blob: 0183b8e70a9e6dfa963713305242964bdf04219b [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 */
17#define BPF_ALU64 0x07 /* alu mode in double word width */
18
19/* ld/ldx fields */
Jesper Dangaard Brouercb5f7332018-01-17 12:05:36 +010020#define BPF_DW 0x18 /* double word (64-bit) */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070021#define BPF_XADD 0xc0 /* exclusive add */
22
23/* alu/jmp fields */
24#define BPF_MOV 0xb0 /* mov reg to reg */
25#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
26
27/* change endianness of a register */
28#define BPF_END 0xd0 /* flags for endianness conversion: */
29#define BPF_TO_LE 0x00 /* convert to little-endian */
30#define BPF_TO_BE 0x08 /* convert to big-endian */
31#define BPF_FROM_LE BPF_TO_LE
32#define BPF_FROM_BE BPF_TO_BE
33
Daniel Borkmann92b31a92017-08-10 01:39:55 +020034/* jmp encodings */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070035#define BPF_JNE 0x50 /* jump != */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020036#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070038#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
Daniel Borkmann92b31a92017-08-10 01:39:55 +020040#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -070042#define BPF_CALL 0x80 /* function call */
43#define BPF_EXIT 0x90 /* function return */
44
45/* Register numbers */
46enum {
47 BPF_REG_0 = 0,
48 BPF_REG_1,
49 BPF_REG_2,
50 BPF_REG_3,
51 BPF_REG_4,
52 BPF_REG_5,
53 BPF_REG_6,
54 BPF_REG_7,
55 BPF_REG_8,
56 BPF_REG_9,
57 BPF_REG_10,
58 __MAX_BPF_REG,
59};
60
61/* BPF has 10 general purpose 64-bit registers and stack frame. */
62#define MAX_BPF_REG __MAX_BPF_REG
63
64struct bpf_insn {
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
70};
71
Daniel Mackb95a5c42017-01-21 17:26:11 +010072/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
76};
77
Roman Gushchinde9cbba2018-08-02 14:27:18 -070078struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
81};
82
Daniel Borkmannb2197752015-10-29 14:58:09 +010083/* BPF syscall commands, see bpf(2) man-page for details. */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070084enum bpf_cmd {
Alexei Starovoitov99c55f72014-09-26 00:16:57 -070085 BPF_MAP_CREATE,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070086 BPF_MAP_LOOKUP_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070087 BPF_MAP_UPDATE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070088 BPF_MAP_DELETE_ELEM,
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -070089 BPF_MAP_GET_NEXT_KEY,
Alexei Starovoitov09756af2014-09-26 00:17:00 -070090 BPF_PROG_LOAD,
Daniel Borkmannb2197752015-10-29 14:58:09 +010091 BPF_OBJ_PIN,
92 BPF_OBJ_GET,
Daniel Mackf4324552016-11-23 16:52:27 +010093 BPF_PROG_ATTACH,
94 BPF_PROG_DETACH,
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -070095 BPF_PROG_TEST_RUN,
Martin KaFai Lau34ad5582017-06-05 12:15:48 -070096 BPF_PROG_GET_NEXT_ID,
97 BPF_MAP_GET_NEXT_ID,
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -070098 BPF_PROG_GET_FD_BY_ID,
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -070099 BPF_MAP_GET_FD_BY_ID,
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700100 BPF_OBJ_GET_INFO_BY_FD,
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700101 BPF_PROG_QUERY,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700102 BPF_RAW_TRACEPOINT_OPEN,
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700103 BPF_BTF_LOAD,
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700104 BPF_BTF_GET_FD_BY_ID,
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700105 BPF_TASK_FD_QUERY,
Mauricio Vasquez Bbd513cd2018-10-18 15:16:30 +0200106 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700107};
108
109enum bpf_map_type {
110 BPF_MAP_TYPE_UNSPEC,
Alexei Starovoitov0f8e4bd2014-11-13 17:36:45 -0800111 BPF_MAP_TYPE_HASH,
Alexei Starovoitov28fbcfa2014-11-13 17:36:46 -0800112 BPF_MAP_TYPE_ARRAY,
Alexei Starovoitov04fd61ab2015-05-19 16:59:03 -0700113 BPF_MAP_TYPE_PROG_ARRAY,
Kaixu Xiaea317b22015-08-06 07:02:34 +0000114 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
Alexei Starovoitov824bd0c2016-02-01 22:39:53 -0800115 BPF_MAP_TYPE_PERCPU_HASH,
Alexei Starovoitova10423b2016-02-01 22:39:54 -0800116 BPF_MAP_TYPE_PERCPU_ARRAY,
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -0800117 BPF_MAP_TYPE_STACK_TRACE,
Martin KaFai Lau4ed8ec52016-06-30 10:28:43 -0700118 BPF_MAP_TYPE_CGROUP_ARRAY,
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800119 BPF_MAP_TYPE_LRU_HASH,
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800120 BPF_MAP_TYPE_LRU_PERCPU_HASH,
Daniel Mackb95a5c42017-01-21 17:26:11 +0100121 BPF_MAP_TYPE_LPM_TRIE,
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700122 BPF_MAP_TYPE_ARRAY_OF_MAPS,
Martin KaFai Laubcc6b1b2017-03-22 10:00:34 -0700123 BPF_MAP_TYPE_HASH_OF_MAPS,
John Fastabend546ac1f2017-07-17 09:28:56 -0700124 BPF_MAP_TYPE_DEVMAP,
John Fastabend174a79f2017-08-15 22:32:47 -0700125 BPF_MAP_TYPE_SOCKMAP,
Jesper Dangaard Brouer6710e112017-10-16 12:19:28 +0200126 BPF_MAP_TYPE_CPUMAP,
Björn Töpelfbfc504a2018-05-02 13:01:28 +0200127 BPF_MAP_TYPE_XSKMAP,
John Fastabend81110382018-05-14 10:00:17 -0700128 BPF_MAP_TYPE_SOCKHASH,
Roman Gushchinde9cbba2018-08-02 14:27:18 -0700129 BPF_MAP_TYPE_CGROUP_STORAGE,
Martin KaFai Lau5dc4c4b2018-08-08 01:01:24 -0700130 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
Roman Gushchinb741f162018-09-28 14:45:43 +0000131 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +0200132 BPF_MAP_TYPE_QUEUE,
133 BPF_MAP_TYPE_STACK,
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700134};
135
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700136enum bpf_prog_type {
137 BPF_PROG_TYPE_UNSPEC,
Alexei Starovoitovddd872b2014-12-01 15:06:34 -0800138 BPF_PROG_TYPE_SOCKET_FILTER,
Alexei Starovoitov25415172015-03-25 12:49:20 -0700139 BPF_PROG_TYPE_KPROBE,
Daniel Borkmann96be4322015-03-01 12:31:46 +0100140 BPF_PROG_TYPE_SCHED_CLS,
Daniel Borkmann94caee8c2015-03-20 15:11:11 +0100141 BPF_PROG_TYPE_SCHED_ACT,
Alexei Starovoitov98b5c2c2016-04-06 18:43:25 -0700142 BPF_PROG_TYPE_TRACEPOINT,
Brenden Blanco6a773a12016-07-19 12:16:47 -0700143 BPF_PROG_TYPE_XDP,
Alexei Starovoitov0515e592016-09-01 18:37:22 -0700144 BPF_PROG_TYPE_PERF_EVENT,
Daniel Mack0e336612016-11-23 16:52:25 +0100145 BPF_PROG_TYPE_CGROUP_SKB,
David Ahern610236582016-12-01 08:48:04 -0800146 BPF_PROG_TYPE_CGROUP_SOCK,
Thomas Graf3a0af8f2016-11-30 17:10:10 +0100147 BPF_PROG_TYPE_LWT_IN,
148 BPF_PROG_TYPE_LWT_OUT,
149 BPF_PROG_TYPE_LWT_XMIT,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700150 BPF_PROG_TYPE_SOCK_OPS,
John Fastabendb005fd12017-08-15 22:31:58 -0700151 BPF_PROG_TYPE_SK_SKB,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500152 BPF_PROG_TYPE_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700153 BPF_PROG_TYPE_SK_MSG,
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700154 BPF_PROG_TYPE_RAW_TRACEPOINT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700155 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
Mathieu Xhonneux004d4b22018-05-20 14:58:16 +0100156 BPF_PROG_TYPE_LWT_SEG6LOCAL,
Sean Youngf4364dc2018-05-27 12:24:09 +0100157 BPF_PROG_TYPE_LIRC_MODE2,
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -0700158 BPF_PROG_TYPE_SK_REUSEPORT,
Petar Penkovd58e4682018-09-14 07:46:18 -0700159 BPF_PROG_TYPE_FLOW_DISSECTOR,
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700160};
161
Daniel Mack0e336612016-11-23 16:52:25 +0100162enum bpf_attach_type {
163 BPF_CGROUP_INET_INGRESS,
164 BPF_CGROUP_INET_EGRESS,
David Ahern610236582016-12-01 08:48:04 -0800165 BPF_CGROUP_INET_SOCK_CREATE,
Lawrence Brakmo40304b22017-06-30 20:02:40 -0700166 BPF_CGROUP_SOCK_OPS,
John Fastabend464bc0f2017-08-28 07:10:04 -0700167 BPF_SK_SKB_STREAM_PARSER,
168 BPF_SK_SKB_STREAM_VERDICT,
Roman Gushchinebc614f2017-11-05 08:15:32 -0500169 BPF_CGROUP_DEVICE,
John Fastabend4f738ad2018-03-18 12:57:10 -0700170 BPF_SK_MSG_VERDICT,
Andrey Ignatov4fbac772018-03-30 15:08:02 -0700171 BPF_CGROUP_INET4_BIND,
172 BPF_CGROUP_INET6_BIND,
Andrey Ignatovd74bad42018-03-30 15:08:05 -0700173 BPF_CGROUP_INET4_CONNECT,
174 BPF_CGROUP_INET6_CONNECT,
Andrey Ignatovaac3fc32018-03-30 15:08:07 -0700175 BPF_CGROUP_INET4_POST_BIND,
176 BPF_CGROUP_INET6_POST_BIND,
Andrey Ignatov1cedee12018-05-25 08:55:23 -0700177 BPF_CGROUP_UDP4_SENDMSG,
178 BPF_CGROUP_UDP6_SENDMSG,
Sean Youngf4364dc2018-05-27 12:24:09 +0100179 BPF_LIRC_MODE2,
Petar Penkovd58e4682018-09-14 07:46:18 -0700180 BPF_FLOW_DISSECTOR,
Daniel Mack0e336612016-11-23 16:52:25 +0100181 __MAX_BPF_ATTACH_TYPE
182};
183
184#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
185
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700186/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
187 *
188 * NONE(default): No further bpf programs allowed in the subtree.
189 *
190 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
191 * the program in this cgroup yields to sub-cgroup program.
192 *
193 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
194 * that cgroup program gets run in addition to the program in this cgroup.
195 *
196 * Only one program is allowed to be attached to a cgroup with
197 * NONE or BPF_F_ALLOW_OVERRIDE flag.
198 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
199 * release old program and attach the new one. Attach flags has to match.
200 *
201 * Multiple programs are allowed to be attached to a cgroup with
202 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
203 * (those that were attached first, run first)
204 * The programs of sub-cgroup are executed first, then programs of
205 * this cgroup and then programs of parent cgroup.
206 * When children program makes decision (like picking TCP CA or sock bind)
207 * parent program has a chance to override it.
208 *
209 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
210 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
211 * Ex1:
212 * cgrp1 (MULTI progs A, B) ->
213 * cgrp2 (OVERRIDE prog C) ->
214 * cgrp3 (MULTI prog D) ->
215 * cgrp4 (OVERRIDE prog E) ->
216 * cgrp5 (NONE prog F)
217 * the event in cgrp5 triggers execution of F,D,A,B in that order.
218 * if prog F is detached, the execution is E,D,A,B
219 * if prog F and D are detached, the execution is E,A,B
220 * if prog F, E and D are detached, the execution is C,A,B
221 *
222 * All eligible programs are executed regardless of return code from
223 * earlier programs.
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800224 */
225#define BPF_F_ALLOW_OVERRIDE (1U << 0)
Alexei Starovoitov324bda9e62017-10-02 22:50:21 -0700226#define BPF_F_ALLOW_MULTI (1U << 1)
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800227
David S. Millere07b98d2017-05-10 11:38:07 -0700228/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
229 * verifier will perform strict alignment checking as if the kernel
230 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
231 * and NET_IP_ALIGN defined to 2.
232 */
233#define BPF_F_STRICT_ALIGNMENT (1U << 0)
234
David Millere9ee9ef2018-11-30 21:08:14 -0800235/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
236 * verifier will allow any alignment whatsoever. On platforms
237 * with strict alignment requirements for loads ands stores (such
238 * as sparc and mips) the verifier validates that all loads and
239 * stores provably follow this requirement. This flag turns that
240 * checking and enforcement off.
241 *
242 * It is mostly used for testing when we want to validate the
243 * context and memory access aspects of the verifier, but because
244 * of an unaligned access the alignment check would trigger before
245 * the one we are interested in.
246 */
247#define BPF_F_ANY_ALIGNMENT (1U << 1)
248
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800249/* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
Daniel Borkmannf1a66f82015-03-01 12:31:43 +0100250#define BPF_PSEUDO_MAP_FD 1
251
Alexei Starovoitovcc8b0b92017-12-14 17:55:05 -0800252/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
253 * offset to another bpf function
254 */
255#define BPF_PSEUDO_CALL 1
256
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800257/* flags for BPF_MAP_UPDATE_ELEM command */
258#define BPF_ANY 0 /* create new element or update existing */
259#define BPF_NOEXIST 1 /* create new element if it didn't exist */
260#define BPF_EXIST 2 /* update existing element */
261
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700262/* flags for BPF_MAP_CREATE command */
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800263#define BPF_F_NO_PREALLOC (1U << 0)
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800264/* Instead of having one common LRU list in the
Martin KaFai Lau8f844932016-11-11 10:55:10 -0800265 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
Martin KaFai Lau29ba7322016-11-11 10:55:09 -0800266 * which can scale and perform better.
267 * Note, the LRU nodes (including free nodes) cannot be moved
268 * across different LRU lists.
269 */
270#define BPF_F_NO_COMMON_LRU (1U << 1)
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700271/* Specify numa node during map creation */
272#define BPF_F_NUMA_NODE (1U << 2)
Alexei Starovoitov6c905982016-03-07 21:57:15 -0800273
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -0700274#define BPF_OBJ_NAME_LEN 16U
275
Chenbo Feng6e71b042017-10-18 13:00:22 -0700276/* Flags for accessing BPF object */
277#define BPF_F_RDONLY (1U << 3)
278#define BPF_F_WRONLY (1U << 4)
279
Song Liu615755a2018-03-14 10:23:21 -0700280/* Flag for stack_map, store build_id+offset instead of pointer */
281#define BPF_F_STACK_BUILD_ID (1U << 5)
282
Lorenz Bauer96b3b6c2018-11-16 11:41:08 +0000283/* Zero-initialize hash function seed. This should only be used for testing. */
284#define BPF_F_ZERO_SEED (1U << 6)
285
Lorenz Bauer2f183362018-11-16 11:41:09 +0000286/* flags for BPF_PROG_QUERY */
287#define BPF_F_QUERY_EFFECTIVE (1U << 0)
288
Song Liu615755a2018-03-14 10:23:21 -0700289enum bpf_stack_build_id_status {
290 /* user space need an empty entry to identify end of a trace */
291 BPF_STACK_BUILD_ID_EMPTY = 0,
292 /* with valid build_id and offset */
293 BPF_STACK_BUILD_ID_VALID = 1,
294 /* couldn't get build_id, fallback to ip */
295 BPF_STACK_BUILD_ID_IP = 2,
296};
297
298#define BPF_BUILD_ID_SIZE 20
299struct bpf_stack_build_id {
300 __s32 status;
301 unsigned char build_id[BPF_BUILD_ID_SIZE];
302 union {
303 __u64 offset;
304 __u64 ip;
305 };
306};
307
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700308union bpf_attr {
309 struct { /* anonymous struct used by BPF_MAP_CREATE command */
310 __u32 map_type; /* one of enum bpf_map_type */
311 __u32 key_size; /* size of key in bytes */
312 __u32 value_size; /* size of value in bytes */
313 __u32 max_entries; /* max number of entries in a map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700314 __u32 map_flags; /* BPF_MAP_CREATE related
315 * flags defined above.
