blob: 461811e5714092f17619902593b9ac47e8572a6d [file] [log] [blame]
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -03001/* 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>
11#include <linux/bpf_common.h>
12
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
Daniel Borkmann92b31a92017-08-10 01:39:55 +020033/* jmp encodings */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -030034#define BPF_JNE 0x50 /* jump != */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020035#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
36#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -030037#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
38#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020039#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
40#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -030041#define BPF_CALL 0x80 /* function call */
42#define BPF_EXIT 0x90 /* function return */
43
44/* Register numbers */
45enum {
46 BPF_REG_0 = 0,
47 BPF_REG_1,
48 BPF_REG_2,
49 BPF_REG_3,
50 BPF_REG_4,
51 BPF_REG_5,
52 BPF_REG_6,
53 BPF_REG_7,
54 BPF_REG_8,
55 BPF_REG_9,
56 BPF_REG_10,
57 __MAX_BPF_REG,
58};
59
60/* BPF has 10 general purpose 64-bit registers and stack frame. */
61#define MAX_BPF_REG __MAX_BPF_REG
62
63struct bpf_insn {
64 __u8 code; /* opcode */
65 __u8 dst_reg:4; /* dest register */
66 __u8 src_reg:4; /* source register */
67 __s16 off; /* signed offset */
68 __s32 imm; /* signed immediate constant */
69};
70
Mickaël Salaün9a738262017-02-10 00:21:35 +010071/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
72struct bpf_lpm_trie_key {
73 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
74 __u8 data[0]; /* Arbitrary size */
75};
76
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -030077/* BPF syscall commands, see bpf(2) man-page for details. */
78enum bpf_cmd {
79 BPF_MAP_CREATE,
80 BPF_MAP_LOOKUP_ELEM,
81 BPF_MAP_UPDATE_ELEM,
82 BPF_MAP_DELETE_ELEM,
83 BPF_MAP_GET_NEXT_KEY,
84 BPF_PROG_LOAD,
85 BPF_OBJ_PIN,
86 BPF_OBJ_GET,
Joe Stringer0cb34dc2016-12-08 18:46:14 -080087 BPF_PROG_ATTACH,
88 BPF_PROG_DETACH,
Alexei Starovoitov30848872017-03-30 21:45:39 -070089 BPF_PROG_TEST_RUN,
Martin KaFai Lau95b9afd2017-06-05 12:15:53 -070090 BPF_PROG_GET_NEXT_ID,
91 BPF_MAP_GET_NEXT_ID,
92 BPF_PROG_GET_FD_BY_ID,
93 BPF_MAP_GET_FD_BY_ID,
94 BPF_OBJ_GET_INFO_BY_FD,
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -030095};
96
97enum bpf_map_type {
98 BPF_MAP_TYPE_UNSPEC,
99 BPF_MAP_TYPE_HASH,
100 BPF_MAP_TYPE_ARRAY,
101 BPF_MAP_TYPE_PROG_ARRAY,
102 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
103 BPF_MAP_TYPE_PERCPU_HASH,
104 BPF_MAP_TYPE_PERCPU_ARRAY,
105 BPF_MAP_TYPE_STACK_TRACE,
Arnaldo Carvalho de Melo791cceb2016-08-09 11:48:07 -0300106 BPF_MAP_TYPE_CGROUP_ARRAY,
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800107 BPF_MAP_TYPE_LRU_HASH,
108 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Mickaël Salaün9a738262017-02-10 00:21:35 +0100109 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Laufb30d4b2017-03-22 10:00:35 -0700110 BPF_MAP_TYPE_ARRAY_OF_MAPS,
111 BPF_MAP_TYPE_HASH_OF_MAPS,
John Fastabend81f6bf82017-07-26 17:32:07 -0700112 BPF_MAP_TYPE_DEVMAP,
John Fastabend69e8cc12017-08-15 22:33:32 -0700113 BPF_MAP_TYPE_SOCKMAP,
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300114};
115
116enum bpf_prog_type {
117 BPF_PROG_TYPE_UNSPEC,
118 BPF_PROG_TYPE_SOCKET_FILTER,
119 BPF_PROG_TYPE_KPROBE,
120 BPF_PROG_TYPE_SCHED_CLS,
121 BPF_PROG_TYPE_SCHED_ACT,
122 BPF_PROG_TYPE_TRACEPOINT,
