. struct { /* anonymous struct used by BPF_*_GET_*_ID */
  199. union {
  200. __u32 start_id;
  201. __u32 prog_id;
  202. __u32 map_id;
  203. };
  204. __u32 next_id;
  205. };
  206. struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  207. __u32 bpf_fd;
  208. __u32 info_len;
  209. __aligned_u64 info;
  210. } info;
  211. } __attribute__((aligned(8)));
  212. /* BPF helper function descriptions:
  213. *
  214. * void *bpf_map_lookup_elem(&map, &key)
  215. * Return: Map value or NULL
  216. *
  217. * int bpf_map_update_elem(&map, &key, &value, flags)
  218. * Return: 0 on success or negative error
  219. *
  220. * int bpf_map_delete_elem(&map, &key)
  221. * Return: 0 on success or negative error
  222. *
  223. * int bpf_probe_read(void *dst, int size, void *src)
  224. * Return: 0 on success or negative error
  225. *
  226. * u64 bpf_ktime_get_ns(void)
  227. * Return: current ktime
  228. *
  229. * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
  230. * Return: length of buffer written or negative error
  231. *
  232. * u32 bpf_prandom_u32(void)
  233. * Return: random value
  234. *
  235. * u32 bpf_raw_smp_processor_id(void)
  236. * Return: SMP processor ID
  237. *
  238. * int bpf_skb_store_bytes(skb, offset, from, len, flags)
  239. * store bytes into packet
  240. * @skb: pointer to skb
  241. * @offset: offset within packet from skb->mac_header
  242. * @from: pointer where to copy bytes from
  243. * @len: number of bytes to store into packet
  244. * @flags: bit 0 - if true, recompute skb->csum
  245. * other bits - reserved
  246. * Return: 0 on success or negative error
  247. *
  248. * int bpf_l3_csum_replace(skb, offset, from, to, flags)
  249. * recompute IP checksum
  250. * @skb: pointer to skb
  251. * @offset: offset within packet where IP checksum is located
  252. * @from: old value of header field
  253. * @to: new value of header field
  254. * @flags: bits 0-3 - size of header field
  255. * other bits - reserved
  256. * Return: 0 on success or negative error
  257. *
  258. * int bpf_l4_csum_replace(skb, offset, from, to, flags)
  259. * recompute TCP/UDP checksum
  260. * @skb: pointer to skb
  261. * @offset: offset within packet where TCP/UDP checksum is located
  262. * @from: old value of header field
  263. * @to: new value of header field
  264. * @flags: bits 0-3 - size of header field
  265. * bit 4 - is pseudo header
  266. * other bits - reserved
  267. * Return: 0 on success or negative error
  268. *
  269. * int bpf_tail_call(ctx, prog_array_map, index)
  270. * jump into another BPF program
  271. * @ctx: context pointer passed to next program
  272. * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
  273. * @index: index inside array that selects specific program to run
  274. * Return: 0 on success or negative error
  275. *
  276. * int bpf_clone_redirect(skb, ifindex, flags)
  277. * redirect to another netdev
  278. * @skb: pointer to skb
  279. * @ifindex: ifindex of the net device
  280. * @flags: bit 0 - if set, redirect to ingress instead of egress
  281. * other bits - reserved
  282. * Return: 0 on success or negative error
  283. *
  284. * u64 bpf_get_current_pid_tgid(void)
  285. * Return: current->tgid << 32 | current->pid
  286. *
  287. * u64 bpf_get_current_uid_gid(void)
  288. * Return: current_gid << 32 | current_uid
  289. *
  290. * int bpf_get_current_comm(char *buf, int size_of_buf)
  291. * stores current->comm into buf
  292. * Return: 0 on success or negative error
  293. *
  294. * u32 bpf_get_cgroup_classid(skb)
  295. * retrieve a proc's classid
  296. * @skb: pointer to skb
  297. * Return: classid if != 0
  298. *
  299. * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
  300. * Return: 0 on success or negative error
  301. *
  302. * int bpf_skb_vlan_pop(skb)
  303. * Return: 0 on success or negative error
  304. *
