blob: 705e4822f997865b96556048f2325b26e149ae3d [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 */
Gustavo A. R. Silvad7f10df2020-02-26 18:17:44 -060076 __u8 data[]; /* Arbitrary size */
Daniel Mackb95a5c42017-01-21 17:26:11 +010077};
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,
Quentin Monnet1b9ed842019-08-20 10:31:50 +0100109 BPF_BTF_GET_NEXT_ID,
Brian Vazquezcb4d03a2020-01-15 10:43:01 -0800110 BPF_MAP_LOOKUP_BATCH,
Yonghong Song05799632020-01-15 10:43:04 -0800111 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
Brian Vazquezaa2e93b2020-01-15 10:43:02 -0800112 BPF_MAP_UPDATE_BATCH,
113 BPF_MAP_DELETE_BATCH,
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700114 BPF_LINK_CREATE,
Andrii Nakryiko0c991eb2020-03-29 19:59:59 -0700115 BPF_LINK_UPDATE,
Andrii Nakryiko2d602c82020-04-28 17:16:07 -0700116 BPF_LINK_GET_FD_BY_ID,
117 BPF_LINK_GET_NEXT_ID,
Song Liud46edd62020-04-30 00:15:04 -0700118 BPF_ENABLE_STATS,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700119};
120
121enum bpf_map_type {
122 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -0800123 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -0800124 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61ab2015-05-19 16:59:03 -0700125 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +0000126 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -0800127 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -0800128 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800129 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -0700130 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800131 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800132 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +0100133 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700134 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700135 BPF_MAP_TYPE_HASH_OF_MAPS,
John Fastabend546ac1f2017-07-17 09:28:56 -0700136 BPF_MAP_TYPE_DEVMAP,
John Fastabend174a79f2017-08-15 22:32:47 -0700137 BPF_MAP_TYPE_SOCKMAP,
Jesper Dangaard Brouer6710e112017-10-16 12:19:28 +0200138 BPF_MAP_TYPE_CPUMAP,
Björn Töpelfbfc504a2018-05-02 13:01:28 +0200139 BPF_MAP_TYPE_XSKMAP,
John Fastabend81110382018-05-14 10:00:17 -0700140 BPF_MAP_TYPE_SOCKHASH,
Roman Gushchinde9cbba2018-08-02 14:27:18 -0700141 BPF_MAP_TYPE_CGROUP_STORAGE,
Martin KaFai Lau5dc4c4b2018-08-08 01:01:24 -0700142 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
Roman Gushchinb741f162018-09-28 14:45:43 +0000143 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +0200144 BPF_MAP_TYPE_QUEUE,
145 BPF_MAP_TYPE_STACK,
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -0700146 BPF_MAP_TYPE_SK_STORAGE,
Toke Høiland-Jørgensen6f9d4512019-07-26 18:06:55 +0200147 BPF_MAP_TYPE_DEVMAP_HASH,
Martin KaFai Lau85d33df2020-01-08 16:35:05 -0800148 BPF_MAP_TYPE_STRUCT_OPS,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700149};
150
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100151/* Note that tracing related programs such as
152 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
153 * are not subject to a stable API since kernel internal data
154 * structures can change from release to release and may
155 * therefore break existing tracing BPF programs. Tracing BPF
156 * programs correspond to /a/ specific kernel which is to be
157 * analyzed, and not /a/ specific kernel /and/ all future ones.
158 */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700159enum bpf_prog_type {
160 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800161 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700162 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100163 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee8c2015-03-20 15:11:11 +0100164 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700165 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700166 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700167 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100168 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800169 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100170 BPF_PROG_TYPE_LWT_IN,
171 BPF_PROG_TYPE_LWT_OUT,
172 BPF_PROG_TYPE_LWT_XMIT,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700173 BPF_PROG_TYPE_SOCK_OPS,
John Fastabendb005fd12017-08-15 22:31:58 -0700174 BPF_PROG_TYPE_SK_SKB,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500175 BPF_PROG_TYPE_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700176 BPF_PROG_TYPE_SK_MSG,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700177 BPF_PROG_TYPE_RAW_TRACEPOINT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700178 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
Mathieu Xhonneux004d4b22018-05-20 14:58:16 +0100179 BPF_PROG_TYPE_LWT_SEG6LOCAL,
Sean Youngf4364dc2018-05-27 12:24:09 +0100180 BPF_PROG_TYPE_LIRC_MODE2,
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -0700181 BPF_PROG_TYPE_SK_REUSEPORT,
Petar Penkovd58e4682018-09-14 07:46:18 -0700182 BPF_PROG_TYPE_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800183 BPF_PROG_TYPE_CGROUP_SYSCTL,
Matt Mullins9df1c282019-04-26 11:49:47 -0700184 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
Stanislav Fomichev0d01da62019-06-27 13:38:47 -0700185 BPF_PROG_TYPE_CGROUP_SOCKOPT,
Alexei Starovoitovf1b95092019-10-30 15:32:11 -0700186 BPF_PROG_TYPE_TRACING,
Martin KaFai Lau27ae79972020-01-08 16:35:03 -0800187 BPF_PROG_TYPE_STRUCT_OPS,
Alexei Starovoitovbe8704f2020-01-20 16:53:46 -0800188 BPF_PROG_TYPE_EXT,
KP Singhfc611f42020-03-29 01:43:49 +0100189 BPF_PROG_TYPE_LSM,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700190};
191
Daniel Mack0e336612016-11-23 16:52:25 +0100192enum bpf_attach_type {
193 BPF_CGROUP_INET_INGRESS,
194 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800195 BPF_CGROUP_INET_SOCK_CREATE,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700196 BPF_CGROUP_SOCK_OPS,
John Fastabend464bc0f2017-08-28 07:10:04 -0700197 BPF_SK_SKB_STREAM_PARSER,
198 BPF_SK_SKB_STREAM_VERDICT,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500199 BPF_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700200 BPF_SK_MSG_VERDICT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700201 BPF_CGROUP_INET4_BIND,
202 BPF_CGROUP_INET6_BIND,
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700203 BPF_CGROUP_INET4_CONNECT,
204 BPF_CGROUP_INET6_CONNECT,
Andrey Ignatovaac3fc32018-03-30 15:08:07 -0700205 BPF_CGROUP_INET4_POST_BIND,
206 BPF_CGROUP_INET6_POST_BIND,
Andrey Ignatov1cedee12018-05-25 08:55:23 -0700207 BPF_CGROUP_UDP4_SENDMSG,
208 BPF_CGROUP_UDP6_SENDMSG,
Sean Youngf4364dc2018-05-27 12:24:09 +0100209 BPF_LIRC_MODE2,
Petar Penkovd58e4682018-09-14 07:46:18 -0700210 BPF_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800211 BPF_CGROUP_SYSCTL,
Daniel Borkmann983695f2019-06-07 01:48:57 +0200212 BPF_CGROUP_UDP4_RECVMSG,
213 BPF_CGROUP_UDP6_RECVMSG,
Stanislav Fomichev0d01da62019-06-27 13:38:47 -0700214 BPF_CGROUP_GETSOCKOPT,
215 BPF_CGROUP_SETSOCKOPT,
Alexei Starovoitovf1b95092019-10-30 15:32:11 -0700216 BPF_TRACE_RAW_TP,
Alexei Starovoitovfec56f52019-11-14 10:57:04 -0800217 BPF_TRACE_FENTRY,
218 BPF_TRACE_FEXIT,
KP Singhae240822020-03-04 20:18:49 +0100219 BPF_MODIFY_RETURN,
KP Singhfc611f42020-03-29 01:43:49 +0100220 BPF_LSM_MAC,
Daniel Mack0e336612016-11-23 16:52:25 +0100221 __MAX_BPF_ATTACH_TYPE
222};
223
224#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
225
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -0700226enum bpf_link_type {
227 BPF_LINK_TYPE_UNSPEC = 0,
228 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
229 BPF_LINK_TYPE_TRACING = 2,
230 BPF_LINK_TYPE_CGROUP = 3,
231
232 MAX_BPF_LINK_TYPE,
233};
234
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700235/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
236 *
237 * NONE(default): No further bpf programs allowed in the subtree.
238 *
239 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
240 * the program in this cgroup yields to sub-cgroup program.
241 *
242 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
243 * that cgroup program gets run in addition to the program in this cgroup.
244 *
245 * Only one program is allowed to be attached to a cgroup with
246 * NONE or BPF_F_ALLOW_OVERRIDE flag.
247 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
248 * release old program and attach the new one. Attach flags has to match.
249 *
250 * Multiple programs are allowed to be attached to a cgroup with
251 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
252 * (those that were attached first, run first)
253 * The programs of sub-cgroup are executed first, then programs of
254 * this cgroup and then programs of parent cgroup.
255 * When children program makes decision (like picking TCP CA or sock bind)
256 * parent program has a chance to override it.
257 *
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800258 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
259 * programs for a cgroup. Though it's possible to replace an old program at
260 * any position by also specifying BPF_F_REPLACE flag and position itself in
261 * replace_bpf_fd attribute. Old program at this position will be released.
262 *
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700263 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
264 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
265 * Ex1:
266 * cgrp1 (MULTI progs A, B) ->
267 * cgrp2 (OVERRIDE prog C) ->
268 * cgrp3 (MULTI prog D) ->
269 * cgrp4 (OVERRIDE prog E) ->
270 * cgrp5 (NONE prog F)
271 * the event in cgrp5 triggers execution of F,D,A,B in that order.
272 * if prog F is detached, the execution is E,D,A,B
273 * if prog F and D are detached, the execution is E,A,B
274 * if prog F, E and D are detached, the execution is C,A,B
275 *
276 * All eligible programs are executed regardless of return code from
277 * earlier programs.
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800278 */
279#define BPF_F_ALLOW_OVERRIDE (1U << 0)
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700280#define BPF_F_ALLOW_MULTI (1U << 1)
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800281#define BPF_F_REPLACE (1U << 2)
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800282
David S. Millere07b98d2017-05-10 11:38:07 -0700283/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
284 * verifier will perform strict alignment checking as if the kernel
285 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
286 * and NET_IP_ALIGN defined to 2.
287 */
288#define BPF_F_STRICT_ALIGNMENT (1U << 0)
289
David Millere9ee9ef2018-11-30 21:08:14 -0800290/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
291 * verifier will allow any alignment whatsoever. On platforms
292 * with strict alignment requirements for loads ands stores (such
293 * as sparc and mips) the verifier validates that all loads and
294 * stores provably follow this requirement. This flag turns that
295 * checking and enforcement off.
296 *
297 * It is mostly used for testing when we want to validate the
298 * context and memory access aspects of the verifier, but because
299 * of an unaligned access the alignment check would trigger before
300 * the one we are interested in.
301 */
302#define BPF_F_ANY_ALIGNMENT (1U << 1)
303
Jiong Wangc240eff2019-05-24 23:25:16 +0100304/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
305 * Verifier does sub-register def/use analysis and identifies instructions whose
306 * def only matters for low 32-bit, high 32-bit is never referenced later
307 * through implicit zero extension. Therefore verifier notifies JIT back-ends
308 * that it is safe to ignore clearing high 32-bit for these instructions. This
309 * saves some back-ends a lot of code-gen. However such optimization is not
310 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
311 * hence hasn't used verifier's analysis result. But, we really want to have a
312 * way to be able to verify the correctness of the described optimization on
313 * x86_64 on which testsuites are frequently exercised.
314 *
315 * So, this flag is introduced. Once it is set, verifier will randomize high
316 * 32-bit for those instructions who has been identified as safe to ignore them.
317 * Then, if verifier is not doing correct analysis, such randomization will
318 * regress tests to expose bugs.
319 */
320#define BPF_F_TEST_RND_HI32 (1U << 2)
321
Alexei Starovoitov10d274e2019-08-22 22:52:12 -0700322/* The verifier internal test flag. Behavior is undefined */
323#define BPF_F_TEST_STATE_FREQ (1U << 3)
324
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200325/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
326 * two extensions:
327 *
328 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
329 * insn[0].imm: map fd map fd
330 * insn[1].imm: 0 offset into value
331 * insn[0].off: 0 0
332 * insn[1].off: 0 0
333 * ldimm64 rewrite: address of map address of map[0]+offset
334 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
335 */
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100336#define BPF_PSEUDO_MAP_FD 1
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200337#define BPF_PSEUDO_MAP_VALUE 2
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100338
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800339/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
340 * offset to another bpf function
341 */
342#define BPF_PSEUDO_CALL 1
343
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800344/* flags for BPF_MAP_UPDATE_ELEM command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800345enum {
346 BPF_ANY = 0, /* create new element or update existing */
347 BPF_NOEXIST = 1, /* create new element if it didn't exist */
348 BPF_EXIST = 2, /* update existing element */
349 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
350};
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800351
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700352/* flags for BPF_MAP_CREATE command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800353enum {
354 BPF_F_NO_PREALLOC = (1U << 0),
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800355/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800356 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800357 * which can scale and perform better.
358 * Note, the LRU nodes (including free nodes) cannot be moved
359 * across different LRU lists.
360 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800361 BPF_F_NO_COMMON_LRU = (1U << 1),
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700362/* Specify numa node during map creation */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800363 BPF_F_NUMA_NODE = (1U << 2),
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -0700364
Daniel Borkmann591fe982019-04-09 23:20:05 +0200365/* Flags for accessing BPF object from syscall side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800366 BPF_F_RDONLY = (1U << 3),
367 BPF_F_WRONLY = (1U << 4),
Chenbo Feng6e71b042017-10-18 13:00:22 -0700368
Song Liu615755a2018-03-14 10:23:21 -0700369/* Flag for stack_map, store build_id+offset instead of pointer */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800370 BPF_F_STACK_BUILD_ID = (1U << 5),
Song Liu615755a2018-03-14 10:23:21 -0700371
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000372/* Zero-initialize hash function seed. This should only be used for testing. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800373 BPF_F_ZERO_SEED = (1U << 6),
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000374
Daniel Borkmann591fe982019-04-09 23:20:05 +0200375/* Flags for accessing BPF object from program side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800376 BPF_F_RDONLY_PROG = (1U << 7),
377 BPF_F_WRONLY_PROG = (1U << 8),
Daniel Borkmann591fe982019-04-09 23:20:05 +0200378
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700379/* Clone map from listener for newly accepted socket */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800380 BPF_F_CLONE = (1U << 9),
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700381
Andrii Nakryikofc970222019-11-17 09:28:04 -0800382/* Enable memory-mapping BPF map */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800383 BPF_F_MMAPABLE = (1U << 10),
384};
Andrii Nakryikofc970222019-11-17 09:28:04 -0800385
Andrey Ignatovf5bfcd92020-01-07 17:40:06 -0800386/* Flags for BPF_PROG_QUERY. */
387
388/* Query effective (directly attached + inherited from ancestor cgroups)
389 * programs that will be executed for events within a cgroup.
390 * attach_flags with this flag are returned only for directly attached programs.
391 */
Lorenz Bauer2f183362018-11-16 11:41:09 +0000392#define BPF_F_QUERY_EFFECTIVE (1U << 0)
393
Song Liud46edd62020-04-30 00:15:04 -0700394/* type for BPF_ENABLE_STATS */
395enum bpf_stats_type {
396 /* enabled run_time_ns and run_cnt */
397 BPF_STATS_RUN_TIME = 0,
398};
399
Song Liu615755a2018-03-14 10:23:21 -0700400enum bpf_stack_build_id_status {
401 /* user space need an empty entry to identify end of a trace */
402 BPF_STACK_BUILD_ID_EMPTY = 0,
403 /* with valid build_id and offset */
404 BPF_STACK_BUILD_ID_VALID = 1,
405 /* couldn't get build_id, fallback to ip */
406 BPF_STACK_BUILD_ID_IP = 2,
407};
408
409#define BPF_BUILD_ID_SIZE 20
410struct bpf_stack_build_id {
411 __s32 status;
412 unsigned char build_id[BPF_BUILD_ID_SIZE];
413 union {
414 __u64 offset;
415 __u64 ip;
416 };
417};
418
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800419#define BPF_OBJ_NAME_LEN 16U
420
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700421union bpf_attr {
422 struct { /* anonymous struct used by BPF_MAP_CREATE command */
423 __u32 map_type; /* one of enum bpf_map_type */
424 __u32 key_size; /* size of key in bytes */
425 __u32 value_size; /* size of value in bytes */
426 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700427 __u32 map_flags; /* BPF_MAP_CREATE related
428 * flags defined above.
429 */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700430 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700431 __u32 numa_node; /* numa node (effective only if
432 * BPF_F_NUMA_NODE is set).
433 */
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700434 char map_name[BPF_OBJ_NAME_LEN];
Jakub Kicinskia3884572018-01-11 20:29:09 -0800435 __u32 map_ifindex; /* ifindex of netdev to create on */
Martin KaFai Laua26ca7c2018-04-18 15:56:03 -0700436 __u32 btf_fd; /* fd pointing to a BTF type data */
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -0700437 __u32 btf_key_type_id; /* BTF type_id of the key */
438 __u32 btf_value_type_id; /* BTF type_id of the value */
Martin KaFai Lau85d33df2020-01-08 16:35:05 -0800439 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
440 * struct stored as the
441 * map value
442 */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700443 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700444
445 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
446 __u32 map_fd;
447 __aligned_u64 key;
448 union {
449 __aligned_u64 value;
450 __aligned_u64 next_key;
451 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800452 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700453 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700454
Brian Vazquezcb4d03a2020-01-15 10:43:01 -0800455 struct { /* struct used by BPF_MAP_*_BATCH commands */
456 __aligned_u64 in_batch; /* start batch,
457 * NULL to start from beginning
458 */
459 __aligned_u64 out_batch; /* output: next start batch */
460 __aligned_u64 keys;
461 __aligned_u64 values;
462 __u32 count; /* input/output:
463 * input: # of key/value
464 * elements
465 * output: # of filled elements
466 */
467 __u32 map_fd;
468 __u64 elem_flags;
469 __u64 flags;
470 } batch;
471
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700472 struct { /* anonymous struct used by BPF_PROG_LOAD command */
473 __u32 prog_type; /* one of enum bpf_prog_type */
474 __u32 insn_cnt;
475 __aligned_u64 insns;
476 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700477 __u32 log_level; /* verbosity level of verifier */
478 __u32 log_size; /* size of user buffer */
479 __aligned_u64 log_buf; /* user supplied buffer */
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100480 __u32 kern_version; /* not used */
David S. Millere07b98d2017-05-10 11:38:07 -0700481 __u32 prog_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700482 char prog_name[BPF_OBJ_NAME_LEN];
Jakub Kicinski1f6f4cb2017-11-20 15:21:53 -0800483 __u32 prog_ifindex; /* ifindex of netdev to prep for */
Andrey Ignatov5e43f892018-03-30 15:08:00 -0700484 /* For some prog types expected attach type must be known at
485 * load time to verify attach type specific parts of prog
486 * (context accesses, allowed helpers, etc).
