blob: 3ba2bbbed80c417ffdd136b53133afcf7456386c [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,
Daniel Mack0e336612016-11-23 16:52:25 +0100229 __MAX_BPF_ATTACH_TYPE
230};
231
232#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
233
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -0700234enum bpf_link_type {
235 BPF_LINK_TYPE_UNSPEC = 0,
236 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
237 BPF_LINK_TYPE_TRACING = 2,
238 BPF_LINK_TYPE_CGROUP = 3,
Yonghong Songde4e05c2020-05-09 10:59:01 -0700239 BPF_LINK_TYPE_ITER = 4,
Jakub Sitnicki7f045a42020-05-31 10:28:38 +0200240 BPF_LINK_TYPE_NETNS = 5,
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -0700241
242 MAX_BPF_LINK_TYPE,
243};
244
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700245/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
246 *
247 * NONE(default): No further bpf programs allowed in the subtree.
248 *
249 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
250 * the program in this cgroup yields to sub-cgroup program.
251 *
252 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
253 * that cgroup program gets run in addition to the program in this cgroup.
254 *
255 * Only one program is allowed to be attached to a cgroup with
256 * NONE or BPF_F_ALLOW_OVERRIDE flag.
257 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
258 * release old program and attach the new one. Attach flags has to match.
259 *
260 * Multiple programs are allowed to be attached to a cgroup with
261 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
262 * (those that were attached first, run first)
263 * The programs of sub-cgroup are executed first, then programs of
264 * this cgroup and then programs of parent cgroup.
265 * When children program makes decision (like picking TCP CA or sock bind)
266 * parent program has a chance to override it.
267 *
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800268 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
269 * programs for a cgroup. Though it's possible to replace an old program at
270 * any position by also specifying BPF_F_REPLACE flag and position itself in
271 * replace_bpf_fd attribute. Old program at this position will be released.
272 *
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700273 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
274 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
275 * Ex1:
276 * cgrp1 (MULTI progs A, B) ->
277 * cgrp2 (OVERRIDE prog C) ->
278 * cgrp3 (MULTI prog D) ->
279 * cgrp4 (OVERRIDE prog E) ->
280 * cgrp5 (NONE prog F)
281 * the event in cgrp5 triggers execution of F,D,A,B in that order.
282 * if prog F is detached, the execution is E,D,A,B
283 * if prog F and D are detached, the execution is E,A,B
284 * if prog F, E and D are detached, the execution is C,A,B
285 *
286 * All eligible programs are executed regardless of return code from
287 * earlier programs.
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800288 */
289#define BPF_F_ALLOW_OVERRIDE (1U << 0)
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700290#define BPF_F_ALLOW_MULTI (1U << 1)
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800291#define BPF_F_REPLACE (1U << 2)
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800292
David S. Millere07b98d2017-05-10 11:38:07 -0700293/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
294 * verifier will perform strict alignment checking as if the kernel
295 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
296 * and NET_IP_ALIGN defined to 2.
297 */
298#define BPF_F_STRICT_ALIGNMENT (1U << 0)
299
David Millere9ee9ef2018-11-30 21:08:14 -0800300/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
301 * verifier will allow any alignment whatsoever. On platforms
302 * with strict alignment requirements for loads ands stores (such
303 * as sparc and mips) the verifier validates that all loads and
304 * stores provably follow this requirement. This flag turns that
305 * checking and enforcement off.
306 *
307 * It is mostly used for testing when we want to validate the
308 * context and memory access aspects of the verifier, but because
309 * of an unaligned access the alignment check would trigger before
310 * the one we are interested in.
311 */
312#define BPF_F_ANY_ALIGNMENT (1U << 1)
313
Jiong Wangc240eff2019-05-24 23:25:16 +0100314/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
315 * Verifier does sub-register def/use analysis and identifies instructions whose
316 * def only matters for low 32-bit, high 32-bit is never referenced later
317 * through implicit zero extension. Therefore verifier notifies JIT back-ends
318 * that it is safe to ignore clearing high 32-bit for these instructions. This
319 * saves some back-ends a lot of code-gen. However such optimization is not
320 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
321 * hence hasn't used verifier's analysis result. But, we really want to have a
322 * way to be able to verify the correctness of the described optimization on
323 * x86_64 on which testsuites are frequently exercised.
324 *
325 * So, this flag is introduced. Once it is set, verifier will randomize high
326 * 32-bit for those instructions who has been identified as safe to ignore them.
327 * Then, if verifier is not doing correct analysis, such randomization will
328 * regress tests to expose bugs.
329 */
330#define BPF_F_TEST_RND_HI32 (1U << 2)
331
Alexei Starovoitov10d274e2019-08-22 22:52:12 -0700332/* The verifier internal test flag. Behavior is undefined */
333#define BPF_F_TEST_STATE_FREQ (1U << 3)
334
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200335/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
336 * two extensions:
337 *
338 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
339 * insn[0].imm: map fd map fd
340 * insn[1].imm: 0 offset into value
341 * insn[0].off: 0 0
342 * insn[1].off: 0 0
343 * ldimm64 rewrite: address of map address of map[0]+offset
344 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
345 */
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100346#define BPF_PSEUDO_MAP_FD 1
Daniel Borkmannd8eca5b2019-04-09 23:20:03 +0200347#define BPF_PSEUDO_MAP_VALUE 2
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100348
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800349/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
350 * offset to another bpf function
351 */
352#define BPF_PSEUDO_CALL 1
353
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800354/* flags for BPF_MAP_UPDATE_ELEM command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800355enum {
356 BPF_ANY = 0, /* create new element or update existing */
357 BPF_NOEXIST = 1, /* create new element if it didn't exist */
358 BPF_EXIST = 2, /* update existing element */
359 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
360};
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800361
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700362/* flags for BPF_MAP_CREATE command */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800363enum {
364 BPF_F_NO_PREALLOC = (1U << 0),
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800365/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800366 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800367 * which can scale and perform better.
368 * Note, the LRU nodes (including free nodes) cannot be moved
369 * across different LRU lists.
370 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800371 BPF_F_NO_COMMON_LRU = (1U << 1),
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700372/* Specify numa node during map creation */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800373 BPF_F_NUMA_NODE = (1U << 2),
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -0700374
Daniel Borkmann591fe982019-04-09 23:20:05 +0200375/* Flags for accessing BPF object from syscall side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800376 BPF_F_RDONLY = (1U << 3),
377 BPF_F_WRONLY = (1U << 4),
Chenbo Feng6e71b042017-10-18 13:00:22 -0700378
Song Liu615755a2018-03-14 10:23:21 -0700379/* Flag for stack_map, store build_id+offset instead of pointer */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800380 BPF_F_STACK_BUILD_ID = (1U << 5),
Song Liu615755a2018-03-14 10:23:21 -0700381
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000382/* Zero-initialize hash function seed. This should only be used for testing. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800383 BPF_F_ZERO_SEED = (1U << 6),
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000384
Daniel Borkmann591fe982019-04-09 23:20:05 +0200385/* Flags for accessing BPF object from program side. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800386 BPF_F_RDONLY_PROG = (1U << 7),
387 BPF_F_WRONLY_PROG = (1U << 8),
Daniel Borkmann591fe982019-04-09 23:20:05 +0200388
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700389/* Clone map from listener for newly accepted socket */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800390 BPF_F_CLONE = (1U << 9),
Stanislav Fomichev8f51dfc2019-08-14 10:37:49 -0700391
Andrii Nakryikofc970222019-11-17 09:28:04 -0800392/* Enable memory-mapping BPF map */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800393 BPF_F_MMAPABLE = (1U << 10),
394};
Andrii Nakryikofc970222019-11-17 09:28:04 -0800395
Andrey Ignatovf5bfcd92020-01-07 17:40:06 -0800396/* Flags for BPF_PROG_QUERY. */
397
398/* Query effective (directly attached + inherited from ancestor cgroups)
399 * programs that will be executed for events within a cgroup.
400 * attach_flags with this flag are returned only for directly attached programs.
401 */
Lorenz Bauer2f183362018-11-16 11:41:09 +0000402#define BPF_F_QUERY_EFFECTIVE (1U << 0)
403
Song Liud46edd62020-04-30 00:15:04 -0700404/* type for BPF_ENABLE_STATS */
405enum bpf_stats_type {
406 /* enabled run_time_ns and run_cnt */
407 BPF_STATS_RUN_TIME = 0,
408};
409
Song Liu615755a2018-03-14 10:23:21 -0700410enum bpf_stack_build_id_status {
411 /* user space need an empty entry to identify end of a trace */
412 BPF_STACK_BUILD_ID_EMPTY = 0,
413 /* with valid build_id and offset */
414 BPF_STACK_BUILD_ID_VALID = 1,
415 /* couldn't get build_id, fallback to ip */
416 BPF_STACK_BUILD_ID_IP = 2,
417};
418
419#define BPF_BUILD_ID_SIZE 20
420struct bpf_stack_build_id {
421 __s32 status;
422 unsigned char build_id[BPF_BUILD_ID_SIZE];
423 union {
424 __u64 offset;
425 __u64 ip;
426 };
427};
428
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -0800429#define BPF_OBJ_NAME_LEN 16U
430
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700431union bpf_attr {
432 struct { /* anonymous struct used by BPF_MAP_CREATE command */
433 __u32 map_type; /* one of enum bpf_map_type */
434 __u32 key_size; /* size of key in bytes */
435 __u32 value_size; /* size of value in bytes */
436 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700437 __u32 map_flags; /* BPF_MAP_CREATE related
438 * flags defined above.
439 */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700440 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700441 __u32 numa_node; /* numa node (effective only if
442 * BPF_F_NUMA_NODE is set).
443 */
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700444 char map_name[BPF_OBJ_NAME_LEN];
Jakub Kicinskia3884572018-01-11 20:29:09 -0800445 __u32 map_ifindex; /* ifindex of netdev to create on */
Martin KaFai Laua26ca7c2018-04-18 15:56:03 -0700446 __u32 btf_fd; /* fd pointing to a BTF type data */
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -0700447 __u32 btf_key_type_id; /* BTF type_id of the key */
448 __u32 btf_value_type_id; /* BTF type_id of the value */
Martin KaFai Lau85d33df2020-01-08 16:35:05 -0800449 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
450 * struct stored as the
451 * map value
452 */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700453 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700454
455 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
456 __u32 map_fd;
457 __aligned_u64 key;
458 union {
459 __aligned_u64 value;
460 __aligned_u64 next_key;
461 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800462 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700463 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700464
Brian Vazquezcb4d03a2020-01-15 10:43:01 -0800465 struct { /* struct used by BPF_MAP_*_BATCH commands */
466 __aligned_u64 in_batch; /* start batch,
467 * NULL to start from beginning
468 */
469 __aligned_u64 out_batch; /* output: next start batch */
470 __aligned_u64 keys;
471 __aligned_u64 values;
472 __u32 count; /* input/output:
473 * input: # of key/value
474 * elements
475 * output: # of filled elements
476 */
477 __u32 map_fd;
478 __u64 elem_flags;
479 __u64 flags;
480 } batch;
481
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700482 struct { /* anonymous struct used by BPF_PROG_LOAD command */
483 __u32 prog_type; /* one of enum bpf_prog_type */
484 __u32 insn_cnt;
485 __aligned_u64 insns;
486 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700487 __u32 log_level; /* verbosity level of verifier */
488 __u32 log_size; /* size of user buffer */
489 __aligned_u64 log_buf; /* user supplied buffer */
Daniel Borkmann6c4fc202018-12-16 00:49:47 +0100490 __u32 kern_version; /* not used */
David S. Millere07b98d2017-05-10 11:38:07 -0700491 __u32 prog_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700492 char prog_name[BPF_OBJ_NAME_LEN];
Jakub Kicinski1f6f4cb2017-11-20 15:21:53 -0800493 __u32 prog_ifindex; /* ifindex of netdev to prep for */
Andrey Ignatov5e43f892018-03-30 15:08:00 -0700494 /* For some prog types expected attach type must be known at
495 * load time to verify attach type specific parts of prog
496 * (context accesses, allowed helpers, etc).
497 */
498 __u32 expected_attach_type;
Yonghong Song838e9692018-11-19 15:29:11 -0800499 __u32 prog_btf_fd; /* fd pointing to BTF type data */
500 __u32 func_info_rec_size; /* userspace bpf_func_info size */
501 __aligned_u64 func_info; /* func info */
502 __u32 func_info_cnt; /* number of bpf_func_info records */
Martin KaFai Lauc454a462018-12-07 16:42:25 -0800503 __u32 line_info_rec_size; /* userspace bpf_line_info size */
504 __aligned_u64 line_info; /* line info */
505 __u32 line_info_cnt; /* number of bpf_line_info records */
Alexei Starovoitovccfe29eb22019-10-15 20:24:58 -0700506 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
Alexei Starovoitov5b92a282019-11-14 10:57:17 -0800507 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700508 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100509
510 struct { /* anonymous struct used by BPF_OBJ_* commands */
511 __aligned_u64 pathname;
512 __u32 bpf_fd;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700513 __u32 file_flags;
Daniel Borkmannb2197752015-10-29 14:58:09 +0100514 };
Daniel Mackf4324552016-11-23 16:52:27 +0100515
516 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
517 __u32 target_fd; /* container object to attach to */
518 __u32 attach_bpf_fd; /* eBPF program to attach */
519 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800520 __u32 attach_flags;
Andrey Ignatov7dd68b32019-12-18 23:44:35 -0800521 __u32 replace_bpf_fd; /* previously attached eBPF
522 * program to replace if
523 * BPF_F_REPLACE is used
524 */
Daniel Mackf4324552016-11-23 16:52:27 +0100525 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700526
527 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
528 __u32 prog_fd;
529 __u32 retval;
Lorenz Bauerb5a36b12018-12-03 11:31:23 +0000530 __u32 data_size_in; /* input: len of data_in */
531 __u32 data_size_out; /* input/output: len of data_out
532 * returns ENOSPC if data_out
533 * is too small.
534 */
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700535 __aligned_u64 data_in;
536 __aligned_u64 data_out;
537 __u32 repeat;
538 __u32 duration;
Stanislav Fomichevb0b93952019-04-09 11:49:09 -0700539 __u32 ctx_size_in; /* input: len of ctx_in */
540 __u32 ctx_size_out; /* input/output: len of ctx_out
541 * returns ENOSPC if ctx_out
542 * is too small.
543 */
544 __aligned_u64 ctx_in;
545 __aligned_u64 ctx_out;
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700546 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700547
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700548 struct { /* anonymous struct used by BPF_*_GET_*_ID */
549 union {
550 __u32 start_id;
551 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700552 __u32 map_id;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700553 __u32 btf_id;
Andrii Nakryikoa3b80e12020-04-28 17:16:06 -0700554 __u32 link_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700555 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700556 __u32 next_id;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700557 __u32 open_flags;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700558 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700559
560 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
561 __u32 bpf_fd;
562 __u32 info_len;
563 __aligned_u64 info;
564 } info;
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700565
566 struct { /* anonymous struct used by BPF_PROG_QUERY command */
567 __u32 target_fd; /* container object to query */
568 __u32 attach_type;
569 __u32 query_flags;
570 __u32 attach_flags;
571 __aligned_u64 prog_ids;
572 __u32 prog_cnt;
573 } query;
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700574
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700575 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700576 __u64 name;
577 __u32 prog_fd;
578 } raw_tracepoint;
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700579
580 struct { /* anonymous struct for BPF_BTF_LOAD */
581 __aligned_u64 btf;
582 __aligned_u64 btf_log_buf;
583 __u32 btf_size;
584 __u32 btf_log_size;
585 __u32 btf_log_level;
586 };
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700587
588 struct {
589 __u32 pid; /* input: pid */
590 __u32 fd; /* input: fd */
591 __u32 flags; /* input: flags */
592 __u32 buf_len; /* input/output: buf len */
593 __aligned_u64 buf; /* input/output:
594 * tp_name for tracepoint
595 * symbol for kprobe
596 * filename for uprobe
597 */
598 __u32 prog_id; /* output: prod_id */
599 __u32 fd_type; /* output: BPF_FD_TYPE_* */
600 __u64 probe_offset; /* output: probe_offset */
601 __u64 probe_addr; /* output: probe_addr */
602 } task_fd_query;
Andrii Nakryikoaf6eea52020-03-29 19:59:58 -0700603
604 struct { /* struct used by BPF_LINK_CREATE command */
605 __u32 prog_fd; /* eBPF program to attach */
606 __u32 target_fd; /* object to attach to */
607 __u32 attach_type; /* attach type */
608 __u32 flags; /* extra flags */
609 } link_create;
Andrii Nakryiko0c991eb2020-03-29 19:59:59 -0700610
611 struct { /* struct used by BPF_LINK_UPDATE command */
612 __u32 link_fd; /* link fd */
613 /* new program fd to update link with */
614 __u32 new_prog_fd;
615 __u32 flags; /* extra flags */
616 /* expected link's program fd; is specified only if
617 * BPF_F_REPLACE flag is set in flags */
618 __u32 old_prog_fd;
619 } link_update;
620
Song Liud46edd62020-04-30 00:15:04 -0700621 struct { /* struct used by BPF_ENABLE_STATS command */
622 __u32 type;
623 } enable_stats;
624
Yonghong Songac51d992020-05-09 10:59:05 -0700625 struct { /* struct used by BPF_ITER_CREATE command */
626 __u32 link_fd;
627 __u32 flags;
628 } iter_create;
629
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700630} __attribute__((aligned(8)));
631
Quentin Monnet56a092c2018-04-25 18:16:52 +0100632/* The description below is an attempt at providing documentation to eBPF
633 * developers about the multiple available eBPF helper functions. It can be
634 * parsed and used to produce a manual page. The workflow is the following,
635 * and requires the rst2man utility:
Thomas Grafebb676d2016-10-27 11:23:51 +0200636 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100637 * $ ./scripts/bpf_helpers_doc.py \
638 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
639 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
640 * $ man /tmp/bpf-helpers.7
Thomas Grafebb676d2016-10-27 11:23:51 +0200641 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100642 * Note that in order to produce this external documentation, some RST
643 * formatting is used in the descriptions to get "bold" and "italics" in
644 * manual pages. Also note that the few trailing white spaces are
645 * intentional, removing them would break paragraphs for rst2man.
Thomas Grafebb676d2016-10-27 11:23:51 +0200646 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100647 * Start of BPF helper function descriptions:
Quentin Monnetad4a5222018-04-25 18:16:53 +0100648 *
649 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
650 * Description
651 * Perform a lookup in *map* for an entry associated to *key*.
652 * Return
653 * Map value associated to *key*, or **NULL** if no entry was
654 * found.
655 *
656 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
657 * Description
658 * Add or update the value of the entry associated to *key* in
659 * *map* with *value*. *flags* is one of:
660 *
661 * **BPF_NOEXIST**
662 * The entry for *key* must not exist in the map.
663 * **BPF_EXIST**
664 * The entry for *key* must already exist in the map.
665 * **BPF_ANY**
666 * No condition on the existence of the entry for *key*.
667 *
668 * Flag value **BPF_NOEXIST** cannot be used for maps of types
669 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
670 * elements always exist), the helper would return an error.
671 * Return
672 * 0 on success, or a negative error in case of failure.
673 *
674 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
675 * Description
676 * Delete entry with *key* from *map*.
677 * Return
678 * 0 on success, or a negative error in case of failure.
679 *
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +0100680 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100681 * Description
682 * For tracing programs, safely attempt to read *size* bytes from
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +0100683 * kernel space address *unsafe_ptr* and store the data in *dst*.
684 *
Quentin Monnetab8d7802020-05-11 17:15:35 +0100685 * Generally, use **bpf_probe_read_user**\ () or
686 * **bpf_probe_read_kernel**\ () instead.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100687 * Return
688 * 0 on success, or a negative error in case of failure.
689 *
690 * u64 bpf_ktime_get_ns(void)
691 * Description
692 * Return the time elapsed since system boot, in nanoseconds.
Maciej Żenczykowski71d19212020-04-26 09:15:25 -0700693 * Does not include time the system was suspended.
Quentin Monnetab8d7802020-05-11 17:15:35 +0100694 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
Quentin Monnetad4a5222018-04-25 18:16:53 +0100695 * Return
696 * Current *ktime*.
697 *
698 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
699 * Description
700 * This helper is a "printk()-like" facility for debugging. It
701 * prints a message defined by format *fmt* (of size *fmt_size*)
702 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
703 * available. It can take up to three additional **u64**
704 * arguments (as an eBPF helpers, the total number of arguments is
705 * limited to five).
706 *
707 * Each time the helper is called, it appends a line to the trace.
