blob: 704bb69514a20318adc1188e440cd18e0e1c5e0d [file] [log] [blame]
Greg Kroah-Hartmane2be04c2017-11-01 15:09:13 +01001/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07002/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 */
8#ifndef _UAPI__LINUX_BPF_H__
9#define _UAPI__LINUX_BPF_H__
10
11#include <linux/types.h>
Alexei Starovoitovc15952d2014-10-14 02:08:54 -070012#include <linux/bpf_common.h>
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070013
14/* Extended instruction set based on top of classic BPF */
15
16/* instruction classes */
Jiong Wangd405c742019-01-26 12:25:59 -050017#define BPF_JMP32 0x06 /* jmp mode in word width */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070018#define BPF_ALU64 0x07 /* alu mode in double word width */
19
20/* ld/ldx fields */
Jesper Dangaard Brouercb5f7332018-01-17 12:05:36 +010021#define BPF_DW 0x18 /* double word (64-bit) */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070022#define BPF_XADD 0xc0 /* exclusive add */
23
24/* alu/jmp fields */
25#define BPF_MOV 0xb0 /* mov reg to reg */
26#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27
28/* change endianness of a register */
29#define BPF_END 0xd0 /* flags for endianness conversion: */
30#define BPF_TO_LE 0x00 /* convert to little-endian */
31#define BPF_TO_BE 0x08 /* convert to big-endian */
32#define BPF_FROM_LE BPF_TO_LE
33#define BPF_FROM_BE BPF_TO_BE
34
Daniel Borkmann92b31a92017-08-10 01:39:55 +020035/* jmp encodings */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070036#define BPF_JNE 0x50 /* jump != */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020037#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070039#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020041#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070043#define BPF_CALL 0x80 /* function call */
44#define BPF_EXIT 0x90 /* function return */
45
46/* Register numbers */
47enum {
48 BPF_REG_0 = 0,
49 BPF_REG_1,
50 BPF_REG_2,
51 BPF_REG_3,
52 BPF_REG_4,
53 BPF_REG_5,
54 BPF_REG_6,
55 BPF_REG_7,
56 BPF_REG_8,
57 BPF_REG_9,
58 BPF_REG_10,
59 __MAX_BPF_REG,
60};
61
62/* BPF has 10 general purpose 64-bit registers and stack frame. */
63#define MAX_BPF_REG __MAX_BPF_REG
64
65struct bpf_insn {
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
71};
72
Daniel Mackb95a5c42017-01-21 17:26:11 +010073/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
76 __u8 data[0]; /* Arbitrary size */
77};
78
Roman Gushchinde9cbba2018-08-02 14:27:18 -070079struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
82};
83
Daniel Borkmannb2197752015-10-29 14:58:09 +010084/* BPF syscall commands, see bpf(2) man-page for details. */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070085enum bpf_cmd {
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070086 BPF_MAP_CREATE,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070087 BPF_MAP_LOOKUP_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070088 BPF_MAP_UPDATE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070089 BPF_MAP_DELETE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070090 BPF_MAP_GET_NEXT_KEY,
Alexei Starovoitov09756af2014-09-26 00:17:00 -070091 BPF_PROG_LOAD,
Daniel Borkmannb2197752015-10-29 14:58:09 +010092 BPF_OBJ_PIN,
93 BPF_OBJ_GET,
Daniel Mackf4324552016-11-23 16:52:27 +010094 BPF_PROG_ATTACH,
95 BPF_PROG_DETACH,
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -070096 BPF_PROG_TEST_RUN,
Martin KaFai Lau34ad5582017-06-05 12:15:48 -070097 BPF_PROG_GET_NEXT_ID,
98 BPF_MAP_GET_NEXT_ID,
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -070099 BPF_PROG_GET_FD_BY_ID,
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700100 BPF_MAP_GET_FD_BY_ID,
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700101 BPF_OBJ_GET_INFO_BY_FD,
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700102 BPF_PROG_QUERY,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700103 BPF_RAW_TRACEPOINT_OPEN,
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700104 BPF_BTF_LOAD,
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700105 BPF_BTF_GET_FD_BY_ID,
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700106 BPF_TASK_FD_QUERY,
Mauricio Vasquez Bbd513cd2018-10-18 15:16:30 +0200107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
Daniel Borkmann87df15d2019-04-09 23:20:06 +0200108 BPF_MAP_FREEZE,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700109};
110
111enum bpf_map_type {
112 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -0800113 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -0800114 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61ab2015-05-19 16:59:03 -0700115 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +0000116 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -0800117 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -0800118 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800119 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -0700120 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800121 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800122 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +0100123 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700124 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700125 BPF_MAP_TYPE_HASH_OF_MAPS,
John Fastabend546ac1f2017-07-17 09:28:56 -0700126 BPF_MAP_TYPE_DEVMAP,
John Fastabend174a79f2017-08-15 22:32:47 -0700127 BPF_MAP_TYPE_SOCKMAP,
Jesper Dangaard Brouer6710e112017-10-16 12:19:28 +0200128 BPF_MAP_TYPE_CPUMAP,
Björn Töpelfbfc504a2018-05-02 13:01:28 +0200129 BPF_MAP_TYPE_XSKMAP,
John Fastabend81110382018-05-14 10:00:17 -0700130 BPF_MAP_TYPE_SOCKHASH,
Roman Gushchinde9cbba2018-08-02 14:27:18 -0700131 BPF_MAP_TYPE_CGROUP_STORAGE,
Martin KaFai Lau5dc4c4b2018-08-08 01:01:24 -0700132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
Roman Gushchinb741f162018-09-28 14:45:43 +0000133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +0200134 BPF_MAP_TYPE_QUEUE,
135 BPF_MAP_TYPE_STACK,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700136};
137
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100138/* Note that tracing related programs such as
139 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
140 * are not subject to a stable API since kernel internal data
141 * structures can change from release to release and may
142 * therefore break existing tracing BPF programs. Tracing BPF
143 * programs correspond to /a/ specific kernel which is to be
144 * analyzed, and not /a/ specific kernel /and/ all future ones.
145 */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700146enum bpf_prog_type {
147 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800148 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700149 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100150 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee8c2015-03-20 15:11:11 +0100151 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700152 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700153 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700154 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100155 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800156 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100157 BPF_PROG_TYPE_LWT_IN,
158 BPF_PROG_TYPE_LWT_OUT,
159 BPF_PROG_TYPE_LWT_XMIT,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700160 BPF_PROG_TYPE_SOCK_OPS,
John Fastabendb005fd12017-08-15 22:31:58 -0700161 BPF_PROG_TYPE_SK_SKB,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500162 BPF_PROG_TYPE_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700163 BPF_PROG_TYPE_SK_MSG,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700164 BPF_PROG_TYPE_RAW_TRACEPOINT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700165 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
Mathieu Xhonneux004d4b22018-05-20 14:58:16 +0100166 BPF_PROG_TYPE_LWT_SEG6LOCAL,
Sean Youngf4364dc2018-05-27 12:24:09 +0100167 BPF_PROG_TYPE_LIRC_MODE2,
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -0700168 BPF_PROG_TYPE_SK_REUSEPORT,
Petar Penkovd58e4682018-09-14 07:46:18 -0700169 BPF_PROG_TYPE_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800170 BPF_PROG_TYPE_CGROUP_SYSCTL,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700171};
172
Daniel Mack0e336612016-11-23 16:52:25 +0100173enum bpf_attach_type {
174 BPF_CGROUP_INET_INGRESS,
175 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800176 BPF_CGROUP_INET_SOCK_CREATE,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700177 BPF_CGROUP_SOCK_OPS,
John Fastabend464bc0f2017-08-28 07:10:04 -0700178 BPF_SK_SKB_STREAM_PARSER,
179 BPF_SK_SKB_STREAM_VERDICT,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500180 BPF_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700181 BPF_SK_MSG_VERDICT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700182 BPF_CGROUP_INET4_BIND,
183 BPF_CGROUP_INET6_BIND,
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700184 BPF_CGROUP_INET4_CONNECT,
185 BPF_CGROUP_INET6_CONNECT,
Andrey Ignatovaac3fc32018-03-30 15:08:07 -0700186 BPF_CGROUP_INET4_POST_BIND,
187 BPF_CGROUP_INET6_POST_BIND,
Andrey Ignatov1cedee12018-05-25 08:55:23 -0700188 BPF_CGROUP_UDP4_SENDMSG,
189 BPF_CGROUP_UDP6_SENDMSG,
Sean Youngf4364dc2018-05-27 12:24:09 +0100190 BPF_LIRC_MODE2,
Petar Penkovd58e4682018-09-14 07:46:18 -0700191 BPF_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800192 BPF_CGROUP_SYSCTL,
Daniel Mack0e336612016-11-23 16:52:25 +0100193 __MAX_BPF_ATTACH_TYPE
194};
195
196#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
197
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700198/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
199 *
200 * NONE(default): No further bpf programs allowed in the subtree.
201 *
202 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
203 * the program in this cgroup yields to sub-cgroup program.
204 *
205 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
206 * that cgroup program gets run in addition to the program in this cgroup.
207 *
208 * Only one program is allowed to be attached to a cgroup with
209 * NONE or BPF_F_ALLOW_OVERRIDE flag.
210 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
211 * release old program and attach the new one. Attach flags has to match.
212 *
213 * Multiple programs are allowed to be attached to a cgroup with
214 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
215 * (those that were attached first, run first)
216 * The programs of sub-cgroup are executed first, then programs of
217 * this cgroup and then programs of parent cgroup.
218 * When children program makes decision (like picking TCP CA or sock bind)
219 * parent program has a chance to override it.
220 *
221 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
222 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
223 * Ex1:
224 * cgrp1 (MULTI progs A, B) ->
225 * cgrp2 (OVERRIDE prog C) ->
226 * cgrp3 (MULTI prog D) ->
227 * cgrp4 (OVERRIDE prog E) ->
228 * cgrp5 (NONE prog F)
229 * the event in cgrp5 triggers execution of F,D,A,B in that order.
230 * if prog F is detached, the execution is E,D,A,B
231 * if prog F and D are detached, the execution is E,A,B
232 * if prog F, E and D are detached, the execution is C,A,B
233 *
234 * All eligible programs are executed regardless of return code from
235 * earlier programs.
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800236 */
237#define BPF_F_ALLOW_OVERRIDE (1U << 0)
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700238#define BPF_F_ALLOW_MULTI (1U << 1)
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800239
David S. Millere07b98d2017-05-10 11:38:07 -0700240/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
241 * verifier will perform strict alignment checking as if the kernel
242 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
243 * and NET_IP_ALIGN defined to 2.
244 */
245#define BPF_F_STRICT_ALIGNMENT (1U << 0)
246
David Millere9ee9ef2018-11-30 21:08:14 -0800247/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
248 * verifier will allow any alignment whatsoever. On platforms
249 * with strict alignment requirements for loads ands stores (such
250 * as sparc and mips) the verifier validates that all loads and
251 * stores provably follow this requirement. This flag turns that
252 * checking and enforcement off.
253 *
254 * It is mostly used for testing when we want to validate the
255 * context and memory access aspects of the verifier, but because
256 * of an unaligned access the alignment check would trigger before
257 * the one we are interested in.
258 */
259#define BPF_F_ANY_ALIGNMENT (1U << 1)
260
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200261/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
262 * two extensions:
263 *
264 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
265 * insn[0].imm: map fd map fd
266 * insn[1].imm: 0 offset into value
267 * insn[0].off: 0 0
268 * insn[1].off: 0 0
269 * ldimm64 rewrite: address of map address of map[0]+offset
270 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
271 */
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100272#define BPF_PSEUDO_MAP_FD 1
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200273#define BPF_PSEUDO_MAP_VALUE 2
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100274
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800275/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
276 * offset to another bpf function
277 */
278#define BPF_PSEUDO_CALL 1
279
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800280/* flags for BPF_MAP_UPDATE_ELEM command */
281#define BPF_ANY 0 /* create new element or update existing */
282#define BPF_NOEXIST 1 /* create new element if it didn't exist */
283#define BPF_EXIST 2 /* update existing element */
Alexei Starovoitov96049f32019-01-31 15:40:09 -0800284#define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800285
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700286/* flags for BPF_MAP_CREATE command */
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800287#define BPF_F_NO_PREALLOC (1U << 0)
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800288/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800289 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800290 * which can scale and perform better.
291 * Note, the LRU nodes (including free nodes) cannot be moved
292 * across different LRU lists.
293 */
294#define BPF_F_NO_COMMON_LRU (1U << 1)
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700295/* Specify numa node during map creation */
296#define BPF_F_NUMA_NODE (1U << 2)
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800297
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -0700298#define BPF_OBJ_NAME_LEN 16U
299
Daniel Borkmann591fe982019-04-09 23:20:05 +0200300/* Flags for accessing BPF object from syscall side. */
Chenbo Feng6e71b042017-10-18 13:00:22 -0700301#define BPF_F_RDONLY (1U << 3)
302#define BPF_F_WRONLY (1U << 4)
303
Song Liu615755a2018-03-14 10:23:21 -0700304/* Flag for stack_map, store build_id+offset instead of pointer */
305#define BPF_F_STACK_BUILD_ID (1U << 5)
306
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000307/* Zero-initialize hash function seed. This should only be used for testing. */
308#define BPF_F_ZERO_SEED (1U << 6)
309
Daniel Borkmann591fe982019-04-09 23:20:05 +0200310/* Flags for accessing BPF object from program side. */
311#define BPF_F_RDONLY_PROG (1U << 7)
312#define BPF_F_WRONLY_PROG (1U << 8)
313
Lorenz Bauer2f183362018-11-16 11:41:09 +0000314/* flags for BPF_PROG_QUERY */
315#define BPF_F_QUERY_EFFECTIVE (1U << 0)
316
Song Liu615755a2018-03-14 10:23:21 -0700317enum bpf_stack_build_id_status {
318 /* user space need an empty entry to identify end of a trace */
319 BPF_STACK_BUILD_ID_EMPTY = 0,
320 /* with valid build_id and offset */
321 BPF_STACK_BUILD_ID_VALID = 1,
322 /* couldn't get build_id, fallback to ip */
323 BPF_STACK_BUILD_ID_IP = 2,
324};
325
326#define BPF_BUILD_ID_SIZE 20
327struct bpf_stack_build_id {
328 __s32 status;
329 unsigned char build_id[BPF_BUILD_ID_SIZE];
330 union {
331 __u64 offset;
332 __u64 ip;
333 };
334};
335
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700336union bpf_attr {
337 struct { /* anonymous struct used by BPF_MAP_CREATE command */
338 __u32 map_type; /* one of enum bpf_map_type */
339 __u32 key_size; /* size of key in bytes */
340 __u32 value_size; /* size of value in bytes */
341 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700342 __u32 map_flags; /* BPF_MAP_CREATE related
343 * flags defined above.
344 */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700345 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700346 __u32 numa_node; /* numa node (effective only if
347 * BPF_F_NUMA_NODE is set).
348 */
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700349 char map_name[BPF_OBJ_NAME_LEN];
Jakub Kicinskia3884572018-01-11 20:29:09 -0800350 __u32 map_ifindex; /* ifindex of netdev to create on */
Martin KaFai Laua26ca7c2018-04-18 15:56:03 -0700351 __u32 btf_fd; /* fd pointing to a BTF type data */
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -0700352 __u32 btf_key_type_id; /* BTF type_id of the key */
353 __u32 btf_value_type_id; /* BTF type_id of the value */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700354 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700355
356 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
357 __u32 map_fd;
358 __aligned_u64 key;
359 union {
360 __aligned_u64 value;
361 __aligned_u64 next_key;
362 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800363 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700364 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700365
366 struct { /* anonymous struct used by BPF_PROG_LOAD command */
367 __u32 prog_type; /* one of enum bpf_prog_type */
368 __u32 insn_cnt;
369 __aligned_u64 insns;
370 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700371 __u32 log_level; /* verbosity level of verifier */
372 __u32 log_size; /* size of user buffer */
373 __aligned_u64 log_buf; /* user supplied buffer */
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100374 __u32 kern_version; /* not used */
David S. Millere07b98d2017-05-10 11:38:07 -0700375 __u32 prog_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700376 char prog_name[BPF_OBJ_NAME_LEN];
Jakub Kicinski1f6f4cb2017-11-20 15:21:53 -0800377 __u32 prog_ifindex; /* ifindex of netdev to prep for */
Andrey Ignatov5e43f892018-03-30 15:08:00 -0700378 /* For some prog types expected attach type must be known at
379 * load time to verify attach type specific parts of prog
380 * (context accesses, allowed helpers, etc).
381 */
382 __u32 expected_attach_type;
Yonghong Song838e9692018-11-19 15:29:11 -0800383 __u32 prog_btf_fd; /* fd pointing to BTF type data */
384 __u32 func_info_rec_size; /* userspace bpf_func_info size */
385 __aligned_u64 func_info; /* func info */
386 __u32 func_info_cnt; /* number of bpf_func_info records */
Martin KaFai Lauc454a462018-12-07 16:42:25 -0800387 __u32 line_info_rec_size; /* userspace bpf_line_info size */
388 __aligned_u64 line_info; /* line info */
389 __u32 line_info_cnt; /* number of bpf_line_info records */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700390 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100391
392 struct { /* anonymous struct used by BPF_OBJ_* commands */
393 __aligned_u64 pathname;
394 __u32 bpf_fd;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700395 __u32 file_flags;
Daniel Borkmannb2197752015-10-29 14:58:09 +0100396 };
Daniel Mackf4324552016-11-23 16:52:27 +0100397
398 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
399 __u32 target_fd; /* container object to attach to */
400 __u32 attach_bpf_fd; /* eBPF program to attach */
401 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800402 __u32 attach_flags;
Daniel Mackf4324552016-11-23 16:52:27 +0100403 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700404
405 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
406 __u32 prog_fd;
407 __u32 retval;
Lorenz Bauerb5a36b12018-12-03 11:31:23 +0000408 __u32 data_size_in; /* input: len of data_in */
409 __u32 data_size_out; /* input/output: len of data_out
410 * returns ENOSPC if data_out
411 * is too small.