316 */
Martin KaFai Lau56f668d2017-03-22 10:00:33 -0700317 __u32 inner_map_fd; /* fd pointing to the inner map */
Martin KaFai Lau96eabe72017-08-18 11:28:00 -0700318 __u32 numa_node; /* numa node (effective only if
319 * BPF_F_NUMA_NODE is set).
320 */
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700321 char map_name[BPF_OBJ_NAME_LEN];
Jakub Kicinskia3884572018-01-11 20:29:09 -0800322 __u32 map_ifindex; /* ifindex of netdev to create on */
Martin KaFai Laua26ca7c2018-04-18 15:56:03 -0700323 __u32 btf_fd; /* fd pointing to a BTF type data */
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -0700324 __u32 btf_key_type_id; /* BTF type_id of the key */
325 __u32 btf_value_type_id; /* BTF type_id of the value */
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700326 };
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700327
328 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
329 __u32 map_fd;
330 __aligned_u64 key;
331 union {
332 __aligned_u64 value;
333 __aligned_u64 next_key;
334 };
Alexei Starovoitov3274f522014-11-13 17:36:44 -0800335 __u64 flags;
Alexei Starovoitovdb20fd22014-09-26 00:16:59 -0700336 };
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700337
338 struct { /* anonymous struct used by BPF_PROG_LOAD command */
339 __u32 prog_type; /* one of enum bpf_prog_type */
340 __u32 insn_cnt;
341 __aligned_u64 insns;
342 __aligned_u64 license;
Alexei Starovoitovcbd35702014-09-26 00:17:03 -0700343 __u32 log_level; /* verbosity level of verifier */
344 __u32 log_size; /* size of user buffer */
345 __aligned_u64 log_buf; /* user supplied buffer */
Alexei Starovoitov25415172015-03-25 12:49:20 -0700346 __u32 kern_version; /* checked when prog_type=kprobe */
David S. Millere07b98d2017-05-10 11:38:07 -0700347 __u32 prog_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -0700348 char prog_name[BPF_OBJ_NAME_LEN];
Jakub Kicinski1f6f4cb2017-11-20 15:21:53 -0800349 __u32 prog_ifindex; /* ifindex of netdev to prep for */
Andrey Ignatov5e43f892018-03-30 15:08:00 -0700350 /* For some prog types expected attach type must be known at
351 * load time to verify attach type specific parts of prog
352 * (context accesses, allowed helpers, etc).
353 */
354 __u32 expected_attach_type;
Yonghong Song838e9692018-11-19 15:29:11 -0800355 __u32 prog_btf_fd; /* fd pointing to BTF type data */
356 __u32 func_info_rec_size; /* userspace bpf_func_info size */
357 __aligned_u64 func_info; /* func info */
358 __u32 func_info_cnt; /* number of bpf_func_info records */
Alexei Starovoitov09756af2014-09-26 00:17:00 -0700359 };
Daniel Borkmannb2197752015-10-29 14:58:09 +0100360
361 struct { /* anonymous struct used by BPF_OBJ_* commands */
362 __aligned_u64 pathname;
363 __u32 bpf_fd;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700364 __u32 file_flags;
Daniel Borkmannb2197752015-10-29 14:58:09 +0100365 };
Daniel Mackf4324552016-11-23 16:52:27 +0100366
367 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
368 __u32 target_fd; /* container object to attach to */
369 __u32 attach_bpf_fd; /* eBPF program to attach */
370 __u32 attach_type;
Alexei Starovoitov7f677632017-02-10 20:28:24 -0800371 __u32 attach_flags;
Daniel Mackf4324552016-11-23 16:52:27 +0100372 };
Alexei Starovoitov1cf1cae2017-03-30 21:45:38 -0700373
374 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
375 __u32 prog_fd;
376 __u32 retval;
377 __u32 data_size_in;
378 __u32 data_size_out;
379 __aligned_u64 data_in;
380 __aligned_u64 data_out;
381 __u32 repeat;
382 __u32 duration;
383 } test;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700384
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700385 struct { /* anonymous struct used by BPF_*_GET_*_ID */
386 union {
387 __u32 start_id;
388 __u32 prog_id;
Martin KaFai Laubd5f5f4e2017-06-05 12:15:50 -0700389 __u32 map_id;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -0700390 __u32 btf_id;
Martin KaFai Laub16d9aa2017-06-05 12:15:49 -0700391 };
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700392 __u32 next_id;
Chenbo Feng6e71b042017-10-18 13:00:22 -0700393 __u32 open_flags;
Martin KaFai Lau34ad5582017-06-05 12:15:48 -0700394 };
Martin KaFai Lau1e270972017-06-05 12:15:52 -0700395
396 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
397 __u32 bpf_fd;
398 __u32 info_len;
399 __aligned_u64 info;
400 } info;
Alexei Starovoitov468e2f62017-10-02 22:50:22 -0700401
402 struct { /* anonymous struct used by BPF_PROG_QUERY command */
403 __u32 target_fd; /* container object to query */
404 __u32 attach_type;
405 __u32 query_flags;
406 __u32 attach_flags;
407 __aligned_u64 prog_ids;
408 __u32 prog_cnt;
409 } query;
Alexei Starovoitovc4f66992018-03-28 12:05:37 -0700410
411 struct {
412 __u64 name;
413 __u32 prog_fd;
414 } raw_tracepoint;
Martin KaFai Lauf56a6532018-04-18 15:56:01 -0700415
416 struct { /* anonymous struct for BPF_BTF_LOAD */
417 __aligned_u64 btf;
418 __aligned_u64 btf_log_buf;
419 __u32 btf_size;
420 __u32 btf_log_size;
421 __u32 btf_log_level;
422 };
Yonghong Song41bdc4b2018-05-24 11:21:09 -0700423
424 struct {
425 __u32 pid; /* input: pid */
426 __u32 fd; /* input: fd */
427 __u32 flags; /* input: flags */
428 __u32 buf_len; /* input/output: buf len */
429 __aligned_u64 buf; /* input/output:
430 * tp_name for tracepoint
431 * symbol for kprobe
432 * filename for uprobe
433 */
434 __u32 prog_id; /* output: prod_id */
435 __u32 fd_type; /* output: BPF_FD_TYPE_* */
436 __u64 probe_offset; /* output: probe_offset */
437 __u64 probe_addr; /* output: probe_addr */
438 } task_fd_query;
Alexei Starovoitov99c55f72014-09-26 00:16:57 -0700439} __attribute__((aligned(8)));
440
Quentin Monnet56a092c2018-04-25 18:16:52 +0100441/* The description below is an attempt at providing documentation to eBPF
442 * developers about the multiple available eBPF helper functions. It can be
443 * parsed and used to produce a manual page. The workflow is the following,
444 * and requires the rst2man utility:
Thomas Grafebb676d2016-10-27 11:23:51 +0200445 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100446 * $ ./scripts/bpf_helpers_doc.py \
447 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
448 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
449 * $ man /tmp/bpf-helpers.7
Thomas Grafebb676d2016-10-27 11:23:51 +0200450 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100451 * Note that in order to produce this external documentation, some RST
452 * formatting is used in the descriptions to get "bold" and "italics" in
453 * manual pages. Also note that the few trailing white spaces are
454 * intentional, removing them would break paragraphs for rst2man.
Thomas Grafebb676d2016-10-27 11:23:51 +0200455 *
Quentin Monnet56a092c2018-04-25 18:16:52 +0100456 * Start of BPF helper function descriptions:
Quentin Monnetad4a5222018-04-25 18:16:53 +0100457 *
458 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
459 * Description
460 * Perform a lookup in *map* for an entry associated to *key*.
461 * Return
462 * Map value associated to *key*, or **NULL** if no entry was
463 * found.
464 *
465 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
466 * Description
467 * Add or update the value of the entry associated to *key* in
468 * *map* with *value*. *flags* is one of:
469 *
470 * **BPF_NOEXIST**
471 * The entry for *key* must not exist in the map.
472 * **BPF_EXIST**
473 * The entry for *key* must already exist in the map.
474 * **BPF_ANY**
475 * No condition on the existence of the entry for *key*.
476 *
477 * Flag value **BPF_NOEXIST** cannot be used for maps of types
478 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
479 * elements always exist), the helper would return an error.
480 * Return
481 * 0 on success, or a negative error in case of failure.
482 *
483 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
484 * Description
485 * Delete entry with *key* from *map*.
486 * Return
487 * 0 on success, or a negative error in case of failure.
488 *
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +0200489 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
490 * Description
491 * Push an element *value* in *map*. *flags* is one of:
492 *
493 * **BPF_EXIST**
494 * If the queue/stack is full, the oldest element is removed to
495 * make room for this.
496 * Return
497 * 0 on success, or a negative error in case of failure.
498 *
499 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
500 * Description
501 * Pop an element from *map*.
502 * Return
503 * 0 on success, or a negative error in case of failure.
504 *
505 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
506 * Description
507 * Get an element from *map* without removing it.
508 * Return
509 * 0 on success, or a negative error in case of failure.
510 *
Quentin Monnetad4a5222018-04-25 18:16:53 +0100511 * int bpf_probe_read(void *dst, u32 size, const void *src)
512 * Description
513 * For tracing programs, safely attempt to read *size* bytes from
514 * address *src* and store the data in *dst*.
515 * Return
516 * 0 on success, or a negative error in case of failure.
517 *
518 * u64 bpf_ktime_get_ns(void)
519 * Description
520 * Return the time elapsed since system boot, in nanoseconds.
521 * Return
522 * Current *ktime*.
523 *
524 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
525 * Description
526 * This helper is a "printk()-like" facility for debugging. It
527 * prints a message defined by format *fmt* (of size *fmt_size*)
528 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
529 * available. It can take up to three additional **u64**
530 * arguments (as an eBPF helpers, the total number of arguments is
531 * limited to five).
532 *
533 * Each time the helper is called, it appends a line to the trace.
534 * The format of the trace is customizable, and the exact output
535 * one will get depends on the options set in
536 * *\/sys/kernel/debug/tracing/trace_options* (see also the
537 * *README* file under the same directory). However, it usually
538 * defaults to something like:
539 *
540 * ::
541 *
542 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
543 *
544 * In the above:
545 *
546 * * ``telnet`` is the name of the current task.
547 * * ``470`` is the PID of the current task.
548 * * ``001`` is the CPU number on which the task is
549 * running.
550 * * In ``.N..``, each character refers to a set of
551 * options (whether irqs are enabled, scheduling
552 * options, whether hard/softirqs are running, level of
553 * preempt_disabled respectively). **N** means that
554 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
555 * are set.
556 * * ``419421.045894`` is a timestamp.
557 * * ``0x00000001`` is a fake value used by BPF for the
558 * instruction pointer register.
559 * * ``<formatted msg>`` is the message formatted with
560 * *fmt*.
561 *
562 * The conversion specifiers supported by *fmt* are similar, but
563 * more limited than for printk(). They are **%d**, **%i**,
564 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
565 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
566 * of field, padding with zeroes, etc.) is available, and the
567 * helper will return **-EINVAL** (but print nothing) if it
568 * encounters an unknown specifier.
569 *
570 * Also, note that **bpf_trace_printk**\ () is slow, and should
571 * only be used for debugging purposes. For this reason, a notice
572 * bloc (spanning several lines) is printed to kernel logs and
573 * states that the helper should not be used "for production use"
574 * the first time this helper is used (or more precisely, when
575 * **trace_printk**\ () buffers are allocated). For passing values
576 * to user space, perf events should be preferred.
577 * Return
578 * The number of bytes written to the buffer, or a negative error
579 * in case of failure.
580 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100581 * u32 bpf_get_prandom_u32(void)
582 * Description
583 * Get a pseudo-random number.
584 *
585 * From a security point of view, this helper uses its own
586 * pseudo-random internal state, and cannot be used to infer the
587 * seed of other random functions in the kernel. However, it is
588 * essential to note that the generator used by the helper is not
589 * cryptographically secure.
590 * Return
591 * A random 32-bit unsigned value.
592 *
593 * u32 bpf_get_smp_processor_id(void)
594 * Description
595 * Get the SMP (symmetric multiprocessing) processor id. Note that
596 * all programs run with preemption disabled, which means that the
597 * SMP processor id is stable during all the execution of the
598 * program.
599 * Return
600 * The SMP id of the processor running the program.
601 *
Quentin Monnetad4a5222018-04-25 18:16:53 +0100602 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
603 * Description
604 * Store *len* bytes from address *from* into the packet
605 * associated to *skb*, at *offset*. *flags* are a combination of
606 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
607 * checksum for the packet after storing the bytes) and
608 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
609 * **->swhash** and *skb*\ **->l4hash** to 0).
610 *
611 * A call to this helper is susceptible to change the underlaying
612 * packet buffer. Therefore, at load time, all checks on pointers
613 * previously done by the verifier are invalidated and must be
614 * performed again, if the helper is used in combination with
615 * direct packet access.
616 * Return
617 * 0 on success, or a negative error in case of failure.
618 *
619 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
620 * Description
621 * Recompute the layer 3 (e.g. IP) checksum for the packet
622 * associated to *skb*. Computation is incremental, so the helper
623 * must know the former value of the header field that was
624 * modified (*from*), the new value of this field (*to*), and the
625 * number of bytes (2 or 4) for this field, stored in *size*.
626 * Alternatively, it is possible to store the difference between
627 * the previous and the new values of the header field in *to*, by
628 * setting *from* and *size* to 0. For both methods, *offset*
629 * indicates the location of the IP checksum within the packet.
630 *
631 * This helper works in combination with **bpf_csum_diff**\ (),
632 * which does not update the checksum in-place, but offers more
633 * flexibility and can handle sizes larger than 2 or 4 for the
634 * checksum to update.
635 *
636 * A call to this helper is susceptible to change the underlaying
637 * packet buffer. Therefore, at load time, all checks on pointers
638 * previously done by the verifier are invalidated and must be
639 * performed again, if the helper is used in combination with
640 * direct packet access.
641 * Return
642 * 0 on success, or a negative error in case of failure.
643 *
644 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
645 * Description
646 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
647 * packet associated to *skb*. Computation is incremental, so the
648 * helper must know the former value of the header field that was
649 * modified (*from*), the new value of this field (*to*), and the
650 * number of bytes (2 or 4) for this field, stored on the lowest
651 * four bits of *flags*. Alternatively, it is possible to store
652 * the difference between the previous and the new values of the
653 * header field in *to*, by setting *from* and the four lowest
654 * bits of *flags* to 0. For both methods, *offset* indicates the
655 * location of the IP checksum within the packet. In addition to
656 * the size of the field, *flags* can be added (bitwise OR) actual
657 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
658 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
659 * for updates resulting in a null checksum the value is set to
660 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
661 * the checksum is to be computed against a pseudo-header.
662 *
663 * This helper works in combination with **bpf_csum_diff**\ (),
664 * which does not update the checksum in-place, but offers more
665 * flexibility and can handle sizes larger than 2 or 4 for the
666 * checksum to update.
667 *
668 * A call to this helper is susceptible to change the underlaying
669 * packet buffer. Therefore, at load time, all checks on pointers
670 * previously done by the verifier are invalidated and must be
671 * performed again, if the helper is used in combination with
672 * direct packet access.
673 * Return
674 * 0 on success, or a negative error in case of failure.
675 *
676 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
677 * Description
678 * This special helper is used to trigger a "tail call", or in
679 * other words, to jump into another eBPF program. The same stack
680 * frame is used (but values on stack and in registers for the
681 * caller are not accessible to the callee). This mechanism allows
682 * for program chaining, either for raising the maximum number of
683 * available eBPF instructions, or to execute given programs in
684 * conditional blocks. For security reasons, there is an upper
685 * limit to the number of successive tail calls that can be
686 * performed.
687 *
688 * Upon call of this helper, the program attempts to jump into a
689 * program referenced at index *index* in *prog_array_map*, a
690 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
691 * *ctx*, a pointer to the context.
692 *
693 * If the call succeeds, the kernel immediately runs the first
694 * instruction of the new program. This is not a function call,
695 * and it never returns to the previous program. If the call
696 * fails, then the helper has no effect, and the caller continues
697 * to run its subsequent instructions. A call can fail if the
698 * destination program for the jump does not exist (i.e. *index*
699 * is superior to the number of entries in *prog_array_map*), or
700 * if the maximum number of tail calls has been reached for this
701 * chain of programs. This limit is defined in the kernel by the
702 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
703 * which is currently set to 32.