Arnaldo Carvalho de Melo791cceb2016-08-09 11:48:07 -0300123 BPF_PROG_TYPE_XDP,
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800124 BPF_PROG_TYPE_PERF_EVENT,
125 BPF_PROG_TYPE_CGROUP_SKB,
126 BPF_PROG_TYPE_CGROUP_SOCK,
127 BPF_PROG_TYPE_LWT_IN,
128 BPF_PROG_TYPE_LWT_OUT,
129 BPF_PROG_TYPE_LWT_XMIT,
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700130 BPF_PROG_TYPE_SOCK_OPS,
John Fastabend69e8cc12017-08-15 22:33:32 -0700131 BPF_PROG_TYPE_SK_SKB,
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300132};
133
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800134enum bpf_attach_type {
135 BPF_CGROUP_INET_INGRESS,
136 BPF_CGROUP_INET_EGRESS,
137 BPF_CGROUP_INET_SOCK_CREATE,
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700138 BPF_CGROUP_SOCK_OPS,
John Fastabend464bc0f2017-08-28 07:10:04 -0700139 BPF_SK_SKB_STREAM_PARSER,
140 BPF_SK_SKB_STREAM_VERDICT,
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800141 __MAX_BPF_ATTACH_TYPE
142};
143
144#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
145
John Fastabend69e8cc12017-08-15 22:33:32 -0700146enum bpf_sockmap_flags {
147 BPF_SOCKMAP_UNSPEC,
148 BPF_SOCKMAP_STRPARSER,
149 __MAX_BPF_SOCKMAP_FLAG
150};
151
Stephen Rothwell5463b3d2017-02-14 08:22:20 +1100152/* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
153 * to the given target_fd cgroup the descendent cgroup will be able to
154 * override effective bpf program that was inherited from this cgroup
155 */
156#define BPF_F_ALLOW_OVERRIDE (1U << 0)
157
David S. Millere07b98d2017-05-10 11:38:07 -0700158/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
159 * verifier will perform strict alignment checking as if the kernel
160 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
161 * and NET_IP_ALIGN defined to 2.
162 */
163#define BPF_F_STRICT_ALIGNMENT (1U << 0)
164
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300165#define BPF_PSEUDO_MAP_FD 1
166
167/* flags for BPF_MAP_UPDATE_ELEM command */
168#define BPF_ANY 0 /* create new element or update existing */
169#define BPF_NOEXIST 1 /* create new element if it didn't exist */
170#define BPF_EXIST 2 /* update existing element */
171
Martin KaFai Lauad17d0e2017-08-18 11:28:01 -0700172/* flags for BPF_MAP_CREATE command */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300173#define BPF_F_NO_PREALLOC (1U << 0)
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800174/* Instead of having one common LRU list in the
175 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
176 * which can scale and perform better.
177 * Note, the LRU nodes (including free nodes) cannot be moved
178 * across different LRU lists.
179 */
180#define BPF_F_NO_COMMON_LRU (1U << 1)
Martin KaFai Lauad17d0e2017-08-18 11:28:01 -0700181/* Specify numa node during map creation */
182#define BPF_F_NUMA_NODE (1U << 2)
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300183
184union bpf_attr {
185 struct { /* anonymous struct used by BPF_MAP_CREATE command */
186 __u32 map_type; /* one of enum bpf_map_type */
187 __u32 key_size; /* size of key in bytes */
188 __u32 value_size; /* size of value in bytes */
189 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lauad17d0e2017-08-18 11:28:01 -0700190 __u32 map_flags; /* BPF_MAP_CREATE related
191 * flags defined above.
192 */
Martin KaFai Laufb30d4b2017-03-22 10:00:35 -0700193 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lauad17d0e2017-08-18 11:28:01 -0700194 __u32 numa_node; /* numa node (effective only if
195 * BPF_F_NUMA_NODE is set).
