]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/uapi/linux/bpf.h
bpf: Rename BPF_XADD and prepare to encode other atomics in .imm
[mirror_ubuntu-jammy-kernel.git] / include / uapi / linux / bpf.h
CommitLineData
e2be04c7 1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
daedfb22
AS
2/* 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>
c15952dc 12#include <linux/bpf_common.h>
daedfb22
AS
13
14/* Extended instruction set based on top of classic BPF */
15
16/* instruction classes */
d405c740 17#define BPF_JMP32 0x06 /* jmp mode in word width */
daedfb22
AS
18#define BPF_ALU64 0x07 /* alu mode in double word width */
19
20/* ld/ldx fields */
cb5f7334 21#define BPF_DW 0x18 /* double word (64-bit) */
91c960b0
BJ
22#define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
23#define BPF_XADD 0xc0 /* exclusive add - legacy name */
daedfb22
AS
24
25/* alu/jmp fields */
26#define BPF_MOV 0xb0 /* mov reg to reg */
27#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28
29/* change endianness of a register */
30#define BPF_END 0xd0 /* flags for endianness conversion: */
31#define BPF_TO_LE 0x00 /* convert to little-endian */
32#define BPF_TO_BE 0x08 /* convert to big-endian */
33#define BPF_FROM_LE BPF_TO_LE
34#define BPF_FROM_BE BPF_TO_BE
35
92b31a9a 36/* jmp encodings */
daedfb22 37#define BPF_JNE 0x50 /* jump != */
92b31a9a
DB
38#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
39#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
daedfb22
AS
40#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
41#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
92b31a9a
DB
42#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
43#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
daedfb22
AS
44#define BPF_CALL 0x80 /* function call */
45#define BPF_EXIT 0x90 /* function return */
46
47/* Register numbers */
48enum {
49 BPF_REG_0 = 0,
50 BPF_REG_1,
51 BPF_REG_2,
52 BPF_REG_3,
53 BPF_REG_4,
54 BPF_REG_5,
55 BPF_REG_6,
56 BPF_REG_7,
57 BPF_REG_8,
58 BPF_REG_9,
59 BPF_REG_10,
60 __MAX_BPF_REG,
61};
62
63/* BPF has 10 general purpose 64-bit registers and stack frame. */
64#define MAX_BPF_REG __MAX_BPF_REG
65
66struct bpf_insn {
67 __u8 code; /* opcode */
68 __u8 dst_reg:4; /* dest register */
69 __u8 src_reg:4; /* source register */
70 __s16 off; /* signed offset */
71 __s32 imm; /* signed immediate constant */
72};
73
b95a5c4d
DM
74/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
75struct bpf_lpm_trie_key {
76 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
1e6e9d0f 77 __u8 data[0]; /* Arbitrary size */
b95a5c4d
DM
78};
79
de9cbbaa
RG
80struct bpf_cgroup_storage_key {
81 __u64 cgroup_inode_id; /* cgroup inode id */
82 __u32 attach_type; /* program attach type */
83};
84
5e7b3020
YS
85union bpf_iter_link_info {
86 struct {
87 __u32 map_fd;
88 } map;
89};
90
b2197755 91/* BPF syscall commands, see bpf(2) man-page for details. */
99c55f7d 92enum bpf_cmd {
99c55f7d 93 BPF_MAP_CREATE,
db20fd2b 94 BPF_MAP_LOOKUP_ELEM,
db20fd2b 95 BPF_MAP_UPDATE_ELEM,
db20fd2b 96 BPF_MAP_DELETE_ELEM,
db20fd2b 97 BPF_MAP_GET_NEXT_KEY,
09756af4 98 BPF_PROG_LOAD,
b2197755
DB
99 BPF_OBJ_PIN,
100 BPF_OBJ_GET,
f4324551
DM
101 BPF_PROG_ATTACH,
102 BPF_PROG_DETACH,
1cf1cae9 103 BPF_PROG_TEST_RUN,
34ad5580
MKL
104 BPF_PROG_GET_NEXT_ID,
105 BPF_MAP_GET_NEXT_ID,
b16d9aa4 106 BPF_PROG_GET_FD_BY_ID,
bd5f5f4e 107 BPF_MAP_GET_FD_BY_ID,
1e270976 108 BPF_OBJ_GET_INFO_BY_FD,
468e2f64 109 BPF_PROG_QUERY,
c4f6699d 110 BPF_RAW_TRACEPOINT_OPEN,
f56a653c 111 BPF_BTF_LOAD,
78958fca 112 BPF_BTF_GET_FD_BY_ID,
41bdc4b4 113 BPF_TASK_FD_QUERY,
bd513cd0 114 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
87df15de 115 BPF_MAP_FREEZE,
1b9ed84e 116 BPF_BTF_GET_NEXT_ID,
cb4d03ab 117 BPF_MAP_LOOKUP_BATCH,
05799638 118 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
aa2e93b8
BV
119 BPF_MAP_UPDATE_BATCH,
120 BPF_MAP_DELETE_BATCH,
af6eea57 121 BPF_LINK_CREATE,
0c991ebc 122 BPF_LINK_UPDATE,
2d602c8c
AN
123 BPF_LINK_GET_FD_BY_ID,
124 BPF_LINK_GET_NEXT_ID,
d46edd67 125 BPF_ENABLE_STATS,
ac51d99b 126 BPF_ITER_CREATE,
73b11c2a 127 BPF_LINK_DETACH,
ef15314a 128 BPF_PROG_BIND_MAP,
99c55f7d
AS
129};
130
131enum bpf_map_type {
132 BPF_MAP_TYPE_UNSPEC,
0f8e4bd8 133 BPF_MAP_TYPE_HASH,
28fbcfa0 134 BPF_MAP_TYPE_ARRAY,
04fd61ab 135 BPF_MAP_TYPE_PROG_ARRAY,
ea317b26 136 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
824bd0ce 137 BPF_MAP_TYPE_PERCPU_HASH,
a10423b8 138 BPF_MAP_TYPE_PERCPU_ARRAY,
d5a3b1f6 139 BPF_MAP_TYPE_STACK_TRACE,
4ed8ec52 140 BPF_MAP_TYPE_CGROUP_ARRAY,
29ba732a 141 BPF_MAP_TYPE_LRU_HASH,
8f844938 142 BPF_MAP_TYPE_LRU_PERCPU_HASH,
b95a5c4d 143 BPF_MAP_TYPE_LPM_TRIE,
56f668df 144 BPF_MAP_TYPE_ARRAY_OF_MAPS,
bcc6b1b7 145 BPF_MAP_TYPE_HASH_OF_MAPS,
546ac1ff 146 BPF_MAP_TYPE_DEVMAP,
174a79ff 147 BPF_MAP_TYPE_SOCKMAP,
6710e112 148 BPF_MAP_TYPE_CPUMAP,
fbfc504a 149 BPF_MAP_TYPE_XSKMAP,
81110384 150 BPF_MAP_TYPE_SOCKHASH,
de9cbbaa 151 BPF_MAP_TYPE_CGROUP_STORAGE,
5dc4c4b7 152 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
b741f163 153 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
f1a2e44a
MV
154 BPF_MAP_TYPE_QUEUE,
155 BPF_MAP_TYPE_STACK,
6ac99e8f 156 BPF_MAP_TYPE_SK_STORAGE,
6f9d451a 157 BPF_MAP_TYPE_DEVMAP_HASH,
85d33df3 158 BPF_MAP_TYPE_STRUCT_OPS,
457f4436 159 BPF_MAP_TYPE_RINGBUF,
8ea63684 160 BPF_MAP_TYPE_INODE_STORAGE,
4cf1bc1f 161 BPF_MAP_TYPE_TASK_STORAGE,
99c55f7d
AS
162};
163
6c4fc209
DB
164/* Note that tracing related programs such as
165 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
166 * are not subject to a stable API since kernel internal data
167 * structures can change from release to release and may
168 * therefore break existing tracing BPF programs. Tracing BPF
169 * programs correspond to /a/ specific kernel which is to be
170 * analyzed, and not /a/ specific kernel /and/ all future ones.
