blob: 1096235273585b5ff4433289c64db54a7d993cae [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. Silva1e6e9d02020-04-24 10:50:00 -050076 __u8 data[0]; /* 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,
Yonghong Songac51d992020-05-09 10:59:05 -0700119 BPF_ITER_CREATE,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700120};
121
122enum bpf_map_type {
123 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -0800124 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -0800125 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61ab2015-05-19 16:59:03 -0700126 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +0000127 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -0800128 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -0800129 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800130 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -0700131 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800132 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800133 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +0100134 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700135 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700136 BPF_MAP_TYPE_HASH_OF_MAPS,
John Fastabend546ac1f2017-07-17 09:28:56 -0700137 BPF_MAP_TYPE_DEVMAP,
John Fastabend174a79f2017-08-15 22:32:47 -0700138 BPF_MAP_TYPE_SOCKMAP,
Jesper Dangaard Brouer6710e112017-10-16 12:19:28 +0200139 BPF_MAP_TYPE_CPUMAP,
Björn Töpelfbfc504a2018-05-02 13:01:28 +0200140 BPF_MAP_TYPE_XSKMAP,
John Fastabend81110382018-05-14 10:00:17 -0700141 BPF_MAP_TYPE_SOCKHASH,
Roman Gushchinde9cbba2018-08-02 14:27:18 -0700142 BPF_MAP_TYPE_CGROUP_STORAGE,
Martin KaFai Lau5dc4c4b2018-08-08 01:01:24 -0700143 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
Roman Gushchinb741f162018-09-28 14:45:43 +0000144 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +0200145 BPF_MAP_TYPE_QUEUE,
146 BPF_MAP_TYPE_STACK,
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -0700147 BPF_MAP_TYPE_SK_STORAGE,
Toke Høiland-Jørgensen6f9d4512019-07-26 18:06:55 +0200148 BPF_MAP_TYPE_DEVMAP_HASH,
Martin KaFai Lau85d33df2020-01-08 16:35:05 -0800149 BPF_MAP_TYPE_STRUCT_OPS,
Andrii Nakryiko457f4432020-05-29 00:54:20 -0700150 BPF_MAP_TYPE_RINGBUF,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700151};
152
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100153/* Note that tracing related programs such as
154 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
155 * are not subject to a stable API since kernel internal data
156 * structures can change from release to release and may
157 * therefore break existing tracing BPF programs. Tracing BPF
158 * programs correspond to /a/ specific kernel which is to be
159 * analyzed, and not /a/ specific kernel /and/ all future ones.
160 */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700161enum bpf_prog_type {
162 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800163 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700164 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100165 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee8c2015-03-20 15:11:11 +0100166 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700167 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700168 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700169 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100170 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800171 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100172 BPF_PROG_TYPE_LWT_IN,
173 BPF_PROG_TYPE_LWT_OUT,
174 BPF_PROG_TYPE_LWT_XMIT,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700175 BPF_PROG_TYPE_SOCK_OPS,
John Fastabendb005fd12017-08-15 22:31:58 -0700176 BPF_PROG_TYPE_SK_SKB,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500177 BPF_PROG_TYPE_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700178 BPF_PROG_TYPE_SK_MSG,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700179 BPF_PROG_TYPE_RAW_TRACEPOINT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700180 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
Mathieu Xhonneux004d4b22018-05-20 14:58:16 +0100181 BPF_PROG_TYPE_LWT_SEG6LOCAL,
Sean Youngf4364dc2018-05-27 12:24:09 +0100182 BPF_PROG_TYPE_LIRC_MODE2,
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -0700183 BPF_PROG_TYPE_SK_REUSEPORT,
Petar Penkovd58e4682018-09-14 07:46:18 -0700184 BPF_PROG_TYPE_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800185 BPF_PROG_TYPE_CGROUP_SYSCTL,
Matt Mullins9df1c282019-04-26 11:49:47 -0700186 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
Stanislav Fomichev0d01da62019-06-27 13:38:47 -0700187 BPF_PROG_TYPE_CGROUP_SOCKOPT,
Alexei Starovoitovf1b95092019-10-30 15:32:11 -0700188 BPF_PROG_TYPE_TRACING,
Martin KaFai Lau27ae79972020-01-08 16:35:03 -0800189 BPF_PROG_TYPE_STRUCT_OPS,
Alexei Starovoitovbe8704f2020-01-20 16:53:46 -0800190 BPF_PROG_TYPE_EXT,
KP Singhfc611f42020-03-29 01:43:49 +0100191 BPF_PROG_TYPE_LSM,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700192};
193
Daniel Mack0e336612016-11-23 16:52:25 +0100194enum bpf_attach_type {
195 BPF_CGROUP_INET_INGRESS,
196 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800197 BPF_CGROUP_INET_SOCK_CREATE,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700198 BPF_CGROUP_SOCK_OPS,
John Fastabend464bc0f2017-08-28 07:10:04 -0700199 BPF_SK_SKB_STREAM_PARSER,
200 BPF_SK_SKB_STREAM_VERDICT,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500201 BPF_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700202 BPF_SK_MSG_VERDICT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700203 BPF_CGROUP_INET4_BIND,
204 BPF_CGROUP_INET6_BIND,
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700205 BPF_CGROUP_INET4_CONNECT,
206 BPF_CGROUP_INET6_CONNECT,
Andrey Ignatovaac3fc32018-03-30 15:08:07 -0700207 BPF_CGROUP_INET4_POST_BIND,
208 BPF_CGROUP_INET6_POST_BIND,
Andrey Ignatov1cedee12018-05-25 08:55:23 -0700209 BPF_CGROUP_UDP4_SENDMSG,
210 BPF_CGROUP_UDP6_SENDMSG,
Sean Youngf4364dc2018-05-27 12:24:09 +0100211 BPF_LIRC_MODE2,
Petar Penkovd58e4682018-09-14 07:46:18 -0700212 BPF_FLOW_DISSECTOR,
Andrey Ignatov7b146ce2019-02-27 12:59:24 -0800213 BPF_CGROUP_SYSCTL,
Daniel Borkmann983695f2019-06-07 01:48:57 +0200214 BPF_CGROUP_UDP4_RECVMSG,
215 BPF_CGROUP_UDP6_RECVMSG,
Stanislav Fomichev0d01da62019-06-27 13:38:47 -0700216 BPF_CGROUP_GETSOCKOPT,
217 BPF_CGROUP_SETSOCKOPT,
Alexei Starovoitovf1b95092019-10-30 15:32:11 -0700218 BPF_TRACE_RAW_TP,
Alexei Starovoitovfec56f52019-11-14 10:57:04 -0800219 BPF_TRACE_FENTRY,
220 BPF_TRACE_FEXIT,
KP Singhae240822020-03-04 20:18:49 +0100221 BPF_MODIFY_RETURN,
KP Singhfc611f42020-03-29 01:43:49 +0100222 BPF_LSM_MAC,
Yonghong Song15d83c42020-05-09 10:59:00 -0700223 BPF_TRACE_ITER,
Daniel Borkmann1b66d252020-05-19 00:45:45 +0200224 BPF_CGROUP_INET4_GETPEERNAME,
225 BPF_CGROUP_INET6_GETPEERNAME,
226 BPF_CGROUP_INET4_GETSOCKNAME,
227 BPF_CGROUP_INET6_GETSOCKNAME,
David Ahernfbee97f2020-05-29 16:07:13 -0600228 BPF_XDP_DEVMAP,
Stanislav Fomichevf5836742020-07-06 16:01:25 -0700229 BPF_CGROUP_INET_SOCK_RELEASE,
Daniel Mack0e336612016-11-23 16:52:25 +0100230 __MAX_BPF_ATTACH_TYPE
231};
232
233#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
234
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -0700235enum bpf_link_type {
236 BPF_LINK_TYPE_UNSPEC = 0,
237 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
238 BPF_LINK_TYPE_TRACING = 2,
239 BPF_LINK_TYPE_CGROUP = 3,
Yonghong Songde4e05c2020-05-09 10:59:01 -0700240 BPF_LINK_TYPE_ITER = 4,
Jakub Sitnicki7f045a42020-05-31 10:28:38 +0200241 BPF_LINK_TYPE_NETNS = 5,
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -0700242
243 MAX_BPF_LINK_TYPE,
244};
245
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700246/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
247 *
248 * NONE(default): No further bpf programs allowed in the subtree.
249 *
250 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
251 * the program in this cgroup yields to sub-cgroup program.
252 *
253 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
254 * that cgroup program gets run in addition to the program in this cgroup.
255 *
256 * Only one program is allowed to be attached to a cgroup with
257 * NONE or BPF_F_ALLOW_OVERRIDE flag.
258 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
259 * release old program and attach the new one. Attach flags has to match.
260 *
261 * Multiple programs are allowed to be attached to a cgroup with
262 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
263 * (those that were attached first, run first)
264 * The programs of sub-cgroup are executed first, then programs of
265 * this cgroup and then programs of parent cgroup.
266 * When children program makes decision (like picking TCP CA or sock bind)
267 * parent program has a chance to override it.
268 *
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800269 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
270 * programs for a cgroup. Though it's possible to replace an old program at
271 * any position by also specifying BPF_F_REPLACE flag and position itself in
272 * replace_bpf_fd attribute. Old program at this position will be released.
273 *
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700274 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
275 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
276 * Ex1:
277 * cgrp1 (MULTI progs A, B) ->
278 * cgrp2 (OVERRIDE prog C) ->
279 * cgrp3 (MULTI prog D) ->
280 * cgrp4 (OVERRIDE prog E) ->
281 * cgrp5 (NONE prog F)
282 * the event in cgrp5 triggers execution of F,D,A,B in that order.
283 * if prog F is detached, the execution is E,D,A,B
284 * if prog F and D are detached, the execution is E,A,B
285 * if prog F, E and D are detached, the execution is C,A,B
286 *
287 * All eligible programs are executed regardless of return code from
288 * earlier programs.
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800289 */
290#define BPF_F_ALLOW_OVERRIDE (1U << 0)
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700291#define BPF_F_ALLOW_MULTI (1U << 1)
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800292#define BPF_F_REPLACE (1U << 2)
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800293
David S. Millere07b98d2017-05-10 11:38:07 -0700294/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
295 * verifier will perform strict alignment checking as if the kernel
296 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
297 * and NET_IP_ALIGN defined to 2.
298 */
299#define BPF_F_STRICT_ALIGNMENT (1U << 0)
300
David Millere9ee9ef2018-11-30 21:08:14 -0800301/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
302 * verifier will allow any alignment whatsoever. On platforms
303 * with strict alignment requirements for loads ands stores (such
304 * as sparc and mips) the verifier validates that all loads and
305 * stores provably follow this requirement. This flag turns that
306 * checking and enforcement off.
307 *
308 * It is mostly used for testing when we want to validate the
309 * context and memory access aspects of the verifier, but because
310 * of an unaligned access the alignment check would trigger before
311 * the one we are interested in.
312 */
313#define BPF_F_ANY_ALIGNMENT (1U << 1)
314
Jiong Wangc240eff2019-05-24 23:25:16 +0100315/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
316 * Verifier does sub-register def/use analysis and identifies instructions whose
317 * def only matters for low 32-bit, high 32-bit is never referenced later
318 * through implicit zero extension. Therefore verifier notifies JIT back-ends
319 * that it is safe to ignore clearing high 32-bit for these instructions. This
320 * saves some back-ends a lot of code-gen. However such optimization is not
321 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
322 * hence hasn't used verifier's analysis result. But, we really want to have a
323 * way to be able to verify the correctness of the described optimization on
324 * x86_64 on which testsuites are frequently exercised.
325 *
326 * So, this flag is introduced. Once it is set, verifier will randomize high
327 * 32-bit for those instructions who has been identified as safe to ignore them.
328 * Then, if verifier is not doing correct analysis, such randomization will
329 * regress tests to expose bugs.
330 */
331#define BPF_F_TEST_RND_HI32 (1U << 2)
332
Alexei Starovoitov10d274e2019-08-22 22:52:12 -0700333/* The verifier internal test flag. Behavior is undefined */
334#define BPF_F_TEST_STATE_FREQ (1U << 3)
335
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200336/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
337 * two extensions:
338 *
339 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
340 * insn[0].imm: map fd map fd
341 * insn[1].imm: 0 offset into value
342 * insn[0].off: 0 0
343 * insn[1].off: 0 0
344 * ldimm64 rewrite: address of map address of map[0]+offset
345 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
346 */
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100347#define BPF_PSEUDO_MAP_FD 1
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200348#define BPF_PSEUDO_MAP_VALUE 2
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100349
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800350/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
351 * offset to another bpf function
352 */
353#define BPF_PSEUDO_CALL 1
354
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800355/* flags for BPF_MAP_UPDATE_ELEM command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800356enum {
357 BPF_ANY = 0, /* create new element or update existing */
358 BPF_NOEXIST = 1, /* create new element if it didn't exist */
359 BPF_EXIST = 2, /* update existing element */
360 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
361};
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800362
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700363/* flags for BPF_MAP_CREATE command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800364enum {
365 BPF_F_NO_PREALLOC = (1U << 0),
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800366/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800367 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800368 * which can scale and perform better.
369 * Note, the LRU nodes (including free nodes) cannot be moved
370 * across different LRU lists.
371 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800372 BPF_F_NO_COMMON_LRU = (1U << 1),
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700373/* Specify numa node during map creation */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800374 BPF_F_NUMA_NODE = (1U << 2),
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -0700375
Daniel Borkmann591fe982019-04-09 23:20:05 +0200376/* Flags for accessing BPF object from syscall side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800377 BPF_F_RDONLY = (1U << 3),
378 BPF_F_WRONLY = (1U << 4),
Chenbo Feng6e71b042017-10-18 13:00:22 -0700379
Song Liu615755a2018-03-14 10:23:21 -0700380/* Flag for stack_map, store build_id+offset instead of pointer */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800381 BPF_F_STACK_BUILD_ID = (1U << 5),
Song Liu615755a2018-03-14 10:23:21 -0700382
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000383/* Zero-initialize hash function seed. This should only be used for testing. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800384 BPF_F_ZERO_SEED = (1U << 6),
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000385
Daniel Borkmann591fe982019-04-09 23:20:05 +0200386/* Flags for accessing BPF object from program side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800387 BPF_F_RDONLY_PROG = (1U << 7),
388 BPF_F_WRONLY_PROG = (1U << 8),
Daniel Borkmann591fe982019-04-09 23:20:05 +0200389
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700390/* Clone map from listener for newly accepted socket */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800391 BPF_F_CLONE = (1U << 9),
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700392
Andrii Nakryikofc970222019-11-17 09:28:04 -0800393/* Enable memory-mapping BPF map */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800394 BPF_F_MMAPABLE = (1U << 10),
395};
Andrii Nakryikofc970222019-11-17 09:28:04 -0800396
Andrey Ignatovf5bfcd92020-01-07 17:40:06 -0800397/* Flags for BPF_PROG_QUERY. */
398
399/* Query effective (directly attached + inherited from ancestor cgroups)
400 * programs that will be executed for events within a cgroup.
401 * attach_flags with this flag are returned only for directly attached programs.
402 */
Lorenz Bauer2f183362018-11-16 11:41:09 +0000403#define BPF_F_QUERY_EFFECTIVE (1U << 0)
404
Song Liud46edd62020-04-30 00:15:04 -0700405/* type for BPF_ENABLE_STATS */
406enum bpf_stats_type {
407 /* enabled run_time_ns and run_cnt */
408 BPF_STATS_RUN_TIME = 0,
409};
410
Song Liu615755a2018-03-14 10:23:21 -0700411enum bpf_stack_build_id_status {
412 /* user space need an empty entry to identify end of a trace */
413 BPF_STACK_BUILD_ID_EMPTY = 0,
414 /* with valid build_id and offset */
415 BPF_STACK_BUILD_ID_VALID = 1,
416 /* couldn't get build_id, fallback to ip */
417 BPF_STACK_BUILD_ID_IP = 2,
418};
419
420#define BPF_BUILD_ID_SIZE 20
421struct bpf_stack_build_id {
422 __s32 status;
423 unsigned char build_id[BPF_BUILD_ID_SIZE];
424 union {
425 __u64 offset;
426 __u64 ip;
427 };
428};
429
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800430#define BPF_OBJ_NAME_LEN 16U
431
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700432union bpf_attr {
433 struct { /* anonymous struct used by BPF_MAP_CREATE command */
434 __u32 map_type; /* one of enum bpf_map_type */
435 __u32 key_size; /* size of key in bytes */
436 __u32 value_size; /* size of value in bytes */
437 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700438 __u32 map_flags; /* BPF_MAP_CREATE related
439 * flags defined above.
440 */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700441 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700442 __u32 numa_node; /* numa node (effective only if
443 * BPF_F_NUMA_NODE is set).
444 */
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700445 char map_name[BPF_OBJ_NAME_LEN];
Jakub Kicinskia3884572018-01-11 20:29:09 -0800446 __u32 map_ifindex; /* ifindex of netdev to create on */
Martin KaFai Laua26ca7c2018-04-18 15:56:03 -0700447 __u32 btf_fd; /* fd pointing to a BTF type data */
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -0700448 __u32 btf_key_type_id; /* BTF type_id of the key */
449 __u32 btf_value_type_id; /* BTF type_id of the value */
Martin KaFai Lau85d33df2020-01-08 16:35:05 -0800450 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
451 * struct stored as the
452 * map value
453 */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700454 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700455
456 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
457 __u32 map_fd;
458 __aligned_u64 key;
459 union {
460 __aligned_u64 value;
461 __aligned_u64 next_key;
462 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800463 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700464 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700465
Brian Vazquezcb4d03a2020-01-15 10:43:01 -0800466 struct { /* struct used by BPF_MAP_*_BATCH commands */
467 __aligned_u64 in_batch; /* start batch,
468 * NULL to start from beginning
469 */
470 __aligned_u64 out_batch; /* output: next start batch */
471 __aligned_u64 keys;
472 __aligned_u64 values;
473 __u32 count; /* input/output:
474 * input: # of key/value
475 * elements
476 * output: # of filled elements
477 */
478 __u32 map_fd;
479 __u64 elem_flags;
480 __u64 flags;
481 } batch;
482
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700483 struct { /* anonymous struct used by BPF_PROG_LOAD command */
484 __u32 prog_type; /* one of enum bpf_prog_type */
485 __u32 insn_cnt;
486 __aligned_u64 insns;
487 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700488 __u32 log_level; /* verbosity level of verifier */
489 __u32 log_size; /* size of user buffer */
490 __aligned_u64 log_buf; /* user supplied buffer */
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100491 __u32 kern_version; /* not used */
David S. Millere07b98d2017-05-10 11:38:07 -0700492 __u32 prog_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700493 char prog_name[BPF_OBJ_NAME_LEN];
Jakub Kicinski1f6f4cb2017-11-20 15:21:53 -0800494 __u32 prog_ifindex; /* ifindex of netdev to prep for */
Andrey Ignatov5e43f892018-03-30 15:08:00 -0700495 /* For some prog types expected attach type must be known at
496 * load time to verify attach type specific parts of prog
497 * (context accesses, allowed helpers, etc).
498 */
499 __u32 expected_attach_type;
Yonghong Song838e9692018-11-19 15:29:11 -0800500 __u32 prog_btf_fd; /* fd pointing to BTF type data */
501 __u32 func_info_rec_size; /* userspace bpf_func_info size */
502 __aligned_u64 func_info; /* func info */
503 __u32 func_info_cnt; /* number of bpf_func_info records */
Martin KaFai Lauc454a462018-12-07 16:42:25 -0800504 __u32 line_info_rec_size; /* userspace bpf_line_info size */
505 __aligned_u64 line_info; /* line info */
506 __u32 line_info_cnt; /* number of bpf_line_info records */
Alexei Starovoitovccfe29eb22019-10-15 20:24:58 -0700507 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
Alexei Starovoitov5b92a282019-11-14 10:57:17 -0800508 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700509 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100510
511 struct { /* anonymous struct used by BPF_OBJ_* commands */
512 __aligned_u64 pathname;
513 __u32 bpf_fd;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700514 __u32 file_flags;
Daniel Borkmannb2197752015-10-29 14:58:09 +0100515 };
Daniel Mackf4324552016-11-23 16:52:27 +0100516
517 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
518 __u32 target_fd; /* container object to attach to */
519 __u32 attach_bpf_fd; /* eBPF program to attach */
520 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800521 __u32 attach_flags;
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800522 __u32 replace_bpf_fd; /* previously attached eBPF
523 * program to replace if
524 * BPF_F_REPLACE is used
525 */
Daniel Mackf4324552016-11-23 16:52:27 +0100526 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700527
528 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
529 __u32 prog_fd;
530 __u32 retval;
Lorenz Bauerb5a36b12018-12-03 11:31:23 +0000531 __u32 data_size_in; /* input: len of data_in */
532 __u32 data_size_out; /* input/output: len of data_out
533 * returns ENOSPC if data_out
534 * is too small.
535 */
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700536 __aligned_u64 data_in;
537 __aligned_u64 data_out;
538 __u32 repeat;
539 __u32 duration;
Stanislav Fomichevb0b93952019-04-09 11:49:09 -0700540 __u32 ctx_size_in; /* input: len of ctx_in */
541 __u32 ctx_size_out; /* input/output: len of ctx_out
542 * returns ENOSPC if ctx_out
543 * is too small.
