blob: f94b48b168dcc82cbd67b916951a4699f4f675ef [file] [log] [blame]
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07001/* 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
10#include <linux/types.h>
Alexei Starovoitovc15952d2014-10-14 02:08:54 -070011#include <linux/bpf_common.h>
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070012
13/* Extended instruction set based on top of classic BPF */
14
15/* instruction classes */
16#define BPF_ALU64 0x07 /* alu mode in double word width */
17
18/* ld/ldx fields */
19#define BPF_DW 0x18 /* double word */
20#define BPF_XADD 0xc0 /* exclusive add */
21
22/* alu/jmp fields */
23#define BPF_MOV 0xb0 /* mov reg to reg */
24#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
25
26/* change endianness of a register */
27#define BPF_END 0xd0 /* flags for endianness conversion: */
28#define BPF_TO_LE 0x00 /* convert to little-endian */
29#define BPF_TO_BE 0x08 /* convert to big-endian */
30#define BPF_FROM_LE BPF_TO_LE
31#define BPF_FROM_BE BPF_TO_BE
32
33#define BPF_JNE 0x50 /* jump != */
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_CALL 0x80 /* function call */
37#define BPF_EXIT 0x90 /* function return */
38
39/* Register numbers */
40enum {
41 BPF_REG_0 = 0,
42 BPF_REG_1,
43 BPF_REG_2,
44 BPF_REG_3,
45 BPF_REG_4,
46 BPF_REG_5,
47 BPF_REG_6,
48 BPF_REG_7,
49 BPF_REG_8,
50 BPF_REG_9,
51 BPF_REG_10,
52 __MAX_BPF_REG,
53};
54
55/* BPF has 10 general purpose 64-bit registers and stack frame. */
56#define MAX_BPF_REG __MAX_BPF_REG
57
58struct bpf_insn {
59 __u8 code; /* opcode */
60 __u8 dst_reg:4; /* dest register */
61 __u8 src_reg:4; /* source register */
62 __s16 off; /* signed offset */
63 __s32 imm; /* signed immediate constant */
64};
65
Daniel Mackb95a5c42017-01-21 17:26:11 +010066/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
67struct bpf_lpm_trie_key {
68 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
69 __u8 data[0]; /* Arbitrary size */
70};
71
Daniel Borkmannb2197752015-10-29 14:58:09 +010072/* BPF syscall commands, see bpf(2) man-page for details. */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070073enum bpf_cmd {
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070074 BPF_MAP_CREATE,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070075 BPF_MAP_LOOKUP_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070076 BPF_MAP_UPDATE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070077 BPF_MAP_DELETE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070078 BPF_MAP_GET_NEXT_KEY,
Alexei Starovoitov09756af2014-09-26 00:17:00 -070079 BPF_PROG_LOAD,
Daniel Borkmannb2197752015-10-29 14:58:09 +010080 BPF_OBJ_PIN,
81 BPF_OBJ_GET,
Daniel Mackf4324552016-11-23 16:52:27 +010082 BPF_PROG_ATTACH,
83 BPF_PROG_DETACH,
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -070084 BPF_PROG_TEST_RUN,
Martin KaFai Lau34ad5582017-06-05 12:15:48 -070085 BPF_PROG_GET_NEXT_ID,
86 BPF_MAP_GET_NEXT_ID,
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -070087 BPF_PROG_GET_FD_BY_ID,
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -070088 BPF_MAP_GET_FD_BY_ID,
Martin KaFai Lau1e270972017-06-05 12:15:52 -070089 BPF_OBJ_GET_INFO_BY_FD,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070090};
91
92enum bpf_map_type {
93 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -080094 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -080095 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61a2015-05-19 16:59:03 -070096 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +000097 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -080098 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -080099 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800100 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -0700101 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800102 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800103 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +0100104 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700105 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700106 BPF_MAP_TYPE_HASH_OF_MAPS,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700107};
108
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700109enum bpf_prog_type {
110 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800111 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700112 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100113 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee82015-03-20 15:11:11 +0100114 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700115 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700116 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700117 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100118 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800119 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100120 BPF_PROG_TYPE_LWT_IN,
121 BPF_PROG_TYPE_LWT_OUT,
122 BPF_PROG_TYPE_LWT_XMIT,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700123};
124
Daniel Mack0e336612016-11-23 16:52:25 +0100125enum bpf_attach_type {
126 BPF_CGROUP_INET_INGRESS,
127 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800128 BPF_CGROUP_INET_SOCK_CREATE,
Daniel Mack0e336612016-11-23 16:52:25 +0100129 __MAX_BPF_ATTACH_TYPE
130};
131
132#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
133
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800134/* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
135 * to the given target_fd cgroup the descendent cgroup will be able to
136 * override effective bpf program that was inherited from this cgroup
137 */
138#define BPF_F_ALLOW_OVERRIDE (1U << 0)
139
David S. Millere07b98d2017-05-10 11:38:07 -0700140/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
141 * verifier will perform strict alignment checking as if the kernel
142 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
143 * and NET_IP_ALIGN defined to 2.