487 */
488 __u32 expected_attach_type;
Yonghong Song838e9692018-11-19 15:29:11 -0800489 __u32 prog_btf_fd; /* fd pointing to BTF type data */
490 __u32 func_info_rec_size; /* userspace bpf_func_info size */
491 __aligned_u64 func_info; /* func info */
492 __u32 func_info_cnt; /* number of bpf_func_info records */
Martin KaFai Lauc454a462018-12-07 16:42:25 -0800493 __u32 line_info_rec_size; /* userspace bpf_line_info size */
494 __aligned_u64 line_info; /* line info */
495 __u32 line_info_cnt; /* number of bpf_line_info records */
Alexei Starovoitovccfe29eb22019-10-15 20:24:58 -0700496 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
Alexei Starovoitov5b92a282019-11-14 10:57:17 -0800497 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700498 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100499
500 struct { /* anonymous struct used by BPF_OBJ_* commands */
501 __aligned_u64 pathname;
502 __u32 bpf_fd;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700503 __u32 file_flags;
Daniel Borkmannb2197752015-10-29 14:58:09 +0100504 };
Daniel Mackf4324552016-11-23 16:52:27 +0100505
506 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
507 __u32 target_fd; /* container object to attach to */
508 __u32 attach_bpf_fd; /* eBPF program to attach */
509 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800510 __u32 attach_flags;
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800511 __u32 replace_bpf_fd; /* previously attached eBPF
512 * program to replace if
513 * BPF_F_REPLACE is used
514 */
Daniel Mackf4324552016-11-23 16:52:27 +0100515 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700516
517 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
518 __u32 prog_fd;
519 __u32 retval;
Lorenz Bauerb5a36b12018-12-03 11:31:23 +0000520 __u32 data_size_in; /* input: len of data_in */
521 __u32 data_size_out; /* input/output: len of data_out
522 * returns ENOSPC if data_out
523 * is too small.
524 */
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700525 __aligned_u64 data_in;
526 __aligned_u64 data_out;
527 __u32 repeat;
528 __u32 duration;
Stanislav Fomichevb0b93952019-04-09 11:49:09 -0700529 __u32 ctx_size_in; /* input: len of ctx_in */
530 __u32 ctx_size_out; /* input/output: len of ctx_out
531 * returns ENOSPC if ctx_out
532 * is too small.
533 */
534 __aligned_u64 ctx_in;
535 __aligned_u64 ctx_out;
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700536 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700537
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700538 struct { /* anonymous struct used by BPF_*_GET_*_ID */
539 union {
540 __u32 start_id;
541 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700542 __u32 map_id;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700543 __u32 btf_id;
Andrii Nakryikoa3b80e12020-04-28 17:16:06 -0700544 __u32 link_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700545 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700546 __u32 next_id;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700547 __u32 open_flags;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700548 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700549
550 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
551 __u32 bpf_fd;
552 __u32 info_len;
553 __aligned_u64 info;
554 } info;
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700555
556 struct { /* anonymous struct used by BPF_PROG_QUERY command */
557 __u32 target_fd; /* container object to query */
558 __u32 attach_type;
559 __u32 query_flags;
560 __u32 attach_flags;
561 __aligned_u64 prog_ids;
562 __u32 prog_cnt;
563 } query;
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700564
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700565 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700566 __u64 name;
567 __u32 prog_fd;
568 } raw_tracepoint;
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700569
570 struct { /* anonymous struct for BPF_BTF_LOAD */
571 __aligned_u64 btf;
572 __aligned_u64 btf_log_buf;
573 __u32 btf_size;
574 __u32 btf_log_size;
575 __u32 btf_log_level;
576 };
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700577
578 struct {
579 __u32 pid; /* input: pid */
580 __u32 fd; /* input: fd */
581 __u32 flags; /* input: flags */
582 __u32 buf_len; /* input/output: buf len */
583 __aligned_u64 buf; /* input/output:
584 * tp_name for tracepoint
585 * symbol for kprobe
586 * filename for uprobe
587 */
588 __u32 prog_id; /* output: prod_id */
589 __u32 fd_type; /* output: BPF_FD_TYPE_* */
590 __u64 probe_offset; /* output: probe_offset */
591 __u64 probe_addr; /* output: probe_addr */
592 } task_fd_query;
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700593
594 struct { /* struct used by BPF_LINK_CREATE command */
595 __u32 prog_fd; /* eBPF program to attach */
596 __u32 target_fd; /* object to attach to */
597 __u32 attach_type; /* attach type */
598 __u32 flags; /* extra flags */
599 } link_create;
Andrii Nakryiko0c991eb2020-03-29 19:59:59 -0700600
601 struct { /* struct used by BPF_LINK_UPDATE command */
602 __u32 link_fd; /* link fd */
603 /* new program fd to update link with */
604 __u32 new_prog_fd;
605 __u32 flags; /* extra flags */
606 /* expected link's program fd; is specified only if
607 * BPF_F_REPLACE flag is set in flags */
608 __u32 old_prog_fd;
609 } link_update;
610
Song Liud46edd62020-04-30 00:15:04 -0700611 struct { /* struct used by BPF_ENABLE_STATS command */
612 __u32 type;
613 } enable_stats;
614
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700615} __attribute__((aligned(8)));
616
Quentin Monnet56a092c2018-04-25 18:16:52 +0100617/* The description below is an attempt at providing documentation to eBPF
618 * developers about the multiple available eBPF helper functions. It can be
619 * parsed and used to produce a manual page. The workflow is the following,
620 * and requires the rst2man utility:
Thomas Grafebb676d2016-10-27 11:23:51 +0200621 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100622 * $ ./scripts/bpf_helpers_doc.py \
623 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
624 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
625 * $ man /tmp/bpf-helpers.7
Thomas Grafebb676d2016-10-27 11:23:51 +0200626 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100627 * Note that in order to produce this external documentation, some RST
628 * formatting is used in the descriptions to get "bold" and "italics" in
629 * manual pages. Also note that the few trailing white spaces are
630 * intentional, removing them would break paragraphs for rst2man.
Thomas Grafebb676d2016-10-27 11:23:51 +0200631 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100632 * Start of BPF helper function descriptions:
Quentin Monnetad4a5222018-04-25 18:16:53 +0100633 *
634 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
635 * Description
636 * Perform a lookup in *map* for an entry associated to *key*.
637 * Return
638 * Map value associated to *key*, or **NULL** if no entry was
639 * found.
640 *
641 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
642 * Description
643 * Add or update the value of the entry associated to *key* in
644 * *map* with *value*. *flags* is one of:
645 *
646 * **BPF_NOEXIST**
647 * The entry for *key* must not exist in the map.
648 * **BPF_EXIST**
649 * The entry for *key* must already exist in the map.
650 * **BPF_ANY**
651 * No condition on the existence of the entry for *key*.
652 *
653 * Flag value **BPF_NOEXIST** cannot be used for maps of types
654 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
655 * elements always exist), the helper would return an error.
656 * Return
657 * 0 on success, or a negative error in case of failure.
658 *
659 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
660 * Description
661 * Delete entry with *key* from *map*.
662 * Return
663 * 0 on success, or a negative error in case of failure.
664 *
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +0100665 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100666 * Description
667 * For tracing programs, safely attempt to read *size* bytes from
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +0100668 * kernel space address *unsafe_ptr* and store the data in *dst*.
669 *
670 * Generally, use bpf_probe_read_user() or bpf_probe_read_kernel()
671 * instead.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100672 * Return
673 * 0 on success, or a negative error in case of failure.
674 *
675 * u64 bpf_ktime_get_ns(void)
676 * Description
677 * Return the time elapsed since system boot, in nanoseconds.
Maciej Żenczykowski71d19212020-04-26 09:15:25 -0700678 * Does not include time the system was suspended.
679 * See: clock_gettime(CLOCK_MONOTONIC)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100680 * Return
681 * Current *ktime*.
682 *
683 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
684 * Description
685 * This helper is a "printk()-like" facility for debugging. It
686 * prints a message defined by format *fmt* (of size *fmt_size*)
687 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
688 * available. It can take up to three additional **u64**
689 * arguments (as an eBPF helpers, the total number of arguments is
690 * limited to five).
691 *
692 * Each time the helper is called, it appends a line to the trace.
Peter Wu55c33df2019-08-21 00:08:59 +0100693 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
694 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100695 * The format of the trace is customizable, and the exact output
696 * one will get depends on the options set in
697 * *\/sys/kernel/debug/tracing/trace_options* (see also the
698 * *README* file under the same directory). However, it usually
699 * defaults to something like:
700 *
701 * ::
702 *
703 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
704 *
705 * In the above:
706 *
707 * * ``telnet`` is the name of the current task.
708 * * ``470`` is the PID of the current task.
709 * * ``001`` is the CPU number on which the task is
710 * running.
711 * * In ``.N..``, each character refers to a set of
712 * options (whether irqs are enabled, scheduling
713 * options, whether hard/softirqs are running, level of
714 * preempt_disabled respectively). **N** means that
715 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
716 * are set.
717 * * ``419421.045894`` is a timestamp.
718 * * ``0x00000001`` is a fake value used by BPF for the
719 * instruction pointer register.
720 * * ``<formatted msg>`` is the message formatted with
721 * *fmt*.
722 *
723 * The conversion specifiers supported by *fmt* are similar, but
724 * more limited than for printk(). They are **%d**, **%i**,
725 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
726 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
727 * of field, padding with zeroes, etc.) is available, and the
728 * helper will return **-EINVAL** (but print nothing) if it
729 * encounters an unknown specifier.
730 *
731 * Also, note that **bpf_trace_printk**\ () is slow, and should
732 * only be used for debugging purposes. For this reason, a notice
733 * bloc (spanning several lines) is printed to kernel logs and
734 * states that the helper should not be used "for production use"
735 * the first time this helper is used (or more precisely, when
736 * **trace_printk**\ () buffers are allocated). For passing values
737 * to user space, perf events should be preferred.
738 * Return
739 * The number of bytes written to the buffer, or a negative error
740 * in case of failure.
741 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100742 * u32 bpf_get_prandom_u32(void)
743 * Description
744 * Get a pseudo-random number.
745 *
746 * From a security point of view, this helper uses its own
747 * pseudo-random internal state, and cannot be used to infer the
748 * seed of other random functions in the kernel. However, it is
749 * essential to note that the generator used by the helper is not
750 * cryptographically secure.
751 * Return
752 * A random 32-bit unsigned value.
753 *
754 * u32 bpf_get_smp_processor_id(void)
755 * Description
756 * Get the SMP (symmetric multiprocessing) processor id. Note that
757 * all programs run with preemption disabled, which means that the
758 * SMP processor id is stable during all the execution of the
759 * program.
760 * Return
761 * The SMP id of the processor running the program.
762 *
Quentin Monnetad4a5222018-04-25 18:16:53 +0100763 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
764 * Description
765 * Store *len* bytes from address *from* into the packet
766 * associated to *skb*, at *offset*. *flags* are a combination of
767 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
768 * checksum for the packet after storing the bytes) and
769 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
770 * **->swhash** and *skb*\ **->l4hash** to 0).
771 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100772 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100773 * packet buffer. Therefore, at load time, all checks on pointers
774 * previously done by the verifier are invalidated and must be
775 * performed again, if the helper is used in combination with
776 * direct packet access.
777 * Return
778 * 0 on success, or a negative error in case of failure.
779 *
780 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
781 * Description
782 * Recompute the layer 3 (e.g. IP) checksum for the packet
783 * associated to *skb*. Computation is incremental, so the helper
784 * must know the former value of the header field that was
785 * modified (*from*), the new value of this field (*to*), and the
786 * number of bytes (2 or 4) for this field, stored in *size*.
787 * Alternatively, it is possible to store the difference between
788 * the previous and the new values of the header field in *to*, by
789 * setting *from* and *size* to 0. For both methods, *offset*
790 * indicates the location of the IP checksum within the packet.
791 *
792 * This helper works in combination with **bpf_csum_diff**\ (),
793 * which does not update the checksum in-place, but offers more
794 * flexibility and can handle sizes larger than 2 or 4 for the
795 * checksum to update.
796 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100797 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100798 * packet buffer. Therefore, at load time, all checks on pointers
799 * previously done by the verifier are invalidated and must be
800 * performed again, if the helper is used in combination with
801 * direct packet access.
802 * Return
803 * 0 on success, or a negative error in case of failure.
804 *
805 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
806 * Description
807 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
808 * packet associated to *skb*. Computation is incremental, so the
809 * helper must know the former value of the header field that was
810 * modified (*from*), the new value of this field (*to*), and the
811 * number of bytes (2 or 4) for this field, stored on the lowest
812 * four bits of *flags*. Alternatively, it is possible to store
813 * the difference between the previous and the new values of the
814 * header field in *to*, by setting *from* and the four lowest
815 * bits of *flags* to 0. For both methods, *offset* indicates the
816 * location of the IP checksum within the packet. In addition to
817 * the size of the field, *flags* can be added (bitwise OR) actual
818 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
819 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
820 * for updates resulting in a null checksum the value is set to
821 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
822 * the checksum is to be computed against a pseudo-header.
823 *
824 * This helper works in combination with **bpf_csum_diff**\ (),
825 * which does not update the checksum in-place, but offers more
826 * flexibility and can handle sizes larger than 2 or 4 for the
827 * checksum to update.
828 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100829 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100830 * packet buffer. Therefore, at load time, all checks on pointers
831 * previously done by the verifier are invalidated and must be
832 * performed again, if the helper is used in combination with
833 * direct packet access.
834 * Return
835 * 0 on success, or a negative error in case of failure.
836 *
837 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
838 * Description
839 * This special helper is used to trigger a "tail call", or in
840 * other words, to jump into another eBPF program. The same stack
841 * frame is used (but values on stack and in registers for the
842 * caller are not accessible to the callee). This mechanism allows
843 * for program chaining, either for raising the maximum number of
844 * available eBPF instructions, or to execute given programs in
845 * conditional blocks. For security reasons, there is an upper
846 * limit to the number of successive tail calls that can be
847 * performed.
848 *
849 * Upon call of this helper, the program attempts to jump into a
850 * program referenced at index *index* in *prog_array_map*, a
851 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
852 * *ctx*, a pointer to the context.
853 *
854 * If the call succeeds, the kernel immediately runs the first
855 * instruction of the new program. This is not a function call,
856 * and it never returns to the previous program. If the call
857 * fails, then the helper has no effect, and the caller continues
858 * to run its subsequent instructions. A call can fail if the
859 * destination program for the jump does not exist (i.e. *index*
860 * is superior to the number of entries in *prog_array_map*), or
861 * if the maximum number of tail calls has been reached for this
862 * chain of programs. This limit is defined in the kernel by the
863 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
864 * which is currently set to 32.
865 * Return
866 * 0 on success, or a negative error in case of failure.
867 *
868 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
869 * Description
870 * Clone and redirect the packet associated to *skb* to another
871 * net device of index *ifindex*. Both ingress and egress
872 * interfaces can be used for redirection. The **BPF_F_INGRESS**
873 * value in *flags* is used to make the distinction (ingress path
874 * is selected if the flag is present, egress path otherwise).
875 * This is the only flag supported for now.
876 *
877 * In comparison with **bpf_redirect**\ () helper,
878 * **bpf_clone_redirect**\ () has the associated cost of
879 * duplicating the packet buffer, but this can be executed out of
880 * the eBPF program. Conversely, **bpf_redirect**\ () is more
881 * efficient, but it is handled through an action code where the
882 * redirection happens only after the eBPF program has returned.
883 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100884 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100885 * packet buffer. Therefore, at load time, all checks on pointers
886 * previously done by the verifier are invalidated and must be
887 * performed again, if the helper is used in combination with
888 * direct packet access.
889 * Return
890 * 0 on success, or a negative error in case of failure.
Quentin Monnetc456dec2018-04-25 18:16:54 +0100891 *
892 * u64 bpf_get_current_pid_tgid(void)
893 * Return
894 * A 64-bit integer containing the current tgid and pid, and
895 * created as such:
896 * *current_task*\ **->tgid << 32 \|**
897 * *current_task*\ **->pid**.
898 *
899 * u64 bpf_get_current_uid_gid(void)
900 * Return
901 * A 64-bit integer containing the current GID and UID, and
902 * created as such: *current_gid* **<< 32 \|** *current_uid*.
903 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -0700904 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100905 * Description
906 * Copy the **comm** attribute of the current task into *buf* of
907 * *size_of_buf*. The **comm** attribute contains the name of
908 * the executable (excluding the path) for the current task. The
909 * *size_of_buf* must be strictly positive. On success, the
910 * helper makes sure that the *buf* is NUL-terminated. On failure,
911 * it is filled with zeroes.
912 * Return
913 * 0 on success, or a negative error in case of failure.
914 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100915 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
916 * Description
917 * Retrieve the classid for the current task, i.e. for the net_cls
918 * cgroup to which *skb* belongs.
919 *
920 * This helper can be used on TC egress path, but not on ingress.
921 *
922 * The net_cls cgroup provides an interface to tag network packets
923 * based on a user-provided identifier for all traffic coming from
924 * the tasks belonging to the related cgroup. See also the related
925 * kernel documentation, available from the Linux sources in file
Mauro Carvalho Chehabda82c922019-06-27 13:08:35 -0300926 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100927 *
928 * The Linux kernel has two versions for cgroups: there are
929 * cgroups v1 and cgroups v2. Both are available to users, who can
930 * use a mixture of them, but note that the net_cls cgroup is for
931 * cgroup v1 only. This makes it incompatible with BPF programs
932 * run on cgroups, which is a cgroup-v2-only feature (a socket can
933 * only hold data for one version of cgroups at a time).
934 *
935 * This helper is only available is the kernel was compiled with
936 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
937 * "**y**" or to "**m**".
938 * Return
939 * The classid, or 0 for the default unconfigured classid.
940 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100941 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
942 * Description
943 * Push a *vlan_tci* (VLAN tag control information) of protocol
944 * *vlan_proto* to the packet associated to *skb*, then update
945 * the checksum. Note that if *vlan_proto* is different from
946 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
947 * be **ETH_P_8021Q**.
948 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100949 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100950 * packet buffer. Therefore, at load time, all checks on pointers
951 * previously done by the verifier are invalidated and must be
952 * performed again, if the helper is used in combination with
953 * direct packet access.
954 * Return
955 * 0 on success, or a negative error in case of failure.
956 *
957 * int bpf_skb_vlan_pop(struct sk_buff *skb)
958 * Description
959 * Pop a VLAN header from the packet associated to *skb*.
960 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100961 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100962 * packet buffer. Therefore, at load time, all checks on pointers
963 * previously done by the verifier are invalidated and must be
964 * performed again, if the helper is used in combination with
965 * direct packet access.
966 * Return
967 * 0 on success, or a negative error in case of failure.
968 *
969 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
970 * Description
971 * Get tunnel metadata. This helper takes a pointer *key* to an
972 * empty **struct bpf_tunnel_key** of **size**, that will be
973 * filled with tunnel metadata for the packet associated to *skb*.
974 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
975 * indicates that the tunnel is based on IPv6 protocol instead of
976 * IPv4.
977 *
978 * The **struct bpf_tunnel_key** is an object that generalizes the
979 * principal parameters used by various tunneling protocols into a
980 * single struct. This way, it can be used to easily make a
981 * decision based on the contents of the encapsulation header,
982 * "summarized" in this struct. In particular, it holds the IP
983 * address of the remote end (IPv4 or IPv6, depending on the case)
984 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
985 * this struct exposes the *key*\ **->tunnel_id**, which is
986 * generally mapped to a VNI (Virtual Network Identifier), making
987 * it programmable together with the **bpf_skb_set_tunnel_key**\
988 * () helper.
989 *
990 * Let's imagine that the following code is part of a program
991 * attached to the TC ingress interface, on one end of a GRE
992 * tunnel, and is supposed to filter out all messages coming from
993 * remote ends with IPv4 address other than 10.0.0.1:
994 *
995 * ::
996 *
997 * int ret;
998 * struct bpf_tunnel_key key = {};
999 *
1000 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1001 * if (ret < 0)
1002 * return TC_ACT_SHOT; // drop packet
1003 *
1004 * if (key.remote_ipv4 != 0x0a000001)
1005 * return TC_ACT_SHOT; // drop packet
1006 *
1007 * return TC_ACT_OK; // accept packet
1008 *
1009 * This interface can also be used with all encapsulation devices
1010 * that can operate in "collect metadata" mode: instead of having
1011 * one network device per specific configuration, the "collect
1012 * metadata" mode only requires a single device where the
1013 * configuration can be extracted from this helper.