196 */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300197 };
198
199 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
200 __u32 map_fd;
201 __aligned_u64 key;
202 union {
203 __aligned_u64 value;
204 __aligned_u64 next_key;
205 };
206 __u64 flags;
207 };
208
209 struct { /* anonymous struct used by BPF_PROG_LOAD command */
210 __u32 prog_type; /* one of enum bpf_prog_type */
211 __u32 insn_cnt;
212 __aligned_u64 insns;
213 __aligned_u64 license;
214 __u32 log_level; /* verbosity level of verifier */
215 __u32 log_size; /* size of user buffer */
216 __aligned_u64 log_buf; /* user supplied buffer */
217 __u32 kern_version; /* checked when prog_type=kprobe */
David S. Millere07b98d2017-05-10 11:38:07 -0700218 __u32 prog_flags;
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300219 };
220
221 struct { /* anonymous struct used by BPF_OBJ_* commands */
222 __aligned_u64 pathname;
223 __u32 bpf_fd;
224 };
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800225
226 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
227 __u32 target_fd; /* container object to attach to */
228 __u32 attach_bpf_fd; /* eBPF program to attach */
229 __u32 attach_type;
Stephen Rothwell5463b3d2017-02-14 08:22:20 +1100230 __u32 attach_flags;
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800231 };
Alexei Starovoitov30848872017-03-30 21:45:39 -0700232
233 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
234 __u32 prog_fd;
235 __u32 retval;
236 __u32 data_size_in;
237 __u32 data_size_out;
238 __aligned_u64 data_in;
239 __aligned_u64 data_out;
240 __u32 repeat;
241 __u32 duration;
242 } test;
Martin KaFai Lau95b9afd2017-06-05 12:15:53 -0700243
244 struct { /* anonymous struct used by BPF_*_GET_*_ID */
245 union {
246 __u32 start_id;
247 __u32 prog_id;
248 __u32 map_id;
249 };
250 __u32 next_id;
251 };
252
253 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
254 __u32 bpf_fd;
255 __u32 info_len;
256 __aligned_u64 info;
257 } info;
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300258} __attribute__((aligned(8)));
259
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800260/* BPF helper function descriptions:
261 *
262 * void *bpf_map_lookup_elem(&map, &key)
263 * Return: Map value or NULL
264 *
265 * int bpf_map_update_elem(&map, &key, &value, flags)
266 * Return: 0 on success or negative error
267 *
268 * int bpf_map_delete_elem(&map, &key)
269 * Return: 0 on success or negative error
270 *
271 * int bpf_probe_read(void *dst, int size, void *src)
272 * Return: 0 on success or negative error
273 *
274 * u64 bpf_ktime_get_ns(void)
275 * Return: current ktime
276 *
277 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
278 * Return: length of buffer written or negative error
279 *
280 * u32 bpf_prandom_u32(void)
281 * Return: random value
282 *
283 * u32 bpf_raw_smp_processor_id(void)
284 * Return: SMP processor ID
285 *
286 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
287 * store bytes into packet
288 * @skb: pointer to skb
289 * @offset: offset within packet from skb->mac_header
290 * @from: pointer where to copy bytes from
291 * @len: number of bytes to store into packet
292 * @flags: bit 0 - if true, recompute skb->csum
293 * other bits - reserved
294 * Return: 0 on success or negative error
295 *
296 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
297 * recompute IP checksum
298 * @skb: pointer to skb
299 * @offset: offset within packet where IP checksum is located
300 * @from: old value of header field
301 * @to: new value of header field
302 * @flags: bits 0-3 - size of header field
303 * other bits - reserved
304 * Return: 0 on success or negative error
305 *
306 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
307 * recompute TCP/UDP checksum
308 * @skb: pointer to skb
309 * @offset: offset within packet where TCP/UDP checksum is located
310 * @from: old value of header field
311 * @to: new value of header field
312 * @flags: bits 0-3 - size of header field
313 * bit 4 - is pseudo header
314 * other bits - reserved
315 * Return: 0 on success or negative error
316 *
317 * int bpf_tail_call(ctx, prog_array_map, index)
318 * jump into another BPF program
319 * @ctx: context pointer passed to next program