412 */
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700413 __aligned_u64 data_in;
414 __aligned_u64 data_out;
415 __u32 repeat;
416 __u32 duration;
Stanislav Fomichevb0b93952019-04-09 11:49:09 -0700417 __u32 ctx_size_in; /* input: len of ctx_in */
418 __u32 ctx_size_out; /* input/output: len of ctx_out
419 * returns ENOSPC if ctx_out
420 * is too small.
421 */
422 __aligned_u64 ctx_in;
423 __aligned_u64 ctx_out;
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700424 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700425
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700426 struct { /* anonymous struct used by BPF_*_GET_*_ID */
427 union {
428 __u32 start_id;
429 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700430 __u32 map_id;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700431 __u32 btf_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700432 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700433 __u32 next_id;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700434 __u32 open_flags;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700435 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700436
437 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
438 __u32 bpf_fd;
439 __u32 info_len;
440 __aligned_u64 info;
441 } info;
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700442
443 struct { /* anonymous struct used by BPF_PROG_QUERY command */
444 __u32 target_fd; /* container object to query */
445 __u32 attach_type;
446 __u32 query_flags;
447 __u32 attach_flags;
448 __aligned_u64 prog_ids;
449 __u32 prog_cnt;
450 } query;
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700451
452 struct {
453 __u64 name;
454 __u32 prog_fd;
455 } raw_tracepoint;
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700456
457 struct { /* anonymous struct for BPF_BTF_LOAD */
458 __aligned_u64 btf;
459 __aligned_u64 btf_log_buf;
460 __u32 btf_size;
461 __u32 btf_log_size;
462 __u32 btf_log_level;
463 };
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700464
465 struct {
466 __u32 pid; /* input: pid */
467 __u32 fd; /* input: fd */
468 __u32 flags; /* input: flags */
469 __u32 buf_len; /* input/output: buf len */
470 __aligned_u64 buf; /* input/output:
471 * tp_name for tracepoint
472 * symbol for kprobe
473 * filename for uprobe
474 */
475 __u32 prog_id; /* output: prod_id */
476 __u32 fd_type; /* output: BPF_FD_TYPE_* */
477 __u64 probe_offset; /* output: probe_offset */
478 __u64 probe_addr; /* output: probe_addr */
479 } task_fd_query;
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700480} __attribute__((aligned(8)));
481
Quentin Monnet56a092c2018-04-25 18:16:52 +0100482/* The description below is an attempt at providing documentation to eBPF
483 * developers about the multiple available eBPF helper functions. It can be
484 * parsed and used to produce a manual page. The workflow is the following,
485 * and requires the rst2man utility:
Thomas Grafebb676d2016-10-27 11:23:51 +0200486 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100487 * $ ./scripts/bpf_helpers_doc.py \
488 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
489 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
490 * $ man /tmp/bpf-helpers.7
Thomas Grafebb676d2016-10-27 11:23:51 +0200491 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100492 * Note that in order to produce this external documentation, some RST
493 * formatting is used in the descriptions to get "bold" and "italics" in
494 * manual pages. Also note that the few trailing white spaces are
495 * intentional, removing them would break paragraphs for rst2man.
Thomas Grafebb676d2016-10-27 11:23:51 +0200496 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100497 * Start of BPF helper function descriptions:
Quentin Monnetad4a5222018-04-25 18:16:53 +0100498 *
499 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
500 * Description
501 * Perform a lookup in *map* for an entry associated to *key*.
502 * Return
503 * Map value associated to *key*, or **NULL** if no entry was
504 * found.
505 *
506 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
507 * Description
508 * Add or update the value of the entry associated to *key* in
509 * *map* with *value*. *flags* is one of:
510 *
511 * **BPF_NOEXIST**
512 * The entry for *key* must not exist in the map.
513 * **BPF_EXIST**
514 * The entry for *key* must already exist in the map.
515 * **BPF_ANY**
516 * No condition on the existence of the entry for *key*.
517 *
518 * Flag value **BPF_NOEXIST** cannot be used for maps of types
519 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
520 * elements always exist), the helper would return an error.
521 * Return
522 * 0 on success, or a negative error in case of failure.
523 *
524 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
525 * Description
526 * Delete entry with *key* from *map*.
527 * Return
528 * 0 on success, or a negative error in case of failure.
529 *
530 * int bpf_probe_read(void *dst, u32 size, const void *src)
531 * Description
532 * For tracing programs, safely attempt to read *size* bytes from
533 * address *src* and store the data in *dst*.
534 * Return
535 * 0 on success, or a negative error in case of failure.
536 *
537 * u64 bpf_ktime_get_ns(void)
538 * Description
539 * Return the time elapsed since system boot, in nanoseconds.
540 * Return
541 * Current *ktime*.
542 *
543 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
544 * Description
545 * This helper is a "printk()-like" facility for debugging. It
546 * prints a message defined by format *fmt* (of size *fmt_size*)
547 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
548 * available. It can take up to three additional **u64**
549 * arguments (as an eBPF helpers, the total number of arguments is
550 * limited to five).
551 *
552 * Each time the helper is called, it appends a line to the trace.
553 * The format of the trace is customizable, and the exact output
554 * one will get depends on the options set in
555 * *\/sys/kernel/debug/tracing/trace_options* (see also the
556 * *README* file under the same directory). However, it usually
557 * defaults to something like:
558 *
559 * ::
560 *
561 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
562 *
563 * In the above:
564 *
565 * * ``telnet`` is the name of the current task.
566 * * ``470`` is the PID of the current task.
567 * * ``001`` is the CPU number on which the task is
568 * running.
569 * * In ``.N..``, each character refers to a set of
570 * options (whether irqs are enabled, scheduling
571 * options, whether hard/softirqs are running, level of
572 * preempt_disabled respectively). **N** means that
573 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
574 * are set.
575 * * ``419421.045894`` is a timestamp.
576 * * ``0x00000001`` is a fake value used by BPF for the
577 * instruction pointer register.
578 * * ``<formatted msg>`` is the message formatted with
579 * *fmt*.
580 *
581 * The conversion specifiers supported by *fmt* are similar, but
582 * more limited than for printk(). They are **%d**, **%i**,
583 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
584 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
585 * of field, padding with zeroes, etc.) is available, and the
586 * helper will return **-EINVAL** (but print nothing) if it
587 * encounters an unknown specifier.
588 *
589 * Also, note that **bpf_trace_printk**\ () is slow, and should
590 * only be used for debugging purposes. For this reason, a notice
591 * bloc (spanning several lines) is printed to kernel logs and
592 * states that the helper should not be used "for production use"
593 * the first time this helper is used (or more precisely, when
594 * **trace_printk**\ () buffers are allocated). For passing values
595 * to user space, perf events should be preferred.
596 * Return
597 * The number of bytes written to the buffer, or a negative error
598 * in case of failure.
599 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100600 * u32 bpf_get_prandom_u32(void)
601 * Description
602 * Get a pseudo-random number.
603 *
604 * From a security point of view, this helper uses its own
605 * pseudo-random internal state, and cannot be used to infer the
606 * seed of other random functions in the kernel. However, it is
607 * essential to note that the generator used by the helper is not
608 * cryptographically secure.
609 * Return
610 * A random 32-bit unsigned value.
611 *
612 * u32 bpf_get_smp_processor_id(void)
613 * Description
614 * Get the SMP (symmetric multiprocessing) processor id. Note that
615 * all programs run with preemption disabled, which means that the
616 * SMP processor id is stable during all the execution of the
617 * program.
618 * Return
619 * The SMP id of the processor running the program.
620 *
Quentin Monnetad4a5222018-04-25 18:16:53 +0100621 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
622 * Description
623 * Store *len* bytes from address *from* into the packet
624 * associated to *skb*, at *offset*. *flags* are a combination of
625 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
626 * checksum for the packet after storing the bytes) and
627 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
628 * **->swhash** and *skb*\ **->l4hash** to 0).
629 *
630 * A call to this helper is susceptible to change the underlaying
631 * packet buffer. Therefore, at load time, all checks on pointers
632 * previously done by the verifier are invalidated and must be
633 * performed again, if the helper is used in combination with
634 * direct packet access.
635 * Return
636 * 0 on success, or a negative error in case of failure.
637 *
638 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
639 * Description
640 * Recompute the layer 3 (e.g. IP) checksum for the packet
641 * associated to *skb*. Computation is incremental, so the helper
642 * must know the former value of the header field that was
643 * modified (*from*), the new value of this field (*to*), and the
644 * number of bytes (2 or 4) for this field, stored in *size*.
645 * Alternatively, it is possible to store the difference between
646 * the previous and the new values of the header field in *to*, by
647 * setting *from* and *size* to 0. For both methods, *offset*
648 * indicates the location of the IP checksum within the packet.
649 *
650 * This helper works in combination with **bpf_csum_diff**\ (),
651 * which does not update the checksum in-place, but offers more
652 * flexibility and can handle sizes larger than 2 or 4 for the
653 * checksum to update.
654 *
655 * A call to this helper is susceptible to change the underlaying
656 * packet buffer. Therefore, at load time, all checks on pointers
657 * previously done by the verifier are invalidated and must be
658 * performed again, if the helper is used in combination with
659 * direct packet access.
660 * Return
661 * 0 on success, or a negative error in case of failure.
662 *
663 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
664 * Description
665 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
666 * packet associated to *skb*. Computation is incremental, so the
667 * helper must know the former value of the header field that was
668 * modified (*from*), the new value of this field (*to*), and the
669 * number of bytes (2 or 4) for this field, stored on the lowest
670 * four bits of *flags*. Alternatively, it is possible to store
671 * the difference between the previous and the new values of the
672 * header field in *to*, by setting *from* and the four lowest
673 * bits of *flags* to 0. For both methods, *offset* indicates the
674 * location of the IP checksum within the packet. In addition to
675 * the size of the field, *flags* can be added (bitwise OR) actual
676 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
677 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
678 * for updates resulting in a null checksum the value is set to
679 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
680 * the checksum is to be computed against a pseudo-header.
681 *
682 * This helper works in combination with **bpf_csum_diff**\ (),
683 * which does not update the checksum in-place, but offers more
684 * flexibility and can handle sizes larger than 2 or 4 for the
685 * checksum to update.
686 *
687 * A call to this helper is susceptible to change the underlaying
688 * packet buffer. Therefore, at load time, all checks on pointers
689 * previously done by the verifier are invalidated and must be
690 * performed again, if the helper is used in combination with
691 * direct packet access.
692 * Return
693 * 0 on success, or a negative error in case of failure.
694 *
695 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
696 * Description
697 * This special helper is used to trigger a "tail call", or in
698 * other words, to jump into another eBPF program. The same stack
699 * frame is used (but values on stack and in registers for the
700 * caller are not accessible to the callee). This mechanism allows
701 * for program chaining, either for raising the maximum number of
702 * available eBPF instructions, or to execute given programs in
703 * conditional blocks. For security reasons, there is an upper
704 * limit to the number of successive tail calls that can be
705 * performed.
706 *
707 * Upon call of this helper, the program attempts to jump into a
708 * program referenced at index *index* in *prog_array_map*, a
709 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
710 * *ctx*, a pointer to the context.
711 *
712 * If the call succeeds, the kernel immediately runs the first
713 * instruction of the new program. This is not a function call,
714 * and it never returns to the previous program. If the call
715 * fails, then the helper has no effect, and the caller continues
716 * to run its subsequent instructions. A call can fail if the
717 * destination program for the jump does not exist (i.e. *index*
718 * is superior to the number of entries in *prog_array_map*), or
719 * if the maximum number of tail calls has been reached for this
720 * chain of programs. This limit is defined in the kernel by the
721 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
722 * which is currently set to 32.
723 * Return
724 * 0 on success, or a negative error in case of failure.
725 *
726 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
727 * Description
728 * Clone and redirect the packet associated to *skb* to another
729 * net device of index *ifindex*. Both ingress and egress
730 * interfaces can be used for redirection. The **BPF_F_INGRESS**
731 * value in *flags* is used to make the distinction (ingress path
732 * is selected if the flag is present, egress path otherwise).
733 * This is the only flag supported for now.
734 *
735 * In comparison with **bpf_redirect**\ () helper,
736 * **bpf_clone_redirect**\ () has the associated cost of
737 * duplicating the packet buffer, but this can be executed out of
738 * the eBPF program. Conversely, **bpf_redirect**\ () is more
739 * efficient, but it is handled through an action code where the
740 * redirection happens only after the eBPF program has returned.
741 *
742 * A call to this helper is susceptible to change the underlaying
743 * packet buffer. Therefore, at load time, all checks on pointers
744 * previously done by the verifier are invalidated and must be
745 * performed again, if the helper is used in combination with
746 * direct packet access.
747 * Return
748 * 0 on success, or a negative error in case of failure.
Quentin Monnetc456dec2018-04-25 18:16:54 +0100749 *
750 * u64 bpf_get_current_pid_tgid(void)
751 * Return
752 * A 64-bit integer containing the current tgid and pid, and
753 * created as such:
754 * *current_task*\ **->tgid << 32 \|**
755 * *current_task*\ **->pid**.
756 *
757 * u64 bpf_get_current_uid_gid(void)
758 * Return
759 * A 64-bit integer containing the current GID and UID, and
760 * created as such: *current_gid* **<< 32 \|** *current_uid*.
761 *
762 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
763 * Description
764 * Copy the **comm** attribute of the current task into *buf* of
765 * *size_of_buf*. The **comm** attribute contains the name of
766 * the executable (excluding the path) for the current task. The
767 * *size_of_buf* must be strictly positive. On success, the
768 * helper makes sure that the *buf* is NUL-terminated. On failure,
769 * it is filled with zeroes.
770 * Return
771 * 0 on success, or a negative error in case of failure.
772 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100773 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
774 * Description
775 * Retrieve the classid for the current task, i.e. for the net_cls
776 * cgroup to which *skb* belongs.
777 *
778 * This helper can be used on TC egress path, but not on ingress.
779 *
780 * The net_cls cgroup provides an interface to tag network packets
781 * based on a user-provided identifier for all traffic coming from
782 * the tasks belonging to the related cgroup. See also the related
783 * kernel documentation, available from the Linux sources in file
784 * *Documentation/cgroup-v1/net_cls.txt*.
785 *
786 * The Linux kernel has two versions for cgroups: there are
787 * cgroups v1 and cgroups v2. Both are available to users, who can
788 * use a mixture of them, but note that the net_cls cgroup is for
789 * cgroup v1 only. This makes it incompatible with BPF programs
790 * run on cgroups, which is a cgroup-v2-only feature (a socket can
791 * only hold data for one version of cgroups at a time).
792 *
793 * This helper is only available is the kernel was compiled with
794 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
795 * "**y**" or to "**m**".
796 * Return
797 * The classid, or 0 for the default unconfigured classid.
798 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100799 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
800 * Description
801 * Push a *vlan_tci* (VLAN tag control information) of protocol
802 * *vlan_proto* to the packet associated to *skb*, then update
803 * the checksum. Note that if *vlan_proto* is different from
804 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
805 * be **ETH_P_8021Q**.
806 *
807 * A call to this helper is susceptible to change the underlaying
808 * packet buffer. Therefore, at load time, all checks on pointers
809 * previously done by the verifier are invalidated and must be
810 * performed again, if the helper is used in combination with
811 * direct packet access.
812 * Return
813 * 0 on success, or a negative error in case of failure.
814 *
815 * int bpf_skb_vlan_pop(struct sk_buff *skb)
816 * Description
817 * Pop a VLAN header from the packet associated to *skb*.
818 *
819 * A call to this helper is susceptible to change the underlaying
820 * packet buffer. Therefore, at load time, all checks on pointers
821 * previously done by the verifier are invalidated and must be
822 * performed again, if the helper is used in combination with
823 * direct packet access.
824 * Return
825 * 0 on success, or a negative error in case of failure.
826 *
827 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
828 * Description
829 * Get tunnel metadata. This helper takes a pointer *key* to an
830 * empty **struct bpf_tunnel_key** of **size**, that will be
831 * filled with tunnel metadata for the packet associated to *skb*.
832 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
833 * indicates that the tunnel is based on IPv6 protocol instead of
834 * IPv4.
835 *
836 * The **struct bpf_tunnel_key** is an object that generalizes the
837 * principal parameters used by various tunneling protocols into a
838 * single struct. This way, it can be used to easily make a
839 * decision based on the contents of the encapsulation header,
840 * "summarized" in this struct. In particular, it holds the IP
841 * address of the remote end (IPv4 or IPv6, depending on the case)
842 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
843 * this struct exposes the *key*\ **->tunnel_id**, which is
844 * generally mapped to a VNI (Virtual Network Identifier), making
845 * it programmable together with the **bpf_skb_set_tunnel_key**\
846 * () helper.