1014 *
1015 * This can be used together with various tunnels such as VXLan,
1016 * Geneve, GRE or IP in IP (IPIP).
1017 * Return
1018 * 0 on success, or a negative error in case of failure.
1019 *
1020 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1021 * Description
1022 * Populate tunnel metadata for packet associated to *skb.* The
1023 * tunnel metadata is set to the contents of *key*, of *size*. The
1024 * *flags* can be set to a combination of the following values:
1025 *
1026 * **BPF_F_TUNINFO_IPV6**
1027 * Indicate that the tunnel is based on IPv6 protocol
1028 * instead of IPv4.
1029 * **BPF_F_ZERO_CSUM_TX**
1030 * For IPv4 packets, add a flag to tunnel metadata
1031 * indicating that checksum computation should be skipped
1032 * and checksum set to zeroes.
1033 * **BPF_F_DONT_FRAGMENT**
1034 * Add a flag to tunnel metadata indicating that the
1035 * packet should not be fragmented.
1036 * **BPF_F_SEQ_NUMBER**
1037 * Add a flag to tunnel metadata indicating that a
1038 * sequence number should be added to tunnel header before
1039 * sending the packet. This flag was added for GRE
1040 * encapsulation, but might be used with other protocols
1041 * as well in the future.
1042 *
1043 * Here is a typical usage on the transmit path:
1044 *
1045 * ::
1046 *
1047 * struct bpf_tunnel_key key;
1048 * populate key ...
1049 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1050 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1051 *
1052 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1053 * helper for additional information.
1054 * Return
1055 * 0 on success, or a negative error in case of failure.
1056 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001057 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1058 * Description
1059 * Read the value of a perf event counter. This helper relies on a
1060 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1061 * the perf event counter is selected when *map* is updated with
1062 * perf event file descriptors. The *map* is an array whose size
1063 * is the number of available CPUs, and each cell contains a value
1064 * relative to one CPU. The value to retrieve is indicated by
1065 * *flags*, that contains the index of the CPU to look up, masked
1066 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1067 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1068 * current CPU should be retrieved.
1069 *
1070 * Note that before Linux 4.13, only hardware perf event can be
1071 * retrieved.
1072 *
1073 * Also, be aware that the newer helper
1074 * **bpf_perf_event_read_value**\ () is recommended over
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001075 * **bpf_perf_event_read**\ () in general. The latter has some ABI
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001076 * quirks where error and counter value are used as a return code
1077 * (which is wrong to do since ranges may overlap). This issue is
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001078 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1079 * time provides more features over the **bpf_perf_event_read**\
1080 * () interface. Please refer to the description of
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001081 * **bpf_perf_event_read_value**\ () for details.
1082 * Return
1083 * The value of the perf event counter read from the map, or a
1084 * negative error code in case of failure.
1085 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001086 * int bpf_redirect(u32 ifindex, u64 flags)
1087 * Description
1088 * Redirect the packet to another net device of index *ifindex*.
1089 * This helper is somewhat similar to **bpf_clone_redirect**\
1090 * (), except that the packet is not cloned, which provides
1091 * increased performance.
1092 *
1093 * Except for XDP, both ingress and egress interfaces can be used
1094 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1095 * to make the distinction (ingress path is selected if the flag
1096 * is present, egress path otherwise). Currently, XDP only
1097 * supports redirection to the egress interface, and accepts no
1098 * flag at all.
1099 *
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001100 * The same effect can also be attained with the more generic
1101 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1102 * redirect target instead of providing it directly to the helper.
Quentin Monnetc456dec2018-04-25 18:16:54 +01001103 * Return
1104 * For XDP, the helper returns **XDP_REDIRECT** on success or
1105 * **XDP_ABORTED** on error. For other program types, the values
1106 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1107 * error.
1108 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001109 * u32 bpf_get_route_realm(struct sk_buff *skb)
1110 * Description
1111 * Retrieve the realm or the route, that is to say the
1112 * **tclassid** field of the destination for the *skb*. The
1113 * indentifier retrieved is a user-provided tag, similar to the
1114 * one used with the net_cls cgroup (see description for
1115 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1116 * held by a route (a destination entry), not by a task.
1117 *
1118 * Retrieving this identifier works with the clsact TC egress hook
1119 * (see also **tc-bpf(8)**), or alternatively on conventional
1120 * classful egress qdiscs, but not on TC ingress path. In case of
1121 * clsact TC egress hook, this has the advantage that, internally,
1122 * the destination entry has not been dropped yet in the transmit
1123 * path. Therefore, the destination entry does not need to be
1124 * artificially held via **netif_keep_dst**\ () for a classful
1125 * qdisc until the *skb* is freed.
1126 *
1127 * This helper is available only if the kernel was compiled with
1128 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1129 * Return
1130 * The realm of the route for the packet associated to *skb*, or 0
1131 * if none was found.
1132 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001133 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001134 * Description
1135 * Write raw *data* blob into a special BPF perf event held by
1136 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1137 * event must have the following attributes: **PERF_SAMPLE_RAW**
1138 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1139 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1140 *
1141 * The *flags* are used to indicate the index in *map* for which
1142 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1143 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1144 * to indicate that the index of the current CPU core should be
1145 * used.
1146 *
1147 * The value to write, of *size*, is passed through eBPF stack and
1148 * pointed by *data*.
1149 *
1150 * The context of the program *ctx* needs also be passed to the
1151 * helper.
1152 *
1153 * On user space, a program willing to read the values needs to
1154 * call **perf_event_open**\ () on the perf event (either for
1155 * one or for all CPUs) and to store the file descriptor into the
1156 * *map*. This must be done before the eBPF program can send data
1157 * into it. An example is available in file
1158 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1159 * tree (the eBPF program counterpart is in
1160 * *samples/bpf/trace_output_kern.c*).
1161 *
1162 * **bpf_perf_event_output**\ () achieves better performance
1163 * than **bpf_trace_printk**\ () for sharing data with user
1164 * space, and is much better suitable for streaming data from eBPF
1165 * programs.
1166 *
1167 * Note that this helper is not restricted to tracing use cases
1168 * and can be used with programs attached to TC or XDP as well,
1169 * where it allows for passing data to user space listeners. Data
1170 * can be:
1171 *
1172 * * Only custom structs,
1173 * * Only the packet payload, or
1174 * * A combination of both.
1175 * Return
1176 * 0 on success, or a negative error in case of failure.
1177 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001178 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001179 * Description
1180 * This helper was provided as an easy way to load data from a
1181 * packet. It can be used to load *len* bytes from *offset* from
1182 * the packet associated to *skb*, into the buffer pointed by
1183 * *to*.
1184 *
1185 * Since Linux 4.7, usage of this helper has mostly been replaced
1186 * by "direct packet access", enabling packet data to be
1187 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1188 * pointing respectively to the first byte of packet data and to
1189 * the byte after the last byte of packet data. However, it
1190 * remains useful if one wishes to read large quantities of data
1191 * at once from a packet into the eBPF stack.
1192 * Return
1193 * 0 on success, or a negative error in case of failure.
1194 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001195 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001196 * Description
1197 * Walk a user or a kernel stack and return its id. To achieve
1198 * this, the helper needs *ctx*, which is a pointer to the context
1199 * on which the tracing program is executed, and a pointer to a
1200 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1201 *
1202 * The last argument, *flags*, holds the number of stack frames to
1203 * skip (from 0 to 255), masked with
1204 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1205 * a combination of the following flags:
1206 *
1207 * **BPF_F_USER_STACK**
1208 * Collect a user space stack instead of a kernel stack.
1209 * **BPF_F_FAST_STACK_CMP**
1210 * Compare stacks by hash only.
1211 * **BPF_F_REUSE_STACKID**
1212 * If two different stacks hash into the same *stackid*,
1213 * discard the old one.
1214 *
1215 * The stack id retrieved is a 32 bit long integer handle which
1216 * can be further combined with other data (including other stack
1217 * ids) and used as a key into maps. This can be useful for
1218 * generating a variety of graphs (such as flame graphs or off-cpu
1219 * graphs).
1220 *
1221 * For walking a stack, this helper is an improvement over
1222 * **bpf_probe_read**\ (), which can be used with unrolled loops
1223 * but is not efficient and consumes a lot of eBPF instructions.
1224 * Instead, **bpf_get_stackid**\ () can collect up to
1225 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1226 * this limit can be controlled with the **sysctl** program, and
1227 * that it should be manually increased in order to profile long
1228 * user stacks (such as stacks for Java programs). To do so, use:
1229 *
1230 * ::
1231 *
1232 * # sysctl kernel.perf_event_max_stack=<new value>
Quentin Monnetc456dec2018-04-25 18:16:54 +01001233 * Return
1234 * The positive or null stack id on success, or a negative error
1235 * in case of failure.
1236 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001237 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1238 * Description
1239 * Compute a checksum difference, from the raw buffer pointed by
1240 * *from*, of length *from_size* (that must be a multiple of 4),
1241 * towards the raw buffer pointed by *to*, of size *to_size*
1242 * (same remark). An optional *seed* can be added to the value
1243 * (this can be cascaded, the seed may come from a previous call
1244 * to the helper).
1245 *
1246 * This is flexible enough to be used in several ways:
1247 *
1248 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1249 * checksum, it can be used when pushing new data.
1250 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1251 * checksum, it can be used when removing data from a packet.
1252 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1253 * can be used to compute a diff. Note that *from_size* and
1254 * *to_size* do not need to be equal.
1255 *
1256 * This helper can be used in combination with
1257 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1258 * which one can feed in the difference computed with
1259 * **bpf_csum_diff**\ ().
1260 * Return
1261 * The checksum result, or a negative error code in case of
1262 * failure.
1263 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001264 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001265 * Description
1266 * Retrieve tunnel options metadata for the packet associated to
1267 * *skb*, and store the raw tunnel option data to the buffer *opt*
1268 * of *size*.
1269 *
1270 * This helper can be used with encapsulation devices that can
1271 * operate in "collect metadata" mode (please refer to the related
1272 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1273 * more details). A particular example where this can be used is
1274 * in combination with the Geneve encapsulation protocol, where it
1275 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1276 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1277 * the eBPF program. This allows for full customization of these
1278 * headers.
1279 * Return
1280 * The size of the option data retrieved.
1281 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001282 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001283 * Description
1284 * Set tunnel options metadata for the packet associated to *skb*
1285 * to the option data contained in the raw buffer *opt* of *size*.
1286 *
1287 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1288 * helper for additional information.
1289 * Return
1290 * 0 on success, or a negative error in case of failure.
1291 *
1292 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1293 * Description
1294 * Change the protocol of the *skb* to *proto*. Currently
1295 * supported are transition from IPv4 to IPv6, and from IPv6 to
1296 * IPv4. The helper takes care of the groundwork for the
1297 * transition, including resizing the socket buffer. The eBPF
1298 * program is expected to fill the new headers, if any, via
1299 * **skb_store_bytes**\ () and to recompute the checksums with
1300 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1301 * (). The main case for this helper is to perform NAT64
1302 * operations out of an eBPF program.
1303 *
1304 * Internally, the GSO type is marked as dodgy so that headers are
1305 * checked and segments are recalculated by the GSO/GRO engine.
1306 * The size for GSO target is adapted as well.
1307 *
1308 * All values for *flags* are reserved for future usage, and must
1309 * be left at zero.
1310 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001311 * A call to this helper is susceptible to change the underlying
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001312 * packet buffer. Therefore, at load time, all checks on pointers
1313 * previously done by the verifier are invalidated and must be
1314 * performed again, if the helper is used in combination with
1315 * direct packet access.
1316 * Return
1317 * 0 on success, or a negative error in case of failure.
1318 *
1319 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1320 * Description
1321 * Change the packet type for the packet associated to *skb*. This
1322 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1323 * the eBPF program does not have a write access to *skb*\
1324 * **->pkt_type** beside this helper. Using a helper here allows
1325 * for graceful handling of errors.
1326 *
1327 * The major use case is to change incoming *skb*s to
1328 * **PACKET_HOST** in a programmatic way instead of having to
1329 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1330 * example.
1331 *
1332 * Note that *type* only allows certain values. At this time, they
1333 * are:
1334 *
1335 * **PACKET_HOST**
1336 * Packet is for us.
1337 * **PACKET_BROADCAST**
1338 * Send packet to all.
1339 * **PACKET_MULTICAST**
1340 * Send packet to group.
1341 * **PACKET_OTHERHOST**
1342 * Send packet to someone else.
1343 * Return
1344 * 0 on success, or a negative error in case of failure.
1345 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001346 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1347 * Description
1348 * Check whether *skb* is a descendant of the cgroup2 held by
1349 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1350 * Return
1351 * The return value depends on the result of the test, and can be:
1352 *
1353 * * 0, if the *skb* failed the cgroup2 descendant test.
1354 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1355 * * A negative error code, if an error occurred.
1356 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001357 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1358 * Description
1359 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1360 * not set, in particular if the hash was cleared due to mangling,
1361 * recompute this hash. Later accesses to the hash can be done
1362 * directly with *skb*\ **->hash**.
1363 *
1364 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1365 * prototype with **bpf_skb_change_proto**\ (), or calling
1366 * **bpf_skb_store_bytes**\ () with the
1367 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1368 * the hash and to trigger a new computation for the next call to
1369 * **bpf_get_hash_recalc**\ ().
1370 * Return
1371 * The 32-bit hash.
1372 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001373 * u64 bpf_get_current_task(void)
1374 * Return
1375 * A pointer to the current task struct.
Quentin Monnetfa156012018-04-25 18:16:56 +01001376 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001377 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1378 * Description
1379 * Attempt in a safe way to write *len* bytes from the buffer
1380 * *src* to *dst* in memory. It only works for threads that are in
1381 * user context, and *dst* must be a valid user space address.
1382 *
1383 * This helper should not be used to implement any kind of
1384 * security mechanism because of TOC-TOU attacks, but rather to
1385 * debug, divert, and manipulate execution of semi-cooperative
1386 * processes.
1387 *
1388 * Keep in mind that this feature is meant for experiments, and it
1389 * has a risk of crashing the system and running programs.
1390 * Therefore, when an eBPF program using this helper is attached,
1391 * a warning including PID and process name is printed to kernel
1392 * logs.
1393 * Return
1394 * 0 on success, or a negative error in case of failure.
1395 *
1396 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1397 * Description
1398 * Check whether the probe is being run is the context of a given
1399 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1400 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1401 * Return
1402 * The return value depends on the result of the test, and can be:
1403 *
1404 * * 0, if the *skb* task belongs to the cgroup2.
1405 * * 1, if the *skb* task does not belong to the cgroup2.
1406 * * A negative error code, if an error occurred.
1407 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001408 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1409 * Description
1410 * Resize (trim or grow) the packet associated to *skb* to the
1411 * new *len*. The *flags* are reserved for future usage, and must
1412 * be left at zero.
1413 *
1414 * The basic idea is that the helper performs the needed work to
1415 * change the size of the packet, then the eBPF program rewrites
1416 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1417 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1418 * and others. This helper is a slow path utility intended for
1419 * replies with control messages. And because it is targeted for
1420 * slow path, the helper itself can afford to be slow: it
1421 * implicitly linearizes, unclones and drops offloads from the
1422 * *skb*.
1423 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001424 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001425 * packet buffer. Therefore, at load time, all checks on pointers
1426 * previously done by the verifier are invalidated and must be
1427 * performed again, if the helper is used in combination with
1428 * direct packet access.
1429 * Return
1430 * 0 on success, or a negative error in case of failure.
1431 *
1432 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1433 * Description
1434 * Pull in non-linear data in case the *skb* is non-linear and not
1435 * all of *len* are part of the linear section. Make *len* bytes
1436 * from *skb* readable and writable. If a zero value is passed for
1437 * *len*, then the whole length of the *skb* is pulled.
1438 *
1439 * This helper is only needed for reading and writing with direct
1440 * packet access.
1441 *
1442 * For direct packet access, testing that offsets to access
1443 * are within packet boundaries (test on *skb*\ **->data_end**) is
1444 * susceptible to fail if offsets are invalid, or if the requested
1445 * data is in non-linear parts of the *skb*. On failure the
1446 * program can just bail out, or in the case of a non-linear
1447 * buffer, use a helper to make the data available. The
1448 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1449 * the data. Another one consists in using **bpf_skb_pull_data**
1450 * to pull in once the non-linear parts, then retesting and
1451 * eventually access the data.
1452 *
1453 * At the same time, this also makes sure the *skb* is uncloned,
1454 * which is a necessary condition for direct write. As this needs
1455 * to be an invariant for the write part only, the verifier
1456 * detects writes and adds a prologue that is calling
1457 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1458 * the very beginning in case it is indeed cloned.
1459 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001460 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001461 * packet buffer. Therefore, at load time, all checks on pointers
1462 * previously done by the verifier are invalidated and must be
1463 * performed again, if the helper is used in combination with
1464 * direct packet access.
1465 * Return
1466 * 0 on success, or a negative error in case of failure.
1467 *
1468 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1469 * Description
1470 * Add the checksum *csum* into *skb*\ **->csum** in case the
1471 * driver has supplied a checksum for the entire packet into that
1472 * field. Return an error otherwise. This helper is intended to be
1473 * used in combination with **bpf_csum_diff**\ (), in particular
1474 * when the checksum needs to be updated after data has been
1475 * written into the packet through direct packet access.
1476 * Return
1477 * The checksum on success, or a negative error code in case of
1478 * failure.
1479 *
1480 * void bpf_set_hash_invalid(struct sk_buff *skb)
1481 * Description
1482 * Invalidate the current *skb*\ **->hash**. It can be used after
1483 * mangling on headers through direct packet access, in order to
1484 * indicate that the hash is outdated and to trigger a
1485 * recalculation the next time the kernel tries to access this
1486 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1487 *
1488 * int bpf_get_numa_node_id(void)
1489 * Description
1490 * Return the id of the current NUMA node. The primary use case
1491 * for this helper is the selection of sockets for the local NUMA
1492 * node, when the program is attached to sockets using the
1493 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1494 * but the helper is also available to other eBPF program types,
1495 * similarly to **bpf_get_smp_processor_id**\ ().
1496 * Return
1497 * The id of current NUMA node.
1498 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001499 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1500 * Description
1501 * Grows headroom of packet associated to *skb* and adjusts the
1502 * offset of the MAC header accordingly, adding *len* bytes of
1503 * space. It automatically extends and reallocates memory as
1504 * required.
1505 *
1506 * This helper can be used on a layer 3 *skb* to push a MAC header
1507 * for redirection into a layer 2 device.
1508 *
1509 * All values for *flags* are reserved for future usage, and must
1510 * be left at zero.
1511 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001512 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001513 * packet buffer. Therefore, at load time, all checks on pointers
1514 * previously done by the verifier are invalidated and must be
1515 * performed again, if the helper is used in combination with
1516 * direct packet access.
1517 * Return
1518 * 0 on success, or a negative error in case of failure.
1519 *
1520 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1521 * Description
1522 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1523 * it is possible to use a negative value for *delta*. This helper
1524 * can be used to prepare the packet for pushing or popping
1525 * headers.
1526 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001527 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001528 * packet buffer. Therefore, at load time, all checks on pointers
1529 * previously done by the verifier are invalidated and must be
1530 * performed again, if the helper is used in combination with
1531 * direct packet access.
1532 * Return
1533 * 0 on success, or a negative error in case of failure.