320 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
321 * @index: index inside array that selects specific program to run
322 * Return: 0 on success or negative error
323 *
324 * int bpf_clone_redirect(skb, ifindex, flags)
325 * redirect to another netdev
326 * @skb: pointer to skb
327 * @ifindex: ifindex of the net device
328 * @flags: bit 0 - if set, redirect to ingress instead of egress
329 * other bits - reserved
330 * Return: 0 on success or negative error
331 *
332 * u64 bpf_get_current_pid_tgid(void)
333 * Return: current->tgid << 32 | current->pid
334 *
335 * u64 bpf_get_current_uid_gid(void)
336 * Return: current_gid << 32 | current_uid
337 *
338 * int bpf_get_current_comm(char *buf, int size_of_buf)
339 * stores current->comm into buf
340 * Return: 0 on success or negative error
341 *
342 * u32 bpf_get_cgroup_classid(skb)
343 * retrieve a proc's classid
344 * @skb: pointer to skb
345 * Return: classid if != 0
346 *
347 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
348 * Return: 0 on success or negative error
349 *
350 * int bpf_skb_vlan_pop(skb)
351 * Return: 0 on success or negative error
352 *
353 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
354 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
355 * retrieve or populate tunnel metadata
356 * @skb: pointer to skb
357 * @key: pointer to 'struct bpf_tunnel_key'
358 * @size: size of 'struct bpf_tunnel_key'
359 * @flags: room for future extensions
360 * Return: 0 on success or negative error
361 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700362 * u64 bpf_perf_event_read(map, flags)
363 * read perf event counter value
364 * @map: pointer to perf_event_array map
365 * @flags: index of event in the map or bitmask flags
366 * Return: value of perf event counter read or error code
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800367 *
368 * int bpf_redirect(ifindex, flags)
369 * redirect to another netdev
370 * @ifindex: ifindex of the net device
371 * @flags: bit 0 - if set, redirect to ingress instead of egress
372 * other bits - reserved
373 * Return: TC_ACT_REDIRECT
374 *
375 * u32 bpf_get_route_realm(skb)
376 * retrieve a dst's tclassid
377 * @skb: pointer to skb
378 * Return: realm if != 0
379 *
Teng Qinb7d3ed52017-06-02 21:03:54 -0700380 * int bpf_perf_event_output(ctx, map, flags, data, size)
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800381 * output perf raw sample
382 * @ctx: struct pt_regs*
383 * @map: pointer to perf_event_array map
Teng Qinb7d3ed52017-06-02 21:03:54 -0700384 * @flags: index of event in the map or bitmask flags
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800385 * @data: data on stack to be output as raw data
386 * @size: size of data
387 * Return: 0 on success or negative error
388 *
389 * int bpf_get_stackid(ctx, map, flags)
390 * walk user or kernel stack and return id
391 * @ctx: struct pt_regs*
392 * @map: pointer to stack_trace map
393 * @flags: bits 0-7 - numer of stack frames to skip
394 * bit 8 - collect user stack instead of kernel
395 * bit 9 - compare stacks by hash only
396 * bit 10 - if two different stacks hash into the same stackid
397 * discard old
398 * other bits - reserved
399 * Return: >= 0 stackid on success or negative error
400 *
401 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
402 * calculate csum diff
403 * @from: raw from buffer
404 * @from_size: length of from buffer
405 * @to: raw to buffer
406 * @to_size: length of to buffer
407 * @seed: optional seed
408 * Return: csum result or negative error code
409 *
410 * int bpf_skb_get_tunnel_opt(skb, opt, size)
411 * retrieve tunnel options metadata
412 * @skb: pointer to skb
413 * @opt: pointer to raw tunnel option data
414 * @size: size of @opt
415 * Return: option size
416 *
417 * int bpf_skb_set_tunnel_opt(skb, opt, size)
418 * populate tunnel options metadata
419 * @skb: pointer to skb
420 * @opt: pointer to raw tunnel option data
421 * @size: size of @opt
422 * Return: 0 on success or negative error
423 *
424 * int bpf_skb_change_proto(skb, proto, flags)
425 * Change protocol of the skb. Currently supported is v4 -> v6,
426 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
427 * program is expected to fill the new headers via skb_store_bytes
428 * and lX_csum_replace.