847 *
848 * Let's imagine that the following code is part of a program
849 * attached to the TC ingress interface, on one end of a GRE
850 * tunnel, and is supposed to filter out all messages coming from
851 * remote ends with IPv4 address other than 10.0.0.1:
852 *
853 * ::
854 *
855 * int ret;
856 * struct bpf_tunnel_key key = {};
857 *
858 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
859 * if (ret < 0)
860 * return TC_ACT_SHOT; // drop packet
861 *
862 * if (key.remote_ipv4 != 0x0a000001)
863 * return TC_ACT_SHOT; // drop packet
864 *
865 * return TC_ACT_OK; // accept packet
866 *
867 * This interface can also be used with all encapsulation devices
868 * that can operate in "collect metadata" mode: instead of having
869 * one network device per specific configuration, the "collect
870 * metadata" mode only requires a single device where the
871 * configuration can be extracted from this helper.
872 *
873 * This can be used together with various tunnels such as VXLan,
874 * Geneve, GRE or IP in IP (IPIP).
875 * Return
876 * 0 on success, or a negative error in case of failure.
877 *
878 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
879 * Description
880 * Populate tunnel metadata for packet associated to *skb.* The
881 * tunnel metadata is set to the contents of *key*, of *size*. The
882 * *flags* can be set to a combination of the following values:
883 *
884 * **BPF_F_TUNINFO_IPV6**
885 * Indicate that the tunnel is based on IPv6 protocol
886 * instead of IPv4.
887 * **BPF_F_ZERO_CSUM_TX**
888 * For IPv4 packets, add a flag to tunnel metadata
889 * indicating that checksum computation should be skipped
890 * and checksum set to zeroes.
891 * **BPF_F_DONT_FRAGMENT**
892 * Add a flag to tunnel metadata indicating that the
893 * packet should not be fragmented.
894 * **BPF_F_SEQ_NUMBER**
895 * Add a flag to tunnel metadata indicating that a
896 * sequence number should be added to tunnel header before
897 * sending the packet. This flag was added for GRE
898 * encapsulation, but might be used with other protocols
899 * as well in the future.
900 *
901 * Here is a typical usage on the transmit path:
902 *
903 * ::
904 *
905 * struct bpf_tunnel_key key;
906 * populate key ...
907 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
908 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
909 *
910 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
911 * helper for additional information.
912 * Return
913 * 0 on success, or a negative error in case of failure.
914 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100915 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
916 * Description
917 * Read the value of a perf event counter. This helper relies on a
918 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
919 * the perf event counter is selected when *map* is updated with
920 * perf event file descriptors. The *map* is an array whose size
921 * is the number of available CPUs, and each cell contains a value
922 * relative to one CPU. The value to retrieve is indicated by
923 * *flags*, that contains the index of the CPU to look up, masked
924 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
925 * **BPF_F_CURRENT_CPU** to indicate that the value for the
926 * current CPU should be retrieved.
927 *
928 * Note that before Linux 4.13, only hardware perf event can be
929 * retrieved.
930 *
931 * Also, be aware that the newer helper
932 * **bpf_perf_event_read_value**\ () is recommended over
Quentin Monnet3bd5a092018-04-30 11:39:03 +0100933 * **bpf_perf_event_read**\ () in general. The latter has some ABI
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100934 * quirks where error and counter value are used as a return code
935 * (which is wrong to do since ranges may overlap). This issue is
Quentin Monnet3bd5a092018-04-30 11:39:03 +0100936 * fixed with **bpf_perf_event_read_value**\ (), which at the same
937 * time provides more features over the **bpf_perf_event_read**\
938 * () interface. Please refer to the description of
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100939 * **bpf_perf_event_read_value**\ () for details.
940 * Return
941 * The value of the perf event counter read from the map, or a
942 * negative error code in case of failure.
943 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100944 * int bpf_redirect(u32 ifindex, u64 flags)
945 * Description
946 * Redirect the packet to another net device of index *ifindex*.
947 * This helper is somewhat similar to **bpf_clone_redirect**\
948 * (), except that the packet is not cloned, which provides
949 * increased performance.
950 *
951 * Except for XDP, both ingress and egress interfaces can be used
952 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
953 * to make the distinction (ingress path is selected if the flag
954 * is present, egress path otherwise). Currently, XDP only
955 * supports redirection to the egress interface, and accepts no
956 * flag at all.
957 *
958 * The same effect can be attained with the more generic
959 * **bpf_redirect_map**\ (), which requires specific maps to be
960 * used but offers better performance.
961 * Return
962 * For XDP, the helper returns **XDP_REDIRECT** on success or
963 * **XDP_ABORTED** on error. For other program types, the values
964 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
965 * error.
966 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100967 * u32 bpf_get_route_realm(struct sk_buff *skb)
968 * Description
969 * Retrieve the realm or the route, that is to say the
970 * **tclassid** field of the destination for the *skb*. The
971 * indentifier retrieved is a user-provided tag, similar to the
972 * one used with the net_cls cgroup (see description for
973 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
974 * held by a route (a destination entry), not by a task.
975 *
976 * Retrieving this identifier works with the clsact TC egress hook
977 * (see also **tc-bpf(8)**), or alternatively on conventional
978 * classful egress qdiscs, but not on TC ingress path. In case of
979 * clsact TC egress hook, this has the advantage that, internally,
980 * the destination entry has not been dropped yet in the transmit
981 * path. Therefore, the destination entry does not need to be
982 * artificially held via **netif_keep_dst**\ () for a classful
983 * qdisc until the *skb* is freed.
984 *
985 * This helper is available only if the kernel was compiled with
986 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
987 * Return
988 * The realm of the route for the packet associated to *skb*, or 0
989 * if none was found.
990 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100991 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
992 * Description
993 * Write raw *data* blob into a special BPF perf event held by
994 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
995 * event must have the following attributes: **PERF_SAMPLE_RAW**
996 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
997 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
998 *
999 * The *flags* are used to indicate the index in *map* for which
1000 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1001 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1002 * to indicate that the index of the current CPU core should be
1003 * used.
1004 *
1005 * The value to write, of *size*, is passed through eBPF stack and
1006 * pointed by *data*.
1007 *
1008 * The context of the program *ctx* needs also be passed to the
1009 * helper.
1010 *
1011 * On user space, a program willing to read the values needs to
1012 * call **perf_event_open**\ () on the perf event (either for
1013 * one or for all CPUs) and to store the file descriptor into the
1014 * *map*. This must be done before the eBPF program can send data
1015 * into it. An example is available in file
1016 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1017 * tree (the eBPF program counterpart is in
1018 * *samples/bpf/trace_output_kern.c*).
1019 *
1020 * **bpf_perf_event_output**\ () achieves better performance
1021 * than **bpf_trace_printk**\ () for sharing data with user
1022 * space, and is much better suitable for streaming data from eBPF
1023 * programs.
1024 *
1025 * Note that this helper is not restricted to tracing use cases
1026 * and can be used with programs attached to TC or XDP as well,
1027 * where it allows for passing data to user space listeners. Data
1028 * can be:
1029 *
1030 * * Only custom structs,
1031 * * Only the packet payload, or
1032 * * A combination of both.
1033 * Return
1034 * 0 on success, or a negative error in case of failure.
1035 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001036 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1037 * Description
1038 * This helper was provided as an easy way to load data from a
1039 * packet. It can be used to load *len* bytes from *offset* from
1040 * the packet associated to *skb*, into the buffer pointed by
1041 * *to*.
1042 *
1043 * Since Linux 4.7, usage of this helper has mostly been replaced
1044 * by "direct packet access", enabling packet data to be
1045 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1046 * pointing respectively to the first byte of packet data and to
1047 * the byte after the last byte of packet data. However, it
1048 * remains useful if one wishes to read large quantities of data
1049 * at once from a packet into the eBPF stack.
1050 * Return
1051 * 0 on success, or a negative error in case of failure.
1052 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001053 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1054 * Description
1055 * Walk a user or a kernel stack and return its id. To achieve
1056 * this, the helper needs *ctx*, which is a pointer to the context
1057 * on which the tracing program is executed, and a pointer to a
1058 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1059 *
1060 * The last argument, *flags*, holds the number of stack frames to
1061 * skip (from 0 to 255), masked with
1062 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1063 * a combination of the following flags:
1064 *
1065 * **BPF_F_USER_STACK**
1066 * Collect a user space stack instead of a kernel stack.
1067 * **BPF_F_FAST_STACK_CMP**
1068 * Compare stacks by hash only.
1069 * **BPF_F_REUSE_STACKID**
1070 * If two different stacks hash into the same *stackid*,
1071 * discard the old one.
1072 *
1073 * The stack id retrieved is a 32 bit long integer handle which
1074 * can be further combined with other data (including other stack
1075 * ids) and used as a key into maps. This can be useful for
1076 * generating a variety of graphs (such as flame graphs or off-cpu
1077 * graphs).
1078 *
1079 * For walking a stack, this helper is an improvement over
1080 * **bpf_probe_read**\ (), which can be used with unrolled loops
1081 * but is not efficient and consumes a lot of eBPF instructions.
1082 * Instead, **bpf_get_stackid**\ () can collect up to
1083 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1084 * this limit can be controlled with the **sysctl** program, and
1085 * that it should be manually increased in order to profile long
1086 * user stacks (such as stacks for Java programs). To do so, use:
1087 *
1088 * ::
1089 *
1090 * # sysctl kernel.perf_event_max_stack=<new value>
Quentin Monnetc456dec2018-04-25 18:16:54 +01001091 * Return
1092 * The positive or null stack id on success, or a negative error
1093 * in case of failure.
1094 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001095 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1096 * Description
1097 * Compute a checksum difference, from the raw buffer pointed by
1098 * *from*, of length *from_size* (that must be a multiple of 4),
1099 * towards the raw buffer pointed by *to*, of size *to_size*
1100 * (same remark). An optional *seed* can be added to the value
1101 * (this can be cascaded, the seed may come from a previous call
1102 * to the helper).
1103 *
1104 * This is flexible enough to be used in several ways:
1105 *
1106 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1107 * checksum, it can be used when pushing new data.
1108 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1109 * checksum, it can be used when removing data from a packet.
1110 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1111 * can be used to compute a diff. Note that *from_size* and
1112 * *to_size* do not need to be equal.
1113 *
1114 * This helper can be used in combination with
1115 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1116 * which one can feed in the difference computed with
1117 * **bpf_csum_diff**\ ().
1118 * Return
1119 * The checksum result, or a negative error code in case of
1120 * failure.
1121 *
1122 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1123 * Description
1124 * Retrieve tunnel options metadata for the packet associated to
1125 * *skb*, and store the raw tunnel option data to the buffer *opt*
1126 * of *size*.
1127 *
1128 * This helper can be used with encapsulation devices that can
1129 * operate in "collect metadata" mode (please refer to the related
1130 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1131 * more details). A particular example where this can be used is
1132 * in combination with the Geneve encapsulation protocol, where it
1133 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1134 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1135 * the eBPF program. This allows for full customization of these
1136 * headers.
1137 * Return
1138 * The size of the option data retrieved.
1139 *
1140 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1141 * Description
1142 * Set tunnel options metadata for the packet associated to *skb*
1143 * to the option data contained in the raw buffer *opt* of *size*.
1144 *
1145 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1146 * helper for additional information.
1147 * Return
1148 * 0 on success, or a negative error in case of failure.
1149 *
1150 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1151 * Description
1152 * Change the protocol of the *skb* to *proto*. Currently
1153 * supported are transition from IPv4 to IPv6, and from IPv6 to
1154 * IPv4. The helper takes care of the groundwork for the
1155 * transition, including resizing the socket buffer. The eBPF
1156 * program is expected to fill the new headers, if any, via
1157 * **skb_store_bytes**\ () and to recompute the checksums with
1158 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1159 * (). The main case for this helper is to perform NAT64
1160 * operations out of an eBPF program.
1161 *
1162 * Internally, the GSO type is marked as dodgy so that headers are
1163 * checked and segments are recalculated by the GSO/GRO engine.
1164 * The size for GSO target is adapted as well.
1165 *
1166 * All values for *flags* are reserved for future usage, and must
1167 * be left at zero.
1168 *
1169 * A call to this helper is susceptible to change the underlaying
1170 * packet buffer. Therefore, at load time, all checks on pointers
1171 * previously done by the verifier are invalidated and must be
1172 * performed again, if the helper is used in combination with
1173 * direct packet access.
1174 * Return
1175 * 0 on success, or a negative error in case of failure.
1176 *
1177 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1178 * Description
1179 * Change the packet type for the packet associated to *skb*. This
1180 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1181 * the eBPF program does not have a write access to *skb*\
1182 * **->pkt_type** beside this helper. Using a helper here allows
1183 * for graceful handling of errors.
1184 *
1185 * The major use case is to change incoming *skb*s to
1186 * **PACKET_HOST** in a programmatic way instead of having to
1187 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1188 * example.
1189 *
1190 * Note that *type* only allows certain values. At this time, they
1191 * are:
1192 *
1193 * **PACKET_HOST**
1194 * Packet is for us.
1195 * **PACKET_BROADCAST**
1196 * Send packet to all.
1197 * **PACKET_MULTICAST**
1198 * Send packet to group.
1199 * **PACKET_OTHERHOST**
1200 * Send packet to someone else.
1201 * Return
1202 * 0 on success, or a negative error in case of failure.
1203 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001204 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1205 * Description
1206 * Check whether *skb* is a descendant of the cgroup2 held by
1207 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1208 * Return
1209 * The return value depends on the result of the test, and can be:
1210 *
1211 * * 0, if the *skb* failed the cgroup2 descendant test.
1212 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1213 * * A negative error code, if an error occurred.
1214 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001215 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1216 * Description
1217 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1218 * not set, in particular if the hash was cleared due to mangling,
1219 * recompute this hash. Later accesses to the hash can be done
1220 * directly with *skb*\ **->hash**.
1221 *
1222 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1223 * prototype with **bpf_skb_change_proto**\ (), or calling
1224 * **bpf_skb_store_bytes**\ () with the
1225 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1226 * the hash and to trigger a new computation for the next call to
1227 * **bpf_get_hash_recalc**\ ().
1228 * Return
1229 * The 32-bit hash.
1230 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001231 * u64 bpf_get_current_task(void)
1232 * Return
1233 * A pointer to the current task struct.
Quentin Monnetfa156012018-04-25 18:16:56 +01001234 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001235 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1236 * Description
1237 * Attempt in a safe way to write *len* bytes from the buffer
1238 * *src* to *dst* in memory. It only works for threads that are in
1239 * user context, and *dst* must be a valid user space address.
1240 *
1241 * This helper should not be used to implement any kind of
1242 * security mechanism because of TOC-TOU attacks, but rather to
1243 * debug, divert, and manipulate execution of semi-cooperative
1244 * processes.
1245 *
1246 * Keep in mind that this feature is meant for experiments, and it
1247 * has a risk of crashing the system and running programs.
1248 * Therefore, when an eBPF program using this helper is attached,
1249 * a warning including PID and process name is printed to kernel
1250 * logs.
1251 * Return
1252 * 0 on success, or a negative error in case of failure.
1253 *
1254 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1255 * Description
1256 * Check whether the probe is being run is the context of a given
1257 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1258 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1259 * Return
1260 * The return value depends on the result of the test, and can be:
1261 *
1262 * * 0, if the *skb* task belongs to the cgroup2.
1263 * * 1, if the *skb* task does not belong to the cgroup2.
1264 * * A negative error code, if an error occurred.
1265 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001266 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1267 * Description
1268 * Resize (trim or grow) the packet associated to *skb* to the
1269 * new *len*. The *flags* are reserved for future usage, and must
1270 * be left at zero.
1271 *
1272 * The basic idea is that the helper performs the needed work to
1273 * change the size of the packet, then the eBPF program rewrites
1274 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1275 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1276 * and others. This helper is a slow path utility intended for
1277 * replies with control messages. And because it is targeted for
1278 * slow path, the helper itself can afford to be slow: it
1279 * implicitly linearizes, unclones and drops offloads from the
1280 * *skb*.
1281 *
1282 * A call to this helper is susceptible to change the underlaying
1283 * packet buffer. Therefore, at load time, all checks on pointers
1284 * previously done by the verifier are invalidated and must be
1285 * performed again, if the helper is used in combination with
1286 * direct packet access.
1287 * Return
1288 * 0 on success, or a negative error in case of failure.
1289 *
1290 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1291 * Description
1292 * Pull in non-linear data in case the *skb* is non-linear and not
1293 * all of *len* are part of the linear section. Make *len* bytes
1294 * from *skb* readable and writable. If a zero value is passed for
1295 * *len*, then the whole length of the *skb* is pulled.
1296 *
1297 * This helper is only needed for reading and writing with direct
1298 * packet access.
1299 *
1300 * For direct packet access, testing that offsets to access
1301 * are within packet boundaries (test on *skb*\ **->data_end**) is
1302 * susceptible to fail if offsets are invalid, or if the requested
1303 * data is in non-linear parts of the *skb*. On failure the
1304 * program can just bail out, or in the case of a non-linear
1305 * buffer, use a helper to make the data available. The
1306 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1307 * the data. Another one consists in using **bpf_skb_pull_data**
1308 * to pull in once the non-linear parts, then retesting and
1309 * eventually access the data.
1310 *
1311 * At the same time, this also makes sure the *skb* is uncloned,
1312 * which is a necessary condition for direct write. As this needs
1313 * to be an invariant for the write part only, the verifier
1314 * detects writes and adds a prologue that is calling
1315 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1316 * the very beginning in case it is indeed cloned.
1317 *
1318 * A call to this helper is susceptible to change the underlaying
1319 * packet buffer. Therefore, at load time, all checks on pointers
1320 * previously done by the verifier are invalidated and must be
1321 * performed again, if the helper is used in combination with
1322 * direct packet access.
