bpf.h 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542
  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. #define BPF_JNE 0x50 /* jump != */
  27. #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
  28. #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
  29. #define BPF_CALL 0x80 /* function call */
  30. #define BPF_EXIT 0x90 /* function return */
  31. /* Register numbers */
  32. enum {
  33. BPF_REG_0 = 0,
  34. BPF_REG_1,
  35. BPF_REG_2,
  36. BPF_REG_3,
  37. BPF_REG_4,
  38. BPF_REG_5,
  39. BPF_REG_6,
  40. BPF_REG_7,
  41. BPF_REG_8,
  42. BPF_REG_9,
  43. BPF_REG_10,
  44. __MAX_BPF_REG,
  45. };
  46. /* BPF has 10 general purpose 64-bit registers and stack frame. */
  47. #define MAX_BPF_REG __MAX_BPF_REG
  48. struct bpf_insn {
  49. __u8 code; /* opcode */
  50. __u8 dst_reg:4; /* dest register */
  51. __u8 src_reg:4; /* source register */
  52. __s16 off; /* signed offset */
  53. __s32 imm; /* signed immediate constant */
  54. };
  55. /* BPF syscall commands, see bpf(2) man-page for details. */
  56. enum bpf_cmd {
  57. BPF_MAP_CREATE,
  58. BPF_MAP_LOOKUP_ELEM,
  59. BPF_MAP_UPDATE_ELEM,
  60. BPF_MAP_DELETE_ELEM,
  61. BPF_MAP_GET_NEXT_KEY,
  62. BPF_PROG_LOAD,
  63. BPF_OBJ_PIN,
  64. BPF_OBJ_GET,
  65. };
  66. enum bpf_map_type {
  67. BPF_MAP_TYPE_UNSPEC,
  68. BPF_MAP_TYPE_HASH,
  69. BPF_MAP_TYPE_ARRAY,
  70. BPF_MAP_TYPE_PROG_ARRAY,
  71. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  72. BPF_MAP_TYPE_PERCPU_HASH,
  73. BPF_MAP_TYPE_PERCPU_ARRAY,
  74. BPF_MAP_TYPE_STACK_TRACE,
  75. BPF_MAP_TYPE_CGROUP_ARRAY,
  76. BPF_MAP_TYPE_LRU_HASH,
  77. };
  78. enum bpf_prog_type {
  79. BPF_PROG_TYPE_UNSPEC,
  80. BPF_PROG_TYPE_SOCKET_FILTER,
  81. BPF_PROG_TYPE_KPROBE,
  82. BPF_PROG_TYPE_SCHED_CLS,
  83. BPF_PROG_TYPE_SCHED_ACT,
  84. BPF_PROG_TYPE_TRACEPOINT,
  85. BPF_PROG_TYPE_XDP,
  86. BPF_PROG_TYPE_PERF_EVENT,
  87. };
  88. #define BPF_PSEUDO_MAP_FD 1
  89. /* flags for BPF_MAP_UPDATE_ELEM command */
  90. #define BPF_ANY 0 /* create new element or update existing */
  91. #define BPF_NOEXIST 1 /* create new element if it didn't exist */
  92. #define BPF_EXIST 2 /* update existing element */
  93. #define BPF_F_NO_PREALLOC (1U << 0)
  94. /* Instead of having one common LRU list in the
  95. * BPF_MAP_TYPE_LRU_HASH map, use a percpu LRU list
  96. * which can scale and perform better.
  97. * Note, the LRU nodes (including free nodes) cannot be moved
  98. * across different LRU lists.