1534 *
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001535 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001536 * Description
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001537 * Copy a NUL terminated string from an unsafe kernel address
1538 * *unsafe_ptr* to *dst*. See bpf_probe_read_kernel_str() for
1539 * more details.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001540 *
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001541 * Generally, use bpf_probe_read_user_str() or bpf_probe_read_kernel_str()
1542 * instead.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001543 * Return
1544 * On success, the strictly positive length of the string,
1545 * including the trailing NUL character. On error, a negative
1546 * value.
1547 *
1548 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1549 * Description
1550 * If the **struct sk_buff** pointed by *skb* has a known socket,
1551 * retrieve the cookie (generated by the kernel) of this socket.
1552 * If no cookie has been set yet, generate a new cookie. Once
1553 * generated, the socket cookie remains stable for the life of the
1554 * socket. This helper can be useful for monitoring per socket
Daniel Borkmanncd48bdd2019-08-08 13:57:25 +02001555 * networking traffic statistics as it provides a global socket
1556 * identifier that can be assumed unique.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001557 * Return
1558 * A 8-byte long non-decreasing number on success, or 0 if the
1559 * socket field is missing inside *skb*.
1560 *
Andrey Ignatovd692f112018-07-30 17:42:28 -07001561 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1562 * Description
1563 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001564 * *skb*, but gets socket from **struct bpf_sock_addr** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001565 * Return
1566 * A 8-byte long non-decreasing number.
1567 *
1568 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1569 * Description
1570 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001571 * *skb*, but gets socket from **struct bpf_sock_ops** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001572 * Return
1573 * A 8-byte long non-decreasing number.
1574 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001575 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1576 * Return
1577 * The owner UID of the socket associated to *skb*. If the socket
1578 * is **NULL**, or if it is not a full socket (i.e. if it is a
1579 * time-wait or a request socket instead), **overflowuid** value
1580 * is returned (note that **overflowuid** might also be the actual
1581 * UID value for the socket).
1582 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001583 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1584 * Description
1585 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1586 * to value *hash*.
1587 * Return
1588 * 0
1589 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001590 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001591 * Description
1592 * Emulate a call to **setsockopt()** on the socket associated to
1593 * *bpf_socket*, which must be a full socket. The *level* at
1594 * which the option resides and the name *optname* of the option
1595 * must be specified, see **setsockopt(2)** for more information.
1596 * The option value of length *optlen* is pointed by *optval*.
1597 *
1598 * This helper actually implements a subset of **setsockopt()**.
1599 * It supports the following *level*\ s:
1600 *
1601 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1602 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1603 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1604 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1605 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1606 * **TCP_BPF_SNDCWND_CLAMP**.
1607 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1608 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1609 * Return
1610 * 0 on success, or a negative error in case of failure.
1611 *
Nicolas Dichtelb55cbc82018-10-17 16:24:48 +02001612 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001613 * Description
1614 * Grow or shrink the room for data in the packet associated to
1615 * *skb* by *len_diff*, and according to the selected *mode*.
1616 *
Willem de Bruijn14aa3192019-03-22 14:32:54 -04001617 * There are two supported modes at this time:
1618 *
1619 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1620 * (room space is added or removed below the layer 2 header).
Quentin Monnetfa156012018-04-25 18:16:56 +01001621 *
1622 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1623 * (room space is added or removed below the layer 3 header).
1624 *
Willem de Bruijn868d5232019-03-22 14:32:56 -04001625 * The following flags are supported at this time:
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04001626 *
1627 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1628 * Adjusting mss in this way is not allowed for datagrams.
Quentin Monnetfa156012018-04-25 18:16:56 +01001629 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001630 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1631 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001632 * Any new space is reserved to hold a tunnel header.
1633 * Configure skb offsets and other fields accordingly.
1634 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001635 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1636 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001637 * Use with ENCAP_L3 flags to further specify the tunnel type.
1638 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001639 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
Alan Maguire58dfc902019-04-09 15:06:41 +01001640 * Use with ENCAP_L3/L4 flags to further specify the tunnel
Quentin Monnet80867c52019-05-10 15:51:24 +01001641 * type; *len* is the length of the inner MAC header.
Alan Maguire58dfc902019-04-09 15:06:41 +01001642 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001643 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001644 * packet buffer. Therefore, at load time, all checks on pointers
1645 * previously done by the verifier are invalidated and must be
1646 * performed again, if the helper is used in combination with
1647 * direct packet access.
1648 * Return
1649 * 0 on success, or a negative error in case of failure.
1650 *
Quentin Monnetab127042018-04-25 18:16:59 +01001651 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1652 * Description
1653 * Redirect the packet to the endpoint referenced by *map* at
1654 * index *key*. Depending on its type, this *map* can contain
1655 * references to net devices (for forwarding packets through other
1656 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1657 * but this is only implemented for native XDP (with driver
1658 * support) as of this writing).
1659 *
Toke Høiland-Jørgensen43e74c02019-06-28 11:12:34 +02001660 * The lower two bits of *flags* are used as the return code if
1661 * the map lookup fails. This is so that the return value can be
1662 * one of the XDP program return codes up to XDP_TX, as chosen by
1663 * the caller. Any higher bits in the *flags* argument must be
1664 * unset.
Quentin Monnetab127042018-04-25 18:16:59 +01001665 *
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001666 * See also bpf_redirect(), which only supports redirecting to an
1667 * ifindex, but doesn't require a map to do so.
Quentin Monnetab127042018-04-25 18:16:59 +01001668 * Return
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001669 * **XDP_REDIRECT** on success, or the value of the two lower bits
Jakub Wilka33d3142020-04-22 10:23:24 +02001670 * of the *flags* argument on error.
Quentin Monnetab127042018-04-25 18:16:59 +01001671 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001672 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001673 * Description
1674 * Redirect the packet to the socket referenced by *map* (of type
1675 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1676 * egress interfaces can be used for redirection. The
1677 * **BPF_F_INGRESS** value in *flags* is used to make the
1678 * distinction (ingress path is selected if the flag is present,
1679 * egress path otherwise). This is the only flag supported for now.
1680 * Return
1681 * **SK_PASS** on success, or **SK_DROP** on error.
1682 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001683 * 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 +01001684 * Description
1685 * Add an entry to, or update a *map* referencing sockets. The
1686 * *skops* is used as a new value for the entry associated to
1687 * *key*. *flags* is one of:
1688 *
1689 * **BPF_NOEXIST**
1690 * The entry for *key* must not exist in the map.
1691 * **BPF_EXIST**
1692 * The entry for *key* must already exist in the map.
1693 * **BPF_ANY**
1694 * No condition on the existence of the entry for *key*.
1695 *
1696 * If the *map* has eBPF programs (parser and verdict), those will
1697 * be inherited by the socket being added. If the socket is
1698 * already attached to eBPF programs, this results in an error.
1699 * Return
1700 * 0 on success, or a negative error in case of failure.
1701 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001702 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1703 * Description
1704 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1705 * *delta* (which can be positive or negative). Note that this
1706 * operation modifies the address stored in *xdp_md*\ **->data**,
1707 * so the latter must be loaded only after the helper has been
1708 * called.
1709 *
1710 * The use of *xdp_md*\ **->data_meta** is optional and programs
1711 * are not required to use it. The rationale is that when the
1712 * packet is processed with XDP (e.g. as DoS filter), it is
1713 * possible to push further meta data along with it before passing
1714 * to the stack, and to give the guarantee that an ingress eBPF
1715 * program attached as a TC classifier on the same device can pick
1716 * this up for further post-processing. Since TC works with socket
1717 * buffers, it remains possible to set from XDP the **mark** or
1718 * **priority** pointers, or other pointers for the socket buffer.
1719 * Having this scratch space generic and programmable allows for
1720 * more flexibility as the user is free to store whatever meta
1721 * data they need.
1722 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001723 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001724 * packet buffer. Therefore, at load time, all checks on pointers
1725 * previously done by the verifier are invalidated and must be
1726 * performed again, if the helper is used in combination with
1727 * direct packet access.
1728 * Return
1729 * 0 on success, or a negative error in case of failure.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001730 *
1731 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1732 * Description
1733 * Read the value of a perf event counter, and store it into *buf*
1734 * of size *buf_size*. This helper relies on a *map* of type
1735 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1736 * counter is selected when *map* is updated with perf event file
1737 * descriptors. The *map* is an array whose size is the number of
1738 * available CPUs, and each cell contains a value relative to one
1739 * CPU. The value to retrieve is indicated by *flags*, that
1740 * contains the index of the CPU to look up, masked with
1741 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1742 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1743 * current CPU should be retrieved.
1744 *
1745 * This helper behaves in a way close to
1746 * **bpf_perf_event_read**\ () helper, save that instead of
1747 * just returning the value observed, it fills the *buf*
1748 * structure. This allows for additional data to be retrieved: in
1749 * particular, the enabled and running times (in *buf*\
1750 * **->enabled** and *buf*\ **->running**, respectively) are
1751 * copied. In general, **bpf_perf_event_read_value**\ () is
1752 * recommended over **bpf_perf_event_read**\ (), which has some
1753 * ABI issues and provides fewer functionalities.
1754 *
1755 * These values are interesting, because hardware PMU (Performance
1756 * Monitoring Unit) counters are limited resources. When there are
1757 * more PMU based perf events opened than available counters,
1758 * kernel will multiplex these events so each event gets certain
1759 * percentage (but not all) of the PMU time. In case that
1760 * multiplexing happens, the number of samples or counter value
1761 * will not reflect the case compared to when no multiplexing
1762 * occurs. This makes comparison between different runs difficult.
1763 * Typically, the counter value should be normalized before
1764 * comparing to other experiments. The usual normalization is done
1765 * as follows.
1766 *
1767 * ::
1768 *
1769 * normalized_counter = counter * t_enabled / t_running
1770 *
1771 * Where t_enabled is the time enabled for event and t_running is
1772 * the time running for event since last normalization. The
1773 * enabled and running times are accumulated since the perf event
1774 * open. To achieve scaling factor between two invocations of an
1775 * eBPF program, users can can use CPU id as the key (which is
1776 * typical for perf array usage model) to remember the previous
1777 * value and do the calculation inside the eBPF program.
1778 * Return
1779 * 0 on success, or a negative error in case of failure.
1780 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001781 * 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 +01001782 * Description
1783 * For en eBPF program attached to a perf event, retrieve the
1784 * value of the event counter associated to *ctx* and store it in
1785 * the structure pointed by *buf* and of size *buf_size*. Enabled
1786 * and running times are also stored in the structure (see
1787 * description of helper **bpf_perf_event_read_value**\ () for
1788 * more details).
1789 * Return
1790 * 0 on success, or a negative error in case of failure.
1791 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001792 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001793 * Description
1794 * Emulate a call to **getsockopt()** on the socket associated to
1795 * *bpf_socket*, which must be a full socket. The *level* at
1796 * which the option resides and the name *optname* of the option
1797 * must be specified, see **getsockopt(2)** for more information.
1798 * The retrieved value is stored in the structure pointed by
1799 * *opval* and of length *optlen*.
1800 *
1801 * This helper actually implements a subset of **getsockopt()**.
1802 * It supports the following *level*\ s:
1803 *
1804 * * **IPPROTO_TCP**, which supports *optname*
1805 * **TCP_CONGESTION**.
1806 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1807 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1808 * Return
1809 * 0 on success, or a negative error in case of failure.
1810 *
Peter Wu8050a392019-08-21 00:08:58 +01001811 * int bpf_override_return(struct pt_regs *regs, u64 rc)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001812 * Description
1813 * Used for error injection, this helper uses kprobes to override
1814 * the return value of the probed function, and to set it to *rc*.
1815 * The first argument is the context *regs* on which the kprobe
1816 * works.
1817 *
1818 * This helper works by setting setting the PC (program counter)
1819 * to an override function which is run in place of the original
1820 * probed function. This means the probed function is not run at
1821 * all. The replacement function just returns with the required
1822 * value.
1823 *
1824 * This helper has security implications, and thus is subject to
1825 * restrictions. It is only available if the kernel was compiled
1826 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1827 * option, and in this case it only works on functions tagged with
1828 * **ALLOW_ERROR_INJECTION** in the kernel code.
1829 *
1830 * Also, the helper is only available for the architectures having
1831 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1832 * x86 architecture is the only one to support this feature.
1833 * Return
1834 * 0
1835 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001836 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001837 * Description
1838 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1839 * for the full TCP socket associated to *bpf_sock_ops* to
1840 * *argval*.
1841 *
1842 * The primary use of this field is to determine if there should
1843 * be calls to eBPF programs of type
1844 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1845 * code. A program of the same type can change its value, per
1846 * connection and as necessary, when the connection is
1847 * established. This field is directly accessible for reading, but
1848 * this helper must be used for updates in order to return an
1849 * error if an eBPF program tries to set a callback that is not
1850 * supported in the current kernel.
1851 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001852 * *argval* is a flag array which can combine these flags:
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001853 *
1854 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1855 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1856 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07001857 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001858 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001859 * Therefore, this function can be used to clear a callback flag by
1860 * setting the appropriate bit to zero. e.g. to disable the RTO
1861 * callback:
1862 *
1863 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1864 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1865 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001866 * Here are some examples of where one could call such eBPF
1867 * program:
1868 *
1869 * * When RTO fires.
1870 * * When a packet is retransmitted.
1871 * * When the connection terminates.
1872 * * When a packet is sent.
1873 * * When a packet is received.
1874 * Return
1875 * Code **-EINVAL** if the socket is not a full TCP socket;
1876 * otherwise, a positive number containing the bits that could not
1877 * be set is returned (which comes down to 0 if all bits were set
1878 * as required).
1879 *
Quentin Monnetab127042018-04-25 18:16:59 +01001880 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1881 * Description
1882 * This helper is used in programs implementing policies at the
1883 * socket level. If the message *msg* is allowed to pass (i.e. if
1884 * the verdict eBPF program returns **SK_PASS**), redirect it to
1885 * the socket referenced by *map* (of type
1886 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1887 * egress interfaces can be used for redirection. The
1888 * **BPF_F_INGRESS** value in *flags* is used to make the
1889 * distinction (ingress path is selected if the flag is present,
1890 * egress path otherwise). This is the only flag supported for now.
1891 * Return
1892 * **SK_PASS** on success, or **SK_DROP** on error.
1893 *
1894 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1895 * Description
1896 * For socket policies, apply the verdict of the eBPF program to
1897 * the next *bytes* (number of bytes) of message *msg*.
1898 *
1899 * For example, this helper can be used in the following cases:
1900 *
1901 * * A single **sendmsg**\ () or **sendfile**\ () system call
1902 * contains multiple logical messages that the eBPF program is
1903 * supposed to read and for which it should apply a verdict.
1904 * * An eBPF program only cares to read the first *bytes* of a
1905 * *msg*. If the message has a large payload, then setting up
1906 * and calling the eBPF program repeatedly for all bytes, even
1907 * though the verdict is already known, would create unnecessary
1908 * overhead.
1909 *
1910 * When called from within an eBPF program, the helper sets a
1911 * counter internal to the BPF infrastructure, that is used to
1912 * apply the last verdict to the next *bytes*. If *bytes* is
1913 * smaller than the current data being processed from a
1914 * **sendmsg**\ () or **sendfile**\ () system call, the first
1915 * *bytes* will be sent and the eBPF program will be re-run with
1916 * the pointer for start of data pointing to byte number *bytes*
1917 * **+ 1**. If *bytes* is larger than the current data being
1918 * processed, then the eBPF verdict will be applied to multiple
1919 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1920 * consumed.
1921 *
1922 * Note that if a socket closes with the internal counter holding
1923 * a non-zero value, this is not a problem because data is not
1924 * being buffered for *bytes* and is sent as it is received.
1925 * Return
1926 * 0
1927 *
1928 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1929 * Description
1930 * For socket policies, prevent the execution of the verdict eBPF
1931 * program for message *msg* until *bytes* (byte number) have been
1932 * accumulated.
1933 *
1934 * This can be used when one needs a specific number of bytes
1935 * before a verdict can be assigned, even if the data spans
1936 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1937 * case would be a user calling **sendmsg**\ () repeatedly with
1938 * 1-byte long message segments. Obviously, this is bad for
1939 * performance, but it is still valid. If the eBPF program needs
1940 * *bytes* bytes to validate a header, this helper can be used to
1941 * prevent the eBPF program to be called again until *bytes* have
1942 * been accumulated.
1943 * Return
1944 * 0
1945 *
1946 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1947 * Description
1948 * For socket policies, pull in non-linear data from user space
1949 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1950 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1951 * respectively.
1952 *
1953 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1954 * *msg* it can only parse data that the (**data**, **data_end**)
1955 * pointers have already consumed. For **sendmsg**\ () hooks this
1956 * is likely the first scatterlist element. But for calls relying
1957 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1958 * be the range (**0**, **0**) because the data is shared with
1959 * user space and by default the objective is to avoid allowing
1960 * user space to modify data while (or after) eBPF verdict is
1961 * being decided. This helper can be used to pull in data and to
1962 * set the start and end pointer to given values. Data will be
1963 * copied if necessary (i.e. if data was not linear and if start
1964 * and end pointers do not point to the same chunk).
1965 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001966 * A call to this helper is susceptible to change the underlying
Quentin Monnetab127042018-04-25 18:16:59 +01001967 * packet buffer. Therefore, at load time, all checks on pointers
1968 * previously done by the verifier are invalidated and must be
1969 * performed again, if the helper is used in combination with
1970 * direct packet access.
1971 *
1972 * All values for *flags* are reserved for future usage, and must
1973 * be left at zero.
1974 * Return
1975 * 0 on success, or a negative error in case of failure.
1976 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001977 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001978 * Description
1979 * Bind the socket associated to *ctx* to the address pointed by
1980 * *addr*, of length *addr_len*. This allows for making outgoing
1981 * connection from the desired IP address, which can be useful for
1982 * example when all processes inside a cgroup should use one
1983 * single IP address on a host that has multiple IP configured.
1984 *
1985 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1986 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1987 * **AF_INET6**). Looking for a free port to bind to can be
1988 * expensive, therefore binding to port is not permitted by the
1989 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1990 * must be set to zero.
1991 * Return
1992 * 0 on success, or a negative error in case of failure.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01001993 *
1994 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1995 * Description
1996 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1997 * only possible to shrink the packet as of this writing,
1998 * therefore *delta* must be a negative integer.
1999 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002000 * A call to this helper is susceptible to change the underlying
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002001 * packet buffer. Therefore, at load time, all checks on pointers
2002 * previously done by the verifier are invalidated and must be
2003 * performed again, if the helper is used in combination with
2004 * direct packet access.
2005 * Return
2006 * 0 on success, or a negative error in case of failure.
2007 *
2008 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2009 * Description
2010 * Retrieve the XFRM state (IP transform framework, see also
2011 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2012 *
2013 * The retrieved value is stored in the **struct bpf_xfrm_state**
2014 * pointed by *xfrm_state* and of length *size*.
2015 *
2016 * All values for *flags* are reserved for future usage, and must
2017 * be left at zero.
2018 *
2019 * This helper is available only if the kernel was compiled with
2020 * **CONFIG_XFRM** configuration option.
2021 * Return
2022 * 0 on success, or a negative error in case of failure.
Yonghong Songc195651e2018-04-28 22:28:08 -07002023 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002024 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
Yonghong Songc195651e2018-04-28 22:28:08 -07002025 * Description
Quentin Monnet79552fb2018-04-30 11:39:04 +01002026 * Return a user or a kernel stack in bpf program provided buffer.
2027 * To achieve this, the helper needs *ctx*, which is a pointer
2028 * to the context on which the tracing program is executed.
2029 * To store the stacktrace, the bpf program provides *buf* with
2030 * a nonnegative *size*.