171 */
09756af4
AS
172enum bpf_prog_type {
173 BPF_PROG_TYPE_UNSPEC,
ddd872bc 174 BPF_PROG_TYPE_SOCKET_FILTER,
2541517c 175 BPF_PROG_TYPE_KPROBE,
96be4325 176 BPF_PROG_TYPE_SCHED_CLS,
94caee8c 177 BPF_PROG_TYPE_SCHED_ACT,
98b5c2c6 178 BPF_PROG_TYPE_TRACEPOINT,
6a773a15 179 BPF_PROG_TYPE_XDP,
0515e599 180 BPF_PROG_TYPE_PERF_EVENT,
0e33661d 181 BPF_PROG_TYPE_CGROUP_SKB,
61023658 182 BPF_PROG_TYPE_CGROUP_SOCK,
3a0af8fd
TG
183 BPF_PROG_TYPE_LWT_IN,
184 BPF_PROG_TYPE_LWT_OUT,
185 BPF_PROG_TYPE_LWT_XMIT,
40304b2a 186 BPF_PROG_TYPE_SOCK_OPS,
b005fd18 187 BPF_PROG_TYPE_SK_SKB,
ebc614f6 188 BPF_PROG_TYPE_CGROUP_DEVICE,
4f738adb 189 BPF_PROG_TYPE_SK_MSG,
c4f6699d 190 BPF_PROG_TYPE_RAW_TRACEPOINT,
4fbac77d 191 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
004d4b27 192 BPF_PROG_TYPE_LWT_SEG6LOCAL,
f4364dcf 193 BPF_PROG_TYPE_LIRC_MODE2,
2dbb9b9e 194 BPF_PROG_TYPE_SK_REUSEPORT,
d58e468b 195 BPF_PROG_TYPE_FLOW_DISSECTOR,
7b146ceb 196 BPF_PROG_TYPE_CGROUP_SYSCTL,
9df1c28b 197 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
0d01da6a 198 BPF_PROG_TYPE_CGROUP_SOCKOPT,
f1b9509c 199 BPF_PROG_TYPE_TRACING,
27ae7997 200 BPF_PROG_TYPE_STRUCT_OPS,
be8704ff 201 BPF_PROG_TYPE_EXT,
fc611f47 202 BPF_PROG_TYPE_LSM,
e9ddbb77 203 BPF_PROG_TYPE_SK_LOOKUP,
09756af4
AS
204};
205
0e33661d
DM
206enum bpf_attach_type {
207 BPF_CGROUP_INET_INGRESS,
208 BPF_CGROUP_INET_EGRESS,
61023658 209 BPF_CGROUP_INET_SOCK_CREATE,
40304b2a 210 BPF_CGROUP_SOCK_OPS,
464bc0fd
JF
211 BPF_SK_SKB_STREAM_PARSER,
212 BPF_SK_SKB_STREAM_VERDICT,
ebc614f6 213 BPF_CGROUP_DEVICE,
4f738adb 214 BPF_SK_MSG_VERDICT,
4fbac77d
AI
215 BPF_CGROUP_INET4_BIND,
216 BPF_CGROUP_INET6_BIND,
d74bad4e
AI
217 BPF_CGROUP_INET4_CONNECT,
218 BPF_CGROUP_INET6_CONNECT,
aac3fc32
AI
219 BPF_CGROUP_INET4_POST_BIND,
220 BPF_CGROUP_INET6_POST_BIND,
1cedee13
AI
221 BPF_CGROUP_UDP4_SENDMSG,
222 BPF_CGROUP_UDP6_SENDMSG,
f4364dcf 223 BPF_LIRC_MODE2,
d58e468b 224 BPF_FLOW_DISSECTOR,
7b146ceb 225 BPF_CGROUP_SYSCTL,
983695fa
DB
226 BPF_CGROUP_UDP4_RECVMSG,
227 BPF_CGROUP_UDP6_RECVMSG,
0d01da6a
SF
228 BPF_CGROUP_GETSOCKOPT,
229 BPF_CGROUP_SETSOCKOPT,
f1b9509c 230 BPF_TRACE_RAW_TP,
fec56f58
AS
231 BPF_TRACE_FENTRY,
232 BPF_TRACE_FEXIT,
ae240823 233 BPF_MODIFY_RETURN,
fc611f47 234 BPF_LSM_MAC,
15d83c4d 235 BPF_TRACE_ITER,
1b66d253
DB
236 BPF_CGROUP_INET4_GETPEERNAME,
237 BPF_CGROUP_INET6_GETPEERNAME,
238 BPF_CGROUP_INET4_GETSOCKNAME,
239 BPF_CGROUP_INET6_GETSOCKNAME,
fbee97fe 240 BPF_XDP_DEVMAP,
f5836749 241 BPF_CGROUP_INET_SOCK_RELEASE,
92164774 242 BPF_XDP_CPUMAP,
e9ddbb77 243 BPF_SK_LOOKUP,
aa8d3a71 244 BPF_XDP,
0e33661d
DM
245 __MAX_BPF_ATTACH_TYPE
246};
247
248#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
249
f2e10bff
AN
250enum bpf_link_type {
251 BPF_LINK_TYPE_UNSPEC = 0,
252 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
253 BPF_LINK_TYPE_TRACING = 2,
254 BPF_LINK_TYPE_CGROUP = 3,
de4e05ca 255 BPF_LINK_TYPE_ITER = 4,
7f045a49 256 BPF_LINK_TYPE_NETNS = 5,
aa8d3a71 257 BPF_LINK_TYPE_XDP = 6,
f2e10bff
AN
258
259 MAX_BPF_LINK_TYPE,
260};
261
324bda9e
AS
262/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
263 *
264 * NONE(default): No further bpf programs allowed in the subtree.
265 *
266 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
267 * the program in this cgroup yields to sub-cgroup program.
268 *
269 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
270 * that cgroup program gets run in addition to the program in this cgroup.
271 *
272 * Only one program is allowed to be attached to a cgroup with
273 * NONE or BPF_F_ALLOW_OVERRIDE flag.
274 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
275 * release old program and attach the new one. Attach flags has to match.
276 *
277 * Multiple programs are allowed to be attached to a cgroup with
278 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
279 * (those that were attached first, run first)
280 * The programs of sub-cgroup are executed first, then programs of
281 * this cgroup and then programs of parent cgroup.
282 * When children program makes decision (like picking TCP CA or sock bind)
283 * parent program has a chance to override it.
284 *
7dd68b32
AI
285 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
286 * programs for a cgroup. Though it's possible to replace an old program at
287 * any position by also specifying BPF_F_REPLACE flag and position itself in
288 * replace_bpf_fd attribute. Old program at this position will be released.
289 *
324bda9e
AS
290 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
291 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
292 * Ex1:
293 * cgrp1 (MULTI progs A, B) ->
294 * cgrp2 (OVERRIDE prog C) ->
295 * cgrp3 (MULTI prog D) ->
296 * cgrp4 (OVERRIDE prog E) ->
297 * cgrp5 (NONE prog F)
298 * the event in cgrp5 triggers execution of F,D,A,B in that order.
299 * if prog F is detached, the execution is E,D,A,B
300 * if prog F and D are detached, the execution is E,A,B
301 * if prog F, E and D are detached, the execution is C,A,B
302 *
303 * All eligible programs are executed regardless of return code from
304 * earlier programs.
7f677633
AS
305 */
306#define BPF_F_ALLOW_OVERRIDE (1U << 0)
324bda9e 307#define BPF_F_ALLOW_MULTI (1U << 1)
7dd68b32 308#define BPF_F_REPLACE (1U << 2)
7f677633 309
e07b98d9
DM
310/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
311 * verifier will perform strict alignment checking as if the kernel
312 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
313 * and NET_IP_ALIGN defined to 2.
314 */
315#define BPF_F_STRICT_ALIGNMENT (1U << 0)
316
e9ee9efc
DM
317/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
318 * verifier will allow any alignment whatsoever. On platforms
319 * with strict alignment requirements for loads ands stores (such
320 * as sparc and mips) the verifier validates that all loads and
321 * stores provably follow this requirement. This flag turns that
322 * checking and enforcement off.
323 *
324 * It is mostly used for testing when we want to validate the
325 * context and memory access aspects of the verifier, but because
326 * of an unaligned access the alignment check would trigger before
327 * the one we are interested in.
328 */
329#define BPF_F_ANY_ALIGNMENT (1U << 1)
330
c240eff6
JW
331/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
332 * Verifier does sub-register def/use analysis and identifies instructions whose
333 * def only matters for low 32-bit, high 32-bit is never referenced later
334 * through implicit zero extension. Therefore verifier notifies JIT back-ends
335 * that it is safe to ignore clearing high 32-bit for these instructions. This
336 * saves some back-ends a lot of code-gen. However such optimization is not
337 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
338 * hence hasn't used verifier's analysis result. But, we really want to have a
339 * way to be able to verify the correctness of the described optimization on
340 * x86_64 on which testsuites are frequently exercised.
341 *
342 * So, this flag is introduced. Once it is set, verifier will randomize high
343 * 32-bit for those instructions who has been identified as safe to ignore them.
344 * Then, if verifier is not doing correct analysis, such randomization will
345 * regress tests to expose bugs.
346 */
347#define BPF_F_TEST_RND_HI32 (1U << 2)
348
10d274e8
AS
349/* The verifier internal test flag. Behavior is undefined */
350#define BPF_F_TEST_STATE_FREQ (1U << 3)
351
1e6c62a8
AS
352/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
353 * restrict map and helper usage for such programs. Sleepable BPF programs can
354 * only be attached to hooks where kernel execution context allows sleeping.
355 * Such programs are allowed to use helpers that may sleep like
356 * bpf_copy_from_user().
357 */
358#define BPF_F_SLEEPABLE (1U << 4)
359
d8eca5bb 360/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
4976b718 361 * the following extensions:
d8eca5bb 362 *
4976b718
HL
363 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
364 * insn[0].imm: map fd
365 * insn[1].imm: 0
366 * insn[0].off: 0
367 * insn[1].off: 0
368 * ldimm64 rewrite: address of map
369 * verifier type: CONST_PTR_TO_MAP
d8eca5bb 370 */
f1a66f85 371#define BPF_PSEUDO_MAP_FD 1
4976b718
HL
372/* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
373 * insn[0].imm: map fd
374 * insn[1].imm: offset into value
375 * insn[0].off: 0
376 * insn[1].off: 0
377 * ldimm64 rewrite: address of map[0]+offset
378 * verifier type: PTR_TO_MAP_VALUE
379 */
d8eca5bb 380#define BPF_PSEUDO_MAP_VALUE 2
4976b718
HL
381/* insn[0].src_reg: BPF_PSEUDO_BTF_ID
382 * insn[0].imm: kernel btd id of VAR
383 * insn[1].imm: 0
384 * insn[0].off: 0
385 * insn[1].off: 0
386 * ldimm64 rewrite: address of the kernel variable
387 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
388 * is struct/union.
389 */
390#define BPF_PSEUDO_BTF_ID 3
f1a66f85 391
cc8b0b92
AS
392/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
393 * offset to another bpf function
394 */
395#define BPF_PSEUDO_CALL 1
396
3274f520 397/* flags for BPF_MAP_UPDATE_ELEM command */
1aae4bdd
AN
398enum {
399 BPF_ANY = 0, /* create new element or update existing */
400 BPF_NOEXIST = 1, /* create new element if it didn't exist */
401 BPF_EXIST = 2, /* update existing element */
402 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
403};
3274f520 404
96eabe7a 405/* flags for BPF_MAP_CREATE command */
1aae4bdd
AN
406enum {
407 BPF_F_NO_PREALLOC = (1U << 0),
29ba732a 408/* Instead of having one common LRU list in the
8f844938 409 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
29ba732a
MKL
410 * which can scale and perform better.
411 * Note, the LRU nodes (including free nodes) cannot be moved
412 * across different LRU lists.
413 */
1aae4bdd 414 BPF_F_NO_COMMON_LRU = (1U << 1),
96eabe7a 415/* Specify numa node during map creation */
1aae4bdd 416 BPF_F_NUMA_NODE = (1U << 2),
cb4d2b3f 417
591fe988 418/* Flags for accessing BPF object from syscall side. */
1aae4bdd
AN
419 BPF_F_RDONLY = (1U << 3),
420 BPF_F_WRONLY = (1U << 4),
6e71b04a 421
615755a7 422/* Flag for stack_map, store build_id+offset instead of pointer */
1aae4bdd 423 BPF_F_STACK_BUILD_ID = (1U << 5),
615755a7 424
96b3b6c9 425/* Zero-initialize hash function seed. This should only be used for testing. */
1aae4bdd 426 BPF_F_ZERO_SEED = (1U << 6),
96b3b6c9 427
591fe988 428/* Flags for accessing BPF object from program side. */
1aae4bdd
AN
429 BPF_F_RDONLY_PROG = (1U << 7),
430 BPF_F_WRONLY_PROG = (1U << 8),
591fe988 431
8f51dfc7 432/* Clone map from listener for newly accepted socket */
1aae4bdd 433 BPF_F_CLONE = (1U << 9),
8f51dfc7 434
fc970227 435/* Enable memory-mapping BPF map */
1aae4bdd 436 BPF_F_MMAPABLE = (1U << 10),
792caccc
SL
437
438/* Share perf_event among processes */
439 BPF_F_PRESERVE_ELEMS = (1U << 11),
4a8f87e6
DB
440
441/* Create a map that is suitable to be an inner map with dynamic max entries */
442 BPF_F_INNER_MAP = (1U << 12),
1aae4bdd 443};
fc970227 444
f5bfcd95
AI
445/* Flags for BPF_PROG_QUERY. */
446
447/* Query effective (directly attached + inherited from ancestor cgroups)
448 * programs that will be executed for events within a cgroup.