Peter Wu55c33df2019-08-21 00:08:59 +0100708 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
709 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
Quentin Monnetad4a5222018-04-25 18:16:53 +0100710 * The format of the trace is customizable, and the exact output
711 * one will get depends on the options set in
712 * *\/sys/kernel/debug/tracing/trace_options* (see also the
713 * *README* file under the same directory). However, it usually
714 * defaults to something like:
715 *
716 * ::
717 *
718 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
719 *
720 * In the above:
721 *
722 * * ``telnet`` is the name of the current task.
723 * * ``470`` is the PID of the current task.
724 * * ``001`` is the CPU number on which the task is
725 * running.
726 * * In ``.N..``, each character refers to a set of
727 * options (whether irqs are enabled, scheduling
728 * options, whether hard/softirqs are running, level of
729 * preempt_disabled respectively). **N** means that
730 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
731 * are set.
732 * * ``419421.045894`` is a timestamp.
733 * * ``0x00000001`` is a fake value used by BPF for the
734 * instruction pointer register.
735 * * ``<formatted msg>`` is the message formatted with
736 * *fmt*.
737 *
738 * The conversion specifiers supported by *fmt* are similar, but
739 * more limited than for printk(). They are **%d**, **%i**,
740 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
741 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
742 * of field, padding with zeroes, etc.) is available, and the
743 * helper will return **-EINVAL** (but print nothing) if it
744 * encounters an unknown specifier.
745 *
746 * Also, note that **bpf_trace_printk**\ () is slow, and should
747 * only be used for debugging purposes. For this reason, a notice
748 * bloc (spanning several lines) is printed to kernel logs and
749 * states that the helper should not be used "for production use"
750 * the first time this helper is used (or more precisely, when
751 * **trace_printk**\ () buffers are allocated). For passing values
752 * to user space, perf events should be preferred.
753 * Return
754 * The number of bytes written to the buffer, or a negative error
755 * in case of failure.
756 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100757 * u32 bpf_get_prandom_u32(void)
758 * Description
759 * Get a pseudo-random number.
760 *
761 * From a security point of view, this helper uses its own
762 * pseudo-random internal state, and cannot be used to infer the
763 * seed of other random functions in the kernel. However, it is
764 * essential to note that the generator used by the helper is not
765 * cryptographically secure.
766 * Return
767 * A random 32-bit unsigned value.
768 *
769 * u32 bpf_get_smp_processor_id(void)
770 * Description
771 * Get the SMP (symmetric multiprocessing) processor id. Note that
772 * all programs run with preemption disabled, which means that the
773 * SMP processor id is stable during all the execution of the
774 * program.
775 * Return
776 * The SMP id of the processor running the program.
777 *
Quentin Monnetad4a5222018-04-25 18:16:53 +0100778 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
779 * Description
780 * Store *len* bytes from address *from* into the packet
781 * associated to *skb*, at *offset*. *flags* are a combination of
782 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
783 * checksum for the packet after storing the bytes) and
784 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
785 * **->swhash** and *skb*\ **->l4hash** to 0).
786 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100787 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100788 * packet buffer. Therefore, at load time, all checks on pointers
789 * previously done by the verifier are invalidated and must be
790 * performed again, if the helper is used in combination with
791 * direct packet access.
792 * Return
793 * 0 on success, or a negative error in case of failure.
794 *
795 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
796 * Description
797 * Recompute the layer 3 (e.g. IP) checksum for the packet
798 * associated to *skb*. Computation is incremental, so the helper
799 * must know the former value of the header field that was
800 * modified (*from*), the new value of this field (*to*), and the
801 * number of bytes (2 or 4) for this field, stored in *size*.
802 * Alternatively, it is possible to store the difference between
803 * the previous and the new values of the header field in *to*, by
804 * setting *from* and *size* to 0. For both methods, *offset*
805 * indicates the location of the IP checksum within the packet.
806 *
807 * This helper works in combination with **bpf_csum_diff**\ (),
808 * which does not update the checksum in-place, but offers more
809 * flexibility and can handle sizes larger than 2 or 4 for the
810 * checksum to update.
811 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100812 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100813 * packet buffer. Therefore, at load time, all checks on pointers
814 * previously done by the verifier are invalidated and must be
815 * performed again, if the helper is used in combination with
816 * direct packet access.
817 * Return
818 * 0 on success, or a negative error in case of failure.
819 *
820 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
821 * Description
822 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
823 * packet associated to *skb*. Computation is incremental, so the
824 * helper must know the former value of the header field that was
825 * modified (*from*), the new value of this field (*to*), and the
826 * number of bytes (2 or 4) for this field, stored on the lowest
827 * four bits of *flags*. Alternatively, it is possible to store
828 * the difference between the previous and the new values of the
829 * header field in *to*, by setting *from* and the four lowest
830 * bits of *flags* to 0. For both methods, *offset* indicates the
831 * location of the IP checksum within the packet. In addition to
832 * the size of the field, *flags* can be added (bitwise OR) actual
833 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
834 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
835 * for updates resulting in a null checksum the value is set to
836 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
837 * the checksum is to be computed against a pseudo-header.
838 *
839 * This helper works in combination with **bpf_csum_diff**\ (),
840 * which does not update the checksum in-place, but offers more
841 * flexibility and can handle sizes larger than 2 or 4 for the
842 * checksum to update.
843 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100844 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100845 * packet buffer. Therefore, at load time, all checks on pointers
846 * previously done by the verifier are invalidated and must be
847 * performed again, if the helper is used in combination with
848 * direct packet access.
849 * Return
850 * 0 on success, or a negative error in case of failure.
851 *
852 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
853 * Description
854 * This special helper is used to trigger a "tail call", or in
855 * other words, to jump into another eBPF program. The same stack
856 * frame is used (but values on stack and in registers for the
857 * caller are not accessible to the callee). This mechanism allows
858 * for program chaining, either for raising the maximum number of
859 * available eBPF instructions, or to execute given programs in
860 * conditional blocks. For security reasons, there is an upper
861 * limit to the number of successive tail calls that can be
862 * performed.
863 *
864 * Upon call of this helper, the program attempts to jump into a
865 * program referenced at index *index* in *prog_array_map*, a
866 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
867 * *ctx*, a pointer to the context.
868 *
869 * If the call succeeds, the kernel immediately runs the first
870 * instruction of the new program. This is not a function call,
871 * and it never returns to the previous program. If the call
872 * fails, then the helper has no effect, and the caller continues
873 * to run its subsequent instructions. A call can fail if the
874 * destination program for the jump does not exist (i.e. *index*
875 * is superior to the number of entries in *prog_array_map*), or
876 * if the maximum number of tail calls has been reached for this
877 * chain of programs. This limit is defined in the kernel by the
878 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
879 * which is currently set to 32.
880 * Return
881 * 0 on success, or a negative error in case of failure.
882 *
883 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
884 * Description
885 * Clone and redirect the packet associated to *skb* to another
886 * net device of index *ifindex*. Both ingress and egress
887 * interfaces can be used for redirection. The **BPF_F_INGRESS**
888 * value in *flags* is used to make the distinction (ingress path
889 * is selected if the flag is present, egress path otherwise).
890 * This is the only flag supported for now.
891 *
892 * In comparison with **bpf_redirect**\ () helper,
893 * **bpf_clone_redirect**\ () has the associated cost of
894 * duplicating the packet buffer, but this can be executed out of
895 * the eBPF program. Conversely, **bpf_redirect**\ () is more
896 * efficient, but it is handled through an action code where the
897 * redirection happens only after the eBPF program has returned.
898 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100899 * A call to this helper is susceptible to change the underlying
Quentin Monnetad4a5222018-04-25 18:16:53 +0100900 * packet buffer. Therefore, at load time, all checks on pointers
901 * previously done by the verifier are invalidated and must be
902 * performed again, if the helper is used in combination with
903 * direct packet access.
904 * Return
905 * 0 on success, or a negative error in case of failure.
Quentin Monnetc456dec2018-04-25 18:16:54 +0100906 *
907 * u64 bpf_get_current_pid_tgid(void)
908 * Return
909 * A 64-bit integer containing the current tgid and pid, and
910 * created as such:
911 * *current_task*\ **->tgid << 32 \|**
912 * *current_task*\ **->pid**.
913 *
914 * u64 bpf_get_current_uid_gid(void)
915 * Return
916 * A 64-bit integer containing the current GID and UID, and
917 * created as such: *current_gid* **<< 32 \|** *current_uid*.
918 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -0700919 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
Quentin Monnetc456dec2018-04-25 18:16:54 +0100920 * Description
921 * Copy the **comm** attribute of the current task into *buf* of
922 * *size_of_buf*. The **comm** attribute contains the name of
923 * the executable (excluding the path) for the current task. The
924 * *size_of_buf* must be strictly positive. On success, the
925 * helper makes sure that the *buf* is NUL-terminated. On failure,
926 * it is filled with zeroes.
927 * Return
928 * 0 on success, or a negative error in case of failure.
929 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100930 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
931 * Description
932 * Retrieve the classid for the current task, i.e. for the net_cls
933 * cgroup to which *skb* belongs.
934 *
935 * This helper can be used on TC egress path, but not on ingress.
936 *
937 * The net_cls cgroup provides an interface to tag network packets
938 * based on a user-provided identifier for all traffic coming from
939 * the tasks belonging to the related cgroup. See also the related
940 * kernel documentation, available from the Linux sources in file
Mauro Carvalho Chehabda82c922019-06-27 13:08:35 -0300941 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100942 *
943 * The Linux kernel has two versions for cgroups: there are
944 * cgroups v1 and cgroups v2. Both are available to users, who can
945 * use a mixture of them, but note that the net_cls cgroup is for
946 * cgroup v1 only. This makes it incompatible with BPF programs
947 * run on cgroups, which is a cgroup-v2-only feature (a socket can
948 * only hold data for one version of cgroups at a time).
949 *
950 * This helper is only available is the kernel was compiled with
951 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
952 * "**y**" or to "**m**".
953 * Return
954 * The classid, or 0 for the default unconfigured classid.
955 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100956 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
957 * Description
958 * Push a *vlan_tci* (VLAN tag control information) of protocol
959 * *vlan_proto* to the packet associated to *skb*, then update
960 * the checksum. Note that if *vlan_proto* is different from
961 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
962 * be **ETH_P_8021Q**.
963 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100964 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100965 * packet buffer. Therefore, at load time, all checks on pointers
966 * previously done by the verifier are invalidated and must be
967 * performed again, if the helper is used in combination with
968 * direct packet access.
969 * Return
970 * 0 on success, or a negative error in case of failure.
971 *
972 * int bpf_skb_vlan_pop(struct sk_buff *skb)
973 * Description
974 * Pop a VLAN header from the packet associated to *skb*.
975 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +0100976 * A call to this helper is susceptible to change the underlying
Quentin Monnetc456dec2018-04-25 18:16:54 +0100977 * packet buffer. Therefore, at load time, all checks on pointers
978 * previously done by the verifier are invalidated and must be
979 * performed again, if the helper is used in combination with
980 * direct packet access.
981 * Return
982 * 0 on success, or a negative error in case of failure.
983 *
984 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
985 * Description
986 * Get tunnel metadata. This helper takes a pointer *key* to an
987 * empty **struct bpf_tunnel_key** of **size**, that will be
988 * filled with tunnel metadata for the packet associated to *skb*.
989 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
990 * indicates that the tunnel is based on IPv6 protocol instead of
991 * IPv4.
992 *
993 * The **struct bpf_tunnel_key** is an object that generalizes the
994 * principal parameters used by various tunneling protocols into a
995 * single struct. This way, it can be used to easily make a
996 * decision based on the contents of the encapsulation header,
997 * "summarized" in this struct. In particular, it holds the IP
998 * address of the remote end (IPv4 or IPv6, depending on the case)
999 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1000 * this struct exposes the *key*\ **->tunnel_id**, which is
1001 * generally mapped to a VNI (Virtual Network Identifier), making
1002 * it programmable together with the **bpf_skb_set_tunnel_key**\
1003 * () helper.
1004 *
1005 * Let's imagine that the following code is part of a program
1006 * attached to the TC ingress interface, on one end of a GRE
1007 * tunnel, and is supposed to filter out all messages coming from
1008 * remote ends with IPv4 address other than 10.0.0.1:
1009 *
1010 * ::
1011 *
1012 * int ret;
1013 * struct bpf_tunnel_key key = {};
1014 *
1015 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1016 * if (ret < 0)
1017 * return TC_ACT_SHOT; // drop packet
1018 *
1019 * if (key.remote_ipv4 != 0x0a000001)
1020 * return TC_ACT_SHOT; // drop packet
1021 *
1022 * return TC_ACT_OK; // accept packet
1023 *
1024 * This interface can also be used with all encapsulation devices
1025 * that can operate in "collect metadata" mode: instead of having
1026 * one network device per specific configuration, the "collect
1027 * metadata" mode only requires a single device where the
1028 * configuration can be extracted from this helper.
1029 *
1030 * This can be used together with various tunnels such as VXLan,
1031 * Geneve, GRE or IP in IP (IPIP).
1032 * Return
1033 * 0 on success, or a negative error in case of failure.
1034 *
1035 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1036 * Description
1037 * Populate tunnel metadata for packet associated to *skb.* The
1038 * tunnel metadata is set to the contents of *key*, of *size*. The
1039 * *flags* can be set to a combination of the following values:
1040 *
1041 * **BPF_F_TUNINFO_IPV6**
1042 * Indicate that the tunnel is based on IPv6 protocol
1043 * instead of IPv4.
1044 * **BPF_F_ZERO_CSUM_TX**
1045 * For IPv4 packets, add a flag to tunnel metadata
1046 * indicating that checksum computation should be skipped
1047 * and checksum set to zeroes.
1048 * **BPF_F_DONT_FRAGMENT**
1049 * Add a flag to tunnel metadata indicating that the
1050 * packet should not be fragmented.
1051 * **BPF_F_SEQ_NUMBER**
1052 * Add a flag to tunnel metadata indicating that a
1053 * sequence number should be added to tunnel header before
1054 * sending the packet. This flag was added for GRE
1055 * encapsulation, but might be used with other protocols
1056 * as well in the future.
1057 *
1058 * Here is a typical usage on the transmit path:
1059 *
1060 * ::
1061 *
1062 * struct bpf_tunnel_key key;
1063 * populate key ...
1064 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1065 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1066 *
1067 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1068 * helper for additional information.
1069 * Return
1070 * 0 on success, or a negative error in case of failure.
1071 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001072 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1073 * Description
1074 * Read the value of a perf event counter. This helper relies on a
1075 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1076 * the perf event counter is selected when *map* is updated with
1077 * perf event file descriptors. The *map* is an array whose size
1078 * is the number of available CPUs, and each cell contains a value
1079 * relative to one CPU. The value to retrieve is indicated by
1080 * *flags*, that contains the index of the CPU to look up, masked
1081 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1082 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1083 * current CPU should be retrieved.
1084 *
1085 * Note that before Linux 4.13, only hardware perf event can be
1086 * retrieved.
1087 *
1088 * Also, be aware that the newer helper
1089 * **bpf_perf_event_read_value**\ () is recommended over
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001090 * **bpf_perf_event_read**\ () in general. The latter has some ABI
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001091 * quirks where error and counter value are used as a return code
1092 * (which is wrong to do since ranges may overlap). This issue is
Quentin Monnet3bd5a092018-04-30 11:39:03 +01001093 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1094 * time provides more features over the **bpf_perf_event_read**\
1095 * () interface. Please refer to the description of
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001096 * **bpf_perf_event_read_value**\ () for details.
1097 * Return
1098 * The value of the perf event counter read from the map, or a
1099 * negative error code in case of failure.
1100 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001101 * int bpf_redirect(u32 ifindex, u64 flags)
1102 * Description
1103 * Redirect the packet to another net device of index *ifindex*.
1104 * This helper is somewhat similar to **bpf_clone_redirect**\
1105 * (), except that the packet is not cloned, which provides
1106 * increased performance.
1107 *
1108 * Except for XDP, both ingress and egress interfaces can be used
1109 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1110 * to make the distinction (ingress path is selected if the flag
1111 * is present, egress path otherwise). Currently, XDP only
1112 * supports redirection to the egress interface, and accepts no
1113 * flag at all.
1114 *
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001115 * The same effect can also be attained with the more generic
1116 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1117 * redirect target instead of providing it directly to the helper.
Quentin Monnetc456dec2018-04-25 18:16:54 +01001118 * Return
1119 * For XDP, the helper returns **XDP_REDIRECT** on success or
1120 * **XDP_ABORTED** on error. For other program types, the values
1121 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1122 * error.
1123 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001124 * u32 bpf_get_route_realm(struct sk_buff *skb)
1125 * Description
1126 * Retrieve the realm or the route, that is to say the
1127 * **tclassid** field of the destination for the *skb*. The
1128 * indentifier retrieved is a user-provided tag, similar to the
1129 * one used with the net_cls cgroup (see description for
1130 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1131 * held by a route (a destination entry), not by a task.
1132 *
1133 * Retrieving this identifier works with the clsact TC egress hook
1134 * (see also **tc-bpf(8)**), or alternatively on conventional
1135 * classful egress qdiscs, but not on TC ingress path. In case of
1136 * clsact TC egress hook, this has the advantage that, internally,
1137 * the destination entry has not been dropped yet in the transmit
1138 * path. Therefore, the destination entry does not need to be
1139 * artificially held via **netif_keep_dst**\ () for a classful
1140 * qdisc until the *skb* is freed.
1141 *
1142 * This helper is available only if the kernel was compiled with
1143 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1144 * Return
1145 * The realm of the route for the packet associated to *skb*, or 0
1146 * if none was found.
1147 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001148 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001149 * Description
1150 * Write raw *data* blob into a special BPF perf event held by
1151 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1152 * event must have the following attributes: **PERF_SAMPLE_RAW**
1153 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1154 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1155 *
1156 * The *flags* are used to indicate the index in *map* for which
1157 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1158 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1159 * to indicate that the index of the current CPU core should be
1160 * used.
1161 *
1162 * The value to write, of *size*, is passed through eBPF stack and
1163 * pointed by *data*.
1164 *
1165 * The context of the program *ctx* needs also be passed to the
1166 * helper.
1167 *
1168 * On user space, a program willing to read the values needs to
1169 * call **perf_event_open**\ () on the perf event (either for
1170 * one or for all CPUs) and to store the file descriptor into the
1171 * *map*. This must be done before the eBPF program can send data
1172 * into it. An example is available in file
1173 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1174 * tree (the eBPF program counterpart is in
1175 * *samples/bpf/trace_output_kern.c*).
1176 *
1177 * **bpf_perf_event_output**\ () achieves better performance
1178 * than **bpf_trace_printk**\ () for sharing data with user
1179 * space, and is much better suitable for streaming data from eBPF
1180 * programs.
1181 *
1182 * Note that this helper is not restricted to tracing use cases
1183 * and can be used with programs attached to TC or XDP as well,
1184 * where it allows for passing data to user space listeners. Data
1185 * can be:
1186 *
1187 * * Only custom structs,
1188 * * Only the packet payload, or
1189 * * A combination of both.
1190 * Return
1191 * 0 on success, or a negative error in case of failure.
1192 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001193 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001194 * Description
1195 * This helper was provided as an easy way to load data from a
1196 * packet. It can be used to load *len* bytes from *offset* from
1197 * the packet associated to *skb*, into the buffer pointed by
1198 * *to*.
1199 *
1200 * Since Linux 4.7, usage of this helper has mostly been replaced
1201 * by "direct packet access", enabling packet data to be
1202 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1203 * pointing respectively to the first byte of packet data and to
1204 * the byte after the last byte of packet data. However, it
1205 * remains useful if one wishes to read large quantities of data
1206 * at once from a packet into the eBPF stack.
1207 * Return
1208 * 0 on success, or a negative error in case of failure.
1209 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001210 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
Quentin Monnetc456dec2018-04-25 18:16:54 +01001211 * Description
1212 * Walk a user or a kernel stack and return its id. To achieve
1213 * this, the helper needs *ctx*, which is a pointer to the context
1214 * on which the tracing program is executed, and a pointer to a
1215 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1216 *
1217 * The last argument, *flags*, holds the number of stack frames to
1218 * skip (from 0 to 255), masked with
1219 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1220 * a combination of the following flags:
1221 *
1222 * **BPF_F_USER_STACK**
1223 * Collect a user space stack instead of a kernel stack.
1224 * **BPF_F_FAST_STACK_CMP**
1225 * Compare stacks by hash only.
1226 * **BPF_F_REUSE_STACKID**
1227 * If two different stacks hash into the same *stackid*,
1228 * discard the old one.