429 * @skb: pointer to skb
430 * @proto: new skb->protocol type
431 * @flags: reserved
432 * Return: 0 on success or negative error
433 *
434 * int bpf_skb_change_type(skb, type)
435 * Change packet type of skb.
436 * @skb: pointer to skb
437 * @type: new skb->pkt_type type
438 * Return: 0 on success or negative error
439 *
440 * int bpf_skb_under_cgroup(skb, map, index)
441 * Check cgroup2 membership of skb
442 * @skb: pointer to skb
443 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
444 * @index: index of the cgroup in the bpf_map
445 * Return:
446 * == 0 skb failed the cgroup2 descendant test
447 * == 1 skb succeeded the cgroup2 descendant test
448 * < 0 error
449 *
450 * u32 bpf_get_hash_recalc(skb)
451 * Retrieve and possibly recalculate skb->hash.
452 * @skb: pointer to skb
453 * Return: hash
454 *
455 * u64 bpf_get_current_task(void)
456 * Returns current task_struct
457 * Return: current
458 *
459 * int bpf_probe_write_user(void *dst, void *src, int len)
460 * safely attempt to write to a location
461 * @dst: destination address in userspace
462 * @src: source address on stack
463 * @len: number of bytes to copy
464 * Return: 0 on success or negative error
465 *
466 * int bpf_current_task_under_cgroup(map, index)
467 * Check cgroup2 membership of current task
468 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
469 * @index: index of the cgroup in the bpf_map
470 * Return:
471 * == 0 current failed the cgroup2 descendant test
472 * == 1 current succeeded the cgroup2 descendant test
473 * < 0 error
474 *
475 * int bpf_skb_change_tail(skb, len, flags)
476 * The helper will resize the skb to the given new size, to be used f.e.
477 * with control messages.
478 * @skb: pointer to skb
479 * @len: new skb length
480 * @flags: reserved
481 * Return: 0 on success or negative error
482 *
483 * int bpf_skb_pull_data(skb, len)
484 * The helper will pull in non-linear data in case the skb is non-linear
485 * and not all of len are part of the linear section. Only needed for
486 * read/write with direct packet access.
487 * @skb: pointer to skb
488 * @len: len to make read/writeable
489 * Return: 0 on success or negative error
490 *
491 * s64 bpf_csum_update(skb, csum)
492 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
493 * @skb: pointer to skb
494 * @csum: csum to add
495 * Return: csum on success or negative error
496 *
497 * void bpf_set_hash_invalid(skb)
498 * Invalidate current skb->hash.
499 * @skb: pointer to skb
500 *
501 * int bpf_get_numa_node_id()
502 * Return: Id of current NUMA node.
503 *
504 * int bpf_skb_change_head()
505 * Grows headroom of skb and adjusts MAC header offset accordingly.
506 * Will extends/reallocae as required automatically.
507 * May change skb data pointer and will thus invalidate any check
508 * performed for direct packet access.
509 * @skb: pointer to skb
510 * @len: length of header to be pushed in front
511 * @flags: Flags (unused for now)
512 * Return: 0 on success or negative error
513 *
514 * int bpf_xdp_adjust_head(xdp_md, delta)
515 * Adjust the xdp_md.data by delta
516 * @xdp_md: pointer to xdp_md
517 * @delta: An positive/negative integer to be added to xdp_md.data
518 * Return: 0 on success or negative on error
Mickaël Salaün9a738262017-02-10 00:21:35 +0100519 *
520 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
521 * Copy a NUL terminated string from unsafe address. In case the string
522 * length is smaller than size, the target is not padded with further NUL
523 * bytes. In case the string length is larger than size, just count-1
524 * bytes are copied and the last byte is set to NUL.
525 * @dst: destination address
526 * @size: maximum number of bytes to copy, including the trailing NUL
527 * @unsafe_ptr: unsafe address
528 * Return:
529 * > 0 length of the string including the trailing NUL on success
530 * < 0 error
Alexei Starovoitov30848872017-03-30 21:45:39 -0700531 *
Alexander Alemayhu3c60a532017-04-08 22:08:10 +0200532 * u64 bpf_get_socket_cookie(skb)
Alexei Starovoitov30848872017-03-30 21:45:39 -0700533 * Get the cookie for the socket stored inside sk_buff.