  305. * int bpf_skb_get_tunnel_key(skb, key, size, flags)
  306. * int bpf_skb_set_tunnel_key(skb, key, size, flags)
  307. * retrieve or populate tunnel metadata
  308. * @skb: pointer to skb
  309. * @key: pointer to 'struct bpf_tunnel_key'
  310. * @size: size of 'struct bpf_tunnel_key'
  311. * @flags: room for future extensions
  312. * Return: 0 on success or negative error
  313. *
  314. * u64 bpf_perf_event_read(map, flags)
  315. * read perf event counter value
  316. * @map: pointer to perf_event_array map
  317. * @flags: index of event in the map or bitmask flags
  318. * Return: value of perf event counter read or error code
  319. *
  320. * int bpf_redirect(ifindex, flags)
  321. * redirect to another netdev
  322. * @ifindex: ifindex of the net device
  323. * @flags: bit 0 - if set, redirect to ingress instead of egress
  324. * other bits - reserved
  325. * Return: TC_ACT_REDIRECT
  326. *
  327. * u32 bpf_get_route_realm(skb)
  328. * retrieve a dst's tclassid
  329. * @skb: pointer to skb
  330. * Return: realm if != 0
  331. *
  332. * int bpf_perf_event_output(ctx, map, flags, data, size)
  333. * output perf raw sample
  334. * @ctx: struct pt_regs*
  335. * @map: pointer to perf_event_array map
  336. * @flags: index of event in the map or bitmask flags
  337. * @data: data on stack to be output as raw data
  338. * @size: size of data
  339. * Return: 0 on success or negative error
  340. *
  341. * int bpf_get_stackid(ctx, map, flags)
  342. * walk user or kernel stack and return id
  343. * @ctx: struct pt_regs*
  344. * @map: pointer to stack_trace map
  345. * @flags: bits 0-7 - numer of stack frames to skip
  346. * bit 8 - collect user stack instead of kernel
  347. * bit 9 - compare stacks by hash only
  348. * bit 10 - if two different stacks hash into the same stackid
  349. * discard old
  350. * other bits - reserved
  351. * Return: >= 0 stackid on success or negative error
  352. *
  353. * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
  354. * calculate csum diff
  355. * @from: raw from buffer
  356. * @from_size: length of from buffer
  357. * @to: raw to buffer
  358. * @to_size: length of to buffer
  359. * @seed: optional seed
  360. * Return: csum result or negative error code
  361. *
  362. * int bpf_skb_get_tunnel_opt(skb, opt, size)
  363. * retrieve tunnel options metadata
  364. * @skb: pointer to skb
  365. * @opt: pointer to raw tunnel option data
  366. * @size: size of @opt
  367. * Return: option size
  368. *
  369. * int bpf_skb_set_tunnel_opt(skb, opt, size)
  370. * populate tunnel options metadata
  371. * @skb: pointer to skb
  372. * @opt: pointer to raw tunnel option data
  373. * @size: size of @opt
  374. * Return: 0 on success or negative error
  375. *
  376. * int bpf_skb_change_proto(skb, proto, flags)
  377. * Change protocol of the skb. Currently supported is v4 -> v6,
  378. * v6 -> v4 transitions. The helper will also resize the skb. eBPF
  379. * program is expected to fill the new headers via skb_store_bytes
  380. * and lX_csum_replace.
  381. * @skb: pointer to skb
  382. * @proto: new skb->protocol type
  383. * @flags: reserved
  384. * Return: 0 on success or negative error
  385. *
  386. * int bpf_skb_change_type(skb, type)
  387. * Change packet type of skb.
  388. * @skb: pointer to skb
  389. * @type: new skb->pkt_type type
  390. * Return: 0 on success or negative error
  391. *
  392. * int bpf_skb_under_cgroup(skb, map, index)
  393. * Check cgroup2 membership of skb
  394. * @skb: pointer to skb
  395. * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
  396. * @index: index of the cgroup in the bpf_map
  397. * Return:
  398. * == 0 skb failed the cgroup2 descendant test
  399. * == 1 skb succeeded the cgroup2 descendant test
  400. * < 0 error
  401. *
  402. * u32 bpf_get_hash_recalc(skb)
  403. * Retrieve and possibly recalculate skb->hash.