1229 *
1230 * The stack id retrieved is a 32 bit long integer handle which
1231 * can be further combined with other data (including other stack
1232 * ids) and used as a key into maps. This can be useful for
1233 * generating a variety of graphs (such as flame graphs or off-cpu
1234 * graphs).
1235 *
1236 * For walking a stack, this helper is an improvement over
1237 * **bpf_probe_read**\ (), which can be used with unrolled loops
1238 * but is not efficient and consumes a lot of eBPF instructions.
1239 * Instead, **bpf_get_stackid**\ () can collect up to
1240 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1241 * this limit can be controlled with the **sysctl** program, and
1242 * that it should be manually increased in order to profile long
1243 * user stacks (such as stacks for Java programs). To do so, use:
1244 *
1245 * ::
1246 *
1247 * # sysctl kernel.perf_event_max_stack=<new value>
Quentin Monnetc456dec2018-04-25 18:16:54 +01001248 * Return
1249 * The positive or null stack id on success, or a negative error
1250 * in case of failure.
1251 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001252 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1253 * Description
1254 * Compute a checksum difference, from the raw buffer pointed by
1255 * *from*, of length *from_size* (that must be a multiple of 4),
1256 * towards the raw buffer pointed by *to*, of size *to_size*
1257 * (same remark). An optional *seed* can be added to the value
1258 * (this can be cascaded, the seed may come from a previous call
1259 * to the helper).
1260 *
1261 * This is flexible enough to be used in several ways:
1262 *
1263 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1264 * checksum, it can be used when pushing new data.
1265 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1266 * checksum, it can be used when removing data from a packet.
1267 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1268 * can be used to compute a diff. Note that *from_size* and
1269 * *to_size* do not need to be equal.
1270 *
1271 * This helper can be used in combination with
1272 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1273 * which one can feed in the difference computed with
1274 * **bpf_csum_diff**\ ().
1275 * Return
1276 * The checksum result, or a negative error code in case of
1277 * failure.
1278 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001279 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001280 * Description
1281 * Retrieve tunnel options metadata for the packet associated to
1282 * *skb*, and store the raw tunnel option data to the buffer *opt*
1283 * of *size*.
1284 *
1285 * This helper can be used with encapsulation devices that can
1286 * operate in "collect metadata" mode (please refer to the related
1287 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1288 * more details). A particular example where this can be used is
1289 * in combination with the Geneve encapsulation protocol, where it
1290 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1291 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1292 * the eBPF program. This allows for full customization of these
1293 * headers.
1294 * Return
1295 * The size of the option data retrieved.
1296 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001297 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001298 * Description
1299 * Set tunnel options metadata for the packet associated to *skb*
1300 * to the option data contained in the raw buffer *opt* of *size*.
1301 *
1302 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1303 * helper for additional information.
1304 * Return
1305 * 0 on success, or a negative error in case of failure.
1306 *
1307 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1308 * Description
1309 * Change the protocol of the *skb* to *proto*. Currently
1310 * supported are transition from IPv4 to IPv6, and from IPv6 to
1311 * IPv4. The helper takes care of the groundwork for the
1312 * transition, including resizing the socket buffer. The eBPF
1313 * program is expected to fill the new headers, if any, via
1314 * **skb_store_bytes**\ () and to recompute the checksums with
1315 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1316 * (). The main case for this helper is to perform NAT64
1317 * operations out of an eBPF program.
1318 *
1319 * Internally, the GSO type is marked as dodgy so that headers are
1320 * checked and segments are recalculated by the GSO/GRO engine.
1321 * The size for GSO target is adapted as well.
1322 *
1323 * All values for *flags* are reserved for future usage, and must
1324 * be left at zero.
1325 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001326 * A call to this helper is susceptible to change the underlying
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001327 * packet buffer. Therefore, at load time, all checks on pointers
1328 * previously done by the verifier are invalidated and must be
1329 * performed again, if the helper is used in combination with
1330 * direct packet access.
1331 * Return
1332 * 0 on success, or a negative error in case of failure.
1333 *
1334 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1335 * Description
1336 * Change the packet type for the packet associated to *skb*. This
1337 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1338 * the eBPF program does not have a write access to *skb*\
1339 * **->pkt_type** beside this helper. Using a helper here allows
1340 * for graceful handling of errors.
1341 *
1342 * The major use case is to change incoming *skb*s to
1343 * **PACKET_HOST** in a programmatic way instead of having to
1344 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1345 * example.
1346 *
1347 * Note that *type* only allows certain values. At this time, they
1348 * are:
1349 *
1350 * **PACKET_HOST**
1351 * Packet is for us.
1352 * **PACKET_BROADCAST**
1353 * Send packet to all.
1354 * **PACKET_MULTICAST**
1355 * Send packet to group.
1356 * **PACKET_OTHERHOST**
1357 * Send packet to someone else.
1358 * Return
1359 * 0 on success, or a negative error in case of failure.
1360 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001361 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1362 * Description
1363 * Check whether *skb* is a descendant of the cgroup2 held by
1364 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1365 * Return
1366 * The return value depends on the result of the test, and can be:
1367 *
1368 * * 0, if the *skb* failed the cgroup2 descendant test.
1369 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1370 * * A negative error code, if an error occurred.
1371 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001372 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1373 * Description
1374 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1375 * not set, in particular if the hash was cleared due to mangling,
1376 * recompute this hash. Later accesses to the hash can be done
1377 * directly with *skb*\ **->hash**.
1378 *
1379 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1380 * prototype with **bpf_skb_change_proto**\ (), or calling
1381 * **bpf_skb_store_bytes**\ () with the
1382 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1383 * the hash and to trigger a new computation for the next call to
1384 * **bpf_get_hash_recalc**\ ().
1385 * Return
1386 * The 32-bit hash.
1387 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001388 * u64 bpf_get_current_task(void)
1389 * Return
1390 * A pointer to the current task struct.
Quentin Monnetfa156012018-04-25 18:16:56 +01001391 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001392 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1393 * Description
1394 * Attempt in a safe way to write *len* bytes from the buffer
1395 * *src* to *dst* in memory. It only works for threads that are in
1396 * user context, and *dst* must be a valid user space address.
1397 *
1398 * This helper should not be used to implement any kind of
1399 * security mechanism because of TOC-TOU attacks, but rather to
1400 * debug, divert, and manipulate execution of semi-cooperative
1401 * processes.
1402 *
1403 * Keep in mind that this feature is meant for experiments, and it
1404 * has a risk of crashing the system and running programs.
1405 * Therefore, when an eBPF program using this helper is attached,
1406 * a warning including PID and process name is printed to kernel
1407 * logs.
1408 * Return
1409 * 0 on success, or a negative error in case of failure.
1410 *
1411 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1412 * Description
1413 * Check whether the probe is being run is the context of a given
1414 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1415 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1416 * Return
1417 * The return value depends on the result of the test, and can be:
1418 *
1419 * * 0, if the *skb* task belongs to the cgroup2.
1420 * * 1, if the *skb* task does not belong to the cgroup2.
1421 * * A negative error code, if an error occurred.
1422 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001423 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1424 * Description
1425 * Resize (trim or grow) the packet associated to *skb* to the
1426 * new *len*. The *flags* are reserved for future usage, and must
1427 * be left at zero.
1428 *
1429 * The basic idea is that the helper performs the needed work to
1430 * change the size of the packet, then the eBPF program rewrites
1431 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1432 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1433 * and others. This helper is a slow path utility intended for
1434 * replies with control messages. And because it is targeted for
1435 * slow path, the helper itself can afford to be slow: it
1436 * implicitly linearizes, unclones and drops offloads from the
1437 * *skb*.
1438 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001439 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001440 * packet buffer. Therefore, at load time, all checks on pointers
1441 * previously done by the verifier are invalidated and must be
1442 * performed again, if the helper is used in combination with
1443 * direct packet access.
1444 * Return
1445 * 0 on success, or a negative error in case of failure.
1446 *
1447 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1448 * Description
1449 * Pull in non-linear data in case the *skb* is non-linear and not
1450 * all of *len* are part of the linear section. Make *len* bytes
1451 * from *skb* readable and writable. If a zero value is passed for
1452 * *len*, then the whole length of the *skb* is pulled.
1453 *
1454 * This helper is only needed for reading and writing with direct
1455 * packet access.
1456 *
1457 * For direct packet access, testing that offsets to access
1458 * are within packet boundaries (test on *skb*\ **->data_end**) is
1459 * susceptible to fail if offsets are invalid, or if the requested
1460 * data is in non-linear parts of the *skb*. On failure the
1461 * program can just bail out, or in the case of a non-linear
1462 * buffer, use a helper to make the data available. The
1463 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1464 * the data. Another one consists in using **bpf_skb_pull_data**
1465 * to pull in once the non-linear parts, then retesting and
1466 * eventually access the data.
1467 *
1468 * At the same time, this also makes sure the *skb* is uncloned,
1469 * which is a necessary condition for direct write. As this needs
1470 * to be an invariant for the write part only, the verifier
1471 * detects writes and adds a prologue that is calling
1472 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1473 * the very beginning in case it is indeed cloned.
1474 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001475 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001476 * packet buffer. Therefore, at load time, all checks on pointers
1477 * previously done by the verifier are invalidated and must be
1478 * performed again, if the helper is used in combination with
1479 * direct packet access.
1480 * Return
1481 * 0 on success, or a negative error in case of failure.
1482 *
1483 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1484 * Description
1485 * Add the checksum *csum* into *skb*\ **->csum** in case the
1486 * driver has supplied a checksum for the entire packet into that
1487 * field. Return an error otherwise. This helper is intended to be
1488 * used in combination with **bpf_csum_diff**\ (), in particular
1489 * when the checksum needs to be updated after data has been
1490 * written into the packet through direct packet access.
1491 * Return
1492 * The checksum on success, or a negative error code in case of
1493 * failure.
1494 *
1495 * void bpf_set_hash_invalid(struct sk_buff *skb)
1496 * Description
1497 * Invalidate the current *skb*\ **->hash**. It can be used after
1498 * mangling on headers through direct packet access, in order to
1499 * indicate that the hash is outdated and to trigger a
1500 * recalculation the next time the kernel tries to access this
1501 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1502 *
1503 * int bpf_get_numa_node_id(void)
1504 * Description
1505 * Return the id of the current NUMA node. The primary use case
1506 * for this helper is the selection of sockets for the local NUMA
1507 * node, when the program is attached to sockets using the
1508 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1509 * but the helper is also available to other eBPF program types,
1510 * similarly to **bpf_get_smp_processor_id**\ ().
1511 * Return
1512 * The id of current NUMA node.
1513 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001514 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1515 * Description
1516 * Grows headroom of packet associated to *skb* and adjusts the
1517 * offset of the MAC header accordingly, adding *len* bytes of
1518 * space. It automatically extends and reallocates memory as
1519 * required.
1520 *
1521 * This helper can be used on a layer 3 *skb* to push a MAC header
1522 * for redirection into a layer 2 device.
1523 *
1524 * All values for *flags* are reserved for future usage, and must
1525 * be left at zero.
1526 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001527 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001528 * packet buffer. Therefore, at load time, all checks on pointers
1529 * previously done by the verifier are invalidated and must be
1530 * performed again, if the helper is used in combination with
1531 * direct packet access.
1532 * Return
1533 * 0 on success, or a negative error in case of failure.
1534 *
1535 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1536 * Description
1537 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1538 * it is possible to use a negative value for *delta*. This helper
1539 * can be used to prepare the packet for pushing or popping
1540 * headers.
1541 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001542 * A call to this helper is susceptible to change the underlying
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001543 * packet buffer. Therefore, at load time, all checks on pointers
1544 * previously done by the verifier are invalidated and must be
1545 * performed again, if the helper is used in combination with
1546 * direct packet access.
1547 * Return
1548 * 0 on success, or a negative error in case of failure.
1549 *
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001550 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001551 * Description
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001552 * Copy a NUL terminated string from an unsafe kernel address
Quentin Monnetab8d7802020-05-11 17:15:35 +01001553 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01001554 * more details.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001555 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001556 * Generally, use **bpf_probe_read_user_str**\ () or
1557 * **bpf_probe_read_kernel_str**\ () instead.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001558 * Return
1559 * On success, the strictly positive length of the string,
1560 * including the trailing NUL character. On error, a negative
1561 * value.
1562 *
1563 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1564 * Description
1565 * If the **struct sk_buff** pointed by *skb* has a known socket,
1566 * retrieve the cookie (generated by the kernel) of this socket.
1567 * If no cookie has been set yet, generate a new cookie. Once
1568 * generated, the socket cookie remains stable for the life of the
1569 * socket. This helper can be useful for monitoring per socket
Daniel Borkmanncd48bdd2019-08-08 13:57:25 +02001570 * networking traffic statistics as it provides a global socket
1571 * identifier that can be assumed unique.
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001572 * Return
1573 * A 8-byte long non-decreasing number on success, or 0 if the
1574 * socket field is missing inside *skb*.
1575 *
Andrey Ignatovd692f112018-07-30 17:42:28 -07001576 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1577 * Description
1578 * Equivalent to bpf_get_socket_cookie() helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001579 * *skb*, but gets socket from **struct bpf_sock_addr** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001580 * Return
1581 * A 8-byte long non-decreasing number.
1582 *
1583 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1584 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01001585 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
Quentin Monnet62369db2019-03-14 12:38:39 +00001586 * *skb*, but gets socket from **struct bpf_sock_ops** context.
Andrey Ignatovd692f112018-07-30 17:42:28 -07001587 * Return
1588 * A 8-byte long non-decreasing number.
1589 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001590 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1591 * Return
1592 * The owner UID of the socket associated to *skb*. If the socket
1593 * is **NULL**, or if it is not a full socket (i.e. if it is a
1594 * time-wait or a request socket instead), **overflowuid** value
1595 * is returned (note that **overflowuid** might also be the actual
1596 * UID value for the socket).
1597 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001598 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1599 * Description
1600 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1601 * to value *hash*.
1602 * Return
1603 * 0
1604 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001605 * int bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001606 * Description
1607 * Emulate a call to **setsockopt()** on the socket associated to
1608 * *bpf_socket*, which must be a full socket. The *level* at
1609 * which the option resides and the name *optname* of the option
1610 * must be specified, see **setsockopt(2)** for more information.
1611 * The option value of length *optlen* is pointed by *optval*.
1612 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001613 * *bpf_socket* should be one of the following:
Quentin Monnetab8d7802020-05-11 17:15:35 +01001614 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001615 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1616 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1617 * and **BPF_CGROUP_INET6_CONNECT**.
1618 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001619 * This helper actually implements a subset of **setsockopt()**.
1620 * It supports the following *level*\ s:
1621 *
1622 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1623 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1624 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1625 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1626 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1627 * **TCP_BPF_SNDCWND_CLAMP**.
1628 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1629 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1630 * Return
1631 * 0 on success, or a negative error in case of failure.
1632 *
Nicolas Dichtelb55cbc82018-10-17 16:24:48 +02001633 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001634 * Description
1635 * Grow or shrink the room for data in the packet associated to
1636 * *skb* by *len_diff*, and according to the selected *mode*.
1637 *
Daniel Borkmann836e66c2020-06-02 16:58:32 +02001638 * By default, the helper will reset any offloaded checksum
1639 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1640 * by the following flag:
1641 *
1642 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1643 * checksum data of the skb to CHECKSUM_NONE.
1644 *
Willem de Bruijn14aa3192019-03-22 14:32:54 -04001645 * There are two supported modes at this time:
1646 *
1647 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1648 * (room space is added or removed below the layer 2 header).
Quentin Monnetfa156012018-04-25 18:16:56 +01001649 *
1650 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1651 * (room space is added or removed below the layer 3 header).
1652 *
Willem de Bruijn868d5232019-03-22 14:32:56 -04001653 * The following flags are supported at this time:
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04001654 *
1655 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1656 * Adjusting mss in this way is not allowed for datagrams.
Quentin Monnetfa156012018-04-25 18:16:56 +01001657 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001658 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1659 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001660 * Any new space is reserved to hold a tunnel header.
1661 * Configure skb offsets and other fields accordingly.
1662 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001663 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1664 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
Willem de Bruijn868d5232019-03-22 14:32:56 -04001665 * Use with ENCAP_L3 flags to further specify the tunnel type.
1666 *
Quentin Monnet80867c52019-05-10 15:51:24 +01001667 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
Alan Maguire58dfc902019-04-09 15:06:41 +01001668 * Use with ENCAP_L3/L4 flags to further specify the tunnel
Quentin Monnet80867c52019-05-10 15:51:24 +01001669 * type; *len* is the length of the inner MAC header.
Alan Maguire58dfc902019-04-09 15:06:41 +01001670 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001671 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001672 * packet buffer. Therefore, at load time, all checks on pointers
1673 * previously done by the verifier are invalidated and must be
1674 * performed again, if the helper is used in combination with
1675 * direct packet access.
1676 * Return
1677 * 0 on success, or a negative error in case of failure.
1678 *
Quentin Monnetab127042018-04-25 18:16:59 +01001679 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1680 * Description
1681 * Redirect the packet to the endpoint referenced by *map* at
1682 * index *key*. Depending on its type, this *map* can contain
1683 * references to net devices (for forwarding packets through other
1684 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1685 * but this is only implemented for native XDP (with driver
1686 * support) as of this writing).
1687 *
Toke Høiland-Jørgensen43e74c02019-06-28 11:12:34 +02001688 * The lower two bits of *flags* are used as the return code if
1689 * the map lookup fails. This is so that the return value can be
Quentin Monnetab8d7802020-05-11 17:15:35 +01001690 * one of the XDP program return codes up to **XDP_TX**, as chosen
1691 * by the caller. Any higher bits in the *flags* argument must be
Toke Høiland-Jørgensen43e74c02019-06-28 11:12:34 +02001692 * unset.
Quentin Monnetab127042018-04-25 18:16:59 +01001693 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001694 * See also **bpf_redirect**\ (), which only supports redirecting
1695 * to an ifindex, but doesn't require a map to do so.
Quentin Monnetab127042018-04-25 18:16:59 +01001696 * Return
Toke Høiland-Jørgensenf25975f2020-02-18 14:03:34 +01001697 * **XDP_REDIRECT** on success, or the value of the two lower bits
Jakub Wilka33d3142020-04-22 10:23:24 +02001698 * of the *flags* argument on error.
Quentin Monnetab127042018-04-25 18:16:59 +01001699 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07001700 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001701 * Description
1702 * Redirect the packet to the socket referenced by *map* (of type
1703 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1704 * egress interfaces can be used for redirection. The
1705 * **BPF_F_INGRESS** value in *flags* is used to make the
1706 * distinction (ingress path is selected if the flag is present,
1707 * egress path otherwise). This is the only flag supported for now.
1708 * Return
1709 * **SK_PASS** on success, or **SK_DROP** on error.
1710 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001711 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
Quentin Monnetab127042018-04-25 18:16:59 +01001712 * Description
1713 * Add an entry to, or update a *map* referencing sockets. The
1714 * *skops* is used as a new value for the entry associated to
1715 * *key*. *flags* is one of:
1716 *
1717 * **BPF_NOEXIST**
1718 * The entry for *key* must not exist in the map.
1719 * **BPF_EXIST**
1720 * The entry for *key* must already exist in the map.
1721 * **BPF_ANY**
1722 * No condition on the existence of the entry for *key*.
1723 *
1724 * If the *map* has eBPF programs (parser and verdict), those will
1725 * be inherited by the socket being added. If the socket is
1726 * already attached to eBPF programs, this results in an error.
1727 * Return
1728 * 0 on success, or a negative error in case of failure.
1729 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001730 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1731 * Description
1732 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1733 * *delta* (which can be positive or negative). Note that this
1734 * operation modifies the address stored in *xdp_md*\ **->data**,
1735 * so the latter must be loaded only after the helper has been
1736 * called.
1737 *
1738 * The use of *xdp_md*\ **->data_meta** is optional and programs
1739 * are not required to use it. The rationale is that when the
1740 * packet is processed with XDP (e.g. as DoS filter), it is
1741 * possible to push further meta data along with it before passing
1742 * to the stack, and to give the guarantee that an ingress eBPF
1743 * program attached as a TC classifier on the same device can pick
1744 * this up for further post-processing. Since TC works with socket
1745 * buffers, it remains possible to set from XDP the **mark** or
1746 * **priority** pointers, or other pointers for the socket buffer.
1747 * Having this scratch space generic and programmable allows for
1748 * more flexibility as the user is free to store whatever meta
1749 * data they need.
1750 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01001751 * A call to this helper is susceptible to change the underlying
Quentin Monnetfa156012018-04-25 18:16:56 +01001752 * packet buffer. Therefore, at load time, all checks on pointers
1753 * previously done by the verifier are invalidated and must be
1754 * performed again, if the helper is used in combination with
1755 * direct packet access.
