blob: e78aece036280292cac154f84349922420cd2800 [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,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070085};
86
87enum bpf_map_type {
88 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -080089 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -080090 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61a2015-05-19 16:59:03 -070091 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +000092 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -080093 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -080094 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -080095 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -070096 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -080097 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -080098 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +010099 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700100 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700101 BPF_MAP_TYPE_HASH_OF_MAPS,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700102};
103
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700104enum bpf_prog_type {
105 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800106 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700107 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100108 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee82015-03-20 15:11:11 +0100109 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700110 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700111 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700112 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100113 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800114 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100115 BPF_PROG_TYPE_LWT_IN,
116 BPF_PROG_TYPE_LWT_OUT,
117 BPF_PROG_TYPE_LWT_XMIT,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700118};
119
Daniel Mack0e336612016-11-23 16:52:25 +0100120enum bpf_attach_type {
121 BPF_CGROUP_INET_INGRESS,
122 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800123 BPF_CGROUP_INET_SOCK_CREATE,
Daniel Mack0e336612016-11-23 16:52:25 +0100124 __MAX_BPF_ATTACH_TYPE
125};
126
127#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
128
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800129/* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
130 * to the given target_fd cgroup the descendent cgroup will be able to
131 * override effective bpf program that was inherited from this cgroup
132 */
133#define BPF_F_ALLOW_OVERRIDE (1U << 0)
134
David S. Millere07b98d2017-05-10 11:38:07 -0700135/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
136 * verifier will perform strict alignment checking as if the kernel
137 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
138 * and NET_IP_ALIGN defined to 2.
139 */
140#define BPF_F_STRICT_ALIGNMENT (1U << 0)
141
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100142#define BPF_PSEUDO_MAP_FD 1
143
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800144/* flags for BPF_MAP_UPDATE_ELEM command */
145#define BPF_ANY 0 /* create new element or update existing */
146#define BPF_NOEXIST 1 /* create new element if it didn't exist */
147#define BPF_EXIST 2 /* update existing element */
148
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800149#define BPF_F_NO_PREALLOC (1U << 0)
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800150/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800151 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800152 * which can scale and perform better.
153 * Note, the LRU nodes (including free nodes) cannot be moved
154 * across different LRU lists.
155 */
156#define BPF_F_NO_COMMON_LRU (1U << 1)
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800157
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700158union bpf_attr {
159 struct { /* anonymous struct used by BPF_MAP_CREATE command */
160 __u32 map_type; /* one of enum bpf_map_type */
161 __u32 key_size; /* size of key in bytes */
162 __u32 value_size; /* size of value in bytes */
163 __u32 max_entries; /* max number of entries in a map */
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800164 __u32 map_flags; /* prealloc or not */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700165 __u32 inner_map_fd; /* fd pointing to the inner map */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700166 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700167
168 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
169 __u32 map_fd;
170 __aligned_u64 key;
171 union {
172 __aligned_u64 value;
173 __aligned_u64 next_key;
174 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800175 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700176 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700177
178 struct { /* anonymous struct used by BPF_PROG_LOAD command */
179 __u32 prog_type; /* one of enum bpf_prog_type */
180 __u32 insn_cnt;
181 __aligned_u64 insns;
182 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700183 __u32 log_level; /* verbosity level of verifier */
184 __u32 log_size; /* size of user buffer */