704 * Return
705 * 0 on success, or a negative error in case of failure.
706 *
707 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
708 * Description
709 * Clone and redirect the packet associated to *skb* to another
710 * net device of index *ifindex*. Both ingress and egress
711 * interfaces can be used for redirection. The **BPF_F_INGRESS**
712 * value in *flags* is used to make the distinction (ingress path
713 * is selected if the flag is present, egress path otherwise).
714 * This is the only flag supported for now.
715 *
716 * In comparison with **bpf_redirect**\ () helper,
717 * **bpf_clone_redirect**\ () has the associated cost of
718 * duplicating the packet buffer, but this can be executed out of
719 * the eBPF program. Conversely, **bpf_redirect**\ () is more
720 * efficient, but it is handled through an action code where the
721 * redirection happens only after the eBPF program has returned.
722 *
723 * A call to this helper is susceptible to change the underlaying
724 * packet buffer. Therefore, at load time, all checks on pointers
725 * previously done by the verifier are invalidated and must be
726 * performed again, if the helper is used in combination with
727 * direct packet access.
728 * Return
729 * 0 on success, or a negative error in case of failure.
Quentin Monnetc456dec2018-04-25 18:16:54 +0100730 *
731 * u64 bpf_get_current_pid_tgid(void)
732 * Return
733 * A 64-bit integer containing the current tgid and pid, and
734 * created as such:
735 * *current_task*\ **->tgid << 32 \|**
736 * *current_task*\ **->pid**.
737 *
738 * u64 bpf_get_current_uid_gid(void)
739 * Return
740 * A 64-bit integer containing the current GID and UID, and
741 * created as such: *current_gid* **<< 32 \|** *current_uid*.
742 *
743 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
744 * Description
745 * Copy the **comm** attribute of the current task into *buf* of
746 * *size_of_buf*. The **comm** attribute contains the name of
747 * the executable (excluding the path) for the current task. The
748 * *size_of_buf* must be strictly positive. On success, the
749 * helper makes sure that the *buf* is NUL-terminated. On failure,
750 * it is filled with zeroes.
751 * Return
752 * 0 on success, or a negative error in case of failure.
753 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100754 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
755 * Description
756 * Retrieve the classid for the current task, i.e. for the net_cls
757 * cgroup to which *skb* belongs.
758 *
759 * This helper can be used on TC egress path, but not on ingress.
760 *
761 * The net_cls cgroup provides an interface to tag network packets
762 * based on a user-provided identifier for all traffic coming from
763 * the tasks belonging to the related cgroup. See also the related
764 * kernel documentation, available from the Linux sources in file
765 * *Documentation/cgroup-v1/net_cls.txt*.
766 *
767 * The Linux kernel has two versions for cgroups: there are
768 * cgroups v1 and cgroups v2. Both are available to users, who can
769 * use a mixture of them, but note that the net_cls cgroup is for
770 * cgroup v1 only. This makes it incompatible with BPF programs
771 * run on cgroups, which is a cgroup-v2-only feature (a socket can
772 * only hold data for one version of cgroups at a time).
773 *
774 * This helper is only available is the kernel was compiled with
775 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
776 * "**y**" or to "**m**".
777 * Return
778 * The classid, or 0 for the default unconfigured classid.
779 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100780 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
781 * Description
782 * Push a *vlan_tci* (VLAN tag control information) of protocol
783 * *vlan_proto* to the packet associated to *skb*, then update
784 * the checksum. Note that if *vlan_proto* is different from
785 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
786 * be **ETH_P_8021Q**.
787 *
788 * A call to this helper is susceptible to change the underlaying
789 * packet buffer. Therefore, at load time, all checks on pointers
790 * previously done by the verifier are invalidated and must be
791 * performed again, if the helper is used in combination with
792 * direct packet access.
793 * Return
794 * 0 on success, or a negative error in case of failure.
795 *
796 * int bpf_skb_vlan_pop(struct sk_buff *skb)
797 * Description
798 * Pop a VLAN header from the packet associated to *skb*.
799 *
800 * A call to this helper is susceptible to change the underlaying
801 * packet buffer. Therefore, at load time, all checks on pointers
802 * previously done by the verifier are invalidated and must be
803 * performed again, if the helper is used in combination with
804 * direct packet access.
805 * Return
806 * 0 on success, or a negative error in case of failure.
807 *
808 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
809 * Description
810 * Get tunnel metadata. This helper takes a pointer *key* to an
811 * empty **struct bpf_tunnel_key** of **size**, that will be
812 * filled with tunnel metadata for the packet associated to *skb*.
813 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
814 * indicates that the tunnel is based on IPv6 protocol instead of
815 * IPv4.
816 *
817 * The **struct bpf_tunnel_key** is an object that generalizes the
818 * principal parameters used by various tunneling protocols into a
819 * single struct. This way, it can be used to easily make a
820 * decision based on the contents of the encapsulation header,
821 * "summarized" in this struct. In particular, it holds the IP
822 * address of the remote end (IPv4 or IPv6, depending on the case)
823 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
824 * this struct exposes the *key*\ **->tunnel_id**, which is
825 * generally mapped to a VNI (Virtual Network Identifier), making
826 * it programmable together with the **bpf_skb_set_tunnel_key**\
827 * () helper.
828 *
829 * Let's imagine that the following code is part of a program
830 * attached to the TC ingress interface, on one end of a GRE
831 * tunnel, and is supposed to filter out all messages coming from
832 * remote ends with IPv4 address other than 10.0.0.1:
833 *
834 * ::
835 *
836 * int ret;
837 * struct bpf_tunnel_key key = {};
838 *
839 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
840 * if (ret < 0)
841 * return TC_ACT_SHOT; // drop packet
842 *
843 * if (key.remote_ipv4 != 0x0a000001)
844 * return TC_ACT_SHOT; // drop packet
845 *
846 * return TC_ACT_OK; // accept packet
847 *
848 * This interface can also be used with all encapsulation devices
849 * that can operate in "collect metadata" mode: instead of having
850 * one network device per specific configuration, the "collect
851 * metadata" mode only requires a single device where the
852 * configuration can be extracted from this helper.
853 *
854 * This can be used together with various tunnels such as VXLan,
855 * Geneve, GRE or IP in IP (IPIP).
856 * Return
857 * 0 on success, or a negative error in case of failure.
858 *
859 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
860 * Description
861 * Populate tunnel metadata for packet associated to *skb.* The
862 * tunnel metadata is set to the contents of *key*, of *size*. The
863 * *flags* can be set to a combination of the following values:
864 *
865 * **BPF_F_TUNINFO_IPV6**
866 * Indicate that the tunnel is based on IPv6 protocol
867 * instead of IPv4.
868 * **BPF_F_ZERO_CSUM_TX**
869 * For IPv4 packets, add a flag to tunnel metadata
870 * indicating that checksum computation should be skipped
871 * and checksum set to zeroes.
872 * **BPF_F_DONT_FRAGMENT**
873 * Add a flag to tunnel metadata indicating that the
874 * packet should not be fragmented.
875 * **BPF_F_SEQ_NUMBER**
876 * Add a flag to tunnel metadata indicating that a
877 * sequence number should be added to tunnel header before
878 * sending the packet. This flag was added for GRE
879 * encapsulation, but might be used with other protocols
880 * as well in the future.
881 *
882 * Here is a typical usage on the transmit path:
883 *
884 * ::
885 *
886 * struct bpf_tunnel_key key;
887 * populate key ...
888 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
889 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
890 *
891 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
892 * helper for additional information.
893 * Return
894 * 0 on success, or a negative error in case of failure.
895 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100896 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
897 * Description
898 * Read the value of a perf event counter. This helper relies on a
899 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
900 * the perf event counter is selected when *map* is updated with
901 * perf event file descriptors. The *map* is an array whose size
902 * is the number of available CPUs, and each cell contains a value
903 * relative to one CPU. The value to retrieve is indicated by
904 * *flags*, that contains the index of the CPU to look up, masked
905 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
906 * **BPF_F_CURRENT_CPU** to indicate that the value for the
907 * current CPU should be retrieved.
908 *
909 * Note that before Linux 4.13, only hardware perf event can be
910 * retrieved.
911 *
912 * Also, be aware that the newer helper
913 * **bpf_perf_event_read_value**\ () is recommended over
Quentin Monnet3bd5a092018-04-30 11:39:03 +0100914 * **bpf_perf_event_read**\ () in general. The latter has some ABI
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100915 * quirks where error and counter value are used as a return code
916 * (which is wrong to do since ranges may overlap). This issue is
Quentin Monnet3bd5a092018-04-30 11:39:03 +0100917 * fixed with **bpf_perf_event_read_value**\ (), which at the same
918 * time provides more features over the **bpf_perf_event_read**\
919 * () interface. Please refer to the description of
Quentin Monnetc6b5fb82018-04-25 18:16:57 +0100920 * **bpf_perf_event_read_value**\ () for details.
921 * Return
922 * The value of the perf event counter read from the map, or a
923 * negative error code in case of failure.
924 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100925 * int bpf_redirect(u32 ifindex, u64 flags)
926 * Description
927 * Redirect the packet to another net device of index *ifindex*.
928 * This helper is somewhat similar to **bpf_clone_redirect**\
929 * (), except that the packet is not cloned, which provides
930 * increased performance.
931 *
932 * Except for XDP, both ingress and egress interfaces can be used
933 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
934 * to make the distinction (ingress path is selected if the flag
935 * is present, egress path otherwise). Currently, XDP only
936 * supports redirection to the egress interface, and accepts no
937 * flag at all.
938 *
939 * The same effect can be attained with the more generic
940 * **bpf_redirect_map**\ (), which requires specific maps to be
941 * used but offers better performance.
942 * Return
943 * For XDP, the helper returns **XDP_REDIRECT** on success or
944 * **XDP_ABORTED** on error. For other program types, the values
945 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
946 * error.
947 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +0100948 * u32 bpf_get_route_realm(struct sk_buff *skb)
949 * Description
950 * Retrieve the realm or the route, that is to say the
951 * **tclassid** field of the destination for the *skb*. The
952 * indentifier retrieved is a user-provided tag, similar to the
953 * one used with the net_cls cgroup (see description for
954 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
955 * held by a route (a destination entry), not by a task.
956 *
957 * Retrieving this identifier works with the clsact TC egress hook
958 * (see also **tc-bpf(8)**), or alternatively on conventional
959 * classful egress qdiscs, but not on TC ingress path. In case of
960 * clsact TC egress hook, this has the advantage that, internally,
961 * the destination entry has not been dropped yet in the transmit
962 * path. Therefore, the destination entry does not need to be
963 * artificially held via **netif_keep_dst**\ () for a classful
964 * qdisc until the *skb* is freed.
965 *
966 * This helper is available only if the kernel was compiled with
967 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
968 * Return
969 * The realm of the route for the packet associated to *skb*, or 0
970 * if none was found.
971 *
Quentin Monnetc456dec2018-04-25 18:16:54 +0100972 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
973 * Description
974 * Write raw *data* blob into a special BPF perf event held by
975 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
976 * event must have the following attributes: **PERF_SAMPLE_RAW**
977 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
978 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
979 *
980 * The *flags* are used to indicate the index in *map* for which
981 * the value must be put, masked with **BPF_F_INDEX_MASK**.
982 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
983 * to indicate that the index of the current CPU core should be
984 * used.
985 *
986 * The value to write, of *size*, is passed through eBPF stack and
987 * pointed by *data*.
988 *
989 * The context of the program *ctx* needs also be passed to the
990 * helper.
991 *
992 * On user space, a program willing to read the values needs to
993 * call **perf_event_open**\ () on the perf event (either for
994 * one or for all CPUs) and to store the file descriptor into the
995 * *map*. This must be done before the eBPF program can send data
996 * into it. An example is available in file
997 * *samples/bpf/trace_output_user.c* in the Linux kernel source
998 * tree (the eBPF program counterpart is in
999 * *samples/bpf/trace_output_kern.c*).
1000 *
1001 * **bpf_perf_event_output**\ () achieves better performance
1002 * than **bpf_trace_printk**\ () for sharing data with user
1003 * space, and is much better suitable for streaming data from eBPF
1004 * programs.
1005 *
1006 * Note that this helper is not restricted to tracing use cases
1007 * and can be used with programs attached to TC or XDP as well,
1008 * where it allows for passing data to user space listeners. Data
1009 * can be:
1010 *
1011 * * Only custom structs,
1012 * * Only the packet payload, or
1013 * * A combination of both.
1014 * Return
1015 * 0 on success, or a negative error in case of failure.
1016 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001017 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1018 * Description
1019 * This helper was provided as an easy way to load data from a
1020 * packet. It can be used to load *len* bytes from *offset* from
1021 * the packet associated to *skb*, into the buffer pointed by
1022 * *to*.
1023 *
1024 * Since Linux 4.7, usage of this helper has mostly been replaced
1025 * by "direct packet access", enabling packet data to be
1026 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1027 * pointing respectively to the first byte of packet data and to
1028 * the byte after the last byte of packet data. However, it
1029 * remains useful if one wishes to read large quantities of data
1030 * at once from a packet into the eBPF stack.
1031 * Return
1032 * 0 on success, or a negative error in case of failure.
1033 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001034 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1035 * Description
1036 * Walk a user or a kernel stack and return its id. To achieve
1037 * this, the helper needs *ctx*, which is a pointer to the context
1038 * on which the tracing program is executed, and a pointer to a
1039 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1040 *
1041 * The last argument, *flags*, holds the number of stack frames to
1042 * skip (from 0 to 255), masked with
1043 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1044 * a combination of the following flags:
1045 *
1046 * **BPF_F_USER_STACK**
1047 * Collect a user space stack instead of a kernel stack.
1048 * **BPF_F_FAST_STACK_CMP**
1049 * Compare stacks by hash only.
1050 * **BPF_F_REUSE_STACKID**
1051 * If two different stacks hash into the same *stackid*,
1052 * discard the old one.
1053 *
1054 * The stack id retrieved is a 32 bit long integer handle which
1055 * can be further combined with other data (including other stack
1056 * ids) and used as a key into maps. This can be useful for
1057 * generating a variety of graphs (such as flame graphs or off-cpu
1058 * graphs).
1059 *
1060 * For walking a stack, this helper is an improvement over
1061 * **bpf_probe_read**\ (), which can be used with unrolled loops
1062 * but is not efficient and consumes a lot of eBPF instructions.
1063 * Instead, **bpf_get_stackid**\ () can collect up to
1064 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1065 * this limit can be controlled with the **sysctl** program, and
1066 * that it should be manually increased in order to profile long
1067 * user stacks (such as stacks for Java programs). To do so, use:
1068 *
1069 * ::
1070 *
1071 * # sysctl kernel.perf_event_max_stack=<new value>
Quentin Monnetc456dec2018-04-25 18:16:54 +01001072 * Return
1073 * The positive or null stack id on success, or a negative error
1074 * in case of failure.
1075 *
Quentin Monnet1fdd08b2018-04-25 18:16:55 +01001076 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1077 * Description
1078 * Compute a checksum difference, from the raw buffer pointed by
1079 * *from*, of length *from_size* (that must be a multiple of 4),
1080 * towards the raw buffer pointed by *to*, of size *to_size*
1081 * (same remark). An optional *seed* can be added to the value
1082 * (this can be cascaded, the seed may come from a previous call
1083 * to the helper).
1084 *
1085 * This is flexible enough to be used in several ways:
1086 *
1087 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1088 * checksum, it can be used when pushing new data.
1089 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1090 * checksum, it can be used when removing data from a packet.
1091 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1092 * can be used to compute a diff. Note that *from_size* and
1093 * *to_size* do not need to be equal.
1094 *
1095 * This helper can be used in combination with
1096 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1097 * which one can feed in the difference computed with
1098 * **bpf_csum_diff**\ ().
1099 * Return
1100 * The checksum result, or a negative error code in case of
1101 * failure.
1102 *
1103 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1104 * Description
1105 * Retrieve tunnel options metadata for the packet associated to
1106 * *skb*, and store the raw tunnel option data to the buffer *opt*
1107 * of *size*.
1108 *
1109 * This helper can be used with encapsulation devices that can
1110 * operate in "collect metadata" mode (please refer to the related
1111 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1112 * more details). A particular example where this can be used is
1113 * in combination with the Geneve encapsulation protocol, where it
1114 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1115 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1116 * the eBPF program. This allows for full customization of these
1117 * headers.