144 */
145#define BPF_F_STRICT_ALIGNMENT (1U << 0)
146
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100147#define BPF_PSEUDO_MAP_FD 1
148
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800149/* flags for BPF_MAP_UPDATE_ELEM command */
150#define BPF_ANY 0 /* create new element or update existing */
151#define BPF_NOEXIST 1 /* create new element if it didn't exist */
152#define BPF_EXIST 2 /* update existing element */
153
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800154#define BPF_F_NO_PREALLOC (1U << 0)
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800155/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800156 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800157 * which can scale and perform better.
158 * Note, the LRU nodes (including free nodes) cannot be moved
159 * across different LRU lists.
160 */
161#define BPF_F_NO_COMMON_LRU (1U << 1)
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800162
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700163union bpf_attr {
164 struct { /* anonymous struct used by BPF_MAP_CREATE command */
165 __u32 map_type; /* one of enum bpf_map_type */
166 __u32 key_size; /* size of key in bytes */
167 __u32 value_size; /* size of value in bytes */
168 __u32 max_entries; /* max number of entries in a map */
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800169 __u32 map_flags; /* prealloc or not */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700170 __u32 inner_map_fd; /* fd pointing to the inner map */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700171 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700172
173 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
174 __u32 map_fd;
175 __aligned_u64 key;
176 union {
177 __aligned_u64 value;
178 __aligned_u64 next_key;
179 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800180 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700181 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700182
183 struct { /* anonymous struct used by BPF_PROG_LOAD command */
184 __u32 prog_type; /* one of enum bpf_prog_type */
185 __u32 insn_cnt;
186 __aligned_u64 insns;
187 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700188 __u32 log_level; /* verbosity level of verifier */
189 __u32 log_size; /* size of user buffer */
190 __aligned_u64 log_buf; /* user supplied buffer */
Alexei Starovoitov25415172015-03-25 12:49:20 -0700191 __u32 kern_version; /* checked when prog_type=kprobe */
David S. Millere07b98d2017-05-10 11:38:07 -0700192 __u32 prog_flags;
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700193 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100194
195 struct { /* anonymous struct used by BPF_OBJ_* commands */
196 __aligned_u64 pathname;
197 __u32 bpf_fd;
198 };
Daniel Mackf4324552016-11-23 16:52:27 +0100199
200 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
201 __u32 target_fd; /* container object to attach to */
202 __u32 attach_bpf_fd; /* eBPF program to attach */
203 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800204 __u32 attach_flags;
Daniel Mackf4324552016-11-23 16:52:27 +0100205 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700206
207 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
208 __u32 prog_fd;
209 __u32 retval;
210 __u32 data_size_in;
211 __u32 data_size_out;
212 __aligned_u64 data_in;
213 __aligned_u64 data_out;
214 __u32 repeat;
215 __u32 duration;
216 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700217
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700218 struct { /* anonymous struct used by BPF_*_GET_*_ID */
219 union {
220 __u32 start_id;
221 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700222 __u32 map_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700223 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700224 __u32 next_id;