  99. */
  100. #define BPF_F_NO_COMMON_LRU (1U << 1)
  101. union bpf_attr {
  102. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  103. __u32 map_type; /* one of enum bpf_map_type */
  104. __u32 key_size; /* size of key in bytes */
  105. __u32 value_size; /* size of value in bytes */
  106. __u32 max_entries; /* max number of entries in a map */
  107. __u32 map_flags; /* prealloc or not */
  108. };
  109. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  110. __u32 map_fd;
  111. __aligned_u64 key;
  112. union {
  113. __aligned_u64 value;
  114. __aligned_u64 next_key;
  115. };
  116. __u64 flags;
  117. };
  118. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  119. __u32 prog_type; /* one of enum bpf_prog_type */
  120. __u32 insn_cnt;
  121. __aligned_u64 insns;
  122. __aligned_u64 license;
  123. __u32 log_level; /* verbosity level of verifier */
  124. __u32 log_size; /* size of user buffer */
  125. __aligned_u64 log_buf; /* user supplied buffer */
  126. __u32 kern_version; /* checked when prog_type=kprobe */
  127. };
  128. struct { /* anonymous struct used by BPF_OBJ_* commands */
  129. __aligned_u64 pathname;
  130. __u32 bpf_fd;
  131. };
  132. } __attribute__((aligned(8)));
  133. /* BPF helper function descriptions:
  134. *
  135. * void *bpf_map_lookup_elem(&map, &key)
  136. * Return: Map value or NULL
  137. *
  138. * int bpf_map_update_elem(&map, &key, &value, flags)
  139. * Return: 0 on success or negative error
  140. *
  141. * int bpf_map_delete_elem(&map, &key)
  142. * Return: 0 on success or negative error
  143. *
  144. * int bpf_probe_read(void *dst, int size, void *src)
  145. * Return: 0 on success or negative error
  146. *
  147. * u64 bpf_ktime_get_ns(void)
  148. * Return: current ktime
  149. *
  150. * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
  151. * Return: length of buffer written or negative error
  152. *
  153. * u32 bpf_prandom_u32(void)
  154. * Return: random value
  155. *
  156. * u32 bpf_raw_smp_processor_id(void)
  157. * Return: SMP processor ID
  158. *
  159. * int bpf_skb_store_bytes(skb, offset, from, len, flags)
  160. * store bytes into packet
  161. * @skb: pointer to skb
  162. * @offset: offset within packet from skb->mac_header
  163. * @from: pointer where to copy bytes from
  164. * @len: number of bytes to store into packet
  165. * @flags: bit 0 - if true, recompute skb->csum
  166. * other bits - reserved
  167. * Return: 0 on success or negative error
  168. *
  169. * int bpf_l3_csum_replace(skb, offset, from, to, flags)
  170. * recompute IP checksum
  171. * @skb: pointer to skb
  172. * @offset: offset within packet where IP checksum is located
  173. * @from: old value of header field
  174. * @to: new value of header field
  175. * @flags: bits 0-3 - size of header field
  176. * other bits - reserved
  177. * Return: 0 on success or negative error
  178. *
  179. * int bpf_l4_csum_replace(skb, offset, from, to, flags)
  180. * recompute TCP/UDP checksum
  181. * @skb: pointer to skb
  182. * @offset: offset within packet where TCP/UDP checksum is located
  183. * @from: old value of header field
  184. * @to: new value of header field
  185. * @flags: bits 0-3 - size of header field
  186. * bit 4 - is pseudo header
  187. * other bits - reserved
  188. * Return: 0 on success or negative error
  189. *
  190. * int bpf_tail_call(ctx, prog_array_map, index)
  191. * jump into another BPF program
  192. * @ctx: context pointer passed to next program
  193. * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
  194. * @index: index inside array that selects specific program to run
  195. * Return: 0 on success or negative error
  196. *
  197. * int bpf_clone_redirect(skb, ifindex, flags)
  198. * redirect to another netdev
  199. * @skb: pointer to skb