Yonghong Songc195651e2018-04-28 22:28:08 -07002031 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002032 * The last argument, *flags*, holds the number of stack frames to
2033 * skip (from 0 to 255), masked with
2034 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2035 * the following flags:
Yonghong Songc195651e2018-04-28 22:28:08 -07002036 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002037 * **BPF_F_USER_STACK**
2038 * Collect a user space stack instead of a kernel stack.
2039 * **BPF_F_USER_BUILD_ID**
2040 * Collect buildid+offset instead of ips for user stack,
2041 * only valid if **BPF_F_USER_STACK** is also specified.
Yonghong Songc195651e2018-04-28 22:28:08 -07002042 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002043 * **bpf_get_stack**\ () can collect up to
2044 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2045 * to sufficient large buffer size. Note that
2046 * this limit can be controlled with the **sysctl** program, and
2047 * that it should be manually increased in order to profile long
2048 * user stacks (such as stacks for Java programs). To do so, use:
Yonghong Songc195651e2018-04-28 22:28:08 -07002049 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002050 * ::
Yonghong Songc195651e2018-04-28 22:28:08 -07002051 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002052 * # sysctl kernel.perf_event_max_stack=<new value>
Yonghong Songc195651e2018-04-28 22:28:08 -07002053 * Return
Quentin Monnet7a279e92018-05-29 12:27:44 +01002054 * A non-negative value equal to or less than *size* on success,
2055 * or a negative error in case of failure.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002056 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002057 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002058 * Description
2059 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2060 * it provides an easy way to load *len* bytes from *offset*
2061 * from the packet associated to *skb*, into the buffer pointed
2062 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2063 * a fifth argument *start_header* exists in order to select a
2064 * base offset to start from. *start_header* can be one of:
2065 *
2066 * **BPF_HDR_START_MAC**
2067 * Base offset to load data from is *skb*'s mac header.
2068 * **BPF_HDR_START_NET**
2069 * Base offset to load data from is *skb*'s network header.
2070 *
2071 * In general, "direct packet access" is the preferred method to
2072 * access packet data, however, this helper is in particular useful
2073 * in socket filters where *skb*\ **->data** does not always point
2074 * to the start of the mac header and where "direct packet access"
2075 * is not available.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002076 * Return
2077 * 0 on success, or a negative error in case of failure.
2078 *
David Ahern87f5fc72018-05-09 20:34:26 -07002079 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2080 * Description
2081 * Do FIB lookup in kernel tables using parameters in *params*.
2082 * If lookup is successful and result shows packet is to be
2083 * forwarded, the neighbor tables are searched for the nexthop.
2084 * If successful (ie., FIB lookup shows forwarding and nexthop
David Ahernfa898d72018-05-29 10:58:07 -07002085 * is resolved), the nexthop address is returned in ipv4_dst
2086 * or ipv6_dst based on family, smac is set to mac address of
2087 * egress device, dmac is set to nexthop mac address, rt_metric
David Ahern4c795792018-06-26 16:21:18 -07002088 * is set to metric from route (IPv4/IPv6 only), and ifindex
2089 * is set to the device index of the nexthop from the FIB lookup.
David Ahern87f5fc72018-05-09 20:34:26 -07002090 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002091 * *plen* argument is the size of the passed in struct.
2092 * *flags* argument can be a combination of one or more of the
2093 * following values:
David Ahern87f5fc72018-05-09 20:34:26 -07002094 *
Quentin Monnet7a279e92018-05-29 12:27:44 +01002095 * **BPF_FIB_LOOKUP_DIRECT**
2096 * Do a direct table lookup vs full lookup using FIB
2097 * rules.
2098 * **BPF_FIB_LOOKUP_OUTPUT**
2099 * Perform lookup from an egress perspective (default is
2100 * ingress).
David Ahern87f5fc72018-05-09 20:34:26 -07002101 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002102 * *ctx* is either **struct xdp_md** for XDP programs or
2103 * **struct sk_buff** tc cls_act programs.
2104 * Return
David Ahern4c795792018-06-26 16:21:18 -07002105 * * < 0 if any input argument is invalid
2106 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2107 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
Quentin Monnet2bae79d2018-07-12 12:52:22 +01002108 * packet is not forwarded or needs assist from full stack
John Fastabend81110382018-05-14 10:00:17 -07002109 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002110 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
John Fastabend81110382018-05-14 10:00:17 -07002111 * Description
2112 * Add an entry to, or update a sockhash *map* referencing sockets.
2113 * The *skops* is used as a new value for the entry associated to
2114 * *key*. *flags* is one of:
2115 *
2116 * **BPF_NOEXIST**
2117 * The entry for *key* must not exist in the map.
2118 * **BPF_EXIST**
2119 * The entry for *key* must already exist in the map.
2120 * **BPF_ANY**
2121 * No condition on the existence of the entry for *key*.
2122 *
2123 * If the *map* has eBPF programs (parser and verdict), those will
2124 * be inherited by the socket being added. If the socket is
2125 * already attached to eBPF programs, this results in an error.
2126 * Return
2127 * 0 on success, or a negative error in case of failure.
2128 *
2129 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2130 * Description
2131 * This helper is used in programs implementing policies at the
2132 * socket level. If the message *msg* is allowed to pass (i.e. if
2133 * the verdict eBPF program returns **SK_PASS**), redirect it to
2134 * the socket referenced by *map* (of type
2135 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2136 * egress interfaces can be used for redirection. The
2137 * **BPF_F_INGRESS** value in *flags* is used to make the
2138 * distinction (ingress path is selected if the flag is present,
2139 * egress path otherwise). This is the only flag supported for now.
2140 * Return
2141 * **SK_PASS** on success, or **SK_DROP** on error.
2142 *
2143 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2144 * Description
2145 * This helper is used in programs implementing policies at the
2146 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2147 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2148 * to the socket referenced by *map* (of type
2149 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2150 * egress interfaces can be used for redirection. The
2151 * **BPF_F_INGRESS** value in *flags* is used to make the
2152 * distinction (ingress path is selected if the flag is present,
2153 * egress otherwise). This is the only flag supported for now.
2154 * Return
2155 * **SK_PASS** on success, or **SK_DROP** on error.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002156 *
2157 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2158 * Description
2159 * Encapsulate the packet associated to *skb* within a Layer 3
2160 * protocol header. This header is provided in the buffer at
2161 * address *hdr*, with *len* its size in bytes. *type* indicates
2162 * the protocol of the header and can be one of:
2163 *
2164 * **BPF_LWT_ENCAP_SEG6**
2165 * IPv6 encapsulation with Segment Routing Header
2166 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2167 * the IPv6 header is computed by the kernel.
2168 * **BPF_LWT_ENCAP_SEG6_INLINE**
2169 * Only works if *skb* contains an IPv6 packet. Insert a
2170 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2171 * the IPv6 header.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002172 * **BPF_LWT_ENCAP_IP**
2173 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2174 * must be IPv4 or IPv6, followed by zero or more
Quentin Monnet80867c52019-05-10 15:51:24 +01002175 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2176 * total bytes in all prepended headers. Please note that
2177 * if **skb_is_gso**\ (*skb*) is true, no more than two
2178 * headers can be prepended, and the inner header, if
2179 * present, should be either GRE or UDP/GUE.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002180 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002181 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2182 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2183 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2184 * **BPF_PROG_TYPE_LWT_XMIT**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002185 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002186 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002187 * packet buffer. Therefore, at load time, all checks on pointers
2188 * previously done by the verifier are invalidated and must be
2189 * performed again, if the helper is used in combination with
2190 * direct packet access.
2191 * Return
2192 * 0 on success, or a negative error in case of failure.
2193 *
2194 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2195 * Description
2196 * Store *len* bytes from address *from* into the packet
2197 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2198 * inside the outermost IPv6 Segment Routing Header can be
2199 * modified through this helper.
2200 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002201 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002202 * packet buffer. Therefore, at load time, all checks on pointers
2203 * previously done by the verifier are invalidated and must be
2204 * performed again, if the helper is used in combination with
2205 * direct packet access.
2206 * Return
2207 * 0 on success, or a negative error in case of failure.
2208 *
2209 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2210 * Description
2211 * Adjust the size allocated to TLVs in the outermost IPv6
2212 * Segment Routing Header contained in the packet associated to
2213 * *skb*, at position *offset* by *delta* bytes. Only offsets
2214 * after the segments are accepted. *delta* can be as well
2215 * positive (growing) as negative (shrinking).
2216 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002217 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002218 * packet buffer. Therefore, at load time, all checks on pointers
2219 * previously done by the verifier are invalidated and must be
2220 * performed again, if the helper is used in combination with
2221 * direct packet access.
2222 * Return
2223 * 0 on success, or a negative error in case of failure.
2224 *
2225 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2226 * Description
2227 * Apply an IPv6 Segment Routing action of type *action* to the
2228 * packet associated to *skb*. Each action takes a parameter
2229 * contained at address *param*, and of length *param_len* bytes.
2230 * *action* can be one of:
2231 *
2232 * **SEG6_LOCAL_ACTION_END_X**
2233 * End.X action: Endpoint with Layer-3 cross-connect.
2234 * Type of *param*: **struct in6_addr**.
2235 * **SEG6_LOCAL_ACTION_END_T**
2236 * End.T action: Endpoint with specific IPv6 table lookup.
2237 * Type of *param*: **int**.
2238 * **SEG6_LOCAL_ACTION_END_B6**
2239 * End.B6 action: Endpoint bound to an SRv6 policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002240 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002241 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2242 * End.B6.Encap action: Endpoint bound to an SRv6
2243 * encapsulation policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002244 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002245 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002246 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002247 * packet buffer. Therefore, at load time, all checks on pointers
2248 * previously done by the verifier are invalidated and must be
2249 * performed again, if the helper is used in combination with
2250 * direct packet access.
2251 * Return
2252 * 0 on success, or a negative error in case of failure.
Sean Youngf4364dc2018-05-27 12:24:09 +01002253 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002254 * int bpf_rc_repeat(void *ctx)
2255 * Description
2256 * This helper is used in programs implementing IR decoding, to
2257 * report a successfully decoded repeat key message. This delays
2258 * the generation of a key up event for previously generated
2259 * key down event.
2260 *
2261 * Some IR protocols like NEC have a special IR message for
2262 * repeating last button, for when a button is held down.
2263 *
2264 * The *ctx* should point to the lirc sample as passed into
2265 * the program.
2266 *
2267 * This helper is only available is the kernel was compiled with
2268 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2269 * "**y**".
2270 * Return
2271 * 0
2272 *
Sean Youngf4364dc2018-05-27 12:24:09 +01002273 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2274 * Description
2275 * This helper is used in programs implementing IR decoding, to
2276 * report a successfully decoded key press with *scancode*,
2277 * *toggle* value in the given *protocol*. The scancode will be
2278 * translated to a keycode using the rc keymap, and reported as
2279 * an input key down event. After a period a key up event is
2280 * generated. This period can be extended by calling either
Quentin Monnet90b10232018-12-03 12:13:35 +00002281 * **bpf_rc_keydown**\ () again with the same values, or calling
2282 * **bpf_rc_repeat**\ ().
Sean Youngf4364dc2018-05-27 12:24:09 +01002283 *
2284 * Some protocols include a toggle bit, in case the button was
2285 * released and pressed again between consecutive scancodes.
2286 *
2287 * The *ctx* should point to the lirc sample as passed into
2288 * the program.
2289 *
2290 * The *protocol* is the decoded protocol number (see
2291 * **enum rc_proto** for some predefined values).
2292 *
2293 * This helper is only available is the kernel was compiled with
2294 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2295 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002296 * Return
2297 * 0
2298 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002299 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002300 * Description
2301 * Return the cgroup v2 id of the socket associated with the *skb*.
2302 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2303 * helper for cgroup v1 by providing a tag resp. identifier that
2304 * can be matched on or used for map lookups e.g. to implement
2305 * policy. The cgroup v2 id of a given path in the hierarchy is
2306 * exposed in user space through the f_handle API in order to get
2307 * to the same 64-bit id.
2308 *
2309 * This helper can be used on TC egress path, but not on ingress,
2310 * and is available only if the kernel was compiled with the
2311 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2312 * Return
2313 * The id is returned or 0 in case the id could not be retrieved.
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002314 *
2315 * u64 bpf_get_current_cgroup_id(void)
2316 * Return
2317 * A 64-bit integer containing the current cgroup id based
2318 * on the cgroup within which the current task is running.
Roman Gushchincd339432018-08-02 14:27:24 -07002319 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002320 * void *bpf_get_local_storage(void *map, u64 flags)
Roman Gushchincd339432018-08-02 14:27:24 -07002321 * Description
2322 * Get the pointer to the local storage area.
2323 * The type and the size of the local storage is defined
2324 * by the *map* argument.
2325 * The *flags* meaning is specific for each map type,
2326 * and has to be 0 for cgroup local storage.
2327 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002328 * Depending on the BPF program type, a local storage area
2329 * can be shared between multiple instances of the BPF program,
Roman Gushchincd339432018-08-02 14:27:24 -07002330 * running simultaneously.
2331 *
2332 * A user should care about the synchronization by himself.
Quentin Monnet90b10232018-12-03 12:13:35 +00002333 * For example, by using the **BPF_STX_XADD** instruction to alter
Roman Gushchincd339432018-08-02 14:27:24 -07002334 * the shared data.
2335 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002336 * A pointer to the local storage area.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002337 *
2338 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2339 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002340 * Select a **SO_REUSEPORT** socket from a
2341 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2342 * It checks the selected socket is matching the incoming
2343 * request in the socket buffer.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002344 * Return
2345 * 0 on success, or a negative error in case of failure.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002346 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002347 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2348 * Description
2349 * Return id of cgroup v2 that is ancestor of cgroup associated
2350 * with the *skb* at the *ancestor_level*. The root cgroup is at
2351 * *ancestor_level* zero and each step down the hierarchy
2352 * increments the level. If *ancestor_level* == level of cgroup
2353 * associated with *skb*, then return value will be same as that
2354 * of **bpf_skb_cgroup_id**\ ().
2355 *
2356 * The helper is useful to implement policies based on cgroups
2357 * that are upper in hierarchy than immediate cgroup associated
2358 * with *skb*.
2359 *
2360 * The format of returned id and helper limitations are same as in
2361 * **bpf_skb_cgroup_id**\ ().
2362 * Return
2363 * The id is returned or 0 in case the id could not be retrieved.
2364 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002365 * 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 -07002366 * Description
2367 * Look for TCP socket matching *tuple*, optionally in a child
2368 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002369 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002370 *
2371 * The *ctx* should point to the context of the program, such as
2372 * the skb or socket (depending on the hook in use). This is used
2373 * to determine the base network namespace for the lookup.
2374 *
2375 * *tuple_size* must be one of:
2376 *
2377 * **sizeof**\ (*tuple*\ **->ipv4**)
2378 * Look for an IPv4 socket.
2379 * **sizeof**\ (*tuple*\ **->ipv6**)
2380 * Look for an IPv6 socket.
2381 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002382 * If the *netns* is a negative signed 32-bit integer, then the
2383 * socket lookup table in the netns associated with the *ctx* will
2384 * will be used. For the TC hooks, this is the netns of the device
2385 * in the skb. For socket hooks, this is the netns of the socket.
2386 * If *netns* is any other signed 32-bit value greater than or
2387 * equal to zero then it specifies the ID of the netns relative to
2388 * the netns associated with the *ctx*. *netns* values beyond the
2389 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002390 *
2391 * All values for *flags* are reserved for future usage, and must
2392 * be left at zero.
2393 *
2394 * This helper is available only if the kernel was compiled with
2395 * **CONFIG_NET** configuration option.
2396 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002397 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2398 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002399 * result is from *reuse*\ **->socks**\ [] using the hash of the
2400 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002401 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002402 * 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 -07002403 * Description
2404 * Look for UDP socket matching *tuple*, optionally in a child
2405 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002406 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002407 *
2408 * The *ctx* should point to the context of the program, such as
2409 * the skb or socket (depending on the hook in use). This is used
2410 * to determine the base network namespace for the lookup.
2411 *
2412 * *tuple_size* must be one of:
2413 *
2414 * **sizeof**\ (*tuple*\ **->ipv4**)
2415 * Look for an IPv4 socket.
2416 * **sizeof**\ (*tuple*\ **->ipv6**)
2417 * Look for an IPv6 socket.
2418 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002419 * If the *netns* is a negative signed 32-bit integer, then the
2420 * socket lookup table in the netns associated with the *ctx* will
2421 * will be used. For the TC hooks, this is the netns of the device
2422 * in the skb. For socket hooks, this is the netns of the socket.
2423 * If *netns* is any other signed 32-bit value greater than or
2424 * equal to zero then it specifies the ID of the netns relative to
2425 * the netns associated with the *ctx*. *netns* values beyond the
2426 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002427 *
2428 * All values for *flags* are reserved for future usage, and must
2429 * be left at zero.
2430 *
2431 * This helper is available only if the kernel was compiled with
2432 * **CONFIG_NET** configuration option.
2433 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002434 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2435 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002436 * result is from *reuse*\ **->socks**\ [] using the hash of the
2437 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002438 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002439 * int bpf_sk_release(struct bpf_sock *sock)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002440 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002441 * Release the reference held by *sock*. *sock* must be a
2442 * non-**NULL** pointer that was returned from
2443 * **bpf_sk_lookup_xxx**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002444 * Return
2445 * 0 on success, or a negative error in case of failure.
John Fastabend6fff6072018-10-19 19:56:49 -07002446 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002447 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2448 * Description
2449 * Push an element *value* in *map*. *flags* is one of:
2450 *
2451 * **BPF_EXIST**
2452 * If the queue/stack is full, the oldest element is
2453 * removed to make room for this.
2454 * Return
2455 * 0 on success, or a negative error in case of failure.
2456 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002457 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2458 * Description
2459 * Pop an element from *map*.
2460 * Return
2461 * 0 on success, or a negative error in case of failure.
2462 *
2463 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2464 * Description
2465 * Get an element from *map* without removing it.
2466 * Return
2467 * 0 on success, or a negative error in case of failure.
2468 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002469 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
John Fastabend6fff6072018-10-19 19:56:49 -07002470 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002471 * For socket policies, insert *len* bytes into *msg* at offset
John Fastabend6fff6072018-10-19 19:56:49 -07002472 * *start*.
2473 *
2474 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
Quentin Monnet90b10232018-12-03 12:13:35 +00002475 * *msg* it may want to insert metadata or options into the *msg*.
John Fastabend6fff6072018-10-19 19:56:49 -07002476 * This can later be read and used by any of the lower layer BPF
2477 * hooks.
2478 *
2479 * This helper may fail if under memory pressure (a malloc
2480 * fails) in these cases BPF programs will get an appropriate
2481 * error and BPF programs will need to handle them.
John Fastabend6fff6072018-10-19 19:56:49 -07002482 * Return
2483 * 0 on success, or a negative error in case of failure.
John Fastabend7246d8e2018-11-26 14:16:17 -08002484 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002485 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
Quentin Monnet90b10232018-12-03 12:13:35 +00002486 * Description
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002487 * Will remove *len* bytes from a *msg* starting at byte *start*.
John Fastabend7246d8e2018-11-26 14:16:17 -08002488 * This may result in **ENOMEM** errors under certain situations if
2489 * an allocation and copy are required due to a full ring buffer.
2490 * However, the helper will try to avoid doing the allocation
2491 * if possible. Other errors can occur if input parameters are
Quentin Monnet90b10232018-12-03 12:13:35 +00002492 * invalid either due to *start* byte not being valid part of *msg*
John Fastabend7246d8e2018-11-26 14:16:17 -08002493 * payload and/or *pop* value being to large.
John Fastabend7246d8e2018-11-26 14:16:17 -08002494 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002495 * 0 on success, or a negative error in case of failure.