544 */
545 __aligned_u64 ctx_in;
546 __aligned_u64 ctx_out;
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700547 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700548
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700549 struct { /* anonymous struct used by BPF_*_GET_*_ID */
550 union {
551 __u32 start_id;
552 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700553 __u32 map_id;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700554 __u32 btf_id;
Andrii Nakryikoa3b80e12020-04-28 17:16:06 -0700555 __u32 link_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700556 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700557 __u32 next_id;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700558 __u32 open_flags;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700559 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700560
561 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
562 __u32 bpf_fd;
563 __u32 info_len;
564 __aligned_u64 info;
565 } info;
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700566
567 struct { /* anonymous struct used by BPF_PROG_QUERY command */
568 __u32 target_fd; /* container object to query */
569 __u32 attach_type;
570 __u32 query_flags;
571 __u32 attach_flags;
572 __aligned_u64 prog_ids;
573 __u32 prog_cnt;
574 } query;
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700575
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700576 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700577 __u64 name;
578 __u32 prog_fd;
579 } raw_tracepoint;
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700580
581 struct { /* anonymous struct for BPF_BTF_LOAD */
582 __aligned_u64 btf;
583 __aligned_u64 btf_log_buf;
584 __u32 btf_size;
585 __u32 btf_log_size;
586 __u32 btf_log_level;
587 };
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700588
589 struct {
590 __u32 pid; /* input: pid */
591 __u32 fd; /* input: fd */
592 __u32 flags; /* input: flags */
593 __u32 buf_len; /* input/output: buf len */
594 __aligned_u64 buf; /* input/output:
595 * tp_name for tracepoint
596 * symbol for kprobe
597 * filename for uprobe
598 */
599 __u32 prog_id; /* output: prod_id */
600 __u32 fd_type; /* output: BPF_FD_TYPE_* */
601 __u64 probe_offset; /* output: probe_offset */
602 __u64 probe_addr; /* output: probe_addr */
603 } task_fd_query;
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700604
605 struct { /* struct used by BPF_LINK_CREATE command */
606 __u32 prog_fd; /* eBPF program to attach */
607 __u32 target_fd; /* object to attach to */
608 __u32 attach_type; /* attach type */
609 __u32 flags; /* extra flags */
610 } link_create;
Andrii Nakryiko0c991eb2020-03-29 19:59:59 -0700611
612 struct { /* struct used by BPF_LINK_UPDATE command */
613 __u32 link_fd; /* link fd */
614 /* new program fd to update link with */
615 __u32 new_prog_fd;
616 __u32 flags; /* extra flags */
617 /* expected link's program fd; is specified only if
618 * BPF_F_REPLACE flag is set in flags */
619 __u32 old_prog_fd;
620 } link_update;
621
Song Liud46edd62020-04-30 00:15:04 -0700622 struct { /* struct used by BPF_ENABLE_STATS command */
623 __u32 type;
624 } enable_stats;
625
Yonghong Songac51d992020-05-09 10:59:05 -0700626 struct { /* struct used by BPF_ITER_CREATE command */
627 __u32 link_fd;
628 __u32 flags;
629 } iter_create;
630
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700631} __attribute__((aligned(8)));
632
Quentin Monnet56a092c2018-04-25 18:16:52 +0100633/* The description below is an attempt at providing documentation to eBPF
634 * developers about the multiple available eBPF helper functions. It can be
635 * parsed and used to produce a manual page. The workflow is the following,
636 * and requires the rst2man utility:
Thomas Grafebb676d2016-10-27 11:23:51 +0200637 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100638 * $ ./scripts/bpf_helpers_doc.py \
639 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
640 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
641 * $ man /tmp/bpf-helpers.7
Thomas Grafebb676d2016-10-27 11:23:51 +0200642 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100643 * Note that in order to produce this external documentation, some RST
644 * formatting is used in the descriptions to get "bold" and "italics" in
645 * manual pages. Also note that the few trailing white spaces are
646 * intentional, removing them would break paragraphs for rst2man.
Thomas Grafebb676d2016-10-27 11:23:51 +0200647 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100648 * Start of BPF helper function descriptions:
Quentin Monnetad4a5222018-04-25 18:16:53 +0100649 *
650 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
651 * Description
652 * Perform a lookup in *map* for an entry associated to *key*.
653 * Return
654 * Map value associated to *key*, or **NULL** if no entry was
655 * found.
656 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700657 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100658 * Description
659 * Add or update the value of the entry associated to *key* in
660 * *map* with *value*. *flags* is one of:
661 *
662 * **BPF_NOEXIST**
663 * The entry for *key* must not exist in the map.
664 * **BPF_EXIST**
665 * The entry for *key* must already exist in the map.
666 * **BPF_ANY**
667 * No condition on the existence of the entry for *key*.
668 *
669 * Flag value **BPF_NOEXIST** cannot be used for maps of types
670 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
671 * elements always exist), the helper would return an error.
672 * Return
673 * 0 on success, or a negative error in case of failure.
674 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700675 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100676 * Description
677 * Delete entry with *key* from *map*.
678 * Return
679 * 0 on success, or a negative error in case of failure.
680 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700681 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100682 * Description
683 * For tracing programs, safely attempt to read *size* bytes from
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +0100684 * kernel space address *unsafe_ptr* and store the data in *dst*.
685 *
Quentin Monnetab8d7802020-05-11 17:15:35 +0100686 * Generally, use **bpf_probe_read_user**\ () or
687 * **bpf_probe_read_kernel**\ () instead.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100688 * Return
689 * 0 on success, or a negative error in case of failure.
690 *
691 * u64 bpf_ktime_get_ns(void)
692 * Description
693 * Return the time elapsed since system boot, in nanoseconds.
Maciej Żenczykowski71d19212020-04-26 09:15:25 -0700694 * Does not include time the system was suspended.
Quentin Monnetab8d7802020-05-11 17:15:35 +0100695 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100696 * Return
697 * Current *ktime*.
698 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700699 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100700 * Description
701 * This helper is a "printk()-like" facility for debugging. It
702 * prints a message defined by format *fmt* (of size *fmt_size*)
703 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
704 * available. It can take up to three additional **u64**
705 * arguments (as an eBPF helpers, the total number of arguments is
706 * limited to five).
707 *
708 * Each time the helper is called, it appends a line to the trace.
Peter Wu55c33df2019-08-21 00:08:59 +0100709 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
710 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100711 * The format of the trace is customizable, and the exact output
712 * one will get depends on the options set in
713 * *\/sys/kernel/debug/tracing/trace_options* (see also the
714 * *README* file under the same directory). However, it usually
715 * defaults to something like:
716 *
717 * ::
718 *
719 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
720 *
721 * In the above:
722 *
723 * * ``telnet`` is the name of the current task.
724 * * ``470`` is the PID of the current task.
725 * * ``001`` is the CPU number on which the task is
726 * running.
727 * * In ``.N..``, each character refers to a set of
728 * options (whether irqs are enabled, scheduling
729 * options, whether hard/softirqs are running, level of
730 * preempt_disabled respectively). **N** means that
731 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
732 * are set.
733 * * ``419421.045894`` is a timestamp.
734 * * ``0x00000001`` is a fake value used by BPF for the
735 * instruction pointer register.
736 * * ``<formatted msg>`` is the message formatted with
737 * *fmt*.
738 *
739 * The conversion specifiers supported by *fmt* are similar, but
740 * more limited than for printk(). They are **%d**, **%i**,
741 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
742 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
743 * of field, padding with zeroes, etc.) is available, and the
744 * helper will return **-EINVAL** (but print nothing) if it
745 * encounters an unknown specifier.
746 *
747 * Also, note that **bpf_trace_printk**\ () is slow, and should
748 * only be used for debugging purposes. For this reason, a notice
749 * bloc (spanning several lines) is printed to kernel logs and
750 * states that the helper should not be used "for production use"
751 * the first time this helper is used (or more precisely, when
752 * **trace_printk**\ () buffers are allocated). For passing values
753 * to user space, perf events should be preferred.
754 * Return
755 * The number of bytes written to the buffer, or a negative error
756 * in case of failure.
757 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100758 * u32 bpf_get_prandom_u32(void)
759 * Description
760 * Get a pseudo-random number.
761 *
762 * From a security point of view, this helper uses its own
763 * pseudo-random internal state, and cannot be used to infer the
764 * seed of other random functions in the kernel. However, it is
765 * essential to note that the generator used by the helper is not
766 * cryptographically secure.
767 * Return
768 * A random 32-bit unsigned value.
769 *
770 * u32 bpf_get_smp_processor_id(void)
771 * Description
772 * Get the SMP (symmetric multiprocessing) processor id. Note that
773 * all programs run with preemption disabled, which means that the
774 * SMP processor id is stable during all the execution of the
775 * program.
776 * Return
777 * The SMP id of the processor running the program.
778 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700779 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100780 * Description
781 * Store *len* bytes from address *from* into the packet
782 * associated to *skb*, at *offset*. *flags* are a combination of
783 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
784 * checksum for the packet after storing the bytes) and
785 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
786 * **->swhash** and *skb*\ **->l4hash** to 0).
787 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100788 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100789 * packet buffer. Therefore, at load time, all checks on pointers
790 * previously done by the verifier are invalidated and must be
791 * performed again, if the helper is used in combination with
792 * direct packet access.
793 * Return
794 * 0 on success, or a negative error in case of failure.
795 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700796 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100797 * Description
798 * Recompute the layer 3 (e.g. IP) checksum for the packet
799 * associated to *skb*. Computation is incremental, so the helper
800 * must know the former value of the header field that was
801 * modified (*from*), the new value of this field (*to*), and the
802 * number of bytes (2 or 4) for this field, stored in *size*.
803 * Alternatively, it is possible to store the difference between
804 * the previous and the new values of the header field in *to*, by
805 * setting *from* and *size* to 0. For both methods, *offset*
806 * indicates the location of the IP checksum within the packet.
807 *
808 * This helper works in combination with **bpf_csum_diff**\ (),
809 * which does not update the checksum in-place, but offers more
810 * flexibility and can handle sizes larger than 2 or 4 for the
811 * checksum to update.
812 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100813 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100814 * packet buffer. Therefore, at load time, all checks on pointers
815 * previously done by the verifier are invalidated and must be
816 * performed again, if the helper is used in combination with
817 * direct packet access.
818 * Return
819 * 0 on success, or a negative error in case of failure.
820 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700821 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100822 * Description
823 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
824 * packet associated to *skb*. Computation is incremental, so the
825 * helper must know the former value of the header field that was
826 * modified (*from*), the new value of this field (*to*), and the
827 * number of bytes (2 or 4) for this field, stored on the lowest
828 * four bits of *flags*. Alternatively, it is possible to store
829 * the difference between the previous and the new values of the
830 * header field in *to*, by setting *from* and the four lowest
831 * bits of *flags* to 0. For both methods, *offset* indicates the
832 * location of the IP checksum within the packet. In addition to
833 * the size of the field, *flags* can be added (bitwise OR) actual
834 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
835 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
836 * for updates resulting in a null checksum the value is set to
837 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
838 * the checksum is to be computed against a pseudo-header.
839 *
840 * This helper works in combination with **bpf_csum_diff**\ (),
841 * which does not update the checksum in-place, but offers more
842 * flexibility and can handle sizes larger than 2 or 4 for the
843 * checksum to update.
844 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100845 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100846 * packet buffer. Therefore, at load time, all checks on pointers
847 * previously done by the verifier are invalidated and must be
848 * performed again, if the helper is used in combination with
849 * direct packet access.
850 * Return
851 * 0 on success, or a negative error in case of failure.
852 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700853 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100854 * Description
855 * This special helper is used to trigger a "tail call", or in
856 * other words, to jump into another eBPF program. The same stack
857 * frame is used (but values on stack and in registers for the
858 * caller are not accessible to the callee). This mechanism allows
859 * for program chaining, either for raising the maximum number of
860 * available eBPF instructions, or to execute given programs in
861 * conditional blocks. For security reasons, there is an upper
862 * limit to the number of successive tail calls that can be
863 * performed.
864 *
865 * Upon call of this helper, the program attempts to jump into a
866 * program referenced at index *index* in *prog_array_map*, a
867 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
868 * *ctx*, a pointer to the context.
869 *
870 * If the call succeeds, the kernel immediately runs the first
871 * instruction of the new program. This is not a function call,
872 * and it never returns to the previous program. If the call
873 * fails, then the helper has no effect, and the caller continues
874 * to run its subsequent instructions. A call can fail if the
875 * destination program for the jump does not exist (i.e. *index*
876 * is superior to the number of entries in *prog_array_map*), or
877 * if the maximum number of tail calls has been reached for this
878 * chain of programs. This limit is defined in the kernel by the
879 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
880 * which is currently set to 32.
881 * Return
882 * 0 on success, or a negative error in case of failure.
883 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700884 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100885 * Description
886 * Clone and redirect the packet associated to *skb* to another
887 * net device of index *ifindex*. Both ingress and egress
888 * interfaces can be used for redirection. The **BPF_F_INGRESS**
889 * value in *flags* is used to make the distinction (ingress path
890 * is selected if the flag is present, egress path otherwise).
891 * This is the only flag supported for now.
892 *
893 * In comparison with **bpf_redirect**\ () helper,
894 * **bpf_clone_redirect**\ () has the associated cost of
895 * duplicating the packet buffer, but this can be executed out of
896 * the eBPF program. Conversely, **bpf_redirect**\ () is more
897 * efficient, but it is handled through an action code where the
898 * redirection happens only after the eBPF program has returned.
899 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100900 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100901 * packet buffer. Therefore, at load time, all checks on pointers
902 * previously done by the verifier are invalidated and must be
903 * performed again, if the helper is used in combination with
904 * direct packet access.
905 * Return
906 * 0 on success, or a negative error in case of failure.
Quentin Monnetc456dec2018-04-25 18:16:54 +0100907 *
908 * u64 bpf_get_current_pid_tgid(void)
909 * Return
910 * A 64-bit integer containing the current tgid and pid, and
911 * created as such:
912 * *current_task*\ **->tgid << 32 \|**
913 * *current_task*\ **->pid**.
914 *
915 * u64 bpf_get_current_uid_gid(void)
916 * Return
917 * A 64-bit integer containing the current GID and UID, and
918 * created as such: *current_gid* **<< 32 \|** *current_uid*.
919 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700920 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100921 * Description
922 * Copy the **comm** attribute of the current task into *buf* of
923 * *size_of_buf*. The **comm** attribute contains the name of
924 * the executable (excluding the path) for the current task. The
925 * *size_of_buf* must be strictly positive. On success, the
926 * helper makes sure that the *buf* is NUL-terminated. On failure,
927 * it is filled with zeroes.
928 * Return
929 * 0 on success, or a negative error in case of failure.
930 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100931 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
932 * Description
933 * Retrieve the classid for the current task, i.e. for the net_cls
934 * cgroup to which *skb* belongs.
935 *
936 * This helper can be used on TC egress path, but not on ingress.
937 *
938 * The net_cls cgroup provides an interface to tag network packets
939 * based on a user-provided identifier for all traffic coming from
940 * the tasks belonging to the related cgroup. See also the related
941 * kernel documentation, available from the Linux sources in file
Mauro Carvalho Chehabda82c922019-06-27 13:08:35 -0300942 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100943 *
944 * The Linux kernel has two versions for cgroups: there are
945 * cgroups v1 and cgroups v2. Both are available to users, who can
946 * use a mixture of them, but note that the net_cls cgroup is for
947 * cgroup v1 only. This makes it incompatible with BPF programs
948 * run on cgroups, which is a cgroup-v2-only feature (a socket can
949 * only hold data for one version of cgroups at a time).
950 *
951 * This helper is only available is the kernel was compiled with
952 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
953 * "**y**" or to "**m**".
954 * Return
955 * The classid, or 0 for the default unconfigured classid.
956 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700957 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100958 * Description
959 * Push a *vlan_tci* (VLAN tag control information) of protocol
960 * *vlan_proto* to the packet associated to *skb*, then update
961 * the checksum. Note that if *vlan_proto* is different from
962 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
963 * be **ETH_P_8021Q**.
964 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100965 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100966 * packet buffer. Therefore, at load time, all checks on pointers
967 * previously done by the verifier are invalidated and must be
968 * performed again, if the helper is used in combination with
969 * direct packet access.
970 * Return
971 * 0 on success, or a negative error in case of failure.
972 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700973 * long bpf_skb_vlan_pop(struct sk_buff *skb)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100974 * Description
975 * Pop a VLAN header from the packet associated to *skb*.
976 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100977 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100978 * packet buffer. Therefore, at load time, all checks on pointers
979 * previously done by the verifier are invalidated and must be
980 * performed again, if the helper is used in combination with
981 * direct packet access.
982 * Return
983 * 0 on success, or a negative error in case of failure.
984 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -0700985 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100986 * Description
987 * Get tunnel metadata. This helper takes a pointer *key* to an
988 * empty **struct bpf_tunnel_key** of **size**, that will be
989 * filled with tunnel metadata for the packet associated to *skb*.
990 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
991 * indicates that the tunnel is based on IPv6 protocol instead of
992 * IPv4.
993 *
994 * The **struct bpf_tunnel_key** is an object that generalizes the
995 * principal parameters used by various tunneling protocols into a
996 * single struct. This way, it can be used to easily make a
997 * decision based on the contents of the encapsulation header,
998 * "summarized" in this struct. In particular, it holds the IP
999 * address of the remote end (IPv4 or IPv6, depending on the case)
1000 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1001 * this struct exposes the *key*\ **->tunnel_id**, which is
1002 * generally mapped to a VNI (Virtual Network Identifier), making
1003 * it programmable together with the **bpf_skb_set_tunnel_key**\
1004 * () helper.
1005 *
1006 * Let's imagine that the following code is part of a program
1007 * attached to the TC ingress interface, on one end of a GRE
1008 * tunnel, and is supposed to filter out all messages coming from
1009 * remote ends with IPv4 address other than 10.0.0.1:
1010 *
1011 * ::
1012 *
1013 * int ret;
1014 * struct bpf_tunnel_key key = {};
1015 *
1016 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1017 * if (ret < 0)
1018 * return TC_ACT_SHOT; // drop packet
1019 *
1020 * if (key.remote_ipv4 != 0x0a000001)
1021 * return TC_ACT_SHOT; // drop packet
1022 *
1023 * return TC_ACT_OK; // accept packet
1024 *
1025 * This interface can also be used with all encapsulation devices
1026 * that can operate in "collect metadata" mode: instead of having
1027 * one network device per specific configuration, the "collect
1028 * metadata" mode only requires a single device where the
1029 * configuration can be extracted from this helper.
1030 *
1031 * This can be used together with various tunnels such as VXLan,
1032 * Geneve, GRE or IP in IP (IPIP).
1033 * Return
1034 * 0 on success, or a negative error in case of failure.
1035 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001036 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001037 * Description
1038 * Populate tunnel metadata for packet associated to *skb.* The
1039 * tunnel metadata is set to the contents of *key*, of *size*. The
1040 * *flags* can be set to a combination of the following values:
1041 *
1042 * **BPF_F_TUNINFO_IPV6**
1043 * Indicate that the tunnel is based on IPv6 protocol
1044 * instead of IPv4.
1045 * **BPF_F_ZERO_CSUM_TX**
1046 * For IPv4 packets, add a flag to tunnel metadata
1047 * indicating that checksum computation should be skipped
1048 * and checksum set to zeroes.
1049 * **BPF_F_DONT_FRAGMENT**
1050 * Add a flag to tunnel metadata indicating that the
1051 * packet should not be fragmented.
1052 * **BPF_F_SEQ_NUMBER**
1053 * Add a flag to tunnel metadata indicating that a
1054 * sequence number should be added to tunnel header before
1055 * sending the packet. This flag was added for GRE
1056 * encapsulation, but might be used with other protocols
1057 * as well in the future.
1058 *
1059 * Here is a typical usage on the transmit path:
1060 *
1061 * ::
1062 *
1063 * struct bpf_tunnel_key key;
1064 * populate key ...
1065 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1066 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1067 *
1068 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1069 * helper for additional information.
1070 * Return
1071 * 0 on success, or a negative error in case of failure.
1072 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001073 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1074 * Description
1075 * Read the value of a perf event counter. This helper relies on a
1076 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1077 * the perf event counter is selected when *map* is updated with
1078 * perf event file descriptors. The *map* is an array whose size
1079 * is the number of available CPUs, and each cell contains a value
1080 * relative to one CPU. The value to retrieve is indicated by
1081 * *flags*, that contains the index of the CPU to look up, masked
1082 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1083 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1084 * current CPU should be retrieved.
1085 *
1086 * Note that before Linux 4.13, only hardware perf event can be
1087 * retrieved.
1088 *
1089 * Also, be aware that the newer helper
1090 * **bpf_perf_event_read_value**\ () is recommended over
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001091 * **bpf_perf_event_read**\ () in general. The latter has some ABI
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001092 * quirks where error and counter value are used as a return code
1093 * (which is wrong to do since ranges may overlap). This issue is
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001094 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1095 * time provides more features over the **bpf_perf_event_read**\
1096 * () interface. Please refer to the description of
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001097 * **bpf_perf_event_read_value**\ () for details.
1098 * Return
1099 * The value of the perf event counter read from the map, or a
1100 * negative error code in case of failure.
1101 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001102 * long bpf_redirect(u32 ifindex, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001103 * Description
1104 * Redirect the packet to another net device of index *ifindex*.
1105 * This helper is somewhat similar to **bpf_clone_redirect**\
1106 * (), except that the packet is not cloned, which provides
1107 * increased performance.
1108 *
1109 * Except for XDP, both ingress and egress interfaces can be used
1110 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1111 * to make the distinction (ingress path is selected if the flag
1112 * is present, egress path otherwise). Currently, XDP only
1113 * supports redirection to the egress interface, and accepts no
1114 * flag at all.