1756 * Return
1757 * 0 on success, or a negative error in case of failure.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001758 *
1759 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1760 * Description
1761 * Read the value of a perf event counter, and store it into *buf*
1762 * of size *buf_size*. This helper relies on a *map* of type
1763 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1764 * counter is selected when *map* is updated with perf event file
1765 * descriptors. The *map* is an array whose size is the number of
1766 * available CPUs, and each cell contains a value relative to one
1767 * CPU. The value to retrieve is indicated by *flags*, that
1768 * contains the index of the CPU to look up, masked with
1769 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1770 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1771 * current CPU should be retrieved.
1772 *
1773 * This helper behaves in a way close to
1774 * **bpf_perf_event_read**\ () helper, save that instead of
1775 * just returning the value observed, it fills the *buf*
1776 * structure. This allows for additional data to be retrieved: in
1777 * particular, the enabled and running times (in *buf*\
1778 * **->enabled** and *buf*\ **->running**, respectively) are
1779 * copied. In general, **bpf_perf_event_read_value**\ () is
1780 * recommended over **bpf_perf_event_read**\ (), which has some
1781 * ABI issues and provides fewer functionalities.
1782 *
1783 * These values are interesting, because hardware PMU (Performance
1784 * Monitoring Unit) counters are limited resources. When there are
1785 * more PMU based perf events opened than available counters,
1786 * kernel will multiplex these events so each event gets certain
1787 * percentage (but not all) of the PMU time. In case that
1788 * multiplexing happens, the number of samples or counter value
1789 * will not reflect the case compared to when no multiplexing
1790 * occurs. This makes comparison between different runs difficult.
1791 * Typically, the counter value should be normalized before
1792 * comparing to other experiments. The usual normalization is done
1793 * as follows.
1794 *
1795 * ::
1796 *
1797 * normalized_counter = counter * t_enabled / t_running
1798 *
1799 * Where t_enabled is the time enabled for event and t_running is
1800 * the time running for event since last normalization. The
1801 * enabled and running times are accumulated since the perf event
1802 * open. To achieve scaling factor between two invocations of an
Quentin Monnetab8d7802020-05-11 17:15:35 +01001803 * eBPF program, users can use CPU id as the key (which is
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001804 * typical for perf array usage model) to remember the previous
1805 * value and do the calculation inside the eBPF program.
1806 * Return
1807 * 0 on success, or a negative error in case of failure.
1808 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001809 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001810 * Description
1811 * For en eBPF program attached to a perf event, retrieve the
1812 * value of the event counter associated to *ctx* and store it in
1813 * the structure pointed by *buf* and of size *buf_size*. Enabled
1814 * and running times are also stored in the structure (see
1815 * description of helper **bpf_perf_event_read_value**\ () for
1816 * more details).
1817 * Return
1818 * 0 on success, or a negative error in case of failure.
1819 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001820 * int bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001821 * Description
1822 * Emulate a call to **getsockopt()** on the socket associated to
1823 * *bpf_socket*, which must be a full socket. The *level* at
1824 * which the option resides and the name *optname* of the option
1825 * must be specified, see **getsockopt(2)** for more information.
1826 * The retrieved value is stored in the structure pointed by
1827 * *opval* and of length *optlen*.
1828 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001829 * *bpf_socket* should be one of the following:
Quentin Monnetab8d7802020-05-11 17:15:35 +01001830 *
Stanislav Fomichevbeecf112020-04-30 16:31:52 -07001831 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1832 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1833 * and **BPF_CGROUP_INET6_CONNECT**.
1834 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001835 * This helper actually implements a subset of **getsockopt()**.
1836 * It supports the following *level*\ s:
1837 *
1838 * * **IPPROTO_TCP**, which supports *optname*
1839 * **TCP_CONGESTION**.
1840 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1841 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1842 * Return
1843 * 0 on success, or a negative error in case of failure.
1844 *
Peter Wu8050a392019-08-21 00:08:58 +01001845 * int bpf_override_return(struct pt_regs *regs, u64 rc)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001846 * Description
1847 * Used for error injection, this helper uses kprobes to override
1848 * the return value of the probed function, and to set it to *rc*.
1849 * The first argument is the context *regs* on which the kprobe
1850 * works.
1851 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01001852 * This helper works by setting the PC (program counter)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001853 * to an override function which is run in place of the original
1854 * probed function. This means the probed function is not run at
1855 * all. The replacement function just returns with the required
1856 * value.
1857 *
1858 * This helper has security implications, and thus is subject to
1859 * restrictions. It is only available if the kernel was compiled
1860 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1861 * option, and in this case it only works on functions tagged with
1862 * **ALLOW_ERROR_INJECTION** in the kernel code.
1863 *
1864 * Also, the helper is only available for the architectures having
1865 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1866 * x86 architecture is the only one to support this feature.
1867 * Return
1868 * 0
1869 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001870 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001871 * Description
1872 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1873 * for the full TCP socket associated to *bpf_sock_ops* to
1874 * *argval*.
1875 *
1876 * The primary use of this field is to determine if there should
1877 * be calls to eBPF programs of type
1878 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1879 * code. A program of the same type can change its value, per
1880 * connection and as necessary, when the connection is
1881 * established. This field is directly accessible for reading, but
1882 * this helper must be used for updates in order to return an
1883 * error if an eBPF program tries to set a callback that is not
1884 * supported in the current kernel.
1885 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001886 * *argval* is a flag array which can combine these flags:
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001887 *
1888 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1889 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1890 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07001891 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001892 *
Viet Hoang Tran725721a2019-04-15 09:54:55 +00001893 * Therefore, this function can be used to clear a callback flag by
1894 * setting the appropriate bit to zero. e.g. to disable the RTO
1895 * callback:
1896 *
1897 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1898 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1899 *
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001900 * Here are some examples of where one could call such eBPF
1901 * program:
1902 *
1903 * * When RTO fires.
1904 * * When a packet is retransmitted.
1905 * * When the connection terminates.
1906 * * When a packet is sent.
1907 * * When a packet is received.
1908 * Return
1909 * Code **-EINVAL** if the socket is not a full TCP socket;
1910 * otherwise, a positive number containing the bits that could not
1911 * be set is returned (which comes down to 0 if all bits were set
1912 * as required).
1913 *
Quentin Monnetab127042018-04-25 18:16:59 +01001914 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1915 * Description
1916 * This helper is used in programs implementing policies at the
1917 * socket level. If the message *msg* is allowed to pass (i.e. if
1918 * the verdict eBPF program returns **SK_PASS**), redirect it to
1919 * the socket referenced by *map* (of type
1920 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1921 * egress interfaces can be used for redirection. The
1922 * **BPF_F_INGRESS** value in *flags* is used to make the
1923 * distinction (ingress path is selected if the flag is present,
1924 * egress path otherwise). This is the only flag supported for now.
1925 * Return
1926 * **SK_PASS** on success, or **SK_DROP** on error.
1927 *
1928 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1929 * Description
1930 * For socket policies, apply the verdict of the eBPF program to
1931 * the next *bytes* (number of bytes) of message *msg*.
1932 *
1933 * For example, this helper can be used in the following cases:
1934 *
1935 * * A single **sendmsg**\ () or **sendfile**\ () system call
1936 * contains multiple logical messages that the eBPF program is
1937 * supposed to read and for which it should apply a verdict.
1938 * * An eBPF program only cares to read the first *bytes* of a
1939 * *msg*. If the message has a large payload, then setting up
1940 * and calling the eBPF program repeatedly for all bytes, even
1941 * though the verdict is already known, would create unnecessary
1942 * overhead.
1943 *
1944 * When called from within an eBPF program, the helper sets a
1945 * counter internal to the BPF infrastructure, that is used to
1946 * apply the last verdict to the next *bytes*. If *bytes* is
1947 * smaller than the current data being processed from a
1948 * **sendmsg**\ () or **sendfile**\ () system call, the first
1949 * *bytes* will be sent and the eBPF program will be re-run with
1950 * the pointer for start of data pointing to byte number *bytes*
1951 * **+ 1**. If *bytes* is larger than the current data being
1952 * processed, then the eBPF verdict will be applied to multiple
1953 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1954 * consumed.
1955 *
1956 * Note that if a socket closes with the internal counter holding
1957 * a non-zero value, this is not a problem because data is not
1958 * being buffered for *bytes* and is sent as it is received.
1959 * Return
1960 * 0
1961 *
1962 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1963 * Description
1964 * For socket policies, prevent the execution of the verdict eBPF
1965 * program for message *msg* until *bytes* (byte number) have been
1966 * accumulated.
1967 *
1968 * This can be used when one needs a specific number of bytes
1969 * before a verdict can be assigned, even if the data spans
1970 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1971 * case would be a user calling **sendmsg**\ () repeatedly with
1972 * 1-byte long message segments. Obviously, this is bad for
1973 * performance, but it is still valid. If the eBPF program needs
1974 * *bytes* bytes to validate a header, this helper can be used to
1975 * prevent the eBPF program to be called again until *bytes* have
1976 * been accumulated.
1977 * Return
1978 * 0
1979 *
1980 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1981 * Description
1982 * For socket policies, pull in non-linear data from user space
1983 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1984 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1985 * respectively.
1986 *
1987 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1988 * *msg* it can only parse data that the (**data**, **data_end**)
1989 * pointers have already consumed. For **sendmsg**\ () hooks this
1990 * is likely the first scatterlist element. But for calls relying
1991 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1992 * be the range (**0**, **0**) because the data is shared with
1993 * user space and by default the objective is to avoid allowing
1994 * user space to modify data while (or after) eBPF verdict is
1995 * being decided. This helper can be used to pull in data and to
1996 * set the start and end pointer to given values. Data will be
1997 * copied if necessary (i.e. if data was not linear and if start
1998 * and end pointers do not point to the same chunk).
1999 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002000 * A call to this helper is susceptible to change the underlying
Quentin Monnetab127042018-04-25 18:16:59 +01002001 * packet buffer. Therefore, at load time, all checks on pointers
2002 * previously done by the verifier are invalidated and must be
2003 * performed again, if the helper is used in combination with
2004 * direct packet access.
2005 *
2006 * All values for *flags* are reserved for future usage, and must
2007 * be left at zero.
2008 * Return
2009 * 0 on success, or a negative error in case of failure.
2010 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07002011 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01002012 * Description
2013 * Bind the socket associated to *ctx* to the address pointed by
2014 * *addr*, of length *addr_len*. This allows for making outgoing
2015 * connection from the desired IP address, which can be useful for
2016 * example when all processes inside a cgroup should use one
2017 * single IP address on a host that has multiple IP configured.
2018 *
2019 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2020 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
Stanislav Fomichev8086fba2020-05-08 10:46:11 -07002021 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2022 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2023 * behavior and lets the kernel efficiently pick up an unused
2024 * port as long as 4-tuple is unique. Passing non-zero port might
2025 * lead to degraded performance.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01002026 * Return
2027 * 0 on success, or a negative error in case of failure.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002028 *
2029 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2030 * Description
2031 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
Jesper Dangaard Brouerc8741e22020-05-14 12:51:25 +02002032 * possible to both shrink and grow the packet tail.
2033 * Shrink done via *delta* being a negative integer.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002034 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002035 * A call to this helper is susceptible to change the underlying
Quentin Monnet2d020dd2018-04-25 18:17:00 +01002036 * packet buffer. Therefore, at load time, all checks on pointers
2037 * previously done by the verifier are invalidated and must be
2038 * performed again, if the helper is used in combination with
2039 * direct packet access.
2040 * Return
2041 * 0 on success, or a negative error in case of failure.
2042 *
2043 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2044 * Description
2045 * Retrieve the XFRM state (IP transform framework, see also
2046 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2047 *
2048 * The retrieved value is stored in the **struct bpf_xfrm_state**
2049 * pointed by *xfrm_state* and of length *size*.
2050 *
2051 * All values for *flags* are reserved for future usage, and must
2052 * be left at zero.
2053 *
2054 * This helper is available only if the kernel was compiled with
2055 * **CONFIG_XFRM** configuration option.
2056 * Return
2057 * 0 on success, or a negative error in case of failure.
Yonghong Songc195651e2018-04-28 22:28:08 -07002058 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002059 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
Yonghong Songc195651e2018-04-28 22:28:08 -07002060 * Description
Quentin Monnet79552fb2018-04-30 11:39:04 +01002061 * Return a user or a kernel stack in bpf program provided buffer.
2062 * To achieve this, the helper needs *ctx*, which is a pointer
2063 * to the context on which the tracing program is executed.
2064 * To store the stacktrace, the bpf program provides *buf* with
2065 * a nonnegative *size*.
Yonghong Songc195651e2018-04-28 22:28:08 -07002066 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002067 * The last argument, *flags*, holds the number of stack frames to
2068 * skip (from 0 to 255), masked with
2069 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2070 * the following flags:
Yonghong Songc195651e2018-04-28 22:28:08 -07002071 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002072 * **BPF_F_USER_STACK**
2073 * Collect a user space stack instead of a kernel stack.
2074 * **BPF_F_USER_BUILD_ID**
2075 * Collect buildid+offset instead of ips for user stack,
2076 * only valid if **BPF_F_USER_STACK** is also specified.
Yonghong Songc195651e2018-04-28 22:28:08 -07002077 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002078 * **bpf_get_stack**\ () can collect up to
2079 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2080 * to sufficient large buffer size. Note that
2081 * this limit can be controlled with the **sysctl** program, and
2082 * that it should be manually increased in order to profile long
2083 * user stacks (such as stacks for Java programs). To do so, use:
Yonghong Songc195651e2018-04-28 22:28:08 -07002084 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002085 * ::
Yonghong Songc195651e2018-04-28 22:28:08 -07002086 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01002087 * # sysctl kernel.perf_event_max_stack=<new value>
Yonghong Songc195651e2018-04-28 22:28:08 -07002088 * Return
Quentin Monnet7a279e92018-05-29 12:27:44 +01002089 * A non-negative value equal to or less than *size* on success,
2090 * or a negative error in case of failure.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002091 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002092 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002093 * Description
2094 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2095 * it provides an easy way to load *len* bytes from *offset*
2096 * from the packet associated to *skb*, into the buffer pointed
2097 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2098 * a fifth argument *start_header* exists in order to select a
2099 * base offset to start from. *start_header* can be one of:
2100 *
2101 * **BPF_HDR_START_MAC**
2102 * Base offset to load data from is *skb*'s mac header.
2103 * **BPF_HDR_START_NET**
2104 * Base offset to load data from is *skb*'s network header.
2105 *
2106 * In general, "direct packet access" is the preferred method to
2107 * access packet data, however, this helper is in particular useful
2108 * in socket filters where *skb*\ **->data** does not always point
2109 * to the start of the mac header and where "direct packet access"
2110 * is not available.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002111 * Return
2112 * 0 on success, or a negative error in case of failure.
2113 *
David Ahern87f5fc72018-05-09 20:34:26 -07002114 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2115 * Description
2116 * Do FIB lookup in kernel tables using parameters in *params*.
2117 * If lookup is successful and result shows packet is to be
2118 * forwarded, the neighbor tables are searched for the nexthop.
2119 * If successful (ie., FIB lookup shows forwarding and nexthop
David Ahernfa898d72018-05-29 10:58:07 -07002120 * is resolved), the nexthop address is returned in ipv4_dst
2121 * or ipv6_dst based on family, smac is set to mac address of
2122 * egress device, dmac is set to nexthop mac address, rt_metric
David Ahern4c795792018-06-26 16:21:18 -07002123 * is set to metric from route (IPv4/IPv6 only), and ifindex
2124 * is set to the device index of the nexthop from the FIB lookup.
David Ahern87f5fc72018-05-09 20:34:26 -07002125 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002126 * *plen* argument is the size of the passed in struct.
2127 * *flags* argument can be a combination of one or more of the
2128 * following values:
David Ahern87f5fc72018-05-09 20:34:26 -07002129 *
Quentin Monnet7a279e92018-05-29 12:27:44 +01002130 * **BPF_FIB_LOOKUP_DIRECT**
2131 * Do a direct table lookup vs full lookup using FIB
2132 * rules.
2133 * **BPF_FIB_LOOKUP_OUTPUT**
2134 * Perform lookup from an egress perspective (default is
2135 * ingress).
David Ahern87f5fc72018-05-09 20:34:26 -07002136 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002137 * *ctx* is either **struct xdp_md** for XDP programs or
2138 * **struct sk_buff** tc cls_act programs.
2139 * Return
David Ahern4c795792018-06-26 16:21:18 -07002140 * * < 0 if any input argument is invalid
2141 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2142 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
Quentin Monnet2bae79d2018-07-12 12:52:22 +01002143 * packet is not forwarded or needs assist from full stack
John Fastabend81110382018-05-14 10:00:17 -07002144 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002145 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
John Fastabend81110382018-05-14 10:00:17 -07002146 * Description
2147 * Add an entry to, or update a sockhash *map* referencing sockets.
2148 * The *skops* is used as a new value for the entry associated to
2149 * *key*. *flags* is one of:
2150 *
2151 * **BPF_NOEXIST**
2152 * The entry for *key* must not exist in the map.
2153 * **BPF_EXIST**
2154 * The entry for *key* must already exist in the map.
2155 * **BPF_ANY**
2156 * No condition on the existence of the entry for *key*.
2157 *
2158 * If the *map* has eBPF programs (parser and verdict), those will
2159 * be inherited by the socket being added. If the socket is
2160 * already attached to eBPF programs, this results in an error.
2161 * Return
2162 * 0 on success, or a negative error in case of failure.
2163 *
2164 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2165 * Description
2166 * This helper is used in programs implementing policies at the
2167 * socket level. If the message *msg* is allowed to pass (i.e. if
2168 * the verdict eBPF program returns **SK_PASS**), redirect it to
2169 * the socket referenced by *map* (of type
2170 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2171 * egress interfaces can be used for redirection. The
2172 * **BPF_F_INGRESS** value in *flags* is used to make the
2173 * distinction (ingress path is selected if the flag is present,
2174 * egress path otherwise). This is the only flag supported for now.
2175 * Return
2176 * **SK_PASS** on success, or **SK_DROP** on error.
2177 *
2178 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2179 * Description
2180 * This helper is used in programs implementing policies at the
2181 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2182 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2183 * to the socket referenced by *map* (of type
2184 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2185 * egress interfaces can be used for redirection. The
2186 * **BPF_F_INGRESS** value in *flags* is used to make the
2187 * distinction (ingress path is selected if the flag is present,
2188 * egress otherwise). This is the only flag supported for now.
2189 * Return
2190 * **SK_PASS** on success, or **SK_DROP** on error.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002191 *
2192 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2193 * Description
2194 * Encapsulate the packet associated to *skb* within a Layer 3
2195 * protocol header. This header is provided in the buffer at
2196 * address *hdr*, with *len* its size in bytes. *type* indicates
2197 * the protocol of the header and can be one of:
2198 *
2199 * **BPF_LWT_ENCAP_SEG6**
2200 * IPv6 encapsulation with Segment Routing Header
2201 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2202 * the IPv6 header is computed by the kernel.
2203 * **BPF_LWT_ENCAP_SEG6_INLINE**
2204 * Only works if *skb* contains an IPv6 packet. Insert a
2205 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2206 * the IPv6 header.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002207 * **BPF_LWT_ENCAP_IP**
2208 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2209 * must be IPv4 or IPv6, followed by zero or more
Quentin Monnet80867c52019-05-10 15:51:24 +01002210 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2211 * total bytes in all prepended headers. Please note that
2212 * if **skb_is_gso**\ (*skb*) is true, no more than two
2213 * headers can be prepended, and the inner header, if
2214 * present, should be either GRE or UDP/GUE.
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08002215 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002216 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2217 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2218 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2219 * **BPF_PROG_TYPE_LWT_XMIT**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002220 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002221 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002222 * packet buffer. Therefore, at load time, all checks on pointers
2223 * previously done by the verifier are invalidated and must be
2224 * performed again, if the helper is used in combination with
2225 * direct packet access.
2226 * Return
2227 * 0 on success, or a negative error in case of failure.
2228 *
2229 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2230 * Description
2231 * Store *len* bytes from address *from* into the packet
2232 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2233 * inside the outermost IPv6 Segment Routing Header can be
2234 * modified through this helper.
2235 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002236 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002237 * packet buffer. Therefore, at load time, all checks on pointers
2238 * previously done by the verifier are invalidated and must be
2239 * performed again, if the helper is used in combination with
2240 * direct packet access.
2241 * Return
2242 * 0 on success, or a negative error in case of failure.