225 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700226
227 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
228 __u32 bpf_fd;
229 __u32 info_len;
230 __aligned_u64 info;
231 } info;
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700232} __attribute__((aligned(8)));
233
Thomas Grafebb676d2016-10-27 11:23:51 +0200234/* BPF helper function descriptions:
235 *
236 * void *bpf_map_lookup_elem(&map, &key)
237 * Return: Map value or NULL
238 *
239 * int bpf_map_update_elem(&map, &key, &value, flags)
240 * Return: 0 on success or negative error
241 *
242 * int bpf_map_delete_elem(&map, &key)
243 * Return: 0 on success or negative error
244 *
245 * int bpf_probe_read(void *dst, int size, void *src)
246 * Return: 0 on success or negative error
247 *
248 * u64 bpf_ktime_get_ns(void)
249 * Return: current ktime
250 *
251 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
252 * Return: length of buffer written or negative error
253 *
254 * u32 bpf_prandom_u32(void)
255 * Return: random value
256 *
257 * u32 bpf_raw_smp_processor_id(void)
258 * Return: SMP processor ID
259 *
260 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
261 * store bytes into packet
262 * @skb: pointer to skb
263 * @offset: offset within packet from skb->mac_header
264 * @from: pointer where to copy bytes from
265 * @len: number of bytes to store into packet
266 * @flags: bit 0 - if true, recompute skb->csum
267 * other bits - reserved
268 * Return: 0 on success or negative error
269 *
270 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
271 * recompute IP checksum
272 * @skb: pointer to skb
273 * @offset: offset within packet where IP checksum is located
274 * @from: old value of header field
275 * @to: new value of header field
276 * @flags: bits 0-3 - size of header field
277 * other bits - reserved
278 * Return: 0 on success or negative error
279 *
280 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
281 * recompute TCP/UDP checksum
282 * @skb: pointer to skb
283 * @offset: offset within packet where TCP/UDP checksum is located
284 * @from: old value of header field
285 * @to: new value of header field
286 * @flags: bits 0-3 - size of header field
287 * bit 4 - is pseudo header
288 * other bits - reserved
289 * Return: 0 on success or negative error
290 *
291 * int bpf_tail_call(ctx, prog_array_map, index)
292 * jump into another BPF program
293 * @ctx: context pointer passed to next program
294 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
295 * @index: index inside array that selects specific program to run
296 * Return: 0 on success or negative error
297 *
298 * int bpf_clone_redirect(skb, ifindex, flags)
299 * redirect to another netdev
300 * @skb: pointer to skb
301 * @ifindex: ifindex of the net device
302 * @flags: bit 0 - if set, redirect to ingress instead of egress
303 * other bits - reserved
304 * Return: 0 on success or negative error
305 *
306 * u64 bpf_get_current_pid_tgid(void)
307 * Return: current->tgid << 32 | current->pid
308 *
309 * u64 bpf_get_current_uid_gid(void)
310 * Return: current_gid << 32 | current_uid
311 *
312 * int bpf_get_current_comm(char *buf, int size_of_buf)
313 * stores current->comm into buf
314 * Return: 0 on success or negative error
315 *
316 * u32 bpf_get_cgroup_classid(skb)
317 * retrieve a proc's classid
318 * @skb: pointer to skb
319 * Return: classid if != 0
320 *
321 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
322 * Return: 0 on success or negative error
323 *
324 * int bpf_skb_vlan_pop(skb)
325 * Return: 0 on success or negative error
326 *
327 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
328 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