1115 *
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001116 * The same effect can also be attained with the more generic
1117 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1118 * redirect target instead of providing it directly to the helper.
Quentin Monnetc456dec2018-04-25 18:16:54 +01001119 * Return
1120 * For XDP, the helper returns **XDP_REDIRECT** on success or
1121 * **XDP_ABORTED** on error. For other program types, the values
1122 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1123 * error.
1124 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001125 * u32 bpf_get_route_realm(struct sk_buff *skb)
1126 * Description
1127 * Retrieve the realm or the route, that is to say the
1128 * **tclassid** field of the destination for the *skb*. The
1129 * indentifier retrieved is a user-provided tag, similar to the
1130 * one used with the net_cls cgroup (see description for
1131 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1132 * held by a route (a destination entry), not by a task.
1133 *
1134 * Retrieving this identifier works with the clsact TC egress hook
1135 * (see also **tc-bpf(8)**), or alternatively on conventional
1136 * classful egress qdiscs, but not on TC ingress path. In case of
1137 * clsact TC egress hook, this has the advantage that, internally,
1138 * the destination entry has not been dropped yet in the transmit
1139 * path. Therefore, the destination entry does not need to be
1140 * artificially held via **netif_keep_dst**\ () for a classful
1141 * qdisc until the *skb* is freed.
1142 *
1143 * This helper is available only if the kernel was compiled with
1144 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1145 * Return
1146 * The realm of the route for the packet associated to *skb*, or 0
1147 * if none was found.
1148 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001149 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001150 * Description
1151 * Write raw *data* blob into a special BPF perf event held by
1152 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1153 * event must have the following attributes: **PERF_SAMPLE_RAW**
1154 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1155 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1156 *
1157 * The *flags* are used to indicate the index in *map* for which
1158 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1159 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1160 * to indicate that the index of the current CPU core should be
1161 * used.
1162 *
1163 * The value to write, of *size*, is passed through eBPF stack and
1164 * pointed by *data*.
1165 *
1166 * The context of the program *ctx* needs also be passed to the
1167 * helper.
1168 *
1169 * On user space, a program willing to read the values needs to
1170 * call **perf_event_open**\ () on the perf event (either for
1171 * one or for all CPUs) and to store the file descriptor into the
1172 * *map*. This must be done before the eBPF program can send data
1173 * into it. An example is available in file
1174 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1175 * tree (the eBPF program counterpart is in
1176 * *samples/bpf/trace_output_kern.c*).
1177 *
1178 * **bpf_perf_event_output**\ () achieves better performance
1179 * than **bpf_trace_printk**\ () for sharing data with user
1180 * space, and is much better suitable for streaming data from eBPF
1181 * programs.
1182 *
1183 * Note that this helper is not restricted to tracing use cases
1184 * and can be used with programs attached to TC or XDP as well,
1185 * where it allows for passing data to user space listeners. Data
1186 * can be:
1187 *
1188 * * Only custom structs,
1189 * * Only the packet payload, or
1190 * * A combination of both.
1191 * Return
1192 * 0 on success, or a negative error in case of failure.
1193 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001194 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001195 * Description
1196 * This helper was provided as an easy way to load data from a
1197 * packet. It can be used to load *len* bytes from *offset* from
1198 * the packet associated to *skb*, into the buffer pointed by
1199 * *to*.
1200 *
1201 * Since Linux 4.7, usage of this helper has mostly been replaced
1202 * by "direct packet access", enabling packet data to be
1203 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1204 * pointing respectively to the first byte of packet data and to
1205 * the byte after the last byte of packet data. However, it
1206 * remains useful if one wishes to read large quantities of data
1207 * at once from a packet into the eBPF stack.
1208 * Return
1209 * 0 on success, or a negative error in case of failure.
1210 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001211 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001212 * Description
1213 * Walk a user or a kernel stack and return its id. To achieve
1214 * this, the helper needs *ctx*, which is a pointer to the context
1215 * on which the tracing program is executed, and a pointer to a
1216 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1217 *
1218 * The last argument, *flags*, holds the number of stack frames to
1219 * skip (from 0 to 255), masked with
1220 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1221 * a combination of the following flags:
1222 *
1223 * **BPF_F_USER_STACK**
1224 * Collect a user space stack instead of a kernel stack.
1225 * **BPF_F_FAST_STACK_CMP**
1226 * Compare stacks by hash only.
1227 * **BPF_F_REUSE_STACKID**
1228 * If two different stacks hash into the same *stackid*,
1229 * discard the old one.
1230 *
1231 * The stack id retrieved is a 32 bit long integer handle which
1232 * can be further combined with other data (including other stack
1233 * ids) and used as a key into maps. This can be useful for
1234 * generating a variety of graphs (such as flame graphs or off-cpu
1235 * graphs).
1236 *
1237 * For walking a stack, this helper is an improvement over
1238 * **bpf_probe_read**\ (), which can be used with unrolled loops
1239 * but is not efficient and consumes a lot of eBPF instructions.
1240 * Instead, **bpf_get_stackid**\ () can collect up to
1241 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1242 * this limit can be controlled with the **sysctl** program, and
1243 * that it should be manually increased in order to profile long
1244 * user stacks (such as stacks for Java programs). To do so, use:
1245 *
1246 * ::
1247 *
1248 * # sysctl kernel.perf_event_max_stack=<new value>
Quentin Monnetc456dec2018-04-25 18:16:54 +01001249 * Return
1250 * The positive or null stack id on success, or a negative error
1251 * in case of failure.
1252 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001253 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1254 * Description
1255 * Compute a checksum difference, from the raw buffer pointed by
1256 * *from*, of length *from_size* (that must be a multiple of 4),
1257 * towards the raw buffer pointed by *to*, of size *to_size*
1258 * (same remark). An optional *seed* can be added to the value
1259 * (this can be cascaded, the seed may come from a previous call
1260 * to the helper).
1261 *
1262 * This is flexible enough to be used in several ways:
1263 *
1264 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1265 * checksum, it can be used when pushing new data.
1266 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1267 * checksum, it can be used when removing data from a packet.
1268 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1269 * can be used to compute a diff. Note that *from_size* and
1270 * *to_size* do not need to be equal.
1271 *
1272 * This helper can be used in combination with
1273 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1274 * which one can feed in the difference computed with
1275 * **bpf_csum_diff**\ ().
1276 * Return
1277 * The checksum result, or a negative error code in case of
1278 * failure.
1279 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001280 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001281 * Description
1282 * Retrieve tunnel options metadata for the packet associated to
1283 * *skb*, and store the raw tunnel option data to the buffer *opt*
1284 * of *size*.
1285 *
1286 * This helper can be used with encapsulation devices that can
1287 * operate in "collect metadata" mode (please refer to the related
1288 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1289 * more details). A particular example where this can be used is
1290 * in combination with the Geneve encapsulation protocol, where it
1291 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1292 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1293 * the eBPF program. This allows for full customization of these
1294 * headers.
1295 * Return
1296 * The size of the option data retrieved.
1297 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001298 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001299 * Description
1300 * Set tunnel options metadata for the packet associated to *skb*
1301 * to the option data contained in the raw buffer *opt* of *size*.
1302 *
1303 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1304 * helper for additional information.
1305 * Return
1306 * 0 on success, or a negative error in case of failure.
1307 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001308 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001309 * Description
1310 * Change the protocol of the *skb* to *proto*. Currently
1311 * supported are transition from IPv4 to IPv6, and from IPv6 to
1312 * IPv4. The helper takes care of the groundwork for the
1313 * transition, including resizing the socket buffer. The eBPF
1314 * program is expected to fill the new headers, if any, via
1315 * **skb_store_bytes**\ () and to recompute the checksums with
1316 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1317 * (). The main case for this helper is to perform NAT64
1318 * operations out of an eBPF program.
1319 *
1320 * Internally, the GSO type is marked as dodgy so that headers are
1321 * checked and segments are recalculated by the GSO/GRO engine.
1322 * The size for GSO target is adapted as well.
1323 *
1324 * All values for *flags* are reserved for future usage, and must
1325 * be left at zero.
1326 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001327 * A call to this helper is susceptible to change the underlying
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001328 * packet buffer. Therefore, at load time, all checks on pointers
1329 * previously done by the verifier are invalidated and must be
1330 * performed again, if the helper is used in combination with
1331 * direct packet access.
1332 * Return
1333 * 0 on success, or a negative error in case of failure.
1334 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001335 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001336 * Description
1337 * Change the packet type for the packet associated to *skb*. This
1338 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1339 * the eBPF program does not have a write access to *skb*\
1340 * **->pkt_type** beside this helper. Using a helper here allows
1341 * for graceful handling of errors.
1342 *
1343 * The major use case is to change incoming *skb*s to
1344 * **PACKET_HOST** in a programmatic way instead of having to
1345 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1346 * example.
1347 *
1348 * Note that *type* only allows certain values. At this time, they
1349 * are:
1350 *
1351 * **PACKET_HOST**
1352 * Packet is for us.
1353 * **PACKET_BROADCAST**
1354 * Send packet to all.
1355 * **PACKET_MULTICAST**
1356 * Send packet to group.
1357 * **PACKET_OTHERHOST**
1358 * Send packet to someone else.
1359 * Return
1360 * 0 on success, or a negative error in case of failure.
1361 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001362 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001363 * Description
1364 * Check whether *skb* is a descendant of the cgroup2 held by
1365 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1366 * Return
1367 * The return value depends on the result of the test, and can be:
1368 *
1369 * * 0, if the *skb* failed the cgroup2 descendant test.
1370 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1371 * * A negative error code, if an error occurred.
1372 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001373 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1374 * Description
1375 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1376 * not set, in particular if the hash was cleared due to mangling,
1377 * recompute this hash. Later accesses to the hash can be done
1378 * directly with *skb*\ **->hash**.
1379 *
1380 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1381 * prototype with **bpf_skb_change_proto**\ (), or calling
1382 * **bpf_skb_store_bytes**\ () with the
1383 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1384 * the hash and to trigger a new computation for the next call to
1385 * **bpf_get_hash_recalc**\ ().
1386 * Return
1387 * The 32-bit hash.
1388 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001389 * u64 bpf_get_current_task(void)
1390 * Return
1391 * A pointer to the current task struct.
Quentin Monnetfa156012018-04-25 18:16:56 +01001392 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001393 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001394 * Description
1395 * Attempt in a safe way to write *len* bytes from the buffer
1396 * *src* to *dst* in memory. It only works for threads that are in
1397 * user context, and *dst* must be a valid user space address.
1398 *
1399 * This helper should not be used to implement any kind of
1400 * security mechanism because of TOC-TOU attacks, but rather to
1401 * debug, divert, and manipulate execution of semi-cooperative
1402 * processes.
1403 *
1404 * Keep in mind that this feature is meant for experiments, and it
1405 * has a risk of crashing the system and running programs.
1406 * Therefore, when an eBPF program using this helper is attached,
1407 * a warning including PID and process name is printed to kernel
1408 * logs.
1409 * Return
1410 * 0 on success, or a negative error in case of failure.
1411 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001412 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001413 * Description
1414 * Check whether the probe is being run is the context of a given
1415 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1416 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1417 * Return
1418 * The return value depends on the result of the test, and can be:
1419 *
1420 * * 0, if the *skb* task belongs to the cgroup2.
1421 * * 1, if the *skb* task does not belong to the cgroup2.
1422 * * A negative error code, if an error occurred.
1423 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001424 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001425 * Description
1426 * Resize (trim or grow) the packet associated to *skb* to the
1427 * new *len*. The *flags* are reserved for future usage, and must
1428 * be left at zero.
1429 *
1430 * The basic idea is that the helper performs the needed work to
1431 * change the size of the packet, then the eBPF program rewrites
1432 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1433 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1434 * and others. This helper is a slow path utility intended for
1435 * replies with control messages. And because it is targeted for
1436 * slow path, the helper itself can afford to be slow: it
1437 * implicitly linearizes, unclones and drops offloads from the
1438 * *skb*.
1439 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001440 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001441 * packet buffer. Therefore, at load time, all checks on pointers
1442 * previously done by the verifier are invalidated and must be
1443 * performed again, if the helper is used in combination with
1444 * direct packet access.
1445 * Return
1446 * 0 on success, or a negative error in case of failure.
1447 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001448 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
Quentin Monnetfa156012018-04-25 18:16:56 +01001449 * Description
1450 * Pull in non-linear data in case the *skb* is non-linear and not
1451 * all of *len* are part of the linear section. Make *len* bytes
1452 * from *skb* readable and writable. If a zero value is passed for
1453 * *len*, then the whole length of the *skb* is pulled.
1454 *
1455 * This helper is only needed for reading and writing with direct
1456 * packet access.
1457 *
1458 * For direct packet access, testing that offsets to access
1459 * are within packet boundaries (test on *skb*\ **->data_end**) is
1460 * susceptible to fail if offsets are invalid, or if the requested
1461 * data is in non-linear parts of the *skb*. On failure the
1462 * program can just bail out, or in the case of a non-linear
1463 * buffer, use a helper to make the data available. The
1464 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1465 * the data. Another one consists in using **bpf_skb_pull_data**
1466 * to pull in once the non-linear parts, then retesting and
1467 * eventually access the data.
1468 *
1469 * At the same time, this also makes sure the *skb* is uncloned,
1470 * which is a necessary condition for direct write. As this needs
1471 * to be an invariant for the write part only, the verifier
1472 * detects writes and adds a prologue that is calling
1473 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1474 * the very beginning in case it is indeed cloned.
1475 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001476 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001477 * packet buffer. Therefore, at load time, all checks on pointers
1478 * previously done by the verifier are invalidated and must be
1479 * performed again, if the helper is used in combination with
1480 * direct packet access.
1481 * Return
1482 * 0 on success, or a negative error in case of failure.
1483 *
1484 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1485 * Description
1486 * Add the checksum *csum* into *skb*\ **->csum** in case the
1487 * driver has supplied a checksum for the entire packet into that
1488 * field. Return an error otherwise. This helper is intended to be
1489 * used in combination with **bpf_csum_diff**\ (), in particular
1490 * when the checksum needs to be updated after data has been
1491 * written into the packet through direct packet access.
1492 * Return
1493 * The checksum on success, or a negative error code in case of
1494 * failure.
1495 *
1496 * void bpf_set_hash_invalid(struct sk_buff *skb)
1497 * Description
1498 * Invalidate the current *skb*\ **->hash**. It can be used after
1499 * mangling on headers through direct packet access, in order to
1500 * indicate that the hash is outdated and to trigger a
1501 * recalculation the next time the kernel tries to access this
1502 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1503 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001504 * long bpf_get_numa_node_id(void)
Quentin Monnetfa156012018-04-25 18:16:56 +01001505 * Description
1506 * Return the id of the current NUMA node. The primary use case
1507 * for this helper is the selection of sockets for the local NUMA
1508 * node, when the program is attached to sockets using the
1509 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1510 * but the helper is also available to other eBPF program types,
1511 * similarly to **bpf_get_smp_processor_id**\ ().
1512 * Return
1513 * The id of current NUMA node.
1514 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001515 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001516 * Description
1517 * Grows headroom of packet associated to *skb* and adjusts the
1518 * offset of the MAC header accordingly, adding *len* bytes of
1519 * space. It automatically extends and reallocates memory as
1520 * required.
1521 *
1522 * This helper can be used on a layer 3 *skb* to push a MAC header
1523 * for redirection into a layer 2 device.
1524 *
1525 * All values for *flags* are reserved for future usage, and must
1526 * be left at zero.
1527 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001528 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001529 * packet buffer. Therefore, at load time, all checks on pointers
1530 * previously done by the verifier are invalidated and must be
1531 * performed again, if the helper is used in combination with
1532 * direct packet access.
1533 * Return
1534 * 0 on success, or a negative error in case of failure.
1535 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001536 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001537 * Description
1538 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1539 * it is possible to use a negative value for *delta*. This helper
1540 * can be used to prepare the packet for pushing or popping
1541 * headers.
1542 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001543 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001544 * packet buffer. Therefore, at load time, all checks on pointers
1545 * previously done by the verifier are invalidated and must be
1546 * performed again, if the helper is used in combination with
1547 * direct packet access.
1548 * Return
1549 * 0 on success, or a negative error in case of failure.
1550 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001551 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001552 * Description
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001553 * Copy a NUL terminated string from an unsafe kernel address
Quentin Monnetab8d7802020-05-11 17:15:35 +01001554 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001555 * more details.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001556 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001557 * Generally, use **bpf_probe_read_user_str**\ () or
1558 * **bpf_probe_read_kernel_str**\ () instead.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001559 * Return
1560 * On success, the strictly positive length of the string,
1561 * including the trailing NUL character. On error, a negative
1562 * value.
1563 *
1564 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1565 * Description
1566 * If the **struct sk_buff** pointed by *skb* has a known socket,
1567 * retrieve the cookie (generated by the kernel) of this socket.
1568 * If no cookie has been set yet, generate a new cookie. Once
1569 * generated, the socket cookie remains stable for the life of the
1570 * socket. This helper can be useful for monitoring per socket
Daniel Borkmanncd48bdd2019-08-08 13:57:25 +02001571 * networking traffic statistics as it provides a global socket
1572 * identifier that can be assumed unique.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001573 * Return
1574 * A 8-byte long non-decreasing number on success, or 0 if the
1575 * socket field is missing inside *skb*.
1576 *
Andrey Ignatovd692f112018-07-30 17:42:28 -07001577 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1578 * Description
1579 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001580 * *skb*, but gets socket from **struct bpf_sock_addr** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001581 * Return
1582 * A 8-byte long non-decreasing number.
1583 *
1584 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1585 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01001586 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001587 * *skb*, but gets socket from **struct bpf_sock_ops** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001588 * Return
1589 * A 8-byte long non-decreasing number.
1590 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001591 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1592 * Return
1593 * The owner UID of the socket associated to *skb*. If the socket
1594 * is **NULL**, or if it is not a full socket (i.e. if it is a
1595 * time-wait or a request socket instead), **overflowuid** value
1596 * is returned (note that **overflowuid** might also be the actual
1597 * UID value for the socket).
1598 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001599 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
Quentin Monnetfa156012018-04-25 18:16:56 +01001600 * Description
1601 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1602 * to value *hash*.
1603 * Return
1604 * 0
1605 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001606 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001607 * Description
1608 * Emulate a call to **setsockopt()** on the socket associated to
1609 * *bpf_socket*, which must be a full socket. The *level* at
1610 * which the option resides and the name *optname* of the option
1611 * must be specified, see **setsockopt(2)** for more information.
1612 * The option value of length *optlen* is pointed by *optval*.
1613 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001614 * *bpf_socket* should be one of the following:
Quentin Monnetab8d7802020-05-11 17:15:35 +01001615 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001616 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1617 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1618 * and **BPF_CGROUP_INET6_CONNECT**.
1619 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001620 * This helper actually implements a subset of **setsockopt()**.
1621 * It supports the following *level*\ s:
1622 *
1623 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1624 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
Dmitry Yakuninf9bcf962020-06-20 18:30:52 +03001625 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1626 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001627 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1628 * **TCP_CONGESTION**, **TCP_BPF_IW**,
Dmitry Yakuninf9bcf962020-06-20 18:30:52 +03001629 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1630 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
1631 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001632 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1633 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1634 * Return
1635 * 0 on success, or a negative error in case of failure.
1636 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001637 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001638 * Description
1639 * Grow or shrink the room for data in the packet associated to
1640 * *skb* by *len_diff*, and according to the selected *mode*.
1641 *
Daniel Borkmann836e66c2020-06-02 16:58:32 +02001642 * By default, the helper will reset any offloaded checksum
1643 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1644 * by the following flag:
1645 *
1646 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1647 * checksum data of the skb to CHECKSUM_NONE.
1648 *
Willem de Bruijn14aa3192019-03-22 14:32:54 -04001649 * There are two supported modes at this time:
1650 *
1651 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1652 * (room space is added or removed below the layer 2 header).
Quentin Monnetfa156012018-04-25 18:16:56 +01001653 *
1654 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1655 * (room space is added or removed below the layer 3 header).
1656 *
Willem de Bruijn868d5232019-03-22 14:32:56 -04001657 * The following flags are supported at this time:
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04001658 *
1659 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1660 * Adjusting mss in this way is not allowed for datagrams.
Quentin Monnetfa156012018-04-25 18:16:56 +01001661 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001662 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1663 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001664 * Any new space is reserved to hold a tunnel header.
1665 * Configure skb offsets and other fields accordingly.
1666 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001667 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1668 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001669 * Use with ENCAP_L3 flags to further specify the tunnel type.
1670 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001671 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
Alan Maguire58dfc902019-04-09 15:06:41 +01001672 * Use with ENCAP_L3/L4 flags to further specify the tunnel
Quentin Monnet80867c52019-05-10 15:51:24 +01001673 * type; *len* is the length of the inner MAC header.
Alan Maguire58dfc902019-04-09 15:06:41 +01001674 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001675 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001676 * packet buffer. Therefore, at load time, all checks on pointers
1677 * previously done by the verifier are invalidated and must be
1678 * performed again, if the helper is used in combination with
1679 * direct packet access.
1680 * Return
1681 * 0 on success, or a negative error in case of failure.
1682 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001683 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001684 * Description
1685 * Redirect the packet to the endpoint referenced by *map* at
1686 * index *key*. Depending on its type, this *map* can contain
1687 * references to net devices (for forwarding packets through other
1688 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1689 * but this is only implemented for native XDP (with driver
1690 * support) as of this writing).