185 __aligned_u64 log_buf; /* user supplied buffer */
Alexei Starovoitov25415172015-03-25 12:49:20 -0700186 __u32 kern_version; /* checked when prog_type=kprobe */
David S. Millere07b98d2017-05-10 11:38:07 -0700187 __u32 prog_flags;
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700188 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100189
190 struct { /* anonymous struct used by BPF_OBJ_* commands */
191 __aligned_u64 pathname;
192 __u32 bpf_fd;
193 };
Daniel Mackf4324552016-11-23 16:52:27 +0100194
195 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
196 __u32 target_fd; /* container object to attach to */
197 __u32 attach_bpf_fd; /* eBPF program to attach */
198 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800199 __u32 attach_flags;
Daniel Mackf4324552016-11-23 16:52:27 +0100200 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700201
202 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
203 __u32 prog_fd;
204 __u32 retval;
205 __u32 data_size_in;
206 __u32 data_size_out;
207 __aligned_u64 data_in;
208 __aligned_u64 data_out;
209 __u32 repeat;
210 __u32 duration;
211 } test;
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700212} __attribute__((aligned(8)));
213
Thomas Grafebb676d2016-10-27 11:23:51 +0200214/* BPF helper function descriptions:
215 *
216 * void *bpf_map_lookup_elem(&map, &key)
217 * Return: Map value or NULL
218 *
219 * int bpf_map_update_elem(&map, &key, &value, flags)
220 * Return: 0 on success or negative error
221 *
222 * int bpf_map_delete_elem(&map, &key)
223 * Return: 0 on success or negative error
224 *
225 * int bpf_probe_read(void *dst, int size, void *src)
226 * Return: 0 on success or negative error
227 *
228 * u64 bpf_ktime_get_ns(void)
229 * Return: current ktime
230 *
231 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
232 * Return: length of buffer written or negative error
233 *
234 * u32 bpf_prandom_u32(void)
235 * Return: random value
236 *
237 * u32 bpf_raw_smp_processor_id(void)
238 * Return: SMP processor ID
239 *
240 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
241 * store bytes into packet
242 * @skb: pointer to skb
243 * @offset: offset within packet from skb->mac_header
244 * @from: pointer where to copy bytes from
245 * @len: number of bytes to store into packet
246 * @flags: bit 0 - if true, recompute skb->csum
247 * other bits - reserved
248 * Return: 0 on success or negative error
249 *
250 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
251 * recompute IP checksum
252 * @skb: pointer to skb
253 * @offset: offset within packet where IP checksum is located
254 * @from: old value of header field
255 * @to: new value of header field
256 * @flags: bits 0-3 - size of header field
257 * other bits - reserved
258 * Return: 0 on success or negative error
259 *
260 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
261 * recompute TCP/UDP checksum
262 * @skb: pointer to skb
263 * @offset: offset within packet where TCP/UDP checksum is located
264 * @from: old value of header field
265 * @to: new value of header field
266 * @flags: bits 0-3 - size of header field
267 * bit 4 - is pseudo header
268 * other bits - reserved
269 * Return: 0 on success or negative error
270 *
271 * int bpf_tail_call(ctx, prog_array_map, index)
272 * jump into another BPF program
273 * @ctx: context pointer passed to next program
274 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
275 * @index: index inside array that selects specific program to run
276 * Return: 0 on success or negative error
277 *
278 * int bpf_clone_redirect(skb, ifindex, flags)
279 * redirect to another netdev
280 * @skb: pointer to skb
281 * @ifindex: ifindex of the net device
282 * @flags: bit 0 - if set, redirect to ingress instead of egress
283 * other bits - reserved
284 * Return: 0 on success or negative error
285 *
286 * u64 bpf_get_current_pid_tgid(void)
287 * Return: current->tgid << 32 | current->pid
288 *
289 * u64 bpf_get_current_uid_gid(void)
290 * Return: current_gid << 32 | current_uid
291 *
292 * int bpf_get_current_comm(char *buf, int size_of_buf)
293 * stores current->comm into buf
294 * Return: 0 on success or negative error
295 *
296 * u32 bpf_get_cgroup_classid(skb)
297 * retrieve a proc's classid
298 * @skb: pointer to skb
299 * Return: classid if != 0
300 *
301 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
302 * Return: 0 on success or negative error
303 *
304 * int bpf_skb_vlan_pop(skb)
305 * Return: 0 on success or negative error
306 *
307 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
308 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
309 * retrieve or populate tunnel metadata
310 * @skb: pointer to skb
311 * @key: pointer to 'struct bpf_tunnel_key'
312 * @size: size of 'struct bpf_tunnel_key'
313 * @flags: room for future extensions
314 * Return: 0 on success or negative error
315 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700316 * u64 bpf_perf_event_read(map, flags)
317 * read perf event counter value
318 * @map: pointer to perf_event_array map
319 * @flags: index of event in the map or bitmask flags
320 * Return: value of perf event counter read or error code