1118 * Return
1119 * The size of the option data retrieved.
1120 *
1121 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1122 * Description
1123 * Set tunnel options metadata for the packet associated to *skb*
1124 * to the option data contained in the raw buffer *opt* of *size*.
1125 *
1126 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1127 * helper for additional information.
1128 * Return
1129 * 0 on success, or a negative error in case of failure.
1130 *
1131 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1132 * Description
1133 * Change the protocol of the *skb* to *proto*. Currently
1134 * supported are transition from IPv4 to IPv6, and from IPv6 to
1135 * IPv4. The helper takes care of the groundwork for the
1136 * transition, including resizing the socket buffer. The eBPF
1137 * program is expected to fill the new headers, if any, via
1138 * **skb_store_bytes**\ () and to recompute the checksums with
1139 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1140 * (). The main case for this helper is to perform NAT64
1141 * operations out of an eBPF program.
1142 *
1143 * Internally, the GSO type is marked as dodgy so that headers are
1144 * checked and segments are recalculated by the GSO/GRO engine.
1145 * The size for GSO target is adapted as well.
1146 *
1147 * All values for *flags* are reserved for future usage, and must
1148 * be left at zero.
1149 *
1150 * A call to this helper is susceptible to change the underlaying
1151 * packet buffer. Therefore, at load time, all checks on pointers
1152 * previously done by the verifier are invalidated and must be
1153 * performed again, if the helper is used in combination with
1154 * direct packet access.
1155 * Return
1156 * 0 on success, or a negative error in case of failure.
1157 *
1158 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1159 * Description
1160 * Change the packet type for the packet associated to *skb*. This
1161 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1162 * the eBPF program does not have a write access to *skb*\
1163 * **->pkt_type** beside this helper. Using a helper here allows
1164 * for graceful handling of errors.
1165 *
1166 * The major use case is to change incoming *skb*s to
1167 * **PACKET_HOST** in a programmatic way instead of having to
1168 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1169 * example.
1170 *
1171 * Note that *type* only allows certain values. At this time, they
1172 * are:
1173 *
1174 * **PACKET_HOST**
1175 * Packet is for us.
1176 * **PACKET_BROADCAST**
1177 * Send packet to all.
1178 * **PACKET_MULTICAST**
1179 * Send packet to group.
1180 * **PACKET_OTHERHOST**
1181 * Send packet to someone else.
1182 * Return
1183 * 0 on success, or a negative error in case of failure.
1184 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001185 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1186 * Description
1187 * Check whether *skb* is a descendant of the cgroup2 held by
1188 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1189 * Return
1190 * The return value depends on the result of the test, and can be:
1191 *
1192 * * 0, if the *skb* failed the cgroup2 descendant test.
1193 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1194 * * A negative error code, if an error occurred.
1195 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001196 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1197 * Description
1198 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1199 * not set, in particular if the hash was cleared due to mangling,
1200 * recompute this hash. Later accesses to the hash can be done
1201 * directly with *skb*\ **->hash**.
1202 *
1203 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1204 * prototype with **bpf_skb_change_proto**\ (), or calling
1205 * **bpf_skb_store_bytes**\ () with the
1206 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1207 * the hash and to trigger a new computation for the next call to
1208 * **bpf_get_hash_recalc**\ ().
1209 * Return
1210 * The 32-bit hash.
1211 *
Quentin Monnetc456dec2018-04-25 18:16:54 +01001212 * u64 bpf_get_current_task(void)
1213 * Return
1214 * A pointer to the current task struct.
Quentin Monnetfa156012018-04-25 18:16:56 +01001215 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001216 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1217 * Description
1218 * Attempt in a safe way to write *len* bytes from the buffer
1219 * *src* to *dst* in memory. It only works for threads that are in
1220 * user context, and *dst* must be a valid user space address.
1221 *
1222 * This helper should not be used to implement any kind of
1223 * security mechanism because of TOC-TOU attacks, but rather to
1224 * debug, divert, and manipulate execution of semi-cooperative
1225 * processes.
1226 *
1227 * Keep in mind that this feature is meant for experiments, and it
1228 * has a risk of crashing the system and running programs.
1229 * Therefore, when an eBPF program using this helper is attached,
1230 * a warning including PID and process name is printed to kernel
1231 * logs.
1232 * Return
1233 * 0 on success, or a negative error in case of failure.
1234 *
1235 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1236 * Description
1237 * Check whether the probe is being run is the context of a given
1238 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1239 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1240 * Return
1241 * The return value depends on the result of the test, and can be:
1242 *
1243 * * 0, if the *skb* task belongs to the cgroup2.
1244 * * 1, if the *skb* task does not belong to the cgroup2.
1245 * * A negative error code, if an error occurred.
1246 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001247 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1248 * Description
1249 * Resize (trim or grow) the packet associated to *skb* to the
1250 * new *len*. The *flags* are reserved for future usage, and must
1251 * be left at zero.
1252 *
1253 * The basic idea is that the helper performs the needed work to
1254 * change the size of the packet, then the eBPF program rewrites
1255 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1256 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1257 * and others. This helper is a slow path utility intended for
1258 * replies with control messages. And because it is targeted for
1259 * slow path, the helper itself can afford to be slow: it
1260 * implicitly linearizes, unclones and drops offloads from the
1261 * *skb*.
1262 *
1263 * A call to this helper is susceptible to change the underlaying
1264 * packet buffer. Therefore, at load time, all checks on pointers
1265 * previously done by the verifier are invalidated and must be
1266 * performed again, if the helper is used in combination with
1267 * direct packet access.
1268 * Return
1269 * 0 on success, or a negative error in case of failure.
1270 *
1271 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1272 * Description
1273 * Pull in non-linear data in case the *skb* is non-linear and not
1274 * all of *len* are part of the linear section. Make *len* bytes
1275 * from *skb* readable and writable. If a zero value is passed for
1276 * *len*, then the whole length of the *skb* is pulled.
1277 *
1278 * This helper is only needed for reading and writing with direct
1279 * packet access.
1280 *
1281 * For direct packet access, testing that offsets to access
1282 * are within packet boundaries (test on *skb*\ **->data_end**) is
1283 * susceptible to fail if offsets are invalid, or if the requested
1284 * data is in non-linear parts of the *skb*. On failure the
1285 * program can just bail out, or in the case of a non-linear
1286 * buffer, use a helper to make the data available. The
1287 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1288 * the data. Another one consists in using **bpf_skb_pull_data**
1289 * to pull in once the non-linear parts, then retesting and
1290 * eventually access the data.
1291 *
1292 * At the same time, this also makes sure the *skb* is uncloned,
1293 * which is a necessary condition for direct write. As this needs
1294 * to be an invariant for the write part only, the verifier
1295 * detects writes and adds a prologue that is calling
1296 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1297 * the very beginning in case it is indeed cloned.
1298 *
1299 * A call to this helper is susceptible to change the underlaying
1300 * packet buffer. Therefore, at load time, all checks on pointers
1301 * previously done by the verifier are invalidated and must be
1302 * performed again, if the helper is used in combination with
1303 * direct packet access.
1304 * Return
1305 * 0 on success, or a negative error in case of failure.
1306 *
1307 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1308 * Description
1309 * Add the checksum *csum* into *skb*\ **->csum** in case the
1310 * driver has supplied a checksum for the entire packet into that
1311 * field. Return an error otherwise. This helper is intended to be
1312 * used in combination with **bpf_csum_diff**\ (), in particular
1313 * when the checksum needs to be updated after data has been
1314 * written into the packet through direct packet access.
1315 * Return
1316 * The checksum on success, or a negative error code in case of
1317 * failure.
1318 *
1319 * void bpf_set_hash_invalid(struct sk_buff *skb)
1320 * Description
1321 * Invalidate the current *skb*\ **->hash**. It can be used after
1322 * mangling on headers through direct packet access, in order to
1323 * indicate that the hash is outdated and to trigger a
1324 * recalculation the next time the kernel tries to access this
1325 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1326 *
1327 * int bpf_get_numa_node_id(void)
1328 * Description
1329 * Return the id of the current NUMA node. The primary use case
1330 * for this helper is the selection of sockets for the local NUMA
1331 * node, when the program is attached to sockets using the
1332 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1333 * but the helper is also available to other eBPF program types,
1334 * similarly to **bpf_get_smp_processor_id**\ ().
1335 * Return
1336 * The id of current NUMA node.
1337 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001338 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1339 * Description
1340 * Grows headroom of packet associated to *skb* and adjusts the
1341 * offset of the MAC header accordingly, adding *len* bytes of
1342 * space. It automatically extends and reallocates memory as
1343 * required.
1344 *
1345 * This helper can be used on a layer 3 *skb* to push a MAC header
1346 * for redirection into a layer 2 device.
1347 *
1348 * All values for *flags* are reserved for future usage, and must
1349 * be left at zero.
1350 *
1351 * A call to this helper is susceptible to change the underlaying
1352 * packet buffer. Therefore, at load time, all checks on pointers
1353 * previously done by the verifier are invalidated and must be
1354 * performed again, if the helper is used in combination with
1355 * direct packet access.
1356 * Return
1357 * 0 on success, or a negative error in case of failure.
1358 *
1359 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1360 * Description
1361 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1362 * it is possible to use a negative value for *delta*. This helper
1363 * can be used to prepare the packet for pushing or popping
1364 * headers.
1365 *
1366 * A call to this helper is susceptible to change the underlaying
1367 * packet buffer. Therefore, at load time, all checks on pointers
1368 * previously done by the verifier are invalidated and must be
1369 * performed again, if the helper is used in combination with
1370 * direct packet access.
1371 * Return
1372 * 0 on success, or a negative error in case of failure.
1373 *
1374 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1375 * Description
1376 * Copy a NUL terminated string from an unsafe address
1377 * *unsafe_ptr* to *dst*. The *size* should include the
1378 * terminating NUL byte. In case the string length is smaller than
1379 * *size*, the target is not padded with further NUL bytes. If the
1380 * string length is larger than *size*, just *size*-1 bytes are
1381 * copied and the last byte is set to NUL.
1382 *
1383 * On success, the length of the copied string is returned. This
1384 * makes this helper useful in tracing programs for reading
1385 * strings, and more importantly to get its length at runtime. See
1386 * the following snippet:
1387 *
1388 * ::
1389 *
1390 * SEC("kprobe/sys_open")
1391 * void bpf_sys_open(struct pt_regs *ctx)
1392 * {
1393 * char buf[PATHLEN]; // PATHLEN is defined to 256
1394 * int res = bpf_probe_read_str(buf, sizeof(buf),
1395 * ctx->di);
1396 *
1397 * // Consume buf, for example push it to
1398 * // userspace via bpf_perf_event_output(); we
1399 * // can use res (the string length) as event
1400 * // size, after checking its boundaries.
1401 * }
1402 *
1403 * In comparison, using **bpf_probe_read()** helper here instead
1404 * to read the string would require to estimate the length at
1405 * compile time, and would often result in copying more memory
1406 * than necessary.
1407 *
1408 * Another useful use case is when parsing individual process
1409 * arguments or individual environment variables navigating
1410 * *current*\ **->mm->arg_start** and *current*\
1411 * **->mm->env_start**: using this helper and the return value,
1412 * one can quickly iterate at the right offset of the memory area.
1413 * Return
1414 * On success, the strictly positive length of the string,
1415 * including the trailing NUL character. On error, a negative
1416 * value.
1417 *
1418 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1419 * Description
1420 * If the **struct sk_buff** pointed by *skb* has a known socket,
1421 * retrieve the cookie (generated by the kernel) of this socket.
1422 * If no cookie has been set yet, generate a new cookie. Once
1423 * generated, the socket cookie remains stable for the life of the
1424 * socket. This helper can be useful for monitoring per socket
1425 * networking traffic statistics as it provides a unique socket
1426 * identifier per namespace.
1427 * Return
1428 * A 8-byte long non-decreasing number on success, or 0 if the
1429 * socket field is missing inside *skb*.
1430 *
Andrey Ignatovd692f112018-07-30 17:42:28 -07001431 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1432 * Description
1433 * Equivalent to bpf_get_socket_cookie() helper that accepts
1434 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1435 * Return
1436 * A 8-byte long non-decreasing number.
1437 *
1438 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1439 * Description
1440 * Equivalent to bpf_get_socket_cookie() helper that accepts
1441 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1442 * Return
1443 * A 8-byte long non-decreasing number.
1444 *
Quentin Monnetc6b5fb82018-04-25 18:16:57 +01001445 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1446 * Return
1447 * The owner UID of the socket associated to *skb*. If the socket
1448 * is **NULL**, or if it is not a full socket (i.e. if it is a
1449 * time-wait or a request socket instead), **overflowuid** value
1450 * is returned (note that **overflowuid** might also be the actual
1451 * UID value for the socket).
1452 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001453 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1454 * Description
1455 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1456 * to value *hash*.
1457 * Return
1458 * 0
1459 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001460 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001461 * Description
1462 * Emulate a call to **setsockopt()** on the socket associated to
1463 * *bpf_socket*, which must be a full socket. The *level* at
1464 * which the option resides and the name *optname* of the option
1465 * must be specified, see **setsockopt(2)** for more information.
1466 * The option value of length *optlen* is pointed by *optval*.
1467 *
1468 * This helper actually implements a subset of **setsockopt()**.
1469 * It supports the following *level*\ s:
1470 *
1471 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1472 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1473 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1474 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1475 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1476 * **TCP_BPF_SNDCWND_CLAMP**.
1477 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1478 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1479 * Return
1480 * 0 on success, or a negative error in case of failure.
1481 *
Nicolas Dichtelb55cbc82018-10-17 16:24:48 +02001482 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
Quentin Monnetfa156012018-04-25 18:16:56 +01001483 * Description
1484 * Grow or shrink the room for data in the packet associated to
1485 * *skb* by *len_diff*, and according to the selected *mode*.
1486 *
1487 * There is a single supported mode at this time:
1488 *
1489 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1490 * (room space is added or removed below the layer 3 header).
1491 *
1492 * All values for *flags* are reserved for future usage, and must
1493 * be left at zero.
1494 *
1495 * A call to this helper is susceptible to change the underlaying
1496 * packet buffer. Therefore, at load time, all checks on pointers
1497 * previously done by the verifier are invalidated and must be
1498 * performed again, if the helper is used in combination with
1499 * direct packet access.
1500 * Return
1501 * 0 on success, or a negative error in case of failure.
1502 *
Quentin Monnetab127042018-04-25 18:16:59 +01001503 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1504 * Description
1505 * Redirect the packet to the endpoint referenced by *map* at
1506 * index *key*. Depending on its type, this *map* can contain
1507 * references to net devices (for forwarding packets through other
1508 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1509 * but this is only implemented for native XDP (with driver
1510 * support) as of this writing).
1511 *
1512 * All values for *flags* are reserved for future usage, and must
1513 * be left at zero.
1514 *
1515 * When used to redirect packets to net devices, this helper
1516 * provides a high performance increase over **bpf_redirect**\ ().
1517 * This is due to various implementation details of the underlying
1518 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1519 * () tries to send packet as a "bulk" to the device.
1520 * Return
1521 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1522 *
1523 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1524 * Description
1525 * Redirect the packet to the socket referenced by *map* (of type
1526 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1527 * egress interfaces can be used for redirection. The
1528 * **BPF_F_INGRESS** value in *flags* is used to make the
1529 * distinction (ingress path is selected if the flag is present,
1530 * egress path otherwise). This is the only flag supported for now.
1531 * Return
1532 * **SK_PASS** on success, or **SK_DROP** on error.
1533 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001534 * 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 +01001535 * Description
1536 * Add an entry to, or update a *map* referencing sockets. The
1537 * *skops* is used as a new value for the entry associated to
1538 * *key*. *flags* is one of:
1539 *
1540 * **BPF_NOEXIST**
1541 * The entry for *key* must not exist in the map.
1542 * **BPF_EXIST**
1543 * The entry for *key* must already exist in the map.
1544 * **BPF_ANY**
1545 * No condition on the existence of the entry for *key*.
1546 *
1547 * If the *map* has eBPF programs (parser and verdict), those will
1548 * be inherited by the socket being added. If the socket is
1549 * already attached to eBPF programs, this results in an error.
1550 * Return
1551 * 0 on success, or a negative error in case of failure.