1691 *
Toke Høiland-Jørgensen43e74c02019-06-28 11:12:34 +02001692 * The lower two bits of *flags* are used as the return code if
1693 * the map lookup fails. This is so that the return value can be
Quentin Monnetab8d7802020-05-11 17:15:35 +01001694 * one of the XDP program return codes up to **XDP_TX**, as chosen
1695 * by the caller. Any higher bits in the *flags* argument must be
Toke Høiland-Jørgensen43e74c02019-06-28 11:12:34 +02001696 * unset.
Quentin Monnetab127042018-04-25 18:16:59 +01001697 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001698 * See also **bpf_redirect**\ (), which only supports redirecting
1699 * to an ifindex, but doesn't require a map to do so.
Quentin Monnetab127042018-04-25 18:16:59 +01001700 * Return
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001701 * **XDP_REDIRECT** on success, or the value of the two lower bits
Jakub Wilka33d3142020-04-22 10:23:24 +02001702 * of the *flags* argument on error.
Quentin Monnetab127042018-04-25 18:16:59 +01001703 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001704 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001705 * Description
1706 * Redirect the packet to the socket referenced by *map* (of type
1707 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1708 * egress interfaces can be used for redirection. The
1709 * **BPF_F_INGRESS** value in *flags* is used to make the
1710 * distinction (ingress path is selected if the flag is present,
1711 * egress path otherwise). This is the only flag supported for now.
1712 * Return
1713 * **SK_PASS** on success, or **SK_DROP** on error.
1714 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001715 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001716 * Description
1717 * Add an entry to, or update a *map* referencing sockets. The
1718 * *skops* is used as a new value for the entry associated to
1719 * *key*. *flags* is one of:
1720 *
1721 * **BPF_NOEXIST**
1722 * The entry for *key* must not exist in the map.
1723 * **BPF_EXIST**
1724 * The entry for *key* must already exist in the map.
1725 * **BPF_ANY**
1726 * No condition on the existence of the entry for *key*.
1727 *
1728 * If the *map* has eBPF programs (parser and verdict), those will
1729 * be inherited by the socket being added. If the socket is
1730 * already attached to eBPF programs, this results in an error.
1731 * Return
1732 * 0 on success, or a negative error in case of failure.
1733 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001734 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
Quentin Monnetfa156012018-04-25 18:16:56 +01001735 * Description
1736 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1737 * *delta* (which can be positive or negative). Note that this
1738 * operation modifies the address stored in *xdp_md*\ **->data**,
1739 * so the latter must be loaded only after the helper has been
1740 * called.
1741 *
1742 * The use of *xdp_md*\ **->data_meta** is optional and programs
1743 * are not required to use it. The rationale is that when the
1744 * packet is processed with XDP (e.g. as DoS filter), it is
1745 * possible to push further meta data along with it before passing
1746 * to the stack, and to give the guarantee that an ingress eBPF
1747 * program attached as a TC classifier on the same device can pick
1748 * this up for further post-processing. Since TC works with socket
1749 * buffers, it remains possible to set from XDP the **mark** or
1750 * **priority** pointers, or other pointers for the socket buffer.
1751 * Having this scratch space generic and programmable allows for
1752 * more flexibility as the user is free to store whatever meta
1753 * data they need.
1754 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001755 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001756 * packet buffer. Therefore, at load time, all checks on pointers
1757 * previously done by the verifier are invalidated and must be
1758 * performed again, if the helper is used in combination with
1759 * direct packet access.
1760 * Return
1761 * 0 on success, or a negative error in case of failure.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001762 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001763 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001764 * Description
1765 * Read the value of a perf event counter, and store it into *buf*
1766 * of size *buf_size*. This helper relies on a *map* of type
1767 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1768 * counter is selected when *map* is updated with perf event file
1769 * descriptors. The *map* is an array whose size is the number of
1770 * available CPUs, and each cell contains a value relative to one
1771 * CPU. The value to retrieve is indicated by *flags*, that
1772 * contains the index of the CPU to look up, masked with
1773 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1774 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1775 * current CPU should be retrieved.
1776 *
1777 * This helper behaves in a way close to
1778 * **bpf_perf_event_read**\ () helper, save that instead of
1779 * just returning the value observed, it fills the *buf*
1780 * structure. This allows for additional data to be retrieved: in
1781 * particular, the enabled and running times (in *buf*\
1782 * **->enabled** and *buf*\ **->running**, respectively) are
1783 * copied. In general, **bpf_perf_event_read_value**\ () is
1784 * recommended over **bpf_perf_event_read**\ (), which has some
1785 * ABI issues and provides fewer functionalities.
1786 *
1787 * These values are interesting, because hardware PMU (Performance
1788 * Monitoring Unit) counters are limited resources. When there are
1789 * more PMU based perf events opened than available counters,
1790 * kernel will multiplex these events so each event gets certain
1791 * percentage (but not all) of the PMU time. In case that
1792 * multiplexing happens, the number of samples or counter value
1793 * will not reflect the case compared to when no multiplexing
1794 * occurs. This makes comparison between different runs difficult.
1795 * Typically, the counter value should be normalized before
1796 * comparing to other experiments. The usual normalization is done
1797 * as follows.
1798 *
1799 * ::
1800 *
1801 * normalized_counter = counter * t_enabled / t_running
1802 *
1803 * Where t_enabled is the time enabled for event and t_running is
1804 * the time running for event since last normalization. The
1805 * enabled and running times are accumulated since the perf event
1806 * open. To achieve scaling factor between two invocations of an
Quentin Monnetab8d7802020-05-11 17:15:35 +01001807 * eBPF program, users can use CPU id as the key (which is
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001808 * typical for perf array usage model) to remember the previous
1809 * value and do the calculation inside the eBPF program.
1810 * Return
1811 * 0 on success, or a negative error in case of failure.
1812 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001813 * long 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 +01001814 * Description
1815 * For en eBPF program attached to a perf event, retrieve the
1816 * value of the event counter associated to *ctx* and store it in
1817 * the structure pointed by *buf* and of size *buf_size*. Enabled
1818 * and running times are also stored in the structure (see
1819 * description of helper **bpf_perf_event_read_value**\ () for
1820 * more details).
1821 * Return
1822 * 0 on success, or a negative error in case of failure.
1823 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001824 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001825 * Description
1826 * Emulate a call to **getsockopt()** on the socket associated to
1827 * *bpf_socket*, which must be a full socket. The *level* at
1828 * which the option resides and the name *optname* of the option
1829 * must be specified, see **getsockopt(2)** for more information.
1830 * The retrieved value is stored in the structure pointed by
1831 * *opval* and of length *optlen*.
1832 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001833 * *bpf_socket* should be one of the following:
Quentin Monnetab8d7802020-05-11 17:15:35 +01001834 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001835 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1836 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1837 * and **BPF_CGROUP_INET6_CONNECT**.
1838 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001839 * This helper actually implements a subset of **getsockopt()**.
1840 * It supports the following *level*\ s:
1841 *
1842 * * **IPPROTO_TCP**, which supports *optname*
1843 * **TCP_CONGESTION**.
1844 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1845 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1846 * Return
1847 * 0 on success, or a negative error in case of failure.
1848 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001849 * long bpf_override_return(struct pt_regs *regs, u64 rc)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001850 * Description
1851 * Used for error injection, this helper uses kprobes to override
1852 * the return value of the probed function, and to set it to *rc*.
1853 * The first argument is the context *regs* on which the kprobe
1854 * works.
1855 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001856 * This helper works by setting the PC (program counter)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001857 * to an override function which is run in place of the original
1858 * probed function. This means the probed function is not run at
1859 * all. The replacement function just returns with the required
1860 * value.
1861 *
1862 * This helper has security implications, and thus is subject to
1863 * restrictions. It is only available if the kernel was compiled
1864 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1865 * option, and in this case it only works on functions tagged with
1866 * **ALLOW_ERROR_INJECTION** in the kernel code.
1867 *
1868 * Also, the helper is only available for the architectures having
1869 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1870 * x86 architecture is the only one to support this feature.
1871 * Return
1872 * 0
1873 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001874 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001875 * Description
1876 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1877 * for the full TCP socket associated to *bpf_sock_ops* to
1878 * *argval*.
1879 *
1880 * The primary use of this field is to determine if there should
1881 * be calls to eBPF programs of type
1882 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1883 * code. A program of the same type can change its value, per
1884 * connection and as necessary, when the connection is
1885 * established. This field is directly accessible for reading, but
1886 * this helper must be used for updates in order to return an
1887 * error if an eBPF program tries to set a callback that is not
1888 * supported in the current kernel.
1889 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001890 * *argval* is a flag array which can combine these flags:
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001891 *
1892 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1893 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1894 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07001895 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001896 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001897 * Therefore, this function can be used to clear a callback flag by
1898 * setting the appropriate bit to zero. e.g. to disable the RTO
1899 * callback:
1900 *
1901 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1902 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1903 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001904 * Here are some examples of where one could call such eBPF
1905 * program:
1906 *
1907 * * When RTO fires.
1908 * * When a packet is retransmitted.
1909 * * When the connection terminates.
1910 * * When a packet is sent.
1911 * * When a packet is received.
1912 * Return
1913 * Code **-EINVAL** if the socket is not a full TCP socket;
1914 * otherwise, a positive number containing the bits that could not
1915 * be set is returned (which comes down to 0 if all bits were set
1916 * as required).
1917 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001918 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001919 * Description
1920 * This helper is used in programs implementing policies at the
1921 * socket level. If the message *msg* is allowed to pass (i.e. if
1922 * the verdict eBPF program returns **SK_PASS**), redirect it to
1923 * the socket referenced by *map* (of type
1924 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1925 * egress interfaces can be used for redirection. The
1926 * **BPF_F_INGRESS** value in *flags* is used to make the
1927 * distinction (ingress path is selected if the flag is present,
1928 * egress path otherwise). This is the only flag supported for now.
1929 * Return
1930 * **SK_PASS** on success, or **SK_DROP** on error.
1931 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001932 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
Quentin Monnetab127042018-04-25 18:16:59 +01001933 * Description
1934 * For socket policies, apply the verdict of the eBPF program to
1935 * the next *bytes* (number of bytes) of message *msg*.
1936 *
1937 * For example, this helper can be used in the following cases:
1938 *
1939 * * A single **sendmsg**\ () or **sendfile**\ () system call
1940 * contains multiple logical messages that the eBPF program is
1941 * supposed to read and for which it should apply a verdict.
1942 * * An eBPF program only cares to read the first *bytes* of a
1943 * *msg*. If the message has a large payload, then setting up
1944 * and calling the eBPF program repeatedly for all bytes, even
1945 * though the verdict is already known, would create unnecessary
1946 * overhead.
1947 *
1948 * When called from within an eBPF program, the helper sets a
1949 * counter internal to the BPF infrastructure, that is used to
1950 * apply the last verdict to the next *bytes*. If *bytes* is
1951 * smaller than the current data being processed from a
1952 * **sendmsg**\ () or **sendfile**\ () system call, the first
1953 * *bytes* will be sent and the eBPF program will be re-run with
1954 * the pointer for start of data pointing to byte number *bytes*
1955 * **+ 1**. If *bytes* is larger than the current data being
1956 * processed, then the eBPF verdict will be applied to multiple
1957 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1958 * consumed.
1959 *
1960 * Note that if a socket closes with the internal counter holding
1961 * a non-zero value, this is not a problem because data is not
1962 * being buffered for *bytes* and is sent as it is received.
1963 * Return
1964 * 0
1965 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001966 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
Quentin Monnetab127042018-04-25 18:16:59 +01001967 * Description
1968 * For socket policies, prevent the execution of the verdict eBPF
1969 * program for message *msg* until *bytes* (byte number) have been
1970 * accumulated.
1971 *
1972 * This can be used when one needs a specific number of bytes
1973 * before a verdict can be assigned, even if the data spans
1974 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1975 * case would be a user calling **sendmsg**\ () repeatedly with
1976 * 1-byte long message segments. Obviously, this is bad for
1977 * performance, but it is still valid. If the eBPF program needs
1978 * *bytes* bytes to validate a header, this helper can be used to
1979 * prevent the eBPF program to be called again until *bytes* have
1980 * been accumulated.
1981 * Return
1982 * 0
1983 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07001984 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001985 * Description
1986 * For socket policies, pull in non-linear data from user space
1987 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1988 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1989 * respectively.
1990 *
1991 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1992 * *msg* it can only parse data that the (**data**, **data_end**)
1993 * pointers have already consumed. For **sendmsg**\ () hooks this
1994 * is likely the first scatterlist element. But for calls relying
1995 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1996 * be the range (**0**, **0**) because the data is shared with
1997 * user space and by default the objective is to avoid allowing
1998 * user space to modify data while (or after) eBPF verdict is
1999 * being decided. This helper can be used to pull in data and to
2000 * set the start and end pointer to given values. Data will be
2001 * copied if necessary (i.e. if data was not linear and if start
2002 * and end pointers do not point to the same chunk).
2003 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002004 * A call to this helper is susceptible to change the underlying
Quentin Monnetab127042018-04-25 18:16:59 +01002005 * packet buffer. Therefore, at load time, all checks on pointers
2006 * previously done by the verifier are invalidated and must be
2007 * performed again, if the helper is used in combination with
2008 * direct packet access.
2009 *
2010 * All values for *flags* are reserved for future usage, and must
2011 * be left at zero.
2012 * Return
2013 * 0 on success, or a negative error in case of failure.
2014 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002015 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01002016 * Description
2017 * Bind the socket associated to *ctx* to the address pointed by
2018 * *addr*, of length *addr_len*. This allows for making outgoing
2019 * connection from the desired IP address, which can be useful for
2020 * example when all processes inside a cgroup should use one
2021 * single IP address on a host that has multiple IP configured.
2022 *
2023 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2024 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
Stanislav Fomichev8086fba2020-05-08 10:46:11 -07002025 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2026 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2027 * behavior and lets the kernel efficiently pick up an unused
2028 * port as long as 4-tuple is unique. Passing non-zero port might
2029 * lead to degraded performance.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01002030 * Return
2031 * 0 on success, or a negative error in case of failure.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002032 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002033 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002034 * Description
2035 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
Jesper Dangaard Brouerc8741e22020-05-14 12:51:25 +02002036 * possible to both shrink and grow the packet tail.
2037 * Shrink done via *delta* being a negative integer.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002038 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002039 * A call to this helper is susceptible to change the underlying
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002040 * packet buffer. Therefore, at load time, all checks on pointers
2041 * previously done by the verifier are invalidated and must be
2042 * performed again, if the helper is used in combination with
2043 * direct packet access.
2044 * Return
2045 * 0 on success, or a negative error in case of failure.
2046 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002047 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002048 * Description
2049 * Retrieve the XFRM state (IP transform framework, see also
2050 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2051 *
2052 * The retrieved value is stored in the **struct bpf_xfrm_state**
2053 * pointed by *xfrm_state* and of length *size*.
2054 *
2055 * All values for *flags* are reserved for future usage, and must
2056 * be left at zero.
2057 *
2058 * This helper is available only if the kernel was compiled with
2059 * **CONFIG_XFRM** configuration option.
2060 * Return
2061 * 0 on success, or a negative error in case of failure.
Yonghong Songc195651e2018-04-28 22:28:08 -07002062 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002063 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
Yonghong Songc195651e2018-04-28 22:28:08 -07002064 * Description
Quentin Monnet79552fb2018-04-30 11:39:04 +01002065 * Return a user or a kernel stack in bpf program provided buffer.
2066 * To achieve this, the helper needs *ctx*, which is a pointer
2067 * to the context on which the tracing program is executed.
2068 * To store the stacktrace, the bpf program provides *buf* with
2069 * a nonnegative *size*.
Yonghong Songc195651e2018-04-28 22:28:08 -07002070 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002071 * The last argument, *flags*, holds the number of stack frames to
2072 * skip (from 0 to 255), masked with
2073 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2074 * the following flags:
Yonghong Songc195651e2018-04-28 22:28:08 -07002075 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002076 * **BPF_F_USER_STACK**
2077 * Collect a user space stack instead of a kernel stack.
2078 * **BPF_F_USER_BUILD_ID**
2079 * Collect buildid+offset instead of ips for user stack,
2080 * only valid if **BPF_F_USER_STACK** is also specified.
Yonghong Songc195651e2018-04-28 22:28:08 -07002081 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002082 * **bpf_get_stack**\ () can collect up to
2083 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2084 * to sufficient large buffer size. Note that
2085 * this limit can be controlled with the **sysctl** program, and
2086 * that it should be manually increased in order to profile long
2087 * user stacks (such as stacks for Java programs). To do so, use:
Yonghong Songc195651e2018-04-28 22:28:08 -07002088 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002089 * ::
Yonghong Songc195651e2018-04-28 22:28:08 -07002090 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002091 * # sysctl kernel.perf_event_max_stack=<new value>
Yonghong Songc195651e2018-04-28 22:28:08 -07002092 * Return
Quentin Monnet7a279e92018-05-29 12:27:44 +01002093 * A non-negative value equal to or less than *size* on success,
2094 * or a negative error in case of failure.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002095 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002096 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002097 * Description
2098 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2099 * it provides an easy way to load *len* bytes from *offset*
2100 * from the packet associated to *skb*, into the buffer pointed
2101 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2102 * a fifth argument *start_header* exists in order to select a
2103 * base offset to start from. *start_header* can be one of:
2104 *
2105 * **BPF_HDR_START_MAC**
2106 * Base offset to load data from is *skb*'s mac header.
2107 * **BPF_HDR_START_NET**
2108 * Base offset to load data from is *skb*'s network header.
2109 *
2110 * In general, "direct packet access" is the preferred method to
2111 * access packet data, however, this helper is in particular useful
2112 * in socket filters where *skb*\ **->data** does not always point
2113 * to the start of the mac header and where "direct packet access"
2114 * is not available.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002115 * Return
2116 * 0 on success, or a negative error in case of failure.
2117 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002118 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
David Ahern87f5fc72018-05-09 20:34:26 -07002119 * Description
2120 * Do FIB lookup in kernel tables using parameters in *params*.
2121 * If lookup is successful and result shows packet is to be
2122 * forwarded, the neighbor tables are searched for the nexthop.
2123 * If successful (ie., FIB lookup shows forwarding and nexthop
David Ahernfa898d72018-05-29 10:58:07 -07002124 * is resolved), the nexthop address is returned in ipv4_dst
2125 * or ipv6_dst based on family, smac is set to mac address of
2126 * egress device, dmac is set to nexthop mac address, rt_metric
David Ahern4c795792018-06-26 16:21:18 -07002127 * is set to metric from route (IPv4/IPv6 only), and ifindex
2128 * is set to the device index of the nexthop from the FIB lookup.
David Ahern87f5fc72018-05-09 20:34:26 -07002129 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002130 * *plen* argument is the size of the passed in struct.
2131 * *flags* argument can be a combination of one or more of the
2132 * following values:
David Ahern87f5fc72018-05-09 20:34:26 -07002133 *
Quentin Monnet7a279e92018-05-29 12:27:44 +01002134 * **BPF_FIB_LOOKUP_DIRECT**
2135 * Do a direct table lookup vs full lookup using FIB
2136 * rules.
2137 * **BPF_FIB_LOOKUP_OUTPUT**
2138 * Perform lookup from an egress perspective (default is
2139 * ingress).
David Ahern87f5fc72018-05-09 20:34:26 -07002140 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002141 * *ctx* is either **struct xdp_md** for XDP programs or
2142 * **struct sk_buff** tc cls_act programs.
2143 * Return
David Ahern4c795792018-06-26 16:21:18 -07002144 * * < 0 if any input argument is invalid
2145 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2146 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
Quentin Monnet2bae79d2018-07-12 12:52:22 +01002147 * packet is not forwarded or needs assist from full stack
John Fastabend81110382018-05-14 10:00:17 -07002148 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002149 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
John Fastabend81110382018-05-14 10:00:17 -07002150 * Description
2151 * Add an entry to, or update a sockhash *map* referencing sockets.
2152 * The *skops* is used as a new value for the entry associated to
2153 * *key*. *flags* is one of:
2154 *
2155 * **BPF_NOEXIST**
2156 * The entry for *key* must not exist in the map.
2157 * **BPF_EXIST**
2158 * The entry for *key* must already exist in the map.
2159 * **BPF_ANY**
2160 * No condition on the existence of the entry for *key*.
2161 *
2162 * If the *map* has eBPF programs (parser and verdict), those will
2163 * be inherited by the socket being added. If the socket is
2164 * already attached to eBPF programs, this results in an error.
2165 * Return
2166 * 0 on success, or a negative error in case of failure.
2167 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002168 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
John Fastabend81110382018-05-14 10:00:17 -07002169 * Description
2170 * This helper is used in programs implementing policies at the
2171 * socket level. If the message *msg* is allowed to pass (i.e. if
2172 * the verdict eBPF program returns **SK_PASS**), redirect it to
2173 * the socket referenced by *map* (of type
2174 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2175 * egress interfaces can be used for redirection. The
2176 * **BPF_F_INGRESS** value in *flags* is used to make the
2177 * distinction (ingress path is selected if the flag is present,
2178 * egress path otherwise). This is the only flag supported for now.
2179 * Return
2180 * **SK_PASS** on success, or **SK_DROP** on error.
2181 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002182 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
John Fastabend81110382018-05-14 10:00:17 -07002183 * Description
2184 * This helper is used in programs implementing policies at the
2185 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2186 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2187 * to the socket referenced by *map* (of type
2188 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2189 * egress interfaces can be used for redirection. The
2190 * **BPF_F_INGRESS** value in *flags* is used to make the
2191 * distinction (ingress path is selected if the flag is present,
2192 * egress otherwise). This is the only flag supported for now.