449 * attach_flags with this flag are returned only for directly attached programs.
450 */
2f183360
LB
451#define BPF_F_QUERY_EFFECTIVE (1U << 0)
452
1b4d60ec
SL
453/* Flags for BPF_PROG_TEST_RUN */
454
455/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
456#define BPF_F_TEST_RUN_ON_CPU (1U << 0)
457
d46edd67
SL
458/* type for BPF_ENABLE_STATS */
459enum bpf_stats_type {
460 /* enabled run_time_ns and run_cnt */
461 BPF_STATS_RUN_TIME = 0,
462};
463
615755a7
SL
464enum bpf_stack_build_id_status {
465 /* user space need an empty entry to identify end of a trace */
466 BPF_STACK_BUILD_ID_EMPTY = 0,
467 /* with valid build_id and offset */
468 BPF_STACK_BUILD_ID_VALID = 1,
469 /* couldn't get build_id, fallback to ip */
470 BPF_STACK_BUILD_ID_IP = 2,
471};
472
473#define BPF_BUILD_ID_SIZE 20
474struct bpf_stack_build_id {
475 __s32 status;
476 unsigned char build_id[BPF_BUILD_ID_SIZE];
477 union {
478 __u64 offset;
479 __u64 ip;
480 };
481};
482
1aae4bdd
AN
483#define BPF_OBJ_NAME_LEN 16U
484
99c55f7d
AS
485union bpf_attr {
486 struct { /* anonymous struct used by BPF_MAP_CREATE command */
487 __u32 map_type; /* one of enum bpf_map_type */
488 __u32 key_size; /* size of key in bytes */
489 __u32 value_size; /* size of value in bytes */
490 __u32 max_entries; /* max number of entries in a map */
96eabe7a
MKL
491 __u32 map_flags; /* BPF_MAP_CREATE related
492 * flags defined above.
493 */
56f668df 494 __u32 inner_map_fd; /* fd pointing to the inner map */
96eabe7a
MKL
495 __u32 numa_node; /* numa node (effective only if
496 * BPF_F_NUMA_NODE is set).
497 */
067cae47 498 char map_name[BPF_OBJ_NAME_LEN];
a3884572 499 __u32 map_ifindex; /* ifindex of netdev to create on */
a26ca7c9 500 __u32 btf_fd; /* fd pointing to a BTF type data */
9b2cf328
MKL
501 __u32 btf_key_type_id; /* BTF type_id of the key */
502 __u32 btf_value_type_id; /* BTF type_id of the value */
85d33df3
MKL
503 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
504 * struct stored as the
505 * map value
506 */
99c55f7d 507 };
db20fd2b
AS
508
509 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
510 __u32 map_fd;
511 __aligned_u64 key;
512 union {
513 __aligned_u64 value;
514 __aligned_u64 next_key;
515 };
3274f520 516 __u64 flags;
db20fd2b 517 };
09756af4 518
cb4d03ab
BV
519 struct { /* struct used by BPF_MAP_*_BATCH commands */
520 __aligned_u64 in_batch; /* start batch,
521 * NULL to start from beginning
522 */
523 __aligned_u64 out_batch; /* output: next start batch */
524 __aligned_u64 keys;
525 __aligned_u64 values;
526 __u32 count; /* input/output:
527 * input: # of key/value
528 * elements
529 * output: # of filled elements
530 */
531 __u32 map_fd;
532 __u64 elem_flags;
533 __u64 flags;
534 } batch;
535
09756af4
AS
536 struct { /* anonymous struct used by BPF_PROG_LOAD command */
537 __u32 prog_type; /* one of enum bpf_prog_type */
538 __u32 insn_cnt;
539 __aligned_u64 insns;
540 __aligned_u64 license;
cbd35700
AS
541 __u32 log_level; /* verbosity level of verifier */
542 __u32 log_size; /* size of user buffer */
543 __aligned_u64 log_buf; /* user supplied buffer */
6c4fc209 544 __u32 kern_version; /* not used */
e07b98d9 545 __u32 prog_flags;
067cae47 546 char prog_name[BPF_OBJ_NAME_LEN];
1f6f4cb7 547 __u32 prog_ifindex; /* ifindex of netdev to prep for */
5e43f899
AI
548 /* For some prog types expected attach type must be known at
549 * load time to verify attach type specific parts of prog
550 * (context accesses, allowed helpers, etc).
551 */
552 __u32 expected_attach_type;
838e9690
YS
553 __u32 prog_btf_fd; /* fd pointing to BTF type data */
554 __u32 func_info_rec_size; /* userspace bpf_func_info size */
555 __aligned_u64 func_info; /* func info */
556 __u32 func_info_cnt; /* number of bpf_func_info records */
c454a46b
MKL
557 __u32 line_info_rec_size; /* userspace bpf_line_info size */
558 __aligned_u64 line_info; /* line info */
559 __u32 line_info_cnt; /* number of bpf_line_info records */
ccfe29eb 560 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
290248a5
AN
561 union {
562 /* valid prog_fd to attach to bpf prog */
563 __u32 attach_prog_fd;
564 /* or valid module BTF object fd or 0 to attach to vmlinux */
565 __u32 attach_btf_obj_fd;
566 };
09756af4 567 };
b2197755
DB
568
569 struct { /* anonymous struct used by BPF_OBJ_* commands */
570 __aligned_u64 pathname;
571 __u32 bpf_fd;
6e71b04a 572 __u32 file_flags;
b2197755 573 };
f4324551
DM
574
575 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
576 __u32 target_fd; /* container object to attach to */
577 __u32 attach_bpf_fd; /* eBPF program to attach */
578 __u32 attach_type;
7f677633 579 __u32 attach_flags;
7dd68b32
AI
580 __u32 replace_bpf_fd; /* previously attached eBPF
581 * program to replace if
582 * BPF_F_REPLACE is used
583 */
f4324551 584 };
1cf1cae9
AS
585
586 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
587 __u32 prog_fd;
588 __u32 retval;
b5a36b1e
LB
589 __u32 data_size_in; /* input: len of data_in */
590 __u32 data_size_out; /* input/output: len of data_out
591 * returns ENOSPC if data_out
592 * is too small.
593 */
1cf1cae9
AS
594 __aligned_u64 data_in;
595 __aligned_u64 data_out;
596 __u32 repeat;
597 __u32 duration;
b0b9395d
SF
598 __u32 ctx_size_in; /* input: len of ctx_in */
599 __u32 ctx_size_out; /* input/output: len of ctx_out
600 * returns ENOSPC if ctx_out
601 * is too small.
602 */
603 __aligned_u64 ctx_in;
604 __aligned_u64 ctx_out;
1b4d60ec
SL
605 __u32 flags;
606 __u32 cpu;
1cf1cae9 607 } test;
34ad5580 608
b16d9aa4
MKL
609 struct { /* anonymous struct used by BPF_*_GET_*_ID */
610 union {
611 __u32 start_id;
612 __u32 prog_id;
bd5f5f4e 613 __u32 map_id;
78958fca 614 __u32 btf_id;
a3b80e10 615 __u32 link_id;
b16d9aa4 616 };
34ad5580 617 __u32 next_id;
6e71b04a 618 __u32 open_flags;
34ad5580 619 };
1e270976
MKL
620
621 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
622 __u32 bpf_fd;
623 __u32 info_len;
624 __aligned_u64 info;
625 } info;
468e2f64
AS
626
627 struct { /* anonymous struct used by BPF_PROG_QUERY command */
628 __u32 target_fd; /* container object to query */
629 __u32 attach_type;
630 __u32 query_flags;
631 __u32 attach_flags;
632 __aligned_u64 prog_ids;
633 __u32 prog_cnt;
634 } query;
c4f6699d 635
af6eea57 636 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
c4f6699d
AS
637 __u64 name;
638 __u32 prog_fd;
639 } raw_tracepoint;
f56a653c
MKL
640
641 struct { /* anonymous struct for BPF_BTF_LOAD */
642 __aligned_u64 btf;
643 __aligned_u64 btf_log_buf;
644 __u32 btf_size;
645 __u32 btf_log_size;
646 __u32 btf_log_level;
647 };
41bdc4b4
YS
648
649 struct {
650 __u32 pid; /* input: pid */
651 __u32 fd; /* input: fd */
652 __u32 flags; /* input: flags */
653 __u32 buf_len; /* input/output: buf len */
654 __aligned_u64 buf; /* input/output:
655 * tp_name for tracepoint
656 * symbol for kprobe
657 * filename for uprobe
658 */
659 __u32 prog_id; /* output: prod_id */
660 __u32 fd_type; /* output: BPF_FD_TYPE_* */
661 __u64 probe_offset; /* output: probe_offset */
662 __u64 probe_addr; /* output: probe_addr */
663 } task_fd_query;
af6eea57
AN
664
665 struct { /* struct used by BPF_LINK_CREATE command */
666 __u32 prog_fd; /* eBPF program to attach */
aa8d3a71
AN
667 union {
668 __u32 target_fd; /* object to attach to */
669 __u32 target_ifindex; /* target ifindex */
670 };
af6eea57
AN
671 __u32 attach_type; /* attach type */
672 __u32 flags; /* extra flags */
4a1e7c0c
THJ
673 union {
674 __u32 target_btf_id; /* btf_id of target to attach to */
675 struct {
676 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
677 __u32 iter_info_len; /* iter_info length */
678 };
679 };
af6eea57 680 } link_create;
0c991ebc
AN
681
682 struct { /* struct used by BPF_LINK_UPDATE command */
683 __u32 link_fd; /* link fd */
684 /* new program fd to update link with */
685 __u32 new_prog_fd;
686 __u32 flags; /* extra flags */
687 /* expected link's program fd; is specified only if
688 * BPF_F_REPLACE flag is set in flags */
689 __u32 old_prog_fd;
690 } link_update;
691
73b11c2a
AN
692 struct {
693 __u32 link_fd;
694 } link_detach;
695
d46edd67
SL
696 struct { /* struct used by BPF_ENABLE_STATS command */
697 __u32 type;
698 } enable_stats;
699
ac51d99b
YS
700 struct { /* struct used by BPF_ITER_CREATE command */
701 __u32 link_fd;
702 __u32 flags;
703 } iter_create;
704
ef15314a
YZ
705 struct { /* struct used by BPF_PROG_BIND_MAP command */
706 __u32 prog_fd;
707 __u32 map_fd;
708 __u32 flags; /* extra flags */
709 } prog_bind_map;
710
99c55f7d
AS
711} __attribute__((aligned(8)));
712
56a092c8
QM
713/* The description below is an attempt at providing documentation to eBPF
714 * developers about the multiple available eBPF helper functions. It can be
715 * parsed and used to produce a manual page. The workflow is the following,
716 * and requires the rst2man utility:
717 *
718 * $ ./scripts/bpf_helpers_doc.py \
719 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
720 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
721 * $ man /tmp/bpf-helpers.7
722 *
723 * Note that in order to produce this external documentation, some RST
724 * formatting is used in the descriptions to get "bold" and "italics" in
725 * manual pages. Also note that the few trailing white spaces are
726 * intentional, removing them would break paragraphs for rst2man.