2243 *
2244 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2245 * Description
2246 * Adjust the size allocated to TLVs in the outermost IPv6
2247 * Segment Routing Header contained in the packet associated to
2248 * *skb*, at position *offset* by *delta* bytes. Only offsets
2249 * after the segments are accepted. *delta* can be as well
2250 * positive (growing) as negative (shrinking).
2251 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002252 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002253 * packet buffer. Therefore, at load time, all checks on pointers
2254 * previously done by the verifier are invalidated and must be
2255 * performed again, if the helper is used in combination with
2256 * direct packet access.
2257 * Return
2258 * 0 on success, or a negative error in case of failure.
2259 *
2260 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2261 * Description
2262 * Apply an IPv6 Segment Routing action of type *action* to the
2263 * packet associated to *skb*. Each action takes a parameter
2264 * contained at address *param*, and of length *param_len* bytes.
2265 * *action* can be one of:
2266 *
2267 * **SEG6_LOCAL_ACTION_END_X**
2268 * End.X action: Endpoint with Layer-3 cross-connect.
2269 * Type of *param*: **struct in6_addr**.
2270 * **SEG6_LOCAL_ACTION_END_T**
2271 * End.T action: Endpoint with specific IPv6 table lookup.
2272 * Type of *param*: **int**.
2273 * **SEG6_LOCAL_ACTION_END_B6**
2274 * End.B6 action: Endpoint bound to an SRv6 policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002275 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002276 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2277 * End.B6.Encap action: Endpoint bound to an SRv6
2278 * encapsulation policy.
Quentin Monnet80867c52019-05-10 15:51:24 +01002279 * Type of *param*: **struct ipv6_sr_hdr**.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002280 *
Quentin Monnet32e7dc22019-05-10 15:51:23 +01002281 * A call to this helper is susceptible to change the underlying
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002282 * packet buffer. Therefore, at load time, all checks on pointers
2283 * previously done by the verifier are invalidated and must be
2284 * performed again, if the helper is used in combination with
2285 * direct packet access.
2286 * Return
2287 * 0 on success, or a negative error in case of failure.
Sean Youngf4364dc2018-05-27 12:24:09 +01002288 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002289 * int bpf_rc_repeat(void *ctx)
2290 * Description
2291 * This helper is used in programs implementing IR decoding, to
2292 * report a successfully decoded repeat key message. This delays
2293 * the generation of a key up event for previously generated
2294 * key down event.
2295 *
2296 * Some IR protocols like NEC have a special IR message for
2297 * repeating last button, for when a button is held down.
2298 *
2299 * The *ctx* should point to the lirc sample as passed into
2300 * the program.
2301 *
2302 * This helper is only available is the kernel was compiled with
2303 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2304 * "**y**".
2305 * Return
2306 * 0
2307 *
Sean Youngf4364dc2018-05-27 12:24:09 +01002308 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2309 * Description
2310 * This helper is used in programs implementing IR decoding, to
2311 * report a successfully decoded key press with *scancode*,
2312 * *toggle* value in the given *protocol*. The scancode will be
2313 * translated to a keycode using the rc keymap, and reported as
2314 * an input key down event. After a period a key up event is
2315 * generated. This period can be extended by calling either
Quentin Monnet90b10232018-12-03 12:13:35 +00002316 * **bpf_rc_keydown**\ () again with the same values, or calling
2317 * **bpf_rc_repeat**\ ().
Sean Youngf4364dc2018-05-27 12:24:09 +01002318 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01002319 * Some protocols include a toggle bit, in case the button was
Sean Youngf4364dc2018-05-27 12:24:09 +01002320 * released and pressed again between consecutive scancodes.
2321 *
2322 * The *ctx* should point to the lirc sample as passed into
2323 * the program.
2324 *
2325 * The *protocol* is the decoded protocol number (see
2326 * **enum rc_proto** for some predefined values).
2327 *
2328 * This helper is only available is the kernel was compiled with
2329 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2330 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002331 * Return
2332 * 0
2333 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002334 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002335 * Description
2336 * Return the cgroup v2 id of the socket associated with the *skb*.
2337 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2338 * helper for cgroup v1 by providing a tag resp. identifier that
2339 * can be matched on or used for map lookups e.g. to implement
2340 * policy. The cgroup v2 id of a given path in the hierarchy is
2341 * exposed in user space through the f_handle API in order to get
2342 * to the same 64-bit id.
2343 *
2344 * This helper can be used on TC egress path, but not on ingress,
2345 * and is available only if the kernel was compiled with the
2346 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2347 * Return
2348 * The id is returned or 0 in case the id could not be retrieved.
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002349 *
2350 * u64 bpf_get_current_cgroup_id(void)
2351 * Return
2352 * A 64-bit integer containing the current cgroup id based
2353 * on the cgroup within which the current task is running.
Roman Gushchincd339432018-08-02 14:27:24 -07002354 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002355 * void *bpf_get_local_storage(void *map, u64 flags)
Roman Gushchincd339432018-08-02 14:27:24 -07002356 * Description
2357 * Get the pointer to the local storage area.
2358 * The type and the size of the local storage is defined
2359 * by the *map* argument.
2360 * The *flags* meaning is specific for each map type,
2361 * and has to be 0 for cgroup local storage.
2362 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002363 * Depending on the BPF program type, a local storage area
2364 * can be shared between multiple instances of the BPF program,
Roman Gushchincd339432018-08-02 14:27:24 -07002365 * running simultaneously.
2366 *
2367 * A user should care about the synchronization by himself.
Quentin Monnet90b10232018-12-03 12:13:35 +00002368 * For example, by using the **BPF_STX_XADD** instruction to alter
Roman Gushchincd339432018-08-02 14:27:24 -07002369 * the shared data.
2370 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002371 * A pointer to the local storage area.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002372 *
2373 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2374 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002375 * Select a **SO_REUSEPORT** socket from a
2376 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2377 * It checks the selected socket is matching the incoming
2378 * request in the socket buffer.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002379 * Return
2380 * 0 on success, or a negative error in case of failure.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002381 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002382 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2383 * Description
2384 * Return id of cgroup v2 that is ancestor of cgroup associated
2385 * with the *skb* at the *ancestor_level*. The root cgroup is at
2386 * *ancestor_level* zero and each step down the hierarchy
2387 * increments the level. If *ancestor_level* == level of cgroup
2388 * associated with *skb*, then return value will be same as that
2389 * of **bpf_skb_cgroup_id**\ ().
2390 *
2391 * The helper is useful to implement policies based on cgroups
2392 * that are upper in hierarchy than immediate cgroup associated
2393 * with *skb*.
2394 *
2395 * The format of returned id and helper limitations are same as in
2396 * **bpf_skb_cgroup_id**\ ().
2397 * Return
2398 * The id is returned or 0 in case the id could not be retrieved.
2399 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002400 * 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 -07002401 * Description
2402 * Look for TCP socket matching *tuple*, optionally in a child
2403 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002404 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002405 *
2406 * The *ctx* should point to the context of the program, such as
2407 * the skb or socket (depending on the hook in use). This is used
2408 * to determine the base network namespace for the lookup.
2409 *
2410 * *tuple_size* must be one of:
2411 *
2412 * **sizeof**\ (*tuple*\ **->ipv4**)
2413 * Look for an IPv4 socket.
2414 * **sizeof**\ (*tuple*\ **->ipv6**)
2415 * Look for an IPv6 socket.
2416 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002417 * If the *netns* is a negative signed 32-bit integer, then the
2418 * socket lookup table in the netns associated with the *ctx* will
2419 * will be used. For the TC hooks, this is the netns of the device
2420 * in the skb. For socket hooks, this is the netns of the socket.
2421 * If *netns* is any other signed 32-bit value greater than or
2422 * equal to zero then it specifies the ID of the netns relative to
2423 * the netns associated with the *ctx*. *netns* values beyond the
2424 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002425 *
2426 * All values for *flags* are reserved for future usage, and must
2427 * be left at zero.
2428 *
2429 * This helper is available only if the kernel was compiled with
2430 * **CONFIG_NET** configuration option.
2431 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002432 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2433 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002434 * result is from *reuse*\ **->socks**\ [] using the hash of the
2435 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002436 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002437 * 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 -07002438 * Description
2439 * Look for UDP socket matching *tuple*, optionally in a child
2440 * network namespace *netns*. The return value must be checked,
Quentin Monnet90b10232018-12-03 12:13:35 +00002441 * and if non-**NULL**, released via **bpf_sk_release**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002442 *
2443 * The *ctx* should point to the context of the program, such as
2444 * the skb or socket (depending on the hook in use). This is used
2445 * to determine the base network namespace for the lookup.
2446 *
2447 * *tuple_size* must be one of:
2448 *
2449 * **sizeof**\ (*tuple*\ **->ipv4**)
2450 * Look for an IPv4 socket.
2451 * **sizeof**\ (*tuple*\ **->ipv6**)
2452 * Look for an IPv6 socket.
2453 *
Joe Stringerf71c6142018-11-30 15:32:20 -08002454 * If the *netns* is a negative signed 32-bit integer, then the
2455 * socket lookup table in the netns associated with the *ctx* will
2456 * will be used. For the TC hooks, this is the netns of the device
2457 * in the skb. For socket hooks, this is the netns of the socket.
2458 * If *netns* is any other signed 32-bit value greater than or
2459 * equal to zero then it specifies the ID of the netns relative to
2460 * the netns associated with the *ctx*. *netns* values beyond the
2461 * range of 32-bit integers are reserved for future use.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002462 *
2463 * All values for *flags* are reserved for future usage, and must
2464 * be left at zero.
2465 *
2466 * This helper is available only if the kernel was compiled with
2467 * **CONFIG_NET** configuration option.
2468 * Return
Daniel Borkmann0bd72112018-12-11 10:26:33 +01002469 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2470 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002471 * result is from *reuse*\ **->socks**\ [] using the hash of the
2472 * tuple.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002473 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002474 * int bpf_sk_release(struct bpf_sock *sock)
Joe Stringer6acc9b42018-10-02 13:35:36 -07002475 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002476 * Release the reference held by *sock*. *sock* must be a
2477 * non-**NULL** pointer that was returned from
2478 * **bpf_sk_lookup_xxx**\ ().
Joe Stringer6acc9b42018-10-02 13:35:36 -07002479 * Return
2480 * 0 on success, or a negative error in case of failure.
John Fastabend6fff6072018-10-19 19:56:49 -07002481 *
Quentin Monnet62369db2019-03-14 12:38:39 +00002482 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2483 * Description
2484 * Push an element *value* in *map*. *flags* is one of:
2485 *
2486 * **BPF_EXIST**
2487 * If the queue/stack is full, the oldest element is
2488 * removed to make room for this.
2489 * Return
2490 * 0 on success, or a negative error in case of failure.
2491 *
Quentin Monnet90b10232018-12-03 12:13:35 +00002492 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2493 * Description
2494 * Pop an element from *map*.
2495 * Return
2496 * 0 on success, or a negative error in case of failure.
2497 *
2498 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2499 * Description
2500 * Get an element from *map* without removing it.
2501 * Return
2502 * 0 on success, or a negative error in case of failure.
2503 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002504 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
John Fastabend6fff6072018-10-19 19:56:49 -07002505 * Description
Quentin Monnet90b10232018-12-03 12:13:35 +00002506 * For socket policies, insert *len* bytes into *msg* at offset
John Fastabend6fff6072018-10-19 19:56:49 -07002507 * *start*.
2508 *
2509 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
Quentin Monnet90b10232018-12-03 12:13:35 +00002510 * *msg* it may want to insert metadata or options into the *msg*.
John Fastabend6fff6072018-10-19 19:56:49 -07002511 * This can later be read and used by any of the lower layer BPF
2512 * hooks.
2513 *
2514 * This helper may fail if under memory pressure (a malloc
2515 * fails) in these cases BPF programs will get an appropriate
2516 * error and BPF programs will need to handle them.
John Fastabend6fff6072018-10-19 19:56:49 -07002517 * Return
2518 * 0 on success, or a negative error in case of failure.
John Fastabend7246d8e2018-11-26 14:16:17 -08002519 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002520 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
Quentin Monnet90b10232018-12-03 12:13:35 +00002521 * Description
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002522 * Will remove *len* bytes from a *msg* starting at byte *start*.
John Fastabend7246d8e2018-11-26 14:16:17 -08002523 * This may result in **ENOMEM** errors under certain situations if
2524 * an allocation and copy are required due to a full ring buffer.
2525 * However, the helper will try to avoid doing the allocation
2526 * if possible. Other errors can occur if input parameters are
Quentin Monnet90b10232018-12-03 12:13:35 +00002527 * invalid either due to *start* byte not being valid part of *msg*
John Fastabend7246d8e2018-11-26 14:16:17 -08002528 * payload and/or *pop* value being to large.
John Fastabend7246d8e2018-11-26 14:16:17 -08002529 * Return
Quentin Monnet90b10232018-12-03 12:13:35 +00002530 * 0 on success, or a negative error in case of failure.
Sean Young01d32402018-12-06 13:01:03 +00002531 *
2532 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2533 * Description
2534 * This helper is used in programs implementing IR decoding, to
2535 * report a successfully decoded pointer movement.
2536 *
2537 * The *ctx* should point to the lirc sample as passed into
2538 * the program.
2539 *
2540 * This helper is only available is the kernel was compiled with
2541 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2542 * "**y**".
2543 * Return
2544 * 0
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002545 *
Quentin Monnet0eb09782019-03-14 12:38:40 +00002546 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2547 * Description
2548 * Acquire a spinlock represented by the pointer *lock*, which is
2549 * stored as part of a value of a map. Taking the lock allows to
2550 * safely update the rest of the fields in that value. The
2551 * spinlock can (and must) later be released with a call to
2552 * **bpf_spin_unlock**\ (\ *lock*\ ).
2553 *
2554 * Spinlocks in BPF programs come with a number of restrictions
2555 * and constraints:
2556 *
2557 * * **bpf_spin_lock** objects are only allowed inside maps of
2558 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2559 * list could be extended in the future).
2560 * * BTF description of the map is mandatory.
2561 * * The BPF program can take ONE lock at a time, since taking two
2562 * or more could cause dead locks.
2563 * * Only one **struct bpf_spin_lock** is allowed per map element.
2564 * * When the lock is taken, calls (either BPF to BPF or helpers)
2565 * are not allowed.
2566 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2567 * allowed inside a spinlock-ed region.
2568 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2569 * the lock, on all execution paths, before it returns.
2570 * * The BPF program can access **struct bpf_spin_lock** only via
2571 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2572 * helpers. Loading or storing data into the **struct
2573 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2574 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2575 * of the map value must be a struct and have **struct
2576 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2577 * Nested lock inside another struct is not allowed.
2578 * * The **struct bpf_spin_lock** *lock* field in a map value must
2579 * be aligned on a multiple of 4 bytes in that value.
2580 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2581 * the **bpf_spin_lock** field to user space.
2582 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2583 * a BPF program, do not update the **bpf_spin_lock** field.
2584 * * **bpf_spin_lock** cannot be on the stack or inside a
2585 * networking packet (it can only be inside of a map values).
2586 * * **bpf_spin_lock** is available to root only.
2587 * * Tracing programs and socket filter programs cannot use
2588 * **bpf_spin_lock**\ () due to insufficient preemption checks
2589 * (but this may change in the future).
2590 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2591 * Return
2592 * 0
2593 *
2594 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2595 * Description
2596 * Release the *lock* previously locked by a call to
2597 * **bpf_spin_lock**\ (\ *lock*\ ).
2598 * Return
2599 * 0
2600 *
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002601 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2602 * Description
2603 * This helper gets a **struct bpf_sock** pointer such
Quentin Monnet62369db2019-03-14 12:38:39 +00002604 * that all the fields in this **bpf_sock** can be accessed.
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002605 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002606 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08002607 * case of failure.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002608 *
2609 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2610 * Description
2611 * This helper gets a **struct bpf_tcp_sock** pointer from a
2612 * **struct bpf_sock** pointer.
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002613 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002614 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08002615 * case of failure.
brakmof7c917b2019-03-01 12:38:46 -08002616 *
Andrii Nakryiko5f0e5412019-10-06 20:07:36 -07002617 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
Quentin Monnet62369db2019-03-14 12:38:39 +00002618 * Description
2619 * Set ECN (Explicit Congestion Notification) field of IP header
2620 * to **CE** (Congestion Encountered) if current value is **ECT**
2621 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2622 * and IPv4.
2623 * Return
2624 * 1 if the **CE** flag is set (either by the current helper call
2625 * or because it was already present), 0 if it is not set.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002626 *
2627 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2628 * Description
Quentin Monnet62369db2019-03-14 12:38:39 +00002629 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2630 * **bpf_sk_release**\ () is unnecessary and not allowed.
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002631 * Return
Quentin Monnet62369db2019-03-14 12:38:39 +00002632 * A **struct bpf_sock** pointer on success, or **NULL** in
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07002633 * case of failure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002634 *
2635 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2636 * Description
2637 * Look for TCP socket matching *tuple*, optionally in a child
2638 * network namespace *netns*. The return value must be checked,
2639 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2640 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002641 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2642 * that it also returns timewait or request sockets. Use
2643 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2644 * full structure.
Lorenz Baueredbf8c02019-03-22 09:54:01 +08002645 *
2646 * This helper is available only if the kernel was compiled with
2647 * **CONFIG_NET** configuration option.
2648 * Return
2649 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2650 * For sockets with reuseport option, the **struct bpf_sock**
Quentin Monnet80867c52019-05-10 15:51:24 +01002651 * result is from *reuse*\ **->socks**\ [] using the hash of the
2652 * tuple.
Lorenz Bauer39904082019-03-22 09:54:02 +08002653 *
2654 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2655 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002656 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2657 * the listening socket in *sk*.
Lorenz Bauer39904082019-03-22 09:54:02 +08002658 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002659 * *iph* points to the start of the IPv4 or IPv6 header, while
2660 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2661 * **sizeof**\ (**struct ip6hdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002662 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002663 * *th* points to the start of the TCP header, while *th_len*
2664 * contains **sizeof**\ (**struct tcphdr**).
Lorenz Bauer39904082019-03-22 09:54:02 +08002665 * Return
Quentin Monnet80867c52019-05-10 15:51:24 +01002666 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2667 * error otherwise.
Andrey Ignatov808649f2019-02-27 13:28:48 -08002668 *
2669 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2670 * Description
2671 * Get name of sysctl in /proc/sys/ and copy it into provided by
2672 * program buffer *buf* of size *buf_len*.
2673 *
2674 * The buffer is always NUL terminated, unless it's zero-sized.
2675 *
2676 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2677 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2678 * only (e.g. "tcp_mem").
2679 * Return
2680 * Number of character copied (not including the trailing NUL).
2681 *
2682 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2683 * truncated name in this case).
Andrey Ignatov1d11b302019-02-28 19:22:15 -08002684 *
2685 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2686 * Description
2687 * Get current value of sysctl as it is presented in /proc/sys
2688 * (incl. newline, etc), and copy it as a string into provided
2689 * by program buffer *buf* of size *buf_len*.
2690 *
2691 * The whole value is copied, no matter what file position user
2692 * space issued e.g. sys_read at.
2693 *
2694 * The buffer is always NUL terminated, unless it's zero-sized.
2695 * Return
2696 * Number of character copied (not including the trailing NUL).
2697 *
2698 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2699 * truncated name in this case).
2700 *
2701 * **-EINVAL** if current value was unavailable, e.g. because
2702 * sysctl is uninitialized and read returns -EIO for it.
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08002703 *
2704 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2705 * Description
2706 * Get new value being written by user space to sysctl (before
2707 * the actual write happens) and copy it as a string into
2708 * provided by program buffer *buf* of size *buf_len*.
2709 *
2710 * User space may write new value at file position > 0.
2711 *
2712 * The buffer is always NUL terminated, unless it's zero-sized.
2713 * Return
2714 * Number of character copied (not including the trailing NUL).
2715 *
2716 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2717 * truncated name in this case).
2718 *
2719 * **-EINVAL** if sysctl is being read.
2720 *
2721 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2722 * Description
2723 * Override new value being written by user space to sysctl with
2724 * value provided by program in buffer *buf* of size *buf_len*.
2725 *
2726 * *buf* should contain a string in same form as provided by user
2727 * space on sysctl write.
2728 *
2729 * User space may write new value at file position > 0. To override
2730 * the whole sysctl value file position should be set to zero.
2731 * Return
2732 * 0 on success.
2733 *
2734 * **-E2BIG** if the *buf_len* is too big.
2735 *
2736 * **-EINVAL** if sysctl is being read.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002737 *
2738 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2739 * Description
2740 * Convert the initial part of the string from buffer *buf* of
2741 * size *buf_len* to a long integer according to the given base
2742 * and save the result in *res*.