2193 * Return
2194 * **SK_PASS** on success, or **SK_DROP** on error.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002195 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002196 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002197 * Description
2198 * Encapsulate the packet associated to *skb* within a Layer 3
2199 * protocol header. This header is provided in the buffer at
2200 * address *hdr*, with *len* its size in bytes. *type* indicates
2201 * the protocol of the header and can be one of:
2202 *
2203 * **BPF_LWT_ENCAP_SEG6**
2204 * IPv6 encapsulation with Segment Routing Header
2205 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2206 * the IPv6 header is computed by the kernel.
2207 * **BPF_LWT_ENCAP_SEG6_INLINE**
2208 * Only works if *skb* contains an IPv6 packet. Insert a
2209 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2210 * the IPv6 header.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002211 * **BPF_LWT_ENCAP_IP**
2212 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2213 * must be IPv4 or IPv6, followed by zero or more
Quentin Monnet80867c52019-05-10 15:51:24 +01002214 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2215 * total bytes in all prepended headers. Please note that
2216 * if **skb_is_gso**\ (*skb*) is true, no more than two
2217 * headers can be prepended, and the inner header, if
2218 * present, should be either GRE or UDP/GUE.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002219 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002220 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2221 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2222 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2223 * **BPF_PROG_TYPE_LWT_XMIT**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002224 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002225 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002226 * packet buffer. Therefore, at load time, all checks on pointers
2227 * previously done by the verifier are invalidated and must be
2228 * performed again, if the helper is used in combination with
2229 * direct packet access.
2230 * Return
2231 * 0 on success, or a negative error in case of failure.
2232 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002233 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002234 * Description
2235 * Store *len* bytes from address *from* into the packet
2236 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2237 * inside the outermost IPv6 Segment Routing Header can be
2238 * modified through this helper.
2239 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002240 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002241 * packet buffer. Therefore, at load time, all checks on pointers
2242 * previously done by the verifier are invalidated and must be
2243 * performed again, if the helper is used in combination with
2244 * direct packet access.
2245 * Return
2246 * 0 on success, or a negative error in case of failure.
2247 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002248 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002249 * Description
2250 * Adjust the size allocated to TLVs in the outermost IPv6
2251 * Segment Routing Header contained in the packet associated to
2252 * *skb*, at position *offset* by *delta* bytes. Only offsets
2253 * after the segments are accepted. *delta* can be as well
2254 * positive (growing) as negative (shrinking).
2255 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002256 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002257 * packet buffer. Therefore, at load time, all checks on pointers
2258 * previously done by the verifier are invalidated and must be
2259 * performed again, if the helper is used in combination with
2260 * direct packet access.
2261 * Return
2262 * 0 on success, or a negative error in case of failure.
2263 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002264 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002265 * Description
2266 * Apply an IPv6 Segment Routing action of type *action* to the
2267 * packet associated to *skb*. Each action takes a parameter
2268 * contained at address *param*, and of length *param_len* bytes.
2269 * *action* can be one of:
2270 *
2271 * **SEG6_LOCAL_ACTION_END_X**
2272 * End.X action: Endpoint with Layer-3 cross-connect.
2273 * Type of *param*: **struct in6_addr**.
2274 * **SEG6_LOCAL_ACTION_END_T**
2275 * End.T action: Endpoint with specific IPv6 table lookup.
2276 * Type of *param*: **int**.
2277 * **SEG6_LOCAL_ACTION_END_B6**
2278 * End.B6 action: Endpoint bound to an SRv6 policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002279 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002280 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2281 * End.B6.Encap action: Endpoint bound to an SRv6
2282 * encapsulation policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002283 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002284 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002285 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002286 * packet buffer. Therefore, at load time, all checks on pointers
2287 * previously done by the verifier are invalidated and must be
2288 * performed again, if the helper is used in combination with
2289 * direct packet access.
2290 * Return
2291 * 0 on success, or a negative error in case of failure.
Sean Youngf4364dc2018-05-27 12:24:09 +01002292 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002293 * long bpf_rc_repeat(void *ctx)
Quentin Monnet62369db2019-03-14 12:38:39 +00002294 * Description
2295 * This helper is used in programs implementing IR decoding, to
2296 * report a successfully decoded repeat key message. This delays
2297 * the generation of a key up event for previously generated
2298 * key down event.
2299 *
2300 * Some IR protocols like NEC have a special IR message for
2301 * repeating last button, for when a button is held down.
2302 *
2303 * The *ctx* should point to the lirc sample as passed into
2304 * the program.
2305 *
2306 * This helper is only available is the kernel was compiled with
2307 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2308 * "**y**".
2309 * Return
2310 * 0
2311 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002312 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
Sean Youngf4364dc2018-05-27 12:24:09 +01002313 * Description
2314 * This helper is used in programs implementing IR decoding, to
2315 * report a successfully decoded key press with *scancode*,
2316 * *toggle* value in the given *protocol*. The scancode will be
2317 * translated to a keycode using the rc keymap, and reported as
2318 * an input key down event. After a period a key up event is
2319 * generated. This period can be extended by calling either
Quentin Monnet90b10232018-12-03 12:13:35 +00002320 * **bpf_rc_keydown**\ () again with the same values, or calling
2321 * **bpf_rc_repeat**\ ().
Sean Youngf4364dc2018-05-27 12:24:09 +01002322 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01002323 * Some protocols include a toggle bit, in case the button was
Sean Youngf4364dc2018-05-27 12:24:09 +01002324 * released and pressed again between consecutive scancodes.
2325 *
2326 * The *ctx* should point to the lirc sample as passed into
2327 * the program.
2328 *
2329 * The *protocol* is the decoded protocol number (see
2330 * **enum rc_proto** for some predefined values).
2331 *
2332 * This helper is only available is the kernel was compiled with
2333 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2334 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002335 * Return
2336 * 0
2337 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002338 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002339 * Description
2340 * Return the cgroup v2 id of the socket associated with the *skb*.
2341 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2342 * helper for cgroup v1 by providing a tag resp. identifier that
2343 * can be matched on or used for map lookups e.g. to implement
2344 * policy. The cgroup v2 id of a given path in the hierarchy is
2345 * exposed in user space through the f_handle API in order to get
2346 * to the same 64-bit id.
2347 *
2348 * This helper can be used on TC egress path, but not on ingress,
2349 * and is available only if the kernel was compiled with the
2350 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2351 * Return
2352 * The id is returned or 0 in case the id could not be retrieved.
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002353 *
2354 * u64 bpf_get_current_cgroup_id(void)
2355 * Return
2356 * A 64-bit integer containing the current cgroup id based
2357 * on the cgroup within which the current task is running.
Roman Gushchincd339432018-08-02 14:27:24 -07002358 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002359 * void *bpf_get_local_storage(void *map, u64 flags)
Roman Gushchincd339432018-08-02 14:27:24 -07002360 * Description
2361 * Get the pointer to the local storage area.
2362 * The type and the size of the local storage is defined
2363 * by the *map* argument.
2364 * The *flags* meaning is specific for each map type,
2365 * and has to be 0 for cgroup local storage.
2366 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002367 * Depending on the BPF program type, a local storage area
2368 * can be shared between multiple instances of the BPF program,
Roman Gushchincd339432018-08-02 14:27:24 -07002369 * running simultaneously.
2370 *
2371 * A user should care about the synchronization by himself.
Quentin Monnet90b10232018-12-03 12:13:35 +00002372 * For example, by using the **BPF_STX_XADD** instruction to alter
Roman Gushchincd339432018-08-02 14:27:24 -07002373 * the shared data.
2374 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002375 * A pointer to the local storage area.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002376 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002377 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002378 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002379 * Select a **SO_REUSEPORT** socket from a
2380 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2381 * It checks the selected socket is matching the incoming
2382 * request in the socket buffer.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002383 * Return
2384 * 0 on success, or a negative error in case of failure.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002385 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002386 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2387 * Description
2388 * Return id of cgroup v2 that is ancestor of cgroup associated
2389 * with the *skb* at the *ancestor_level*. The root cgroup is at
2390 * *ancestor_level* zero and each step down the hierarchy
2391 * increments the level. If *ancestor_level* == level of cgroup
2392 * associated with *skb*, then return value will be same as that
2393 * of **bpf_skb_cgroup_id**\ ().
2394 *
2395 * The helper is useful to implement policies based on cgroups
2396 * that are upper in hierarchy than immediate cgroup associated
2397 * with *skb*.
2398 *
2399 * The format of returned id and helper limitations are same as in
2400 * **bpf_skb_cgroup_id**\ ().
2401 * Return
2402 * The id is returned or 0 in case the id could not be retrieved.
2403 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002404 * 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 -07002405 * Description
2406 * Look for TCP socket matching *tuple*, optionally in a child
2407 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002408 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002409 *
2410 * The *ctx* should point to the context of the program, such as
2411 * the skb or socket (depending on the hook in use). This is used
2412 * to determine the base network namespace for the lookup.
2413 *
2414 * *tuple_size* must be one of:
2415 *
2416 * **sizeof**\ (*tuple*\ **->ipv4**)
2417 * Look for an IPv4 socket.
2418 * **sizeof**\ (*tuple*\ **->ipv6**)
2419 * Look for an IPv6 socket.
2420 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002421 * If the *netns* is a negative signed 32-bit integer, then the
2422 * socket lookup table in the netns associated with the *ctx* will
2423 * will be used. For the TC hooks, this is the netns of the device
2424 * in the skb. For socket hooks, this is the netns of the socket.
2425 * If *netns* is any other signed 32-bit value greater than or
2426 * equal to zero then it specifies the ID of the netns relative to
2427 * the netns associated with the *ctx*. *netns* values beyond the
2428 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002429 *
2430 * All values for *flags* are reserved for future usage, and must
2431 * be left at zero.
2432 *
2433 * This helper is available only if the kernel was compiled with
2434 * **CONFIG_NET** configuration option.
2435 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002436 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2437 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002438 * result is from *reuse*\ **->socks**\ [] using the hash of the
2439 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002440 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002441 * 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 -07002442 * Description
2443 * Look for UDP socket matching *tuple*, optionally in a child
2444 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002445 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002446 *
2447 * The *ctx* should point to the context of the program, such as
2448 * the skb or socket (depending on the hook in use). This is used
2449 * to determine the base network namespace for the lookup.
2450 *
2451 * *tuple_size* must be one of:
2452 *
2453 * **sizeof**\ (*tuple*\ **->ipv4**)
2454 * Look for an IPv4 socket.
2455 * **sizeof**\ (*tuple*\ **->ipv6**)
2456 * Look for an IPv6 socket.
2457 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002458 * If the *netns* is a negative signed 32-bit integer, then the
2459 * socket lookup table in the netns associated with the *ctx* will
2460 * will be used. For the TC hooks, this is the netns of the device
2461 * in the skb. For socket hooks, this is the netns of the socket.
2462 * If *netns* is any other signed 32-bit value greater than or
2463 * equal to zero then it specifies the ID of the netns relative to
2464 * the netns associated with the *ctx*. *netns* values beyond the
2465 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002466 *
2467 * All values for *flags* are reserved for future usage, and must
2468 * be left at zero.
2469 *
2470 * This helper is available only if the kernel was compiled with
2471 * **CONFIG_NET** configuration option.
2472 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002473 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2474 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002475 * result is from *reuse*\ **->socks**\ [] using the hash of the
2476 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002477 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002478 * long bpf_sk_release(struct bpf_sock *sock)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002479 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002480 * Release the reference held by *sock*. *sock* must be a
2481 * non-**NULL** pointer that was returned from
2482 * **bpf_sk_lookup_xxx**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002483 * Return
2484 * 0 on success, or a negative error in case of failure.
John Fastabend6fff6072018-10-19 19:56:49 -07002485 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002486 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
Quentin Monnet62369db2019-03-14 12:38:39 +00002487 * Description
2488 * Push an element *value* in *map*. *flags* is one of:
2489 *
2490 * **BPF_EXIST**
2491 * If the queue/stack is full, the oldest element is
2492 * removed to make room for this.
2493 * Return
2494 * 0 on success, or a negative error in case of failure.
2495 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002496 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
Quentin Monnet90b10232018-12-03 12:13:35 +00002497 * Description
2498 * Pop an element from *map*.
2499 * Return
2500 * 0 on success, or a negative error in case of failure.
2501 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002502 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
Quentin Monnet90b10232018-12-03 12:13:35 +00002503 * Description
2504 * Get an element from *map* without removing it.
2505 * Return
2506 * 0 on success, or a negative error in case of failure.
2507 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002508 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
John Fastabend6fff6072018-10-19 19:56:49 -07002509 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002510 * For socket policies, insert *len* bytes into *msg* at offset
John Fastabend6fff6072018-10-19 19:56:49 -07002511 * *start*.
2512 *
2513 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
Quentin Monnet90b10232018-12-03 12:13:35 +00002514 * *msg* it may want to insert metadata or options into the *msg*.
John Fastabend6fff6072018-10-19 19:56:49 -07002515 * This can later be read and used by any of the lower layer BPF
2516 * hooks.
2517 *
2518 * This helper may fail if under memory pressure (a malloc
2519 * fails) in these cases BPF programs will get an appropriate
2520 * error and BPF programs will need to handle them.
John Fastabend6fff6072018-10-19 19:56:49 -07002521 * Return
2522 * 0 on success, or a negative error in case of failure.
John Fastabend7246d8e2018-11-26 14:16:17 -08002523 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002524 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
Quentin Monnet90b10232018-12-03 12:13:35 +00002525 * Description
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002526 * Will remove *len* bytes from a *msg* starting at byte *start*.
John Fastabend7246d8e2018-11-26 14:16:17 -08002527 * This may result in **ENOMEM** errors under certain situations if
2528 * an allocation and copy are required due to a full ring buffer.
2529 * However, the helper will try to avoid doing the allocation
2530 * if possible. Other errors can occur if input parameters are
Quentin Monnet90b10232018-12-03 12:13:35 +00002531 * invalid either due to *start* byte not being valid part of *msg*
John Fastabend7246d8e2018-11-26 14:16:17 -08002532 * payload and/or *pop* value being to large.
John Fastabend7246d8e2018-11-26 14:16:17 -08002533 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002534 * 0 on success, or a negative error in case of failure.
Sean Young01d32402018-12-06 13:01:03 +00002535 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002536 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
Sean Young01d32402018-12-06 13:01:03 +00002537 * Description
2538 * This helper is used in programs implementing IR decoding, to
2539 * report a successfully decoded pointer movement.
2540 *
2541 * The *ctx* should point to the lirc sample as passed into
2542 * the program.
2543 *
2544 * This helper is only available is the kernel was compiled with
2545 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2546 * "**y**".
2547 * Return
2548 * 0
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002549 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002550 * long bpf_spin_lock(struct bpf_spin_lock *lock)
Quentin Monnet0eb09782019-03-14 12:38:40 +00002551 * Description
2552 * Acquire a spinlock represented by the pointer *lock*, which is
2553 * stored as part of a value of a map. Taking the lock allows to
2554 * safely update the rest of the fields in that value. The
2555 * spinlock can (and must) later be released with a call to
2556 * **bpf_spin_unlock**\ (\ *lock*\ ).
2557 *
2558 * Spinlocks in BPF programs come with a number of restrictions
2559 * and constraints:
2560 *
2561 * * **bpf_spin_lock** objects are only allowed inside maps of
2562 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2563 * list could be extended in the future).
2564 * * BTF description of the map is mandatory.
2565 * * The BPF program can take ONE lock at a time, since taking two
2566 * or more could cause dead locks.
2567 * * Only one **struct bpf_spin_lock** is allowed per map element.
2568 * * When the lock is taken, calls (either BPF to BPF or helpers)
2569 * are not allowed.
2570 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2571 * allowed inside a spinlock-ed region.
2572 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2573 * the lock, on all execution paths, before it returns.
2574 * * The BPF program can access **struct bpf_spin_lock** only via
2575 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2576 * helpers. Loading or storing data into the **struct
2577 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2578 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2579 * of the map value must be a struct and have **struct
2580 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2581 * Nested lock inside another struct is not allowed.
2582 * * The **struct bpf_spin_lock** *lock* field in a map value must
2583 * be aligned on a multiple of 4 bytes in that value.
2584 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2585 * the **bpf_spin_lock** field to user space.
2586 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2587 * a BPF program, do not update the **bpf_spin_lock** field.
2588 * * **bpf_spin_lock** cannot be on the stack or inside a
2589 * networking packet (it can only be inside of a map values).
2590 * * **bpf_spin_lock** is available to root only.
2591 * * Tracing programs and socket filter programs cannot use
2592 * **bpf_spin_lock**\ () due to insufficient preemption checks
2593 * (but this may change in the future).
2594 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2595 * Return
2596 * 0
2597 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002598 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
Quentin Monnet0eb09782019-03-14 12:38:40 +00002599 * Description
2600 * Release the *lock* previously locked by a call to
2601 * **bpf_spin_lock**\ (\ *lock*\ ).
2602 * Return
2603 * 0
2604 *
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002605 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2606 * Description
2607 * This helper gets a **struct bpf_sock** pointer such
Quentin Monnet62369db2019-03-14 12:38:39 +00002608 * that all the fields in this **bpf_sock** can be accessed.
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002609 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002610 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002611 * case of failure.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002612 *
2613 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2614 * Description
2615 * This helper gets a **struct bpf_tcp_sock** pointer from a
2616 * **struct bpf_sock** pointer.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002617 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002618 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002619 * case of failure.
brakmof7c917b2019-03-01 12:38:46 -08002620 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002621 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
Quentin Monnet62369db2019-03-14 12:38:39 +00002622 * Description
2623 * Set ECN (Explicit Congestion Notification) field of IP header
2624 * to **CE** (Congestion Encountered) if current value is **ECT**
2625 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2626 * and IPv4.
2627 * Return
2628 * 1 if the **CE** flag is set (either by the current helper call
2629 * or because it was already present), 0 if it is not set.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002630 *
2631 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2632 * Description
Quentin Monnet62369db2019-03-14 12:38:39 +00002633 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2634 * **bpf_sk_release**\ () is unnecessary and not allowed.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002635 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002636 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002637 * case of failure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002638 *
2639 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2640 * Description
2641 * Look for TCP socket matching *tuple*, optionally in a child
2642 * network namespace *netns*. The return value must be checked,
2643 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2644 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002645 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2646 * that it also returns timewait or request sockets. Use
2647 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2648 * full structure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002649 *
2650 * This helper is available only if the kernel was compiled with
2651 * **CONFIG_NET** configuration option.
2652 * Return
2653 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2654 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002655 * result is from *reuse*\ **->socks**\ [] using the hash of the
2656 * tuple.
Lorenz Bauer39904082019-03-22 09:54:02 +08002657 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002658 * long bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
Lorenz Bauer39904082019-03-22 09:54:02 +08002659 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002660 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2661 * the listening socket in *sk*.
Lorenz Bauer39904082019-03-22 09:54:02 +08002662 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002663 * *iph* points to the start of the IPv4 or IPv6 header, while
2664 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2665 * **sizeof**\ (**struct ip6hdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002666 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002667 * *th* points to the start of the TCP header, while *th_len*
2668 * contains **sizeof**\ (**struct tcphdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002669 * Return
Quentin Monnet80867c52019-05-10 15:51:24 +01002670 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2671 * error otherwise.
Andrey Ignatov808649f2019-02-27 13:28:48 -08002672 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002673 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
Andrey Ignatov808649f2019-02-27 13:28:48 -08002674 * Description
2675 * Get name of sysctl in /proc/sys/ and copy it into provided by
2676 * program buffer *buf* of size *buf_len*.
2677 *
2678 * The buffer is always NUL terminated, unless it's zero-sized.
2679 *
2680 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2681 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2682 * only (e.g. "tcp_mem").
2683 * Return
2684 * Number of character copied (not including the trailing NUL).
2685 *
2686 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2687 * truncated name in this case).
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002688 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002689 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002690 * Description
2691 * Get current value of sysctl as it is presented in /proc/sys
2692 * (incl. newline, etc), and copy it as a string into provided
2693 * by program buffer *buf* of size *buf_len*.
2694 *
2695 * The whole value is copied, no matter what file position user
2696 * space issued e.g. sys_read at.
2697 *
2698 * The buffer is always NUL terminated, unless it's zero-sized.
2699 * Return
2700 * Number of character copied (not including the trailing NUL).
2701 *
2702 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2703 * truncated name in this case).
2704 *
2705 * **-EINVAL** if current value was unavailable, e.g. because
2706 * sysctl is uninitialized and read returns -EIO for it.
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002707 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002708 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002709 * Description
2710 * Get new value being written by user space to sysctl (before
2711 * the actual write happens) and copy it as a string into
2712 * provided by program buffer *buf* of size *buf_len*.
2713 *
2714 * User space may write new value at file position > 0.
2715 *
2716 * The buffer is always NUL terminated, unless it's zero-sized.
2717 * Return
2718 * Number of character copied (not including the trailing NUL).
2719 *
2720 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2721 * truncated name in this case).
2722 *
2723 * **-EINVAL** if sysctl is being read.
2724 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002725 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002726 * Description
2727 * Override new value being written by user space to sysctl with
2728 * value provided by program in buffer *buf* of size *buf_len*.
2729 *
2730 * *buf* should contain a string in same form as provided by user
2731 * space on sysctl write.
2732 *
2733 * User space may write new value at file position > 0. To override
2734 * the whole sysctl value file position should be set to zero.
2735 * Return
2736 * 0 on success.
2737 *
2738 * **-E2BIG** if the *buf_len* is too big.
2739 *
2740 * **-EINVAL** if sysctl is being read.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002741 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002742 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002743 * Description
2744 * Convert the initial part of the string from buffer *buf* of
2745 * size *buf_len* to a long integer according to the given base
2746 * and save the result in *res*.