  200. * @ifindex: ifindex of the net device
  201. * @flags: bit 0 - if set, redirect to ingress instead of egress
  202. * other bits - reserved
  203. * Return: 0 on success or negative error
  204. *
  205. * u64 bpf_get_current_pid_tgid(void)
  206. * Return: current->tgid << 32 | current->pid
  207. *
  208. * u64 bpf_get_current_uid_gid(void)
  209. * Return: current_gid << 32 | current_uid
  210. *
  211. * int bpf_get_current_comm(char *buf, int size_of_buf)
  212. * stores current->comm into buf
  213. * Return: 0 on success or negative error
  214. *
  215. * u32 bpf_get_cgroup_classid(skb)
  216. * retrieve a proc's classid
  217. * @skb: pointer to skb
  218. * Return: classid if != 0
  219. *
  220. * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
  221. * Return: 0 on success or negative error
  222. *
  223. * int bpf_skb_vlan_pop(skb)
  224. * Return: 0 on success or negative error
  225. *
  226. * int bpf_skb_get_tunnel_key(skb, key, size, flags)
  227. * int bpf_skb_set_tunnel_key(skb, key, size, flags)
  228. * retrieve or populate tunnel metadata
  229. * @skb: pointer to skb
  230. * @key: pointer to 'struct bpf_tunnel_key'
  231. * @size: size of 'struct bpf_tunnel_key'
  232. * @flags: room for future extensions
  233. * Return: 0 on success or negative error
  234. *
  235. * u64 bpf_perf_event_read(&map, index)
  236. * Return: Number events read or error code
  237. *
  238. * int bpf_redirect(ifindex, flags)
  239. * redirect to another netdev
  240. * @ifindex: ifindex of the net device
  241. * @flags: bit 0 - if set, redirect to ingress instead of egress
  242. * other bits - reserved
  243. * Return: TC_ACT_REDIRECT
  244. *
  245. * u32 bpf_get_route_realm(skb)
  246. * retrieve a dst's tclassid
  247. * @skb: pointer to skb
  248. * Return: realm if != 0
  249. *
  250. * int bpf_perf_event_output(ctx, map, index, data, size)
  251. * output perf raw sample
  252. * @ctx: struct pt_regs*
  253. * @map: pointer to perf_event_array map
  254. * @index: index of event in the map
  255. * @data: data on stack to be output as raw data
  256. * @size: size of data
  257. * Return: 0 on success or negative error
  258. *
  259. * int bpf_get_stackid(ctx, map, flags)
  260. * walk user or kernel stack and return id
  261. * @ctx: struct pt_regs*
  262. * @map: pointer to stack_trace map
  263. * @flags: bits 0-7 - numer of stack frames to skip
  264. * bit 8 - collect user stack instead of kernel
  265. * bit 9 - compare stacks by hash only
  266. * bit 10 - if two different stacks hash into the same stackid
  267. * discard old
  268. * other bits - reserved
  269. * Return: >= 0 stackid on success or negative error
  270. *
  271. * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
  272. * calculate csum diff
  273. * @from: raw from buffer
  274. * @from_size: length of from buffer
  275. * @to: raw to buffer
  276. * @to_size: length of to buffer
  277. * @seed: optional seed
  278. * Return: csum result or negative error code
  279. *
  280. * int bpf_skb_get_tunnel_opt(skb, opt, size)
  281. * retrieve tunnel options metadata
  282. * @skb: pointer to skb
  283. * @opt: pointer to raw tunnel option data
  284. * @size: size of @opt
  285. * Return: option size
  286. *
  287. * int bpf_skb_set_tunnel_opt(skb, opt, size)
  288. * populate tunnel options metadata
  289. * @skb: pointer to skb
  290. * @opt: pointer to raw tunnel option data
  291. * @size: size of @opt
  292. * Return: 0 on success or negative error
  293. *
  294. * int bpf_skb_change_proto(skb, proto, flags)
  295. * Change protocol of the skb. Currently supported is v4 -> v6,
  296. * v6 -> v4 transitions. The helper will also resize the skb. eBPF
  297. * program is expected to fill the new headers via skb_store_bytes
  298. * and lX_csum_replace.