2743 *
2744 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002745 * (as determined by **isspace**\ (3)) followed by a single
2746 * optional '**-**' sign.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002747 *
2748 * Five least significant bits of *flags* encode base, other bits
2749 * are currently unused.
2750 *
2751 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002752 * similar to user space **strtol**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002753 * Return
2754 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002755 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002756 *
2757 * **-EINVAL** if no valid digits were found or unsupported base
2758 * was provided.
2759 *
2760 * **-ERANGE** if resulting value was out of range.
2761 *
2762 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2763 * Description
2764 * Convert the initial part of the string from buffer *buf* of
2765 * size *buf_len* to an unsigned long integer according to the
2766 * given base and save the result in *res*.
2767 *
2768 * The string may begin with an arbitrary amount of white space
Quentin Monnet80867c52019-05-10 15:51:24 +01002769 * (as determined by **isspace**\ (3)).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002770 *
2771 * Five least significant bits of *flags* encode base, other bits
2772 * are currently unused.
2773 *
2774 * Base must be either 8, 10, 16 or 0 to detect it automatically
Quentin Monnet80867c52019-05-10 15:51:24 +01002775 * similar to user space **strtoul**\ (3).
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002776 * Return
2777 * Number of characters consumed on success. Must be positive but
Quentin Monnet80867c52019-05-10 15:51:24 +01002778 * no more than *buf_len*.
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07002779 *
2780 * **-EINVAL** if no valid digits were found or unsupported base
2781 * was provided.
2782 *
2783 * **-ERANGE** if resulting value was out of range.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002784 *
2785 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2786 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002787 * Get a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002788 *
2789 * Logically, it could be thought of getting the value from
2790 * a *map* with *sk* as the **key**. From this
2791 * perspective, the usage is not much different from
Quentin Monnet80867c52019-05-10 15:51:24 +01002792 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2793 * helper enforces the key must be a full socket and the map must
2794 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002795 *
2796 * Underneath, the value is stored locally at *sk* instead of
Quentin Monnet80867c52019-05-10 15:51:24 +01002797 * the *map*. The *map* is used as the bpf-local-storage
2798 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2799 * searched against all bpf-local-storages residing at *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002800 *
Quentin Monnet80867c52019-05-10 15:51:24 +01002801 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002802 * used such that a new bpf-local-storage will be
2803 * created if one does not exist. *value* can be used
Quentin Monnet80867c52019-05-10 15:51:24 +01002804 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002805 * the initial value of a bpf-local-storage. If *value* is
Quentin Monnet80867c52019-05-10 15:51:24 +01002806 * **NULL**, the new bpf-local-storage will be zero initialized.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002807 * Return
2808 * A bpf-local-storage pointer is returned on success.
2809 *
2810 * **NULL** if not found or there was an error in adding
2811 * a new bpf-local-storage.
2812 *
2813 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2814 * Description
Quentin Monnet80867c52019-05-10 15:51:24 +01002815 * Delete a bpf-local-storage from a *sk*.
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07002816 * Return
2817 * 0 on success.
2818 *
2819 * **-ENOENT** if the bpf-local-storage cannot be found.
Yonghong Song8b401f92019-05-23 14:47:45 -07002820 *
2821 * int bpf_send_signal(u32 sig)
2822 * Description
Yonghong Song84829412020-01-14 19:50:02 -08002823 * Send signal *sig* to the process of the current task.
2824 * The signal may be delivered to any of this process's threads.
Yonghong Song8b401f92019-05-23 14:47:45 -07002825 * Return
2826 * 0 on success or successfully queued.
2827 *
2828 * **-EBUSY** if work queue under nmi is full.
2829 *
2830 * **-EINVAL** if *sig* is invalid.
2831 *
2832 * **-EPERM** if no permission to send the *sig*.
2833 *
2834 * **-EAGAIN** if bpf program can try again.
Petar Penkov70d66242019-07-29 09:59:15 -07002835 *
2836 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2837 * Description
2838 * Try to issue a SYN cookie for the packet with corresponding
2839 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2840 *
2841 * *iph* points to the start of the IPv4 or IPv6 header, while
2842 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2843 * **sizeof**\ (**struct ip6hdr**).
2844 *
2845 * *th* points to the start of the TCP header, while *th_len*
2846 * contains the length of the TCP header.
Petar Penkov70d66242019-07-29 09:59:15 -07002847 * Return
2848 * On success, lower 32 bits hold the generated SYN cookie in
2849 * followed by 16 bits which hold the MSS value for that cookie,
2850 * and the top 16 bits are unused.
2851 *
2852 * On failure, the returned value is one of the following:
2853 *
2854 * **-EINVAL** SYN cookie cannot be issued due to error
2855 *
2856 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2857 *
2858 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2859 *
2860 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
Alexei Starovoitova7658e12019-10-15 20:25:04 -07002861 *
2862 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2863 * Description
2864 * Write raw *data* blob into a special BPF perf event held by
2865 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2866 * event must have the following attributes: **PERF_SAMPLE_RAW**
2867 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2868 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2869 *
2870 * The *flags* are used to indicate the index in *map* for which
2871 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2872 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2873 * to indicate that the index of the current CPU core should be
2874 * used.
2875 *
2876 * The value to write, of *size*, is passed through eBPF stack and
2877 * pointed by *data*.
2878 *
2879 * *ctx* is a pointer to in-kernel struct sk_buff.
2880 *
2881 * This helper is similar to **bpf_perf_event_output**\ () but
2882 * restricted to raw_tracepoint bpf programs.
2883 * Return
2884 * 0 on success, or a negative error in case of failure.
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002885 *
2886 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2887 * Description
2888 * Safely attempt to read *size* bytes from user space address
2889 * *unsafe_ptr* and store the data in *dst*.
2890 * Return
2891 * 0 on success, or a negative error in case of failure.
2892 *
2893 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2894 * Description
2895 * Safely attempt to read *size* bytes from kernel space address
2896 * *unsafe_ptr* and store the data in *dst*.
2897 * Return
2898 * 0 on success, or a negative error in case of failure.
2899 *
2900 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2901 * Description
2902 * Copy a NUL terminated string from an unsafe user address
2903 * *unsafe_ptr* to *dst*. The *size* should include the
2904 * terminating NUL byte. In case the string length is smaller than
2905 * *size*, the target is not padded with further NUL bytes. If the
2906 * string length is larger than *size*, just *size*-1 bytes are
2907 * copied and the last byte is set to NUL.
2908 *
2909 * On success, the length of the copied string is returned. This
2910 * makes this helper useful in tracing programs for reading
2911 * strings, and more importantly to get its length at runtime. See
2912 * the following snippet:
2913 *
2914 * ::
2915 *
2916 * SEC("kprobe/sys_open")
2917 * void bpf_sys_open(struct pt_regs *ctx)
2918 * {
2919 * char buf[PATHLEN]; // PATHLEN is defined to 256
2920 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2921 * ctx->di);
2922 *
2923 * // Consume buf, for example push it to
2924 * // userspace via bpf_perf_event_output(); we
2925 * // can use res (the string length) as event
2926 * // size, after checking its boundaries.
2927 * }
2928 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01002929 * In comparison, using **bpf_probe_read_user**\ () helper here
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002930 * instead to read the string would require to estimate the length
2931 * at compile time, and would often result in copying more memory
2932 * than necessary.
2933 *
2934 * Another useful use case is when parsing individual process
2935 * arguments or individual environment variables navigating
2936 * *current*\ **->mm->arg_start** and *current*\
2937 * **->mm->env_start**: using this helper and the return value,
2938 * one can quickly iterate at the right offset of the memory area.
2939 * Return
2940 * On success, the strictly positive length of the string,
2941 * including the trailing NUL character. On error, a negative
2942 * value.
2943 *
2944 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2945 * Description
2946 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
Quentin Monnetab8d7802020-05-11 17:15:35 +01002947 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002948 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01002949 * On success, the strictly positive length of the string, including
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01002950 * the trailing NUL character. On error, a negative value.
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002951 *
2952 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2953 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01002954 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
Martin KaFai Lau206057f2020-01-08 16:45:51 -08002955 * *rcv_nxt* is the ack_seq to be sent out.
2956 * Return
2957 * 0 on success, or a negative error in case of failure.
2958 *
Yonghong Song84829412020-01-14 19:50:02 -08002959 * int bpf_send_signal_thread(u32 sig)
2960 * Description
2961 * Send signal *sig* to the thread corresponding to the current task.
2962 * Return
2963 * 0 on success or successfully queued.
2964 *
2965 * **-EBUSY** if work queue under nmi is full.
2966 *
2967 * **-EINVAL** if *sig* is invalid.
2968 *
2969 * **-EPERM** if no permission to send the *sig*.
2970 *
2971 * **-EAGAIN** if bpf program can try again.
Martin KaFai Lau5576b992020-01-22 15:36:46 -08002972 *
2973 * u64 bpf_jiffies64(void)
2974 * Description
2975 * Obtain the 64bit jiffies
2976 * Return
2977 * The 64 bit jiffies
Daniel Xufff7b642020-02-17 19:04:31 -08002978 *
2979 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2980 * Description
2981 * For an eBPF program attached to a perf event, retrieve the
Quentin Monnetab8d7802020-05-11 17:15:35 +01002982 * branch records (**struct perf_branch_entry**) associated to *ctx*
2983 * and store it in the buffer pointed by *buf* up to size
Daniel Xufff7b642020-02-17 19:04:31 -08002984 * *size* bytes.
2985 * Return
2986 * On success, number of bytes written to *buf*. On error, a
2987 * negative value.
2988 *
2989 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
Quentin Monnetab8d7802020-05-11 17:15:35 +01002990 * instead return the number of bytes required to store all the
Daniel Xufff7b642020-02-17 19:04:31 -08002991 * branch entries. If this flag is set, *buf* may be NULL.
2992 *
2993 * **-EINVAL** if arguments invalid or **size** not a multiple
Quentin Monnetab8d7802020-05-11 17:15:35 +01002994 * of **sizeof**\ (**struct perf_branch_entry**\ ).
Daniel Xufff7b642020-02-17 19:04:31 -08002995 *
2996 * **-ENOENT** if architecture does not support branch records.
Carlos Neirab4490c52020-03-04 17:41:56 -03002997 *
2998 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2999 * Description
3000 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3001 * *namespace* will be returned in *nsdata*.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003002 * Return
3003 * 0 on success, or one of the following in case of failure:
Carlos Neirab4490c52020-03-04 17:41:56 -03003004 *
3005 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3006 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3007 *
3008 * **-ENOENT** if pidns does not exists for the current task.
3009 *
Eelco Chaudrond831ee82020-03-06 08:59:23 +00003010 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3011 * Description
3012 * Write raw *data* blob into a special BPF perf event held by
3013 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3014 * event must have the following attributes: **PERF_SAMPLE_RAW**
3015 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3016 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3017 *
3018 * The *flags* are used to indicate the index in *map* for which
3019 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3020 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3021 * to indicate that the index of the current CPU core should be
3022 * used.
3023 *
3024 * The value to write, of *size*, is passed through eBPF stack and
3025 * pointed by *data*.
3026 *
3027 * *ctx* is a pointer to in-kernel struct xdp_buff.
3028 *
3029 * This helper is similar to **bpf_perf_eventoutput**\ () but
3030 * restricted to raw_tracepoint bpf programs.
3031 * Return
3032 * 0 on success, or a negative error in case of failure.
Daniel Borkmannf3189032020-03-27 16:58:52 +01003033 *
3034 * u64 bpf_get_netns_cookie(void *ctx)
3035 * Description
3036 * Retrieve the cookie (generated by the kernel) of the network
3037 * namespace the input *ctx* is associated with. The network
3038 * namespace cookie remains stable for its lifetime and provides
3039 * a global identifier that can be assumed unique. If *ctx* is
3040 * NULL, then the helper returns the cookie for the initial
3041 * network namespace. The cookie itself is very similar to that
Quentin Monnetab8d7802020-05-11 17:15:35 +01003042 * of **bpf_get_socket_cookie**\ () helper, but for network
3043 * namespaces instead of sockets.
Daniel Borkmannf3189032020-03-27 16:58:52 +01003044 * Return
3045 * A 8-byte long opaque number.
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003046 *
3047 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3048 * Description
3049 * Return id of cgroup v2 that is ancestor of the cgroup associated
3050 * with the current task at the *ancestor_level*. The root cgroup
3051 * is at *ancestor_level* zero and each step down the hierarchy
3052 * increments the level. If *ancestor_level* == level of cgroup
3053 * associated with the current task, then return value will be the
3054 * same as that of **bpf_get_current_cgroup_id**\ ().
3055 *
3056 * The helper is useful to implement policies based on cgroups
3057 * that are upper in hierarchy than immediate cgroup associated
3058 * with the current task.
3059 *
3060 * The format of returned id and helper limitations are same as in
3061 * **bpf_get_current_cgroup_id**\ ().
3062 * Return
3063 * The id is returned or 0 in case the id could not be retrieved.
Joe Stringercf7fbe62020-03-29 15:53:38 -07003064 *
3065 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3066 * Description
3067 * Assign the *sk* to the *skb*. When combined with appropriate
3068 * routing configuration to receive the packet towards the socket,
3069 * will cause *skb* to be delivered to the specified socket.
3070 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3071 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3072 * interfere with successful delivery to the socket.
3073 *
3074 * This operation is only valid from TC ingress path.
3075 *
3076 * The *flags* argument must be zero.
3077 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003078 * 0 on success, or a negative error in case of failure:
Joe Stringercf7fbe62020-03-29 15:53:38 -07003079 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003080 * **-EINVAL** if specified *flags* are not supported.
3081 *
3082 * **-ENOENT** if the socket is unavailable for assignment.
3083 *
3084 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3085 *
3086 * **-EOPNOTSUPP** if the operation is not supported, for example
3087 * a call from outside of TC ingress.
3088 *
3089 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3090 * (reuseport).
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003091 *
3092 * u64 bpf_ktime_get_boot_ns(void)
3093 * Description
3094 * Return the time elapsed since system boot, in nanoseconds.
3095 * Does include the time the system was suspended.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003096 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003097 * Return
3098 * Current *ktime*.
Yonghong Song492e6392020-05-09 10:59:14 -07003099 *
3100 * int bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3101 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01003102 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3103 * out the format string.
Yonghong Song492e6392020-05-09 10:59:14 -07003104 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3105 * the format string itself. The *data* and *data_len* are format string
Quentin Monnetab8d7802020-05-11 17:15:35 +01003106 * arguments. The *data* are a **u64** array and corresponding format string
Yonghong Song492e6392020-05-09 10:59:14 -07003107 * values are stored in the array. For strings and pointers where pointees
3108 * are accessed, only the pointer values are stored in the *data* array.
Quentin Monnetab8d7802020-05-11 17:15:35 +01003109 * The *data_len* is the size of *data* in bytes.
Yonghong Song492e6392020-05-09 10:59:14 -07003110 *
3111 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3112 * Reading kernel memory may fail due to either invalid address or
3113 * valid address but requiring a major memory fault. If reading kernel memory
3114 * fails, the string for **%s** will be an empty string, and the ip
3115 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
Quentin Monnetab8d7802020-05-11 17:15:35 +01003116 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
Yonghong Song492e6392020-05-09 10:59:14 -07003117 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003118 * 0 on success, or a negative error in case of failure:
Yonghong Song492e6392020-05-09 10:59:14 -07003119 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003120 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3121 * by returning 1 from bpf program.
3122 *
3123 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3124 *
3125 * **-E2BIG** if *fmt* contains too many format specifiers.
3126 *
3127 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
Yonghong Song492e6392020-05-09 10:59:14 -07003128 *
3129 * int bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3130 * Description
Quentin Monnetab8d7802020-05-11 17:15:35 +01003131 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
Yonghong Song492e6392020-05-09 10:59:14 -07003132 * The *m* represents the seq_file. The *data* and *len* represent the
Quentin Monnetab8d7802020-05-11 17:15:35 +01003133 * data to write in bytes.
Yonghong Song492e6392020-05-09 10:59:14 -07003134 * Return
Quentin Monnetab8d7802020-05-11 17:15:35 +01003135 * 0 on success, or a negative error in case of failure:
Yonghong Song492e6392020-05-09 10:59:14 -07003136 *
Quentin Monnetab8d7802020-05-11 17:15:35 +01003137 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
Andrey Ignatovf307fa22020-05-14 13:03:47 -07003138 *
3139 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3140 * Description
3141 * Return the cgroup v2 id of the socket *sk*.
3142 *
3143 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3144 * returned from **bpf_sk_lookup_xxx**\ (),
3145 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3146 * same as in **bpf_skb_cgroup_id**\ ().
3147 *
3148 * This helper is available only if the kernel was compiled with
3149 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3150 * Return
3151 * The id is returned or 0 in case the id could not be retrieved.
3152 *
3153 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3154 * Description
3155 * Return id of cgroup v2 that is ancestor of cgroup associated
3156 * with the *sk* at the *ancestor_level*. The root cgroup is at
3157 * *ancestor_level* zero and each step down the hierarchy
3158 * increments the level. If *ancestor_level* == level of cgroup
3159 * associated with *sk*, then return value will be same as that
3160 * of **bpf_sk_cgroup_id**\ ().
3161 *
3162 * The helper is useful to implement policies based on cgroups
3163 * that are upper in hierarchy than immediate cgroup associated
3164 * with *sk*.
3165 *
3166 * The format of returned id and helper limitations are same as in
3167 * **bpf_sk_cgroup_id**\ ().
3168 * Return
3169 * The id is returned or 0 in case the id could not be retrieved.
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003170 *
3171 * void *bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3172 * Description
3173 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3174 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3175 * new data availability is sent.
3176 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3177 * new data availability is sent unconditionally.
3178 * Return
3179 * 0, on success;
3180 * < 0, on error.
3181 *
3182 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3183 * Description
3184 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3185 * Return
3186 * Valid pointer with *size* bytes of memory available; NULL,
3187 * otherwise.
3188 *
3189 * void bpf_ringbuf_submit(void *data, u64 flags)
3190 * Description
3191 * Submit reserved ring buffer sample, pointed to by *data*.
3192 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3193 * new data availability is sent.
3194 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3195 * new data availability is sent unconditionally.
3196 * Return
3197 * Nothing. Always succeeds.
3198 *
3199 * void bpf_ringbuf_discard(void *data, u64 flags)
3200 * Description
3201 * Discard reserved ring buffer sample, pointed to by *data*.
3202 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3203 * new data availability is sent.
3204 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3205 * new data availability is sent unconditionally.
3206 * Return
3207 * Nothing. Always succeeds.
3208 *
3209 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3210 * Description
3211 * Query various characteristics of provided ring buffer. What
3212 * exactly is queries is determined by *flags*:
3213 * - BPF_RB_AVAIL_DATA - amount of data not yet consumed;
3214 * - BPF_RB_RING_SIZE - the size of ring buffer;
3215 * - BPF_RB_CONS_POS - consumer position (can wrap around);
3216 * - BPF_RB_PROD_POS - producer(s) position (can wrap around);
3217 * Data returned is just a momentary snapshots of actual values
3218 * and could be inaccurate, so this facility should be used to
3219 * power heuristics and for reporting, not to make 100% correct
3220 * calculation.
3221 * Return
3222 * Requested value, or 0, if flags are not recognized.