2747 *
2748 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002749 * (as determined by **isspace**\ (3)) followed by a single
2750 * optional '**-**' sign.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002751 *
2752 * Five least significant bits of *flags* encode base, other bits
2753 * are currently unused.
2754 *
2755 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002756 * similar to user space **strtol**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002757 * Return
2758 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002759 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002760 *
2761 * **-EINVAL** if no valid digits were found or unsupported base
2762 * was provided.
2763 *
2764 * **-ERANGE** if resulting value was out of range.
2765 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002766 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002767 * Description
2768 * Convert the initial part of the string from buffer *buf* of
2769 * size *buf_len* to an unsigned long integer according to the
2770 * given base and save the result in *res*.
2771 *
2772 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002773 * (as determined by **isspace**\ (3)).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002774 *
2775 * Five least significant bits of *flags* encode base, other bits
2776 * are currently unused.
2777 *
2778 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002779 * similar to user space **strtoul**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002780 * Return
2781 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002782 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002783 *
2784 * **-EINVAL** if no valid digits were found or unsupported base
2785 * was provided.
2786 *
2787 * **-ERANGE** if resulting value was out of range.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002788 *
2789 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2790 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002791 * Get a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002792 *
2793 * Logically, it could be thought of getting the value from
2794 * a *map* with *sk* as the **key**. From this
2795 * perspective, the usage is not much different from
Quentin Monnet80867c52019-05-10 15:51:24 +01002796 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2797 * helper enforces the key must be a full socket and the map must
2798 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002799 *
2800 * Underneath, the value is stored locally at *sk* instead of
Quentin Monnet80867c52019-05-10 15:51:24 +01002801 * the *map*. The *map* is used as the bpf-local-storage
2802 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2803 * searched against all bpf-local-storages residing at *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002804 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002805 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002806 * used such that a new bpf-local-storage will be
2807 * created if one does not exist. *value* can be used
Quentin Monnet80867c52019-05-10 15:51:24 +01002808 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002809 * the initial value of a bpf-local-storage. If *value* is
Quentin Monnet80867c52019-05-10 15:51:24 +01002810 * **NULL**, the new bpf-local-storage will be zero initialized.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002811 * Return
2812 * A bpf-local-storage pointer is returned on success.
2813 *
2814 * **NULL** if not found or there was an error in adding
2815 * a new bpf-local-storage.
2816 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002817 * long bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002818 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002819 * Delete a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002820 * Return
2821 * 0 on success.
2822 *
2823 * **-ENOENT** if the bpf-local-storage cannot be found.
Yonghong Song8b401f92019-05-23 14:47:45 -07002824 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002825 * long bpf_send_signal(u32 sig)
Yonghong Song8b401f92019-05-23 14:47:45 -07002826 * Description
Yonghong Song84829412020-01-14 19:50:02 -08002827 * Send signal *sig* to the process of the current task.
2828 * The signal may be delivered to any of this process's threads.
Yonghong Song8b401f92019-05-23 14:47:45 -07002829 * Return
2830 * 0 on success or successfully queued.
2831 *
2832 * **-EBUSY** if work queue under nmi is full.
2833 *
2834 * **-EINVAL** if *sig* is invalid.
2835 *
2836 * **-EPERM** if no permission to send the *sig*.
2837 *
2838 * **-EAGAIN** if bpf program can try again.
Petar Penkov70d66242019-07-29 09:59:15 -07002839 *
2840 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2841 * Description
2842 * Try to issue a SYN cookie for the packet with corresponding
2843 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2844 *
2845 * *iph* points to the start of the IPv4 or IPv6 header, while
2846 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2847 * **sizeof**\ (**struct ip6hdr**).
2848 *
2849 * *th* points to the start of the TCP header, while *th_len*
2850 * contains the length of the TCP header.
Petar Penkov70d66242019-07-29 09:59:15 -07002851 * Return
2852 * On success, lower 32 bits hold the generated SYN cookie in
2853 * followed by 16 bits which hold the MSS value for that cookie,
2854 * and the top 16 bits are unused.
2855 *
2856 * On failure, the returned value is one of the following:
2857 *
2858 * **-EINVAL** SYN cookie cannot be issued due to error
2859 *
2860 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2861 *
2862 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2863 *
2864 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
Alexei Starovoitova7658e12019-10-15 20:25:04 -07002865 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002866 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
Alexei Starovoitova7658e12019-10-15 20:25:04 -07002867 * Description
2868 * Write raw *data* blob into a special BPF perf event held by
2869 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2870 * event must have the following attributes: **PERF_SAMPLE_RAW**
2871 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2872 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2873 *
2874 * The *flags* are used to indicate the index in *map* for which
2875 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2876 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2877 * to indicate that the index of the current CPU core should be
2878 * used.
2879 *
2880 * The value to write, of *size*, is passed through eBPF stack and
2881 * pointed by *data*.
2882 *
2883 * *ctx* is a pointer to in-kernel struct sk_buff.
2884 *
2885 * This helper is similar to **bpf_perf_event_output**\ () but
2886 * restricted to raw_tracepoint bpf programs.
2887 * Return
2888 * 0 on success, or a negative error in case of failure.
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002889 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002890 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002891 * Description
2892 * Safely attempt to read *size* bytes from user space address
2893 * *unsafe_ptr* and store the data in *dst*.
2894 * Return
2895 * 0 on success, or a negative error in case of failure.
2896 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002897 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002898 * Description
2899 * Safely attempt to read *size* bytes from kernel space address
2900 * *unsafe_ptr* and store the data in *dst*.
2901 * Return
2902 * 0 on success, or a negative error in case of failure.
2903 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002904 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002905 * Description
2906 * Copy a NUL terminated string from an unsafe user address
2907 * *unsafe_ptr* to *dst*. The *size* should include the
2908 * terminating NUL byte. In case the string length is smaller than
2909 * *size*, the target is not padded with further NUL bytes. If the
2910 * string length is larger than *size*, just *size*-1 bytes are
2911 * copied and the last byte is set to NUL.
2912 *
2913 * On success, the length of the copied string is returned. This
2914 * makes this helper useful in tracing programs for reading
2915 * strings, and more importantly to get its length at runtime. See
2916 * the following snippet:
2917 *
2918 * ::
2919 *
2920 * SEC("kprobe/sys_open")
2921 * void bpf_sys_open(struct pt_regs *ctx)
2922 * {
2923 * char buf[PATHLEN]; // PATHLEN is defined to 256
2924 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2925 * ctx->di);
2926 *
2927 * // Consume buf, for example push it to
2928 * // userspace via bpf_perf_event_output(); we
2929 * // can use res (the string length) as event
2930 * // size, after checking its boundaries.
2931 * }
2932 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01002933 * In comparison, using **bpf_probe_read_user**\ () helper here
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002934 * instead to read the string would require to estimate the length
2935 * at compile time, and would often result in copying more memory
2936 * than necessary.
2937 *
2938 * Another useful use case is when parsing individual process
2939 * arguments or individual environment variables navigating
2940 * *current*\ **->mm->arg_start** and *current*\
2941 * **->mm->env_start**: using this helper and the return value,
2942 * one can quickly iterate at the right offset of the memory area.
2943 * Return
2944 * On success, the strictly positive length of the string,
2945 * including the trailing NUL character. On error, a negative
2946 * value.
2947 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002948 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002949 * Description
2950 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
Quentin Monnetab8d7802020-05-11 17:15:35 +01002951 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002952 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01002953 * On success, the strictly positive length of the string, including
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002954 * the trailing NUL character. On error, a negative value.
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002955 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002956 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002957 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01002958 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002959 * *rcv_nxt* is the ack_seq to be sent out.
2960 * Return
2961 * 0 on success, or a negative error in case of failure.
2962 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002963 * long bpf_send_signal_thread(u32 sig)
Yonghong Song84829412020-01-14 19:50:02 -08002964 * Description
2965 * Send signal *sig* to the thread corresponding to the current task.
2966 * Return
2967 * 0 on success or successfully queued.
2968 *
2969 * **-EBUSY** if work queue under nmi is full.
2970 *
2971 * **-EINVAL** if *sig* is invalid.
2972 *
2973 * **-EPERM** if no permission to send the *sig*.
2974 *
2975 * **-EAGAIN** if bpf program can try again.
Martin KaFai Lau5576b992020-01-22 15:36:46 -08002976 *
2977 * u64 bpf_jiffies64(void)
2978 * Description
2979 * Obtain the 64bit jiffies
2980 * Return
2981 * The 64 bit jiffies
Daniel Xufff7b642020-02-17 19:04:31 -08002982 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07002983 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
Daniel Xufff7b642020-02-17 19:04:31 -08002984 * Description
2985 * For an eBPF program attached to a perf event, retrieve the
Quentin Monnetab8d7802020-05-11 17:15:35 +01002986 * branch records (**struct perf_branch_entry**) associated to *ctx*
2987 * and store it in the buffer pointed by *buf* up to size
Daniel Xufff7b642020-02-17 19:04:31 -08002988 * *size* bytes.
2989 * Return
2990 * On success, number of bytes written to *buf*. On error, a
2991 * negative value.
2992 *
2993 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
Quentin Monnetab8d7802020-05-11 17:15:35 +01002994 * instead return the number of bytes required to store all the
Daniel Xufff7b642020-02-17 19:04:31 -08002995 * branch entries. If this flag is set, *buf* may be NULL.
2996 *
2997 * **-EINVAL** if arguments invalid or **size** not a multiple
Quentin Monnetab8d7802020-05-11 17:15:35 +01002998 * of **sizeof**\ (**struct perf_branch_entry**\ ).
Daniel Xufff7b642020-02-17 19:04:31 -08002999 *
3000 * **-ENOENT** if architecture does not support branch records.
Carlos Neirab4490c52020-03-04 17:41:56 -03003001 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003002 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
Carlos Neirab4490c52020-03-04 17:41:56 -03003003 * Description
3004 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3005 * *namespace* will be returned in *nsdata*.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003006 * Return
3007 * 0 on success, or one of the following in case of failure:
Carlos Neirab4490c52020-03-04 17:41:56 -03003008 *
3009 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3010 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3011 *
3012 * **-ENOENT** if pidns does not exists for the current task.
3013 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003014 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
Eelco Chaudrond831ee82020-03-06 08:59:23 +00003015 * Description
3016 * Write raw *data* blob into a special BPF perf event held by
3017 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3018 * event must have the following attributes: **PERF_SAMPLE_RAW**
3019 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3020 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3021 *
3022 * The *flags* are used to indicate the index in *map* for which
3023 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3024 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3025 * to indicate that the index of the current CPU core should be
3026 * used.
3027 *
3028 * The value to write, of *size*, is passed through eBPF stack and
3029 * pointed by *data*.
3030 *
3031 * *ctx* is a pointer to in-kernel struct xdp_buff.
3032 *
3033 * This helper is similar to **bpf_perf_eventoutput**\ () but
3034 * restricted to raw_tracepoint bpf programs.
3035 * Return
3036 * 0 on success, or a negative error in case of failure.
Daniel Borkmannf3189032020-03-27 16:58:52 +01003037 *
3038 * u64 bpf_get_netns_cookie(void *ctx)
3039 * Description
3040 * Retrieve the cookie (generated by the kernel) of the network
3041 * namespace the input *ctx* is associated with. The network
3042 * namespace cookie remains stable for its lifetime and provides
3043 * a global identifier that can be assumed unique. If *ctx* is
3044 * NULL, then the helper returns the cookie for the initial
3045 * network namespace. The cookie itself is very similar to that
Quentin Monnetab8d7802020-05-11 17:15:35 +01003046 * of **bpf_get_socket_cookie**\ () helper, but for network
3047 * namespaces instead of sockets.
Daniel Borkmannf3189032020-03-27 16:58:52 +01003048 * Return
3049 * A 8-byte long opaque number.
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003050 *
3051 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3052 * Description
3053 * Return id of cgroup v2 that is ancestor of the cgroup associated
3054 * with the current task at the *ancestor_level*. The root cgroup
3055 * is at *ancestor_level* zero and each step down the hierarchy
3056 * increments the level. If *ancestor_level* == level of cgroup
3057 * associated with the current task, then return value will be the
3058 * same as that of **bpf_get_current_cgroup_id**\ ().
3059 *
3060 * The helper is useful to implement policies based on cgroups
3061 * that are upper in hierarchy than immediate cgroup associated
3062 * with the current task.
3063 *
3064 * The format of returned id and helper limitations are same as in
3065 * **bpf_get_current_cgroup_id**\ ().
3066 * Return
3067 * The id is returned or 0 in case the id could not be retrieved.
Joe Stringercf7fbe62020-03-29 15:53:38 -07003068 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003069 * long bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
Joe Stringercf7fbe62020-03-29 15:53:38 -07003070 * Description
3071 * Assign the *sk* to the *skb*. When combined with appropriate
3072 * routing configuration to receive the packet towards the socket,
3073 * will cause *skb* to be delivered to the specified socket.
3074 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3075 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3076 * interfere with successful delivery to the socket.
3077 *
3078 * This operation is only valid from TC ingress path.
3079 *
3080 * The *flags* argument must be zero.
3081 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003082 * 0 on success, or a negative error in case of failure:
Joe Stringercf7fbe62020-03-29 15:53:38 -07003083 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003084 * **-EINVAL** if specified *flags* are not supported.
3085 *
3086 * **-ENOENT** if the socket is unavailable for assignment.
3087 *
3088 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3089 *
3090 * **-EOPNOTSUPP** if the operation is not supported, for example
3091 * a call from outside of TC ingress.
3092 *
3093 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3094 * (reuseport).
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003095 *
3096 * u64 bpf_ktime_get_boot_ns(void)
3097 * Description
3098 * Return the time elapsed since system boot, in nanoseconds.
3099 * Does include the time the system was suspended.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003100 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003101 * Return
3102 * Current *ktime*.
Yonghong Song492e6392020-05-09 10:59:14 -07003103 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003104 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
Yonghong Song492e6392020-05-09 10:59:14 -07003105 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01003106 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3107 * out the format string.
Yonghong Song492e6392020-05-09 10:59:14 -07003108 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3109 * the format string itself. The *data* and *data_len* are format string
Quentin Monnetab8d7802020-05-11 17:15:35 +01003110 * arguments. The *data* are a **u64** array and corresponding format string
Yonghong Song492e6392020-05-09 10:59:14 -07003111 * values are stored in the array. For strings and pointers where pointees
3112 * are accessed, only the pointer values are stored in the *data* array.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003113 * The *data_len* is the size of *data* in bytes.
Yonghong Song492e6392020-05-09 10:59:14 -07003114 *
3115 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3116 * Reading kernel memory may fail due to either invalid address or
3117 * valid address but requiring a major memory fault. If reading kernel memory
3118 * fails, the string for **%s** will be an empty string, and the ip
3119 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
Quentin Monnetab8d7802020-05-11 17:15:35 +01003120 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
Yonghong Song492e6392020-05-09 10:59:14 -07003121 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003122 * 0 on success, or a negative error in case of failure:
Yonghong Song492e6392020-05-09 10:59:14 -07003123 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003124 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3125 * by returning 1 from bpf program.
3126 *
3127 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3128 *
3129 * **-E2BIG** if *fmt* contains too many format specifiers.
3130 *
3131 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
Yonghong Song492e6392020-05-09 10:59:14 -07003132 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003133 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
Yonghong Song492e6392020-05-09 10:59:14 -07003134 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01003135 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
Yonghong Song492e6392020-05-09 10:59:14 -07003136 * The *m* represents the seq_file. The *data* and *len* represent the
Quentin Monnetab8d7802020-05-11 17:15:35 +01003137 * data to write in bytes.
Yonghong Song492e6392020-05-09 10:59:14 -07003138 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003139 * 0 on success, or a negative error in case of failure:
Yonghong Song492e6392020-05-09 10:59:14 -07003140 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003141 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
Andrey Ignatovf307fa22020-05-14 13:03:47 -07003142 *
3143 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3144 * Description
3145 * Return the cgroup v2 id of the socket *sk*.
3146 *
3147 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3148 * returned from **bpf_sk_lookup_xxx**\ (),
3149 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3150 * same as in **bpf_skb_cgroup_id**\ ().
3151 *
3152 * This helper is available only if the kernel was compiled with
3153 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3154 * Return
3155 * The id is returned or 0 in case the id could not be retrieved.
3156 *
3157 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3158 * Description
3159 * Return id of cgroup v2 that is ancestor of cgroup associated
3160 * with the *sk* at the *ancestor_level*. The root cgroup is at
3161 * *ancestor_level* zero and each step down the hierarchy
3162 * increments the level. If *ancestor_level* == level of cgroup
3163 * associated with *sk*, then return value will be same as that
3164 * of **bpf_sk_cgroup_id**\ ().
3165 *
3166 * The helper is useful to implement policies based on cgroups
3167 * that are upper in hierarchy than immediate cgroup associated
3168 * with *sk*.
3169 *
3170 * The format of returned id and helper limitations are same as in
3171 * **bpf_sk_cgroup_id**\ ().
3172 * Return
3173 * The id is returned or 0 in case the id could not be retrieved.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003174 *
Andrii Nakryikob0659d82020-06-15 14:49:26 -07003175 * int bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003176 * Description
3177 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003178 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3179 * of new data availability is sent.
3180 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3181 * of new data availability is sent unconditionally.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003182 * Return
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003183 * 0 on success, or a negative error in case of failure.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003184 *
3185 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3186 * Description
3187 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3188 * Return
3189 * Valid pointer with *size* bytes of memory available; NULL,
3190 * otherwise.
3191 *
3192 * void bpf_ringbuf_submit(void *data, u64 flags)
3193 * Description
3194 * Submit reserved ring buffer sample, pointed to by *data*.
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003195 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3196 * of new data availability is sent.
3197 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3198 * of new data availability is sent unconditionally.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003199 * Return
3200 * Nothing. Always succeeds.
3201 *
3202 * void bpf_ringbuf_discard(void *data, u64 flags)
3203 * Description
3204 * Discard reserved ring buffer sample, pointed to by *data*.
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003205 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3206 * of new data availability is sent.
3207 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3208 * of new data availability is sent unconditionally.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003209 * Return
3210 * Nothing. Always succeeds.
3211 *
3212 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3213 * Description
3214 * Query various characteristics of provided ring buffer. What
3215 * exactly is queries is determined by *flags*:
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003216 *
3217 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3218 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3219 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3220 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3221 *
3222 * Data returned is just a momentary snapshot of actual values
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003223 * and could be inaccurate, so this facility should be used to
3224 * power heuristics and for reporting, not to make 100% correct
3225 * calculation.
3226 * Return
Quentin Monnetbcc7f552020-06-23 16:39:35 +01003227 * Requested value, or 0, if *flags* are not recognized.
Daniel Borkmann7cdec542020-06-02 16:58:33 +02003228 *
Andrii Nakryikobdb7b792020-06-22 20:22:21 -07003229 * long bpf_csum_level(struct sk_buff *skb, u64 level)
Daniel Borkmann7cdec542020-06-02 16:58:33 +02003230 * Description
3231 * Change the skbs checksum level by one layer up or down, or
3232 * reset it entirely to none in order to have the stack perform
3233 * checksum validation. The level is applicable to the following
3234 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3235 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3236 * through **bpf_skb_adjust_room**\ () helper with passing in
3237 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3238 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3239 * the UDP header is removed. Similarly, an encap of the latter
3240 * into the former could be accompanied by a helper call to
3241 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3242 * skb is still intended to be processed in higher layers of the
3243 * stack instead of just egressing at tc.
3244 *
3245 * There are three supported level settings at this time:
3246 *
3247 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3248 * with CHECKSUM_UNNECESSARY.
3249 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3250 * with CHECKSUM_UNNECESSARY.
3251 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3252 * sets CHECKSUM_NONE to force checksum validation by the stack.
3253 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3254 * skb->csum_level.
3255 * Return
3256 * 0 on success, or a negative error in case of failure. In the
3257 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3258 * is returned or the error code -EACCES in case the skb is not
3259 * subject to CHECKSUM_UNNECESSARY.
Yonghong Songaf7ec132020-06-23 16:08:09 -07003260 *
3261 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3262 * Description
3263 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3264 * Return
3265 * *sk* if casting is valid, or NULL otherwise.
Yonghong Song478cfbd2020-06-23 16:08:11 -07003266 *
3267 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3268 * Description
3269 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3270 * Return
3271 * *sk* if casting is valid, or NULL otherwise.
3272 *
3273 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3274 * Description
3275 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3276 * Return
3277 * *sk* if casting is valid, or NULL otherwise.
3278 *
3279 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3280 * Description
3281 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3282 * Return
3283 * *sk* if casting is valid, or NULL otherwise.
Yonghong Song0d4fad32020-06-23 16:08:15 -07003284 *
3285 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3286 * Description
3287 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3288 * Return
3289 * *sk* if casting is valid, or NULL otherwise.
Song Liufa28dcb2020-06-29 23:28:44 -07003290 *
3291 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3292 * Description
3293 * Return a user or a kernel stack in bpf program provided buffer.
3294 * To achieve this, the helper needs *task*, which is a valid
3295 * pointer to struct task_struct. To store the stacktrace, the
3296 * bpf program provides *buf* with a nonnegative *size*.
3297 *
3298 * The last argument, *flags*, holds the number of stack frames to
3299 * skip (from 0 to 255), masked with
3300 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3301 * the following flags:
3302 *
3303 * **BPF_F_USER_STACK**
3304 * Collect a user space stack instead of a kernel stack.
3305 * **BPF_F_USER_BUILD_ID**
3306 * Collect buildid+offset instead of ips for user stack,
3307 * only valid if **BPF_F_USER_STACK** is also specified.