1552 *
Quentin Monnetfa156012018-04-25 18:16:56 +01001553 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1554 * Description
1555 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1556 * *delta* (which can be positive or negative). Note that this
1557 * operation modifies the address stored in *xdp_md*\ **->data**,
1558 * so the latter must be loaded only after the helper has been
1559 * called.
1560 *
1561 * The use of *xdp_md*\ **->data_meta** is optional and programs
1562 * are not required to use it. The rationale is that when the
1563 * packet is processed with XDP (e.g. as DoS filter), it is
1564 * possible to push further meta data along with it before passing
1565 * to the stack, and to give the guarantee that an ingress eBPF
1566 * program attached as a TC classifier on the same device can pick
1567 * this up for further post-processing. Since TC works with socket
1568 * buffers, it remains possible to set from XDP the **mark** or
1569 * **priority** pointers, or other pointers for the socket buffer.
1570 * Having this scratch space generic and programmable allows for
1571 * more flexibility as the user is free to store whatever meta
1572 * data they need.
1573 *
1574 * A call to this helper is susceptible to change the underlaying
1575 * packet buffer. Therefore, at load time, all checks on pointers
1576 * previously done by the verifier are invalidated and must be
1577 * performed again, if the helper is used in combination with
1578 * direct packet access.
1579 * Return
1580 * 0 on success, or a negative error in case of failure.
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001581 *
1582 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1583 * Description
1584 * Read the value of a perf event counter, and store it into *buf*
1585 * of size *buf_size*. This helper relies on a *map* of type
1586 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1587 * counter is selected when *map* is updated with perf event file
1588 * descriptors. The *map* is an array whose size is the number of
1589 * available CPUs, and each cell contains a value relative to one
1590 * CPU. The value to retrieve is indicated by *flags*, that
1591 * contains the index of the CPU to look up, masked with
1592 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1593 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1594 * current CPU should be retrieved.
1595 *
1596 * This helper behaves in a way close to
1597 * **bpf_perf_event_read**\ () helper, save that instead of
1598 * just returning the value observed, it fills the *buf*
1599 * structure. This allows for additional data to be retrieved: in
1600 * particular, the enabled and running times (in *buf*\
1601 * **->enabled** and *buf*\ **->running**, respectively) are
1602 * copied. In general, **bpf_perf_event_read_value**\ () is
1603 * recommended over **bpf_perf_event_read**\ (), which has some
1604 * ABI issues and provides fewer functionalities.
1605 *
1606 * These values are interesting, because hardware PMU (Performance
1607 * Monitoring Unit) counters are limited resources. When there are
1608 * more PMU based perf events opened than available counters,
1609 * kernel will multiplex these events so each event gets certain
1610 * percentage (but not all) of the PMU time. In case that
1611 * multiplexing happens, the number of samples or counter value
1612 * will not reflect the case compared to when no multiplexing
1613 * occurs. This makes comparison between different runs difficult.
1614 * Typically, the counter value should be normalized before
1615 * comparing to other experiments. The usual normalization is done
1616 * as follows.
1617 *
1618 * ::
1619 *
1620 * normalized_counter = counter * t_enabled / t_running
1621 *
1622 * Where t_enabled is the time enabled for event and t_running is
1623 * the time running for event since last normalization. The
1624 * enabled and running times are accumulated since the perf event
1625 * open. To achieve scaling factor between two invocations of an
1626 * eBPF program, users can can use CPU id as the key (which is
1627 * typical for perf array usage model) to remember the previous
1628 * value and do the calculation inside the eBPF program.
1629 * Return
1630 * 0 on success, or a negative error in case of failure.
1631 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001632 * 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 +01001633 * Description
1634 * For en eBPF program attached to a perf event, retrieve the
1635 * value of the event counter associated to *ctx* and store it in
1636 * the structure pointed by *buf* and of size *buf_size*. Enabled
1637 * and running times are also stored in the structure (see
1638 * description of helper **bpf_perf_event_read_value**\ () for
1639 * more details).
1640 * Return
1641 * 0 on success, or a negative error in case of failure.
1642 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001643 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001644 * Description
1645 * Emulate a call to **getsockopt()** on the socket associated to
1646 * *bpf_socket*, which must be a full socket. The *level* at
1647 * which the option resides and the name *optname* of the option
1648 * must be specified, see **getsockopt(2)** for more information.
1649 * The retrieved value is stored in the structure pointed by
1650 * *opval* and of length *optlen*.
1651 *
1652 * This helper actually implements a subset of **getsockopt()**.
1653 * It supports the following *level*\ s:
1654 *
1655 * * **IPPROTO_TCP**, which supports *optname*
1656 * **TCP_CONGESTION**.
1657 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1658 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1659 * Return
1660 * 0 on success, or a negative error in case of failure.
1661 *
1662 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1663 * Description
1664 * Used for error injection, this helper uses kprobes to override
1665 * the return value of the probed function, and to set it to *rc*.
1666 * The first argument is the context *regs* on which the kprobe
1667 * works.
1668 *
1669 * This helper works by setting setting the PC (program counter)
1670 * to an override function which is run in place of the original
1671 * probed function. This means the probed function is not run at
1672 * all. The replacement function just returns with the required
1673 * value.
1674 *
1675 * This helper has security implications, and thus is subject to
1676 * restrictions. It is only available if the kernel was compiled
1677 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1678 * option, and in this case it only works on functions tagged with
1679 * **ALLOW_ERROR_INJECTION** in the kernel code.
1680 *
1681 * Also, the helper is only available for the architectures having
1682 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1683 * x86 architecture is the only one to support this feature.
1684 * Return
1685 * 0
1686 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001687 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001688 * Description
1689 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1690 * for the full TCP socket associated to *bpf_sock_ops* to
1691 * *argval*.
1692 *
1693 * The primary use of this field is to determine if there should
1694 * be calls to eBPF programs of type
1695 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1696 * code. A program of the same type can change its value, per
1697 * connection and as necessary, when the connection is
1698 * established. This field is directly accessible for reading, but
1699 * this helper must be used for updates in order to return an
1700 * error if an eBPF program tries to set a callback that is not
1701 * supported in the current kernel.
1702 *
1703 * The supported callback values that *argval* can combine are:
1704 *
1705 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1706 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1707 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1708 *
1709 * Here are some examples of where one could call such eBPF
1710 * program:
1711 *
1712 * * When RTO fires.
1713 * * When a packet is retransmitted.
1714 * * When the connection terminates.
1715 * * When a packet is sent.
1716 * * When a packet is received.
1717 * Return
1718 * Code **-EINVAL** if the socket is not a full TCP socket;
1719 * otherwise, a positive number containing the bits that could not
1720 * be set is returned (which comes down to 0 if all bits were set
1721 * as required).
1722 *
Quentin Monnetab127042018-04-25 18:16:59 +01001723 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1724 * Description
1725 * This helper is used in programs implementing policies at the
1726 * socket level. If the message *msg* is allowed to pass (i.e. if
1727 * the verdict eBPF program returns **SK_PASS**), redirect it to
1728 * the socket referenced by *map* (of type
1729 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1730 * egress interfaces can be used for redirection. The
1731 * **BPF_F_INGRESS** value in *flags* is used to make the
1732 * distinction (ingress path is selected if the flag is present,
1733 * egress path otherwise). This is the only flag supported for now.
1734 * Return
1735 * **SK_PASS** on success, or **SK_DROP** on error.
1736 *
1737 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1738 * Description
1739 * For socket policies, apply the verdict of the eBPF program to
1740 * the next *bytes* (number of bytes) of message *msg*.
1741 *
1742 * For example, this helper can be used in the following cases:
1743 *
1744 * * A single **sendmsg**\ () or **sendfile**\ () system call
1745 * contains multiple logical messages that the eBPF program is
1746 * supposed to read and for which it should apply a verdict.
1747 * * An eBPF program only cares to read the first *bytes* of a
1748 * *msg*. If the message has a large payload, then setting up
1749 * and calling the eBPF program repeatedly for all bytes, even
1750 * though the verdict is already known, would create unnecessary
1751 * overhead.
1752 *
1753 * When called from within an eBPF program, the helper sets a
1754 * counter internal to the BPF infrastructure, that is used to
1755 * apply the last verdict to the next *bytes*. If *bytes* is
1756 * smaller than the current data being processed from a
1757 * **sendmsg**\ () or **sendfile**\ () system call, the first
1758 * *bytes* will be sent and the eBPF program will be re-run with
1759 * the pointer for start of data pointing to byte number *bytes*
1760 * **+ 1**. If *bytes* is larger than the current data being
1761 * processed, then the eBPF verdict will be applied to multiple
1762 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1763 * consumed.
1764 *
1765 * Note that if a socket closes with the internal counter holding
1766 * a non-zero value, this is not a problem because data is not
1767 * being buffered for *bytes* and is sent as it is received.
1768 * Return
1769 * 0
1770 *
1771 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1772 * Description
1773 * For socket policies, prevent the execution of the verdict eBPF
1774 * program for message *msg* until *bytes* (byte number) have been
1775 * accumulated.
1776 *
1777 * This can be used when one needs a specific number of bytes
1778 * before a verdict can be assigned, even if the data spans
1779 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1780 * case would be a user calling **sendmsg**\ () repeatedly with
1781 * 1-byte long message segments. Obviously, this is bad for
1782 * performance, but it is still valid. If the eBPF program needs
1783 * *bytes* bytes to validate a header, this helper can be used to
1784 * prevent the eBPF program to be called again until *bytes* have
1785 * been accumulated.
1786 * Return
1787 * 0
1788 *
1789 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1790 * Description
1791 * For socket policies, pull in non-linear data from user space
1792 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1793 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1794 * respectively.
1795 *
1796 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1797 * *msg* it can only parse data that the (**data**, **data_end**)
1798 * pointers have already consumed. For **sendmsg**\ () hooks this
1799 * is likely the first scatterlist element. But for calls relying
1800 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1801 * be the range (**0**, **0**) because the data is shared with
1802 * user space and by default the objective is to avoid allowing
1803 * user space to modify data while (or after) eBPF verdict is
1804 * being decided. This helper can be used to pull in data and to
1805 * set the start and end pointer to given values. Data will be
1806 * copied if necessary (i.e. if data was not linear and if start
1807 * and end pointers do not point to the same chunk).
1808 *
1809 * A call to this helper is susceptible to change the underlaying
1810 * packet buffer. Therefore, at load time, all checks on pointers
1811 * previously done by the verifier are invalidated and must be
1812 * performed again, if the helper is used in combination with
1813 * direct packet access.
1814 *
1815 * All values for *flags* are reserved for future usage, and must
1816 * be left at zero.
1817 * Return
1818 * 0 on success, or a negative error in case of failure.
1819 *
Andrey Ignatova3ef8e92018-04-28 16:06:19 -07001820 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
Quentin Monnet7aa79a82018-04-25 18:16:58 +01001821 * Description
1822 * Bind the socket associated to *ctx* to the address pointed by
1823 * *addr*, of length *addr_len*. This allows for making outgoing
1824 * connection from the desired IP address, which can be useful for
1825 * example when all processes inside a cgroup should use one
1826 * single IP address on a host that has multiple IP configured.
1827 *
1828 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1829 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1830 * **AF_INET6**). Looking for a free port to bind to can be
1831 * expensive, therefore binding to port is not permitted by the
1832 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1833 * must be set to zero.
1834 * Return
1835 * 0 on success, or a negative error in case of failure.
Quentin Monnet2d020dd2018-04-25 18:17:00 +01001836 *
1837 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1838 * Description
1839 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1840 * only possible to shrink the packet as of this writing,
1841 * therefore *delta* must be a negative integer.
1842 *
1843 * A call to this helper is susceptible to change the underlaying
1844 * packet buffer. Therefore, at load time, all checks on pointers
1845 * previously done by the verifier are invalidated and must be
1846 * performed again, if the helper is used in combination with
1847 * direct packet access.
1848 * Return
1849 * 0 on success, or a negative error in case of failure.
1850 *
1851 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1852 * Description
1853 * Retrieve the XFRM state (IP transform framework, see also
1854 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1855 *
1856 * The retrieved value is stored in the **struct bpf_xfrm_state**
1857 * pointed by *xfrm_state* and of length *size*.
1858 *
1859 * All values for *flags* are reserved for future usage, and must
1860 * be left at zero.
1861 *
1862 * This helper is available only if the kernel was compiled with
1863 * **CONFIG_XFRM** configuration option.
1864 * Return
1865 * 0 on success, or a negative error in case of failure.
Yonghong Songc195651e2018-04-28 22:28:08 -07001866 *
1867 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1868 * Description
Quentin Monnet79552fb2018-04-30 11:39:04 +01001869 * Return a user or a kernel stack in bpf program provided buffer.
1870 * To achieve this, the helper needs *ctx*, which is a pointer
1871 * to the context on which the tracing program is executed.
1872 * To store the stacktrace, the bpf program provides *buf* with
1873 * a nonnegative *size*.
Yonghong Songc195651e2018-04-28 22:28:08 -07001874 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001875 * The last argument, *flags*, holds the number of stack frames to
1876 * skip (from 0 to 255), masked with
1877 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1878 * the following flags:
Yonghong Songc195651e2018-04-28 22:28:08 -07001879 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001880 * **BPF_F_USER_STACK**
1881 * Collect a user space stack instead of a kernel stack.
1882 * **BPF_F_USER_BUILD_ID**
1883 * Collect buildid+offset instead of ips for user stack,
1884 * only valid if **BPF_F_USER_STACK** is also specified.
Yonghong Songc195651e2018-04-28 22:28:08 -07001885 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001886 * **bpf_get_stack**\ () can collect up to
1887 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1888 * to sufficient large buffer size. Note that
1889 * this limit can be controlled with the **sysctl** program, and
1890 * that it should be manually increased in order to profile long
1891 * user stacks (such as stacks for Java programs). To do so, use:
Yonghong Songc195651e2018-04-28 22:28:08 -07001892 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001893 * ::
Yonghong Songc195651e2018-04-28 22:28:08 -07001894 *
Quentin Monnet79552fb2018-04-30 11:39:04 +01001895 * # sysctl kernel.perf_event_max_stack=<new value>
Yonghong Songc195651e2018-04-28 22:28:08 -07001896 * Return
Quentin Monnet7a279e92018-05-29 12:27:44 +01001897 * A non-negative value equal to or less than *size* on success,
1898 * or a negative error in case of failure.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001899 *
Quentin Monnet2bae79d2018-07-12 12:52:22 +01001900 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001901 * Description
1902 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1903 * it provides an easy way to load *len* bytes from *offset*
1904 * from the packet associated to *skb*, into the buffer pointed
1905 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1906 * a fifth argument *start_header* exists in order to select a
1907 * base offset to start from. *start_header* can be one of:
1908 *
1909 * **BPF_HDR_START_MAC**
1910 * Base offset to load data from is *skb*'s mac header.
1911 * **BPF_HDR_START_NET**
1912 * Base offset to load data from is *skb*'s network header.
1913 *
1914 * In general, "direct packet access" is the preferred method to
1915 * access packet data, however, this helper is in particular useful
1916 * in socket filters where *skb*\ **->data** does not always point
1917 * to the start of the mac header and where "direct packet access"
1918 * is not available.
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02001919 * Return
1920 * 0 on success, or a negative error in case of failure.
1921 *
David Ahern87f5fc72018-05-09 20:34:26 -07001922 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1923 * Description
1924 * Do FIB lookup in kernel tables using parameters in *params*.
1925 * If lookup is successful and result shows packet is to be
1926 * forwarded, the neighbor tables are searched for the nexthop.
1927 * If successful (ie., FIB lookup shows forwarding and nexthop
David Ahernfa898d72018-05-29 10:58:07 -07001928 * is resolved), the nexthop address is returned in ipv4_dst
1929 * or ipv6_dst based on family, smac is set to mac address of
1930 * egress device, dmac is set to nexthop mac address, rt_metric
David Ahern4c795792018-06-26 16:21:18 -07001931 * is set to metric from route (IPv4/IPv6 only), and ifindex
1932 * is set to the device index of the nexthop from the FIB lookup.
David Ahern87f5fc72018-05-09 20:34:26 -07001933 *
1934 * *plen* argument is the size of the passed in struct.