727 *
728 * Start of BPF helper function descriptions:
ad4a5223
QM
729 *
730 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
731 * Description
732 * Perform a lookup in *map* for an entry associated to *key*.
733 * Return
734 * Map value associated to *key*, or **NULL** if no entry was
735 * found.
736 *
bdb7b79b 737 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
ad4a5223
QM
738 * Description
739 * Add or update the value of the entry associated to *key* in
740 * *map* with *value*. *flags* is one of:
741 *
742 * **BPF_NOEXIST**
743 * The entry for *key* must not exist in the map.
744 * **BPF_EXIST**
745 * The entry for *key* must already exist in the map.
746 * **BPF_ANY**
747 * No condition on the existence of the entry for *key*.
748 *
749 * Flag value **BPF_NOEXIST** cannot be used for maps of types
750 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
751 * elements always exist), the helper would return an error.
752 * Return
753 * 0 on success, or a negative error in case of failure.
754 *
bdb7b79b 755 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
ad4a5223
QM
756 * Description
757 * Delete entry with *key* from *map*.
758 * Return
759 * 0 on success, or a negative error in case of failure.
760 *
bdb7b79b 761 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
ad4a5223
QM
762 * Description
763 * For tracing programs, safely attempt to read *size* bytes from
6ae08ae3
DB
764 * kernel space address *unsafe_ptr* and store the data in *dst*.
765 *
ab8d7809
QM
766 * Generally, use **bpf_probe_read_user**\ () or
767 * **bpf_probe_read_kernel**\ () instead.
ad4a5223
QM
768 * Return
769 * 0 on success, or a negative error in case of failure.
770 *
771 * u64 bpf_ktime_get_ns(void)
772 * Description
773 * Return the time elapsed since system boot, in nanoseconds.
71d19214 774 * Does not include time the system was suspended.
ab8d7809 775 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
ad4a5223
QM
776 * Return
777 * Current *ktime*.
778 *
bdb7b79b 779 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
ad4a5223
QM
780 * Description
781 * This helper is a "printk()-like" facility for debugging. It
782 * prints a message defined by format *fmt* (of size *fmt_size*)
783 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
784 * available. It can take up to three additional **u64**
785 * arguments (as an eBPF helpers, the total number of arguments is
786 * limited to five).
787 *
788 * Each time the helper is called, it appends a line to the trace.
55c33dfb
PW
789 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
790 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
ad4a5223
QM
791 * The format of the trace is customizable, and the exact output
792 * one will get depends on the options set in
793 * *\/sys/kernel/debug/tracing/trace_options* (see also the
794 * *README* file under the same directory). However, it usually
795 * defaults to something like:
796 *
797 * ::
798 *
799 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
800 *
801 * In the above:
802 *
803 * * ``telnet`` is the name of the current task.
804 * * ``470`` is the PID of the current task.
805 * * ``001`` is the CPU number on which the task is
806 * running.
807 * * In ``.N..``, each character refers to a set of
808 * options (whether irqs are enabled, scheduling
809 * options, whether hard/softirqs are running, level of
810 * preempt_disabled respectively). **N** means that
811 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
812 * are set.
813 * * ``419421.045894`` is a timestamp.
814 * * ``0x00000001`` is a fake value used by BPF for the
815 * instruction pointer register.
816 * * ``<formatted msg>`` is the message formatted with
817 * *fmt*.
818 *
819 * The conversion specifiers supported by *fmt* are similar, but
820 * more limited than for printk(). They are **%d**, **%i**,
821 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
822 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
823 * of field, padding with zeroes, etc.) is available, and the
824 * helper will return **-EINVAL** (but print nothing) if it
825 * encounters an unknown specifier.
826 *
827 * Also, note that **bpf_trace_printk**\ () is slow, and should
828 * only be used for debugging purposes. For this reason, a notice
b16fc097 829 * block (spanning several lines) is printed to kernel logs and
ad4a5223
QM
830 * states that the helper should not be used "for production use"
831 * the first time this helper is used (or more precisely, when
832 * **trace_printk**\ () buffers are allocated). For passing values
833 * to user space, perf events should be preferred.
834 * Return
835 * The number of bytes written to the buffer, or a negative error
836 * in case of failure.
837 *
1fdd08be
QM
838 * u32 bpf_get_prandom_u32(void)
839 * Description
840 * Get a pseudo-random number.
841 *
842 * From a security point of view, this helper uses its own
843 * pseudo-random internal state, and cannot be used to infer the
844 * seed of other random functions in the kernel. However, it is
845 * essential to note that the generator used by the helper is not
846 * cryptographically secure.
847 * Return
848 * A random 32-bit unsigned value.
849 *
850 * u32 bpf_get_smp_processor_id(void)
851 * Description
852 * Get the SMP (symmetric multiprocessing) processor id. Note that
853 * all programs run with preemption disabled, which means that the
854 * SMP processor id is stable during all the execution of the
855 * program.
856 * Return
857 * The SMP id of the processor running the program.
858 *
bdb7b79b 859 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
ad4a5223
QM
860 * Description
861 * Store *len* bytes from address *from* into the packet
862 * associated to *skb*, at *offset*. *flags* are a combination of
863 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
864 * checksum for the packet after storing the bytes) and
865 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
866 * **->swhash** and *skb*\ **->l4hash** to 0).
867 *
32e7dc28 868 * A call to this helper is susceptible to change the underlying
ad4a5223
QM
869 * packet buffer. Therefore, at load time, all checks on pointers
870 * previously done by the verifier are invalidated and must be
871 * performed again, if the helper is used in combination with
872 * direct packet access.
873 * Return
874 * 0 on success, or a negative error in case of failure.
875 *
bdb7b79b 876 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
ad4a5223
QM
877 * Description
878 * Recompute the layer 3 (e.g. IP) checksum for the packet
879 * associated to *skb*. Computation is incremental, so the helper
880 * must know the former value of the header field that was
881 * modified (*from*), the new value of this field (*to*), and the
882 * number of bytes (2 or 4) for this field, stored in *size*.
883 * Alternatively, it is possible to store the difference between
884 * the previous and the new values of the header field in *to*, by
885 * setting *from* and *size* to 0. For both methods, *offset*
886 * indicates the location of the IP checksum within the packet.
887 *
888 * This helper works in combination with **bpf_csum_diff**\ (),
889 * which does not update the checksum in-place, but offers more
890 * flexibility and can handle sizes larger than 2 or 4 for the
891 * checksum to update.
892 *
32e7dc28 893 * A call to this helper is susceptible to change the underlying
ad4a5223
QM
894 * packet buffer. Therefore, at load time, all checks on pointers
895 * previously done by the verifier are invalidated and must be
896 * performed again, if the helper is used in combination with
897 * direct packet access.
898 * Return
899 * 0 on success, or a negative error in case of failure.
900 *
bdb7b79b 901 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
ad4a5223
QM
902 * Description
903 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
904 * packet associated to *skb*. Computation is incremental, so the
905 * helper must know the former value of the header field that was
906 * modified (*from*), the new value of this field (*to*), and the
907 * number of bytes (2 or 4) for this field, stored on the lowest
908 * four bits of *flags*. Alternatively, it is possible to store
909 * the difference between the previous and the new values of the
910 * header field in *to*, by setting *from* and the four lowest
911 * bits of *flags* to 0. For both methods, *offset* indicates the
912 * location of the IP checksum within the packet. In addition to
913 * the size of the field, *flags* can be added (bitwise OR) actual
914 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
915 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
916 * for updates resulting in a null checksum the value is set to
917 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
918 * the checksum is to be computed against a pseudo-header.
919 *
920 * This helper works in combination with **bpf_csum_diff**\ (),
921 * which does not update the checksum in-place, but offers more
922 * flexibility and can handle sizes larger than 2 or 4 for the
923 * checksum to update.
924 *
32e7dc28 925 * A call to this helper is susceptible to change the underlying
ad4a5223
QM
926 * packet buffer. Therefore, at load time, all checks on pointers
927 * previously done by the verifier are invalidated and must be
928 * performed again, if the helper is used in combination with
929 * direct packet access.
930 * Return
931 * 0 on success, or a negative error in case of failure.
932 *
bdb7b79b 933 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
ad4a5223
QM
934 * Description
935 * This special helper is used to trigger a "tail call", or in
936 * other words, to jump into another eBPF program. The same stack
937 * frame is used (but values on stack and in registers for the
938 * caller are not accessible to the callee). This mechanism allows
939 * for program chaining, either for raising the maximum number of
940 * available eBPF instructions, or to execute given programs in
941 * conditional blocks. For security reasons, there is an upper
942 * limit to the number of successive tail calls that can be
943 * performed.
944 *
945 * Upon call of this helper, the program attempts to jump into a
946 * program referenced at index *index* in *prog_array_map*, a
947 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
948 * *ctx*, a pointer to the context.
949 *
950 * If the call succeeds, the kernel immediately runs the first
951 * instruction of the new program. This is not a function call,
952 * and it never returns to the previous program. If the call
953 * fails, then the helper has no effect, and the caller continues
954 * to run its subsequent instructions. A call can fail if the
955 * destination program for the jump does not exist (i.e. *index*
956 * is superior to the number of entries in *prog_array_map*), or
957 * if the maximum number of tail calls has been reached for this
958 * chain of programs. This limit is defined in the kernel by the
959 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
960 * which is currently set to 32.
961 * Return
962 * 0 on success, or a negative error in case of failure.
963 *
bdb7b79b 964 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
ad4a5223
QM
965 * Description
966 * Clone and redirect the packet associated to *skb* to another
967 * net device of index *ifindex*. Both ingress and egress
968 * interfaces can be used for redirection. The **BPF_F_INGRESS**
969 * value in *flags* is used to make the distinction (ingress path
970 * is selected if the flag is present, egress path otherwise).
971 * This is the only flag supported for now.
972 *
973 * In comparison with **bpf_redirect**\ () helper,
974 * **bpf_clone_redirect**\ () has the associated cost of
975 * duplicating the packet buffer, but this can be executed out of
976 * the eBPF program. Conversely, **bpf_redirect**\ () is more
977 * efficient, but it is handled through an action code where the
978 * redirection happens only after the eBPF program has returned.
979 *
32e7dc28 980 * A call to this helper is susceptible to change the underlying
ad4a5223
QM
981 * packet buffer. Therefore, at load time, all checks on pointers
982 * previously done by the verifier are invalidated and must be
983 * performed again, if the helper is used in combination with
984 * direct packet access.