534 * @skb: pointer to skb
535 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
536 * field is missing inside sk_buff
537 *
538 * u32 bpf_get_socket_uid(skb)
539 * Get the owner uid of the socket stored inside sk_buff.
540 * @skb: pointer to skb
David S. Millere07b98d2017-05-10 11:38:07 -0700541 * Return: uid of the socket owner on success or overflowuid if failed.
Daniel Borkmannded092c2017-06-11 00:50:47 +0200542 *
543 * u32 bpf_set_hash(skb, hash)
544 * Set full skb->hash.
545 * @skb: pointer to skb
546 * @hash: hash to set
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700547 *
548 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
549 * Calls setsockopt. Not all opts are available, only those with
550 * integer optvals plus TCP_CONGESTION.
551 * Supported levels: SOL_SOCKET and IPROTO_TCP
552 * @bpf_socket: pointer to bpf_socket
553 * @level: SOL_SOCKET or IPROTO_TCP
554 * @optname: option name
555 * @optval: pointer to option value
556 * @optlen: length of optval in byes
557 * Return: 0 or negative error
Daniel Borkmann2be7e212017-07-02 02:13:26 +0200558 *
559 * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
560 * Grow or shrink room in sk_buff.
561 * @skb: pointer to skb
562 * @len_diff: (signed) amount of room to grow/shrink
563 * @mode: operation mode (enum bpf_adj_room_mode)
564 * @flags: reserved for future use
565 * Return: 0 on success or negative error code
John Fastabend69e8cc12017-08-15 22:33:32 -0700566 *
567 * int bpf_sk_redirect_map(map, key, flags)
568 * Redirect skb to a sock in map using key as a lookup key for the
569 * sock in map.
570 * @map: pointer to sockmap
571 * @key: key to lookup sock in map
572 * @flags: reserved for future use
573 * Return: SK_REDIRECT
574 *
John Fastabend464bc0f2017-08-28 07:10:04 -0700575 * int bpf_sock_map_update(skops, map, key, flags)
John Fastabend69e8cc12017-08-15 22:33:32 -0700576 * @skops: pointer to bpf_sock_ops
577 * @map: pointer to sockmap to update
578 * @key: key to insert/update sock in map
579 * @flags: same flags as map update elem
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800580 */
581#define __BPF_FUNC_MAPPER(FN) \
582 FN(unspec), \
583 FN(map_lookup_elem), \
584 FN(map_update_elem), \
585 FN(map_delete_elem), \
586 FN(probe_read), \
587 FN(ktime_get_ns), \
588 FN(trace_printk), \
589 FN(get_prandom_u32), \
590 FN(get_smp_processor_id), \
591 FN(skb_store_bytes), \
592 FN(l3_csum_replace), \
593 FN(l4_csum_replace), \
594 FN(tail_call), \
595 FN(clone_redirect), \
596 FN(get_current_pid_tgid), \
597 FN(get_current_uid_gid), \
598 FN(get_current_comm), \
599 FN(get_cgroup_classid), \
600 FN(skb_vlan_push), \
601 FN(skb_vlan_pop), \
602 FN(skb_get_tunnel_key), \
603 FN(skb_set_tunnel_key), \
604 FN(perf_event_read), \
605 FN(redirect), \
606 FN(get_route_realm), \
607 FN(perf_event_output), \
608 FN(skb_load_bytes), \
609 FN(get_stackid), \
610 FN(csum_diff), \
611 FN(skb_get_tunnel_opt), \
612 FN(skb_set_tunnel_opt), \
613 FN(skb_change_proto), \
614 FN(skb_change_type), \
615 FN(skb_under_cgroup), \
616 FN(get_hash_recalc), \
617 FN(get_current_task), \
618 FN(probe_write_user), \
619 FN(current_task_under_cgroup), \
620 FN(skb_change_tail), \
621 FN(skb_pull_data), \
622 FN(csum_update), \
623 FN(set_hash_invalid), \
624 FN(get_numa_node_id), \
625 FN(skb_change_head), \
Mickaël Salaün9a738262017-02-10 00:21:35 +0100626 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -0700627 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -0700628 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +0200629 FN(get_socket_uid), \
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700630 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +0200631 FN(setsockopt), \
William Tu996139e2017-08-07 13:14:42 -0700632 FN(skb_adjust_room), \
John Fastabend69e8cc12017-08-15 22:33:32 -0700633 FN(redirect_map), \
634 FN(sk_redirect_map), \
635 FN(sock_map_update),
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800636
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300637/* integer value in 'imm' field of BPF_CALL instruction selects which helper
638 * function eBPF program intends to call
639 */
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800640#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300641enum bpf_func_id {
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800642 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300643 __BPF_FUNC_MAX_ID,
644};
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800645#undef __BPF_ENUM_FN
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300646
647/* All flags used by eBPF helper functions, placed here. */
648
649/* BPF_FUNC_skb_store_bytes flags. */
650#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
651#define BPF_F_INVALIDATE_HASH (1ULL << 1)
652
653/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
654 * First 4 bits are for passing the header field size.