Quentin Monnet7a279e92018-05-29 12:27:44 +01001935 * *flags* argument can be a combination of one or more of the
1936 * following values:
David Ahern87f5fc72018-05-09 20:34:26 -07001937 *
Quentin Monnet7a279e92018-05-29 12:27:44 +01001938 * **BPF_FIB_LOOKUP_DIRECT**
1939 * Do a direct table lookup vs full lookup using FIB
1940 * rules.
1941 * **BPF_FIB_LOOKUP_OUTPUT**
1942 * Perform lookup from an egress perspective (default is
1943 * ingress).
David Ahern87f5fc72018-05-09 20:34:26 -07001944 *
1945 * *ctx* is either **struct xdp_md** for XDP programs or
1946 * **struct sk_buff** tc cls_act programs.
David Ahern87f5fc72018-05-09 20:34:26 -07001947 * Return
David Ahern4c795792018-06-26 16:21:18 -07001948 * * < 0 if any input argument is invalid
1949 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1950 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
Quentin Monnet2bae79d2018-07-12 12:52:22 +01001951 * packet is not forwarded or needs assist from full stack
John Fastabend81110382018-05-14 10:00:17 -07001952 *
1953 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1954 * Description
1955 * Add an entry to, or update a sockhash *map* referencing sockets.
1956 * The *skops* is used as a new value for the entry associated to
1957 * *key*. *flags* is one of:
1958 *
1959 * **BPF_NOEXIST**
1960 * The entry for *key* must not exist in the map.
1961 * **BPF_EXIST**
1962 * The entry for *key* must already exist in the map.
1963 * **BPF_ANY**
1964 * No condition on the existence of the entry for *key*.
1965 *
1966 * If the *map* has eBPF programs (parser and verdict), those will
1967 * be inherited by the socket being added. If the socket is
1968 * already attached to eBPF programs, this results in an error.
1969 * Return
1970 * 0 on success, or a negative error in case of failure.
1971 *
1972 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1973 * Description
1974 * This helper is used in programs implementing policies at the
1975 * socket level. If the message *msg* is allowed to pass (i.e. if
1976 * the verdict eBPF program returns **SK_PASS**), redirect it to
1977 * the socket referenced by *map* (of type
1978 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1979 * egress interfaces can be used for redirection. The
1980 * **BPF_F_INGRESS** value in *flags* is used to make the
1981 * distinction (ingress path is selected if the flag is present,
1982 * egress path otherwise). This is the only flag supported for now.
1983 * Return
1984 * **SK_PASS** on success, or **SK_DROP** on error.
1985 *
1986 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1987 * Description
1988 * This helper is used in programs implementing policies at the
1989 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1990 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1991 * to the socket referenced by *map* (of type
1992 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1993 * egress interfaces can be used for redirection. The
1994 * **BPF_F_INGRESS** value in *flags* is used to make the
1995 * distinction (ingress path is selected if the flag is present,
1996 * egress otherwise). This is the only flag supported for now.
1997 * Return
1998 * **SK_PASS** on success, or **SK_DROP** on error.
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01001999 *
2000 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2001 * Description
2002 * Encapsulate the packet associated to *skb* within a Layer 3
2003 * protocol header. This header is provided in the buffer at
2004 * address *hdr*, with *len* its size in bytes. *type* indicates
2005 * the protocol of the header and can be one of:
2006 *
2007 * **BPF_LWT_ENCAP_SEG6**
2008 * IPv6 encapsulation with Segment Routing Header
2009 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2010 * the IPv6 header is computed by the kernel.
2011 * **BPF_LWT_ENCAP_SEG6_INLINE**
2012 * Only works if *skb* contains an IPv6 packet. Insert a
2013 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2014 * the IPv6 header.
2015 *
2016 * A call to this helper is susceptible to change the underlaying
2017 * packet buffer. Therefore, at load time, all checks on pointers
2018 * previously done by the verifier are invalidated and must be
2019 * performed again, if the helper is used in combination with
2020 * direct packet access.
2021 * Return
2022 * 0 on success, or a negative error in case of failure.
2023 *
2024 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2025 * Description
2026 * Store *len* bytes from address *from* into the packet
2027 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2028 * inside the outermost IPv6 Segment Routing Header can be
2029 * modified through this helper.
2030 *
2031 * A call to this helper is susceptible to change the underlaying
2032 * packet buffer. Therefore, at load time, all checks on pointers
2033 * previously done by the verifier are invalidated and must be
2034 * performed again, if the helper is used in combination with
2035 * direct packet access.
2036 * Return
2037 * 0 on success, or a negative error in case of failure.
2038 *
2039 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2040 * Description
2041 * Adjust the size allocated to TLVs in the outermost IPv6
2042 * Segment Routing Header contained in the packet associated to
2043 * *skb*, at position *offset* by *delta* bytes. Only offsets
2044 * after the segments are accepted. *delta* can be as well
2045 * positive (growing) as negative (shrinking).
2046 *
2047 * A call to this helper is susceptible to change the underlaying
2048 * packet buffer. Therefore, at load time, all checks on pointers
2049 * previously done by the verifier are invalidated and must be
2050 * performed again, if the helper is used in combination with
2051 * direct packet access.
2052 * Return
2053 * 0 on success, or a negative error in case of failure.
2054 *
2055 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2056 * Description
2057 * Apply an IPv6 Segment Routing action of type *action* to the
2058 * packet associated to *skb*. Each action takes a parameter
2059 * contained at address *param*, and of length *param_len* bytes.
2060 * *action* can be one of:
2061 *
2062 * **SEG6_LOCAL_ACTION_END_X**
2063 * End.X action: Endpoint with Layer-3 cross-connect.
2064 * Type of *param*: **struct in6_addr**.
2065 * **SEG6_LOCAL_ACTION_END_T**
2066 * End.T action: Endpoint with specific IPv6 table lookup.
2067 * Type of *param*: **int**.
2068 * **SEG6_LOCAL_ACTION_END_B6**
2069 * End.B6 action: Endpoint bound to an SRv6 policy.
2070 * Type of param: **struct ipv6_sr_hdr**.
2071 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2072 * End.B6.Encap action: Endpoint bound to an SRv6
2073 * encapsulation policy.
2074 * Type of param: **struct ipv6_sr_hdr**.
2075 *
2076 * A call to this helper is susceptible to change the underlaying
2077 * packet buffer. Therefore, at load time, all checks on pointers
2078 * previously done by the verifier are invalidated and must be
2079 * performed again, if the helper is used in combination with
2080 * direct packet access.
2081 * Return
2082 * 0 on success, or a negative error in case of failure.
Sean Youngf4364dc2018-05-27 12:24:09 +01002083 *
2084 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2085 * Description
2086 * This helper is used in programs implementing IR decoding, to
2087 * report a successfully decoded key press with *scancode*,
2088 * *toggle* value in the given *protocol*. The scancode will be
2089 * translated to a keycode using the rc keymap, and reported as
2090 * an input key down event. After a period a key up event is
2091 * generated. This period can be extended by calling either
2092 * **bpf_rc_keydown** () again with the same values, or calling
2093 * **bpf_rc_repeat** ().
2094 *
2095 * Some protocols include a toggle bit, in case the button was
2096 * released and pressed again between consecutive scancodes.
2097 *
2098 * The *ctx* should point to the lirc sample as passed into
2099 * the program.
2100 *
2101 * The *protocol* is the decoded protocol number (see
2102 * **enum rc_proto** for some predefined values).
2103 *
2104 * This helper is only available is the kernel was compiled with
2105 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2106 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002107 * Return
2108 * 0
2109 *
2110 * int bpf_rc_repeat(void *ctx)
2111 * Description
2112 * This helper is used in programs implementing IR decoding, to
2113 * report a successfully decoded repeat key message. This delays
2114 * the generation of a key up event for previously generated
2115 * key down event.
2116 *
2117 * Some IR protocols like NEC have a special IR message for
2118 * repeating last button, for when a button is held down.
2119 *
2120 * The *ctx* should point to the lirc sample as passed into
2121 * the program.
2122 *
2123 * This helper is only available is the kernel was compiled with
2124 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2125 * "**y**".
Sean Youngf4364dc2018-05-27 12:24:09 +01002126 * Return
2127 * 0
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002128 *
2129 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2130 * Description
2131 * Return the cgroup v2 id of the socket associated with the *skb*.
2132 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2133 * helper for cgroup v1 by providing a tag resp. identifier that
2134 * can be matched on or used for map lookups e.g. to implement
2135 * policy. The cgroup v2 id of a given path in the hierarchy is
2136 * exposed in user space through the f_handle API in order to get
2137 * to the same 64-bit id.
2138 *
2139 * This helper can be used on TC egress path, but not on ingress,
2140 * and is available only if the kernel was compiled with the
2141 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2142 * Return
2143 * The id is returned or 0 in case the id could not be retrieved.
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002144 *
Andrey Ignatov77236282018-08-12 10:49:27 -07002145 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2146 * Description
2147 * Return id of cgroup v2 that is ancestor of cgroup associated
2148 * with the *skb* at the *ancestor_level*. The root cgroup is at
2149 * *ancestor_level* zero and each step down the hierarchy
2150 * increments the level. If *ancestor_level* == level of cgroup
2151 * associated with *skb*, then return value will be same as that
2152 * of **bpf_skb_cgroup_id**\ ().
2153 *
2154 * The helper is useful to implement policies based on cgroups
2155 * that are upper in hierarchy than immediate cgroup associated
2156 * with *skb*.
2157 *
2158 * The format of returned id and helper limitations are same as in
2159 * **bpf_skb_cgroup_id**\ ().
2160 * Return
2161 * The id is returned or 0 in case the id could not be retrieved.
2162 *
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002163 * u64 bpf_get_current_cgroup_id(void)
2164 * Return
2165 * A 64-bit integer containing the current cgroup id based
2166 * on the cgroup within which the current task is running.
Roman Gushchincd339432018-08-02 14:27:24 -07002167 *
2168 * void* get_local_storage(void *map, u64 flags)
2169 * Description
2170 * Get the pointer to the local storage area.
2171 * The type and the size of the local storage is defined
2172 * by the *map* argument.
2173 * The *flags* meaning is specific for each map type,
2174 * and has to be 0 for cgroup local storage.
2175 *
2176 * Depending on the bpf program type, a local storage area
2177 * can be shared between multiple instances of the bpf program,
2178 * running simultaneously.
2179 *
2180 * A user should care about the synchronization by himself.
2181 * For example, by using the BPF_STX_XADD instruction to alter
2182 * the shared data.
2183 * Return
2184 * Pointer to the local storage area.
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002185 *
2186 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2187 * Description
2188 * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
2189 * It checks the selected sk is matching the incoming
2190 * request in the skb.
2191 * Return
2192 * 0 on success, or a negative error in case of failure.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002193 *
2194 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2195 * Description
2196 * Look for TCP socket matching *tuple*, optionally in a child
2197 * network namespace *netns*. The return value must be checked,
2198 * and if non-NULL, released via **bpf_sk_release**\ ().
2199 *
2200 * The *ctx* should point to the context of the program, such as
2201 * the skb or socket (depending on the hook in use). This is used
2202 * to determine the base network namespace for the lookup.
2203 *
2204 * *tuple_size* must be one of:
2205 *
2206 * **sizeof**\ (*tuple*\ **->ipv4**)
2207 * Look for an IPv4 socket.
2208 * **sizeof**\ (*tuple*\ **->ipv6**)
2209 * Look for an IPv6 socket.
2210 *
2211 * If the *netns* is zero, then the socket lookup table in the
2212 * netns associated with the *ctx* will be used. For the TC hooks,
2213 * this in the netns of the device in the skb. For socket hooks,
2214 * this in the netns of the socket. If *netns* is non-zero, then
2215 * it specifies the ID of the netns relative to the netns
2216 * associated with the *ctx*.
2217 *
2218 * All values for *flags* are reserved for future usage, and must
2219 * be left at zero.
2220 *
2221 * This helper is available only if the kernel was compiled with
2222 * **CONFIG_NET** configuration option.
2223 * Return
2224 * Pointer to *struct bpf_sock*, or NULL in case of failure.
Nitin Handec8123ea2018-10-28 21:02:45 -07002225 * For sockets with reuseport option, *struct bpf_sock*
2226 * return is from reuse->socks[] using hash of the packet.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002227 *
2228 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2229 * Description
2230 * Look for UDP socket matching *tuple*, optionally in a child
2231 * network namespace *netns*. The return value must be checked,
2232 * and if non-NULL, released via **bpf_sk_release**\ ().
2233 *
2234 * The *ctx* should point to the context of the program, such as
2235 * the skb or socket (depending on the hook in use). This is used
2236 * to determine the base network namespace for the lookup.
2237 *
2238 * *tuple_size* must be one of:
2239 *
2240 * **sizeof**\ (*tuple*\ **->ipv4**)
2241 * Look for an IPv4 socket.
2242 * **sizeof**\ (*tuple*\ **->ipv6**)
2243 * Look for an IPv6 socket.
2244 *
2245 * If the *netns* is zero, then the socket lookup table in the
2246 * netns associated with the *ctx* will be used. For the TC hooks,
2247 * this in the netns of the device in the skb. For socket hooks,
2248 * this in the netns of the socket. If *netns* is non-zero, then
2249 * it specifies the ID of the netns relative to the netns
2250 * associated with the *ctx*.
2251 *
2252 * All values for *flags* are reserved for future usage, and must
2253 * be left at zero.
2254 *
2255 * This helper is available only if the kernel was compiled with
2256 * **CONFIG_NET** configuration option.
2257 * Return
2258 * Pointer to *struct bpf_sock*, or NULL in case of failure.
Nitin Handec8123ea2018-10-28 21:02:45 -07002259 * For sockets with reuseport option, *struct bpf_sock*
2260 * return is from reuse->socks[] using hash of the packet.
Joe Stringer6acc9b42018-10-02 13:35:36 -07002261 *
2262 * int bpf_sk_release(struct bpf_sock *sk)
2263 * Description
2264 * Release the reference held by *sock*. *sock* must be a non-NULL
2265 * pointer that was returned from bpf_sk_lookup_xxx\ ().
2266 * Return
2267 * 0 on success, or a negative error in case of failure.
John Fastabend6fff6072018-10-19 19:56:49 -07002268 *
2269 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2270 * Description
2271 * For socket policies, insert *len* bytes into msg at offset
2272 * *start*.
2273 *
2274 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2275 * *msg* it may want to insert metadata or options into the msg.
2276 * This can later be read and used by any of the lower layer BPF
2277 * hooks.
2278 *
2279 * This helper may fail if under memory pressure (a malloc
2280 * fails) in these cases BPF programs will get an appropriate
2281 * error and BPF programs will need to handle them.
2282 *
2283 * Return
2284 * 0 on success, or a negative error in case of failure.
John Fastabend7246d8e2018-11-26 14:16:17 -08002285 *
2286 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2287 * Description
2288 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2289 * This may result in **ENOMEM** errors under certain situations if
2290 * an allocation and copy are required due to a full ring buffer.
2291 * However, the helper will try to avoid doing the allocation
2292 * if possible. Other errors can occur if input parameters are
2293 * invalid either due to *start* byte not being valid part of msg
2294 * payload and/or *pop* value being to large.
2295 *
2296 * Return
2297 * 0 on success, or a negative erro in case of failure.