985 * Return
986 * 0 on success, or a negative error in case of failure.
c456dec4
QM
987 *
988 * u64 bpf_get_current_pid_tgid(void)
989 * Return
990 * A 64-bit integer containing the current tgid and pid, and
991 * created as such:
992 * *current_task*\ **->tgid << 32 \|**
993 * *current_task*\ **->pid**.
994 *
995 * u64 bpf_get_current_uid_gid(void)
996 * Return
997 * A 64-bit integer containing the current GID and UID, and
998 * created as such: *current_gid* **<< 32 \|** *current_uid*.
999 *
bdb7b79b 1000 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
c456dec4
QM
1001 * Description
1002 * Copy the **comm** attribute of the current task into *buf* of
1003 * *size_of_buf*. The **comm** attribute contains the name of
1004 * the executable (excluding the path) for the current task. The
1005 * *size_of_buf* must be strictly positive. On success, the
1006 * helper makes sure that the *buf* is NUL-terminated. On failure,
1007 * it is filled with zeroes.
1008 * Return
1009 * 0 on success, or a negative error in case of failure.
1010 *
1fdd08be
QM
1011 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1012 * Description
1013 * Retrieve the classid for the current task, i.e. for the net_cls
1014 * cgroup to which *skb* belongs.
1015 *
1016 * This helper can be used on TC egress path, but not on ingress.
1017 *
1018 * The net_cls cgroup provides an interface to tag network packets
1019 * based on a user-provided identifier for all traffic coming from
1020 * the tasks belonging to the related cgroup. See also the related
1021 * kernel documentation, available from the Linux sources in file
da82c92f 1022 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1fdd08be
QM
1023 *
1024 * The Linux kernel has two versions for cgroups: there are
1025 * cgroups v1 and cgroups v2. Both are available to users, who can
1026 * use a mixture of them, but note that the net_cls cgroup is for
1027 * cgroup v1 only. This makes it incompatible with BPF programs
1028 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1029 * only hold data for one version of cgroups at a time).
1030 *
1031 * This helper is only available is the kernel was compiled with
1032 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1033 * "**y**" or to "**m**".
1034 * Return
1035 * The classid, or 0 for the default unconfigured classid.
1036 *
bdb7b79b 1037 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
c456dec4
QM
1038 * Description
1039 * Push a *vlan_tci* (VLAN tag control information) of protocol
1040 * *vlan_proto* to the packet associated to *skb*, then update
1041 * the checksum. Note that if *vlan_proto* is different from
1042 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1043 * be **ETH_P_8021Q**.
1044 *
32e7dc28 1045 * A call to this helper is susceptible to change the underlying
c456dec4
QM
1046 * packet buffer. Therefore, at load time, all checks on pointers
1047 * previously done by the verifier are invalidated and must be
1048 * performed again, if the helper is used in combination with
1049 * direct packet access.
1050 * Return
1051 * 0 on success, or a negative error in case of failure.
1052 *
bdb7b79b 1053 * long bpf_skb_vlan_pop(struct sk_buff *skb)
c456dec4
QM
1054 * Description
1055 * Pop a VLAN header from the packet associated to *skb*.
1056 *
32e7dc28 1057 * A call to this helper is susceptible to change the underlying
c456dec4
QM
1058 * packet buffer. Therefore, at load time, all checks on pointers
1059 * previously done by the verifier are invalidated and must be
1060 * performed again, if the helper is used in combination with
1061 * direct packet access.
1062 * Return
1063 * 0 on success, or a negative error in case of failure.
1064 *
bdb7b79b 1065 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
c456dec4
QM
1066 * Description
1067 * Get tunnel metadata. This helper takes a pointer *key* to an
1068 * empty **struct bpf_tunnel_key** of **size**, that will be
1069 * filled with tunnel metadata for the packet associated to *skb*.
1070 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1071 * indicates that the tunnel is based on IPv6 protocol instead of
1072 * IPv4.
1073 *
1074 * The **struct bpf_tunnel_key** is an object that generalizes the
1075 * principal parameters used by various tunneling protocols into a
1076 * single struct. This way, it can be used to easily make a
1077 * decision based on the contents of the encapsulation header,
1078 * "summarized" in this struct. In particular, it holds the IP
1079 * address of the remote end (IPv4 or IPv6, depending on the case)
1080 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1081 * this struct exposes the *key*\ **->tunnel_id**, which is
1082 * generally mapped to a VNI (Virtual Network Identifier), making
1083 * it programmable together with the **bpf_skb_set_tunnel_key**\
1084 * () helper.
1085 *
1086 * Let's imagine that the following code is part of a program
1087 * attached to the TC ingress interface, on one end of a GRE
1088 * tunnel, and is supposed to filter out all messages coming from
1089 * remote ends with IPv4 address other than 10.0.0.1:
1090 *
1091 * ::
1092 *
1093 * int ret;
1094 * struct bpf_tunnel_key key = {};
b16fc097 1095 *
c456dec4
QM
1096 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1097 * if (ret < 0)
1098 * return TC_ACT_SHOT; // drop packet
b16fc097 1099 *
c456dec4
QM
1100 * if (key.remote_ipv4 != 0x0a000001)
1101 * return TC_ACT_SHOT; // drop packet
b16fc097 1102 *
c456dec4
QM
1103 * return TC_ACT_OK; // accept packet
1104 *
1105 * This interface can also be used with all encapsulation devices
1106 * that can operate in "collect metadata" mode: instead of having
1107 * one network device per specific configuration, the "collect
1108 * metadata" mode only requires a single device where the
1109 * configuration can be extracted from this helper.
1110 *
1111 * This can be used together with various tunnels such as VXLan,
1112 * Geneve, GRE or IP in IP (IPIP).
1113 * Return
1114 * 0 on success, or a negative error in case of failure.
1115 *
bdb7b79b 1116 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
c456dec4
QM
1117 * Description
1118 * Populate tunnel metadata for packet associated to *skb.* The
1119 * tunnel metadata is set to the contents of *key*, of *size*. The
1120 * *flags* can be set to a combination of the following values:
1121 *
1122 * **BPF_F_TUNINFO_IPV6**
1123 * Indicate that the tunnel is based on IPv6 protocol
1124 * instead of IPv4.
1125 * **BPF_F_ZERO_CSUM_TX**
1126 * For IPv4 packets, add a flag to tunnel metadata
1127 * indicating that checksum computation should be skipped
1128 * and checksum set to zeroes.
1129 * **BPF_F_DONT_FRAGMENT**
1130 * Add a flag to tunnel metadata indicating that the
1131 * packet should not be fragmented.
1132 * **BPF_F_SEQ_NUMBER**
1133 * Add a flag to tunnel metadata indicating that a
1134 * sequence number should be added to tunnel header before
1135 * sending the packet. This flag was added for GRE
1136 * encapsulation, but might be used with other protocols
1137 * as well in the future.
1138 *
1139 * Here is a typical usage on the transmit path:
1140 *
1141 * ::
1142 *
1143 * struct bpf_tunnel_key key;
1144 * populate key ...
1145 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1146 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1147 *
1148 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1149 * helper for additional information.
1150 * Return
1151 * 0 on success, or a negative error in case of failure.
1152 *
c6b5fb86
QM
1153 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1154 * Description
1155 * Read the value of a perf event counter. This helper relies on a
1156 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1157 * the perf event counter is selected when *map* is updated with
1158 * perf event file descriptors. The *map* is an array whose size
1159 * is the number of available CPUs, and each cell contains a value
1160 * relative to one CPU. The value to retrieve is indicated by
1161 * *flags*, that contains the index of the CPU to look up, masked
1162 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1163 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1164 * current CPU should be retrieved.
1165 *
1166 * Note that before Linux 4.13, only hardware perf event can be
1167 * retrieved.
1168 *
1169 * Also, be aware that the newer helper
1170 * **bpf_perf_event_read_value**\ () is recommended over
3bd5a09b 1171 * **bpf_perf_event_read**\ () in general. The latter has some ABI
c6b5fb86
QM
1172 * quirks where error and counter value are used as a return code
1173 * (which is wrong to do since ranges may overlap). This issue is
3bd5a09b
QM
1174 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1175 * time provides more features over the **bpf_perf_event_read**\
1176 * () interface. Please refer to the description of
c6b5fb86
QM
1177 * **bpf_perf_event_read_value**\ () for details.
1178 * Return
1179 * The value of the perf event counter read from the map, or a
1180 * negative error code in case of failure.
1181 *
bdb7b79b 1182 * long bpf_redirect(u32 ifindex, u64 flags)
c456dec4
QM
1183 * Description
1184 * Redirect the packet to another net device of index *ifindex*.
1185 * This helper is somewhat similar to **bpf_clone_redirect**\
1186 * (), except that the packet is not cloned, which provides
1187 * increased performance.
1188 *
1189 * Except for XDP, both ingress and egress interfaces can be used
1190 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1191 * to make the distinction (ingress path is selected if the flag
1192 * is present, egress path otherwise). Currently, XDP only
1193 * supports redirection to the egress interface, and accepts no
1194 * flag at all.
1195 *
f25975f4
THJ
1196 * The same effect can also be attained with the more generic
1197 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1198 * redirect target instead of providing it directly to the helper.
c456dec4
QM
1199 * Return
1200 * For XDP, the helper returns **XDP_REDIRECT** on success or
1201 * **XDP_ABORTED** on error. For other program types, the values
1202 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1203 * error.
1204 *
1fdd08be
QM
1205 * u32 bpf_get_route_realm(struct sk_buff *skb)
1206 * Description
1207 * Retrieve the realm or the route, that is to say the
1208 * **tclassid** field of the destination for the *skb*. The
b16fc097 1209 * identifier retrieved is a user-provided tag, similar to the
1fdd08be
QM
1210 * one used with the net_cls cgroup (see description for
1211 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1212 * held by a route (a destination entry), not by a task.
1213 *
1214 * Retrieving this identifier works with the clsact TC egress hook
1215 * (see also **tc-bpf(8)**), or alternatively on conventional
1216 * classful egress qdiscs, but not on TC ingress path. In case of
1217 * clsact TC egress hook, this has the advantage that, internally,
1218 * the destination entry has not been dropped yet in the transmit
1219 * path. Therefore, the destination entry does not need to be
1220 * artificially held via **netif_keep_dst**\ () for a classful
1221 * qdisc until the *skb* is freed.
1222 *
1223 * This helper is available only if the kernel was compiled with
1224 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1225 * Return
1226 * The realm of the route for the packet associated to *skb*, or 0
1227 * if none was found.
1228 *
bdb7b79b 1229 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
c456dec4
QM
1230 * Description
1231 * Write raw *data* blob into a special BPF perf event held by
1232 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1233 * event must have the following attributes: **PERF_SAMPLE_RAW**
1234 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1235 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1236 *
1237 * The *flags* are used to indicate the index in *map* for which
1238 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1239 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1240 * to indicate that the index of the current CPU core should be
1241 * used.