Thomas Grafebb676d2016-10-27 11:23:51 +02003223 */
3224#define __BPF_FUNC_MAPPER(FN) \
3225 FN(unspec), \
3226 FN(map_lookup_elem), \
3227 FN(map_update_elem), \
3228 FN(map_delete_elem), \
3229 FN(probe_read), \
3230 FN(ktime_get_ns), \
3231 FN(trace_printk), \
3232 FN(get_prandom_u32), \
3233 FN(get_smp_processor_id), \
3234 FN(skb_store_bytes), \
3235 FN(l3_csum_replace), \
3236 FN(l4_csum_replace), \
3237 FN(tail_call), \
3238 FN(clone_redirect), \
3239 FN(get_current_pid_tgid), \
3240 FN(get_current_uid_gid), \
3241 FN(get_current_comm), \
3242 FN(get_cgroup_classid), \
3243 FN(skb_vlan_push), \
3244 FN(skb_vlan_pop), \
3245 FN(skb_get_tunnel_key), \
3246 FN(skb_set_tunnel_key), \
3247 FN(perf_event_read), \
3248 FN(redirect), \
3249 FN(get_route_realm), \
3250 FN(perf_event_output), \
3251 FN(skb_load_bytes), \
3252 FN(get_stackid), \
3253 FN(csum_diff), \
3254 FN(skb_get_tunnel_opt), \
3255 FN(skb_set_tunnel_opt), \
3256 FN(skb_change_proto), \
3257 FN(skb_change_type), \
3258 FN(skb_under_cgroup), \
3259 FN(get_hash_recalc), \
3260 FN(get_current_task), \
3261 FN(probe_write_user), \
3262 FN(current_task_under_cgroup), \
3263 FN(skb_change_tail), \
3264 FN(skb_pull_data), \
3265 FN(csum_update), \
3266 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003267 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003268 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +00003269 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -07003270 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -07003271 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +02003272 FN(get_socket_uid), \
Lawrence Brakmo8c4b4c72017-06-30 20:02:46 -07003273 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003274 FN(setsockopt), \
John Fastabend97f91a72017-07-17 09:29:18 -07003275 FN(skb_adjust_room), \
John Fastabend174a79f2017-08-15 22:32:47 -07003276 FN(redirect_map), \
3277 FN(sk_redirect_map), \
3278 FN(sock_map_update), \
Yonghong Song908432c2017-10-05 09:19:20 -07003279 FN(xdp_adjust_meta), \
Yonghong Song4bebdc72017-10-05 09:19:22 -07003280 FN(perf_event_read_value), \
Lawrence Brakmocd86d1f2017-10-20 11:05:40 -07003281 FN(perf_prog_read_value), \
Josef Bacik9802d862017-12-11 11:36:48 -05003282 FN(getsockopt), \
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003283 FN(override_return), \
John Fastabend4f738ad2018-03-18 12:57:10 -07003284 FN(sock_ops_cb_flags_set), \
John Fastabend2a100312018-03-18 12:57:15 -07003285 FN(msg_redirect_map), \
John Fastabend91843d52018-03-18 12:57:20 -07003286 FN(msg_apply_bytes), \
John Fastabend015632b2018-03-18 12:57:25 -07003287 FN(msg_cork_bytes), \
Andrey Ignatovd74bad42018-03-30 15:08:05 -07003288 FN(msg_pull_data), \
Nikita V. Shirokovb32cc5b2018-04-17 21:42:13 -07003289 FN(bind), \
Eyal Birger12bed762018-04-24 17:50:29 +03003290 FN(xdp_adjust_tail), \
Yonghong Songc195651e2018-04-28 22:28:08 -07003291 FN(skb_get_xfrm_state), \
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003292 FN(get_stack), \
David Ahern87f5fc72018-05-09 20:34:26 -07003293 FN(skb_load_bytes_relative), \
John Fastabend81110382018-05-14 10:00:17 -07003294 FN(fib_lookup), \
3295 FN(sock_hash_update), \
3296 FN(msg_redirect_hash), \
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003297 FN(sk_redirect_hash), \
3298 FN(lwt_push_encap), \
3299 FN(lwt_seg6_store_bytes), \
3300 FN(lwt_seg6_adjust_srh), \
Sean Youngf4364dc2018-05-27 12:24:09 +01003301 FN(lwt_seg6_action), \
3302 FN(rc_repeat), \
Daniel Borkmanncb20b082018-06-02 23:06:36 +02003303 FN(rc_keydown), \
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07003304 FN(skb_cgroup_id), \
Roman Gushchincd339432018-08-02 14:27:24 -07003305 FN(get_current_cgroup_id), \
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003306 FN(get_local_storage), \
Andrey Ignatov77236282018-08-12 10:49:27 -07003307 FN(sk_select_reuseport), \
Joe Stringer6acc9b42018-10-02 13:35:36 -07003308 FN(skb_ancestor_cgroup_id), \
3309 FN(sk_lookup_tcp), \
3310 FN(sk_lookup_udp), \
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +02003311 FN(sk_release), \
3312 FN(map_push_elem), \
3313 FN(map_pop_elem), \
John Fastabend6fff6072018-10-19 19:56:49 -07003314 FN(map_peek_elem), \
John Fastabend7246d8e2018-11-26 14:16:17 -08003315 FN(msg_push_data), \
Sean Young01d32402018-12-06 13:01:03 +00003316 FN(msg_pop_data), \
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08003317 FN(rc_pointer_rel), \
3318 FN(spin_lock), \
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003319 FN(spin_unlock), \
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003320 FN(sk_fullsock), \
brakmof7c917b2019-03-01 12:38:46 -08003321 FN(tcp_sock), \
Martin KaFai Laudbafd7d2019-03-12 10:23:04 -07003322 FN(skb_ecn_set_ce), \
Lorenz Baueredbf8c02019-03-22 09:54:01 +08003323 FN(get_listener_sock), \
Lorenz Bauer39904082019-03-22 09:54:02 +08003324 FN(skc_lookup_tcp), \
Andrey Ignatov808649f2019-02-27 13:28:48 -08003325 FN(tcp_check_syncookie), \
Andrey Ignatov1d11b302019-02-28 19:22:15 -08003326 FN(sysctl_get_name), \
Andrey Ignatov4e63acd2019-03-07 18:38:43 -08003327 FN(sysctl_get_current_value), \
3328 FN(sysctl_get_new_value), \
Andrey Ignatovd7a4cb92019-03-18 17:55:26 -07003329 FN(sysctl_set_new_value), \
3330 FN(strtol), \
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003331 FN(strtoul), \
3332 FN(sk_storage_get), \
Yonghong Song8b401f92019-05-23 14:47:45 -07003333 FN(sk_storage_delete), \
Petar Penkov70d66242019-07-29 09:59:15 -07003334 FN(send_signal), \
Alexei Starovoitova7658e12019-10-15 20:25:04 -07003335 FN(tcp_gen_syncookie), \
Daniel Borkmann6ae08ae2019-11-02 00:17:59 +01003336 FN(skb_output), \
3337 FN(probe_read_user), \
3338 FN(probe_read_kernel), \
3339 FN(probe_read_user_str), \
Martin KaFai Lau206057f2020-01-08 16:45:51 -08003340 FN(probe_read_kernel_str), \
Yonghong Song84829412020-01-14 19:50:02 -08003341 FN(tcp_send_ack), \
Martin KaFai Lau5576b992020-01-22 15:36:46 -08003342 FN(send_signal_thread), \
Daniel Xufff7b642020-02-17 19:04:31 -08003343 FN(jiffies64), \
Carlos Neirab4490c52020-03-04 17:41:56 -03003344 FN(read_branch_records), \
Eelco Chaudrond831ee82020-03-06 08:59:23 +00003345 FN(get_ns_current_pid_tgid), \
Daniel Borkmannf3189032020-03-27 16:58:52 +01003346 FN(xdp_output), \
Daniel Borkmann0f09abd2020-03-27 16:58:54 +01003347 FN(get_netns_cookie), \
Joe Stringercf7fbe62020-03-29 15:53:38 -07003348 FN(get_current_ancestor_cgroup_id), \
Maciej Żenczykowski71d19212020-04-26 09:15:25 -07003349 FN(sk_assign), \
Yonghong Song492e6392020-05-09 10:59:14 -07003350 FN(ktime_get_boot_ns), \
3351 FN(seq_printf), \
Andrey Ignatovf307fa22020-05-14 13:03:47 -07003352 FN(seq_write), \
3353 FN(sk_cgroup_id), \
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003354 FN(sk_ancestor_cgroup_id), \
3355 FN(ringbuf_output), \
3356 FN(ringbuf_reserve), \
3357 FN(ringbuf_submit), \
3358 FN(ringbuf_discard), \
3359 FN(ringbuf_query),
Thomas Grafebb676d2016-10-27 11:23:51 +02003360
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003361/* integer value in 'imm' field of BPF_CALL instruction selects which helper
3362 * function eBPF program intends to call
3363 */
Thomas Grafebb676d2016-10-27 11:23:51 +02003364#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003365enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +02003366 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003367 __BPF_FUNC_MAX_ID,
3368};
Thomas Grafebb676d2016-10-27 11:23:51 +02003369#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -07003370
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003371/* All flags used by eBPF helper functions, placed here. */
3372
3373/* BPF_FUNC_skb_store_bytes flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003374enum {
3375 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3376 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3377};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003378
3379/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3380 * First 4 bits are for passing the header field size.
3381 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003382enum {
3383 BPF_F_HDR_FIELD_MASK = 0xfULL,
3384};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003385
3386/* BPF_FUNC_l4_csum_replace flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003387enum {
3388 BPF_F_PSEUDO_HDR = (1ULL << 4),
3389 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3390 BPF_F_MARK_ENFORCE = (1ULL << 6),
3391};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003392
3393/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003394enum {
3395 BPF_F_INGRESS = (1ULL << 0),
3396};
Daniel Borkmann781c53b2016-01-11 01:16:38 +01003397
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003398/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003399enum {
3400 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3401};
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003402
Yonghong Songc195651e2018-04-28 22:28:08 -07003403/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003404enum {
3405 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3406 BPF_F_USER_STACK = (1ULL << 8),
Yonghong Songc195651e2018-04-28 22:28:08 -07003407/* flags used by BPF_FUNC_get_stackid only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003408 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3409 BPF_F_REUSE_STACKID = (1ULL << 10),
Yonghong Songc195651e2018-04-28 22:28:08 -07003410/* flags used by BPF_FUNC_get_stack only. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003411 BPF_F_USER_BUILD_ID = (1ULL << 11),
3412};
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08003413
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003414/* BPF_FUNC_skb_set_tunnel_key flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003415enum {
3416 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3417 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3418 BPF_F_SEQ_NUMBER = (1ULL << 3),
3419};
Daniel Borkmann2da897e2016-02-23 02:05:26 +01003420
Yonghong Song908432c2017-10-05 09:19:20 -07003421/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3422 * BPF_FUNC_perf_event_read_value flags.
3423 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003424enum {
3425 BPF_F_INDEX_MASK = 0xffffffffULL,
3426 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
Daniel Borkmann555c8a82016-07-14 18:08:05 +02003427/* BPF_FUNC_perf_event_output for sk_buff input context. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003428 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3429};
Daniel Borkmann1e337592016-04-18 21:01:23 +02003430
Joe Stringerf71c6142018-11-30 15:32:20 -08003431/* Current network namespace */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003432enum {
3433 BPF_F_CURRENT_NETNS = (-1L),
3434};
Joe Stringerf71c6142018-11-30 15:32:20 -08003435
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003436/* BPF_FUNC_skb_adjust_room flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003437enum {
3438 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3439 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3440 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3441 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3442 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
Daniel Borkmann836e66c2020-06-02 16:58:32 +02003443 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003444};
Willem de Bruijn2278f6c2019-03-22 14:32:55 -04003445
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003446enum {
3447 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3448 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3449};
Alan Maguire58dfc902019-04-09 15:06:41 +01003450
Alan Maguirebfb35c22019-04-12 12:27:34 +01003451#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
Alan Maguire58dfc902019-04-09 15:06:41 +01003452 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3453 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Willem de Bruijn868d5232019-03-22 14:32:56 -04003454
Andrey Ignatov808649f2019-02-27 13:28:48 -08003455/* BPF_FUNC_sysctl_get_name flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003456enum {
3457 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3458};
Andrey Ignatov808649f2019-02-27 13:28:48 -08003459
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003460/* BPF_FUNC_sk_storage_get flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003461enum {
3462 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3463};
Martin KaFai Lau6ac99e82019-04-26 16:39:39 -07003464
Daniel Xufff7b642020-02-17 19:04:31 -08003465/* BPF_FUNC_read_branch_records flags. */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003466enum {
3467 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3468};
Daniel Xufff7b642020-02-17 19:04:31 -08003469
Andrii Nakryiko457f4432020-05-29 00:54:20 -07003470/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3471 * BPF_FUNC_bpf_ringbuf_output flags.
3472 */
3473enum {
3474 BPF_RB_NO_WAKEUP = (1ULL << 0),
3475 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3476};
3477
3478/* BPF_FUNC_bpf_ringbuf_query flags */
3479enum {
3480 BPF_RB_AVAIL_DATA = 0,
3481 BPF_RB_RING_SIZE = 1,
3482 BPF_RB_CONS_POS = 2,
3483 BPF_RB_PROD_POS = 3,
3484};
3485
3486/* BPF ring buffer constants */
3487enum {
3488 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3489 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3490 BPF_RINGBUF_HDR_SZ = 8,
3491};
3492
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003493/* Mode for BPF_FUNC_skb_adjust_room helper. */
3494enum bpf_adj_room_mode {
3495 BPF_ADJ_ROOM_NET,
Willem de Bruijn14aa3192019-03-22 14:32:54 -04003496 BPF_ADJ_ROOM_MAC,
Daniel Borkmann2be7e212017-07-02 02:13:26 +02003497};
3498
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02003499/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3500enum bpf_hdr_start_off {
3501 BPF_HDR_START_MAC,
3502 BPF_HDR_START_NET,
3503};
3504
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003505/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3506enum bpf_lwt_encap_mode {
3507 BPF_LWT_ENCAP_SEG6,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003508 BPF_LWT_ENCAP_SEG6_INLINE,
3509 BPF_LWT_ENCAP_IP,
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01003510};
3511
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003512#define __bpf_md_ptr(type, name) \
3513union { \
3514 type name; \
3515 __u64 :64; \
3516} __attribute__((aligned(8)))
3517
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003518/* user accessible mirror of in-kernel sk_buff.
3519 * new fields can only be added to the end of this structure
3520 */
3521struct __sk_buff {
3522 __u32 len;
3523 __u32 pkt_type;
3524 __u32 mark;
3525 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -07003526 __u32 protocol;
3527 __u32 vlan_present;
3528 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +01003529 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +02003530 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -07003531 __u32 ingress_ifindex;
3532 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -07003533 __u32 tc_index;
3534 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +02003535 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -07003536 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -07003537 __u32 data;
3538 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +02003539 __u32 napi_id;
John Fastabend8a31db52017-08-15 22:33:09 -07003540
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003541 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
John Fastabend8a31db52017-08-15 22:33:09 -07003542 __u32 family;
3543 __u32 remote_ip4; /* Stored in network byte order */
3544 __u32 local_ip4; /* Stored in network byte order */
3545 __u32 remote_ip6[4]; /* Stored in network byte order */
3546 __u32 local_ip6[4]; /* Stored in network byte order */
3547 __u32 remote_port; /* Stored in network byte order */
3548 __u32 local_port; /* stored in host byte order */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003549 /* ... here. */
3550
3551 __u32 data_meta;
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003552 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
Vlad Dumitrescuf11216b2018-11-22 14:39:16 -05003553 __u64 tstamp;
Petar Penkove3da08d2018-12-02 20:18:19 -05003554 __u32 wire_len;
Eric Dumazetd9ff2862019-01-23 09:22:27 -08003555 __u32 gso_segs;
Martin KaFai Lau46f8bc92019-02-09 23:22:20 -08003556 __bpf_md_ptr(struct bpf_sock *, sk);
Willem de Bruijncf620892020-03-03 15:05:01 -05003557 __u32 gso_size;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07003558};
3559
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003560struct bpf_tunnel_key {
3561 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +01003562 union {
3563 __u32 remote_ipv4;
3564 __u32 remote_ipv6[4];
3565 };
3566 __u8 tunnel_tos;
3567 __u8 tunnel_ttl;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003568 __u16 tunnel_ext; /* Padding, future use. */
Daniel Borkmann4018ab12016-03-09 03:00:05 +01003569 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07003570};
3571
Eyal Birger12bed762018-04-24 17:50:29 +03003572/* user accessible mirror of in-kernel xfrm_state.
3573 * new fields can only be added to the end of this structure
3574 */
3575struct bpf_xfrm_state {
3576 __u32 reqid;
3577 __u32 spi; /* Stored in network byte order */
3578 __u16 family;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02003579 __u16 ext; /* Padding, future use. */
Eyal Birger12bed762018-04-24 17:50:29 +03003580 union {
3581 __u32 remote_ipv4; /* Stored in network byte order */
3582 __u32 remote_ipv6[4]; /* Stored in network byte order */
3583 };
3584};
3585
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003586/* Generic BPF return codes which all BPF program types may support.
3587 * The values are binary compatible with their TC_ACT_* counter-part to
3588 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3589 * programs.
3590 *
3591 * XDP is handled seprately, see XDP_*.
3592 */
3593enum bpf_ret_code {
3594 BPF_OK = 0,
3595 /* 1 reserved */
3596 BPF_DROP = 2,
3597 /* 3-6 reserved */
3598 BPF_REDIRECT = 7,
Peter Oskolkov3e0bd372019-02-13 11:53:35 -08003599 /* >127 are reserved for prog type specific return codes.
3600 *
3601 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3602 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3603 * changed and should be routed based on its new L3 header.
3604 * (This is an L3 redirect, as opposed to L2 redirect
3605 * represented by BPF_REDIRECT above).
3606 */
3607 BPF_LWT_REROUTE = 128,
Thomas Graf3a0af8f2016-11-30 17:10:10 +01003608};
3609
David Ahern610236582016-12-01 08:48:04 -08003610struct bpf_sock {
3611 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -08003612 __u32 family;
3613 __u32 type;
3614 __u32 protocol;
David Ahern482dca92017-08-31 15:05:44 -07003615 __u32 mark;
3616 __u32 priority;
Martin KaFai Lauaa65d692019-02-09 23:22:21 -08003617 /* IP address also allows 1 and 2 bytes access */
3618 __u32 src_ip4;
3619 __u32 src_ip6[4];
3620 __u32 src_port; /* host byte order */
3621 __u32 dst_port; /* network byte order */
3622 __u32 dst_ip4;
3623 __u32 dst_ip6[4];
3624 __u32 state;
Amritha Nambiarc3c16f22020-05-26 17:34:36 -07003625 __s32 rx_queue_mapping;
David Ahern610236582016-12-01 08:48:04 -08003626};
3627
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003628struct bpf_tcp_sock {
3629 __u32 snd_cwnd; /* Sending congestion window */
3630 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3631 __u32 rtt_min;
3632 __u32 snd_ssthresh; /* Slow start size threshold */
3633 __u32 rcv_nxt; /* What we want to receive next */
3634 __u32 snd_nxt; /* Next sequence we send */
3635 __u32 snd_una; /* First byte we want an ack for */
3636 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3637 __u32 ecn_flags; /* ECN status bits. */
3638 __u32 rate_delivered; /* saved rate sample: packets delivered */
3639 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3640 __u32 packets_out; /* Packets which are "in flight" */
3641 __u32 retrans_out; /* Retransmitted packets out */
3642 __u32 total_retrans; /* Total retransmits for entire connection */
3643 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3644 * total number of segments in.
3645 */
3646 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3647 * total number of data segments in.
3648 */
3649 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3650 * The total number of segments sent.
3651 */
3652 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3653 * total number of data segments sent.
3654 */
3655 __u32 lost_out; /* Lost packets */
3656 __u32 sacked_out; /* SACK'd packets */
3657 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3658 * sum(delta(rcv_nxt)), or how many bytes
3659 * were acked.
3660 */
3661 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3662 * sum(delta(snd_una)), or how many bytes
3663 * were acked.
3664 */
Stanislav Fomichev03577462019-07-02 09:13:58 -07003665 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3666 * total number of DSACK blocks received
3667 */
3668 __u32 delivered; /* Total data packets delivered incl. rexmits */
3669 __u32 delivered_ce; /* Like the above but only ECE marked packets */
Stanislav Fomichevc2cb5e82019-07-02 09:13:59 -07003670 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
Martin KaFai Lau655a51e2019-02-09 23:22:24 -08003671};
3672
Joe Stringer6acc9b42018-10-02 13:35:36 -07003673struct bpf_sock_tuple {
3674 union {
3675 struct {
3676 __be32 saddr;
3677 __be32 daddr;
3678 __be16 sport;
3679 __be16 dport;
3680 } ipv4;
3681 struct {
3682 __be32 saddr[4];
3683 __be32 daddr[4];
3684 __be16 sport;
3685 __be16 dport;
3686 } ipv6;
3687 };
3688};
3689
Jonathan Lemonfada7fd2019-06-06 13:59:40 -07003690struct bpf_xdp_sock {
3691 __u32 queue_id;
3692};
3693
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08003694#define XDP_PACKET_HEADROOM 256
3695
Brenden Blanco6a773a12016-07-19 12:16:47 -07003696/* User return codes for XDP prog type.