  404. * @skb: pointer to skb
  405. * Return: hash
  406. *
  407. * u64 bpf_get_current_task(void)
  408. * Returns current task_struct
  409. * Return: current
  410. *
  411. * int bpf_probe_write_user(void *dst, void *src, int len)
  412. * safely attempt to write to a location
  413. * @dst: destination address in userspace
  414. * @src: source address on stack
  415. * @len: number of bytes to copy
  416. * Return: 0 on success or negative error
  417. *
  418. * int bpf_current_task_under_cgroup(map, index)
  419. * Check cgroup2 membership of current task
  420. * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
  421. * @index: index of the cgroup in the bpf_map
  422. * Return:
  423. * == 0 current failed the cgroup2 descendant test
  424. * == 1 current succeeded the cgroup2 descendant test
  425. * < 0 error
  426. *
  427. * int bpf_skb_change_tail(skb, len, flags)
  428. * The helper will resize the skb to the given new size, to be used f.e.
  429. * with control messages.
  430. * @skb: pointer to skb
  431. * @len: new skb length
  432. * @flags: reserved
  433. * Return: 0 on success or negative error
  434. *
  435. * int bpf_skb_pull_data(skb, len)
  436. * The helper will pull in non-linear data in case the skb is non-linear
  437. * and not all of len are part of the linear section. Only needed for
  438. * read/write with direct packet access.
  439. * @skb: pointer to skb
  440. * @len: len to make read/writeable
  441. * Return: 0 on success or negative error
  442. *
  443. * s64 bpf_csum_update(skb, csum)
  444. * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
  445. * @skb: pointer to skb
  446. * @csum: csum to add
  447. * Return: csum on success or negative error
  448. *
  449. * void bpf_set_hash_invalid(skb)
  450. * Invalidate current skb->hash.
  451. * @skb: pointer to skb
  452. *
  453. * int bpf_get_numa_node_id()
  454. * Return: Id of current NUMA node.
  455. *
  456. * int bpf_skb_change_head()
  457. * Grows headroom of skb and adjusts MAC header offset accordingly.
  458. * Will extends/reallocae as required automatically.
  459. * May change skb data pointer and will thus invalidate any check
  460. * performed for direct packet access.
  461. * @skb: pointer to skb
  462. * @len: length of header to be pushed in front
  463. * @flags: Flags (unused for now)
  464. * Return: 0 on success or negative error
  465. *
  466. * int bpf_xdp_adjust_head(xdp_md, delta)
  467. * Adjust the xdp_md.data by delta
  468. * @xdp_md: pointer to xdp_md
  469. * @delta: An positive/negative integer to be added to xdp_md.data
  470. * Return: 0 on success or negative on error
  471. *
  472. * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
  473. * Copy a NUL terminated string from unsafe address. In case the string
  474. * length is smaller than size, the target is not padded with further NUL
  475. * bytes. In case the string length is larger than size, just count-1
  476. * bytes are copied and the last byte is set to NUL.
  477. * @dst: destination address
  478. * @size: maximum number of bytes to copy, including the trailing NUL
  479. * @unsafe_ptr: unsafe address
  480. * Return:
  481. * > 0 length of the string including the trailing NUL on success
  482. * < 0 error
  483. *
  484. * u64 bpf_get_socket_cookie(skb)
  485. * Get the cookie for the socket stored inside sk_buff.
  486. * @skb: pointer to skb
  487. * Return: 8 Bytes non-decreasing number on success or 0 if the socket
  488. * field is missing inside sk_buff
  489. *
  490. * u32 bpf_get_socket_uid(skb)
  491. * Get the owner uid of the socket stored inside sk_buff.
  492. * @skb: pointer to skb
  493. * Return: uid of the socket owner on success or overflowuid if failed.
  494. *
  495. * u32 bpf_set_hash(skb, hash)
  496. * Set full skb->hash.
  497. * @skb: pointer to skb
  498. * @hash: hash to set
  499. *
  500. * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
  501. * Calls setsockopt. Not all opts are available, only those with
  502. * integer optvals plus TCP_CONGESTION.
  503. * Supported levels: SOL_SOCKET and IPROTO_TCP
  504. * @bpf_socket: pointer to bpf_socket
  505. * @level: SOL_SOCKET or IPROTO_TCP
  506. * @optname: option name
  507. * @optval: pointer to option value
  508. * @optlen: length of optval in byes
  509. * Return: 0 or negative error
  510. *
  511. * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
  512. * Grow or shrink room in sk_buff.