Sean Young01d32402018-12-06 13:01:03 +00002496 *
2497 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2498 * Description
2499 * This helper is used in programs implementing IR decoding, to
2500 * report a successfully decoded pointer movement.
2501 *
2502 * The *ctx* should point to the lirc sample as passed into
2503 * the program.
2504 *
2505 * This helper is only available is the kernel was compiled with
2506 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2507 * "**y**".
2508 * Return
2509 * 0
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002510 *
Quentin Monnet0eb09782019-03-14 12:38:40 +00002511 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2512 * Description
2513 * Acquire a spinlock represented by the pointer *lock*, which is
2514 * stored as part of a value of a map. Taking the lock allows to
2515 * safely update the rest of the fields in that value. The
2516 * spinlock can (and must) later be released with a call to
2517 * **bpf_spin_unlock**\ (\ *lock*\ ).
2518 *
2519 * Spinlocks in BPF programs come with a number of restrictions
2520 * and constraints:
2521 *
2522 * * **bpf_spin_lock** objects are only allowed inside maps of
2523 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2524 * list could be extended in the future).
2525 * * BTF description of the map is mandatory.
2526 * * The BPF program can take ONE lock at a time, since taking two
2527 * or more could cause dead locks.
2528 * * Only one **struct bpf_spin_lock** is allowed per map element.
2529 * * When the lock is taken, calls (either BPF to BPF or helpers)
2530 * are not allowed.
2531 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2532 * allowed inside a spinlock-ed region.
2533 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2534 * the lock, on all execution paths, before it returns.
2535 * * The BPF program can access **struct bpf_spin_lock** only via
2536 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2537 * helpers. Loading or storing data into the **struct
2538 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2539 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2540 * of the map value must be a struct and have **struct
2541 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2542 * Nested lock inside another struct is not allowed.
2543 * * The **struct bpf_spin_lock** *lock* field in a map value must
2544 * be aligned on a multiple of 4 bytes in that value.
2545 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2546 * the **bpf_spin_lock** field to user space.
2547 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2548 * a BPF program, do not update the **bpf_spin_lock** field.
2549 * * **bpf_spin_lock** cannot be on the stack or inside a
2550 * networking packet (it can only be inside of a map values).
2551 * * **bpf_spin_lock** is available to root only.
2552 * * Tracing programs and socket filter programs cannot use
2553 * **bpf_spin_lock**\ () due to insufficient preemption checks
2554 * (but this may change in the future).
2555 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2556 * Return
2557 * 0
2558 *
2559 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2560 * Description
2561 * Release the *lock* previously locked by a call to
2562 * **bpf_spin_lock**\ (\ *lock*\ ).
2563 * Return
2564 * 0
2565 *
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002566 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2567 * Description
2568 * This helper gets a **struct bpf_sock** pointer such
Quentin Monnet62369db2019-03-14 12:38:39 +00002569 * that all the fields in this **bpf_sock** can be accessed.
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002570 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002571 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002572 * case of failure.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002573 *
2574 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2575 * Description
2576 * This helper gets a **struct bpf_tcp_sock** pointer from a
2577 * **struct bpf_sock** pointer.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002578 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002579 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002580 * case of failure.
brakmof7c917b2019-03-01 12:38:46 -08002581 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002582 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
Quentin Monnet62369db2019-03-14 12:38:39 +00002583 * Description
2584 * Set ECN (Explicit Congestion Notification) field of IP header
2585 * to **CE** (Congestion Encountered) if current value is **ECT**
2586 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2587 * and IPv4.
2588 * Return
2589 * 1 if the **CE** flag is set (either by the current helper call
2590 * or because it was already present), 0 if it is not set.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002591 *
2592 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2593 * Description
Quentin Monnet62369db2019-03-14 12:38:39 +00002594 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2595 * **bpf_sk_release**\ () is unnecessary and not allowed.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002596 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002597 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002598 * case of failure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002599 *
2600 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2601 * Description
2602 * Look for TCP socket matching *tuple*, optionally in a child
2603 * network namespace *netns*. The return value must be checked,
2604 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2605 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002606 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2607 * that it also returns timewait or request sockets. Use
2608 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2609 * full structure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002610 *
2611 * This helper is available only if the kernel was compiled with
2612 * **CONFIG_NET** configuration option.
2613 * Return
2614 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2615 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002616 * result is from *reuse*\ **->socks**\ [] using the hash of the
2617 * tuple.
Lorenz Bauer39904082019-03-22 09:54:02 +08002618 *
2619 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2620 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002621 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2622 * the listening socket in *sk*.
Lorenz Bauer39904082019-03-22 09:54:02 +08002623 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002624 * *iph* points to the start of the IPv4 or IPv6 header, while
2625 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2626 * **sizeof**\ (**struct ip6hdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002627 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002628 * *th* points to the start of the TCP header, while *th_len*
2629 * contains **sizeof**\ (**struct tcphdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002630 *
2631 * Return
Quentin Monnet80867c52019-05-10 15:51:24 +01002632 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2633 * error otherwise.
Andrey Ignatov808649f2019-02-27 13:28:48 -08002634 *
2635 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2636 * Description
2637 * Get name of sysctl in /proc/sys/ and copy it into provided by
2638 * program buffer *buf* of size *buf_len*.
2639 *
2640 * The buffer is always NUL terminated, unless it's zero-sized.
2641 *
2642 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2643 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2644 * only (e.g. "tcp_mem").
2645 * Return
2646 * Number of character copied (not including the trailing NUL).
2647 *
2648 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2649 * truncated name in this case).
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002650 *
2651 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2652 * Description
2653 * Get current value of sysctl as it is presented in /proc/sys
2654 * (incl. newline, etc), and copy it as a string into provided
2655 * by program buffer *buf* of size *buf_len*.
2656 *
2657 * The whole value is copied, no matter what file position user
2658 * space issued e.g. sys_read at.
2659 *
2660 * The buffer is always NUL terminated, unless it's zero-sized.
2661 * Return
2662 * Number of character copied (not including the trailing NUL).
2663 *
2664 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2665 * truncated name in this case).
2666 *
2667 * **-EINVAL** if current value was unavailable, e.g. because
2668 * sysctl is uninitialized and read returns -EIO for it.
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002669 *
2670 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2671 * Description
2672 * Get new value being written by user space to sysctl (before
2673 * the actual write happens) and copy it as a string into
2674 * provided by program buffer *buf* of size *buf_len*.
2675 *
2676 * User space may write new value at file position > 0.
2677 *
2678 * The buffer is always NUL terminated, unless it's zero-sized.
2679 * Return
2680 * Number of character copied (not including the trailing NUL).
2681 *
2682 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2683 * truncated name in this case).
2684 *
2685 * **-EINVAL** if sysctl is being read.
2686 *
2687 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2688 * Description
2689 * Override new value being written by user space to sysctl with
2690 * value provided by program in buffer *buf* of size *buf_len*.
2691 *
2692 * *buf* should contain a string in same form as provided by user
2693 * space on sysctl write.
2694 *
2695 * User space may write new value at file position > 0. To override
2696 * the whole sysctl value file position should be set to zero.
2697 * Return
2698 * 0 on success.
2699 *
2700 * **-E2BIG** if the *buf_len* is too big.
2701 *
2702 * **-EINVAL** if sysctl is being read.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002703 *
2704 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2705 * Description
2706 * Convert the initial part of the string from buffer *buf* of
2707 * size *buf_len* to a long integer according to the given base
2708 * and save the result in *res*.
2709 *
2710 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002711 * (as determined by **isspace**\ (3)) followed by a single
2712 * optional '**-**' sign.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002713 *
2714 * Five least significant bits of *flags* encode base, other bits
2715 * are currently unused.
2716 *
2717 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002718 * similar to user space **strtol**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002719 * Return
2720 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002721 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002722 *
2723 * **-EINVAL** if no valid digits were found or unsupported base
2724 * was provided.
2725 *
2726 * **-ERANGE** if resulting value was out of range.
2727 *
2728 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2729 * Description
2730 * Convert the initial part of the string from buffer *buf* of
2731 * size *buf_len* to an unsigned long integer according to the
2732 * given base and save the result in *res*.
2733 *
2734 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002735 * (as determined by **isspace**\ (3)).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002736 *
2737 * Five least significant bits of *flags* encode base, other bits
2738 * are currently unused.
2739 *
2740 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002741 * similar to user space **strtoul**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002742 * Return
2743 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002744 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002745 *
2746 * **-EINVAL** if no valid digits were found or unsupported base
2747 * was provided.
2748 *
2749 * **-ERANGE** if resulting value was out of range.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002750 *
2751 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2752 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002753 * Get a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002754 *
2755 * Logically, it could be thought of getting the value from
2756 * a *map* with *sk* as the **key**. From this
2757 * perspective, the usage is not much different from
Quentin Monnet80867c52019-05-10 15:51:24 +01002758 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2759 * helper enforces the key must be a full socket and the map must
2760 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002761 *
2762 * Underneath, the value is stored locally at *sk* instead of
Quentin Monnet80867c52019-05-10 15:51:24 +01002763 * the *map*. The *map* is used as the bpf-local-storage
2764 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2765 * searched against all bpf-local-storages residing at *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002766 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002767 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002768 * used such that a new bpf-local-storage will be
2769 * created if one does not exist. *value* can be used
Quentin Monnet80867c52019-05-10 15:51:24 +01002770 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002771 * the initial value of a bpf-local-storage. If *value* is
Quentin Monnet80867c52019-05-10 15:51:24 +01002772 * **NULL**, the new bpf-local-storage will be zero initialized.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002773 * Return
2774 * A bpf-local-storage pointer is returned on success.
2775 *
2776 * **NULL** if not found or there was an error in adding
2777 * a new bpf-local-storage.
2778 *
2779 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2780 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002781 * Delete a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002782 * Return
2783 * 0 on success.
2784 *
2785 * **-ENOENT** if the bpf-local-storage cannot be found.
Yonghong Song8b401f92019-05-23 14:47:45 -07002786 *
2787 * int bpf_send_signal(u32 sig)
2788 * Description
Yonghong Song84829412020-01-14 19:50:02 -08002789 * Send signal *sig* to the process of the current task.
2790 * The signal may be delivered to any of this process's threads.
Yonghong Song8b401f92019-05-23 14:47:45 -07002791 * Return
2792 * 0 on success or successfully queued.
2793 *
2794 * **-EBUSY** if work queue under nmi is full.
2795 *
2796 * **-EINVAL** if *sig* is invalid.
2797 *
2798 * **-EPERM** if no permission to send the *sig*.
2799 *
2800 * **-EAGAIN** if bpf program can try again.
Petar Penkov70d66242019-07-29 09:59:15 -07002801 *
2802 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2803 * Description
2804 * Try to issue a SYN cookie for the packet with corresponding
2805 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2806 *
2807 * *iph* points to the start of the IPv4 or IPv6 header, while
2808 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2809 * **sizeof**\ (**struct ip6hdr**).
2810 *
2811 * *th* points to the start of the TCP header, while *th_len*
2812 * contains the length of the TCP header.
2813 *
2814 * Return
2815 * On success, lower 32 bits hold the generated SYN cookie in
2816 * followed by 16 bits which hold the MSS value for that cookie,
2817 * and the top 16 bits are unused.
2818 *
2819 * On failure, the returned value is one of the following:
2820 *
2821 * **-EINVAL** SYN cookie cannot be issued due to error
2822 *
2823 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2824 *
2825 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2826 *
2827 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
Alexei Starovoitova7658e12019-10-15 20:25:04 -07002828 *
2829 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2830 * Description
2831 * Write raw *data* blob into a special BPF perf event held by
2832 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2833 * event must have the following attributes: **PERF_SAMPLE_RAW**
2834 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2835 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2836 *
2837 * The *flags* are used to indicate the index in *map* for which
2838 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2839 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2840 * to indicate that the index of the current CPU core should be
2841 * used.
2842 *
2843 * The value to write, of *size*, is passed through eBPF stack and
2844 * pointed by *data*.
2845 *
2846 * *ctx* is a pointer to in-kernel struct sk_buff.
2847 *
2848 * This helper is similar to **bpf_perf_event_output**\ () but
2849 * restricted to raw_tracepoint bpf programs.
2850 * Return
2851 * 0 on success, or a negative error in case of failure.
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002852 *
2853 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2854 * Description
2855 * Safely attempt to read *size* bytes from user space address
2856 * *unsafe_ptr* and store the data in *dst*.
2857 * Return
2858 * 0 on success, or a negative error in case of failure.
2859 *
2860 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2861 * Description
2862 * Safely attempt to read *size* bytes from kernel space address
2863 * *unsafe_ptr* and store the data in *dst*.
2864 * Return
2865 * 0 on success, or a negative error in case of failure.
2866 *
2867 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2868 * Description
2869 * Copy a NUL terminated string from an unsafe user address
2870 * *unsafe_ptr* to *dst*. The *size* should include the
2871 * terminating NUL byte. In case the string length is smaller than
2872 * *size*, the target is not padded with further NUL bytes. If the
2873 * string length is larger than *size*, just *size*-1 bytes are
2874 * copied and the last byte is set to NUL.
2875 *
2876 * On success, the length of the copied string is returned. This
2877 * makes this helper useful in tracing programs for reading
2878 * strings, and more importantly to get its length at runtime. See
2879 * the following snippet:
2880 *
2881 * ::
2882 *
2883 * SEC("kprobe/sys_open")
2884 * void bpf_sys_open(struct pt_regs *ctx)
2885 * {
2886 * char buf[PATHLEN]; // PATHLEN is defined to 256
2887 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2888 * ctx->di);
2889 *
2890 * // Consume buf, for example push it to
2891 * // userspace via bpf_perf_event_output(); we
2892 * // can use res (the string length) as event
2893 * // size, after checking its boundaries.
2894 * }
2895 *
2896 * In comparison, using **bpf_probe_read_user()** helper here
2897 * instead to read the string would require to estimate the length
2898 * at compile time, and would often result in copying more memory
2899 * than necessary.
2900 *
2901 * Another useful use case is when parsing individual process
2902 * arguments or individual environment variables navigating
2903 * *current*\ **->mm->arg_start** and *current*\
2904 * **->mm->env_start**: using this helper and the return value,
2905 * one can quickly iterate at the right offset of the memory area.
2906 * Return
2907 * On success, the strictly positive length of the string,
2908 * including the trailing NUL character. On error, a negative
2909 * value.
2910 *
2911 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2912 * Description
2913 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
2914 * to *dst*. Same semantics as with bpf_probe_read_user_str() apply.
2915 * Return
2916 * On success, the strictly positive length of the string, including
2917 * the trailing NUL character. On error, a negative value.
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002918 *
2919 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2920 * Description
2921 * Send out a tcp-ack. *tp* is the in-kernel struct tcp_sock.
2922 * *rcv_nxt* is the ack_seq to be sent out.
2923 * Return
2924 * 0 on success, or a negative error in case of failure.
2925 *
Yonghong Song84829412020-01-14 19:50:02 -08002926 * int bpf_send_signal_thread(u32 sig)
2927 * Description
2928 * Send signal *sig* to the thread corresponding to the current task.
2929 * Return
2930 * 0 on success or successfully queued.
2931 *
2932 * **-EBUSY** if work queue under nmi is full.
2933 *
2934 * **-EINVAL** if *sig* is invalid.
2935 *
2936 * **-EPERM** if no permission to send the *sig*.
2937 *
2938 * **-EAGAIN** if bpf program can try again.
Martin KaFai Lau5576b992020-01-22 15:36:46 -08002939 *
2940 * u64 bpf_jiffies64(void)
2941 * Description
2942 * Obtain the 64bit jiffies
2943 * Return
2944 * The 64 bit jiffies
Daniel Xufff7b642020-02-17 19:04:31 -08002945 *
2946 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2947 * Description
2948 * For an eBPF program attached to a perf event, retrieve the
2949 * branch records (struct perf_branch_entry) associated to *ctx*
2950 * and store it in the buffer pointed by *buf* up to size
2951 * *size* bytes.
2952 * Return
2953 * On success, number of bytes written to *buf*. On error, a
2954 * negative value.
2955 *
2956 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
2957 * instead return the number of bytes required to store all the
2958 * branch entries. If this flag is set, *buf* may be NULL.
2959 *
2960 * **-EINVAL** if arguments invalid or **size** not a multiple
2961 * of sizeof(struct perf_branch_entry).
2962 *
2963 * **-ENOENT** if architecture does not support branch records.
Carlos Neirab4490c52020-03-04 17:41:56 -03002964 *
2965 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2966 * Description
2967 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
2968 * *namespace* will be returned in *nsdata*.
2969 *
2970 * On failure, the returned value is one of the following:
2971 *
2972 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
2973 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
2974 *
2975 * **-ENOENT** if pidns does not exists for the current task.
2976 *
Eelco Chaudrond831ee82020-03-06 08:59:23 +00002977 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2978 * Description
2979 * Write raw *data* blob into a special BPF perf event held by
2980 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2981 * event must have the following attributes: **PERF_SAMPLE_RAW**
2982 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2983 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2984 *
2985 * The *flags* are used to indicate the index in *map* for which
2986 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2987 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2988 * to indicate that the index of the current CPU core should be
2989 * used.
2990 *
2991 * The value to write, of *size*, is passed through eBPF stack and
2992 * pointed by *data*.
2993 *
2994 * *ctx* is a pointer to in-kernel struct xdp_buff.
2995 *
2996 * This helper is similar to **bpf_perf_eventoutput**\ () but
2997 * restricted to raw_tracepoint bpf programs.
2998 * Return
2999 * 0 on success, or a negative error in case of failure.
Daniel Borkmannf3189032020-03-27 16:58:52 +01003000 *
3001 * u64 bpf_get_netns_cookie(void *ctx)
3002 * Description
3003 * Retrieve the cookie (generated by the kernel) of the network
3004 * namespace the input *ctx* is associated with. The network
3005 * namespace cookie remains stable for its lifetime and provides
3006 * a global identifier that can be assumed unique. If *ctx* is
3007 * NULL, then the helper returns the cookie for the initial
3008 * network namespace. The cookie itself is very similar to that
3009 * of bpf_get_socket_cookie() helper, but for network namespaces
3010 * instead of sockets.
3011 * Return
3012 * A 8-byte long opaque number.
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003013 *
3014 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3015 * Description
3016 * Return id of cgroup v2 that is ancestor of the cgroup associated
3017 * with the current task at the *ancestor_level*. The root cgroup
3018 * is at *ancestor_level* zero and each step down the hierarchy
3019 * increments the level. If *ancestor_level* == level of cgroup
3020 * associated with the current task, then return value will be the
3021 * same as that of **bpf_get_current_cgroup_id**\ ().
3022 *
3023 * The helper is useful to implement policies based on cgroups
3024 * that are upper in hierarchy than immediate cgroup associated
3025 * with the current task.
3026 *
3027 * The format of returned id and helper limitations are same as in
3028 * **bpf_get_current_cgroup_id**\ ().
3029 * Return
3030 * The id is returned or 0 in case the id could not be retrieved.
Joe Stringercf7fbe62020-03-29 15:53:38 -07003031 *
3032 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3033 * Description
3034 * Assign the *sk* to the *skb*. When combined with appropriate
3035 * routing configuration to receive the packet towards the socket,
3036 * will cause *skb* to be delivered to the specified socket.
3037 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3038 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3039 * interfere with successful delivery to the socket.
3040 *
3041 * This operation is only valid from TC ingress path.
3042 *
3043 * The *flags* argument must be zero.
3044 * Return
3045 * 0 on success, or a negative errno in case of failure.
3046 *
3047 * * **-EINVAL** Unsupported flags specified.