Thomas Grafebb676d2016-10-27 11:23:51 +02002298 */
2299#define __BPF_FUNC_MAPPER(FN) \
2300 FN(unspec), \
2301 FN(map_lookup_elem), \
2302 FN(map_update_elem), \
2303 FN(map_delete_elem), \
2304 FN(probe_read), \
2305 FN(ktime_get_ns), \
2306 FN(trace_printk), \
2307 FN(get_prandom_u32), \
2308 FN(get_smp_processor_id), \
2309 FN(skb_store_bytes), \
2310 FN(l3_csum_replace), \
2311 FN(l4_csum_replace), \
2312 FN(tail_call), \
2313 FN(clone_redirect), \
2314 FN(get_current_pid_tgid), \
2315 FN(get_current_uid_gid), \
2316 FN(get_current_comm), \
2317 FN(get_cgroup_classid), \
2318 FN(skb_vlan_push), \
2319 FN(skb_vlan_pop), \
2320 FN(skb_get_tunnel_key), \
2321 FN(skb_set_tunnel_key), \
2322 FN(perf_event_read), \
2323 FN(redirect), \
2324 FN(get_route_realm), \
2325 FN(perf_event_output), \
2326 FN(skb_load_bytes), \
2327 FN(get_stackid), \
2328 FN(csum_diff), \
2329 FN(skb_get_tunnel_opt), \
2330 FN(skb_set_tunnel_opt), \
2331 FN(skb_change_proto), \
2332 FN(skb_change_type), \
2333 FN(skb_under_cgroup), \
2334 FN(get_hash_recalc), \
2335 FN(get_current_task), \
2336 FN(probe_write_user), \
2337 FN(current_task_under_cgroup), \
2338 FN(skb_change_tail), \
2339 FN(skb_pull_data), \
2340 FN(csum_update), \
2341 FN(set_hash_invalid), \
Thomas Graf3a0af8f2016-11-30 17:10:10 +01002342 FN(get_numa_node_id), \
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08002343 FN(skb_change_head), \
Gianluca Borelloa5e8c072017-01-18 17:55:49 +00002344 FN(xdp_adjust_head), \
Chenbo Feng91b82702017-03-22 17:27:34 -07002345 FN(probe_read_str), \
Chenbo Feng6acc5c22017-03-22 17:27:35 -07002346 FN(get_socket_cookie), \
Daniel Borkmannded092c2017-06-11 00:50:47 +02002347 FN(get_socket_uid), \
Lawrence Brakmo8c4b4c72017-06-30 20:02:46 -07002348 FN(set_hash), \
Daniel Borkmann2be7e212017-07-02 02:13:26 +02002349 FN(setsockopt), \
John Fastabend97f91a72017-07-17 09:29:18 -07002350 FN(skb_adjust_room), \
John Fastabend174a79f2017-08-15 22:32:47 -07002351 FN(redirect_map), \
2352 FN(sk_redirect_map), \
2353 FN(sock_map_update), \
Yonghong Song908432c2017-10-05 09:19:20 -07002354 FN(xdp_adjust_meta), \
Yonghong Song4bebdc72017-10-05 09:19:22 -07002355 FN(perf_event_read_value), \
Lawrence Brakmocd86d1f2017-10-20 11:05:40 -07002356 FN(perf_prog_read_value), \
Josef Bacik9802d862017-12-11 11:36:48 -05002357 FN(getsockopt), \
Lawrence Brakmob13d8802018-01-25 16:14:10 -08002358 FN(override_return), \
John Fastabend4f738ad2018-03-18 12:57:10 -07002359 FN(sock_ops_cb_flags_set), \
John Fastabend2a100312018-03-18 12:57:15 -07002360 FN(msg_redirect_map), \
John Fastabend91843d52018-03-18 12:57:20 -07002361 FN(msg_apply_bytes), \
John Fastabend015632b2018-03-18 12:57:25 -07002362 FN(msg_cork_bytes), \
Andrey Ignatovd74bad42018-03-30 15:08:05 -07002363 FN(msg_pull_data), \
Nikita V. Shirokovb32cc5b2018-04-17 21:42:13 -07002364 FN(bind), \
Eyal Birger12bed762018-04-24 17:50:29 +03002365 FN(xdp_adjust_tail), \
Yonghong Songc195651e2018-04-28 22:28:08 -07002366 FN(skb_get_xfrm_state), \
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002367 FN(get_stack), \
David Ahern87f5fc72018-05-09 20:34:26 -07002368 FN(skb_load_bytes_relative), \
John Fastabend81110382018-05-14 10:00:17 -07002369 FN(fib_lookup), \
2370 FN(sock_hash_update), \
2371 FN(msg_redirect_hash), \
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002372 FN(sk_redirect_hash), \
2373 FN(lwt_push_encap), \
2374 FN(lwt_seg6_store_bytes), \
2375 FN(lwt_seg6_adjust_srh), \
Sean Youngf4364dc2018-05-27 12:24:09 +01002376 FN(lwt_seg6_action), \
2377 FN(rc_repeat), \
Daniel Borkmanncb20b082018-06-02 23:06:36 +02002378 FN(rc_keydown), \
Yonghong Songbf6fa2c2018-06-03 15:59:41 -07002379 FN(skb_cgroup_id), \
Roman Gushchincd339432018-08-02 14:27:24 -07002380 FN(get_current_cgroup_id), \
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002381 FN(get_local_storage), \
Andrey Ignatov77236282018-08-12 10:49:27 -07002382 FN(sk_select_reuseport), \
Joe Stringer6acc9b42018-10-02 13:35:36 -07002383 FN(skb_ancestor_cgroup_id), \
2384 FN(sk_lookup_tcp), \
2385 FN(sk_lookup_udp), \
Mauricio Vasquez Bf1a2e442018-10-18 15:16:25 +02002386 FN(sk_release), \
2387 FN(map_push_elem), \
2388 FN(map_pop_elem), \
John Fastabend6fff6072018-10-19 19:56:49 -07002389 FN(map_peek_elem), \
John Fastabend7246d8e2018-11-26 14:16:17 -08002390 FN(msg_push_data), \
2391 FN(msg_pop_data),
Thomas Grafebb676d2016-10-27 11:23:51 +02002392
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002393/* integer value in 'imm' field of BPF_CALL instruction selects which helper
2394 * function eBPF program intends to call
2395 */
Thomas Grafebb676d2016-10-27 11:23:51 +02002396#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002397enum bpf_func_id {
Thomas Grafebb676d2016-10-27 11:23:51 +02002398 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002399 __BPF_FUNC_MAX_ID,
2400};
Thomas Grafebb676d2016-10-27 11:23:51 +02002401#undef __BPF_ENUM_FN
Alexei Starovoitov09756af2014-09-26 00:17:00 -07002402
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002403/* All flags used by eBPF helper functions, placed here. */
2404
2405/* BPF_FUNC_skb_store_bytes flags. */
2406#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
Daniel Borkmann8afd54c2016-03-04 15:15:03 +01002407#define BPF_F_INVALIDATE_HASH (1ULL << 1)
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002408
2409/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2410 * First 4 bits are for passing the header field size.
2411 */
2412#define BPF_F_HDR_FIELD_MASK 0xfULL
2413
2414/* BPF_FUNC_l4_csum_replace flags. */
2415#define BPF_F_PSEUDO_HDR (1ULL << 4)
Daniel Borkmann2f729592016-02-19 23:05:26 +01002416#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
Daniel Borkmannd1b662a2017-01-24 01:06:28 +01002417#define BPF_F_MARK_ENFORCE (1ULL << 6)
Daniel Borkmann781c53b2016-01-11 01:16:38 +01002418
2419/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2420#define BPF_F_INGRESS (1ULL << 0)
2421
Daniel Borkmannc6c33452016-01-11 01:16:39 +01002422/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2423#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2424
Yonghong Songc195651e2018-04-28 22:28:08 -07002425/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002426#define BPF_F_SKIP_FIELD_MASK 0xffULL
2427#define BPF_F_USER_STACK (1ULL << 8)
Yonghong Songc195651e2018-04-28 22:28:08 -07002428/* flags used by BPF_FUNC_get_stackid only. */
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002429#define BPF_F_FAST_STACK_CMP (1ULL << 9)
2430#define BPF_F_REUSE_STACKID (1ULL << 10)
Yonghong Songc195651e2018-04-28 22:28:08 -07002431/* flags used by BPF_FUNC_get_stack only. */
2432#define BPF_F_USER_BUILD_ID (1ULL << 11)
Alexei Starovoitovd5a3b1f2016-02-17 19:58:58 -08002433
Daniel Borkmann2da897e2016-02-23 02:05:26 +01002434/* BPF_FUNC_skb_set_tunnel_key flags. */
2435#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
Daniel Borkmann22080872016-03-04 15:15:05 +01002436#define BPF_F_DONT_FRAGMENT (1ULL << 2)
William Tu77a51962018-03-01 13:49:57 -08002437#define BPF_F_SEQ_NUMBER (1ULL << 3)
Daniel Borkmann2da897e2016-02-23 02:05:26 +01002438
Yonghong Song908432c2017-10-05 09:19:20 -07002439/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2440 * BPF_FUNC_perf_event_read_value flags.
2441 */
Daniel Borkmann1e337592016-04-18 21:01:23 +02002442#define BPF_F_INDEX_MASK 0xffffffffULL
2443#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
Daniel Borkmann555c8a82016-07-14 18:08:05 +02002444/* BPF_FUNC_perf_event_output for sk_buff input context. */
2445#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
Daniel Borkmann1e337592016-04-18 21:01:23 +02002446
Daniel Borkmann2be7e212017-07-02 02:13:26 +02002447/* Mode for BPF_FUNC_skb_adjust_room helper. */
2448enum bpf_adj_room_mode {
2449 BPF_ADJ_ROOM_NET,
2450};
2451
Daniel Borkmann4e1ec562018-05-04 01:08:15 +02002452/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2453enum bpf_hdr_start_off {
2454 BPF_HDR_START_MAC,
2455 BPF_HDR_START_NET,
2456};
2457
Mathieu Xhonneuxfe94cc22018-05-20 14:58:14 +01002458/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2459enum bpf_lwt_encap_mode {
2460 BPF_LWT_ENCAP_SEG6,
2461 BPF_LWT_ENCAP_SEG6_INLINE
2462};
2463
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07002464/* user accessible mirror of in-kernel sk_buff.
2465 * new fields can only be added to the end of this structure
2466 */
2467struct __sk_buff {
2468 __u32 len;
2469 __u32 pkt_type;
2470 __u32 mark;
2471 __u32 queue_mapping;
Alexei Starovoitovc2497392015-03-16 18:06:02 -07002472 __u32 protocol;
2473 __u32 vlan_present;
2474 __u32 vlan_tci;
Michal Sekletar27cd5452015-03-24 14:48:41 +01002475 __u32 vlan_proto;
Daniel Borkmannbcad5712015-04-03 20:52:24 +02002476 __u32 priority;
Alexei Starovoitov37e82c22015-05-27 15:30:39 -07002477 __u32 ingress_ifindex;
2478 __u32 ifindex;
Alexei Starovoitovd691f9e2015-06-04 10:11:54 -07002479 __u32 tc_index;
2480 __u32 cb[5];
Daniel Borkmannba7591d2015-08-01 00:46:29 +02002481 __u32 hash;
Daniel Borkmann045efa82015-09-15 23:05:42 -07002482 __u32 tc_classid;
Alexei Starovoitov969bf052016-05-05 19:49:10 -07002483 __u32 data;
2484 __u32 data_end;
Daniel Borkmannb1d9fc42017-04-19 23:01:17 +02002485 __u32 napi_id;
John Fastabend8a31db52017-08-15 22:33:09 -07002486
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02002487 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
John Fastabend8a31db52017-08-15 22:33:09 -07002488 __u32 family;
2489 __u32 remote_ip4; /* Stored in network byte order */
2490 __u32 local_ip4; /* Stored in network byte order */
2491 __u32 remote_ip6[4]; /* Stored in network byte order */
2492 __u32 local_ip6[4]; /* Stored in network byte order */
2493 __u32 remote_port; /* Stored in network byte order */
2494 __u32 local_port; /* stored in host byte order */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02002495 /* ... here. */
2496
2497 __u32 data_meta;
Petar Penkovd58e4682018-09-14 07:46:18 -07002498 struct bpf_flow_keys *flow_keys;
Vlad Dumitrescuf11216b2018-11-22 14:39:16 -05002499 __u64 tstamp;
Petar Penkove3da08d2018-12-02 20:18:19 -05002500 __u32 wire_len;
Alexei Starovoitov9bac3d62015-03-13 11:57:42 -07002501};
2502
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07002503struct bpf_tunnel_key {
2504 __u32 tunnel_id;
Daniel Borkmannc6c33452016-01-11 01:16:39 +01002505 union {
2506 __u32 remote_ipv4;
2507 __u32 remote_ipv6[4];
2508 };
2509 __u8 tunnel_tos;
2510 __u8 tunnel_ttl;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02002511 __u16 tunnel_ext; /* Padding, future use. */
Daniel Borkmann4018ab12016-03-09 03:00:05 +01002512 __u32 tunnel_label;
Alexei Starovoitovd3aa45c2015-07-30 15:36:57 -07002513};
2514
Eyal Birger12bed762018-04-24 17:50:29 +03002515/* user accessible mirror of in-kernel xfrm_state.
2516 * new fields can only be added to the end of this structure
2517 */
2518struct bpf_xfrm_state {
2519 __u32 reqid;
2520 __u32 spi; /* Stored in network byte order */
2521 __u16 family;
Daniel Borkmann1fbc2e02018-06-02 23:06:37 +02002522 __u16 ext; /* Padding, future use. */
Eyal Birger12bed762018-04-24 17:50:29 +03002523 union {
2524 __u32 remote_ipv4; /* Stored in network byte order */
2525 __u32 remote_ipv6[4]; /* Stored in network byte order */
2526 };
2527};
2528
Thomas Graf3a0af8f2016-11-30 17:10:10 +01002529/* Generic BPF return codes which all BPF program types may support.
2530 * The values are binary compatible with their TC_ACT_* counter-part to
2531 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2532 * programs.
2533 *
2534 * XDP is handled seprately, see XDP_*.
2535 */
2536enum bpf_ret_code {
2537 BPF_OK = 0,
2538 /* 1 reserved */
2539 BPF_DROP = 2,
2540 /* 3-6 reserved */
2541 BPF_REDIRECT = 7,
2542 /* >127 are reserved for prog type specific return codes */
2543};
2544
David Ahern610236582016-12-01 08:48:04 -08002545struct bpf_sock {
2546 __u32 bound_dev_if;
David Ahernaa4c1032016-12-01 08:48:06 -08002547 __u32 family;
2548 __u32 type;
2549 __u32 protocol;
David Ahern482dca92017-08-31 15:05:44 -07002550 __u32 mark;
2551 __u32 priority;
Andrey Ignatovaac3fc32018-03-30 15:08:07 -07002552 __u32 src_ip4; /* Allows 1,2,4-byte read.
2553 * Stored in network byte order.
2554 */
2555 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2556 * Stored in network byte order.
2557 */
2558 __u32 src_port; /* Allows 4-byte read.
2559 * Stored in host byte order
2560 */
David Ahern610236582016-12-01 08:48:04 -08002561};
2562
Joe Stringer6acc9b42018-10-02 13:35:36 -07002563struct bpf_sock_tuple {
2564 union {
2565 struct {
2566 __be32 saddr;
2567 __be32 daddr;
2568 __be16 sport;
2569 __be16 dport;
2570 } ipv4;
2571 struct {
2572 __be32 saddr[4];
2573 __be32 daddr[4];
2574 __be16 sport;
2575 __be16 dport;
2576 } ipv6;
2577 };
2578};
2579
Martin KaFai Lau17bedab2016-12-07 15:53:11 -08002580#define XDP_PACKET_HEADROOM 256
2581
Brenden Blanco6a773a12016-07-19 12:16:47 -07002582/* User return codes for XDP prog type.
2583 * A valid XDP program must return one of these defined values. All other
Daniel Borkmann9beb8be2017-09-09 01:40:35 +02002584 * return codes are reserved for future use. Unknown return codes will
2585 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
Brenden Blanco6a773a12016-07-19 12:16:47 -07002586 */
2587enum xdp_action {
2588 XDP_ABORTED = 0,
2589 XDP_DROP,
2590 XDP_PASS,
Brenden Blanco6ce96ca2016-07-19 12:16:53 -07002591 XDP_TX,
John Fastabend814abfa2017-07-17 09:27:07 -07002592 XDP_REDIRECT,
Brenden Blanco6a773a12016-07-19 12:16:47 -07002593};
2594
2595/* user accessible metadata for XDP packet hook
2596 * new fields must be added to the end of this structure
2597 */
2598struct xdp_md {
2599 __u32 data;
2600 __u32 data_end;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02002601 __u32 data_meta;
Jesper Dangaard Brouerdaaf24c2018-01-11 17:39:09 +01002602 /* Below access go through struct xdp_rxq_info */
Jesper Dangaard Brouer02dd3292018-01-03 11:26:14 +01002603 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2604 __u32 rx_queue_index; /* rxq->queue_index */
Brenden Blanco6a773a12016-07-19 12:16:47 -07002605};
2606
John Fastabend174a79f2017-08-15 22:32:47 -07002607enum sk_action {
John Fastabendbfa640752017-10-27 09:45:53 -07002608 SK_DROP = 0,
2609 SK_PASS,
John Fastabend174a79f2017-08-15 22:32:47 -07002610};
2611
John Fastabend4f738ad2018-03-18 12:57:10 -07002612/* user accessible metadata for SK_MSG packet hook, new fields must
2613 * be added to the end of this structure
2614 */
2615struct sk_msg_md {
2616 void *data;
2617 void *data_end;
John Fastabend303def32018-05-17 14:16:58 -07002618
2619 __u32 family;
2620 __u32 remote_ip4; /* Stored in network byte order */
2621 __u32 local_ip4; /* Stored in network byte order */
2622 __u32 remote_ip6[4]; /* Stored in network byte order */
2623 __u32 local_ip6[4]; /* Stored in network byte order */
2624 __u32 remote_port; /* Stored in network byte order */
2625 __u32 local_port; /* stored in host byte order */
John Fastabend4f738ad2018-03-18 12:57:10 -07002626};
2627
Martin KaFai Lau2dbb9b92018-08-08 01:01:25 -07002628struct sk_reuseport_md {
2629 /*
2630 * Start of directly accessible data. It begins from
2631 * the tcp/udp header.