1242 *
1243 * The value to write, of *size*, is passed through eBPF stack and
1244 * pointed by *data*.
1245 *
1246 * The context of the program *ctx* needs also be passed to the
1247 * helper.
1248 *
1249 * On user space, a program willing to read the values needs to
1250 * call **perf_event_open**\ () on the perf event (either for
1251 * one or for all CPUs) and to store the file descriptor into the
1252 * *map*. This must be done before the eBPF program can send data
1253 * into it. An example is available in file
1254 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1255 * tree (the eBPF program counterpart is in
1256 * *samples/bpf/trace_output_kern.c*).
1257 *
1258 * **bpf_perf_event_output**\ () achieves better performance
1259 * than **bpf_trace_printk**\ () for sharing data with user
1260 * space, and is much better suitable for streaming data from eBPF
1261 * programs.
1262 *
1263 * Note that this helper is not restricted to tracing use cases
1264 * and can be used with programs attached to TC or XDP as well,
1265 * where it allows for passing data to user space listeners. Data
1266 * can be:
1267 *
1268 * * Only custom structs,
1269 * * Only the packet payload, or
1270 * * A combination of both.
1271 * Return
1272 * 0 on success, or a negative error in case of failure.
1273 *
bdb7b79b 1274 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
1fdd08be
QM
1275 * Description
1276 * This helper was provided as an easy way to load data from a
1277 * packet. It can be used to load *len* bytes from *offset* from
1278 * the packet associated to *skb*, into the buffer pointed by
1279 * *to*.
1280 *
1281 * Since Linux 4.7, usage of this helper has mostly been replaced
1282 * by "direct packet access", enabling packet data to be
1283 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1284 * pointing respectively to the first byte of packet data and to
1285 * the byte after the last byte of packet data. However, it
1286 * remains useful if one wishes to read large quantities of data
1287 * at once from a packet into the eBPF stack.
1288 * Return
1289 * 0 on success, or a negative error in case of failure.
1290 *
bdb7b79b 1291 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
c456dec4
QM
1292 * Description
1293 * Walk a user or a kernel stack and return its id. To achieve
1294 * this, the helper needs *ctx*, which is a pointer to the context
1295 * on which the tracing program is executed, and a pointer to a
1296 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1297 *
1298 * The last argument, *flags*, holds the number of stack frames to
1299 * skip (from 0 to 255), masked with
1300 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1301 * a combination of the following flags:
1302 *
1303 * **BPF_F_USER_STACK**
1304 * Collect a user space stack instead of a kernel stack.
1305 * **BPF_F_FAST_STACK_CMP**
1306 * Compare stacks by hash only.
1307 * **BPF_F_REUSE_STACKID**
1308 * If two different stacks hash into the same *stackid*,
1309 * discard the old one.
1310 *
1311 * The stack id retrieved is a 32 bit long integer handle which
1312 * can be further combined with other data (including other stack
1313 * ids) and used as a key into maps. This can be useful for
1314 * generating a variety of graphs (such as flame graphs or off-cpu
1315 * graphs).
1316 *
1317 * For walking a stack, this helper is an improvement over
1318 * **bpf_probe_read**\ (), which can be used with unrolled loops
1319 * but is not efficient and consumes a lot of eBPF instructions.
1320 * Instead, **bpf_get_stackid**\ () can collect up to
1321 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1322 * this limit can be controlled with the **sysctl** program, and
1323 * that it should be manually increased in order to profile long
1324 * user stacks (such as stacks for Java programs). To do so, use:
1325 *
1326 * ::
1327 *
1328 * # sysctl kernel.perf_event_max_stack=<new value>
c456dec4
QM
1329 * Return
1330 * The positive or null stack id on success, or a negative error
1331 * in case of failure.
1332 *
1fdd08be
QM
1333 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1334 * Description
1335 * Compute a checksum difference, from the raw buffer pointed by
1336 * *from*, of length *from_size* (that must be a multiple of 4),
1337 * towards the raw buffer pointed by *to*, of size *to_size*
1338 * (same remark). An optional *seed* can be added to the value
1339 * (this can be cascaded, the seed may come from a previous call
1340 * to the helper).
1341 *
1342 * This is flexible enough to be used in several ways:
1343 *
1344 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1345 * checksum, it can be used when pushing new data.
1346 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1347 * checksum, it can be used when removing data from a packet.
1348 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1349 * can be used to compute a diff. Note that *from_size* and
1350 * *to_size* do not need to be equal.
1351 *
1352 * This helper can be used in combination with
1353 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1354 * which one can feed in the difference computed with
1355 * **bpf_csum_diff**\ ().
1356 * Return
1357 * The checksum result, or a negative error code in case of
1358 * failure.
1359 *
bdb7b79b 1360 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1fdd08be
QM
1361 * Description
1362 * Retrieve tunnel options metadata for the packet associated to
1363 * *skb*, and store the raw tunnel option data to the buffer *opt*
1364 * of *size*.
1365 *
1366 * This helper can be used with encapsulation devices that can
1367 * operate in "collect metadata" mode (please refer to the related
1368 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1369 * more details). A particular example where this can be used is
1370 * in combination with the Geneve encapsulation protocol, where it
1371 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1372 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1373 * the eBPF program. This allows for full customization of these
1374 * headers.
1375 * Return
1376 * The size of the option data retrieved.
1377 *
bdb7b79b 1378 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
1fdd08be
QM
1379 * Description
1380 * Set tunnel options metadata for the packet associated to *skb*
1381 * to the option data contained in the raw buffer *opt* of *size*.
1382 *
1383 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1384 * helper for additional information.
1385 * Return
1386 * 0 on success, or a negative error in case of failure.
1387 *
bdb7b79b 1388 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1fdd08be
QM
1389 * Description
1390 * Change the protocol of the *skb* to *proto*. Currently
1391 * supported are transition from IPv4 to IPv6, and from IPv6 to
1392 * IPv4. The helper takes care of the groundwork for the
1393 * transition, including resizing the socket buffer. The eBPF
1394 * program is expected to fill the new headers, if any, via
1395 * **skb_store_bytes**\ () and to recompute the checksums with
1396 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1397 * (). The main case for this helper is to perform NAT64
1398 * operations out of an eBPF program.
1399 *
1400 * Internally, the GSO type is marked as dodgy so that headers are
1401 * checked and segments are recalculated by the GSO/GRO engine.
1402 * The size for GSO target is adapted as well.
1403 *
1404 * All values for *flags* are reserved for future usage, and must
1405 * be left at zero.
1406 *
32e7dc28 1407 * A call to this helper is susceptible to change the underlying
1fdd08be
QM
1408 * packet buffer. Therefore, at load time, all checks on pointers
1409 * previously done by the verifier are invalidated and must be
1410 * performed again, if the helper is used in combination with
1411 * direct packet access.
1412 * Return
1413 * 0 on success, or a negative error in case of failure.
1414 *
bdb7b79b 1415 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
1fdd08be
QM
1416 * Description
1417 * Change the packet type for the packet associated to *skb*. This
1418 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1419 * the eBPF program does not have a write access to *skb*\
1420 * **->pkt_type** beside this helper. Using a helper here allows
1421 * for graceful handling of errors.
1422 *
1423 * The major use case is to change incoming *skb*s to
1424 * **PACKET_HOST** in a programmatic way instead of having to
1425 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1426 * example.
1427 *
1428 * Note that *type* only allows certain values. At this time, they
1429 * are:
1430 *
1431 * **PACKET_HOST**
1432 * Packet is for us.
1433 * **PACKET_BROADCAST**
1434 * Send packet to all.
1435 * **PACKET_MULTICAST**
1436 * Send packet to group.
1437 * **PACKET_OTHERHOST**
1438 * Send packet to someone else.
1439 * Return
1440 * 0 on success, or a negative error in case of failure.
1441 *
bdb7b79b 1442 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
c6b5fb86
QM
1443 * Description
1444 * Check whether *skb* is a descendant of the cgroup2 held by
1445 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1446 * Return
1447 * The return value depends on the result of the test, and can be:
1448 *
1449 * * 0, if the *skb* failed the cgroup2 descendant test.
1450 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1451 * * A negative error code, if an error occurred.
1452 *
fa15601a
QM
1453 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1454 * Description
1455 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1456 * not set, in particular if the hash was cleared due to mangling,
1457 * recompute this hash. Later accesses to the hash can be done
1458 * directly with *skb*\ **->hash**.
1459 *
1460 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1461 * prototype with **bpf_skb_change_proto**\ (), or calling
1462 * **bpf_skb_store_bytes**\ () with the
1463 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1464 * the hash and to trigger a new computation for the next call to
1465 * **bpf_get_hash_recalc**\ ().
1466 * Return
1467 * The 32-bit hash.
1468 *
c456dec4
QM
1469 * u64 bpf_get_current_task(void)
1470 * Return
1471 * A pointer to the current task struct.
fa15601a 1472 *
bdb7b79b 1473 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
c6b5fb86
QM
1474 * Description
1475 * Attempt in a safe way to write *len* bytes from the buffer
1476 * *src* to *dst* in memory. It only works for threads that are in
1477 * user context, and *dst* must be a valid user space address.
1478 *
1479 * This helper should not be used to implement any kind of
1480 * security mechanism because of TOC-TOU attacks, but rather to
1481 * debug, divert, and manipulate execution of semi-cooperative
1482 * processes.
1483 *
1484 * Keep in mind that this feature is meant for experiments, and it
1485 * has a risk of crashing the system and running programs.
1486 * Therefore, when an eBPF program using this helper is attached,
1487 * a warning including PID and process name is printed to kernel
1488 * logs.
1489 * Return
1490 * 0 on success, or a negative error in case of failure.
1491 *
bdb7b79b 1492 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
c6b5fb86
QM
1493 * Description
1494 * Check whether the probe is being run is the context of a given
1495 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1496 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1497 * Return
1498 * The return value depends on the result of the test, and can be:
1499 *
1aef5b43
SL
1500 * * 0, if current task belongs to the cgroup2.
1501 * * 1, if current task does not belong to the cgroup2.
c6b5fb86
QM
1502 * * A negative error code, if an error occurred.
1503 *
bdb7b79b 1504 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
fa15601a
QM
1505 * Description
1506 * Resize (trim or grow) the packet associated to *skb* to the
1507 * new *len*. The *flags* are reserved for future usage, and must
1508 * be left at zero.
1509 *
1510 * The basic idea is that the helper performs the needed work to
1511 * change the size of the packet, then the eBPF program rewrites
1512 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1513 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1514 * and others. This helper is a slow path utility intended for
1515 * replies with control messages. And because it is targeted for
1516 * slow path, the helper itself can afford to be slow: it
1517 * implicitly linearizes, unclones and drops offloads from the
1518 * *skb*.