  299. * @skb: pointer to skb
  300. * @proto: new skb->protocol type
  301. * @flags: reserved
  302. * Return: 0 on success or negative error
  303. *
  304. * int bpf_skb_change_type(skb, type)
  305. * Change packet type of skb.
  306. * @skb: pointer to skb
  307. * @type: new skb->pkt_type type
  308. * Return: 0 on success or negative error
  309. *
  310. * int bpf_skb_under_cgroup(skb, map, index)
  311. * Check cgroup2 membership of skb
  312. * @skb: pointer to skb
  313. * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
  314. * @index: index of the cgroup in the bpf_map
  315. * Return:
  316. * == 0 skb failed the cgroup2 descendant test
  317. * == 1 skb succeeded the cgroup2 descendant test
  318. * < 0 error
  319. *
  320. * u32 bpf_get_hash_recalc(skb)
  321. * Retrieve and possibly recalculate skb->hash.
  322. * @skb: pointer to skb
  323. * Return: hash
  324. *
  325. * u64 bpf_get_current_task(void)
  326. * Returns current task_struct
  327. * Return: current
  328. *
  329. * int bpf_probe_write_user(void *dst, void *src, int len)
  330. * safely attempt to write to a location
  331. * @dst: destination address in userspace
  332. * @src: source address on stack
  333. * @len: number of bytes to copy
  334. * Return: 0 on success or negative error
  335. *
  336. * int bpf_current_task_under_cgroup(map, index)
  337. * Check cgroup2 membership of current task
  338. * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
  339. * @index: index of the cgroup in the bpf_map
  340. * Return:
  341. * == 0 current failed the cgroup2 descendant test
  342. * == 1 current succeeded the cgroup2 descendant test
  343. * < 0 error
  344. *
  345. * int bpf_skb_change_tail(skb, len, flags)
  346. * The helper will resize the skb to the given new size, to be used f.e.
  347. * with control messages.
  348. * @skb: pointer to skb
  349. * @len: new skb length
  350. * @flags: reserved
  351. * Return: 0 on success or negative error
  352. *
  353. * int bpf_skb_pull_data(skb, len)
  354. * The helper will pull in non-linear data in case the skb is non-linear
  355. * and not all of len are part of the linear section. Only needed for
  356. * read/write with direct packet access.
  357. * @skb: pointer to skb
  358. * @len: len to make read/writeable
  359. * Return: 0 on success or negative error
  360. *
  361. * s64 bpf_csum_update(skb, csum)
  362. * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
  363. * @skb: pointer to skb
  364. * @csum: csum to add
  365. * Return: csum on success or negative error
  366. *
  367. * void bpf_set_hash_invalid(skb)
  368. * Invalidate current skb->hash.
  369. * @skb: pointer to skb
  370. *
  371. * int bpf_get_numa_node_id()
  372. * Return: Id of current NUMA node.
  373. */
  374. #define __BPF_FUNC_MAPPER(FN) \
  375. FN(unspec), \
  376. FN(map_lookup_elem), \
  377. FN(map_update_elem), \
  378. FN(map_delete_elem), \
  379. FN(probe_read), \
  380. FN(ktime_get_ns), \
  381. FN(trace_printk), \
  382. FN(get_prandom_u32), \
  383. FN(get_smp_processor_id), \
  384. FN(skb_store_bytes), \
  385. FN(l3_csum_replace), \
  386. FN(l4_csum_replace), \
  387. FN(tail_call), \
  388. FN(clone_redirect), \
  389. FN(get_current_pid_tgid), \
  390. FN(get_current_uid_gid), \
  391. FN(get_current_comm), \
  392. FN(get_cgroup_classid), \
  393. FN(skb_vlan_push), \
  394. FN(skb_vlan_pop), \
  395. FN(skb_get_tunnel_key), \
  396. FN(skb_set_tunnel_key), \
  397. FN(perf_event_read), \
  398. FN(redirect), \
  399. FN(get_route_realm), \
  400. FN(perf_event_output), \
  401. FN(skb_load_bytes), \
  402. FN(get_stackid), \
  403. FN(csum_diff), \
  404. FN(skb_get_tunnel_opt), \
  405. FN(skb_set_tunnel_opt), \
  406. FN(skb_change_proto), \
  407. FN(skb_change_type), \
  408. FN(skb_under_cgroup), \
  409. FN(get_hash_recalc), \
  410. FN(get_current_task), \
  411. FN(probe_write_user), \
  412. FN(current_task_under_cgroup), \
  413. FN(skb_change_tail), \
  414. FN(skb_pull_data), \
  415. FN(csum_update), \
  416. FN(set_hash_invalid), \
  417. FN(get_numa_node_id),
  418. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  419. * function eBPF program intends to call
  420. */
  421. #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
  422. enum bpf_func_id {
  423. __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  424. __BPF_FUNC_MAX_ID,
  425. };
  426. #undef __BPF_ENUM_FN
  427. /* All flags used by eBPF helper functions, placed here. */
  428. /* BPF_FUNC_skb_store_bytes flags. */
  429. #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
  430. #define BPF_F_INVALIDATE_HASH (1ULL << 1)
  431. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  432. * First 4 bits are for passing the header field size.