329 * retrieve or populate tunnel metadata
330 * @skb: pointer to skb
331 * @key: pointer to 'struct bpf_tunnel_key'
332 * @size: size of 'struct bpf_tunnel_key'
333 * @flags: room for future extensions
334 * Return: 0 on success or negative error
335 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700336 * u64 bpf_perf_event_read(map, flags)
337 * read perf event counter value
338 * @map: pointer to perf_event_array map
339 * @flags: index of event in the map or bitmask flags
340 * Return: value of perf event counter read or error code
Thomas Grafebb676d2016-10-27 11:23:51 +0200341 *
342 * int bpf_redirect(ifindex, flags)
343 * redirect to another netdev
344 * @ifindex: ifindex of the net device
345 * @flags: bit 0 - if set, redirect to ingress instead of egress
346 * other bits - reserved
347 * Return: TC_ACT_REDIRECT
348 *
349 * u32 bpf_get_route_realm(skb)
350 * retrieve a dst's tclassid
351 * @skb: pointer to skb
352 * Return: realm if != 0
353 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700354 * int bpf_perf_event_output(ctx, map, flags, data, size)
Thomas Grafebb676d2016-10-27 11:23:51 +0200355 * output perf raw sample
356 * @ctx: struct pt_regs*
357 * @map: pointer to perf_event_array map
Teng Qinb7d3ed52017-06-02 21:03:54 -0700358 * @flags: index of event in the map or bitmask flags
Thomas Grafebb676d2016-10-27 11:23:51 +0200359 * @data: data on stack to be output as raw data
360 * @size: size of data
361 * Return: 0 on success or negative error
362 *
363 * int bpf_get_stackid(ctx, map, flags)
364 * walk user or kernel stack and return id
365 * @ctx: struct pt_regs*
366 * @map: pointer to stack_trace map
367 * @flags: bits 0-7 - numer of stack frames to skip
368 * bit 8 - collect user stack instead of kernel
369 * bit 9 - compare stacks by hash only
370 * bit 10 - if two different stacks hash into the same stackid
371 * discard old
372 * other bits - reserved
373 * Return: >= 0 stackid on success or negative error
374 *
375 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
376 * calculate csum diff
377 * @from: raw from buffer
378 * @from_size: length of from buffer
379 * @to: raw to buffer
380 * @to_size: length of to buffer
381 * @seed: optional seed
382 * Return: csum result or negative error code
383 *
384 * int bpf_skb_get_tunnel_opt(skb, opt, size)
385 * retrieve tunnel options metadata
386 * @skb: pointer to skb
387 * @opt: pointer to raw tunnel option data
388 * @size: size of @opt
389 * Return: option size
390 *
391 * int bpf_skb_set_tunnel_opt(skb, opt, size)
392 * populate tunnel options metadata
393 * @skb: pointer to skb
394 * @opt: pointer to raw tunnel option data
395 * @size: size of @opt
396 * Return: 0 on success or negative error
397 *
398 * int bpf_skb_change_proto(skb, proto, flags)
399 * Change protocol of the skb. Currently supported is v4 -> v6,
400 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
401 * program is expected to fill the new headers via skb_store_bytes
402 * and lX_csum_replace.
403 * @skb: pointer to skb
404 * @proto: new skb->protocol type
405 * @flags: reserved
406 * Return: 0 on success or negative error
407 *
408 * int bpf_skb_change_type(skb, type)
409 * Change packet type of skb.
410 * @skb: pointer to skb
411 * @type: new skb->pkt_type type
412 * Return: 0 on success or negative error
413 *
414 * int bpf_skb_under_cgroup(skb, map, index)
415 * Check cgroup2 membership of skb
416 * @skb: pointer to skb
417 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
418 * @index: index of the cgroup in the bpf_map
419 * Return:
420 * == 0 skb failed the cgroup2 descendant test
421 * == 1 skb succeeded the cgroup2 descendant test
422 * < 0 error
423 *
424 * u32 bpf_get_hash_recalc(skb)
425 * Retrieve and possibly recalculate skb->hash.