2632 */
2633 void *data;
2634 void *data_end; /* End of directly accessible data */
2635 /*
2636 * Total length of packet (starting from the tcp/udp header).
2637 * Note that the directly accessible bytes (data_end - data)
2638 * could be less than this "len". Those bytes could be
2639 * indirectly read by a helper "bpf_skb_load_bytes()".
2640 */
2641 __u32 len;
2642 /*
2643 * Eth protocol in the mac header (network byte order). e.g.
2644 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2645 */
2646 __u32 eth_protocol;
2647 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2648 __u32 bind_inany; /* Is sock bound to an INANY address? */
2649 __u32 hash; /* A hash of the packet 4 tuples */
2650};
2651
Martin KaFai Lau1e270972017-06-05 12:15:52 -07002652#define BPF_TAG_SIZE 8
2653
2654struct bpf_prog_info {
2655 __u32 type;
2656 __u32 id;
2657 __u8 tag[BPF_TAG_SIZE];
2658 __u32 jited_prog_len;
2659 __u32 xlated_prog_len;
2660 __aligned_u64 jited_prog_insns;
2661 __aligned_u64 xlated_prog_insns;
Martin KaFai Laucb4d2b32017-09-27 14:37:52 -07002662 __u64 load_time; /* ns since boottime */
2663 __u32 created_by_uid;
2664 __u32 nr_map_ids;
2665 __aligned_u64 map_ids;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07002666 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski675fc272017-12-27 18:39:09 -08002667 __u32 ifindex;
Jiri Olsab85fab02018-04-25 19:41:06 +02002668 __u32 gpl_compatible:1;
Jakub Kicinski675fc272017-12-27 18:39:09 -08002669 __u64 netns_dev;
2670 __u64 netns_ino;
Sandipan Dasdbecd732018-05-24 12:26:48 +05302671 __u32 nr_jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05302672 __u32 nr_jited_func_lens;
Sandipan Dasdbecd732018-05-24 12:26:48 +05302673 __aligned_u64 jited_ksyms;
Sandipan Das815581c2018-05-24 12:26:52 +05302674 __aligned_u64 jited_func_lens;
Yonghong Song838e9692018-11-19 15:29:11 -08002675 __u32 btf_id;
2676 __u32 func_info_rec_size;
2677 __aligned_u64 func_info;
2678 __u32 func_info_cnt;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07002679} __attribute__((aligned(8)));
2680
2681struct bpf_map_info {
2682 __u32 type;
2683 __u32 id;
2684 __u32 key_size;
2685 __u32 value_size;
2686 __u32 max_entries;
2687 __u32 map_flags;
Martin KaFai Lau067cae42017-10-05 21:52:12 -07002688 char name[BPF_OBJ_NAME_LEN];
Jakub Kicinski52775b32018-01-17 19:13:28 -08002689 __u32 ifindex;
Daniel Borkmann36f98142018-06-02 05:21:59 +02002690 __u32 :32;
Jakub Kicinski52775b32018-01-17 19:13:28 -08002691 __u64 netns_dev;
2692 __u64 netns_ino;
Martin KaFai Lau78958fc2018-05-04 14:49:51 -07002693 __u32 btf_id;
Martin KaFai Lau9b2cf322018-05-22 14:57:21 -07002694 __u32 btf_key_type_id;
2695 __u32 btf_value_type_id;
Martin KaFai Lau1e270972017-06-05 12:15:52 -07002696} __attribute__((aligned(8)));
2697
Martin KaFai Lau62dab842018-05-04 14:49:52 -07002698struct bpf_btf_info {
2699 __aligned_u64 btf;
2700 __u32 btf_size;
2701 __u32 id;
2702} __attribute__((aligned(8)));
2703
Andrey Ignatov4fbac772018-03-30 15:08:02 -07002704/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2705 * by user and intended to be used by socket (e.g. to bind to, depends on
2706 * attach attach type).
2707 */
2708struct bpf_sock_addr {
2709 __u32 user_family; /* Allows 4-byte read, but no write. */
2710 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2711 * Stored in network byte order.
2712 */
2713 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2714 * Stored in network byte order.
2715 */
2716 __u32 user_port; /* Allows 4-byte read and write.
2717 * Stored in network byte order
2718 */
2719 __u32 family; /* Allows 4-byte read, but no write */
2720 __u32 type; /* Allows 4-byte read, but no write */
2721 __u32 protocol; /* Allows 4-byte read, but no write */
Andrey Ignatov1cedee12018-05-25 08:55:23 -07002722 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2723 * Stored in network byte order.
2724 */
2725 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2726 * Stored in network byte order.
2727 */
Andrey Ignatov4fbac772018-03-30 15:08:02 -07002728};
2729
Lawrence Brakmo40304b22017-06-30 20:02:40 -07002730/* User bpf_sock_ops struct to access socket values and specify request ops
2731 * and their replies.
2732 * Some of this fields are in network (bigendian) byte order and may need
2733 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2734 * New fields can only be added at the end of this structure
2735 */
2736struct bpf_sock_ops {
2737 __u32 op;
2738 union {
Lawrence Brakmode525be2018-01-25 16:14:09 -08002739 __u32 args[4]; /* Optionally passed to bpf program */
2740 __u32 reply; /* Returned by bpf program */
2741 __u32 replylong[4]; /* Optionally returned by bpf prog */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07002742 };
2743 __u32 family;
2744 __u32 remote_ip4; /* Stored in network byte order */
2745 __u32 local_ip4; /* Stored in network byte order */
2746 __u32 remote_ip6[4]; /* Stored in network byte order */
2747 __u32 local_ip6[4]; /* Stored in network byte order */
2748 __u32 remote_port; /* Stored in network byte order */
2749 __u32 local_port; /* stored in host byte order */
Lawrence Brakmof19397a2017-12-01 10:15:04 -08002750 __u32 is_fullsock; /* Some TCP fields are only valid if
2751 * there is a full socket. If not, the
2752 * fields read as zero.
2753 */
2754 __u32 snd_cwnd;
2755 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
Lawrence Brakmob13d8802018-01-25 16:14:10 -08002756 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
Lawrence Brakmo44f0e432018-01-25 16:14:12 -08002757 __u32 state;
2758 __u32 rtt_min;
2759 __u32 snd_ssthresh;
2760 __u32 rcv_nxt;
2761 __u32 snd_nxt;
2762 __u32 snd_una;
2763 __u32 mss_cache;
2764 __u32 ecn_flags;
2765 __u32 rate_delivered;
2766 __u32 rate_interval_us;
2767 __u32 packets_out;
2768 __u32 retrans_out;
2769 __u32 total_retrans;
2770 __u32 segs_in;
2771 __u32 data_segs_in;
2772 __u32 segs_out;
2773 __u32 data_segs_out;
2774 __u32 lost_out;
2775 __u32 sacked_out;
2776 __u32 sk_txhash;
2777 __u64 bytes_received;
2778 __u64 bytes_acked;
Lawrence Brakmo40304b22017-06-30 20:02:40 -07002779};
2780
Lawrence Brakmob13d8802018-01-25 16:14:10 -08002781/* Definitions for bpf_sock_ops_cb_flags */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08002782#define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08002783#define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
Lawrence Brakmod4487492018-01-25 16:14:15 -08002784#define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2785#define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
Lawrence Brakmob13d8802018-01-25 16:14:10 -08002786 * supported cb flags
2787 */
2788
Lawrence Brakmo40304b22017-06-30 20:02:40 -07002789/* List of known BPF sock_ops operators.
2790 * New entries can only be added at the end
2791 */
2792enum {
2793 BPF_SOCK_OPS_VOID,
Lawrence Brakmo8550f322017-06-30 20:02:42 -07002794 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2795 * -1 if default value should be used
2796 */
Lawrence Brakmo13d3b1e2017-06-30 20:02:44 -07002797 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2798 * window (in packets) or -1 if default
2799 * value should be used
2800 */
Lawrence Brakmo9872a4b2017-06-30 20:02:47 -07002801 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2802 * active connection is initialized
2803 */
2804 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2805 * active connection is
2806 * established
2807 */
2808 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2809 * passive connection is
2810 * established
2811 */
Lawrence Brakmo91b5b212017-06-30 20:02:49 -07002812 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2813 * needs ECN
2814 */
Lawrence Brakmoe6546ef2017-10-20 11:05:39 -07002815 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2816 * based on the path and may be
2817 * dependent on the congestion control
2818 * algorithm. In general it indicates
2819 * a congestion threshold. RTTs above
2820 * this indicate congestion
2821 */
Lawrence Brakmof89013f2018-01-25 16:14:11 -08002822 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2823 * Arg1: value of icsk_retransmits
2824 * Arg2: value of icsk_rto
2825 * Arg3: whether RTO has expired
2826 */
Lawrence Brakmoa31ad292018-01-25 16:14:14 -08002827 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2828 * Arg1: sequence number of 1st byte
2829 * Arg2: # segments
2830 * Arg3: return value of
2831 * tcp_transmit_skb (0 => success)
2832 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08002833 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2834 * Arg1: old_state
2835 * Arg2: new_state
2836 */
Andrey Ignatovf333ee02018-07-11 17:33:32 -07002837 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2838 * socket transition to LISTEN state.
2839 */
Lawrence Brakmod4487492018-01-25 16:14:15 -08002840};
2841
2842/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2843 * changes between the TCP and BPF versions. Ideally this should never happen.
2844 * If it does, we need to add code to convert them before calling
2845 * the BPF sock_ops function.
2846 */
2847enum {
2848 BPF_TCP_ESTABLISHED = 1,
2849 BPF_TCP_SYN_SENT,
2850 BPF_TCP_SYN_RECV,
2851 BPF_TCP_FIN_WAIT1,
2852 BPF_TCP_FIN_WAIT2,
2853 BPF_TCP_TIME_WAIT,
2854 BPF_TCP_CLOSE,
2855 BPF_TCP_CLOSE_WAIT,
2856 BPF_TCP_LAST_ACK,
2857 BPF_TCP_LISTEN,
2858 BPF_TCP_CLOSING, /* Now a valid state */
2859 BPF_TCP_NEW_SYN_RECV,
2860
2861 BPF_TCP_MAX_STATES /* Leave at the end! */
Lawrence Brakmo40304b22017-06-30 20:02:40 -07002862};
2863
Lawrence Brakmofc747812017-06-30 20:02:51 -07002864#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
Lawrence Brakmo13bf9642017-06-30 20:02:53 -07002865#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
Lawrence Brakmofc747812017-06-30 20:02:51 -07002866
Yonghong Song908432c2017-10-05 09:19:20 -07002867struct bpf_perf_event_value {
2868 __u64 counter;
2869 __u64 enabled;
2870 __u64 running;
2871};
2872
Roman Gushchinebc614f2017-11-05 08:15:32 -05002873#define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2874#define BPF_DEVCG_ACC_READ (1ULL << 1)
2875#define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2876
2877#define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2878#define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2879
2880struct bpf_cgroup_dev_ctx {
Yonghong Song06ef0cc2017-12-18 10:13:44 -08002881 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2882 __u32 access_type;
Roman Gushchinebc614f2017-11-05 08:15:32 -05002883 __u32 major;
2884 __u32 minor;
2885};
2886
Alexei Starovoitovc4f66992018-03-28 12:05:37 -07002887struct bpf_raw_tracepoint_args {
2888 __u64 args[0];
2889};
2890
David Ahern87f5fc72018-05-09 20:34:26 -07002891/* DIRECT: Skip the FIB rules and go to FIB table associated with device
2892 * OUTPUT: Do lookup from egress perspective; default is ingress
2893 */
2894#define BPF_FIB_LOOKUP_DIRECT BIT(0)
2895#define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2896
David Ahern4c795792018-06-26 16:21:18 -07002897enum {
2898 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2899 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2900 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2901 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2902 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2903 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2904 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2905 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2906 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2907};
2908
David Ahern87f5fc72018-05-09 20:34:26 -07002909struct bpf_fib_lookup {
David Ahernfa898d72018-05-29 10:58:07 -07002910 /* input: network family for lookup (AF_INET, AF_INET6)
2911 * output: network family of egress nexthop
2912 */
2913 __u8 family;
David Ahern87f5fc72018-05-09 20:34:26 -07002914
2915 /* set if lookup is to consider L4 data - e.g., FIB rules */
2916 __u8 l4_protocol;
2917 __be16 sport;
2918 __be16 dport;
2919
2920 /* total length of packet from network header - used for MTU check */
2921 __u16 tot_len;
David Ahern4c795792018-06-26 16:21:18 -07002922
2923 /* input: L3 device index for lookup
2924 * output: device index from FIB lookup
2925 */
2926 __u32 ifindex;
David Ahern87f5fc72018-05-09 20:34:26 -07002927
2928 union {
2929 /* inputs to lookup */
2930 __u8 tos; /* AF_INET */
David Ahernbd3a08a2018-06-03 08:15:19 -07002931 __be32 flowinfo; /* AF_INET6, flow_label + priority */
David Ahern87f5fc72018-05-09 20:34:26 -07002932
David Ahernfa898d72018-05-29 10:58:07 -07002933 /* output: metric of fib result (IPv4/IPv6 only) */
2934 __u32 rt_metric;
David Ahern87f5fc72018-05-09 20:34:26 -07002935 };
2936
2937 union {
David Ahern87f5fc72018-05-09 20:34:26 -07002938 __be32 ipv4_src;
2939 __u32 ipv6_src[4]; /* in6_addr; network order */
2940 };
2941
David Ahernfa898d72018-05-29 10:58:07 -07002942 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2943 * network header. output: bpf_fib_lookup sets to gateway address
2944 * if FIB lookup returns gateway route
David Ahern87f5fc72018-05-09 20:34:26 -07002945 */
2946 union {
David Ahern87f5fc72018-05-09 20:34:26 -07002947 __be32 ipv4_dst;
2948 __u32 ipv6_dst[4]; /* in6_addr; network order */
2949 };
2950
2951 /* output */
2952 __be16 h_vlan_proto;
2953 __be16 h_vlan_TCI;
2954 __u8 smac[6]; /* ETH_ALEN */
2955 __u8 dmac[6]; /* ETH_ALEN */
2956};
2957
Yonghong Song41bdc4b2018-05-24 11:21:09 -07002958enum bpf_task_fd_type {
2959 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2960 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2961 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2962 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2963 BPF_FD_TYPE_UPROBE, /* filename + offset */
2964 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2965};
2966
Petar Penkovd58e4682018-09-14 07:46:18 -07002967struct bpf_flow_keys {
2968 __u16 nhoff;
2969 __u16 thoff;
2970 __u16 addr_proto; /* ETH_P_* of valid addrs */
2971 __u8 is_frag;
2972 __u8 is_first_frag;
2973 __u8 is_encap;
2974 __u8 ip_proto;
2975 __be16 n_proto;
2976 __be16 sport;
2977 __be16 dport;
2978 union {
2979 struct {
2980 __be32 ipv4_src;
2981 __be32 ipv4_dst;
2982 };
2983 struct {
2984 __u32 ipv6_src[4]; /* in6_addr; network order */
2985 __u32 ipv6_dst[4]; /* in6_addr; network order */
2986 };
2987 };
2988};
2989
Yonghong Song838e9692018-11-19 15:29:11 -08002990struct bpf_func_info {
2991 __u32 insn_offset;
2992 __u32 type_id;
2993};
2994
Alexei Starovoitovdaedfb22014-09-04 22:17:18 -07002995#endif /* _UAPI__LINUX_BPF_H__ */