Thomas Grafebb676d2016-10-27 11:23:51 +0200321 *
322 * int bpf_redirect(ifindex, flags)
323 * redirect to another netdev
324 * @ifindex: ifindex of the net device
325 * @flags: bit 0 - if set, redirect to ingress instead of egress
326 * other bits - reserved
327 * Return: TC_ACT_REDIRECT
328 *
329 * u32 bpf_get_route_realm(skb)
330 * retrieve a dst's tclassid
331 * @skb: pointer to skb
332 * Return: realm if != 0
333 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700334 * int bpf_perf_event_output(ctx, map, flags, data, size)
Thomas Grafebb676d2016-10-27 11:23:51 +0200335 * output perf raw sample
336 * @ctx: struct pt_regs*
337 * @map: pointer to perf_event_array map
Teng Qinb7d3ed52017-06-02 21:03:54 -0700338 * @flags: index of event in the map or bitmask flags
Thomas Grafebb676d2016-10-27 11:23:51 +0200339 * @data: data on stack to be output as raw data
340 * @size: size of data
341 * Return: 0 on success or negative error
342 *
343 * int bpf_get_stackid(ctx, map, flags)
344 * walk user or kernel stack and return id
345 * @ctx: struct pt_regs*
346 * @map: pointer to stack_trace map
347 * @flags: bits 0-7 - numer of stack frames to skip
348 * bit 8 - collect user stack instead of kernel
349 * bit 9 - compare stacks by hash only
350 * bit 10 - if two different stacks hash into the same stackid
351 * discard old
352 * other bits - reserved
353 * Return: >= 0 stackid on success or negative error
354 *
355 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
356 * calculate csum diff
357 * @from: raw from buffer
358 * @from_size: length of from buffer
359 * @to: raw to buffer
360 * @to_size: length of to buffer
361 * @seed: optional seed
362 * Return: csum result or negative error code
363 *
364 * int bpf_skb_get_tunnel_opt(skb, opt, size)
365 * retrieve tunnel options metadata
366 * @skb: pointer to skb
367 * @opt: pointer to raw tunnel option data
368 * @size: size of @opt
369 * Return: option size
370 *
371 * int bpf_skb_set_tunnel_opt(skb, opt, size)
372 * populate tunnel options metadata
373 * @skb: pointer to skb
374 * @opt: pointer to raw tunnel option data
375 * @size: size of @opt
376 * Return: 0 on success or negative error
377 *
378 * int bpf_skb_change_proto(skb, proto, flags)
379 * Change protocol of the skb. Currently supported is v4 -> v6,
380 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
381 * program is expected to fill the new headers via skb_store_bytes
382 * and lX_csum_replace.
383 * @skb: pointer to skb
384 * @proto: new skb->protocol type
385 * @flags: reserved
386 * Return: 0 on success or negative error
387 *
388 * int bpf_skb_change_type(skb, type)
389 * Change packet type of skb.
390 * @skb: pointer to skb
391 * @type: new skb->pkt_type type
392 * Return: 0 on success or negative error
393 *
394 * int bpf_skb_under_cgroup(skb, map, index)
395 * Check cgroup2 membership of skb
396 * @skb: pointer to skb
397 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
398 * @index: index of the cgroup in the bpf_map
399 * Return:
400 * == 0 skb failed the cgroup2 descendant test
401 * == 1 skb succeeded the cgroup2 descendant test
402 * < 0 error
403 *
404 * u32 bpf_get_hash_recalc(skb)
405 * Retrieve and possibly recalculate skb->hash.
406 * @skb: pointer to skb
407 * Return: hash
408 *
409 * u64 bpf_get_current_task(void)
410 * Returns current task_struct
411 * Return: current
412 *
413 * int bpf_probe_write_user(void *dst, void *src, int len)
414 * safely attempt to write to a location
415 * @dst: destination address in userspace
416 * @src: source address on stack
417 * @len: number of bytes to copy
418 * Return: 0 on success or negative error
419 *
420 * int bpf_current_task_under_cgroup(map, index)
421 * Check cgroup2 membership of current task
422 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
423 * @index: index of the cgroup in the bpf_map
424 * Return:
425 * == 0 current failed the cgroup2 descendant test
426 * == 1 current succeeded the cgroup2 descendant test
427 * < 0 error
428 *
429 * int bpf_skb_change_tail(skb, len, flags)
430 * The helper will resize the skb to the given new size, to be used f.e.
431 * with control messages.
432 * @skb: pointer to skb
433 * @len: new skb length
434 * @flags: reserved
435 * Return: 0 on success or negative error
436 *
437 * int bpf_skb_pull_data(skb, len)
438 * The helper will pull in non-linear data in case the skb is non-linear
439 * and not all of len are part of the linear section. Only needed for
440 * read/write with direct packet access.
441 * @skb: pointer to skb
442 * @len: len to make read/writeable
443 * Return: 0 on success or negative error
444 *
445 * s64 bpf_csum_update(skb, csum)
446 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
447 * @skb: pointer to skb
448 * @csum: csum to add
449 * Return: csum on success or negative error
450 *
451 * void bpf_set_hash_invalid(skb)
452 * Invalidate current skb->hash.
453 * @skb: pointer to skb
454 *
455 * int bpf_get_numa_node_id()
456 * Return: Id of current NUMA node.
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100457 *
458 * int bpf_skb_change_head()
459 * Grows headroom of skb and adjusts MAC header offset accordingly.