1519 *
32e7dc28 1520 * A call to this helper is susceptible to change the underlying
fa15601a
QM
1521 * packet buffer. Therefore, at load time, all checks on pointers
1522 * previously done by the verifier are invalidated and must be
1523 * performed again, if the helper is used in combination with
1524 * direct packet access.
1525 * Return
1526 * 0 on success, or a negative error in case of failure.
1527 *
bdb7b79b 1528 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
fa15601a
QM
1529 * Description
1530 * Pull in non-linear data in case the *skb* is non-linear and not
1531 * all of *len* are part of the linear section. Make *len* bytes
1532 * from *skb* readable and writable. If a zero value is passed for
1533 * *len*, then the whole length of the *skb* is pulled.
1534 *
1535 * This helper is only needed for reading and writing with direct
1536 * packet access.
1537 *
1538 * For direct packet access, testing that offsets to access
1539 * are within packet boundaries (test on *skb*\ **->data_end**) is
1540 * susceptible to fail if offsets are invalid, or if the requested
1541 * data is in non-linear parts of the *skb*. On failure the
1542 * program can just bail out, or in the case of a non-linear
1543 * buffer, use a helper to make the data available. The
1544 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1545 * the data. Another one consists in using **bpf_skb_pull_data**
1546 * to pull in once the non-linear parts, then retesting and
1547 * eventually access the data.
1548 *
1549 * At the same time, this also makes sure the *skb* is uncloned,
1550 * which is a necessary condition for direct write. As this needs
1551 * to be an invariant for the write part only, the verifier
1552 * detects writes and adds a prologue that is calling
1553 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1554 * the very beginning in case it is indeed cloned.
1555 *
32e7dc28 1556 * A call to this helper is susceptible to change the underlying
fa15601a
QM
1557 * packet buffer. Therefore, at load time, all checks on pointers
1558 * previously done by the verifier are invalidated and must be
1559 * performed again, if the helper is used in combination with
1560 * direct packet access.
1561 * Return
1562 * 0 on success, or a negative error in case of failure.
1563 *
1564 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1565 * Description
1566 * Add the checksum *csum* into *skb*\ **->csum** in case the
1567 * driver has supplied a checksum for the entire packet into that
1568 * field. Return an error otherwise. This helper is intended to be
1569 * used in combination with **bpf_csum_diff**\ (), in particular
1570 * when the checksum needs to be updated after data has been
1571 * written into the packet through direct packet access.
1572 * Return
1573 * The checksum on success, or a negative error code in case of
1574 * failure.
1575 *
1576 * void bpf_set_hash_invalid(struct sk_buff *skb)
1577 * Description
1578 * Invalidate the current *skb*\ **->hash**. It can be used after
1579 * mangling on headers through direct packet access, in order to
1580 * indicate that the hash is outdated and to trigger a
1581 * recalculation the next time the kernel tries to access this
1582 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1583 *
bdb7b79b 1584 * long bpf_get_numa_node_id(void)
fa15601a
QM
1585 * Description
1586 * Return the id of the current NUMA node. The primary use case
1587 * for this helper is the selection of sockets for the local NUMA
1588 * node, when the program is attached to sockets using the
1589 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1590 * but the helper is also available to other eBPF program types,
1591 * similarly to **bpf_get_smp_processor_id**\ ().
1592 * Return
1593 * The id of current NUMA node.
1594 *
bdb7b79b 1595 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
c6b5fb86
QM
1596 * Description
1597 * Grows headroom of packet associated to *skb* and adjusts the
1598 * offset of the MAC header accordingly, adding *len* bytes of
1599 * space. It automatically extends and reallocates memory as
1600 * required.
1601 *
1602 * This helper can be used on a layer 3 *skb* to push a MAC header
1603 * for redirection into a layer 2 device.
1604 *
1605 * All values for *flags* are reserved for future usage, and must
1606 * be left at zero.
1607 *
32e7dc28 1608 * A call to this helper is susceptible to change the underlying
c6b5fb86
QM
1609 * packet buffer. Therefore, at load time, all checks on pointers
1610 * previously done by the verifier are invalidated and must be
1611 * performed again, if the helper is used in combination with
1612 * direct packet access.
1613 * Return
1614 * 0 on success, or a negative error in case of failure.
1615 *
bdb7b79b 1616 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
c6b5fb86
QM
1617 * Description
1618 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1619 * it is possible to use a negative value for *delta*. This helper
1620 * can be used to prepare the packet for pushing or popping
1621 * headers.
1622 *
32e7dc28 1623 * A call to this helper is susceptible to change the underlying
c6b5fb86
QM
1624 * packet buffer. Therefore, at load time, all checks on pointers
1625 * previously done by the verifier are invalidated and must be
1626 * performed again, if the helper is used in combination with
1627 * direct packet access.
1628 * Return
1629 * 0 on success, or a negative error in case of failure.
1630 *
bdb7b79b 1631 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
c6b5fb86 1632 * Description
6ae08ae3 1633 * Copy a NUL terminated string from an unsafe kernel address
ab8d7809 1634 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
6ae08ae3 1635 * more details.
c6b5fb86 1636 *
ab8d7809
QM
1637 * Generally, use **bpf_probe_read_user_str**\ () or
1638 * **bpf_probe_read_kernel_str**\ () instead.
c6b5fb86
QM
1639 * Return
1640 * On success, the strictly positive length of the string,
1641 * including the trailing NUL character. On error, a negative
1642 * value.
1643 *
1644 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1645 * Description
1646 * If the **struct sk_buff** pointed by *skb* has a known socket,
1647 * retrieve the cookie (generated by the kernel) of this socket.
1648 * If no cookie has been set yet, generate a new cookie. Once
1649 * generated, the socket cookie remains stable for the life of the
1650 * socket. This helper can be useful for monitoring per socket
cd48bdda
DB
1651 * networking traffic statistics as it provides a global socket
1652 * identifier that can be assumed unique.
c6b5fb86
QM
1653 * Return
1654 * A 8-byte long non-decreasing number on success, or 0 if the
1655 * socket field is missing inside *skb*.
1656 *
d692f113
AI
1657 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1658 * Description
1659 * Equivalent to bpf_get_socket_cookie() helper that accepts
62369db2 1660 * *skb*, but gets socket from **struct bpf_sock_addr** context.
d692f113
AI
1661 * Return
1662 * A 8-byte long non-decreasing number.
1663 *
1664 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1665 * Description
ab8d7809 1666 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
62369db2 1667 * *skb*, but gets socket from **struct bpf_sock_ops** context.
d692f113
AI
1668 * Return
1669 * A 8-byte long non-decreasing number.
1670 *
c6b5fb86
QM
1671 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1672 * Return
1673 * The owner UID of the socket associated to *skb*. If the socket
1674 * is **NULL**, or if it is not a full socket (i.e. if it is a
1675 * time-wait or a request socket instead), **overflowuid** value
1676 * is returned (note that **overflowuid** might also be the actual
1677 * UID value for the socket).
1678 *
bdb7b79b 1679 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
fa15601a
QM
1680 * Description
1681 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1682 * to value *hash*.
1683 * Return
1684 * 0
1685 *
bdb7b79b 1686 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
7aa79a86
QM
1687 * Description
1688 * Emulate a call to **setsockopt()** on the socket associated to
1689 * *bpf_socket*, which must be a full socket. The *level* at
1690 * which the option resides and the name *optname* of the option
1691 * must be specified, see **setsockopt(2)** for more information.
1692 * The option value of length *optlen* is pointed by *optval*.
1693 *
beecf11b 1694 * *bpf_socket* should be one of the following:
ab8d7809 1695 *
beecf11b
SF
1696 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1697 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1698 * and **BPF_CGROUP_INET6_CONNECT**.
1699 *
7aa79a86
QM
1700 * This helper actually implements a subset of **setsockopt()**.
1701 * It supports the following *level*\ s:
1702 *
1703 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1704 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
f9bcf968
DY
1705 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
1706 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
7aa79a86
QM
1707 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1708 * **TCP_CONGESTION**, **TCP_BPF_IW**,
f9bcf968
DY
1709 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
1710 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
eca43ee6 1711 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
7aa79a86
QM
1712 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1713 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1714 * Return
1715 * 0 on success, or a negative error in case of failure.
1716 *
bdb7b79b 1717 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
fa15601a
QM
1718 * Description
1719 * Grow or shrink the room for data in the packet associated to
1720 * *skb* by *len_diff*, and according to the selected *mode*.
1721 *
836e66c2
DB
1722 * By default, the helper will reset any offloaded checksum
1723 * indicator of the skb to CHECKSUM_NONE. This can be avoided
1724 * by the following flag:
1725 *
1726 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
1727 * checksum data of the skb to CHECKSUM_NONE.
1728 *
14aa3192
WB
1729 * There are two supported modes at this time:
1730 *
1731 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1732 * (room space is added or removed below the layer 2 header).
fa15601a
QM
1733 *
1734 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1735 * (room space is added or removed below the layer 3 header).
1736 *
868d5235 1737 * The following flags are supported at this time:
2278f6cc
WB
1738 *
1739 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1740 * Adjusting mss in this way is not allowed for datagrams.
fa15601a 1741 *
80867c5e
QM
1742 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1743 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
868d5235
WB
1744 * Any new space is reserved to hold a tunnel header.
1745 * Configure skb offsets and other fields accordingly.
1746 *
80867c5e
QM
1747 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1748 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
868d5235
WB
1749 * Use with ENCAP_L3 flags to further specify the tunnel type.
1750 *
80867c5e 1751 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
58dfc900 1752 * Use with ENCAP_L3/L4 flags to further specify the tunnel
80867c5e 1753 * type; *len* is the length of the inner MAC header.
58dfc900 1754 *
32e7dc28 1755 * A call to this helper is susceptible to change the underlying
fa15601a
QM
1756 * packet buffer. Therefore, at load time, all checks on pointers
1757 * previously done by the verifier are invalidated and must be
1758 * performed again, if the helper is used in combination with
1759 * direct packet access.
1760 * Return
1761 * 0 on success, or a negative error in case of failure.
1762 *
bdb7b79b 1763 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
ab127040
QM
1764 * Description
1765 * Redirect the packet to the endpoint referenced by *map* at
1766 * index *key*. Depending on its type, this *map* can contain
1767 * references to net devices (for forwarding packets through other
1768 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1769 * but this is only implemented for native XDP (with driver
1770 * support) as of this writing).