426 * @skb: pointer to skb
427 * Return: hash
428 *
429 * u64 bpf_get_current_task(void)
430 * Returns current task_struct
431 * Return: current
432 *
433 * int bpf_probe_write_user(void *dst, void *src, int len)
434 * safely attempt to write to a location
435 * @dst: destination address in userspace
436 * @src: source address on stack
437 * @len: number of bytes to copy
438 * Return: 0 on success or negative error
439 *
440 * int bpf_current_task_under_cgroup(map, index)
441 * Check cgroup2 membership of current task
442 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
443 * @index: index of the cgroup in the bpf_map
444 * Return:
445 * == 0 current failed the cgroup2 descendant test
446 * == 1 current succeeded the cgroup2 descendant test
447 * < 0 error
448 *
449 * int bpf_skb_change_tail(skb, len, flags)
450 * The helper will resize the skb to the given new size, to be used f.e.
451 * with control messages.
452 * @skb: pointer to skb
453 * @len: new skb length
454 * @flags: reserved
455 * Return: 0 on success or negative error
456 *
457 * int bpf_skb_pull_data(skb, len)
458 * The helper will pull in non-linear data in case the skb is non-linear
459 * and not all of len are part of the linear section. Only needed for
460 * read/write with direct packet access.
461 * @skb: pointer to skb
462 * @len: len to make read/writeable
463 * Return: 0 on success or negative error
464 *
465 * s64 bpf_csum_update(skb, csum)
466 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
467 * @skb: pointer to skb
468 * @csum: csum to add
469 * Return: csum on success or negative error
470 *
471 * void bpf_set_hash_invalid(skb)
472 * Invalidate current skb->hash.
473 * @skb: pointer to skb
474 *
475 * int bpf_get_numa_node_id()
476 * Return: Id of current NUMA node.
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100477 *
478 * int bpf_skb_change_head()
479 * Grows headroom of skb and adjusts MAC header offset accordingly.
480 * Will extends/reallocae as required automatically.
481 * May change skb data pointer and will thus invalidate any check
482 * performed for direct packet access.
483 * @skb: pointer to skb
484 * @len: length of header to be pushed in front
485 * @flags: Flags (unused for now)
486 * Return: 0 on success or negative error
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800487 *
488 * int bpf_xdp_adjust_head(xdp_md, delta)
489 * Adjust the xdp_md.data by delta
490 * @xdp_md: pointer to xdp_md
491 * @delta: An positive/negative integer to be added to xdp_md.data
492 * Return: 0 on success or negative on error
Gianluca Borelloa5e8c072017-01-18 17:55:49 +0000493 *
494 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
495 * Copy a NUL terminated string from unsafe address. In case the string
496 * length is smaller than size, the target is not padded with further NUL
497 * bytes. In case the string length is larger than size, just count-1
498 * bytes are copied and the last byte is set to NUL.
499 * @dst: destination address
500 * @size: maximum number of bytes to copy, including the trailing NUL
501 * @unsafe_ptr: unsafe address
502 * Return:
503 * > 0 length of the string including the trailing NUL on success
504 * < 0 error
Chenbo Feng91b82702017-03-22 17:27:34 -0700505 *
Alexander Alemayhu3c60a532017-04-08 22:08:10 +0200506 * u64 bpf_get_socket_cookie(skb)
Chenbo Feng91b82702017-03-22 17:27:34 -0700507 * Get the cookie for the socket stored inside sk_buff.
508 * @skb: pointer to skb
509 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
510 * field is missing inside sk_buff
Chenbo Feng6acc5c22017-03-22 17:27:35 -0700511 *
512 * u32 bpf_get_socket_uid(skb)
513 * Get the owner uid of the socket stored inside sk_buff.
514 * @skb: pointer to skb
Chenbo Feng5d4e3442017-04-26 16:41:23 -0700515 * Return: uid of the socket owner on success or overflowuid if failed.