1323 * Return
1324 * 0 on success, or a negative error in case of failure.
1325 *
1326 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1327 * Description
1328 * Add the checksum *csum* into *skb*\ **->csum** in case the
1329 * driver has supplied a checksum for the entire packet into that
1330 * field. Return an error otherwise. This helper is intended to be
1331 * used in combination with **bpf_csum_diff**\ (), in particular
1332 * when the checksum needs to be updated after data has been
1333 * written into the packet through direct packet access.
1334 * Return
1335 * The checksum on success, or a negative error code in case of
1336 * failure.
1337 *
1338 * void bpf_set_hash_invalid(struct sk_buff *skb)
1339 * Description
1340 * Invalidate the current *skb*\ **->hash**. It can be used after
1341 * mangling on headers through direct packet access, in order to
1342 * indicate that the hash is outdated and to trigger a
1343 * recalculation the next time the kernel tries to access this
1344 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1345 *
1346 * int bpf_get_numa_node_id(void)
1347 * Description
1348 * Return the id of the current NUMA node. The primary use case
1349 * for this helper is the selection of sockets for the local NUMA
1350 * node, when the program is attached to sockets using the
1351 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1352 * but the helper is also available to other eBPF program types,
1353 * similarly to **bpf_get_smp_processor_id**\ ().
1354 * Return
1355 * The id of current NUMA node.
1356 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001357 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1358 * Description
1359 * Grows headroom of packet associated to *skb* and adjusts the
1360 * offset of the MAC header accordingly, adding *len* bytes of
1361 * space. It automatically extends and reallocates memory as
1362 * required.
1363 *
1364 * This helper can be used on a layer 3 *skb* to push a MAC header
1365 * for redirection into a layer 2 device.
1366 *
1367 * All values for *flags* are reserved for future usage, and must
1368 * be left at zero.
1369 *
1370 * A call to this helper is susceptible to change the underlaying
1371 * packet buffer. Therefore, at load time, all checks on pointers
1372 * previously done by the verifier are invalidated and must be
1373 * performed again, if the helper is used in combination with
1374 * direct packet access.
1375 * Return
1376 * 0 on success, or a negative error in case of failure.
1377 *
1378 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1379 * Description
1380 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1381 * it is possible to use a negative value for *delta*. This helper
1382 * can be used to prepare the packet for pushing or popping
1383 * headers.
1384 *
1385 * A call to this helper is susceptible to change the underlaying
1386 * packet buffer. Therefore, at load time, all checks on pointers
1387 * previously done by the verifier are invalidated and must be
1388 * performed again, if the helper is used in combination with
1389 * direct packet access.
1390 * Return
1391 * 0 on success, or a negative error in case of failure.
1392 *
1393 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1394 * Description
1395 * Copy a NUL terminated string from an unsafe address
1396 * *unsafe_ptr* to *dst*. The *size* should include the
1397 * terminating NUL byte. In case the string length is smaller than
1398 * *size*, the target is not padded with further NUL bytes. If the
1399 * string length is larger than *size*, just *size*-1 bytes are
1400 * copied and the last byte is set to NUL.
1401 *
1402 * On success, the length of the copied string is returned. This
1403 * makes this helper useful in tracing programs for reading
1404 * strings, and more importantly to get its length at runtime. See
1405 * the following snippet:
1406 *
1407 * ::
1408 *
1409 * SEC("kprobe/sys_open")
1410 * void bpf_sys_open(struct pt_regs *ctx)
1411 * {
1412 * char buf[PATHLEN]; // PATHLEN is defined to 256
1413 * int res = bpf_probe_read_str(buf, sizeof(buf),
1414 * ctx->di);
1415 *
1416 * // Consume buf, for example push it to
1417 * // userspace via bpf_perf_event_output(); we
1418 * // can use res (the string length) as event
1419 * // size, after checking its boundaries.
1420 * }
1421 *
1422 * In comparison, using **bpf_probe_read()** helper here instead
1423 * to read the string would require to estimate the length at
1424 * compile time, and would often result in copying more memory
1425 * than necessary.
1426 *
1427 * Another useful use case is when parsing individual process
1428 * arguments or individual environment variables navigating
1429 * *current*\ **->mm->arg_start** and *current*\
1430 * **->mm->env_start**: using this helper and the return value,
1431 * one can quickly iterate at the right offset of the memory area.
1432 * Return
1433 * On success, the strictly positive length of the string,
1434 * including the trailing NUL character. On error, a negative
1435 * value.
1436 *
1437 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1438 * Description
1439 * If the **struct sk_buff** pointed by *skb* has a known socket,
1440 * retrieve the cookie (generated by the kernel) of this socket.
1441 * If no cookie has been set yet, generate a new cookie. Once
1442 * generated, the socket cookie remains stable for the life of the
1443 * socket. This helper can be useful for monitoring per socket
1444 * networking traffic statistics as it provides a unique socket
1445 * identifier per namespace.
1446 * Return
1447 * A 8-byte long non-decreasing number on success, or 0 if the
1448 * socket field is missing inside *skb*.
1449 *
Andrey Ignatovd692f112018-07-30 17:42:28 -07001450 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1451 * Description
1452 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001453 * *skb*, but gets socket from **struct bpf_sock_addr** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001454 * Return
1455 * A 8-byte long non-decreasing number.
1456 *
1457 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1458 * Description
1459 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001460 * *skb*, but gets socket from **struct bpf_sock_ops** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001461 * Return
1462 * A 8-byte long non-decreasing number.
1463 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001464 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1465 * Return
1466 * The owner UID of the socket associated to *skb*. If the socket
1467 * is **NULL**, or if it is not a full socket (i.e. if it is a
1468 * time-wait or a request socket instead), **overflowuid** value
1469 * is returned (note that **overflowuid** might also be the actual
1470 * UID value for the socket).
1471 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001472 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1473 * Description
1474 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1475 * to value *hash*.
1476 * Return
1477 * 0
1478 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001479 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001480 * Description
1481 * Emulate a call to **setsockopt()** on the socket associated to
1482 * *bpf_socket*, which must be a full socket. The *level* at
1483 * which the option resides and the name *optname* of the option
1484 * must be specified, see **setsockopt(2)** for more information.
1485 * The option value of length *optlen* is pointed by *optval*.
1486 *
1487 * This helper actually implements a subset of **setsockopt()**.
1488 * It supports the following *level*\ s:
1489 *
1490 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1491 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1492 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1493 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1494 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1495 * **TCP_BPF_SNDCWND_CLAMP**.
1496 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1497 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1498 * Return
1499 * 0 on success, or a negative error in case of failure.
1500 *
Nicolas Dichtelb55cbc82018-10-17 16:24:48 +02001501 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001502 * Description
1503 * Grow or shrink the room for data in the packet associated to
1504 * *skb* by *len_diff*, and according to the selected *mode*.
1505 *
Willem de Bruijn14aa3192019-03-22 14:32:54 -04001506 * There are two supported modes at this time:
1507 *
1508 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1509 * (room space is added or removed below the layer 2 header).
Quentin Monnetfa156012018-04-25 18:16:56 +01001510 *
1511 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1512 * (room space is added or removed below the layer 3 header).
1513 *
Willem de Bruijn868d5232019-03-22 14:32:56 -04001514 * The following flags are supported at this time:
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04001515 *
1516 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1517 * Adjusting mss in this way is not allowed for datagrams.
Quentin Monnetfa156012018-04-25 18:16:56 +01001518 *
Willem de Bruijn868d5232019-03-22 14:32:56 -04001519 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **:
1520 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **:
1521 * Any new space is reserved to hold a tunnel header.
1522 * Configure skb offsets and other fields accordingly.
1523 *
1524 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **:
1525 * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
1526 * Use with ENCAP_L3 flags to further specify the tunnel type.
1527 *
Alan Maguire58dfc902019-04-09 15:06:41 +01001528 * * **BPF_F_ADJ_ROOM_ENCAP_L2(len) **:
1529 * Use with ENCAP_L3/L4 flags to further specify the tunnel
1530 * type; **len** is the length of the inner MAC header.
1531 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001532 * A call to this helper is susceptible to change the underlaying
1533 * packet buffer. Therefore, at load time, all checks on pointers
1534 * previously done by the verifier are invalidated and must be
1535 * performed again, if the helper is used in combination with
1536 * direct packet access.
1537 * Return
1538 * 0 on success, or a negative error in case of failure.
1539 *
Quentin Monnetab127042018-04-25 18:16:59 +01001540 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1541 * Description
1542 * Redirect the packet to the endpoint referenced by *map* at
1543 * index *key*. Depending on its type, this *map* can contain
1544 * references to net devices (for forwarding packets through other
1545 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1546 * but this is only implemented for native XDP (with driver
1547 * support) as of this writing).
1548 *
1549 * All values for *flags* are reserved for future usage, and must
1550 * be left at zero.
1551 *
1552 * When used to redirect packets to net devices, this helper
1553 * provides a high performance increase over **bpf_redirect**\ ().
1554 * This is due to various implementation details of the underlying
1555 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1556 * () tries to send packet as a "bulk" to the device.
1557 * Return
1558 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1559 *
1560 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1561 * Description
1562 * Redirect the packet to the socket referenced by *map* (of type
1563 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1564 * egress interfaces can be used for redirection. The
1565 * **BPF_F_INGRESS** value in *flags* is used to make the
1566 * distinction (ingress path is selected if the flag is present,
1567 * egress path otherwise). This is the only flag supported for now.
1568 * Return
1569 * **SK_PASS** on success, or **SK_DROP** on error.
1570 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001571 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001572 * Description
1573 * Add an entry to, or update a *map* referencing sockets. The
1574 * *skops* is used as a new value for the entry associated to
1575 * *key*. *flags* is one of:
1576 *
1577 * **BPF_NOEXIST**
1578 * The entry for *key* must not exist in the map.
1579 * **BPF_EXIST**
1580 * The entry for *key* must already exist in the map.
1581 * **BPF_ANY**
1582 * No condition on the existence of the entry for *key*.
1583 *
1584 * If the *map* has eBPF programs (parser and verdict), those will
1585 * be inherited by the socket being added. If the socket is
1586 * already attached to eBPF programs, this results in an error.
1587 * Return
1588 * 0 on success, or a negative error in case of failure.
1589 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001590 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1591 * Description
1592 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1593 * *delta* (which can be positive or negative). Note that this
1594 * operation modifies the address stored in *xdp_md*\ **->data**,
1595 * so the latter must be loaded only after the helper has been
1596 * called.
1597 *
1598 * The use of *xdp_md*\ **->data_meta** is optional and programs
1599 * are not required to use it. The rationale is that when the
1600 * packet is processed with XDP (e.g. as DoS filter), it is
1601 * possible to push further meta data along with it before passing
1602 * to the stack, and to give the guarantee that an ingress eBPF
1603 * program attached as a TC classifier on the same device can pick
1604 * this up for further post-processing. Since TC works with socket
1605 * buffers, it remains possible to set from XDP the **mark** or
1606 * **priority** pointers, or other pointers for the socket buffer.
1607 * Having this scratch space generic and programmable allows for
1608 * more flexibility as the user is free to store whatever meta
1609 * data they need.
1610 *
1611 * A call to this helper is susceptible to change the underlaying
1612 * packet buffer. Therefore, at load time, all checks on pointers
1613 * previously done by the verifier are invalidated and must be
1614 * performed again, if the helper is used in combination with
1615 * direct packet access.
1616 * Return
1617 * 0 on success, or a negative error in case of failure.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001618 *
1619 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1620 * Description
1621 * Read the value of a perf event counter, and store it into *buf*
1622 * of size *buf_size*. This helper relies on a *map* of type
1623 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1624 * counter is selected when *map* is updated with perf event file
1625 * descriptors. The *map* is an array whose size is the number of
1626 * available CPUs, and each cell contains a value relative to one
1627 * CPU. The value to retrieve is indicated by *flags*, that
1628 * contains the index of the CPU to look up, masked with
1629 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1630 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1631 * current CPU should be retrieved.
1632 *
1633 * This helper behaves in a way close to
1634 * **bpf_perf_event_read**\ () helper, save that instead of
1635 * just returning the value observed, it fills the *buf*
1636 * structure. This allows for additional data to be retrieved: in
1637 * particular, the enabled and running times (in *buf*\
1638 * **->enabled** and *buf*\ **->running**, respectively) are
1639 * copied. In general, **bpf_perf_event_read_value**\ () is
1640 * recommended over **bpf_perf_event_read**\ (), which has some
1641 * ABI issues and provides fewer functionalities.
1642 *
1643 * These values are interesting, because hardware PMU (Performance
1644 * Monitoring Unit) counters are limited resources. When there are
1645 * more PMU based perf events opened than available counters,
1646 * kernel will multiplex these events so each event gets certain
1647 * percentage (but not all) of the PMU time. In case that
1648 * multiplexing happens, the number of samples or counter value
1649 * will not reflect the case compared to when no multiplexing
1650 * occurs. This makes comparison between different runs difficult.
1651 * Typically, the counter value should be normalized before
1652 * comparing to other experiments. The usual normalization is done
1653 * as follows.
1654 *
1655 * ::
1656 *
1657 * normalized_counter = counter * t_enabled / t_running
1658 *
1659 * Where t_enabled is the time enabled for event and t_running is
1660 * the time running for event since last normalization. The
1661 * enabled and running times are accumulated since the perf event
1662 * open. To achieve scaling factor between two invocations of an
1663 * eBPF program, users can can use CPU id as the key (which is
1664 * typical for perf array usage model) to remember the previous
1665 * value and do the calculation inside the eBPF program.
1666 * Return
1667 * 0 on success, or a negative error in case of failure.
1668 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001669 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001670 * Description
1671 * For en eBPF program attached to a perf event, retrieve the
1672 * value of the event counter associated to *ctx* and store it in
1673 * the structure pointed by *buf* and of size *buf_size*. Enabled
1674 * and running times are also stored in the structure (see
1675 * description of helper **bpf_perf_event_read_value**\ () for
1676 * more details).
1677 * Return
1678 * 0 on success, or a negative error in case of failure.
1679 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001680 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001681 * Description
1682 * Emulate a call to **getsockopt()** on the socket associated to
1683 * *bpf_socket*, which must be a full socket. The *level* at
1684 * which the option resides and the name *optname* of the option
1685 * must be specified, see **getsockopt(2)** for more information.
1686 * The retrieved value is stored in the structure pointed by
1687 * *opval* and of length *optlen*.
1688 *
1689 * This helper actually implements a subset of **getsockopt()**.
1690 * It supports the following *level*\ s:
1691 *
1692 * * **IPPROTO_TCP**, which supports *optname*
1693 * **TCP_CONGESTION**.
1694 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1695 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1696 * Return
1697 * 0 on success, or a negative error in case of failure.
1698 *
1699 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1700 * Description
1701 * Used for error injection, this helper uses kprobes to override
1702 * the return value of the probed function, and to set it to *rc*.
1703 * The first argument is the context *regs* on which the kprobe
1704 * works.
1705 *
1706 * This helper works by setting setting the PC (program counter)
1707 * to an override function which is run in place of the original
1708 * probed function. This means the probed function is not run at
1709 * all. The replacement function just returns with the required
1710 * value.
1711 *
1712 * This helper has security implications, and thus is subject to
1713 * restrictions. It is only available if the kernel was compiled
1714 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1715 * option, and in this case it only works on functions tagged with
1716 * **ALLOW_ERROR_INJECTION** in the kernel code.
1717 *
1718 * Also, the helper is only available for the architectures having
1719 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1720 * x86 architecture is the only one to support this feature.
1721 * Return
1722 * 0
1723 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001724 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001725 * Description
1726 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1727 * for the full TCP socket associated to *bpf_sock_ops* to
1728 * *argval*.
1729 *
1730 * The primary use of this field is to determine if there should
1731 * be calls to eBPF programs of type
1732 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1733 * code. A program of the same type can change its value, per
1734 * connection and as necessary, when the connection is
1735 * established. This field is directly accessible for reading, but
1736 * this helper must be used for updates in order to return an
1737 * error if an eBPF program tries to set a callback that is not
1738 * supported in the current kernel.
1739 *
1740 * The supported callback values that *argval* can combine are:
1741 *
1742 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1743 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1744 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1745 *
1746 * Here are some examples of where one could call such eBPF
1747 * program:
1748 *
1749 * * When RTO fires.
1750 * * When a packet is retransmitted.
1751 * * When the connection terminates.
1752 * * When a packet is sent.
1753 * * When a packet is received.
1754 * Return
1755 * Code **-EINVAL** if the socket is not a full TCP socket;
1756 * otherwise, a positive number containing the bits that could not
1757 * be set is returned (which comes down to 0 if all bits were set
1758 * as required).
1759 *
Quentin Monnetab127042018-04-25 18:16:59 +01001760 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1761 * Description
1762 * This helper is used in programs implementing policies at the
1763 * socket level. If the message *msg* is allowed to pass (i.e. if
1764 * the verdict eBPF program returns **SK_PASS**), redirect it to
1765 * the socket referenced by *map* (of type
1766 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1767 * egress interfaces can be used for redirection. The
1768 * **BPF_F_INGRESS** value in *flags* is used to make the
1769 * distinction (ingress path is selected if the flag is present,
1770 * egress path otherwise). This is the only flag supported for now.
1771 * Return
1772 * **SK_PASS** on success, or **SK_DROP** on error.