3308 *
3309 * **bpf_get_task_stack**\ () can collect up to
3310 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3311 * to sufficient large buffer size. Note that
3312 * this limit can be controlled with the **sysctl** program, and
3313 * that it should be manually increased in order to profile long
3314 * user stacks (such as stacks for Java programs). To do so, use:
3315 *
3316 * ::
3317 *
3318 * # sysctl kernel.perf_event_max_stack=<new value>
3319 * Return
3320 * A non-negative value equal to or less than *size* on success,
3321 * or a negative error in case of failure.
3322 *
Thomas Grafebb676d2016-10-27 11:23:51 +02003323 */
3324#define __BPF_FUNC_MAPPER(FN) \
3325 FN(unspec), \
3326 FN(map_lookup_elem), \
3327 FN(map_update_elem), \
3328 FN(map_delete_elem), \
3329 FN(probe_read), \
3330 FN(ktime_get_ns), \
3331 FN(trace_printk), \
3332 FN(get_prandom_u32), \
3333 FN(get_smp_processor_id), \
3334 FN(skb_store_bytes), \
3335 FN(l3_csum_replace), \
3336 FN(l4_csum_replace), \
3337 FN(tail_call), \
3338 FN(clone_redirect), \
3339 FN(get_current_pid_tgid), \
3340 FN(get_current_uid_gid), \
3341 FN(get_current_comm), \
3342 FN(get_cgroup_classid), \
3343 FN(skb_vlan_push), \
3344 FN(skb_vlan_pop), \
3345 FN(skb_get_tunnel_key), \
3346 FN(skb_set_tunnel_key), \
3347 FN(perf_event_read), \
3348 FN(redirect), \
3349 FN(get_route_realm), \
3350 FN(perf_event_output), \
3351 FN(skb_load_bytes), \
3352 FN(get_stackid), \
3353 FN(csum_diff), \
3354 FN(skb_get_tunnel_opt), \
3355 FN(skb_set_tunnel_opt), \
3356 FN(skb_change_proto), \
3357 FN(skb_change_type), \
3358 FN(skb_under_cgroup), \
3359 FN(get_hash_recalc), \
3360 FN(get_current_task), \
3361 FN(probe_write_user), \
3362 FN(current_task_under_cgroup), \
3363 FN(skb_change_tail), \
3364 FN(skb_pull_data), \
3365 FN(csum_update), \
3366 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003367 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003368 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +00003369 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -07003370 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -07003371 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +02003372 FN(get_socket_uid), \
Lawrence Brakmo8c4b4c72017-06-30 20:02:46 -07003373 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003374 FN(setsockopt), \
John Fastabend97f91a72017-07-17 09:29:18 -07003375 FN(skb_adjust_room), \
John Fastabend174a79f2017-08-15 22:32:47 -07003376 FN(redirect_map), \
3377 FN(sk_redirect_map), \
3378 FN(sock_map_update), \
Yonghong Song908432c2017-10-05 09:19:20 -07003379 FN(xdp_adjust_meta), \
Yonghong Song4bebdc72017-10-05 09:19:22 -07003380 FN(perf_event_read_value), \
Lawrence Brakmocd86d1f2017-10-20 11:05:40 -07003381 FN(perf_prog_read_value), \
Josef Bacik9802d862017-12-11 11:36:48 -05003382 FN(getsockopt), \
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003383 FN(override_return), \
John Fastabend4f738ad2018-03-18 12:57:10 -07003384 FN(sock_ops_cb_flags_set), \
John Fastabend2a100312018-03-18 12:57:15 -07003385 FN(msg_redirect_map), \
John Fastabend91843d52018-03-18 12:57:20 -07003386 FN(msg_apply_bytes), \
John Fastabend015632b2018-03-18 12:57:25 -07003387 FN(msg_cork_bytes), \
Andrey Ignatovd74bad42018-03-30 15:08:05 -07003388 FN(msg_pull_data), \
Nikita V. Shirokovb32cc5b2018-04-17 21:42:13 -07003389 FN(bind), \
Eyal Birger12bed762018-04-24 17:50:29 +03003390 FN(xdp_adjust_tail), \
Yonghong Songc195651e2018-04-28 22:28:08 -07003391 FN(skb_get_xfrm_state), \
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003392 FN(get_stack), \
David Ahern87f5fc72018-05-09 20:34:26 -07003393 FN(skb_load_bytes_relative), \
John Fastabend81110382018-05-14 10:00:17 -07003394 FN(fib_lookup), \
3395 FN(sock_hash_update), \
3396 FN(msg_redirect_hash), \
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003397 FN(sk_redirect_hash), \
3398 FN(lwt_push_encap), \
3399 FN(lwt_seg6_store_bytes), \
3400 FN(lwt_seg6_adjust_srh), \
Sean Youngf4364dc2018-05-27 12:24:09 +01003401 FN(lwt_seg6_action), \
3402 FN(rc_repeat), \
Daniel Borkmanncb20b082018-06-02 23:06:36 +02003403 FN(rc_keydown), \
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07003404 FN(skb_cgroup_id), \
Roman Gushchincd339432018-08-02 14:27:24 -07003405 FN(get_current_cgroup_id), \
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003406 FN(get_local_storage), \
Andrey Ignatov77236282018-08-12 10:49:27 -07003407 FN(sk_select_reuseport), \
Joe Stringer6acc9b42018-10-02 13:35:36 -07003408 FN(skb_ancestor_cgroup_id), \
3409 FN(sk_lookup_tcp), \
3410 FN(sk_lookup_udp), \
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +02003411 FN(sk_release), \
3412 FN(map_push_elem), \
3413 FN(map_pop_elem), \
John Fastabend6fff6072018-10-19 19:56:49 -07003414 FN(map_peek_elem), \
John Fastabend7246d8e2018-11-26 14:16:17 -08003415 FN(msg_push_data), \
Sean Young01d32402018-12-06 13:01:03 +00003416 FN(msg_pop_data), \
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08003417 FN(rc_pointer_rel), \
3418 FN(spin_lock), \
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003419 FN(spin_unlock), \
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003420 FN(sk_fullsock), \
brakmof7c917b2019-03-01 12:38:46 -08003421 FN(tcp_sock), \
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07003422 FN(skb_ecn_set_ce), \
Lorenz Baueredbf8c02019-03-22 09:54:01 +08003423 FN(get_listener_sock), \
Lorenz Bauer39904082019-03-22 09:54:02 +08003424 FN(skc_lookup_tcp), \
Andrey Ignatov808649f2019-02-27 13:28:48 -08003425 FN(tcp_check_syncookie), \
Andrey Ignatov1d11b302019-02-28 19:22:15 -08003426 FN(sysctl_get_name), \
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08003427 FN(sysctl_get_current_value), \
3428 FN(sysctl_get_new_value), \
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07003429 FN(sysctl_set_new_value), \
3430 FN(strtol), \
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003431 FN(strtoul), \
3432 FN(sk_storage_get), \
Yonghong Song8b401f92019-05-23 14:47:45 -07003433 FN(sk_storage_delete), \
Petar Penkov70d66242019-07-29 09:59:15 -07003434 FN(send_signal), \
Alexei Starovoitova7658e12019-10-15 20:25:04 -07003435 FN(tcp_gen_syncookie), \
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01003436 FN(skb_output), \
3437 FN(probe_read_user), \
3438 FN(probe_read_kernel), \
3439 FN(probe_read_user_str), \
Martin KaFai Lau206057f2020-01-08 16:45:51 -08003440 FN(probe_read_kernel_str), \
Yonghong Song84829412020-01-14 19:50:02 -08003441 FN(tcp_send_ack), \
Martin KaFai Lau5576b992020-01-22 15:36:46 -08003442 FN(send_signal_thread), \
Daniel Xufff7b642020-02-17 19:04:31 -08003443 FN(jiffies64), \
Carlos Neirab4490c52020-03-04 17:41:56 -03003444 FN(read_branch_records), \
Eelco Chaudrond831ee82020-03-06 08:59:23 +00003445 FN(get_ns_current_pid_tgid), \
Daniel Borkmannf3189032020-03-27 16:58:52 +01003446 FN(xdp_output), \
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003447 FN(get_netns_cookie), \
Joe Stringercf7fbe62020-03-29 15:53:38 -07003448 FN(get_current_ancestor_cgroup_id), \
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003449 FN(sk_assign), \
Yonghong Song492e6392020-05-09 10:59:14 -07003450 FN(ktime_get_boot_ns), \
3451 FN(seq_printf), \
Andrey Ignatovf307fa22020-05-14 13:03:47 -07003452 FN(seq_write), \
3453 FN(sk_cgroup_id), \
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003454 FN(sk_ancestor_cgroup_id), \
3455 FN(ringbuf_output), \
3456 FN(ringbuf_reserve), \
3457 FN(ringbuf_submit), \
3458 FN(ringbuf_discard), \
Daniel Borkmann7cdec542020-06-02 16:58:33 +02003459 FN(ringbuf_query), \
Yonghong Songaf7ec132020-06-23 16:08:09 -07003460 FN(csum_level), \
Yonghong Song478cfbd2020-06-23 16:08:11 -07003461 FN(skc_to_tcp6_sock), \
3462 FN(skc_to_tcp_sock), \
3463 FN(skc_to_tcp_timewait_sock), \
Yonghong Song0d4fad32020-06-23 16:08:15 -07003464 FN(skc_to_tcp_request_sock), \
Song Liufa28dcb2020-06-29 23:28:44 -07003465 FN(skc_to_udp6_sock), \
3466 FN(get_task_stack), \
3467 /* */
Thomas Grafebb676d2016-10-27 11:23:51 +02003468
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003469/* integer value in 'imm' field of BPF_CALL instruction selects which helper
3470 * function eBPF program intends to call
3471 */
Thomas Grafebb676d2016-10-27 11:23:51 +02003472#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003473enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +02003474 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003475 __BPF_FUNC_MAX_ID,
3476};
Thomas Grafebb676d2016-10-27 11:23:51 +02003477#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003478
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003479/* All flags used by eBPF helper functions, placed here. */
3480
3481/* BPF_FUNC_skb_store_bytes flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003482enum {
3483 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3484 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3485};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003486
3487/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3488 * First 4 bits are for passing the header field size.
3489 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003490enum {
3491 BPF_F_HDR_FIELD_MASK = 0xfULL,
3492};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003493
3494/* BPF_FUNC_l4_csum_replace flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003495enum {
3496 BPF_F_PSEUDO_HDR = (1ULL << 4),
3497 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3498 BPF_F_MARK_ENFORCE = (1ULL << 6),
3499};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003500
3501/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003502enum {
3503 BPF_F_INGRESS = (1ULL << 0),
3504};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003505
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003506/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003507enum {
3508 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3509};
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003510
Yonghong Songc195651e2018-04-28 22:28:08 -07003511/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003512enum {
3513 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3514 BPF_F_USER_STACK = (1ULL << 8),
Yonghong Songc195651e2018-04-28 22:28:08 -07003515/* flags used by BPF_FUNC_get_stackid only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003516 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3517 BPF_F_REUSE_STACKID = (1ULL << 10),
Yonghong Songc195651e2018-04-28 22:28:08 -07003518/* flags used by BPF_FUNC_get_stack only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003519 BPF_F_USER_BUILD_ID = (1ULL << 11),
3520};
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08003521
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003522/* BPF_FUNC_skb_set_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003523enum {
3524 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3525 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3526 BPF_F_SEQ_NUMBER = (1ULL << 3),
3527};
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003528
Yonghong Song908432c2017-10-05 09:19:20 -07003529/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3530 * BPF_FUNC_perf_event_read_value flags.
3531 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003532enum {
3533 BPF_F_INDEX_MASK = 0xffffffffULL,
3534 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
Daniel Borkmann555c8a82016-07-14 18:08:05 +02003535/* BPF_FUNC_perf_event_output for sk_buff input context. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003536 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3537};
Daniel Borkmann1e337592016-04-18 21:01:23 +02003538
Joe Stringerf71c6142018-11-30 15:32:20 -08003539/* Current network namespace */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003540enum {
3541 BPF_F_CURRENT_NETNS = (-1L),
3542};
Joe Stringerf71c6142018-11-30 15:32:20 -08003543
Daniel Borkmann7cdec542020-06-02 16:58:33 +02003544/* BPF_FUNC_csum_level level values. */
3545enum {
3546 BPF_CSUM_LEVEL_QUERY,
3547 BPF_CSUM_LEVEL_INC,
3548 BPF_CSUM_LEVEL_DEC,
3549 BPF_CSUM_LEVEL_RESET,
3550};
3551
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003552/* BPF_FUNC_skb_adjust_room flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003553enum {
3554 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3555 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3556 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3557 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3558 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
Daniel Borkmann836e66c2020-06-02 16:58:32 +02003559 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003560};
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003561
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003562enum {
3563 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3564 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3565};
Alan Maguire58dfc902019-04-09 15:06:41 +01003566
Alan Maguirebfb35c22019-04-12 12:27:34 +01003567#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
Alan Maguire58dfc902019-04-09 15:06:41 +01003568 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3569 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Willem de Bruijn868d5232019-03-22 14:32:56 -04003570
Andrey Ignatov808649f2019-02-27 13:28:48 -08003571/* BPF_FUNC_sysctl_get_name flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003572enum {
3573 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3574};
Andrey Ignatov808649f2019-02-27 13:28:48 -08003575
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003576/* BPF_FUNC_sk_storage_get flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003577enum {
3578 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3579};
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003580
Daniel Xufff7b642020-02-17 19:04:31 -08003581/* BPF_FUNC_read_branch_records flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003582enum {
3583 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3584};
Daniel Xufff7b642020-02-17 19:04:31 -08003585
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003586/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3587 * BPF_FUNC_bpf_ringbuf_output flags.
3588 */
3589enum {
3590 BPF_RB_NO_WAKEUP = (1ULL << 0),
3591 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3592};
3593
3594/* BPF_FUNC_bpf_ringbuf_query flags */
3595enum {
3596 BPF_RB_AVAIL_DATA = 0,
3597 BPF_RB_RING_SIZE = 1,
3598 BPF_RB_CONS_POS = 2,
3599 BPF_RB_PROD_POS = 3,
3600};
3601
3602/* BPF ring buffer constants */
3603enum {
3604 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3605 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3606 BPF_RINGBUF_HDR_SZ = 8,
3607};
3608
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003609/* Mode for BPF_FUNC_skb_adjust_room helper. */
3610enum bpf_adj_room_mode {
3611 BPF_ADJ_ROOM_NET,
Willem de Bruijn14aa3192019-03-22 14:32:54 -04003612 BPF_ADJ_ROOM_MAC,
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003613};
3614
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003615/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3616enum bpf_hdr_start_off {
3617 BPF_HDR_START_MAC,
3618 BPF_HDR_START_NET,
3619};
3620
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003621/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3622enum bpf_lwt_encap_mode {
3623 BPF_LWT_ENCAP_SEG6,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003624 BPF_LWT_ENCAP_SEG6_INLINE,
3625 BPF_LWT_ENCAP_IP,
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003626};
3627
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003628#define __bpf_md_ptr(type, name) \
3629union { \
3630 type name; \
3631 __u64 :64; \
3632} __attribute__((aligned(8)))
3633
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003634/* user accessible mirror of in-kernel sk_buff.
3635 * new fields can only be added to the end of this structure
3636 */
3637struct __sk_buff {
3638 __u32 len;
3639 __u32 pkt_type;
3640 __u32 mark;
3641 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -07003642 __u32 protocol;
3643 __u32 vlan_present;
3644 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +01003645 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +02003646 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -07003647 __u32 ingress_ifindex;
3648 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -07003649 __u32 tc_index;
3650 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +02003651 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -07003652 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -07003653 __u32 data;
3654 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +02003655 __u32 napi_id;
John Fastabend8a31db52017-08-15 22:33:09 -07003656
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003657 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
John Fastabend8a31db52017-08-15 22:33:09 -07003658 __u32 family;
3659 __u32 remote_ip4; /* Stored in network byte order */
3660 __u32 local_ip4; /* Stored in network byte order */
3661 __u32 remote_ip6[4]; /* Stored in network byte order */
3662 __u32 local_ip6[4]; /* Stored in network byte order */
3663 __u32 remote_port; /* Stored in network byte order */
3664 __u32 local_port; /* stored in host byte order */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003665 /* ... here. */
3666
3667 __u32 data_meta;
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003668 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
Vlad Dumitrescuf11216b2018-11-22 14:39:16 -05003669 __u64 tstamp;
Petar Penkove3da08d2018-12-02 20:18:19 -05003670 __u32 wire_len;
Eric Dumazetd9ff2862019-01-23 09:22:27 -08003671 __u32 gso_segs;
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003672 __bpf_md_ptr(struct bpf_sock *, sk);
Willem de Bruijncf620892020-03-03 15:05:01 -05003673 __u32 gso_size;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003674};
3675
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003676struct bpf_tunnel_key {
3677 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003678 union {
3679 __u32 remote_ipv4;
3680 __u32 remote_ipv6[4];
3681 };
3682 __u8 tunnel_tos;
3683 __u8 tunnel_ttl;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003684 __u16 tunnel_ext; /* Padding, future use. */
Daniel Borkmann4018ab12016-03-09 03:00:05 +01003685 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003686};
3687
Eyal Birger12bed762018-04-24 17:50:29 +03003688/* user accessible mirror of in-kernel xfrm_state.
3689 * new fields can only be added to the end of this structure
3690 */
3691struct bpf_xfrm_state {
3692 __u32 reqid;
3693 __u32 spi; /* Stored in network byte order */
3694 __u16 family;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003695 __u16 ext; /* Padding, future use. */
Eyal Birger12bed762018-04-24 17:50:29 +03003696 union {
3697 __u32 remote_ipv4; /* Stored in network byte order */
3698 __u32 remote_ipv6[4]; /* Stored in network byte order */
3699 };
3700};
3701
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003702/* Generic BPF return codes which all BPF program types may support.
3703 * The values are binary compatible with their TC_ACT_* counter-part to
3704 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3705 * programs.
3706 *
3707 * XDP is handled seprately, see XDP_*.
3708 */
3709enum bpf_ret_code {
3710 BPF_OK = 0,
3711 /* 1 reserved */
3712 BPF_DROP = 2,
3713 /* 3-6 reserved */
3714 BPF_REDIRECT = 7,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003715 /* >127 are reserved for prog type specific return codes.
3716 *
3717 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3718 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3719 * changed and should be routed based on its new L3 header.
3720 * (This is an L3 redirect, as opposed to L2 redirect
3721 * represented by BPF_REDIRECT above).
3722 */
3723 BPF_LWT_REROUTE = 128,
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003724};
3725
David Ahern610236582016-12-01 08:48:04 -08003726struct bpf_sock {
3727 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -08003728 __u32 family;
3729 __u32 type;
3730 __u32 protocol;
David Ahern482dca92017-08-31 15:05:44 -07003731 __u32 mark;
3732 __u32 priority;
Martin KaFai Lauaa65d692019-02-09 23:22:21 -08003733 /* IP address also allows 1 and 2 bytes access */
3734 __u32 src_ip4;
3735 __u32 src_ip6[4];
3736 __u32 src_port; /* host byte order */
3737 __u32 dst_port; /* network byte order */
3738 __u32 dst_ip4;
3739 __u32 dst_ip6[4];
3740 __u32 state;
Amritha Nambiarc3c16f22020-05-26 17:34:36 -07003741 __s32 rx_queue_mapping;
David Ahern610236582016-12-01 08:48:04 -08003742};
3743
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003744struct bpf_tcp_sock {
3745 __u32 snd_cwnd; /* Sending congestion window */
3746 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3747 __u32 rtt_min;
3748 __u32 snd_ssthresh; /* Slow start size threshold */
3749 __u32 rcv_nxt; /* What we want to receive next */
3750 __u32 snd_nxt; /* Next sequence we send */
3751 __u32 snd_una; /* First byte we want an ack for */
3752 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3753 __u32 ecn_flags; /* ECN status bits. */
3754 __u32 rate_delivered; /* saved rate sample: packets delivered */
3755 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3756 __u32 packets_out; /* Packets which are "in flight" */
3757 __u32 retrans_out; /* Retransmitted packets out */
3758 __u32 total_retrans; /* Total retransmits for entire connection */
3759 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3760 * total number of segments in.
3761 */
3762 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3763 * total number of data segments in.
3764 */
3765 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3766 * The total number of segments sent.
3767 */
3768 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3769 * total number of data segments sent.
3770 */
3771 __u32 lost_out; /* Lost packets */
3772 __u32 sacked_out; /* SACK'd packets */
3773 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3774 * sum(delta(rcv_nxt)), or how many bytes
3775 * were acked.
3776 */
3777 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3778 * sum(delta(snd_una)), or how many bytes
3779 * were acked.
3780 */
Stanislav Fomichev03577462019-07-02 09:13:58 -07003781 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3782 * total number of DSACK blocks received
3783 */
3784 __u32 delivered; /* Total data packets delivered incl. rexmits */
3785 __u32 delivered_ce; /* Like the above but only ECE marked packets */
Stanislav Fomichevc2cb5e82019-07-02 09:13:59 -07003786 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003787};
3788
Joe Stringer6acc9b42018-10-02 13:35:36 -07003789struct bpf_sock_tuple {
3790 union {
3791 struct {
3792 __be32 saddr;
3793 __be32 daddr;
3794 __be16 sport;
3795 __be16 dport;
3796 } ipv4;
3797 struct {
3798 __be32 saddr[4];
3799 __be32 daddr[4];
3800 __be16 sport;
3801 __be16 dport;
3802 } ipv6;
3803 };
3804};
3805
Jonathan Lemonfada7fd2019-06-06 13:59:40 -07003806struct bpf_xdp_sock {
3807 __u32 queue_id;
3808};
3809
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003810#define XDP_PACKET_HEADROOM 256
3811
Brenden Blanco6a773a12016-07-19 12:16:47 -07003812/* User return codes for XDP prog type.