Daniel Borkmannded092c2017-06-11 00:50:47 +0200516 *
517 * u32 bpf_set_hash(skb, hash)
518 * Set full skb->hash.
519 * @skb: pointer to skb
520 * @hash: hash to set
Thomas Grafebb676d2016-10-27 11:23:51 +0200521 */
522#define __BPF_FUNC_MAPPER(FN) \
523 FN(unspec), \
524 FN(map_lookup_elem), \
525 FN(map_update_elem), \
526 FN(map_delete_elem), \
527 FN(probe_read), \
528 FN(ktime_get_ns), \
529 FN(trace_printk), \
530 FN(get_prandom_u32), \
531 FN(get_smp_processor_id), \
532 FN(skb_store_bytes), \
533 FN(l3_csum_replace), \
534 FN(l4_csum_replace), \
535 FN(tail_call), \
536 FN(clone_redirect), \
537 FN(get_current_pid_tgid), \
538 FN(get_current_uid_gid), \
539 FN(get_current_comm), \
540 FN(get_cgroup_classid), \
541 FN(skb_vlan_push), \
542 FN(skb_vlan_pop), \
543 FN(skb_get_tunnel_key), \
544 FN(skb_set_tunnel_key), \
545 FN(perf_event_read), \
546 FN(redirect), \
547 FN(get_route_realm), \
548 FN(perf_event_output), \
549 FN(skb_load_bytes), \
550 FN(get_stackid), \
551 FN(csum_diff), \
552 FN(skb_get_tunnel_opt), \
553 FN(skb_set_tunnel_opt), \
554 FN(skb_change_proto), \
555 FN(skb_change_type), \
556 FN(skb_under_cgroup), \
557 FN(get_hash_recalc), \
558 FN(get_current_task), \
559 FN(probe_write_user), \
560 FN(current_task_under_cgroup), \
561 FN(skb_change_tail), \
562 FN(skb_pull_data), \
563 FN(csum_update), \
564 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100565 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800566 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +0000567 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -0700568 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -0700569 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +0200570 FN(get_socket_uid), \
571 FN(set_hash),
Thomas Grafebb676d2016-10-27 11:23:51 +0200572
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700573/* integer value in 'imm' field of BPF_CALL instruction selects which helper
574 * function eBPF program intends to call
575 */
Thomas Grafebb676d2016-10-27 11:23:51 +0200576#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700577enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +0200578 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700579 __BPF_FUNC_MAX_ID,
580};
Thomas Grafebb676d2016-10-27 11:23:51 +0200581#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700582
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100583/* All flags used by eBPF helper functions, placed here. */
584
585/* BPF_FUNC_skb_store_bytes flags. */
586#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
Daniel Borkmann8afd54c2016-03-04 15:15:03 +0100587#define BPF_F_INVALIDATE_HASH (1ULL << 1)
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100588
589/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
590 * First 4 bits are for passing the header field size.
591 */
592#define BPF_F_HDR_FIELD_MASK 0xfULL
593
594/* BPF_FUNC_l4_csum_replace flags. */
595#define BPF_F_PSEUDO_HDR (1ULL << 4)
Daniel Borkmann2f729592016-02-19 23:05:26 +0100596#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
Daniel Borkmannd1b662a2017-01-24 01:06:28 +0100597#define BPF_F_MARK_ENFORCE (1ULL << 6)
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100598
599/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
600#define BPF_F_INGRESS (1ULL << 0)
601
Daniel Borkmannc6c33452016-01-11 01:16:39 +0100602/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
603#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
604
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800605/* BPF_FUNC_get_stackid flags. */
606#define BPF_F_SKIP_FIELD_MASK 0xffULL
607#define BPF_F_USER_STACK (1ULL << 8)
608#define BPF_F_FAST_STACK_CMP (1ULL << 9)
609#define BPF_F_REUSE_STACKID (1ULL << 10)
610
Daniel Borkmann2da897e2016-02-23 02:05:26 +0100611/* BPF_FUNC_skb_set_tunnel_key flags. */
612#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
Daniel Borkmann22080872016-03-04 15:15:05 +0100613#define BPF_F_DONT_FRAGMENT (1ULL << 2)
Daniel Borkmann2da897e2016-02-23 02:05:26 +0100614
Daniel Borkmann6816a7f2016-06-28 12:18:25 +0200615/* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
Daniel Borkmann1e337592016-04-18 21:01:23 +0200616#define BPF_F_INDEX_MASK 0xffffffffULL
617#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
Daniel Borkmann555c8a82016-07-14 18:08:05 +0200618/* BPF_FUNC_perf_event_output for sk_buff input context. */
619#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
Daniel Borkmann1e337592016-04-18 21:01:23 +0200620
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -0700621/* user accessible mirror of in-kernel sk_buff.