1773 *
1774 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1775 * Description
1776 * For socket policies, apply the verdict of the eBPF program to
1777 * the next *bytes* (number of bytes) of message *msg*.
1778 *
1779 * For example, this helper can be used in the following cases:
1780 *
1781 * * A single **sendmsg**\ () or **sendfile**\ () system call
1782 * contains multiple logical messages that the eBPF program is
1783 * supposed to read and for which it should apply a verdict.
1784 * * An eBPF program only cares to read the first *bytes* of a
1785 * *msg*. If the message has a large payload, then setting up
1786 * and calling the eBPF program repeatedly for all bytes, even
1787 * though the verdict is already known, would create unnecessary
1788 * overhead.
1789 *
1790 * When called from within an eBPF program, the helper sets a
1791 * counter internal to the BPF infrastructure, that is used to
1792 * apply the last verdict to the next *bytes*. If *bytes* is
1793 * smaller than the current data being processed from a
1794 * **sendmsg**\ () or **sendfile**\ () system call, the first
1795 * *bytes* will be sent and the eBPF program will be re-run with
1796 * the pointer for start of data pointing to byte number *bytes*
1797 * **+ 1**. If *bytes* is larger than the current data being
1798 * processed, then the eBPF verdict will be applied to multiple
1799 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1800 * consumed.
1801 *
1802 * Note that if a socket closes with the internal counter holding
1803 * a non-zero value, this is not a problem because data is not
1804 * being buffered for *bytes* and is sent as it is received.
1805 * Return
1806 * 0
1807 *
1808 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1809 * Description
1810 * For socket policies, prevent the execution of the verdict eBPF
1811 * program for message *msg* until *bytes* (byte number) have been
1812 * accumulated.
1813 *
1814 * This can be used when one needs a specific number of bytes
1815 * before a verdict can be assigned, even if the data spans
1816 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1817 * case would be a user calling **sendmsg**\ () repeatedly with
1818 * 1-byte long message segments. Obviously, this is bad for
1819 * performance, but it is still valid. If the eBPF program needs
1820 * *bytes* bytes to validate a header, this helper can be used to
1821 * prevent the eBPF program to be called again until *bytes* have
1822 * been accumulated.
1823 * Return
1824 * 0
1825 *
1826 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1827 * Description
1828 * For socket policies, pull in non-linear data from user space
1829 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1830 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1831 * respectively.
1832 *
1833 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1834 * *msg* it can only parse data that the (**data**, **data_end**)
1835 * pointers have already consumed. For **sendmsg**\ () hooks this
1836 * is likely the first scatterlist element. But for calls relying
1837 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1838 * be the range (**0**, **0**) because the data is shared with
1839 * user space and by default the objective is to avoid allowing
1840 * user space to modify data while (or after) eBPF verdict is
1841 * being decided. This helper can be used to pull in data and to
1842 * set the start and end pointer to given values. Data will be
1843 * copied if necessary (i.e. if data was not linear and if start
1844 * and end pointers do not point to the same chunk).
1845 *
1846 * A call to this helper is susceptible to change the underlaying
1847 * packet buffer. Therefore, at load time, all checks on pointers
1848 * previously done by the verifier are invalidated and must be
1849 * performed again, if the helper is used in combination with
1850 * direct packet access.
1851 *
1852 * All values for *flags* are reserved for future usage, and must
1853 * be left at zero.
1854 * Return
1855 * 0 on success, or a negative error in case of failure.
1856 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001857 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001858 * Description
1859 * Bind the socket associated to *ctx* to the address pointed by
1860 * *addr*, of length *addr_len*. This allows for making outgoing
1861 * connection from the desired IP address, which can be useful for
1862 * example when all processes inside a cgroup should use one
1863 * single IP address on a host that has multiple IP configured.
1864 *
1865 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1866 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1867 * **AF_INET6**). Looking for a free port to bind to can be
1868 * expensive, therefore binding to port is not permitted by the
1869 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1870 * must be set to zero.
1871 * Return
1872 * 0 on success, or a negative error in case of failure.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01001873 *
1874 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1875 * Description
1876 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1877 * only possible to shrink the packet as of this writing,
1878 * therefore *delta* must be a negative integer.
1879 *
1880 * A call to this helper is susceptible to change the underlaying
1881 * packet buffer. Therefore, at load time, all checks on pointers
1882 * previously done by the verifier are invalidated and must be
1883 * performed again, if the helper is used in combination with
1884 * direct packet access.
1885 * Return
1886 * 0 on success, or a negative error in case of failure.
1887 *
1888 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1889 * Description
1890 * Retrieve the XFRM state (IP transform framework, see also
1891 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1892 *
1893 * The retrieved value is stored in the **struct bpf_xfrm_state**
1894 * pointed by *xfrm_state* and of length *size*.
1895 *
1896 * All values for *flags* are reserved for future usage, and must
1897 * be left at zero.
1898 *
1899 * This helper is available only if the kernel was compiled with
1900 * **CONFIG_XFRM** configuration option.
1901 * Return
1902 * 0 on success, or a negative error in case of failure.
Yonghong Songc195651e2018-04-28 22:28:08 -07001903 *
1904 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1905 * Description
Quentin Monnet79552fb2018-04-30 11:39:04 +01001906 * Return a user or a kernel stack in bpf program provided buffer.
1907 * To achieve this, the helper needs *ctx*, which is a pointer
1908 * to the context on which the tracing program is executed.
1909 * To store the stacktrace, the bpf program provides *buf* with
1910 * a nonnegative *size*.
Yonghong Songc195651e2018-04-28 22:28:08 -07001911 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001912 * The last argument, *flags*, holds the number of stack frames to
1913 * skip (from 0 to 255), masked with
1914 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1915 * the following flags:
Yonghong Songc195651e2018-04-28 22:28:08 -07001916 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001917 * **BPF_F_USER_STACK**
1918 * Collect a user space stack instead of a kernel stack.
1919 * **BPF_F_USER_BUILD_ID**
1920 * Collect buildid+offset instead of ips for user stack,
1921 * only valid if **BPF_F_USER_STACK** is also specified.
Yonghong Songc195651e2018-04-28 22:28:08 -07001922 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001923 * **bpf_get_stack**\ () can collect up to
1924 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1925 * to sufficient large buffer size. Note that
1926 * this limit can be controlled with the **sysctl** program, and
1927 * that it should be manually increased in order to profile long
1928 * user stacks (such as stacks for Java programs). To do so, use:
Yonghong Songc195651e2018-04-28 22:28:08 -07001929 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001930 * ::
Yonghong Songc195651e2018-04-28 22:28:08 -07001931 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001932 * # sysctl kernel.perf_event_max_stack=<new value>
Yonghong Songc195651e2018-04-28 22:28:08 -07001933 * Return
Quentin Monnet7a279e92018-05-29 12:27:44 +01001934 * A non-negative value equal to or less than *size* on success,
1935 * or a negative error in case of failure.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001936 *
Quentin Monnet2bae79d2018-07-12 12:52:22 +01001937 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001938 * Description
1939 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1940 * it provides an easy way to load *len* bytes from *offset*
1941 * from the packet associated to *skb*, into the buffer pointed
1942 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1943 * a fifth argument *start_header* exists in order to select a
1944 * base offset to start from. *start_header* can be one of:
1945 *
1946 * **BPF_HDR_START_MAC**
1947 * Base offset to load data from is *skb*'s mac header.
1948 * **BPF_HDR_START_NET**
1949 * Base offset to load data from is *skb*'s network header.
1950 *
1951 * In general, "direct packet access" is the preferred method to
1952 * access packet data, however, this helper is in particular useful
1953 * in socket filters where *skb*\ **->data** does not always point
1954 * to the start of the mac header and where "direct packet access"
1955 * is not available.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001956 * Return
1957 * 0 on success, or a negative error in case of failure.
1958 *
David Ahern87f5fc72018-05-09 20:34:26 -07001959 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1960 * Description
1961 * Do FIB lookup in kernel tables using parameters in *params*.
1962 * If lookup is successful and result shows packet is to be
1963 * forwarded, the neighbor tables are searched for the nexthop.
1964 * If successful (ie., FIB lookup shows forwarding and nexthop
David Ahernfa898d72018-05-29 10:58:07 -07001965 * is resolved), the nexthop address is returned in ipv4_dst
1966 * or ipv6_dst based on family, smac is set to mac address of
1967 * egress device, dmac is set to nexthop mac address, rt_metric
David Ahern4c795792018-06-26 16:21:18 -07001968 * is set to metric from route (IPv4/IPv6 only), and ifindex
1969 * is set to the device index of the nexthop from the FIB lookup.
David Ahern87f5fc72018-05-09 20:34:26 -07001970 *
Quentin Monnet90b10232018-12-03 12:13:35 +00001971 * *plen* argument is the size of the passed in struct.
1972 * *flags* argument can be a combination of one or more of the
1973 * following values:
David Ahern87f5fc72018-05-09 20:34:26 -07001974 *
Quentin Monnet7a279e92018-05-29 12:27:44 +01001975 * **BPF_FIB_LOOKUP_DIRECT**
1976 * Do a direct table lookup vs full lookup using FIB
1977 * rules.
1978 * **BPF_FIB_LOOKUP_OUTPUT**
1979 * Perform lookup from an egress perspective (default is
1980 * ingress).
David Ahern87f5fc72018-05-09 20:34:26 -07001981 *
Quentin Monnet90b10232018-12-03 12:13:35 +00001982 * *ctx* is either **struct xdp_md** for XDP programs or
1983 * **struct sk_buff** tc cls_act programs.
1984 * Return
David Ahern4c795792018-06-26 16:21:18 -07001985 * * < 0 if any input argument is invalid
1986 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1987 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
Quentin Monnet2bae79d2018-07-12 12:52:22 +01001988 * packet is not forwarded or needs assist from full stack
John Fastabend81110382018-05-14 10:00:17 -07001989 *
1990 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1991 * Description
1992 * Add an entry to, or update a sockhash *map* referencing sockets.
1993 * The *skops* is used as a new value for the entry associated to
1994 * *key*. *flags* is one of:
1995 *
1996 * **BPF_NOEXIST**
1997 * The entry for *key* must not exist in the map.
1998 * **BPF_EXIST**
1999 * The entry for *key* must already exist in the map.
2000 * **BPF_ANY**
2001 * No condition on the existence of the entry for *key*.
2002 *
2003 * If the *map* has eBPF programs (parser and verdict), those will
2004 * be inherited by the socket being added. If the socket is
2005 * already attached to eBPF programs, this results in an error.
2006 * Return
2007 * 0 on success, or a negative error in case of failure.
2008 *
2009 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2010 * Description
2011 * This helper is used in programs implementing policies at the
2012 * socket level. If the message *msg* is allowed to pass (i.e. if
2013 * the verdict eBPF program returns **SK_PASS**), redirect it to
2014 * the socket referenced by *map* (of type
2015 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2016 * egress interfaces can be used for redirection. The
2017 * **BPF_F_INGRESS** value in *flags* is used to make the
2018 * distinction (ingress path is selected if the flag is present,
2019 * egress path otherwise). This is the only flag supported for now.
2020 * Return
2021 * **SK_PASS** on success, or **SK_DROP** on error.
2022 *
2023 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2024 * Description
2025 * This helper is used in programs implementing policies at the
2026 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2027 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2028 * to the socket referenced by *map* (of type
2029 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2030 * egress interfaces can be used for redirection. The
2031 * **BPF_F_INGRESS** value in *flags* is used to make the
2032 * distinction (ingress path is selected if the flag is present,
2033 * egress otherwise). This is the only flag supported for now.
2034 * Return
2035 * **SK_PASS** on success, or **SK_DROP** on error.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002036 *
2037 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2038 * Description
2039 * Encapsulate the packet associated to *skb* within a Layer 3
2040 * protocol header. This header is provided in the buffer at
2041 * address *hdr*, with *len* its size in bytes. *type* indicates
2042 * the protocol of the header and can be one of:
2043 *
2044 * **BPF_LWT_ENCAP_SEG6**
2045 * IPv6 encapsulation with Segment Routing Header
2046 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2047 * the IPv6 header is computed by the kernel.
2048 * **BPF_LWT_ENCAP_SEG6_INLINE**
2049 * Only works if *skb* contains an IPv6 packet. Insert a
2050 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2051 * the IPv6 header.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002052 * **BPF_LWT_ENCAP_IP**
2053 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2054 * must be IPv4 or IPv6, followed by zero or more
2055 * additional headers, up to LWT_BPF_MAX_HEADROOM total
2056 * bytes in all prepended headers. Please note that
2057 * if skb_is_gso(skb) is true, no more than two headers
2058 * can be prepended, and the inner header, if present,
2059 * should be either GRE or UDP/GUE.
2060 *
2061 * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of
2062 * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called
2063 * by bpf programs of types BPF_PROG_TYPE_LWT_IN and
2064 * BPF_PROG_TYPE_LWT_XMIT.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002065 *
2066 * A call to this helper is susceptible to change the underlaying
2067 * packet buffer. Therefore, at load time, all checks on pointers
2068 * previously done by the verifier are invalidated and must be
2069 * performed again, if the helper is used in combination with
2070 * direct packet access.
2071 * Return
2072 * 0 on success, or a negative error in case of failure.
2073 *
2074 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2075 * Description
2076 * Store *len* bytes from address *from* into the packet
2077 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2078 * inside the outermost IPv6 Segment Routing Header can be
2079 * modified through this helper.
2080 *
2081 * A call to this helper is susceptible to change the underlaying
2082 * packet buffer. Therefore, at load time, all checks on pointers
2083 * previously done by the verifier are invalidated and must be
2084 * performed again, if the helper is used in combination with
2085 * direct packet access.
2086 * Return
2087 * 0 on success, or a negative error in case of failure.
2088 *
2089 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2090 * Description
2091 * Adjust the size allocated to TLVs in the outermost IPv6
2092 * Segment Routing Header contained in the packet associated to
2093 * *skb*, at position *offset* by *delta* bytes. Only offsets
2094 * after the segments are accepted. *delta* can be as well
2095 * positive (growing) as negative (shrinking).
2096 *
2097 * A call to this helper is susceptible to change the underlaying
2098 * packet buffer. Therefore, at load time, all checks on pointers
2099 * previously done by the verifier are invalidated and must be
2100 * performed again, if the helper is used in combination with
2101 * direct packet access.
2102 * Return
2103 * 0 on success, or a negative error in case of failure.
2104 *
2105 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2106 * Description
2107 * Apply an IPv6 Segment Routing action of type *action* to the
2108 * packet associated to *skb*. Each action takes a parameter
2109 * contained at address *param*, and of length *param_len* bytes.
2110 * *action* can be one of:
2111 *
2112 * **SEG6_LOCAL_ACTION_END_X**
2113 * End.X action: Endpoint with Layer-3 cross-connect.
2114 * Type of *param*: **struct in6_addr**.
2115 * **SEG6_LOCAL_ACTION_END_T**
2116 * End.T action: Endpoint with specific IPv6 table lookup.
2117 * Type of *param*: **int**.
2118 * **SEG6_LOCAL_ACTION_END_B6**
2119 * End.B6 action: Endpoint bound to an SRv6 policy.
2120 * Type of param: **struct ipv6_sr_hdr**.
2121 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2122 * End.B6.Encap action: Endpoint bound to an SRv6
2123 * encapsulation policy.
2124 * Type of param: **struct ipv6_sr_hdr**.
2125 *
2126 * A call to this helper is susceptible to change the underlaying
2127 * packet buffer. Therefore, at load time, all checks on pointers
2128 * previously done by the verifier are invalidated and must be
2129 * performed again, if the helper is used in combination with
2130 * direct packet access.
2131 * Return
2132 * 0 on success, or a negative error in case of failure.
Sean Youngf4364dc2018-05-27 12:24:09 +01002133 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002134 * int bpf_rc_repeat(void *ctx)
2135 * Description
2136 * This helper is used in programs implementing IR decoding, to
2137 * report a successfully decoded repeat key message. This delays
2138 * the generation of a key up event for previously generated
2139 * key down event.
2140 *
2141 * Some IR protocols like NEC have a special IR message for
2142 * repeating last button, for when a button is held down.
2143 *
2144 * The *ctx* should point to the lirc sample as passed into
2145 * the program.
2146 *
2147 * This helper is only available is the kernel was compiled with
2148 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2149 * "**y**".
2150 * Return
2151 * 0
2152 *
Sean Youngf4364dc2018-05-27 12:24:09 +01002153 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2154 * Description
2155 * This helper is used in programs implementing IR decoding, to
2156 * report a successfully decoded key press with *scancode*,
2157 * *toggle* value in the given *protocol*. The scancode will be
2158 * translated to a keycode using the rc keymap, and reported as
2159 * an input key down event. After a period a key up event is
2160 * generated. This period can be extended by calling either
Quentin Monnet90b10232018-12-03 12:13:35 +00002161 * **bpf_rc_keydown**\ () again with the same values, or calling
2162 * **bpf_rc_repeat**\ ().
Sean Youngf4364dc2018-05-27 12:24:09 +01002163 *
2164 * Some protocols include a toggle bit, in case the button was
2165 * released and pressed again between consecutive scancodes.
2166 *
2167 * The *ctx* should point to the lirc sample as passed into
2168 * the program.
2169 *
2170 * The *protocol* is the decoded protocol number (see
2171 * **enum rc_proto** for some predefined values).
2172 *
2173 * This helper is only available is the kernel was compiled with
2174 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2175 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002176 * Return
2177 * 0
2178 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002179 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002180 * Description
2181 * Return the cgroup v2 id of the socket associated with the *skb*.