3048 * * **-ENOENT** Socket is unavailable for assignment.
3049 * * **-ENETUNREACH** Socket is unreachable (wrong netns).
3050 * * **-EOPNOTSUPP** Unsupported operation, for example a
3051 * call from outside of TC ingress.
3052 * * **-ESOCKTNOSUPPORT** Socket type not supported (reuseport).
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003053 *
3054 * u64 bpf_ktime_get_boot_ns(void)
3055 * Description
3056 * Return the time elapsed since system boot, in nanoseconds.
3057 * Does include the time the system was suspended.
3058 * See: clock_gettime(CLOCK_BOOTTIME)
3059 * Return
3060 * Current *ktime*.
Thomas Grafebb676d2016-10-27 11:23:51 +02003061 */
3062#define __BPF_FUNC_MAPPER(FN) \
3063 FN(unspec), \
3064 FN(map_lookup_elem), \
3065 FN(map_update_elem), \
3066 FN(map_delete_elem), \
3067 FN(probe_read), \
3068 FN(ktime_get_ns), \
3069 FN(trace_printk), \
3070 FN(get_prandom_u32), \
3071 FN(get_smp_processor_id), \
3072 FN(skb_store_bytes), \
3073 FN(l3_csum_replace), \
3074 FN(l4_csum_replace), \
3075 FN(tail_call), \
3076 FN(clone_redirect), \
3077 FN(get_current_pid_tgid), \
3078 FN(get_current_uid_gid), \
3079 FN(get_current_comm), \
3080 FN(get_cgroup_classid), \
3081 FN(skb_vlan_push), \
3082 FN(skb_vlan_pop), \
3083 FN(skb_get_tunnel_key), \
3084 FN(skb_set_tunnel_key), \
3085 FN(perf_event_read), \
3086 FN(redirect), \
3087 FN(get_route_realm), \
3088 FN(perf_event_output), \
3089 FN(skb_load_bytes), \
3090 FN(get_stackid), \
3091 FN(csum_diff), \
3092 FN(skb_get_tunnel_opt), \
3093 FN(skb_set_tunnel_opt), \
3094 FN(skb_change_proto), \
3095 FN(skb_change_type), \
3096 FN(skb_under_cgroup), \
3097 FN(get_hash_recalc), \
3098 FN(get_current_task), \
3099 FN(probe_write_user), \
3100 FN(current_task_under_cgroup), \
3101 FN(skb_change_tail), \
3102 FN(skb_pull_data), \
3103 FN(csum_update), \
3104 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003105 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003106 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +00003107 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -07003108 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -07003109 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +02003110 FN(get_socket_uid), \
Lawrence Brakmo8c4b4c72017-06-30 20:02:46 -07003111 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003112 FN(setsockopt), \
John Fastabend97f91a72017-07-17 09:29:18 -07003113 FN(skb_adjust_room), \
John Fastabend174a79f2017-08-15 22:32:47 -07003114 FN(redirect_map), \
3115 FN(sk_redirect_map), \
3116 FN(sock_map_update), \
Yonghong Song908432c2017-10-05 09:19:20 -07003117 FN(xdp_adjust_meta), \
Yonghong Song4bebdc72017-10-05 09:19:22 -07003118 FN(perf_event_read_value), \
Lawrence Brakmocd86d1f2017-10-20 11:05:40 -07003119 FN(perf_prog_read_value), \
Josef Bacik9802d862017-12-11 11:36:48 -05003120 FN(getsockopt), \
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003121 FN(override_return), \
John Fastabend4f738ad2018-03-18 12:57:10 -07003122 FN(sock_ops_cb_flags_set), \
John Fastabend2a100312018-03-18 12:57:15 -07003123 FN(msg_redirect_map), \
John Fastabend91843d52018-03-18 12:57:20 -07003124 FN(msg_apply_bytes), \
John Fastabend015632b2018-03-18 12:57:25 -07003125 FN(msg_cork_bytes), \
Andrey Ignatovd74bad42018-03-30 15:08:05 -07003126 FN(msg_pull_data), \
Nikita V. Shirokovb32cc5b2018-04-17 21:42:13 -07003127 FN(bind), \
Eyal Birger12bed762018-04-24 17:50:29 +03003128 FN(xdp_adjust_tail), \
Yonghong Songc195651e2018-04-28 22:28:08 -07003129 FN(skb_get_xfrm_state), \
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003130 FN(get_stack), \
David Ahern87f5fc72018-05-09 20:34:26 -07003131 FN(skb_load_bytes_relative), \
John Fastabend81110382018-05-14 10:00:17 -07003132 FN(fib_lookup), \
3133 FN(sock_hash_update), \
3134 FN(msg_redirect_hash), \
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003135 FN(sk_redirect_hash), \
3136 FN(lwt_push_encap), \
3137 FN(lwt_seg6_store_bytes), \
3138 FN(lwt_seg6_adjust_srh), \
Sean Youngf4364dc2018-05-27 12:24:09 +01003139 FN(lwt_seg6_action), \
3140 FN(rc_repeat), \
Daniel Borkmanncb20b082018-06-02 23:06:36 +02003141 FN(rc_keydown), \
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07003142 FN(skb_cgroup_id), \
Roman Gushchincd339432018-08-02 14:27:24 -07003143 FN(get_current_cgroup_id), \
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003144 FN(get_local_storage), \
Andrey Ignatov77236282018-08-12 10:49:27 -07003145 FN(sk_select_reuseport), \
Joe Stringer6acc9b42018-10-02 13:35:36 -07003146 FN(skb_ancestor_cgroup_id), \
3147 FN(sk_lookup_tcp), \
3148 FN(sk_lookup_udp), \
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +02003149 FN(sk_release), \
3150 FN(map_push_elem), \
3151 FN(map_pop_elem), \
John Fastabend6fff6072018-10-19 19:56:49 -07003152 FN(map_peek_elem), \
John Fastabend7246d8e2018-11-26 14:16:17 -08003153 FN(msg_push_data), \
Sean Young01d32402018-12-06 13:01:03 +00003154 FN(msg_pop_data), \
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08003155 FN(rc_pointer_rel), \
3156 FN(spin_lock), \
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003157 FN(spin_unlock), \
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003158 FN(sk_fullsock), \
brakmof7c917b2019-03-01 12:38:46 -08003159 FN(tcp_sock), \
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07003160 FN(skb_ecn_set_ce), \
Lorenz Baueredbf8c02019-03-22 09:54:01 +08003161 FN(get_listener_sock), \
Lorenz Bauer39904082019-03-22 09:54:02 +08003162 FN(skc_lookup_tcp), \
Andrey Ignatov808649f2019-02-27 13:28:48 -08003163 FN(tcp_check_syncookie), \
Andrey Ignatov1d11b302019-02-28 19:22:15 -08003164 FN(sysctl_get_name), \
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08003165 FN(sysctl_get_current_value), \
3166 FN(sysctl_get_new_value), \
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07003167 FN(sysctl_set_new_value), \
3168 FN(strtol), \
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003169 FN(strtoul), \
3170 FN(sk_storage_get), \
Yonghong Song8b401f92019-05-23 14:47:45 -07003171 FN(sk_storage_delete), \
Petar Penkov70d66242019-07-29 09:59:15 -07003172 FN(send_signal), \
Alexei Starovoitova7658e12019-10-15 20:25:04 -07003173 FN(tcp_gen_syncookie), \
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01003174 FN(skb_output), \
3175 FN(probe_read_user), \
3176 FN(probe_read_kernel), \
3177 FN(probe_read_user_str), \
Martin KaFai Lau206057f2020-01-08 16:45:51 -08003178 FN(probe_read_kernel_str), \
Yonghong Song84829412020-01-14 19:50:02 -08003179 FN(tcp_send_ack), \
Martin KaFai Lau5576b992020-01-22 15:36:46 -08003180 FN(send_signal_thread), \
Daniel Xufff7b642020-02-17 19:04:31 -08003181 FN(jiffies64), \
Carlos Neirab4490c52020-03-04 17:41:56 -03003182 FN(read_branch_records), \
Eelco Chaudrond831ee82020-03-06 08:59:23 +00003183 FN(get_ns_current_pid_tgid), \
Daniel Borkmannf3189032020-03-27 16:58:52 +01003184 FN(xdp_output), \
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003185 FN(get_netns_cookie), \
Joe Stringercf7fbe62020-03-29 15:53:38 -07003186 FN(get_current_ancestor_cgroup_id), \
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003187 FN(sk_assign), \
3188 FN(ktime_get_boot_ns),
Thomas Grafebb676d2016-10-27 11:23:51 +02003189
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003190/* integer value in 'imm' field of BPF_CALL instruction selects which helper
3191 * function eBPF program intends to call
3192 */
Thomas Grafebb676d2016-10-27 11:23:51 +02003193#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003194enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +02003195 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003196 __BPF_FUNC_MAX_ID,
3197};
Thomas Grafebb676d2016-10-27 11:23:51 +02003198#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003199
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003200/* All flags used by eBPF helper functions, placed here. */
3201
3202/* BPF_FUNC_skb_store_bytes flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003203enum {
3204 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3205 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3206};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003207
3208/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3209 * First 4 bits are for passing the header field size.
3210 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003211enum {
3212 BPF_F_HDR_FIELD_MASK = 0xfULL,
3213};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003214
3215/* BPF_FUNC_l4_csum_replace flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003216enum {
3217 BPF_F_PSEUDO_HDR = (1ULL << 4),
3218 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3219 BPF_F_MARK_ENFORCE = (1ULL << 6),
3220};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003221
3222/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003223enum {
3224 BPF_F_INGRESS = (1ULL << 0),
3225};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003226
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003227/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003228enum {
3229 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3230};
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003231
Yonghong Songc195651e2018-04-28 22:28:08 -07003232/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003233enum {
3234 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3235 BPF_F_USER_STACK = (1ULL << 8),
Yonghong Songc195651e2018-04-28 22:28:08 -07003236/* flags used by BPF_FUNC_get_stackid only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003237 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3238 BPF_F_REUSE_STACKID = (1ULL << 10),
Yonghong Songc195651e2018-04-28 22:28:08 -07003239/* flags used by BPF_FUNC_get_stack only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003240 BPF_F_USER_BUILD_ID = (1ULL << 11),
3241};
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08003242
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003243/* BPF_FUNC_skb_set_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003244enum {
3245 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3246 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3247 BPF_F_SEQ_NUMBER = (1ULL << 3),
3248};
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003249
Yonghong Song908432c2017-10-05 09:19:20 -07003250/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3251 * BPF_FUNC_perf_event_read_value flags.
3252 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003253enum {
3254 BPF_F_INDEX_MASK = 0xffffffffULL,
3255 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
Daniel Borkmann555c8a82016-07-14 18:08:05 +02003256/* BPF_FUNC_perf_event_output for sk_buff input context. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003257 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3258};
Daniel Borkmann1e337592016-04-18 21:01:23 +02003259
Joe Stringerf71c6142018-11-30 15:32:20 -08003260/* Current network namespace */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003261enum {
3262 BPF_F_CURRENT_NETNS = (-1L),
3263};
Joe Stringerf71c6142018-11-30 15:32:20 -08003264
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003265/* BPF_FUNC_skb_adjust_room flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003266enum {
3267 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3268 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3269 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3270 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3271 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3272};
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003273
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003274enum {
3275 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3276 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3277};
Alan Maguire58dfc902019-04-09 15:06:41 +01003278
Alan Maguirebfb35c22019-04-12 12:27:34 +01003279#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
Alan Maguire58dfc902019-04-09 15:06:41 +01003280 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3281 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Willem de Bruijn868d5232019-03-22 14:32:56 -04003282
Andrey Ignatov808649f2019-02-27 13:28:48 -08003283/* BPF_FUNC_sysctl_get_name flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003284enum {
3285 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3286};
Andrey Ignatov808649f2019-02-27 13:28:48 -08003287
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003288/* BPF_FUNC_sk_storage_get flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003289enum {
3290 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3291};
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003292
Daniel Xufff7b642020-02-17 19:04:31 -08003293/* BPF_FUNC_read_branch_records flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003294enum {
3295 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3296};
Daniel Xufff7b642020-02-17 19:04:31 -08003297
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003298/* Mode for BPF_FUNC_skb_adjust_room helper. */
3299enum bpf_adj_room_mode {
3300 BPF_ADJ_ROOM_NET,
Willem de Bruijn14aa3192019-03-22 14:32:54 -04003301 BPF_ADJ_ROOM_MAC,
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003302};
3303
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003304/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3305enum bpf_hdr_start_off {
3306 BPF_HDR_START_MAC,
3307 BPF_HDR_START_NET,
3308};
3309
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003310/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3311enum bpf_lwt_encap_mode {
3312 BPF_LWT_ENCAP_SEG6,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003313 BPF_LWT_ENCAP_SEG6_INLINE,
3314 BPF_LWT_ENCAP_IP,
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003315};
3316
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003317#define __bpf_md_ptr(type, name) \
3318union { \
3319 type name; \
3320 __u64 :64; \
3321} __attribute__((aligned(8)))
3322
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003323/* user accessible mirror of in-kernel sk_buff.
3324 * new fields can only be added to the end of this structure
3325 */
3326struct __sk_buff {
3327 __u32 len;
3328 __u32 pkt_type;
3329 __u32 mark;
3330 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -07003331 __u32 protocol;
3332 __u32 vlan_present;
3333 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +01003334 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +02003335 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -07003336 __u32 ingress_ifindex;
3337 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -07003338 __u32 tc_index;
3339 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +02003340 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -07003341 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -07003342 __u32 data;
3343 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +02003344 __u32 napi_id;
John Fastabend8a31db52017-08-15 22:33:09 -07003345
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003346 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
John Fastabend8a31db52017-08-15 22:33:09 -07003347 __u32 family;
3348 __u32 remote_ip4; /* Stored in network byte order */
3349 __u32 local_ip4; /* Stored in network byte order */
3350 __u32 remote_ip6[4]; /* Stored in network byte order */
3351 __u32 local_ip6[4]; /* Stored in network byte order */
3352 __u32 remote_port; /* Stored in network byte order */
3353 __u32 local_port; /* stored in host byte order */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003354 /* ... here. */
3355
3356 __u32 data_meta;
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003357 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
Vlad Dumitrescuf11216b2018-11-22 14:39:16 -05003358 __u64 tstamp;
Petar Penkove3da08d2018-12-02 20:18:19 -05003359 __u32 wire_len;
Eric Dumazetd9ff2862019-01-23 09:22:27 -08003360 __u32 gso_segs;
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003361 __bpf_md_ptr(struct bpf_sock *, sk);
Willem de Bruijncf620892020-03-03 15:05:01 -05003362 __u32 gso_size;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003363};
3364
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003365struct bpf_tunnel_key {
3366 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003367 union {
3368 __u32 remote_ipv4;
3369 __u32 remote_ipv6[4];
3370 };
3371 __u8 tunnel_tos;
3372 __u8 tunnel_ttl;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003373 __u16 tunnel_ext; /* Padding, future use. */
Daniel Borkmann4018ab12016-03-09 03:00:05 +01003374 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003375};
3376
Eyal Birger12bed762018-04-24 17:50:29 +03003377/* user accessible mirror of in-kernel xfrm_state.
3378 * new fields can only be added to the end of this structure
3379 */
3380struct bpf_xfrm_state {
3381 __u32 reqid;
3382 __u32 spi; /* Stored in network byte order */
3383 __u16 family;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003384 __u16 ext; /* Padding, future use. */
Eyal Birger12bed762018-04-24 17:50:29 +03003385 union {
3386 __u32 remote_ipv4; /* Stored in network byte order */
3387 __u32 remote_ipv6[4]; /* Stored in network byte order */
3388 };
3389};
3390
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003391/* Generic BPF return codes which all BPF program types may support.
3392 * The values are binary compatible with their TC_ACT_* counter-part to
3393 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3394 * programs.
3395 *
3396 * XDP is handled seprately, see XDP_*.
3397 */
3398enum bpf_ret_code {
3399 BPF_OK = 0,
3400 /* 1 reserved */
3401 BPF_DROP = 2,
3402 /* 3-6 reserved */
3403 BPF_REDIRECT = 7,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003404 /* >127 are reserved for prog type specific return codes.
3405 *
3406 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3407 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3408 * changed and should be routed based on its new L3 header.
3409 * (This is an L3 redirect, as opposed to L2 redirect
3410 * represented by BPF_REDIRECT above).
3411 */
3412 BPF_LWT_REROUTE = 128,
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003413};
3414
David Ahern610236582016-12-01 08:48:04 -08003415struct bpf_sock {
3416 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -08003417 __u32 family;
3418 __u32 type;
3419 __u32 protocol;
David Ahern482dca92017-08-31 15:05:44 -07003420 __u32 mark;
3421 __u32 priority;
Martin KaFai Lauaa65d692019-02-09 23:22:21 -08003422 /* IP address also allows 1 and 2 bytes access */
3423 __u32 src_ip4;
3424 __u32 src_ip6[4];
3425 __u32 src_port; /* host byte order */
3426 __u32 dst_port; /* network byte order */
3427 __u32 dst_ip4;
3428 __u32 dst_ip6[4];
3429 __u32 state;
David Ahern610236582016-12-01 08:48:04 -08003430};
3431
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003432struct bpf_tcp_sock {
3433 __u32 snd_cwnd; /* Sending congestion window */
3434 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3435 __u32 rtt_min;
3436 __u32 snd_ssthresh; /* Slow start size threshold */
3437 __u32 rcv_nxt; /* What we want to receive next */
3438 __u32 snd_nxt; /* Next sequence we send */
3439 __u32 snd_una; /* First byte we want an ack for */
3440 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3441 __u32 ecn_flags; /* ECN status bits. */
3442 __u32 rate_delivered; /* saved rate sample: packets delivered */
3443 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3444 __u32 packets_out; /* Packets which are "in flight" */
3445 __u32 retrans_out; /* Retransmitted packets out */
3446 __u32 total_retrans; /* Total retransmits for entire connection */
3447 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3448 * total number of segments in.
3449 */
3450 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3451 * total number of data segments in.
3452 */
3453 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3454 * The total number of segments sent.
3455 */
3456 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3457 * total number of data segments sent.
3458 */
3459 __u32 lost_out; /* Lost packets */
3460 __u32 sacked_out; /* SACK'd packets */
3461 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3462 * sum(delta(rcv_nxt)), or how many bytes
3463 * were acked.
3464 */
3465 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3466 * sum(delta(snd_una)), or how many bytes
3467 * were acked.
3468 */
Stanislav Fomichev03577462019-07-02 09:13:58 -07003469 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3470 * total number of DSACK blocks received
3471 */
3472 __u32 delivered; /* Total data packets delivered incl. rexmits */
3473 __u32 delivered_ce; /* Like the above but only ECE marked packets */
Stanislav Fomichevc2cb5e82019-07-02 09:13:59 -07003474 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003475};
3476
Joe Stringer6acc9b42018-10-02 13:35:36 -07003477struct bpf_sock_tuple {
3478 union {
3479 struct {
3480 __be32 saddr;
3481 __be32 daddr;
3482 __be16 sport;
3483 __be16 dport;
3484 } ipv4;
3485 struct {
3486 __be32 saddr[4];
3487 __be32 daddr[4];
3488 __be16 sport;
3489 __be16 dport;
3490 } ipv6;
3491 };
3492};
3493
Jonathan Lemonfada7fd2019-06-06 13:59:40 -07003494struct bpf_xdp_sock {
3495 __u32 queue_id;
3496};
3497
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003498#define XDP_PACKET_HEADROOM 256
3499
Brenden Blanco6a773a12016-07-19 12:16:47 -07003500/* User return codes for XDP prog type.