460 * Will extends/reallocae as required automatically.
461 * May change skb data pointer and will thus invalidate any check
462 * performed for direct packet access.
463 * @skb: pointer to skb
464 * @len: length of header to be pushed in front
465 * @flags: Flags (unused for now)
466 * Return: 0 on success or negative error
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800467 *
468 * int bpf_xdp_adjust_head(xdp_md, delta)
469 * Adjust the xdp_md.data by delta
470 * @xdp_md: pointer to xdp_md
471 * @delta: An positive/negative integer to be added to xdp_md.data
472 * Return: 0 on success or negative on error
Gianluca Borelloa5e8c072017-01-18 17:55:49 +0000473 *
474 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
475 * Copy a NUL terminated string from unsafe address. In case the string
476 * length is smaller than size, the target is not padded with further NUL
477 * bytes. In case the string length is larger than size, just count-1
478 * bytes are copied and the last byte is set to NUL.
479 * @dst: destination address
480 * @size: maximum number of bytes to copy, including the trailing NUL
481 * @unsafe_ptr: unsafe address
482 * Return:
483 * > 0 length of the string including the trailing NUL on success
484 * < 0 error
Chenbo Feng91b82702017-03-22 17:27:34 -0700485 *
Alexander Alemayhu3c60a532017-04-08 22:08:10 +0200486 * u64 bpf_get_socket_cookie(skb)
Chenbo Feng91b82702017-03-22 17:27:34 -0700487 * Get the cookie for the socket stored inside sk_buff.
488 * @skb: pointer to skb
489 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
490 * field is missing inside sk_buff
Chenbo Feng6acc5c22017-03-22 17:27:35 -0700491 *
492 * u32 bpf_get_socket_uid(skb)
493 * Get the owner uid of the socket stored inside sk_buff.
494 * @skb: pointer to skb
Chenbo Feng5d4e3442017-04-26 16:41:23 -0700495 * Return: uid of the socket owner on success or overflowuid if failed.
Thomas Grafebb676d2016-10-27 11:23:51 +0200496 */
497#define __BPF_FUNC_MAPPER(FN) \
498 FN(unspec), \
499 FN(map_lookup_elem), \
500 FN(map_update_elem), \
501 FN(map_delete_elem), \
502 FN(probe_read), \
503 FN(ktime_get_ns), \
504 FN(trace_printk), \
505 FN(get_prandom_u32), \
506 FN(get_smp_processor_id), \
507 FN(skb_store_bytes), \
508 FN(l3_csum_replace), \
509 FN(l4_csum_replace), \
510 FN(tail_call), \
511 FN(clone_redirect), \
512 FN(get_current_pid_tgid), \
513 FN(get_current_uid_gid), \
514 FN(get_current_comm), \
515 FN(get_cgroup_classid), \
516 FN(skb_vlan_push), \
517 FN(skb_vlan_pop), \
518 FN(skb_get_tunnel_key), \
519 FN(skb_set_tunnel_key), \
520 FN(perf_event_read), \
521 FN(redirect), \
522 FN(get_route_realm), \
523 FN(perf_event_output), \
524 FN(skb_load_bytes), \
525 FN(get_stackid), \
526 FN(csum_diff), \
527 FN(skb_get_tunnel_opt), \
528 FN(skb_set_tunnel_opt), \
529 FN(skb_change_proto), \
530 FN(skb_change_type), \
531 FN(skb_under_cgroup), \
532 FN(get_hash_recalc), \
533 FN(get_current_task), \
534 FN(probe_write_user), \
535 FN(current_task_under_cgroup), \
536 FN(skb_change_tail), \
537 FN(skb_pull_data), \
538 FN(csum_update), \
539 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100540 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800541 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +0000542 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -0700543 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -0700544 FN(get_socket_cookie), \
545 FN(get_socket_uid),
Thomas Grafebb676d2016-10-27 11:23:51 +0200546
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700547/* integer value in 'imm' field of BPF_CALL instruction selects which helper
548 * function eBPF program intends to call
549 */
Thomas Grafebb676d2016-10-27 11:23:51 +0200550#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700551enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +0200552 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700553 __BPF_FUNC_MAX_ID,
554};
Thomas Grafebb676d2016-10-27 11:23:51 +0200555#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700556
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100557/* All flags used by eBPF helper functions, placed here. */
558
559/* BPF_FUNC_skb_store_bytes flags. */
560#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
Daniel Borkmann8afd54c2016-03-04 15:15:03 +0100561#define BPF_F_INVALIDATE_HASH (1ULL << 1)
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100562
563/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
564 * First 4 bits are for passing the header field size.