2182 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2183 * helper for cgroup v1 by providing a tag resp. identifier that
2184 * can be matched on or used for map lookups e.g. to implement
2185 * policy. The cgroup v2 id of a given path in the hierarchy is
2186 * exposed in user space through the f_handle API in order to get
2187 * to the same 64-bit id.
2188 *
2189 * This helper can be used on TC egress path, but not on ingress,
2190 * and is available only if the kernel was compiled with the
2191 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2192 * Return
2193 * The id is returned or 0 in case the id could not be retrieved.
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002194 *
2195 * u64 bpf_get_current_cgroup_id(void)
2196 * Return
2197 * A 64-bit integer containing the current cgroup id based
2198 * on the cgroup within which the current task is running.
Roman Gushchincd339432018-08-02 14:27:24 -07002199 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002200 * void *bpf_get_local_storage(void *map, u64 flags)
Roman Gushchincd339432018-08-02 14:27:24 -07002201 * Description
2202 * Get the pointer to the local storage area.
2203 * The type and the size of the local storage is defined
2204 * by the *map* argument.
2205 * The *flags* meaning is specific for each map type,
2206 * and has to be 0 for cgroup local storage.
2207 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002208 * Depending on the BPF program type, a local storage area
2209 * can be shared between multiple instances of the BPF program,
Roman Gushchincd339432018-08-02 14:27:24 -07002210 * running simultaneously.
2211 *
2212 * A user should care about the synchronization by himself.
Quentin Monnet90b10232018-12-03 12:13:35 +00002213 * For example, by using the **BPF_STX_XADD** instruction to alter
Roman Gushchincd339432018-08-02 14:27:24 -07002214 * the shared data.
2215 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002216 * A pointer to the local storage area.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002217 *
2218 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2219 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002220 * Select a **SO_REUSEPORT** socket from a
2221 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2222 * It checks the selected socket is matching the incoming
2223 * request in the socket buffer.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002224 * Return
2225 * 0 on success, or a negative error in case of failure.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002226 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002227 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2228 * Description
2229 * Return id of cgroup v2 that is ancestor of cgroup associated
2230 * with the *skb* at the *ancestor_level*. The root cgroup is at
2231 * *ancestor_level* zero and each step down the hierarchy
2232 * increments the level. If *ancestor_level* == level of cgroup
2233 * associated with *skb*, then return value will be same as that
2234 * of **bpf_skb_cgroup_id**\ ().
2235 *
2236 * The helper is useful to implement policies based on cgroups
2237 * that are upper in hierarchy than immediate cgroup associated
2238 * with *skb*.
2239 *
2240 * The format of returned id and helper limitations are same as in
2241 * **bpf_skb_cgroup_id**\ ().
2242 * Return
2243 * The id is returned or 0 in case the id could not be retrieved.
2244 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002245 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002246 * Description
2247 * Look for TCP socket matching *tuple*, optionally in a child
2248 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002249 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002250 *
2251 * The *ctx* should point to the context of the program, such as
2252 * the skb or socket (depending on the hook in use). This is used
2253 * to determine the base network namespace for the lookup.
2254 *
2255 * *tuple_size* must be one of:
2256 *
2257 * **sizeof**\ (*tuple*\ **->ipv4**)
2258 * Look for an IPv4 socket.
2259 * **sizeof**\ (*tuple*\ **->ipv6**)
2260 * Look for an IPv6 socket.
2261 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002262 * If the *netns* is a negative signed 32-bit integer, then the
2263 * socket lookup table in the netns associated with the *ctx* will
2264 * will be used. For the TC hooks, this is the netns of the device
2265 * in the skb. For socket hooks, this is the netns of the socket.
2266 * If *netns* is any other signed 32-bit value greater than or
2267 * equal to zero then it specifies the ID of the netns relative to
2268 * the netns associated with the *ctx*. *netns* values beyond the
2269 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002270 *
2271 * All values for *flags* are reserved for future usage, and must
2272 * be left at zero.
2273 *
2274 * This helper is available only if the kernel was compiled with
2275 * **CONFIG_NET** configuration option.
2276 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002277 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2278 * For sockets with reuseport option, the **struct bpf_sock**
2279 * result is from **reuse->socks**\ [] using the hash of the tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002280 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002281 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002282 * Description
2283 * Look for UDP socket matching *tuple*, optionally in a child
2284 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002285 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002286 *
2287 * The *ctx* should point to the context of the program, such as
2288 * the skb or socket (depending on the hook in use). This is used
2289 * to determine the base network namespace for the lookup.
2290 *
2291 * *tuple_size* must be one of:
2292 *
2293 * **sizeof**\ (*tuple*\ **->ipv4**)
2294 * Look for an IPv4 socket.
2295 * **sizeof**\ (*tuple*\ **->ipv6**)
2296 * Look for an IPv6 socket.
2297 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002298 * If the *netns* is a negative signed 32-bit integer, then the
2299 * socket lookup table in the netns associated with the *ctx* will
2300 * will be used. For the TC hooks, this is the netns of the device
2301 * in the skb. For socket hooks, this is the netns of the socket.
2302 * If *netns* is any other signed 32-bit value greater than or
2303 * equal to zero then it specifies the ID of the netns relative to
2304 * the netns associated with the *ctx*. *netns* values beyond the
2305 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002306 *
2307 * All values for *flags* are reserved for future usage, and must
2308 * be left at zero.
2309 *
2310 * This helper is available only if the kernel was compiled with
2311 * **CONFIG_NET** configuration option.
2312 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002313 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2314 * For sockets with reuseport option, the **struct bpf_sock**
2315 * result is from **reuse->socks**\ [] using the hash of the tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002316 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002317 * int bpf_sk_release(struct bpf_sock *sock)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002318 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002319 * Release the reference held by *sock*. *sock* must be a
2320 * non-**NULL** pointer that was returned from
2321 * **bpf_sk_lookup_xxx**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002322 * Return
2323 * 0 on success, or a negative error in case of failure.
John Fastabend6fff6072018-10-19 19:56:49 -07002324 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002325 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2326 * Description
2327 * Push an element *value* in *map*. *flags* is one of:
2328 *
2329 * **BPF_EXIST**
2330 * If the queue/stack is full, the oldest element is
2331 * removed to make room for this.
2332 * Return
2333 * 0 on success, or a negative error in case of failure.
2334 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002335 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2336 * Description
2337 * Pop an element from *map*.
2338 * Return
2339 * 0 on success, or a negative error in case of failure.
2340 *
2341 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2342 * Description
2343 * Get an element from *map* without removing it.
2344 * Return
2345 * 0 on success, or a negative error in case of failure.
2346 *
John Fastabend6fff6072018-10-19 19:56:49 -07002347 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2348 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002349 * For socket policies, insert *len* bytes into *msg* at offset
John Fastabend6fff6072018-10-19 19:56:49 -07002350 * *start*.
2351 *
2352 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
Quentin Monnet90b10232018-12-03 12:13:35 +00002353 * *msg* it may want to insert metadata or options into the *msg*.
John Fastabend6fff6072018-10-19 19:56:49 -07002354 * This can later be read and used by any of the lower layer BPF
2355 * hooks.
2356 *
2357 * This helper may fail if under memory pressure (a malloc
2358 * fails) in these cases BPF programs will get an appropriate
2359 * error and BPF programs will need to handle them.
John Fastabend6fff6072018-10-19 19:56:49 -07002360 * Return
2361 * 0 on success, or a negative error in case of failure.
John Fastabend7246d8e2018-11-26 14:16:17 -08002362 *
2363 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
Quentin Monnet90b10232018-12-03 12:13:35 +00002364 * Description
John Fastabend7246d8e2018-11-26 14:16:17 -08002365 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2366 * This may result in **ENOMEM** errors under certain situations if
2367 * an allocation and copy are required due to a full ring buffer.
2368 * However, the helper will try to avoid doing the allocation
2369 * if possible. Other errors can occur if input parameters are
Quentin Monnet90b10232018-12-03 12:13:35 +00002370 * invalid either due to *start* byte not being valid part of *msg*
John Fastabend7246d8e2018-11-26 14:16:17 -08002371 * payload and/or *pop* value being to large.
John Fastabend7246d8e2018-11-26 14:16:17 -08002372 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002373 * 0 on success, or a negative error in case of failure.
Sean Young01d32402018-12-06 13:01:03 +00002374 *
2375 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2376 * Description
2377 * This helper is used in programs implementing IR decoding, to
2378 * report a successfully decoded pointer movement.
2379 *
2380 * The *ctx* should point to the lirc sample as passed into
2381 * the program.
2382 *
2383 * This helper is only available is the kernel was compiled with
2384 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2385 * "**y**".
2386 * Return
2387 * 0
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002388 *
Quentin Monnet0eb09782019-03-14 12:38:40 +00002389 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2390 * Description
2391 * Acquire a spinlock represented by the pointer *lock*, which is
2392 * stored as part of a value of a map. Taking the lock allows to
2393 * safely update the rest of the fields in that value. The
2394 * spinlock can (and must) later be released with a call to
2395 * **bpf_spin_unlock**\ (\ *lock*\ ).
2396 *
2397 * Spinlocks in BPF programs come with a number of restrictions
2398 * and constraints:
2399 *
2400 * * **bpf_spin_lock** objects are only allowed inside maps of
2401 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2402 * list could be extended in the future).
2403 * * BTF description of the map is mandatory.
2404 * * The BPF program can take ONE lock at a time, since taking two
2405 * or more could cause dead locks.
2406 * * Only one **struct bpf_spin_lock** is allowed per map element.
2407 * * When the lock is taken, calls (either BPF to BPF or helpers)
2408 * are not allowed.
2409 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2410 * allowed inside a spinlock-ed region.
2411 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2412 * the lock, on all execution paths, before it returns.
2413 * * The BPF program can access **struct bpf_spin_lock** only via
2414 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2415 * helpers. Loading or storing data into the **struct
2416 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2417 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2418 * of the map value must be a struct and have **struct
2419 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2420 * Nested lock inside another struct is not allowed.
2421 * * The **struct bpf_spin_lock** *lock* field in a map value must
2422 * be aligned on a multiple of 4 bytes in that value.
2423 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2424 * the **bpf_spin_lock** field to user space.
2425 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2426 * a BPF program, do not update the **bpf_spin_lock** field.
2427 * * **bpf_spin_lock** cannot be on the stack or inside a
2428 * networking packet (it can only be inside of a map values).
2429 * * **bpf_spin_lock** is available to root only.
2430 * * Tracing programs and socket filter programs cannot use
2431 * **bpf_spin_lock**\ () due to insufficient preemption checks
2432 * (but this may change in the future).
2433 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2434 * Return
2435 * 0
2436 *
2437 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2438 * Description
2439 * Release the *lock* previously locked by a call to
2440 * **bpf_spin_lock**\ (\ *lock*\ ).
2441 * Return
2442 * 0
2443 *
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002444 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2445 * Description
2446 * This helper gets a **struct bpf_sock** pointer such
Quentin Monnet62369db2019-03-14 12:38:39 +00002447 * that all the fields in this **bpf_sock** can be accessed.
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002448 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002449 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002450 * case of failure.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002451 *
2452 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2453 * Description
2454 * This helper gets a **struct bpf_tcp_sock** pointer from a
2455 * **struct bpf_sock** pointer.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002456 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002457 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002458 * case of failure.
brakmof7c917b2019-03-01 12:38:46 -08002459 *
2460 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
Quentin Monnet62369db2019-03-14 12:38:39 +00002461 * Description
2462 * Set ECN (Explicit Congestion Notification) field of IP header
2463 * to **CE** (Congestion Encountered) if current value is **ECT**
2464 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2465 * and IPv4.
2466 * Return
2467 * 1 if the **CE** flag is set (either by the current helper call
2468 * or because it was already present), 0 if it is not set.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002469 *
2470 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2471 * Description
Quentin Monnet62369db2019-03-14 12:38:39 +00002472 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2473 * **bpf_sk_release**\ () is unnecessary and not allowed.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002474 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002475 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002476 * case of failure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002477 *
2478 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2479 * Description
2480 * Look for TCP socket matching *tuple*, optionally in a child
2481 * network namespace *netns*. The return value must be checked,
2482 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2483 *
2484 * This function is identical to bpf_sk_lookup_tcp, except that it
2485 * also returns timewait or request sockets. Use bpf_sk_fullsock
2486 * or bpf_tcp_socket to access the full structure.
2487 *
2488 * This helper is available only if the kernel was compiled with
2489 * **CONFIG_NET** configuration option.
2490 * Return
2491 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2492 * For sockets with reuseport option, the **struct bpf_sock**
2493 * result is from **reuse->socks**\ [] using the hash of the tuple.
Lorenz Bauer39904082019-03-22 09:54:02 +08002494 *
2495 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2496 * Description
2497 * Check whether iph and th contain a valid SYN cookie ACK for
2498 * the listening socket in sk.
2499 *
2500 * iph points to the start of the IPv4 or IPv6 header, while
2501 * iph_len contains sizeof(struct iphdr) or sizeof(struct ip6hdr).
2502 *
2503 * th points to the start of the TCP header, while th_len contains
2504 * sizeof(struct tcphdr).
2505 *
2506 * Return
2507 * 0 if iph and th are a valid SYN cookie ACK, or a negative error
2508 * otherwise.
Andrey Ignatov808649f2019-02-27 13:28:48 -08002509 *
2510 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2511 * Description
2512 * Get name of sysctl in /proc/sys/ and copy it into provided by
2513 * program buffer *buf* of size *buf_len*.
2514 *
2515 * The buffer is always NUL terminated, unless it's zero-sized.
2516 *
2517 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2518 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2519 * only (e.g. "tcp_mem").
2520 * Return
2521 * Number of character copied (not including the trailing NUL).
2522 *
2523 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2524 * truncated name in this case).
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002525 *
2526 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2527 * Description
2528 * Get current value of sysctl as it is presented in /proc/sys
2529 * (incl. newline, etc), and copy it as a string into provided
2530 * by program buffer *buf* of size *buf_len*.
2531 *
2532 * The whole value is copied, no matter what file position user
2533 * space issued e.g. sys_read at.
2534 *
2535 * The buffer is always NUL terminated, unless it's zero-sized.
2536 * Return
2537 * Number of character copied (not including the trailing NUL).
2538 *
2539 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2540 * truncated name in this case).
2541 *
2542 * **-EINVAL** if current value was unavailable, e.g. because
2543 * sysctl is uninitialized and read returns -EIO for it.
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002544 *
2545 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2546 * Description
2547 * Get new value being written by user space to sysctl (before
2548 * the actual write happens) and copy it as a string into
2549 * provided by program buffer *buf* of size *buf_len*.
2550 *
2551 * User space may write new value at file position > 0.
2552 *
2553 * The buffer is always NUL terminated, unless it's zero-sized.
2554 * Return
2555 * Number of character copied (not including the trailing NUL).
2556 *
2557 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2558 * truncated name in this case).
2559 *
2560 * **-EINVAL** if sysctl is being read.
2561 *
2562 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2563 * Description
2564 * Override new value being written by user space to sysctl with
2565 * value provided by program in buffer *buf* of size *buf_len*.
2566 *
2567 * *buf* should contain a string in same form as provided by user
2568 * space on sysctl write.
2569 *
2570 * User space may write new value at file position > 0. To override
2571 * the whole sysctl value file position should be set to zero.
2572 * Return
2573 * 0 on success.
2574 *
2575 * **-E2BIG** if the *buf_len* is too big.
2576 *
2577 * **-EINVAL** if sysctl is being read.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002578 *
2579 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2580 * Description
2581 * Convert the initial part of the string from buffer *buf* of
2582 * size *buf_len* to a long integer according to the given base
2583 * and save the result in *res*.
2584 *
2585 * The string may begin with an arbitrary amount of white space
2586 * (as determined by isspace(3)) followed by a single optional '-'
2587 * sign.
2588 *
2589 * Five least significant bits of *flags* encode base, other bits
2590 * are currently unused.
2591 *
2592 * Base must be either 8, 10, 16 or 0 to detect it automatically
2593 * similar to user space strtol(3).
2594 * Return
2595 * Number of characters consumed on success. Must be positive but
2596 * no more than buf_len.
2597 *
2598 * **-EINVAL** if no valid digits were found or unsupported base
2599 * was provided.
2600 *
2601 * **-ERANGE** if resulting value was out of range.
2602 *
2603 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2604 * Description
2605 * Convert the initial part of the string from buffer *buf* of
2606 * size *buf_len* to an unsigned long integer according to the
2607 * given base and save the result in *res*.
2608 *
2609 * The string may begin with an arbitrary amount of white space
2610 * (as determined by isspace(3)).
2611 *
2612 * Five least significant bits of *flags* encode base, other bits
2613 * are currently unused.
2614 *
2615 * Base must be either 8, 10, 16 or 0 to detect it automatically
2616 * similar to user space strtoul(3).
2617 * Return
2618 * Number of characters consumed on success. Must be positive but
2619 * no more than buf_len.
2620 *
2621 * **-EINVAL** if no valid digits were found or unsupported base
2622 * was provided.
2623 *
2624 * **-ERANGE** if resulting value was out of range.