3501 * A valid XDP program must return one of these defined values. All other
Daniel Borkmann9beb8be2017-09-09 01:40:35 +02003502 * return codes are reserved for future use. Unknown return codes will
3503 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
Brenden Blanco6a773a12016-07-19 12:16:47 -07003504 */
3505enum xdp_action {
3506 XDP_ABORTED = 0,
3507 XDP_DROP,
3508 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -07003509 XDP_TX,
John Fastabend814abfa2017-07-17 09:27:07 -07003510 XDP_REDIRECT,
Brenden Blanco6a773a12016-07-19 12:16:47 -07003511};
3512
3513/* user accessible metadata for XDP packet hook
3514 * new fields must be added to the end of this structure
3515 */
3516struct xdp_md {
3517 __u32 data;
3518 __u32 data_end;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003519 __u32 data_meta;
Jesper Dangaard Brouerdaaf24c2018-01-11 17:39:09 +01003520 /* Below access go through struct xdp_rxq_info */
Jesper Dangaard Brouer02dd3292018-01-03 11:26:14 +01003521 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3522 __u32 rx_queue_index; /* rxq->queue_index */
Brenden Blanco6a773a12016-07-19 12:16:47 -07003523};
3524
John Fastabend174a79f2017-08-15 22:32:47 -07003525enum sk_action {
John Fastabendbfa640752017-10-27 09:45:53 -07003526 SK_DROP = 0,
3527 SK_PASS,
John Fastabend174a79f2017-08-15 22:32:47 -07003528};
3529
John Fastabend4f738ad2018-03-18 12:57:10 -07003530/* user accessible metadata for SK_MSG packet hook, new fields must
3531 * be added to the end of this structure
3532 */
3533struct sk_msg_md {
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003534 __bpf_md_ptr(void *, data);
3535 __bpf_md_ptr(void *, data_end);
John Fastabend303def32018-05-17 14:16:58 -07003536
3537 __u32 family;
3538 __u32 remote_ip4; /* Stored in network byte order */
3539 __u32 local_ip4; /* Stored in network byte order */
3540 __u32 remote_ip6[4]; /* Stored in network byte order */
3541 __u32 local_ip6[4]; /* Stored in network byte order */
3542 __u32 remote_port; /* Stored in network byte order */
3543 __u32 local_port; /* stored in host byte order */
John Fastabend3bdbd022018-12-16 15:47:04 -08003544 __u32 size; /* Total size of sk_msg */
John Fastabend4f738ad2018-03-18 12:57:10 -07003545};
3546
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003547struct sk_reuseport_md {
3548 /*
3549 * Start of directly accessible data. It begins from
3550 * the tcp/udp header.
3551 */
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003552 __bpf_md_ptr(void *, data);
3553 /* End of directly accessible data */
3554 __bpf_md_ptr(void *, data_end);
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003555 /*
3556 * Total length of packet (starting from the tcp/udp header).
3557 * Note that the directly accessible bytes (data_end - data)
3558 * could be less than this "len". Those bytes could be
3559 * indirectly read by a helper "bpf_skb_load_bytes()".
3560 */
3561 __u32 len;
3562 /*
3563 * Eth protocol in the mac header (network byte order). e.g.
3564 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3565 */
3566 __u32 eth_protocol;
3567 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3568 __u32 bind_inany; /* Is sock bound to an INANY address? */
3569 __u32 hash; /* A hash of the packet 4 tuples */
3570};
3571
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003572#define BPF_TAG_SIZE 8
3573
3574struct bpf_prog_info {
3575 __u32 type;
3576 __u32 id;
3577 __u8 tag[BPF_TAG_SIZE];
3578 __u32 jited_prog_len;
3579 __u32 xlated_prog_len;
3580 __aligned_u64 jited_prog_insns;
3581 __aligned_u64 xlated_prog_insns;
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -07003582 __u64 load_time; /* ns since boottime */
3583 __u32 created_by_uid;
3584 __u32 nr_map_ids;
3585 __aligned_u64 map_ids;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003586 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski675fc272017-12-27 18:39:09 -08003587 __u32 ifindex;
Jiri Olsab85fab02018-04-25 19:41:06 +02003588 __u32 gpl_compatible:1;
Baruch Siach04723012019-06-28 07:08:45 +03003589 __u32 :31; /* alignment pad */
Jakub Kicinski675fc272017-12-27 18:39:09 -08003590 __u64 netns_dev;
3591 __u64 netns_ino;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303592 __u32 nr_jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303593 __u32 nr_jited_func_lens;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303594 __aligned_u64 jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303595 __aligned_u64 jited_func_lens;
Yonghong Song838e9692018-11-19 15:29:11 -08003596 __u32 btf_id;
3597 __u32 func_info_rec_size;
3598 __aligned_u64 func_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003599 __u32 nr_func_info;
3600 __u32 nr_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003601 __aligned_u64 line_info;
3602 __aligned_u64 jited_line_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003603 __u32 nr_jited_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003604 __u32 line_info_rec_size;
3605 __u32 jited_line_info_rec_size;
Song Liuc872bdb2018-12-12 09:37:46 -08003606 __u32 nr_prog_tags;
3607 __aligned_u64 prog_tags;
Alexei Starovoitov5f8f8b92019-02-25 14:28:40 -08003608 __u64 run_time_ns;
3609 __u64 run_cnt;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003610} __attribute__((aligned(8)));
3611
3612struct bpf_map_info {
3613 __u32 type;
3614 __u32 id;
3615 __u32 key_size;
3616 __u32 value_size;
3617 __u32 max_entries;
3618 __u32 map_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003619 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski52775b32018-01-17 19:13:28 -08003620 __u32 ifindex;
Martin KaFai Lau85d33df2020-01-08 16:35:05 -08003621 __u32 btf_vmlinux_value_type_id;
Jakub Kicinski52775b32018-01-17 19:13:28 -08003622 __u64 netns_dev;
3623 __u64 netns_ino;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -07003624 __u32 btf_id;
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -07003625 __u32 btf_key_type_id;
3626 __u32 btf_value_type_id;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003627} __attribute__((aligned(8)));
3628
Martin KaFai Lau62dab842018-05-04 14:49:52 -07003629struct bpf_btf_info {
3630 __aligned_u64 btf;
3631 __u32 btf_size;
3632 __u32 id;
3633} __attribute__((aligned(8)));
3634
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -07003635struct bpf_link_info {
3636 __u32 type;
3637 __u32 id;
3638 __u32 prog_id;
3639 union {
3640 struct {
3641 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3642 __u32 tp_name_len; /* in/out: tp_name buffer len */
3643 } raw_tracepoint;
3644 struct {
3645 __u32 attach_type;
3646 } tracing;
3647 struct {
3648 __u64 cgroup_id;
3649 __u32 attach_type;
3650 } cgroup;
3651 };
3652} __attribute__((aligned(8)));
3653
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003654/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3655 * by user and intended to be used by socket (e.g. to bind to, depends on
3656 * attach attach type).
3657 */
3658struct bpf_sock_addr {
3659 __u32 user_family; /* Allows 4-byte read, but no write. */
3660 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3661 * Stored in network byte order.
3662 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07003663 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003664 * Stored in network byte order.
3665 */
3666 __u32 user_port; /* Allows 4-byte read and write.
3667 * Stored in network byte order
3668 */
3669 __u32 family; /* Allows 4-byte read, but no write */
3670 __u32 type; /* Allows 4-byte read, but no write */
3671 __u32 protocol; /* Allows 4-byte read, but no write */
Stanislav Fomichev600c70b2019-07-01 10:38:39 -07003672 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07003673 * Stored in network byte order.
3674 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07003675 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07003676 * Stored in network byte order.
3677 */
Stanislav Fomichevfb85c4a2019-06-12 10:30:37 -07003678 __bpf_md_ptr(struct bpf_sock *, sk);
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003679};
3680
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003681/* User bpf_sock_ops struct to access socket values and specify request ops
3682 * and their replies.
3683 * Some of this fields are in network (bigendian) byte order and may need
3684 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3685 * New fields can only be added at the end of this structure
3686 */
3687struct bpf_sock_ops {
3688 __u32 op;
3689 union {
Lawrence Brakmode525be2018-01-25 16:14:09 -08003690 __u32 args[4]; /* Optionally passed to bpf program */
3691 __u32 reply; /* Returned by bpf program */
3692 __u32 replylong[4]; /* Optionally returned by bpf prog */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003693 };
3694 __u32 family;
3695 __u32 remote_ip4; /* Stored in network byte order */
3696 __u32 local_ip4; /* Stored in network byte order */
3697 __u32 remote_ip6[4]; /* Stored in network byte order */
3698 __u32 local_ip6[4]; /* Stored in network byte order */
3699 __u32 remote_port; /* Stored in network byte order */
3700 __u32 local_port; /* stored in host byte order */
Lawrence Brakmof19397a2017-12-01 10:15:04 -08003701 __u32 is_fullsock; /* Some TCP fields are only valid if
3702 * there is a full socket. If not, the
3703 * fields read as zero.
3704 */
3705 __u32 snd_cwnd;
3706 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003707 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
Lawrence Brakmo44f0e432018-01-25 16:14:12 -08003708 __u32 state;
3709 __u32 rtt_min;
3710 __u32 snd_ssthresh;
3711 __u32 rcv_nxt;
3712 __u32 snd_nxt;
3713 __u32 snd_una;
3714 __u32 mss_cache;
3715 __u32 ecn_flags;
3716 __u32 rate_delivered;
3717 __u32 rate_interval_us;
3718 __u32 packets_out;
3719 __u32 retrans_out;
3720 __u32 total_retrans;
3721 __u32 segs_in;
3722 __u32 data_segs_in;
3723 __u32 segs_out;
3724 __u32 data_segs_out;
3725 __u32 lost_out;
3726 __u32 sacked_out;
3727 __u32 sk_txhash;
3728 __u64 bytes_received;
3729 __u64 bytes_acked;
Stanislav Fomichev1314ef52019-06-12 10:30:38 -07003730 __bpf_md_ptr(struct bpf_sock *, sk);
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003731};
3732
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003733/* Definitions for bpf_sock_ops_cb_flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003734enum {
3735 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3736 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3737 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3738 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3739/* Mask of all currently supported cb flags */
3740 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3741};
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003742
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003743/* List of known BPF sock_ops operators.
3744 * New entries can only be added at the end
3745 */
3746enum {
3747 BPF_SOCK_OPS_VOID,
Lawrence Brakmo8550f322017-06-30 20:02:42 -07003748 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3749 * -1 if default value should be used
3750 */
Lawrence Brakmo13d3b1e2017-06-30 20:02:44 -07003751 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3752 * window (in packets) or -1 if default
3753 * value should be used
3754 */
Lawrence Brakmo9872a4b2017-06-30 20:02:47 -07003755 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3756 * active connection is initialized
3757 */
3758 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3759 * active connection is
3760 * established
3761 */
3762 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3763 * passive connection is
3764 * established
3765 */
Lawrence Brakmo91b5b212017-06-30 20:02:49 -07003766 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3767 * needs ECN
3768 */
Lawrence Brakmoe6546ef2017-10-20 11:05:39 -07003769 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3770 * based on the path and may be
3771 * dependent on the congestion control
3772 * algorithm. In general it indicates
3773 * a congestion threshold. RTTs above
3774 * this indicate congestion
3775 */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08003776 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3777 * Arg1: value of icsk_retransmits
3778 * Arg2: value of icsk_rto
3779 * Arg3: whether RTO has expired
3780 */
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08003781 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3782 * Arg1: sequence number of 1st byte
3783 * Arg2: # segments
3784 * Arg3: return value of
3785 * tcp_transmit_skb (0 => success)
3786 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08003787 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3788 * Arg1: old_state
3789 * Arg2: new_state
3790 */
Andrey Ignatovf333ee02018-07-11 17:33:32 -07003791 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3792 * socket transition to LISTEN state.
3793 */
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07003794 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3795 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08003796};
3797
3798/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3799 * changes between the TCP and BPF versions. Ideally this should never happen.
3800 * If it does, we need to add code to convert them before calling
3801 * the BPF sock_ops function.
3802 */
3803enum {
3804 BPF_TCP_ESTABLISHED = 1,
3805 BPF_TCP_SYN_SENT,
3806 BPF_TCP_SYN_RECV,
3807 BPF_TCP_FIN_WAIT1,
3808 BPF_TCP_FIN_WAIT2,
3809 BPF_TCP_TIME_WAIT,
3810 BPF_TCP_CLOSE,
3811 BPF_TCP_CLOSE_WAIT,
3812 BPF_TCP_LAST_ACK,
3813 BPF_TCP_LISTEN,
3814 BPF_TCP_CLOSING, /* Now a valid state */
3815 BPF_TCP_NEW_SYN_RECV,
3816
3817 BPF_TCP_MAX_STATES /* Leave at the end! */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003818};
3819
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003820enum {
3821 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
3822 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
3823};
Lawrence Brakmofc747812017-06-30 20:02:51 -07003824
Yonghong Song908432c2017-10-05 09:19:20 -07003825struct bpf_perf_event_value {
3826 __u64 counter;
3827 __u64 enabled;
3828 __u64 running;
3829};
3830
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003831enum {
3832 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
3833 BPF_DEVCG_ACC_READ = (1ULL << 1),
3834 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
3835};
Roman Gushchinebc614f2017-11-05 08:15:32 -05003836
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003837enum {
3838 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
3839 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
3840};
Roman Gushchinebc614f2017-11-05 08:15:32 -05003841
3842struct bpf_cgroup_dev_ctx {
Yonghong Song06ef0cc2017-12-18 10:13:44 -08003843 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3844 __u32 access_type;
Roman Gushchinebc614f2017-11-05 08:15:32 -05003845 __u32 major;
3846 __u32 minor;
3847};
3848
Alexei Starovoitovc4f66992018-03-28 12:05:37 -07003849struct bpf_raw_tracepoint_args {
3850 __u64 args[0];
3851};
3852
David Ahern87f5fc72018-05-09 20:34:26 -07003853/* DIRECT: Skip the FIB rules and go to FIB table associated with device
3854 * OUTPUT: Do lookup from egress perspective; default is ingress
3855 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003856enum {
3857 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
3858 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
3859};
David Ahern87f5fc72018-05-09 20:34:26 -07003860
David Ahern4c795792018-06-26 16:21:18 -07003861enum {
3862 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3863 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3864 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3865 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3866 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3867 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3868 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3869 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3870 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3871};
3872
David Ahern87f5fc72018-05-09 20:34:26 -07003873struct bpf_fib_lookup {
David Ahernfa898d72018-05-29 10:58:07 -07003874 /* input: network family for lookup (AF_INET, AF_INET6)
3875 * output: network family of egress nexthop
3876 */
3877 __u8 family;
David Ahern87f5fc72018-05-09 20:34:26 -07003878
3879 /* set if lookup is to consider L4 data - e.g., FIB rules */
3880 __u8 l4_protocol;
3881 __be16 sport;
3882 __be16 dport;
3883
3884 /* total length of packet from network header - used for MTU check */
3885 __u16 tot_len;
David Ahern4c795792018-06-26 16:21:18 -07003886
3887 /* input: L3 device index for lookup
3888 * output: device index from FIB lookup
3889 */
3890 __u32 ifindex;
David Ahern87f5fc72018-05-09 20:34:26 -07003891
3892 union {
3893 /* inputs to lookup */
3894 __u8 tos; /* AF_INET */
David Ahernbd3a08a2018-06-03 08:15:19 -07003895 __be32 flowinfo; /* AF_INET6, flow_label + priority */
David Ahern87f5fc72018-05-09 20:34:26 -07003896
David Ahernfa898d72018-05-29 10:58:07 -07003897 /* output: metric of fib result (IPv4/IPv6 only) */
3898 __u32 rt_metric;
David Ahern87f5fc72018-05-09 20:34:26 -07003899 };
3900
3901 union {
David Ahern87f5fc72018-05-09 20:34:26 -07003902 __be32 ipv4_src;
3903 __u32 ipv6_src[4]; /* in6_addr; network order */
3904 };
3905
David Ahernfa898d72018-05-29 10:58:07 -07003906 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3907 * network header. output: bpf_fib_lookup sets to gateway address
3908 * if FIB lookup returns gateway route
David Ahern87f5fc72018-05-09 20:34:26 -07003909 */
3910 union {
David Ahern87f5fc72018-05-09 20:34:26 -07003911 __be32 ipv4_dst;
3912 __u32 ipv6_dst[4]; /* in6_addr; network order */
3913 };
3914
3915 /* output */
3916 __be16 h_vlan_proto;
3917 __be16 h_vlan_TCI;
3918 __u8 smac[6]; /* ETH_ALEN */
3919 __u8 dmac[6]; /* ETH_ALEN */
3920};
3921
Yonghong Song41bdc4b2018-05-24 11:21:09 -07003922enum bpf_task_fd_type {
3923 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3924 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3925 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3926 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3927 BPF_FD_TYPE_UPROBE, /* filename + offset */
3928 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3929};
3930
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003931enum {
3932 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
3933 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
3934 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
3935};
Stanislav Fomichev086f9562019-07-25 15:52:25 -07003936
Petar Penkovd58e4682018-09-14 07:46:18 -07003937struct bpf_flow_keys {
3938 __u16 nhoff;
3939 __u16 thoff;
3940 __u16 addr_proto; /* ETH_P_* of valid addrs */
3941 __u8 is_frag;
3942 __u8 is_first_frag;
3943 __u8 is_encap;
3944 __u8 ip_proto;
3945 __be16 n_proto;
3946 __be16 sport;
3947 __be16 dport;
3948 union {
3949 struct {
3950 __be32 ipv4_src;
3951 __be32 ipv4_dst;
3952 };
3953 struct {
3954 __u32 ipv6_src[4]; /* in6_addr; network order */
3955 __u32 ipv6_dst[4]; /* in6_addr; network order */
3956 };
3957 };
Stanislav Fomichev086f9562019-07-25 15:52:25 -07003958 __u32 flags;
Stanislav Fomichev71c99e32019-07-25 15:52:30 -07003959 __be32 flow_label;
Petar Penkovd58e4682018-09-14 07:46:18 -07003960};
3961
Yonghong Song838e9692018-11-19 15:29:11 -08003962struct bpf_func_info {
Martin KaFai Laud30d42e2018-12-05 17:35:44 -08003963 __u32 insn_off;
Yonghong Song838e9692018-11-19 15:29:11 -08003964 __u32 type_id;
3965};
3966
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003967#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3968#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3969
3970struct bpf_line_info {
3971 __u32 insn_off;
3972 __u32 file_name_off;
3973 __u32 line_off;
3974 __u32 line_col;
3975};
3976
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08003977struct bpf_spin_lock {
3978 __u32 val;
3979};
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08003980
3981struct bpf_sysctl {
3982 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3983 * Allows 1,2,4-byte read, but no write.
3984 */
Andrey Ignatove1550bf2019-03-07 18:50:52 -08003985 __u32 file_pos; /* Sysctl file position to read from, write to.
3986 * Allows 1,2,4-byte read an 4-byte write.
3987 */
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08003988};
3989
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07003990struct bpf_sockopt {
3991 __bpf_md_ptr(struct bpf_sock *, sk);
3992 __bpf_md_ptr(void *, optval);
3993 __bpf_md_ptr(void *, optval_end);
3994
3995 __s32 level;
3996 __s32 optname;
3997 __s32 optlen;
3998 __s32 retval;
3999};
4000
Carlos Neirab4490c52020-03-04 17:41:56 -03004001struct bpf_pidns_info {
4002 __u32 pid;
4003 __u32 tgid;
4004};
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07004005#endif /* _UAPI__LINUX_BPF_H__ */