  513. * @skb: pointer to skb
  514. * @len_diff: (signed) amount of room to grow/shrink
  515. * @mode: operation mode (enum bpf_adj_room_mode)
  516. * @flags: reserved for future use
  517. * Return: 0 on success or negative error code
  518. */
  519. #define __BPF_FUNC_MAPPER(FN) \
  520. FN(unspec), \
  521. FN(map_lookup_elem), \
  522. FN(map_update_elem), \
  523. FN(map_delete_elem), \
  524. FN(probe_read), \
  525. FN(ktime_get_ns), \
  526. FN(trace_printk), \
  527. FN(get_prandom_u32), \
  528. FN(get_smp_processor_id), \
  529. FN(skb_store_bytes), \
  530. FN(l3_csum_replace), \
  531. FN(l4_csum_replace), \
  532. FN(tail_call), \
  533. FN(clone_redirect), \
  534. FN(get_current_pid_tgid), \
  535. FN(get_current_uid_gid), \
  536. FN(get_current_comm), \
  537. FN(get_cgroup_classid), \
  538. FN(skb_vlan_push), \
  539. FN(skb_vlan_pop), \
  540. FN(skb_get_tunnel_key), \
  541. FN(skb_set_tunnel_key), \
  542. FN(perf_event_read), \
  543. FN(redirect), \
  544. FN(get_route_realm), \
  545. FN(perf_event_output), \
  546. FN(skb_load_bytes), \
  547. FN(get_stackid), \
  548. FN(csum_diff), \
  549. FN(skb_get_tunnel_opt), \
  550. FN(skb_set_tunnel_opt), \
  551. FN(skb_change_proto), \
  552. FN(skb_change_type), \
  553. FN(skb_under_cgroup), \
  554. FN(get_hash_recalc), \
  555. FN(get_current_task), \
  556. FN(probe_write_user), \
  557. FN(current_task_under_cgroup), \
  558. FN(skb_change_tail), \
  559. FN(skb_pull_data), \
  560. FN(csum_update), \
  561. FN(set_hash_invalid), \
  562. FN(get_numa_node_id), \
  563. FN(skb_change_head), \
  564. FN(xdp_adjust_head), \
  565. FN(probe_read_str), \
  566. FN(get_socket_cookie), \
  567. FN(get_socket_uid), \
  568. FN(set_hash), \
  569. FN(setsockopt), \
  570. FN(skb_adjust_room), \
  571. FN(redirect_map),
  572. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  573. * function eBPF program intends to call
  574. */
  575. #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
  576. enum bpf_func_id {
  577. __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  578. __BPF_FUNC_MAX_ID,
  579. };
  580. #undef __BPF_ENUM_FN
  581. /* All flags used by eBPF helper functions, placed here. */
  582. /* BPF_FUNC_skb_store_bytes flags. */
  583. #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
  584. #define BPF_F_INVALIDATE_HASH (1ULL << 1)
  585. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  586. * First 4 bits are for passing the header field size.
  587. */
  588. #define BPF_F_HDR_FIELD_MASK 0xfULL
  589. /* BPF_FUNC_l4_csum_replace flags. */
  590. #define BPF_F_PSEUDO_HDR (1ULL << 4)
  591. #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
  592. #define BPF_F_MARK_ENFORCE (1ULL << 6)
  593. /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
  594. #define BPF_F_INGRESS (1ULL << 0)
  595. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  596. #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
  597. /* BPF_FUNC_get_stackid flags. */
  598. #define BPF_F_SKIP_FIELD_MASK 0xffULL
  599. #define BPF_F_USER_STACK (1ULL << 8)
  600. #define BPF_F_FAST_STACK_CMP (1ULL << 9)
  601. #define BPF_F_REUSE_STACKID (1ULL << 10)
  602. /* BPF_FUNC_skb_set_tunnel_key flags. */
  603. #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
  604. #define BPF_F_DONT_FRAGMENT (1ULL << 2)
  605. /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
  606. #define BPF_F_INDEX_MASK 0xffffffffULL
  607. #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
  608. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  609. #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
  610. /* Mode for BPF_FUNC_skb_adjust_room helper. */
  611. enum bpf_adj_room_mode {
  612. BPF_ADJ_ROOM_NET_OPTS,
  613. };
  614. /* user accessible mirror of in-kernel sk_buff.