Thomas Grafebb676d2016-10-27 11:23:51 +02002625 */
2626#define __BPF_FUNC_MAPPER(FN) \
2627 FN(unspec), \
2628 FN(map_lookup_elem), \
2629 FN(map_update_elem), \
2630 FN(map_delete_elem), \
2631 FN(probe_read), \
2632 FN(ktime_get_ns), \
2633 FN(trace_printk), \
2634 FN(get_prandom_u32), \
2635 FN(get_smp_processor_id), \
2636 FN(skb_store_bytes), \
2637 FN(l3_csum_replace), \
2638 FN(l4_csum_replace), \
2639 FN(tail_call), \
2640 FN(clone_redirect), \
2641 FN(get_current_pid_tgid), \
2642 FN(get_current_uid_gid), \
2643 FN(get_current_comm), \
2644 FN(get_cgroup_classid), \
2645 FN(skb_vlan_push), \
2646 FN(skb_vlan_pop), \
2647 FN(skb_get_tunnel_key), \
2648 FN(skb_set_tunnel_key), \
2649 FN(perf_event_read), \
2650 FN(redirect), \
2651 FN(get_route_realm), \
2652 FN(perf_event_output), \
2653 FN(skb_load_bytes), \
2654 FN(get_stackid), \
2655 FN(csum_diff), \
2656 FN(skb_get_tunnel_opt), \
2657 FN(skb_set_tunnel_opt), \
2658 FN(skb_change_proto), \
2659 FN(skb_change_type), \
2660 FN(skb_under_cgroup), \
2661 FN(get_hash_recalc), \
2662 FN(get_current_task), \
2663 FN(probe_write_user), \
2664 FN(current_task_under_cgroup), \
2665 FN(skb_change_tail), \
2666 FN(skb_pull_data), \
2667 FN(csum_update), \
2668 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +01002669 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08002670 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +00002671 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -07002672 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -07002673 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +02002674 FN(get_socket_uid), \
Lawrence Brakmo8c4b4c72017-06-30 20:02:46 -07002675 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +02002676 FN(setsockopt), \
John Fastabend97f91a72017-07-17 09:29:18 -07002677 FN(skb_adjust_room), \
John Fastabend174a79f2017-08-15 22:32:47 -07002678 FN(redirect_map), \
2679 FN(sk_redirect_map), \
2680 FN(sock_map_update), \
Yonghong Song908432c2017-10-05 09:19:20 -07002681 FN(xdp_adjust_meta), \
Yonghong Song4bebdc72017-10-05 09:19:22 -07002682 FN(perf_event_read_value), \
Lawrence Brakmocd86d1f2017-10-20 11:05:40 -07002683 FN(perf_prog_read_value), \
Josef Bacik9802d862017-12-11 11:36:48 -05002684 FN(getsockopt), \
Lawrence Brakmob13d8802018-01-25 16:14:10 -08002685 FN(override_return), \
John Fastabend4f738ad2018-03-18 12:57:10 -07002686 FN(sock_ops_cb_flags_set), \
John Fastabend2a100312018-03-18 12:57:15 -07002687 FN(msg_redirect_map), \
John Fastabend91843d52018-03-18 12:57:20 -07002688 FN(msg_apply_bytes), \
John Fastabend015632b2018-03-18 12:57:25 -07002689 FN(msg_cork_bytes), \
Andrey Ignatovd74bad42018-03-30 15:08:05 -07002690 FN(msg_pull_data), \
Nikita V. Shirokovb32cc5b2018-04-17 21:42:13 -07002691 FN(bind), \
Eyal Birger12bed762018-04-24 17:50:29 +03002692 FN(xdp_adjust_tail), \
Yonghong Songc195651e2018-04-28 22:28:08 -07002693 FN(skb_get_xfrm_state), \
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002694 FN(get_stack), \
David Ahern87f5fc72018-05-09 20:34:26 -07002695 FN(skb_load_bytes_relative), \
John Fastabend81110382018-05-14 10:00:17 -07002696 FN(fib_lookup), \
2697 FN(sock_hash_update), \
2698 FN(msg_redirect_hash), \
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002699 FN(sk_redirect_hash), \
2700 FN(lwt_push_encap), \
2701 FN(lwt_seg6_store_bytes), \
2702 FN(lwt_seg6_adjust_srh), \
Sean Youngf4364dc2018-05-27 12:24:09 +01002703 FN(lwt_seg6_action), \
2704 FN(rc_repeat), \
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002705 FN(rc_keydown), \
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002706 FN(skb_cgroup_id), \
Roman Gushchincd339432018-08-02 14:27:24 -07002707 FN(get_current_cgroup_id), \
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002708 FN(get_local_storage), \
Andrey Ignatov77236282018-08-12 10:49:27 -07002709 FN(sk_select_reuseport), \
Joe Stringer6acc9b42018-10-02 13:35:36 -07002710 FN(skb_ancestor_cgroup_id), \
2711 FN(sk_lookup_tcp), \
2712 FN(sk_lookup_udp), \
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +02002713 FN(sk_release), \
2714 FN(map_push_elem), \
2715 FN(map_pop_elem), \
John Fastabend6fff6072018-10-19 19:56:49 -07002716 FN(map_peek_elem), \
John Fastabend7246d8e2018-11-26 14:16:17 -08002717 FN(msg_push_data), \
Sean Young01d32402018-12-06 13:01:03 +00002718 FN(msg_pop_data), \
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08002719 FN(rc_pointer_rel), \
2720 FN(spin_lock), \
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002721 FN(spin_unlock), \
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002722 FN(sk_fullsock), \
brakmof7c917b2019-03-01 12:38:46 -08002723 FN(tcp_sock), \
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002724 FN(skb_ecn_set_ce), \
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002725 FN(get_listener_sock), \
Lorenz Bauer39904082019-03-22 09:54:02 +08002726 FN(skc_lookup_tcp), \
Andrey Ignatov808649f2019-02-27 13:28:48 -08002727 FN(tcp_check_syncookie), \
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002728 FN(sysctl_get_name), \
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002729 FN(sysctl_get_current_value), \
2730 FN(sysctl_get_new_value), \
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002731 FN(sysctl_set_new_value), \
2732 FN(strtol), \
2733 FN(strtoul),
Thomas Grafebb676d2016-10-27 11:23:51 +02002734
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002735/* integer value in 'imm' field of BPF_CALL instruction selects which helper
2736 * function eBPF program intends to call
2737 */
Thomas Grafebb676d2016-10-27 11:23:51 +02002738#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002739enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +02002740 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002741 __BPF_FUNC_MAX_ID,
2742};
Thomas Grafebb676d2016-10-27 11:23:51 +02002743#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002744
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002745/* All flags used by eBPF helper functions, placed here. */
2746
2747/* BPF_FUNC_skb_store_bytes flags. */
2748#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
Daniel Borkmann8afd54c2016-03-04 15:15:03 +01002749#define BPF_F_INVALIDATE_HASH (1ULL << 1)
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002750
2751/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2752 * First 4 bits are for passing the header field size.
2753 */
2754#define BPF_F_HDR_FIELD_MASK 0xfULL
2755
2756/* BPF_FUNC_l4_csum_replace flags. */
2757#define BPF_F_PSEUDO_HDR (1ULL << 4)
Daniel Borkmann2f729592016-02-19 23:05:26 +01002758#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
Daniel Borkmannd1b662a2017-01-24 01:06:28 +01002759#define BPF_F_MARK_ENFORCE (1ULL << 6)
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002760
2761/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2762#define BPF_F_INGRESS (1ULL << 0)
2763
Daniel Borkmannc6c33452016-01-11 01:16:39 +01002764/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2765#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2766
Yonghong Songc195651e2018-04-28 22:28:08 -07002767/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002768#define BPF_F_SKIP_FIELD_MASK 0xffULL
2769#define BPF_F_USER_STACK (1ULL << 8)
Yonghong Songc195651e2018-04-28 22:28:08 -07002770/* flags used by BPF_FUNC_get_stackid only. */
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002771#define BPF_F_FAST_STACK_CMP (1ULL << 9)
2772#define BPF_F_REUSE_STACKID (1ULL << 10)
Yonghong Songc195651e2018-04-28 22:28:08 -07002773/* flags used by BPF_FUNC_get_stack only. */
2774#define BPF_F_USER_BUILD_ID (1ULL << 11)
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002775
Daniel Borkmann2da897e2016-02-23 02:05:26 +01002776/* BPF_FUNC_skb_set_tunnel_key flags. */
2777#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
Daniel Borkmann22080872016-03-04 15:15:05 +01002778#define BPF_F_DONT_FRAGMENT (1ULL << 2)
William Tu77a51962018-03-01 13:49:57 -08002779#define BPF_F_SEQ_NUMBER (1ULL << 3)
Daniel Borkmann2da897e2016-02-23 02:05:26 +01002780
Yonghong Song908432c2017-10-05 09:19:20 -07002781/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2782 * BPF_FUNC_perf_event_read_value flags.
2783 */
Daniel Borkmann1e337592016-04-18 21:01:23 +02002784#define BPF_F_INDEX_MASK 0xffffffffULL
2785#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
Daniel Borkmann555c8a82016-07-14 18:08:05 +02002786/* BPF_FUNC_perf_event_output for sk_buff input context. */
2787#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
Daniel Borkmann1e337592016-04-18 21:01:23 +02002788
Joe Stringerf71c6142018-11-30 15:32:20 -08002789/* Current network namespace */
2790#define BPF_F_CURRENT_NETNS (-1L)
2791
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04002792/* BPF_FUNC_skb_adjust_room flags. */
2793#define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2794
Alan Maguirebfb35c22019-04-12 12:27:34 +01002795#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2796#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
Alan Maguire58dfc902019-04-09 15:06:41 +01002797
Willem de Bruijn868d5232019-03-22 14:32:56 -04002798#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2799#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2800#define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2801#define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
Alan Maguirebfb35c22019-04-12 12:27:34 +01002802#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
Alan Maguire58dfc902019-04-09 15:06:41 +01002803 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2804 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Willem de Bruijn868d5232019-03-22 14:32:56 -04002805
Andrey Ignatov808649f2019-02-27 13:28:48 -08002806/* BPF_FUNC_sysctl_get_name flags. */
2807#define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2808
Daniel Borkmann2be7e212017-07-02 02:13:26 +02002809/* Mode for BPF_FUNC_skb_adjust_room helper. */
2810enum bpf_adj_room_mode {
2811 BPF_ADJ_ROOM_NET,
Willem de Bruijn14aa3192019-03-22 14:32:54 -04002812 BPF_ADJ_ROOM_MAC,
Daniel Borkmann2be7e212017-07-02 02:13:26 +02002813};
2814
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002815/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2816enum bpf_hdr_start_off {
2817 BPF_HDR_START_MAC,
2818 BPF_HDR_START_NET,
2819};
2820
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002821/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2822enum bpf_lwt_encap_mode {
2823 BPF_LWT_ENCAP_SEG6,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002824 BPF_LWT_ENCAP_SEG6_INLINE,
2825 BPF_LWT_ENCAP_IP,
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002826};
2827
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01002828#define __bpf_md_ptr(type, name) \
2829union { \
2830 type name; \
2831 __u64 :64; \
2832} __attribute__((aligned(8)))
2833
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07002834/* user accessible mirror of in-kernel sk_buff.
2835 * new fields can only be added to the end of this structure
2836 */
2837struct __sk_buff {
2838 __u32 len;
2839 __u32 pkt_type;
2840 __u32 mark;
2841 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -07002842 __u32 protocol;
2843 __u32 vlan_present;
2844 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +01002845 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +02002846 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -07002847 __u32 ingress_ifindex;
2848 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -07002849 __u32 tc_index;
2850 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +02002851 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -07002852 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -07002853 __u32 data;
2854 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +02002855 __u32 napi_id;
John Fastabend8a31db52017-08-15 22:33:09 -07002856
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02002857 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
John Fastabend8a31db52017-08-15 22:33:09 -07002858 __u32 family;
2859 __u32 remote_ip4; /* Stored in network byte order */
2860 __u32 local_ip4; /* Stored in network byte order */
2861 __u32 remote_ip6[4]; /* Stored in network byte order */
2862 __u32 local_ip6[4]; /* Stored in network byte order */
2863 __u32 remote_port; /* Stored in network byte order */
2864 __u32 local_port; /* stored in host byte order */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02002865 /* ... here. */
2866
2867 __u32 data_meta;
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01002868 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
Vlad Dumitrescuf11216b2018-11-22 14:39:16 -05002869 __u64 tstamp;
Petar Penkove3da08d2018-12-02 20:18:19 -05002870 __u32 wire_len;
Eric Dumazetd9ff2862019-01-23 09:22:27 -08002871 __u32 gso_segs;
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002872 __bpf_md_ptr(struct bpf_sock *, sk);
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07002873};
2874
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07002875struct bpf_tunnel_key {
2876 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +01002877 union {
2878 __u32 remote_ipv4;
2879 __u32 remote_ipv6[4];
2880 };
2881 __u8 tunnel_tos;
2882 __u8 tunnel_ttl;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02002883 __u16 tunnel_ext; /* Padding, future use. */
Daniel Borkmann4018ab12016-03-09 03:00:05 +01002884 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07002885};
2886
Eyal Birger12bed762018-04-24 17:50:29 +03002887/* user accessible mirror of in-kernel xfrm_state.
2888 * new fields can only be added to the end of this structure
2889 */
2890struct bpf_xfrm_state {
2891 __u32 reqid;
2892 __u32 spi; /* Stored in network byte order */
2893 __u16 family;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02002894 __u16 ext; /* Padding, future use. */
Eyal Birger12bed762018-04-24 17:50:29 +03002895 union {
2896 __u32 remote_ipv4; /* Stored in network byte order */
2897 __u32 remote_ipv6[4]; /* Stored in network byte order */
2898 };
2899};
2900
Thomas Graf3a0af8f2016-11-30 17:10:10 +01002901/* Generic BPF return codes which all BPF program types may support.
2902 * The values are binary compatible with their TC_ACT_* counter-part to
2903 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2904 * programs.
2905 *
2906 * XDP is handled seprately, see XDP_*.
2907 */
2908enum bpf_ret_code {
2909 BPF_OK = 0,
2910 /* 1 reserved */
2911 BPF_DROP = 2,
2912 /* 3-6 reserved */
2913 BPF_REDIRECT = 7,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002914 /* >127 are reserved for prog type specific return codes.
2915 *
2916 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
2917 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
2918 * changed and should be routed based on its new L3 header.
2919 * (This is an L3 redirect, as opposed to L2 redirect
2920 * represented by BPF_REDIRECT above).
2921 */
2922 BPF_LWT_REROUTE = 128,
Thomas Graf3a0af8f2016-11-30 17:10:10 +01002923};
2924
David Ahern610236582016-12-01 08:48:04 -08002925struct bpf_sock {
2926 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -08002927 __u32 family;
2928 __u32 type;
2929 __u32 protocol;
David Ahern482dca92017-08-31 15:05:44 -07002930 __u32 mark;
2931 __u32 priority;
Martin KaFai Lauaa65d692019-02-09 23:22:21 -08002932 /* IP address also allows 1 and 2 bytes access */
2933 __u32 src_ip4;
2934 __u32 src_ip6[4];
2935 __u32 src_port; /* host byte order */
2936 __u32 dst_port; /* network byte order */
2937 __u32 dst_ip4;
2938 __u32 dst_ip6[4];
2939 __u32 state;
David Ahern610236582016-12-01 08:48:04 -08002940};
2941
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002942struct bpf_tcp_sock {
2943 __u32 snd_cwnd; /* Sending congestion window */
2944 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
2945 __u32 rtt_min;
2946 __u32 snd_ssthresh; /* Slow start size threshold */
2947 __u32 rcv_nxt; /* What we want to receive next */
2948 __u32 snd_nxt; /* Next sequence we send */
2949 __u32 snd_una; /* First byte we want an ack for */
2950 __u32 mss_cache; /* Cached effective mss, not including SACKS */
2951 __u32 ecn_flags; /* ECN status bits. */
2952 __u32 rate_delivered; /* saved rate sample: packets delivered */
2953 __u32 rate_interval_us; /* saved rate sample: time elapsed */
2954 __u32 packets_out; /* Packets which are "in flight" */
2955 __u32 retrans_out; /* Retransmitted packets out */
2956 __u32 total_retrans; /* Total retransmits for entire connection */
2957 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
2958 * total number of segments in.
2959 */
2960 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
2961 * total number of data segments in.
2962 */
2963 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
2964 * The total number of segments sent.
2965 */
2966 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
2967 * total number of data segments sent.
2968 */
2969 __u32 lost_out; /* Lost packets */
2970 __u32 sacked_out; /* SACK'd packets */
2971 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
2972 * sum(delta(rcv_nxt)), or how many bytes
2973 * were acked.
2974 */
2975 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
2976 * sum(delta(snd_una)), or how many bytes
2977 * were acked.
2978 */
2979};
2980
Joe Stringer6acc9b42018-10-02 13:35:36 -07002981struct bpf_sock_tuple {
2982 union {
2983 struct {
2984 __be32 saddr;
2985 __be32 daddr;
2986 __be16 sport;
2987 __be16 dport;
2988 } ipv4;
2989 struct {
2990 __be32 saddr[4];
2991 __be32 daddr[4];
2992 __be16 sport;
2993 __be16 dport;
2994 } ipv6;
2995 };
2996};
2997
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08002998#define XDP_PACKET_HEADROOM 256
2999
Brenden Blanco6a773a12016-07-19 12:16:47 -07003000/* User return codes for XDP prog type.