655 */
656#define BPF_F_HDR_FIELD_MASK 0xfULL
657
658/* BPF_FUNC_l4_csum_replace flags. */
659#define BPF_F_PSEUDO_HDR (1ULL << 4)
660#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
Mickaël Salaün9a738262017-02-10 00:21:35 +0100661#define BPF_F_MARK_ENFORCE (1ULL << 6)
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300662
663/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
664#define BPF_F_INGRESS (1ULL << 0)
665
666/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
667#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
668
669/* BPF_FUNC_get_stackid flags. */
670#define BPF_F_SKIP_FIELD_MASK 0xffULL
671#define BPF_F_USER_STACK (1ULL << 8)
672#define BPF_F_FAST_STACK_CMP (1ULL << 9)
673#define BPF_F_REUSE_STACKID (1ULL << 10)
674
675/* BPF_FUNC_skb_set_tunnel_key flags. */
676#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
677#define BPF_F_DONT_FRAGMENT (1ULL << 2)
678
Arnaldo Carvalho de Melo791cceb2016-08-09 11:48:07 -0300679/* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300680#define BPF_F_INDEX_MASK 0xffffffffULL
681#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
Arnaldo Carvalho de Melo791cceb2016-08-09 11:48:07 -0300682/* BPF_FUNC_perf_event_output for sk_buff input context. */
683#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300684
Daniel Borkmann2be7e212017-07-02 02:13:26 +0200685/* Mode for BPF_FUNC_skb_adjust_room helper. */
686enum bpf_adj_room_mode {
Arnaldo Carvalho de Melod62c1d72017-07-31 10:45:07 -0300687 BPF_ADJ_ROOM_NET,
Daniel Borkmann2be7e212017-07-02 02:13:26 +0200688};
689
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300690/* user accessible mirror of in-kernel sk_buff.
691 * new fields can only be added to the end of this structure
692 */
693struct __sk_buff {
694 __u32 len;
695 __u32 pkt_type;
696 __u32 mark;
697 __u32 queue_mapping;
698 __u32 protocol;
699 __u32 vlan_present;
700 __u32 vlan_tci;
701 __u32 vlan_proto;
702 __u32 priority;
703 __u32 ingress_ifindex;
704 __u32 ifindex;
705 __u32 tc_index;
706 __u32 cb[5];
707 __u32 hash;
708 __u32 tc_classid;
709 __u32 data;
710 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +0200711 __u32 napi_id;
John Fastabend69e8cc12017-08-15 22:33:32 -0700712
713 /* accessed by BPF_PROG_TYPE_sk_skb types */
714 __u32 family;
715 __u32 remote_ip4; /* Stored in network byte order */
716 __u32 local_ip4; /* Stored in network byte order */
717 __u32 remote_ip6[4]; /* Stored in network byte order */
718 __u32 local_ip6[4]; /* Stored in network byte order */
719 __u32 remote_port; /* Stored in network byte order */
720 __u32 local_port; /* stored in host byte order */
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300721};
722
723struct bpf_tunnel_key {
724 __u32 tunnel_id;
725 union {
726 __u32 remote_ipv4;
727 __u32 remote_ipv6[4];
728 };
729 __u8 tunnel_tos;
730 __u8 tunnel_ttl;
731 __u16 tunnel_ext;
732 __u32 tunnel_label;
733};
734
Joe Stringer0cb34dc2016-12-08 18:46:14 -0800735/* Generic BPF return codes which all BPF program types may support.