  615. * new fields can only be added to the end of this structure
  616. */
  617. struct __sk_buff {
  618. __u32 len;
  619. __u32 pkt_type;
  620. __u32 mark;
  621. __u32 queue_mapping;
  622. __u32 protocol;
  623. __u32 vlan_present;
  624. __u32 vlan_tci;
  625. __u32 vlan_proto;
  626. __u32 priority;
  627. __u32 ingress_ifindex;
  628. __u32 ifindex;
  629. __u32 tc_index;
  630. __u32 cb[5];
  631. __u32 hash;
  632. __u32 tc_classid;
  633. __u32 data;
  634. __u32 data_end;
  635. __u32 napi_id;
  636. };
  637. struct bpf_tunnel_key {
  638. __u32 tunnel_id;
  639. union {
  640. __u32 remote_ipv4;
  641. __u32 remote_ipv6[4];
  642. };
  643. __u8 tunnel_tos;
  644. __u8 tunnel_ttl;
  645. __u16 tunnel_ext;
  646. __u32 tunnel_label;
  647. };
  648. /* Generic BPF return codes which all BPF program types may support.
  649. * The values are binary compatible with their TC_ACT_* counter-part to
  650. * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  651. * programs.
  652. *
  653. * XDP is handled seprately, see XDP_*.
  654. */
  655. enum bpf_ret_code {
  656. BPF_OK = 0,
  657. /* 1 reserved */
  658. BPF_DROP = 2,
  659. /* 3-6 reserved */
  660. BPF_REDIRECT = 7,
  661. /* >127 are reserved for prog type specific return codes */
  662. };
  663. struct bpf_sock {
  664. __u32 bound_dev_if;
  665. __u32 family;
  666. __u32 type;
  667. __u32 protocol;
  668. };
  669. #define XDP_PACKET_HEADROOM 256
  670. /* User return codes for XDP prog type.
  671. * A valid XDP program must return one of these defined values. All other
  672. * return codes are reserved for future use. Unknown return codes will result
  673. * in packet drop.
  674. */
  675. enum xdp_action {
  676. XDP_ABORTED = 0,
  677. XDP_DROP,
  678. XDP_PASS,
  679. XDP_TX,
  680. };
  681. /* user accessible metadata for XDP packet hook
  682. * new fields must be added to the end of this structure
  683. */
  684. struct xdp_md {
  685. __u32 data;
  686. __u32 data_end;
  687. };
  688. #define BPF_TAG_SIZE 8
  689. struct bpf_prog_info {
  690. __u32 type;
  691. __u32 id;
  692. __u8 tag[BPF_TAG_SIZE];
  693. __u32 jited_prog_len;
  694. __u32 xlated_prog_len;
  695. __aligned_u64 jited_prog_insns;
  696. __aligned_u64 xlated_prog_insns;
  697. } __attribute__((aligned(8)));
  698. struct bpf_map_info {
  699. __u32 type;
  700. __u32 id;
  701. __u32 key_size;
  702. __u32 value_size;
  703. __u32 max_entries;
  704. __u32 map_flags;
  705. } __attribute__((aligned(8)));
  706. /* User bpf_sock_ops struct to access socket values and specify request ops
  707. * and their replies.
  708. * New fields can only be added at the end of this structure
  709. */
  710. struct bpf_sock_ops {
  711. __u32 op;
  712. union {
  713. __u32 reply;
  714. __u32 replylong[4];
  715. };
  716. __u32 family;
  717. __u32 remote_ip4;
  718. __u32 local_ip4;
  719. __u32 remote_ip6[4];
  720. __u32 local_ip6[4];
  721. __u32 remote_port;
  722. __u32 local_port;
  723. };
  724. /* List of known BPF sock_ops operators.
  725. * New entries can only be added at the end
  726. */
  727. enum {
  728. BPF_SOCK_OPS_VOID,
  729. BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
  730. * -1 if default value should be used
  731. */
  732. BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
  733. * window (in packets) or -1 if default
  734. * value should be used
  735. */
  736. BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
  737. * active connection is initialized
  738. */
  739. BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
  740. * active connection is
  741. * established
  742. */
  743. BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
  744. * passive connection is
  745. * established
  746. */
  747. BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
  748. * needs ECN
  749. */
  750. };
  751. #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
  752. #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
  753. #endif /* _UAPI__LINUX_BPF_H__ */