565 */
566#define BPF_F_HDR_FIELD_MASK 0xfULL
567
568/* BPF_FUNC_l4_csum_replace flags. */
569#define BPF_F_PSEUDO_HDR (1ULL << 4)
Daniel Borkmann2f729592016-02-19 23:05:26 +0100570#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
Daniel Borkmannd1b662a2017-01-24 01:06:28 +0100571#define BPF_F_MARK_ENFORCE (1ULL << 6)
Daniel Borkmann781c53b2016-01-11 01:16:38 +0100572
573/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
574#define BPF_F_INGRESS (1ULL << 0)
575
Daniel Borkmannc6c33452016-01-11 01:16:39 +0100576/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
577#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
578
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800579/* BPF_FUNC_get_stackid flags. */
580#define BPF_F_SKIP_FIELD_MASK 0xffULL
581#define BPF_F_USER_STACK (1ULL << 8)
582#define BPF_F_FAST_STACK_CMP (1ULL << 9)
583#define BPF_F_REUSE_STACKID (1ULL << 10)
584
Daniel Borkmann2da897e2016-02-23 02:05:26 +0100585/* BPF_FUNC_skb_set_tunnel_key flags. */
586#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
Daniel Borkmann22080872016-03-04 15:15:05 +0100587#define BPF_F_DONT_FRAGMENT (1ULL << 2)
Daniel Borkmann2da897e2016-02-23 02:05:26 +0100588
Daniel Borkmann6816a7f2016-06-28 12:18:25 +0200589/* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
Daniel Borkmann1e337592016-04-18 21:01:23 +0200590#define BPF_F_INDEX_MASK 0xffffffffULL
591#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
Daniel Borkmann555c8a82016-07-14 18:08:05 +0200592/* BPF_FUNC_perf_event_output for sk_buff input context. */
593#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
Daniel Borkmann1e337592016-04-18 21:01:23 +0200594
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -0700595/* user accessible mirror of in-kernel sk_buff.
596 * new fields can only be added to the end of this structure
597 */
598struct __sk_buff {
599 __u32 len;
600 __u32 pkt_type;
601 __u32 mark;
602 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -0700603 __u32 protocol;
604 __u32 vlan_present;
605 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +0100606 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +0200607 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -0700608 __u32 ingress_ifindex;
609 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -0700610 __u32 tc_index;
611 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +0200612 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -0700613 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -0700614 __u32 data;
615 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +0200616 __u32 napi_id;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -0700617};
618
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -0700619struct bpf_tunnel_key {
620 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +0100621 union {
622 __u32 remote_ipv4;
623 __u32 remote_ipv6[4];
624 };
625 __u8 tunnel_tos;
626 __u8 tunnel_ttl;
Daniel Borkmannc0e760c2016-03-30 00:02:00 +0200627 __u16 tunnel_ext;
Daniel Borkmann4018ab12016-03-09 03:00:05 +0100628 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -0700629};
630
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100631/* Generic BPF return codes which all BPF program types may support.
632 * The values are binary compatible with their TC_ACT_* counter-part to
633 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
634 * programs.
635 *
636 * XDP is handled seprately, see XDP_*.
637 */
638enum bpf_ret_code {
639 BPF_OK = 0,
640 /* 1 reserved */
641 BPF_DROP = 2,
642 /* 3-6 reserved */
643 BPF_REDIRECT = 7,
644 /* >127 are reserved for prog type specific return codes */
645};
646
David Ahern610236582016-12-01 08:48:04 -0800647struct bpf_sock {
648 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -0800649 __u32 family;
650 __u32 type;
651 __u32 protocol;
David Ahern610236582016-12-01 08:48:04 -0800652};
653
Martin KaFai Lau17bedab2016-12-07 15:53:11 -0800654#define XDP_PACKET_HEADROOM 256
655
Brenden Blanco6a773a12016-07-19 12:16:47 -0700656/* User return codes for XDP prog type.
657 * A valid XDP program must return one of these defined values. All other
658 * return codes are reserved for future use. Unknown return codes will result
659 * in packet drop.
660 */
661enum xdp_action {
662 XDP_ABORTED = 0,
663 XDP_DROP,
664 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -0700665 XDP_TX,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700666};
667
668/* user accessible metadata for XDP packet hook
669 * new fields must be added to the end of this structure
670 */
671struct xdp_md {
672 __u32 data;
673 __u32 data_end;
674};
675
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -0700676#endif /* _UAPI__LINUX_BPF_H__ */