  433. */
  434. #define BPF_F_HDR_FIELD_MASK 0xfULL
  435. /* BPF_FUNC_l4_csum_replace flags. */
  436. #define BPF_F_PSEUDO_HDR (1ULL << 4)
  437. #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
  438. /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
  439. #define BPF_F_INGRESS (1ULL << 0)
  440. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  441. #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
  442. /* BPF_FUNC_get_stackid flags. */
  443. #define BPF_F_SKIP_FIELD_MASK 0xffULL
  444. #define BPF_F_USER_STACK (1ULL << 8)
  445. #define BPF_F_FAST_STACK_CMP (1ULL << 9)
  446. #define BPF_F_REUSE_STACKID (1ULL << 10)
  447. /* BPF_FUNC_skb_set_tunnel_key flags. */
  448. #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
  449. #define BPF_F_DONT_FRAGMENT (1ULL << 2)
  450. /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
  451. #define BPF_F_INDEX_MASK 0xffffffffULL
  452. #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
  453. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  454. #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
  455. /* user accessible mirror of in-kernel sk_buff.
  456. * new fields can only be added to the end of this structure
  457. */
  458. struct __sk_buff {
  459. __u32 len;
  460. __u32 pkt_type;
  461. __u32 mark;
  462. __u32 queue_mapping;
  463. __u32 protocol;
  464. __u32 vlan_present;
  465. __u32 vlan_tci;
  466. __u32 vlan_proto;
  467. __u32 priority;
  468. __u32 ingress_ifindex;
  469. __u32 ifindex;
  470. __u32 tc_index;
  471. __u32 cb[5];
  472. __u32 hash;
  473. __u32 tc_classid;
  474. __u32 data;
  475. __u32 data_end;
  476. };
  477. struct bpf_tunnel_key {
  478. __u32 tunnel_id;
  479. union {
  480. __u32 remote_ipv4;
  481. __u32 remote_ipv6[4];
  482. };
  483. __u8 tunnel_tos;
  484. __u8 tunnel_ttl;
  485. __u16 tunnel_ext;
  486. __u32 tunnel_label;
  487. };
  488. /* User return codes for XDP prog type.
  489. * A valid XDP program must return one of these defined values. All other
  490. * return codes are reserved for future use. Unknown return codes will result
  491. * in packet drop.
  492. */
  493. enum xdp_action {
  494. XDP_ABORTED = 0,
  495. XDP_DROP,
  496. XDP_PASS,
  497. XDP_TX,
  498. };
  499. /* user accessible metadata for XDP packet hook
  500. * new fields must be added to the end of this structure
  501. */
  502. struct xdp_md {
  503. __u32 data;
  504. __u32 data_end;
  505. };
  506. #endif /* _UAPI__LINUX_BPF_H__ */