3697 * A valid XDP program must return one of these defined values. All other
Daniel Borkmann9beb8be2017-09-09 01:40:35 +02003698 * return codes are reserved for future use. Unknown return codes will
3699 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
Brenden Blanco6a773a12016-07-19 12:16:47 -07003700 */
3701enum xdp_action {
3702 XDP_ABORTED = 0,
3703 XDP_DROP,
3704 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -07003705 XDP_TX,
John Fastabend814abfa2017-07-17 09:27:07 -07003706 XDP_REDIRECT,
Brenden Blanco6a773a12016-07-19 12:16:47 -07003707};
3708
3709/* user accessible metadata for XDP packet hook
3710 * new fields must be added to the end of this structure
3711 */
3712struct xdp_md {
3713 __u32 data;
3714 __u32 data_end;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003715 __u32 data_meta;
Jesper Dangaard Brouerdaaf24c2018-01-11 17:39:09 +01003716 /* Below access go through struct xdp_rxq_info */
Jesper Dangaard Brouer02dd3292018-01-03 11:26:14 +01003717 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3718 __u32 rx_queue_index; /* rxq->queue_index */
David Ahern64b59022020-05-29 16:07:14 -06003719
3720 __u32 egress_ifindex; /* txq->dev->ifindex */
Brenden Blanco6a773a12016-07-19 12:16:47 -07003721};
3722
John Fastabend174a79f2017-08-15 22:32:47 -07003723enum sk_action {
John Fastabendbfa640752017-10-27 09:45:53 -07003724 SK_DROP = 0,
3725 SK_PASS,
John Fastabend174a79f2017-08-15 22:32:47 -07003726};
3727
John Fastabend4f738ad2018-03-18 12:57:10 -07003728/* user accessible metadata for SK_MSG packet hook, new fields must
3729 * be added to the end of this structure
3730 */
3731struct sk_msg_md {
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003732 __bpf_md_ptr(void *, data);
3733 __bpf_md_ptr(void *, data_end);
John Fastabend303def32018-05-17 14:16:58 -07003734
3735 __u32 family;
3736 __u32 remote_ip4; /* Stored in network byte order */
3737 __u32 local_ip4; /* Stored in network byte order */
3738 __u32 remote_ip6[4]; /* Stored in network byte order */
3739 __u32 local_ip6[4]; /* Stored in network byte order */
3740 __u32 remote_port; /* Stored in network byte order */
3741 __u32 local_port; /* stored in host byte order */
John Fastabend3bdbd022018-12-16 15:47:04 -08003742 __u32 size; /* Total size of sk_msg */
John Fastabend13d70f52020-05-24 09:51:15 -07003743
3744 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
John Fastabend4f738ad2018-03-18 12:57:10 -07003745};
3746
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003747struct sk_reuseport_md {
3748 /*
3749 * Start of directly accessible data. It begins from
3750 * the tcp/udp header.
3751 */
Daniel Borkmannb7df9ada2018-12-01 01:18:53 +01003752 __bpf_md_ptr(void *, data);
3753 /* End of directly accessible data */
3754 __bpf_md_ptr(void *, data_end);
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07003755 /*
3756 * Total length of packet (starting from the tcp/udp header).
3757 * Note that the directly accessible bytes (data_end - data)
3758 * could be less than this "len". Those bytes could be
3759 * indirectly read by a helper "bpf_skb_load_bytes()".
3760 */
3761 __u32 len;
3762 /*
3763 * Eth protocol in the mac header (network byte order). e.g.
3764 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3765 */
3766 __u32 eth_protocol;
3767 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3768 __u32 bind_inany; /* Is sock bound to an INANY address? */
3769 __u32 hash; /* A hash of the packet 4 tuples */
3770};
3771
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003772#define BPF_TAG_SIZE 8
3773
3774struct bpf_prog_info {
3775 __u32 type;
3776 __u32 id;
3777 __u8 tag[BPF_TAG_SIZE];
3778 __u32 jited_prog_len;
3779 __u32 xlated_prog_len;
3780 __aligned_u64 jited_prog_insns;
3781 __aligned_u64 xlated_prog_insns;
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -07003782 __u64 load_time; /* ns since boottime */
3783 __u32 created_by_uid;
3784 __u32 nr_map_ids;
3785 __aligned_u64 map_ids;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003786 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski675fc272017-12-27 18:39:09 -08003787 __u32 ifindex;
Jiri Olsab85fab02018-04-25 19:41:06 +02003788 __u32 gpl_compatible:1;
Baruch Siach04723012019-06-28 07:08:45 +03003789 __u32 :31; /* alignment pad */
Jakub Kicinski675fc272017-12-27 18:39:09 -08003790 __u64 netns_dev;
3791 __u64 netns_ino;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303792 __u32 nr_jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303793 __u32 nr_jited_func_lens;
Sandipan Dasdbecd732018-05-24 12:26:48 +05303794 __aligned_u64 jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05303795 __aligned_u64 jited_func_lens;
Yonghong Song838e9692018-11-19 15:29:11 -08003796 __u32 btf_id;
3797 __u32 func_info_rec_size;
3798 __aligned_u64 func_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003799 __u32 nr_func_info;
3800 __u32 nr_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003801 __aligned_u64 line_info;
3802 __aligned_u64 jited_line_info;
Yonghong Song11d8b822018-12-10 14:14:08 -08003803 __u32 nr_jited_line_info;
Martin KaFai Lauc454a462018-12-07 16:42:25 -08003804 __u32 line_info_rec_size;
3805 __u32 jited_line_info_rec_size;
Song Liuc872bdb2018-12-12 09:37:46 -08003806 __u32 nr_prog_tags;
3807 __aligned_u64 prog_tags;
Alexei Starovoitov5f8f8b92019-02-25 14:28:40 -08003808 __u64 run_time_ns;
3809 __u64 run_cnt;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003810} __attribute__((aligned(8)));
3811
3812struct bpf_map_info {
3813 __u32 type;
3814 __u32 id;
3815 __u32 key_size;
3816 __u32 value_size;
3817 __u32 max_entries;
3818 __u32 map_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07003819 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski52775b32018-01-17 19:13:28 -08003820 __u32 ifindex;
Martin KaFai Lau85d33df2020-01-08 16:35:05 -08003821 __u32 btf_vmlinux_value_type_id;
Jakub Kicinski52775b32018-01-17 19:13:28 -08003822 __u64 netns_dev;
3823 __u64 netns_ino;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -07003824 __u32 btf_id;
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -07003825 __u32 btf_key_type_id;
3826 __u32 btf_value_type_id;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07003827} __attribute__((aligned(8)));
3828
Martin KaFai Lau62dab842018-05-04 14:49:52 -07003829struct bpf_btf_info {
3830 __aligned_u64 btf;
3831 __u32 btf_size;
3832 __u32 id;
3833} __attribute__((aligned(8)));
3834
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -07003835struct bpf_link_info {
3836 __u32 type;
3837 __u32 id;
3838 __u32 prog_id;
3839 union {
3840 struct {
3841 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3842 __u32 tp_name_len; /* in/out: tp_name buffer len */
3843 } raw_tracepoint;
3844 struct {
3845 __u32 attach_type;
3846 } tracing;
3847 struct {
3848 __u64 cgroup_id;
3849 __u32 attach_type;
3850 } cgroup;
Jakub Sitnicki7f045a42020-05-31 10:28:38 +02003851 struct {
3852 __u32 netns_ino;
3853 __u32 attach_type;
3854 } netns;
Andrii Nakryikof2e10bf2020-04-28 17:16:08 -07003855 };
3856} __attribute__((aligned(8)));
3857
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003858/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3859 * by user and intended to be used by socket (e.g. to bind to, depends on
3860 * attach attach type).
3861 */
3862struct bpf_sock_addr {
3863 __u32 user_family; /* Allows 4-byte read, but no write. */
3864 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3865 * Stored in network byte order.
3866 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07003867 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003868 * Stored in network byte order.
3869 */
Andrey Ignatov7aebfa12020-05-13 18:50:27 -07003870 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003871 * Stored in network byte order
3872 */
3873 __u32 family; /* Allows 4-byte read, but no write */
3874 __u32 type; /* Allows 4-byte read, but no write */
3875 __u32 protocol; /* Allows 4-byte read, but no write */
Stanislav Fomichev600c70b2019-07-01 10:38:39 -07003876 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07003877 * Stored in network byte order.
3878 */
Stanislav Fomichevd4ecfeb2019-07-15 09:39:53 -07003879 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
Andrey Ignatov1cedee12018-05-25 08:55:23 -07003880 * Stored in network byte order.
3881 */
Stanislav Fomichevfb85c4a2019-06-12 10:30:37 -07003882 __bpf_md_ptr(struct bpf_sock *, sk);
Andrey Ignatov4fbac772018-03-30 15:08:02 -07003883};
3884
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003885/* User bpf_sock_ops struct to access socket values and specify request ops
3886 * and their replies.
3887 * Some of this fields are in network (bigendian) byte order and may need
3888 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3889 * New fields can only be added at the end of this structure
3890 */
3891struct bpf_sock_ops {
3892 __u32 op;
3893 union {
Lawrence Brakmode525be2018-01-25 16:14:09 -08003894 __u32 args[4]; /* Optionally passed to bpf program */
3895 __u32 reply; /* Returned by bpf program */
3896 __u32 replylong[4]; /* Optionally returned by bpf prog */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003897 };
3898 __u32 family;
3899 __u32 remote_ip4; /* Stored in network byte order */
3900 __u32 local_ip4; /* Stored in network byte order */
3901 __u32 remote_ip6[4]; /* Stored in network byte order */
3902 __u32 local_ip6[4]; /* Stored in network byte order */
3903 __u32 remote_port; /* Stored in network byte order */
3904 __u32 local_port; /* stored in host byte order */
Lawrence Brakmof19397a2017-12-01 10:15:04 -08003905 __u32 is_fullsock; /* Some TCP fields are only valid if
3906 * there is a full socket. If not, the
3907 * fields read as zero.
3908 */
3909 __u32 snd_cwnd;
3910 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003911 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
Lawrence Brakmo44f0e432018-01-25 16:14:12 -08003912 __u32 state;
3913 __u32 rtt_min;
3914 __u32 snd_ssthresh;
3915 __u32 rcv_nxt;
3916 __u32 snd_nxt;
3917 __u32 snd_una;
3918 __u32 mss_cache;
3919 __u32 ecn_flags;
3920 __u32 rate_delivered;
3921 __u32 rate_interval_us;
3922 __u32 packets_out;
3923 __u32 retrans_out;
3924 __u32 total_retrans;
3925 __u32 segs_in;
3926 __u32 data_segs_in;
3927 __u32 segs_out;
3928 __u32 data_segs_out;
3929 __u32 lost_out;
3930 __u32 sacked_out;
3931 __u32 sk_txhash;
3932 __u64 bytes_received;
3933 __u64 bytes_acked;
Stanislav Fomichev1314ef52019-06-12 10:30:38 -07003934 __bpf_md_ptr(struct bpf_sock *, sk);
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003935};
3936
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003937/* Definitions for bpf_sock_ops_cb_flags */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08003938enum {
3939 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3940 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3941 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3942 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3943/* Mask of all currently supported cb flags */
3944 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3945};
Lawrence Brakmob13d8802018-01-25 16:14:10 -08003946
Lawrence Brakmo40304b22017-06-30 20:02:40 -07003947/* List of known BPF sock_ops operators.
3948 * New entries can only be added at the end
3949 */
3950enum {
3951 BPF_SOCK_OPS_VOID,
Lawrence Brakmo8550f322017-06-30 20:02:42 -07003952 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3953 * -1 if default value should be used
3954 */
Lawrence Brakmo13d3b1e2017-06-30 20:02:44 -07003955 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3956 * window (in packets) or -1 if default
3957 * value should be used
3958 */
Lawrence Brakmo9872a4b2017-06-30 20:02:47 -07003959 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3960 * active connection is initialized
3961 */
3962 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3963 * active connection is
3964 * established
3965 */
3966 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3967 * passive connection is
3968 * established
3969 */
Lawrence Brakmo91b5b212017-06-30 20:02:49 -07003970 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3971 * needs ECN
3972 */
Lawrence Brakmoe6546ef2017-10-20 11:05:39 -07003973 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3974 * based on the path and may be
3975 * dependent on the congestion control
3976 * algorithm. In general it indicates
3977 * a congestion threshold. RTTs above
3978 * this indicate congestion
3979 */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08003980 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3981 * Arg1: value of icsk_retransmits
3982 * Arg2: value of icsk_rto
3983 * Arg3: whether RTO has expired
3984 */
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08003985 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3986 * Arg1: sequence number of 1st byte
3987 * Arg2: # segments
3988 * Arg3: return value of
3989 * tcp_transmit_skb (0 => success)
3990 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08003991 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3992 * Arg1: old_state
3993 * Arg2: new_state
3994 */
Andrey Ignatovf333ee02018-07-11 17:33:32 -07003995 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3996 * socket transition to LISTEN state.
3997 */
Stanislav Fomichev23729ff2019-07-02 09:13:56 -07003998 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3999 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08004000};
4001
4002/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4003 * changes between the TCP and BPF versions. Ideally this should never happen.
4004 * If it does, we need to add code to convert them before calling
4005 * the BPF sock_ops function.
4006 */
4007enum {
4008 BPF_TCP_ESTABLISHED = 1,
4009 BPF_TCP_SYN_SENT,
4010 BPF_TCP_SYN_RECV,
4011 BPF_TCP_FIN_WAIT1,
4012 BPF_TCP_FIN_WAIT2,
4013 BPF_TCP_TIME_WAIT,
4014 BPF_TCP_CLOSE,
4015 BPF_TCP_CLOSE_WAIT,
4016 BPF_TCP_LAST_ACK,
4017 BPF_TCP_LISTEN,
4018 BPF_TCP_CLOSING, /* Now a valid state */
4019 BPF_TCP_NEW_SYN_RECV,
4020
4021 BPF_TCP_MAX_STATES /* Leave at the end! */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07004022};
4023
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004024enum {
4025 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4026 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4027};
Lawrence Brakmofc747812017-06-30 20:02:51 -07004028
Yonghong Song908432c2017-10-05 09:19:20 -07004029struct bpf_perf_event_value {
4030 __u64 counter;
4031 __u64 enabled;
4032 __u64 running;
4033};
4034
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004035enum {
4036 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4037 BPF_DEVCG_ACC_READ = (1ULL << 1),
4038 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4039};
Roman Gushchinebc614f2017-11-05 08:15:32 -05004040
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004041enum {
4042 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4043 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4044};
Roman Gushchinebc614f2017-11-05 08:15:32 -05004045
4046struct bpf_cgroup_dev_ctx {
Yonghong Song06ef0cc2017-12-18 10:13:44 -08004047 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4048 __u32 access_type;
Roman Gushchinebc614f2017-11-05 08:15:32 -05004049 __u32 major;
4050 __u32 minor;
4051};
4052
Alexei Starovoitovc4f66992018-03-28 12:05:37 -07004053struct bpf_raw_tracepoint_args {
4054 __u64 args[0];
4055};
4056
David Ahern87f5fc72018-05-09 20:34:26 -07004057/* DIRECT: Skip the FIB rules and go to FIB table associated with device
4058 * OUTPUT: Do lookup from egress perspective; default is ingress
4059 */
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004060enum {
4061 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4062 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4063};
David Ahern87f5fc72018-05-09 20:34:26 -07004064
David Ahern4c795792018-06-26 16:21:18 -07004065enum {
4066 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4067 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4068 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4069 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4070 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4071 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4072 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4073 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4074 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4075};
4076
David Ahern87f5fc72018-05-09 20:34:26 -07004077struct bpf_fib_lookup {
David Ahernfa898d72018-05-29 10:58:07 -07004078 /* input: network family for lookup (AF_INET, AF_INET6)
4079 * output: network family of egress nexthop
4080 */
4081 __u8 family;
David Ahern87f5fc72018-05-09 20:34:26 -07004082
4083 /* set if lookup is to consider L4 data - e.g., FIB rules */
4084 __u8 l4_protocol;
4085 __be16 sport;
4086 __be16 dport;
4087
4088 /* total length of packet from network header - used for MTU check */
4089 __u16 tot_len;
David Ahern4c795792018-06-26 16:21:18 -07004090
4091 /* input: L3 device index for lookup
4092 * output: device index from FIB lookup
4093 */
4094 __u32 ifindex;
David Ahern87f5fc72018-05-09 20:34:26 -07004095
4096 union {
4097 /* inputs to lookup */
4098 __u8 tos; /* AF_INET */
David Ahernbd3a08a2018-06-03 08:15:19 -07004099 __be32 flowinfo; /* AF_INET6, flow_label + priority */
David Ahern87f5fc72018-05-09 20:34:26 -07004100
David Ahernfa898d72018-05-29 10:58:07 -07004101 /* output: metric of fib result (IPv4/IPv6 only) */
4102 __u32 rt_metric;
David Ahern87f5fc72018-05-09 20:34:26 -07004103 };
4104
4105 union {
David Ahern87f5fc72018-05-09 20:34:26 -07004106 __be32 ipv4_src;
4107 __u32 ipv6_src[4]; /* in6_addr; network order */
4108 };
4109
David Ahernfa898d72018-05-29 10:58:07 -07004110 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4111 * network header. output: bpf_fib_lookup sets to gateway address
4112 * if FIB lookup returns gateway route
David Ahern87f5fc72018-05-09 20:34:26 -07004113 */
4114 union {
David Ahern87f5fc72018-05-09 20:34:26 -07004115 __be32 ipv4_dst;
4116 __u32 ipv6_dst[4]; /* in6_addr; network order */
4117 };
4118
4119 /* output */
4120 __be16 h_vlan_proto;
4121 __be16 h_vlan_TCI;
4122 __u8 smac[6]; /* ETH_ALEN */
4123 __u8 dmac[6]; /* ETH_ALEN */
4124};
4125
Yonghong Song41bdc4b2018-05-24 11:21:09 -07004126enum bpf_task_fd_type {
4127 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4128 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4129 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4130 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4131 BPF_FD_TYPE_UPROBE, /* filename + offset */
4132 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4133};
4134
Andrii Nakryiko1aae4bd2020-03-02 16:32:31 -08004135enum {
4136 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4137 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4138 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4139};
Stanislav Fomichev086f9562019-07-25 15:52:25 -07004140
Petar Penkovd58e4682018-09-14 07:46:18 -07004141struct bpf_flow_keys {
4142 __u16 nhoff;
4143 __u16 thoff;
4144 __u16 addr_proto; /* ETH_P_* of valid addrs */
4145 __u8 is_frag;
4146 __u8 is_first_frag;
4147 __u8 is_encap;
4148 __u8 ip_proto;
4149 __be16 n_proto;
4150 __be16 sport;
4151 __be16 dport;
4152 union {
4153 struct {
4154 __be32 ipv4_src;
4155 __be32 ipv4_dst;
4156 };
4157 struct {
4158 __u32 ipv6_src[4]; /* in6_addr; network order */
4159 __u32 ipv6_dst[4]; /* in6_addr; network order */
4160 };
4161 };
Stanislav Fomichev086f9562019-07-25 15:52:25 -07004162 __u32 flags;
Stanislav Fomichev71c99e32019-07-25 15:52:30 -07004163 __be32 flow_label;
Petar Penkovd58e4682018-09-14 07:46:18 -07004164};
4165
Yonghong Song838e9692018-11-19 15:29:11 -08004166struct bpf_func_info {
Martin KaFai Laud30d42e2018-12-05 17:35:44 -08004167 __u32 insn_off;
Yonghong Song838e9692018-11-19 15:29:11 -08004168 __u32 type_id;
4169};
4170
Martin KaFai Lauc454a462018-12-07 16:42:25 -08004171#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4172#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4173
4174struct bpf_line_info {
4175 __u32 insn_off;
4176 __u32 file_name_off;
4177 __u32 line_off;
4178 __u32 line_col;
4179};
4180
Alexei Starovoitovd83525c2019-01-31 15:40:04 -08004181struct bpf_spin_lock {
4182 __u32 val;
4183};
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08004184
4185struct bpf_sysctl {
4186 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4187 * Allows 1,2,4-byte read, but no write.
4188 */
Andrey Ignatove1550bf2019-03-07 18:50:52 -08004189 __u32 file_pos; /* Sysctl file position to read from, write to.
4190 * Allows 1,2,4-byte read an 4-byte write.
4191 */
Andrey Ignatov7b146ce2019-02-27 12:59:24 -08004192};
4193
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07004194struct bpf_sockopt {
4195 __bpf_md_ptr(struct bpf_sock *, sk);
4196 __bpf_md_ptr(void *, optval);
4197 __bpf_md_ptr(void *, optval_end);
4198
4199 __s32 level;
4200 __s32 optname;
4201 __s32 optlen;
4202 __s32 retval;
4203};
4204
Carlos Neirab4490c52020-03-04 17:41:56 -03004205struct bpf_pidns_info {
4206 __u32 pid;
4207 __u32 tgid;
4208};
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07004209#endif /* _UAPI__LINUX_BPF_H__ */