736 * The values are binary compatible with their TC_ACT_* counter-part to
737 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
738 * programs.
739 *
740 * XDP is handled seprately, see XDP_*.
741 */
742enum bpf_ret_code {
743 BPF_OK = 0,
744 /* 1 reserved */
745 BPF_DROP = 2,
746 /* 3-6 reserved */
747 BPF_REDIRECT = 7,
748 /* >127 are reserved for prog type specific return codes */
749};
750
751struct bpf_sock {
752 __u32 bound_dev_if;
753 __u32 family;
754 __u32 type;
755 __u32 protocol;
756};
757
758#define XDP_PACKET_HEADROOM 256
759
Arnaldo Carvalho de Melo791cceb2016-08-09 11:48:07 -0300760/* User return codes for XDP prog type.
761 * A valid XDP program must return one of these defined values. All other
762 * return codes are reserved for future use. Unknown return codes will result
763 * in packet drop.
764 */
765enum xdp_action {
766 XDP_ABORTED = 0,
767 XDP_DROP,
768 XDP_PASS,
769 XDP_TX,
770};
771
772/* user accessible metadata for XDP packet hook
773 * new fields must be added to the end of this structure
774 */
775struct xdp_md {
776 __u32 data;
777 __u32 data_end;
778};
779
John Fastabend69e8cc12017-08-15 22:33:32 -0700780enum sk_action {
781 SK_ABORTED = 0,
782 SK_DROP,
783 SK_REDIRECT,
784};
785
Martin KaFai Lau95b9afd2017-06-05 12:15:53 -0700786#define BPF_TAG_SIZE 8
787
788struct bpf_prog_info {
789 __u32 type;
790 __u32 id;
791 __u8 tag[BPF_TAG_SIZE];
792 __u32 jited_prog_len;
793 __u32 xlated_prog_len;
794 __aligned_u64 jited_prog_insns;
795 __aligned_u64 xlated_prog_insns;
796} __attribute__((aligned(8)));
797
798struct bpf_map_info {
799 __u32 type;
800 __u32 id;
801 __u32 key_size;
802 __u32 value_size;
803 __u32 max_entries;
804 __u32 map_flags;
805} __attribute__((aligned(8)));
806
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700807/* User bpf_sock_ops struct to access socket values and specify request ops
808 * and their replies.
Arnaldo Carvalho de Melof1d6cb22017-07-31 10:45:07 -0300809 * Some of this fields are in network (bigendian) byte order and may need
810 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700811 * New fields can only be added at the end of this structure
812 */
813struct bpf_sock_ops {
814 __u32 op;
815 union {
816 __u32 reply;
817 __u32 replylong[4];
818 };
819 __u32 family;
Arnaldo Carvalho de Melof1d6cb22017-07-31 10:45:07 -0300820 __u32 remote_ip4; /* Stored in network byte order */
821 __u32 local_ip4; /* Stored in network byte order */
822 __u32 remote_ip6[4]; /* Stored in network byte order */
823 __u32 local_ip6[4]; /* Stored in network byte order */
824 __u32 remote_port; /* Stored in network byte order */
825 __u32 local_port; /* stored in host byte order */
Lawrence Brakmo04df41e2017-06-30 20:02:55 -0700826};
827
828/* List of known BPF sock_ops operators.
829 * New entries can only be added at the end
830 */
831enum {
832 BPF_SOCK_OPS_VOID,
833 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
834 * -1 if default value should be used
835 */
836 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
837 * window (in packets) or -1 if default
838 * value should be used
839 */
840 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
841 * active connection is initialized
842 */
843 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
844 * active connection is
845 * established
846 */
847 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
848 * passive connection is
849 * established
850 */
851 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
852 * needs ECN
853 */
854};
855
856#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
857#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
858
Arnaldo Carvalho de Melo971e8272016-07-11 16:38:05 -0300859#endif /* _UAPI__LINUX_BPF_H__ */