3813 * A valid XDP program must return one of these defined values. All other
Daniel Borkmann9beb8be2017-09-09 01:40:35 +02003814 * return codes are reserved for future use. Unknown return codes will
3815 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
Brenden Blanco6a773a12016-07-19 12:16:47 -07003816 */
3817enum xdp_action {
3818 XDP_ABORTED = 0,
3819 XDP_DROP,
3820 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -07003821 XDP_TX,
John Fastabend814abfa2017-07-17 09:27:07 -07003822 XDP_REDIRECT,
Brenden Blanco6a773a12016-07-19 12:16:47 -07003823};
3824
3825/* user accessible metadata for XDP packet hook
3826 * new fields must be added to the end of this structure
3827 */
3828struct xdp_md {
3829 __u32 data;
3830 __u32 data_end;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003831 __u32 data_meta;
Jesper Dangaard Brouerdaaf24c2018-01-11 17:39:09 +01003832 /* Below access go through struct xdp_rxq_info */
Jesper Dangaard Brouer02dd3292018-01-03 11:26:14 +01003833 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3834 __u32 rx_queue_index; /* rxq->queue_index */
David Ahern64b59022020-05-29 16:07:14 -06003835
3836 __u32 egress_ifindex; /* txq->dev->ifindex */
Brenden Blanco6a773a12016-07-19 12:16:47 -07003837};
3838
Jesper Dangaard Brouer281920b2020-06-09 15:31:46 +02003839/* DEVMAP map-value layout
3840 *
3841 * The struct data-layout of map-value is a configuration interface.
3842 * New members can only be added to the end of this structure.
3843 */
3844struct bpf_devmap_val {
3845 __u32 ifindex; /* device index */
3846 union {
3847 int fd; /* prog fd on map write */
3848 __u32 id; /* prog id on map read */
3849 } bpf_prog;
3850};
3851
Lorenzo Bianconi644bfe52020-07-14 15:56:37 +02003852/* CPUMAP map-value layout
3853 *
3854 * The struct data-layout of map-value is a configuration interface.
3855 * New members can only be added to the end of this structure.
3856 */
3857struct bpf_cpumap_val {
3858 __u32 qsize; /* queue size to remote target CPU */
3859};
3860
John Fastabend174a79f2017-08-15 22:32:47 -07003861enum sk_action {
John Fastabendbfa640752017-10-27 09:45:53 -07003862 SK_DROP = 0,
3863 SK_PASS,
John Fastabend174a79f2017-08-15 22:32:47 -07003864};
3865
John Fastabend4f738ad2018-03-18 12:57:10 -07003866/* user accessible metadata for SK_MSG packet hook, new fields must
3867 * be added to the end of this structure
3868 */
3869struct sk_msg_md {
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003870 __bpf_md_ptr(void *, data);
3871 __bpf_md_ptr(void *, data_end);
John Fastabend303def32018-05-17 14:16:58 -07003872
3873 __u32 family;
3874 __u32 remote_ip4; /* Stored in network byte order */
3875 __u32 local_ip4; /* Stored in network byte order */
3876 __u32 remote_ip6[4]; /* Stored in network byte order */
3877 __u32 local_ip6[4]; /* Stored in network byte order */
3878 __u32 remote_port; /* Stored in network byte order */
3879 __u32 local_port; /* stored in host byte order */
John Fastabend3bdbd022018-12-16 15:47:04 -08003880 __u32 size; /* Total size of sk_msg */
John Fastabend13d70f52020-05-24 09:51:15 -07003881
3882 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
John Fastabend4f738ad2018-03-18 12:57:10 -07003883};
3884
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003885struct sk_reuseport_md {
3886 /*
3887 * Start of directly accessible data. It begins from
3888 * the tcp/udp header.
3889 */
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003890 __bpf_md_ptr(void *, data);
3891 /* End of directly accessible data */
3892 __bpf_md_ptr(void *, data_end);
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003893 /*
3894 * Total length of packet (starting from the tcp/udp header).
3895 * Note that the directly accessible bytes (data_end - data)
3896 * could be less than this "len". Those bytes could be
3897 * indirectly read by a helper "bpf_skb_load_bytes()".
3898 */
3899 __u32 len;
3900 /*
3901 * Eth protocol in the mac header (network byte order). e.g.
3902 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3903 */
3904 __u32 eth_protocol;
3905 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3906 __u32 bind_inany; /* Is sock bound to an INANY address? */
3907 __u32 hash; /* A hash of the packet 4 tuples */
3908};
3909
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003910#define BPF_TAG_SIZE 8
3911
3912struct bpf_prog_info {
3913 __u32 type;
3914 __u32 id;
3915 __u8 tag[BPF_TAG_SIZE];
3916 __u32 jited_prog_len;
3917 __u32 xlated_prog_len;
3918 __aligned_u64 jited_prog_insns;
3919 __aligned_u64 xlated_prog_insns;
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -07003920 __u64 load_time; /* ns since boottime */
3921 __u32 created_by_uid;
3922 __u32 nr_map_ids;
3923 __aligned_u64 map_ids;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003924 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski675fc272017-12-27 18:39:09 -08003925 __u32 ifindex;
Jiri Olsab85fab02018-04-25 19:41:06 +02003926 __u32 gpl_compatible:1;
Baruch Siach04723012019-06-28 07:08:45 +03003927 __u32 :31; /* alignment pad */
Jakub Kicinski675fc272017-12-27 18:39:09 -08003928 __u64 netns_dev;
3929 __u64 netns_ino;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303930 __u32 nr_jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303931 __u32 nr_jited_func_lens;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303932 __aligned_u64 jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303933 __aligned_u64 jited_func_lens;
Yonghong Song838e9692018-11-19 15:29:11 -08003934 __u32 btf_id;
3935 __u32 func_info_rec_size;
3936 __aligned_u64 func_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003937 __u32 nr_func_info;
3938 __u32 nr_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003939 __aligned_u64 line_info;
3940 __aligned_u64 jited_line_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003941 __u32 nr_jited_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003942 __u32 line_info_rec_size;
3943 __u32 jited_line_info_rec_size;
Song Liuc872bdb2018-12-12 09:37:46 -08003944 __u32 nr_prog_tags;
3945 __aligned_u64 prog_tags;
Alexei Starovoitov5f8f8b92019-02-25 14:28:40 -08003946 __u64 run_time_ns;
3947 __u64 run_cnt;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003948} __attribute__((aligned(8)));
3949
3950struct bpf_map_info {
3951 __u32 type;
3952 __u32 id;
3953 __u32 key_size;
3954 __u32 value_size;
3955 __u32 max_entries;
3956 __u32 map_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003957 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski52775b32018-01-17 19:13:28 -08003958 __u32 ifindex;
Martin KaFai Lau85d33df2020-01-08 16:35:05 -08003959 __u32 btf_vmlinux_value_type_id;
Jakub Kicinski52775b32018-01-17 19:13:28 -08003960 __u64 netns_dev;
3961 __u64 netns_ino;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -07003962 __u32 btf_id;
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -07003963 __u32 btf_key_type_id;
3964 __u32 btf_value_type_id;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003965} __attribute__((aligned(8)));
3966
Martin KaFai Lau62dab842018-05-04 14:49:52 -07003967struct bpf_btf_info {
3968 __aligned_u64 btf;
3969 __u32 btf_size;
3970 __u32 id;
3971} __attribute__((aligned(8)));
3972
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -07003973struct bpf_link_info {
3974 __u32 type;
3975 __u32 id;
3976 __u32 prog_id;
3977 union {
3978 struct {
3979 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3980 __u32 tp_name_len; /* in/out: tp_name buffer len */
3981 } raw_tracepoint;
3982 struct {
3983 __u32 attach_type;
3984 } tracing;
3985 struct {
3986 __u64 cgroup_id;
3987 __u32 attach_type;
3988 } cgroup;
Jakub Sitnicki7f045a42020-05-31 10:28:38 +02003989 struct {
3990 __u32 netns_ino;
3991 __u32 attach_type;
3992 } netns;
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -07003993 };
3994} __attribute__((aligned(8)));
3995
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003996/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3997 * by user and intended to be used by socket (e.g. to bind to, depends on
3998 * attach attach type).
3999 */
4000struct bpf_sock_addr {
4001 __u32 user_family; /* Allows 4-byte read, but no write. */
4002 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4003 * Stored in network byte order.
4004 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07004005 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov4fbac772018-03-30 15:08:02 -07004006 * Stored in network byte order.
4007 */
Andrey Ignatov7aebfa12020-05-13 18:50:27 -07004008 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
Andrey Ignatov4fbac772018-03-30 15:08:02 -07004009 * Stored in network byte order
4010 */
4011 __u32 family; /* Allows 4-byte read, but no write */
4012 __u32 type; /* Allows 4-byte read, but no write */
4013 __u32 protocol; /* Allows 4-byte read, but no write */
Stanislav Fomichev600c70b2019-07-01 10:38:39 -07004014 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07004015 * Stored in network byte order.
4016 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07004017 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07004018 * Stored in network byte order.
4019 */
Stanislav Fomichevfb85c4a2019-06-12 10:30:37 -07004020 __bpf_md_ptr(struct bpf_sock *, sk);
Andrey Ignatov4fbac772018-03-30 15:08:02 -07004021};
4022
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004023/* User bpf_sock_ops struct to access socket values and specify request ops
4024 * and their replies.
4025 * Some of this fields are in network (bigendian) byte order and may need
4026 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4027 * New fields can only be added at the end of this structure
4028 */
4029struct bpf_sock_ops {
4030 __u32 op;
4031 union {
Lawrence Brakmode525be2018-01-25 16:14:09 -08004032 __u32 args[4]; /* Optionally passed to bpf program */
4033 __u32 reply; /* Returned by bpf program */
4034 __u32 replylong[4]; /* Optionally returned by bpf prog */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004035 };
4036 __u32 family;
4037 __u32 remote_ip4; /* Stored in network byte order */
4038 __u32 local_ip4; /* Stored in network byte order */
4039 __u32 remote_ip6[4]; /* Stored in network byte order */
4040 __u32 local_ip6[4]; /* Stored in network byte order */
4041 __u32 remote_port; /* Stored in network byte order */
4042 __u32 local_port; /* stored in host byte order */
Lawrence Brakmof19397a2017-12-01 10:15:04 -08004043 __u32 is_fullsock; /* Some TCP fields are only valid if
4044 * there is a full socket. If not, the
4045 * fields read as zero.
4046 */
4047 __u32 snd_cwnd;
4048 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
Lawrence Brakmob13d8802018-01-25 16:14:10 -08004049 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
Lawrence Brakmo44f0e432018-01-25 16:14:12 -08004050 __u32 state;
4051 __u32 rtt_min;
4052 __u32 snd_ssthresh;
4053 __u32 rcv_nxt;
4054 __u32 snd_nxt;
4055 __u32 snd_una;
4056 __u32 mss_cache;
4057 __u32 ecn_flags;
4058 __u32 rate_delivered;
4059 __u32 rate_interval_us;
4060 __u32 packets_out;
4061 __u32 retrans_out;
4062 __u32 total_retrans;
4063 __u32 segs_in;
4064 __u32 data_segs_in;
4065 __u32 segs_out;
4066 __u32 data_segs_out;
4067 __u32 lost_out;
4068 __u32 sacked_out;
4069 __u32 sk_txhash;
4070 __u64 bytes_received;
4071 __u64 bytes_acked;
Stanislav Fomichev1314ef52019-06-12 10:30:38 -07004072 __bpf_md_ptr(struct bpf_sock *, sk);
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004073};
4074
Lawrence Brakmob13d8802018-01-25 16:14:10 -08004075/* Definitions for bpf_sock_ops_cb_flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004076enum {
4077 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4078 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4079 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4080 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
4081/* Mask of all currently supported cb flags */
4082 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
4083};
Lawrence Brakmob13d8802018-01-25 16:14:10 -08004084
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004085/* List of known BPF sock_ops operators.
4086 * New entries can only be added at the end
4087 */
4088enum {
4089 BPF_SOCK_OPS_VOID,
Lawrence Brakmo8550f322017-06-30 20:02:42 -07004090 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4091 * -1 if default value should be used
4092 */
Lawrence Brakmo13d3b1e2017-06-30 20:02:44 -07004093 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4094 * window (in packets) or -1 if default
4095 * value should be used
4096 */
Lawrence Brakmo9872a4b2017-06-30 20:02:47 -07004097 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4098 * active connection is initialized
4099 */
4100 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4101 * active connection is
4102 * established
4103 */
4104 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4105 * passive connection is
4106 * established
4107 */
Lawrence Brakmo91b5b212017-06-30 20:02:49 -07004108 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4109 * needs ECN
4110 */
Lawrence Brakmoe6546ef2017-10-20 11:05:39 -07004111 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4112 * based on the path and may be
4113 * dependent on the congestion control
4114 * algorithm. In general it indicates
4115 * a congestion threshold. RTTs above
4116 * this indicate congestion
4117 */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08004118 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4119 * Arg1: value of icsk_retransmits
4120 * Arg2: value of icsk_rto
4121 * Arg3: whether RTO has expired
4122 */
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08004123 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4124 * Arg1: sequence number of 1st byte
4125 * Arg2: # segments
4126 * Arg3: return value of
4127 * tcp_transmit_skb (0 => success)
4128 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08004129 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4130 * Arg1: old_state
4131 * Arg2: new_state
4132 */
Andrey Ignatovf333ee02018-07-11 17:33:32 -07004133 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4134 * socket transition to LISTEN state.
4135 */
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07004136 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4137 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08004138};
4139
4140/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4141 * changes between the TCP and BPF versions. Ideally this should never happen.
4142 * If it does, we need to add code to convert them before calling
4143 * the BPF sock_ops function.
4144 */
4145enum {
4146 BPF_TCP_ESTABLISHED = 1,
4147 BPF_TCP_SYN_SENT,
4148 BPF_TCP_SYN_RECV,
4149 BPF_TCP_FIN_WAIT1,
4150 BPF_TCP_FIN_WAIT2,
4151 BPF_TCP_TIME_WAIT,
4152 BPF_TCP_CLOSE,
4153 BPF_TCP_CLOSE_WAIT,
4154 BPF_TCP_LAST_ACK,
4155 BPF_TCP_LISTEN,
4156 BPF_TCP_CLOSING, /* Now a valid state */
4157 BPF_TCP_NEW_SYN_RECV,
4158
4159 BPF_TCP_MAX_STATES /* Leave at the end! */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004160};
4161
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004162enum {
4163 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4164 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4165};
Lawrence Brakmofc747812017-06-30 20:02:51 -07004166
Yonghong Song908432c2017-10-05 09:19:20 -07004167struct bpf_perf_event_value {
4168 __u64 counter;
4169 __u64 enabled;
4170 __u64 running;
4171};
4172
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004173enum {
4174 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4175 BPF_DEVCG_ACC_READ = (1ULL << 1),
4176 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4177};
Roman Gushchinebc614f2017-11-05 08:15:32 -05004178
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004179enum {
4180 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4181 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4182};
Roman Gushchinebc614f2017-11-05 08:15:32 -05004183
4184struct bpf_cgroup_dev_ctx {
Yonghong Song06ef0cc2017-12-18 10:13:44 -08004185 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4186 __u32 access_type;
Roman Gushchinebc614f2017-11-05 08:15:32 -05004187 __u32 major;
4188 __u32 minor;
4189};
4190
Alexei Starovoitovc4f66992018-03-28 12:05:37 -07004191struct bpf_raw_tracepoint_args {
4192 __u64 args[0];
4193};
4194
David Ahern87f5fc72018-05-09 20:34:26 -07004195/* DIRECT: Skip the FIB rules and go to FIB table associated with device
4196 * OUTPUT: Do lookup from egress perspective; default is ingress
4197 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004198enum {
4199 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4200 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4201};
David Ahern87f5fc72018-05-09 20:34:26 -07004202
David Ahern4c795792018-06-26 16:21:18 -07004203enum {
4204 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4205 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4206 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4207 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4208 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4209 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4210 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4211 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4212 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4213};
4214
David Ahern87f5fc72018-05-09 20:34:26 -07004215struct bpf_fib_lookup {
David Ahernfa898d72018-05-29 10:58:07 -07004216 /* input: network family for lookup (AF_INET, AF_INET6)
4217 * output: network family of egress nexthop
4218 */
4219 __u8 family;
David Ahern87f5fc72018-05-09 20:34:26 -07004220
4221 /* set if lookup is to consider L4 data - e.g., FIB rules */
4222 __u8 l4_protocol;
4223 __be16 sport;
4224 __be16 dport;
4225
4226 /* total length of packet from network header - used for MTU check */
4227 __u16 tot_len;
David Ahern4c795792018-06-26 16:21:18 -07004228
4229 /* input: L3 device index for lookup
4230 * output: device index from FIB lookup
4231 */
4232 __u32 ifindex;
David Ahern87f5fc72018-05-09 20:34:26 -07004233
4234 union {
4235 /* inputs to lookup */
4236 __u8 tos; /* AF_INET */
David Ahernbd3a08a2018-06-03 08:15:19 -07004237 __be32 flowinfo; /* AF_INET6, flow_label + priority */
David Ahern87f5fc72018-05-09 20:34:26 -07004238
David Ahernfa898d72018-05-29 10:58:07 -07004239 /* output: metric of fib result (IPv4/IPv6 only) */
4240 __u32 rt_metric;
David Ahern87f5fc72018-05-09 20:34:26 -07004241 };
4242
4243 union {
David Ahern87f5fc72018-05-09 20:34:26 -07004244 __be32 ipv4_src;
4245 __u32 ipv6_src[4]; /* in6_addr; network order */
4246 };
4247
David Ahernfa898d72018-05-29 10:58:07 -07004248 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4249 * network header. output: bpf_fib_lookup sets to gateway address
4250 * if FIB lookup returns gateway route
David Ahern87f5fc72018-05-09 20:34:26 -07004251 */
4252 union {
David Ahern87f5fc72018-05-09 20:34:26 -07004253 __be32 ipv4_dst;
4254 __u32 ipv6_dst[4]; /* in6_addr; network order */
4255 };
4256
4257 /* output */
4258 __be16 h_vlan_proto;
4259 __be16 h_vlan_TCI;
4260 __u8 smac[6]; /* ETH_ALEN */
4261 __u8 dmac[6]; /* ETH_ALEN */
4262};
4263
Yonghong Song41bdc4b2018-05-24 11:21:09 -07004264enum bpf_task_fd_type {
4265 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4266 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4267 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4268 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4269 BPF_FD_TYPE_UPROBE, /* filename + offset */
4270 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4271};
4272
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004273enum {
4274 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4275 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4276 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4277};
Stanislav Fomichev086f9562019-07-25 15:52:25 -07004278
Petar Penkovd58e4682018-09-14 07:46:18 -07004279struct bpf_flow_keys {
4280 __u16 nhoff;
4281 __u16 thoff;
4282 __u16 addr_proto; /* ETH_P_* of valid addrs */
4283 __u8 is_frag;
4284 __u8 is_first_frag;
4285 __u8 is_encap;
4286 __u8 ip_proto;
4287 __be16 n_proto;
4288 __be16 sport;
4289 __be16 dport;
4290 union {
4291 struct {
4292 __be32 ipv4_src;
4293 __be32 ipv4_dst;
4294 };
4295 struct {
4296 __u32 ipv6_src[4]; /* in6_addr; network order */
4297 __u32 ipv6_dst[4]; /* in6_addr; network order */
4298 };
4299 };
Stanislav Fomichev086f9562019-07-25 15:52:25 -07004300 __u32 flags;
Stanislav Fomichev71c99e32019-07-25 15:52:30 -07004301 __be32 flow_label;
Petar Penkovd58e4682018-09-14 07:46:18 -07004302};
4303
Yonghong Song838e9692018-11-19 15:29:11 -08004304struct bpf_func_info {
Martin KaFai Laud30d42e2018-12-05 17:35:44 -08004305 __u32 insn_off;
Yonghong Song838e9692018-11-19 15:29:11 -08004306 __u32 type_id;
4307};
4308
Martin KaFai Lauc454a462018-12-07 16:42:25 -08004309#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4310#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4311
4312struct bpf_line_info {
4313 __u32 insn_off;
4314 __u32 file_name_off;
4315 __u32 line_off;
4316 __u32 line_col;
4317};
4318
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08004319struct bpf_spin_lock {
4320 __u32 val;
4321};
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08004322
4323struct bpf_sysctl {
4324 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4325 * Allows 1,2,4-byte read, but no write.
4326 */
Andrey Ignatove1550bf2019-03-07 18:50:52 -08004327 __u32 file_pos; /* Sysctl file position to read from, write to.
4328 * Allows 1,2,4-byte read an 4-byte write.
4329 */
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08004330};
4331
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07004332struct bpf_sockopt {
4333 __bpf_md_ptr(struct bpf_sock *, sk);
4334 __bpf_md_ptr(void *, optval);
4335 __bpf_md_ptr(void *, optval_end);
4336
4337 __s32 level;
4338 __s32 optname;
4339 __s32 optlen;
4340 __s32 retval;
4341};
4342
Carlos Neirab4490c52020-03-04 17:41:56 -03004343struct bpf_pidns_info {
4344 __u32 pid;
4345 __u32 tgid;
4346};
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07004347#endif /* _UAPI__LINUX_BPF_H__ */