3001 * A valid XDP program must return one of these defined values. All other
Daniel Borkmann9beb8be2017-09-09 01:40:35 +02003002 * return codes are reserved for future use. Unknown return codes will
3003 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
Brenden Blanco6a773a12016-07-19 12:16:47 -07003004 */
3005enum xdp_action {
3006 XDP_ABORTED = 0,
3007 XDP_DROP,
3008 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -07003009 XDP_TX,
John Fastabend814abfa2017-07-17 09:27:07 -07003010 XDP_REDIRECT,
Brenden Blanco6a773a12016-07-19 12:16:47 -07003011};
3012
3013/* user accessible metadata for XDP packet hook
3014 * new fields must be added to the end of this structure
3015 */
3016struct xdp_md {
3017 __u32 data;
3018 __u32 data_end;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003019 __u32 data_meta;
Jesper Dangaard Brouerdaaf24c2018-01-11 17:39:09 +01003020 /* Below access go through struct xdp_rxq_info */
Jesper Dangaard Brouer02dd3292018-01-03 11:26:14 +01003021 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3022 __u32 rx_queue_index; /* rxq->queue_index */
Brenden Blanco6a773a12016-07-19 12:16:47 -07003023};
3024
John Fastabend174a79f2017-08-15 22:32:47 -07003025enum sk_action {
John Fastabendbfa640752017-10-27 09:45:53 -07003026 SK_DROP = 0,
3027 SK_PASS,
John Fastabend174a79f2017-08-15 22:32:47 -07003028};
3029
John Fastabend4f738ad2018-03-18 12:57:10 -07003030/* user accessible metadata for SK_MSG packet hook, new fields must
3031 * be added to the end of this structure
3032 */
3033struct sk_msg_md {
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003034 __bpf_md_ptr(void *, data);
3035 __bpf_md_ptr(void *, data_end);
John Fastabend303def32018-05-17 14:16:58 -07003036
3037 __u32 family;
3038 __u32 remote_ip4; /* Stored in network byte order */
3039 __u32 local_ip4; /* Stored in network byte order */
3040 __u32 remote_ip6[4]; /* Stored in network byte order */
3041 __u32 local_ip6[4]; /* Stored in network byte order */
3042 __u32 remote_port; /* Stored in network byte order */
3043 __u32 local_port; /* stored in host byte order */
John Fastabend3bdbd022018-12-16 15:47:04 -08003044 __u32 size; /* Total size of sk_msg */
John Fastabend4f738ad2018-03-18 12:57:10 -07003045};
3046
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003047struct sk_reuseport_md {
3048 /*
3049 * Start of directly accessible data. It begins from
3050 * the tcp/udp header.
3051 */
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003052 __bpf_md_ptr(void *, data);
3053 /* End of directly accessible data */
3054 __bpf_md_ptr(void *, data_end);
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003055 /*
3056 * Total length of packet (starting from the tcp/udp header).
3057 * Note that the directly accessible bytes (data_end - data)
3058 * could be less than this "len". Those bytes could be
3059 * indirectly read by a helper "bpf_skb_load_bytes()".
3060 */
3061 __u32 len;
3062 /*
3063 * Eth protocol in the mac header (network byte order). e.g.
3064 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3065 */
3066 __u32 eth_protocol;
3067 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3068 __u32 bind_inany; /* Is sock bound to an INANY address? */
3069 __u32 hash; /* A hash of the packet 4 tuples */
3070};
3071
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003072#define BPF_TAG_SIZE 8
3073
3074struct bpf_prog_info {
3075 __u32 type;
3076 __u32 id;
3077 __u8 tag[BPF_TAG_SIZE];
3078 __u32 jited_prog_len;
3079 __u32 xlated_prog_len;
3080 __aligned_u64 jited_prog_insns;
3081 __aligned_u64 xlated_prog_insns;
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -07003082 __u64 load_time; /* ns since boottime */
3083 __u32 created_by_uid;
3084 __u32 nr_map_ids;
3085 __aligned_u64 map_ids;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003086 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski675fc272017-12-27 18:39:09 -08003087 __u32 ifindex;
Jiri Olsab85fab02018-04-25 19:41:06 +02003088 __u32 gpl_compatible:1;
Jakub Kicinski675fc272017-12-27 18:39:09 -08003089 __u64 netns_dev;
3090 __u64 netns_ino;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303091 __u32 nr_jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303092 __u32 nr_jited_func_lens;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303093 __aligned_u64 jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303094 __aligned_u64 jited_func_lens;
Yonghong Song838e9692018-11-19 15:29:11 -08003095 __u32 btf_id;
3096 __u32 func_info_rec_size;
3097 __aligned_u64 func_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003098 __u32 nr_func_info;
3099 __u32 nr_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003100 __aligned_u64 line_info;
3101 __aligned_u64 jited_line_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003102 __u32 nr_jited_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003103 __u32 line_info_rec_size;
3104 __u32 jited_line_info_rec_size;
Song Liuc872bdb2018-12-12 09:37:46 -08003105 __u32 nr_prog_tags;
3106 __aligned_u64 prog_tags;
Alexei Starovoitov5f8f8b92019-02-25 14:28:40 -08003107 __u64 run_time_ns;
3108 __u64 run_cnt;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003109} __attribute__((aligned(8)));
3110
3111struct bpf_map_info {
3112 __u32 type;
3113 __u32 id;
3114 __u32 key_size;
3115 __u32 value_size;
3116 __u32 max_entries;
3117 __u32 map_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003118 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski52775b32018-01-17 19:13:28 -08003119 __u32 ifindex;
Daniel Borkmann36f98142018-06-02 05:21:59 +02003120 __u32 :32;
Jakub Kicinski52775b32018-01-17 19:13:28 -08003121 __u64 netns_dev;
3122 __u64 netns_ino;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -07003123 __u32 btf_id;
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -07003124 __u32 btf_key_type_id;
3125 __u32 btf_value_type_id;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003126} __attribute__((aligned(8)));
3127
Martin KaFai Lau62dab842018-05-04 14:49:52 -07003128struct bpf_btf_info {
3129 __aligned_u64 btf;
3130 __u32 btf_size;
3131 __u32 id;
3132} __attribute__((aligned(8)));
3133
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003134/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3135 * by user and intended to be used by socket (e.g. to bind to, depends on
3136 * attach attach type).
3137 */
3138struct bpf_sock_addr {
3139 __u32 user_family; /* Allows 4-byte read, but no write. */
3140 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3141 * Stored in network byte order.
3142 */
3143 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3144 * Stored in network byte order.
3145 */
3146 __u32 user_port; /* Allows 4-byte read and write.
3147 * Stored in network byte order
3148 */
3149 __u32 family; /* Allows 4-byte read, but no write */
3150 __u32 type; /* Allows 4-byte read, but no write */
3151 __u32 protocol; /* Allows 4-byte read, but no write */
Andrey Ignatov1cedee12018-05-25 08:55:23 -07003152 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
3153 * Stored in network byte order.
3154 */
3155 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3156 * Stored in network byte order.
3157 */
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003158};
3159
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003160/* User bpf_sock_ops struct to access socket values and specify request ops
3161 * and their replies.
3162 * Some of this fields are in network (bigendian) byte order and may need
3163 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3164 * New fields can only be added at the end of this structure
3165 */
3166struct bpf_sock_ops {
3167 __u32 op;
3168 union {
Lawrence Brakmode525be2018-01-25 16:14:09 -08003169 __u32 args[4]; /* Optionally passed to bpf program */
3170 __u32 reply; /* Returned by bpf program */
3171 __u32 replylong[4]; /* Optionally returned by bpf prog */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003172 };
3173 __u32 family;
3174 __u32 remote_ip4; /* Stored in network byte order */
3175 __u32 local_ip4; /* Stored in network byte order */
3176 __u32 remote_ip6[4]; /* Stored in network byte order */
3177 __u32 local_ip6[4]; /* Stored in network byte order */
3178 __u32 remote_port; /* Stored in network byte order */
3179 __u32 local_port; /* stored in host byte order */
Lawrence Brakmof19397a2017-12-01 10:15:04 -08003180 __u32 is_fullsock; /* Some TCP fields are only valid if
3181 * there is a full socket. If not, the
3182 * fields read as zero.
3183 */
3184 __u32 snd_cwnd;
3185 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003186 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
Lawrence Brakmo44f0e432018-01-25 16:14:12 -08003187 __u32 state;
3188 __u32 rtt_min;
3189 __u32 snd_ssthresh;
3190 __u32 rcv_nxt;
3191 __u32 snd_nxt;
3192 __u32 snd_una;
3193 __u32 mss_cache;
3194 __u32 ecn_flags;
3195 __u32 rate_delivered;
3196 __u32 rate_interval_us;
3197 __u32 packets_out;
3198 __u32 retrans_out;
3199 __u32 total_retrans;
3200 __u32 segs_in;
3201 __u32 data_segs_in;
3202 __u32 segs_out;
3203 __u32 data_segs_out;
3204 __u32 lost_out;
3205 __u32 sacked_out;
3206 __u32 sk_txhash;
3207 __u64 bytes_received;
3208 __u64 bytes_acked;
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003209};
3210
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003211/* Definitions for bpf_sock_ops_cb_flags */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08003212#define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08003213#define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
Lawrence Brakmod4487492018-01-25 16:14:15 -08003214#define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3215#define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003216 * supported cb flags
3217 */
3218
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003219/* List of known BPF sock_ops operators.
3220 * New entries can only be added at the end
3221 */
3222enum {
3223 BPF_SOCK_OPS_VOID,
Lawrence Brakmo8550f322017-06-30 20:02:42 -07003224 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3225 * -1 if default value should be used
3226 */
Lawrence Brakmo13d3b1e2017-06-30 20:02:44 -07003227 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3228 * window (in packets) or -1 if default
3229 * value should be used
3230 */
Lawrence Brakmo9872a4b2017-06-30 20:02:47 -07003231 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3232 * active connection is initialized
3233 */
3234 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3235 * active connection is
3236 * established
3237 */
3238 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3239 * passive connection is
3240 * established
3241 */
Lawrence Brakmo91b5b212017-06-30 20:02:49 -07003242 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3243 * needs ECN
3244 */
Lawrence Brakmoe6546ef2017-10-20 11:05:39 -07003245 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3246 * based on the path and may be
3247 * dependent on the congestion control
3248 * algorithm. In general it indicates
3249 * a congestion threshold. RTTs above
3250 * this indicate congestion
3251 */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08003252 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3253 * Arg1: value of icsk_retransmits
3254 * Arg2: value of icsk_rto
3255 * Arg3: whether RTO has expired
3256 */
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08003257 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3258 * Arg1: sequence number of 1st byte
3259 * Arg2: # segments
3260 * Arg3: return value of
3261 * tcp_transmit_skb (0 => success)
3262 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08003263 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3264 * Arg1: old_state
3265 * Arg2: new_state
3266 */
Andrey Ignatovf333ee02018-07-11 17:33:32 -07003267 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3268 * socket transition to LISTEN state.
3269 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08003270};
3271
3272/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3273 * changes between the TCP and BPF versions. Ideally this should never happen.
3274 * If it does, we need to add code to convert them before calling
3275 * the BPF sock_ops function.
3276 */
3277enum {
3278 BPF_TCP_ESTABLISHED = 1,
3279 BPF_TCP_SYN_SENT,
3280 BPF_TCP_SYN_RECV,
3281 BPF_TCP_FIN_WAIT1,
3282 BPF_TCP_FIN_WAIT2,
3283 BPF_TCP_TIME_WAIT,
3284 BPF_TCP_CLOSE,
3285 BPF_TCP_CLOSE_WAIT,
3286 BPF_TCP_LAST_ACK,
3287 BPF_TCP_LISTEN,
3288 BPF_TCP_CLOSING, /* Now a valid state */
3289 BPF_TCP_NEW_SYN_RECV,
3290
3291 BPF_TCP_MAX_STATES /* Leave at the end! */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003292};
3293
Lawrence Brakmofc747812017-06-30 20:02:51 -07003294#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
Lawrence Brakmo13bf9642017-06-30 20:02:53 -07003295#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
Lawrence Brakmofc747812017-06-30 20:02:51 -07003296
Yonghong Song908432c2017-10-05 09:19:20 -07003297struct bpf_perf_event_value {
3298 __u64 counter;
3299 __u64 enabled;
3300 __u64 running;
3301};
3302
Roman Gushchinebc614f2017-11-05 08:15:32 -05003303#define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3304#define BPF_DEVCG_ACC_READ (1ULL << 1)
3305#define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3306
3307#define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3308#define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3309
3310struct bpf_cgroup_dev_ctx {
Yonghong Song06ef0cc2017-12-18 10:13:44 -08003311 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3312 __u32 access_type;
Roman Gushchinebc614f2017-11-05 08:15:32 -05003313 __u32 major;
3314 __u32 minor;
3315};
3316
Alexei Starovoitovc4f66992018-03-28 12:05:37 -07003317struct bpf_raw_tracepoint_args {
3318 __u64 args[0];
3319};
3320
David Ahern87f5fc72018-05-09 20:34:26 -07003321/* DIRECT: Skip the FIB rules and go to FIB table associated with device
3322 * OUTPUT: Do lookup from egress perspective; default is ingress
3323 */
3324#define BPF_FIB_LOOKUP_DIRECT BIT(0)
3325#define BPF_FIB_LOOKUP_OUTPUT BIT(1)
3326
David Ahern4c795792018-06-26 16:21:18 -07003327enum {
3328 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3329 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3330 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3331 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3332 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3333 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3334 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3335 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3336 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3337};
3338
David Ahern87f5fc72018-05-09 20:34:26 -07003339struct bpf_fib_lookup {
David Ahernfa898d72018-05-29 10:58:07 -07003340 /* input: network family for lookup (AF_INET, AF_INET6)
3341 * output: network family of egress nexthop
3342 */
3343 __u8 family;
David Ahern87f5fc72018-05-09 20:34:26 -07003344
3345 /* set if lookup is to consider L4 data - e.g., FIB rules */
3346 __u8 l4_protocol;
3347 __be16 sport;
3348 __be16 dport;
3349
3350 /* total length of packet from network header - used for MTU check */
3351 __u16 tot_len;
David Ahern4c795792018-06-26 16:21:18 -07003352
3353 /* input: L3 device index for lookup
3354 * output: device index from FIB lookup
3355 */
3356 __u32 ifindex;
David Ahern87f5fc72018-05-09 20:34:26 -07003357
3358 union {
3359 /* inputs to lookup */
3360 __u8 tos; /* AF_INET */
David Ahernbd3a08a2018-06-03 08:15:19 -07003361 __be32 flowinfo; /* AF_INET6, flow_label + priority */
David Ahern87f5fc72018-05-09 20:34:26 -07003362
David Ahernfa898d72018-05-29 10:58:07 -07003363 /* output: metric of fib result (IPv4/IPv6 only) */
3364 __u32 rt_metric;
David Ahern87f5fc72018-05-09 20:34:26 -07003365 };
3366
3367 union {
David Ahern87f5fc72018-05-09 20:34:26 -07003368 __be32 ipv4_src;
3369 __u32 ipv6_src[4]; /* in6_addr; network order */
3370 };
3371
David Ahernfa898d72018-05-29 10:58:07 -07003372 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3373 * network header. output: bpf_fib_lookup sets to gateway address
3374 * if FIB lookup returns gateway route
David Ahern87f5fc72018-05-09 20:34:26 -07003375 */
3376 union {
David Ahern87f5fc72018-05-09 20:34:26 -07003377 __be32 ipv4_dst;
3378 __u32 ipv6_dst[4]; /* in6_addr; network order */
3379 };
3380
3381 /* output */
3382 __be16 h_vlan_proto;
3383 __be16 h_vlan_TCI;
3384 __u8 smac[6]; /* ETH_ALEN */
3385 __u8 dmac[6]; /* ETH_ALEN */
3386};
3387
Yonghong Song41bdc4b2018-05-24 11:21:09 -07003388enum bpf_task_fd_type {
3389 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3390 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3391 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3392 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3393 BPF_FD_TYPE_UPROBE, /* filename + offset */
3394 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3395};
3396
Petar Penkovd58e4682018-09-14 07:46:18 -07003397struct bpf_flow_keys {
3398 __u16 nhoff;
3399 __u16 thoff;
3400 __u16 addr_proto; /* ETH_P_* of valid addrs */
3401 __u8 is_frag;
3402 __u8 is_first_frag;
3403 __u8 is_encap;
3404 __u8 ip_proto;
3405 __be16 n_proto;
3406 __be16 sport;
3407 __be16 dport;
3408 union {
3409 struct {
3410 __be32 ipv4_src;
3411 __be32 ipv4_dst;
3412 };
3413 struct {
3414 __u32 ipv6_src[4]; /* in6_addr; network order */
3415 __u32 ipv6_dst[4]; /* in6_addr; network order */
3416 };
3417 };
3418};
3419
Yonghong Song838e9692018-11-19 15:29:11 -08003420struct bpf_func_info {
Martin KaFai Laud30d42e2018-12-05 17:35:44 -08003421 __u32 insn_off;
Yonghong Song838e9692018-11-19 15:29:11 -08003422 __u32 type_id;
3423};
3424
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003425#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3426#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3427
3428struct bpf_line_info {
3429 __u32 insn_off;
3430 __u32 file_name_off;
3431 __u32 line_off;
3432 __u32 line_col;
3433};
3434
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08003435struct bpf_spin_lock {
3436 __u32 val;
3437};
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08003438
3439struct bpf_sysctl {
3440 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3441 * Allows 1,2,4-byte read, but no write.
3442 */
Andrey Ignatove1550bf2019-03-07 18:50:52 -08003443 __u32 file_pos; /* Sysctl file position to read from, write to.
3444 * Allows 1,2,4-byte read an 4-byte write.
3445 */
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08003446};
3447
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07003448#endif /* _UAPI__LINUX_BPF_H__ */