622 * new fields can only be added to the end of this structure
623 */
624struct __sk_buff {
625 __u32 len;
626 __u32 pkt_type;
627 __u32 mark;
628 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -0700629 __u32 protocol;
630 __u32 vlan_present;
631 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +0100632 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +0200633 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -0700634 __u32 ingress_ifindex;
635 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -0700636 __u32 tc_index;
637 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +0200638 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -0700639 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -0700640 __u32 data;
641 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +0200642 __u32 napi_id;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -0700643};
644
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -0700645struct bpf_tunnel_key {
646 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +0100647 union {
648 __u32 remote_ipv4;
649 __u32 remote_ipv6[4];
650 };
651 __u8 tunnel_tos;
652 __u8 tunnel_ttl;
Daniel Borkmannc0e760c2016-03-30 00:02:00 +0200653 __u16 tunnel_ext;
Daniel Borkmann4018ab12016-03-09 03:00:05 +0100654 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -0700655};
656
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100657/* Generic BPF return codes which all BPF program types may support.
658 * The values are binary compatible with their TC_ACT_* counter-part to
659 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
660 * programs.
661 *
662 * XDP is handled seprately, see XDP_*.
663 */
664enum bpf_ret_code {
665 BPF_OK = 0,
666 /* 1 reserved */
667 BPF_DROP = 2,
668 /* 3-6 reserved */
669 BPF_REDIRECT = 7,
670 /* >127 are reserved for prog type specific return codes */
671};
672
David Ahern610236582016-12-01 08:48:04 -0800673struct bpf_sock {
674 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -0800675 __u32 family;
676 __u32 type;
677 __u32 protocol;
David Ahern610236582016-12-01 08:48:04 -0800678};
679
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800680#define XDP_PACKET_HEADROOM 256
681
Brenden Blanco6a773a12016-07-19 12:16:47 -0700682/* User return codes for XDP prog type.
683 * A valid XDP program must return one of these defined values. All other
684 * return codes are reserved for future use. Unknown return codes will result
685 * in packet drop.
686 */
687enum xdp_action {
688 XDP_ABORTED = 0,
689 XDP_DROP,
690 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -0700691 XDP_TX,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700692};
693
694/* user accessible metadata for XDP packet hook
695 * new fields must be added to the end of this structure
696 */
697struct xdp_md {
698 __u32 data;
699 __u32 data_end;
700};
701
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700702#define BPF_TAG_SIZE 8
703
704struct bpf_prog_info {
705 __u32 type;
706 __u32 id;
707 __u8 tag[BPF_TAG_SIZE];
708 __u32 jited_prog_len;
709 __u32 xlated_prog_len;
710 __aligned_u64 jited_prog_insns;
711 __aligned_u64 xlated_prog_insns;
712} __attribute__((aligned(8)));
713
714struct bpf_map_info {
715 __u32 type;
716 __u32 id;
717 __u32 key_size;
718 __u32 value_size;
719 __u32 max_entries;
720 __u32 map_flags;
721} __attribute__((aligned(8)));
722
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -0700723#endif /* _UAPI__LINUX_BPF_H__ */