1771 *
43e74c02
THJ
1772 * The lower two bits of *flags* are used as the return code if
1773 * the map lookup fails. This is so that the return value can be
ab8d7809
QM
1774 * one of the XDP program return codes up to **XDP_TX**, as chosen
1775 * by the caller. Any higher bits in the *flags* argument must be
43e74c02 1776 * unset.
ab127040 1777 *
ab8d7809
QM
1778 * See also **bpf_redirect**\ (), which only supports redirecting
1779 * to an ifindex, but doesn't require a map to do so.
ab127040 1780 * Return
f25975f4 1781 * **XDP_REDIRECT** on success, or the value of the two lower bits
a33d3147 1782 * of the *flags* argument on error.
ab127040 1783 *
bdb7b79b 1784 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
ab127040
QM
1785 * Description
1786 * Redirect the packet to the socket referenced by *map* (of type
1787 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1788 * egress interfaces can be used for redirection. The
1789 * **BPF_F_INGRESS** value in *flags* is used to make the
1790 * distinction (ingress path is selected if the flag is present,
1791 * egress path otherwise). This is the only flag supported for now.
1792 * Return
1793 * **SK_PASS** on success, or **SK_DROP** on error.
1794 *
bdb7b79b 1795 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
ab127040
QM
1796 * Description
1797 * Add an entry to, or update a *map* referencing sockets. The
1798 * *skops* is used as a new value for the entry associated to
1799 * *key*. *flags* is one of:
1800 *
1801 * **BPF_NOEXIST**
1802 * The entry for *key* must not exist in the map.
1803 * **BPF_EXIST**
1804 * The entry for *key* must already exist in the map.
1805 * **BPF_ANY**
1806 * No condition on the existence of the entry for *key*.
1807 *
1808 * If the *map* has eBPF programs (parser and verdict), those will
1809 * be inherited by the socket being added. If the socket is
1810 * already attached to eBPF programs, this results in an error.
1811 * Return
1812 * 0 on success, or a negative error in case of failure.
1813 *
bdb7b79b 1814 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
fa15601a
QM
1815 * Description
1816 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1817 * *delta* (which can be positive or negative). Note that this
1818 * operation modifies the address stored in *xdp_md*\ **->data**,
1819 * so the latter must be loaded only after the helper has been
1820 * called.
1821 *
1822 * The use of *xdp_md*\ **->data_meta** is optional and programs
1823 * are not required to use it. The rationale is that when the
1824 * packet is processed with XDP (e.g. as DoS filter), it is
1825 * possible to push further meta data along with it before passing
1826 * to the stack, and to give the guarantee that an ingress eBPF
1827 * program attached as a TC classifier on the same device can pick
1828 * this up for further post-processing. Since TC works with socket
1829 * buffers, it remains possible to set from XDP the **mark** or
1830 * **priority** pointers, or other pointers for the socket buffer.
1831 * Having this scratch space generic and programmable allows for
1832 * more flexibility as the user is free to store whatever meta
1833 * data they need.
1834 *
32e7dc28 1835 * A call to this helper is susceptible to change the underlying
fa15601a
QM
1836 * packet buffer. Therefore, at load time, all checks on pointers
1837 * previously done by the verifier are invalidated and must be
1838 * performed again, if the helper is used in combination with
1839 * direct packet access.
1840 * Return
1841 * 0 on success, or a negative error in case of failure.
7aa79a86 1842 *
bdb7b79b 1843 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
7aa79a86
QM
1844 * Description
1845 * Read the value of a perf event counter, and store it into *buf*
1846 * of size *buf_size*. This helper relies on a *map* of type
1847 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1848 * counter is selected when *map* is updated with perf event file
1849 * descriptors. The *map* is an array whose size is the number of
1850 * available CPUs, and each cell contains a value relative to one
1851 * CPU. The value to retrieve is indicated by *flags*, that
1852 * contains the index of the CPU to look up, masked with
1853 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1854 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1855 * current CPU should be retrieved.
1856 *
1857 * This helper behaves in a way close to
1858 * **bpf_perf_event_read**\ () helper, save that instead of
1859 * just returning the value observed, it fills the *buf*
1860 * structure. This allows for additional data to be retrieved: in
1861 * particular, the enabled and running times (in *buf*\
1862 * **->enabled** and *buf*\ **->running**, respectively) are
1863 * copied. In general, **bpf_perf_event_read_value**\ () is
1864 * recommended over **bpf_perf_event_read**\ (), which has some
1865 * ABI issues and provides fewer functionalities.
1866 *
1867 * These values are interesting, because hardware PMU (Performance
1868 * Monitoring Unit) counters are limited resources. When there are
1869 * more PMU based perf events opened than available counters,
1870 * kernel will multiplex these events so each event gets certain
1871 * percentage (but not all) of the PMU time. In case that
1872 * multiplexing happens, the number of samples or counter value
1873 * will not reflect the case compared to when no multiplexing
1874 * occurs. This makes comparison between different runs difficult.
1875 * Typically, the counter value should be normalized before
1876 * comparing to other experiments. The usual normalization is done
1877 * as follows.
1878 *
1879 * ::
1880 *
1881 * normalized_counter = counter * t_enabled / t_running
1882 *
1883 * Where t_enabled is the time enabled for event and t_running is
1884 * the time running for event since last normalization. The
1885 * enabled and running times are accumulated since the perf event
1886 * open. To achieve scaling factor between two invocations of an
ab8d7809 1887 * eBPF program, users can use CPU id as the key (which is
7aa79a86
QM
1888 * typical for perf array usage model) to remember the previous
1889 * value and do the calculation inside the eBPF program.
1890 * Return
1891 * 0 on success, or a negative error in case of failure.
1892 *
bdb7b79b 1893 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
7aa79a86
QM
1894 * Description
1895 * For en eBPF program attached to a perf event, retrieve the
1896 * value of the event counter associated to *ctx* and store it in
1897 * the structure pointed by *buf* and of size *buf_size*. Enabled
1898 * and running times are also stored in the structure (see
1899 * description of helper **bpf_perf_event_read_value**\ () for
1900 * more details).
1901 * Return
1902 * 0 on success, or a negative error in case of failure.
1903 *
bdb7b79b 1904 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
7aa79a86
QM
1905 * Description
1906 * Emulate a call to **getsockopt()** on the socket associated to
1907 * *bpf_socket*, which must be a full socket. The *level* at
1908 * which the option resides and the name *optname* of the option
1909 * must be specified, see **getsockopt(2)** for more information.
1910 * The retrieved value is stored in the structure pointed by
1911 * *opval* and of length *optlen*.
1912 *
beecf11b 1913 * *bpf_socket* should be one of the following:
ab8d7809 1914 *
beecf11b
SF
1915 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1916 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1917 * and **BPF_CGROUP_INET6_CONNECT**.
1918 *
7aa79a86
QM
1919 * This helper actually implements a subset of **getsockopt()**.
1920 * It supports the following *level*\ s:
1921 *
1922 * * **IPPROTO_TCP**, which supports *optname*
1923 * **TCP_CONGESTION**.
1924 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1925 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1926 * Return
1927 * 0 on success, or a negative error in case of failure.
1928 *
bdb7b79b 1929 * long bpf_override_return(struct pt_regs *regs, u64 rc)
7aa79a86
QM
1930 * Description
1931 * Used for error injection, this helper uses kprobes to override
1932 * the return value of the probed function, and to set it to *rc*.
1933 * The first argument is the context *regs* on which the kprobe
1934 * works.
1935 *
ab8d7809 1936 * This helper works by setting the PC (program counter)
7aa79a86
QM
1937 * to an override function which is run in place of the original
1938 * probed function. This means the probed function is not run at
1939 * all. The replacement function just returns with the required
1940 * value.
1941 *
1942 * This helper has security implications, and thus is subject to
1943 * restrictions. It is only available if the kernel was compiled
1944 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1945 * option, and in this case it only works on functions tagged with
1946 * **ALLOW_ERROR_INJECTION** in the kernel code.
1947 *
1948 * Also, the helper is only available for the architectures having
1949 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1950 * x86 architecture is the only one to support this feature.
1951 * Return
1952 * 0
1953 *
bdb7b79b 1954 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
7aa79a86
QM
1955 * Description
1956 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1957 * for the full TCP socket associated to *bpf_sock_ops* to
1958 * *argval*.
1959 *
1960 * The primary use of this field is to determine if there should
1961 * be calls to eBPF programs of type
1962 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1963 * code. A program of the same type can change its value, per
1964 * connection and as necessary, when the connection is
1965 * established. This field is directly accessible for reading, but
1966 * this helper must be used for updates in order to return an
1967 * error if an eBPF program tries to set a callback that is not
1968 * supported in the current kernel.
1969 *
725721a6 1970 * *argval* is a flag array which can combine these flags:
7aa79a86
QM
1971 *
1972 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1973 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1974 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
23729ff2 1975 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
7aa79a86 1976 *
725721a6
VHT
1977 * Therefore, this function can be used to clear a callback flag by
1978 * setting the appropriate bit to zero. e.g. to disable the RTO
1979 * callback:
1980 *
1981 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1982 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1983 *
7aa79a86
QM
1984 * Here are some examples of where one could call such eBPF
1985 * program:
1986 *
1987 * * When RTO fires.
1988 * * When a packet is retransmitted.
1989 * * When the connection terminates.
1990 * * When a packet is sent.
1991 * * When a packet is received.
1992 * Return
1993 * Code **-EINVAL** if the socket is not a full TCP socket;
1994 * otherwise, a positive number containing the bits that could not
1995 * be set is returned (which comes down to 0 if all bits were set
1996 * as required).
1997 *
bdb7b79b 1998 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
ab127040
QM
1999 * Description
2000 * This helper is used in programs implementing policies at the
2001 * socket level. If the message *msg* is allowed to pass (i.e. if
2002 * the verdict eBPF program returns **SK_PASS**), redirect it to
2003 * the socket referenced by *map* (of type
2004 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2005 * egress interfaces can be used for redirection. The
2006 * **BPF_F_INGRESS** value in *flags* is used to make the
2007 * distinction (ingress path is selected if the flag is present,
2008 * egress path otherwise). This is the only flag supported for now.
2009 * Return
2010 * **SK_PASS** on success, or **SK_DROP** on error.
2011 *
bdb7b79b 2012 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
ab127040
QM
2013 * Description
2014 * For socket policies, apply the verdict of the eBPF program to
2015 * the next *bytes* (number of bytes) of message *msg*.
2016 *
2017 * For example, this helper can be used in the following cases:
2018 *
2019 * * A single **sendmsg**\ () or **sendfile**\ () system call
2020 * contains multiple logical messages that the eBPF program is
2021 * supposed to read and for which it should apply a verdict.
2022 * * An eBPF program only cares to read the first *bytes* of a
2023 * *msg*. If the message has a large payload, then setting up
2024 * and calling the eBPF program repeatedly for all bytes, even
2025 * though the verdict is already known, would create unnecessary
2026 * overhead.
2027 *
2028 * When called from within an eBPF program, the helper sets a
2029 * counter internal to the BPF infrastructure, that is used to
2030 * apply the last verdict to the next *bytes*. If *bytes* is
2031 * smaller than the current data being processed from a
2032 * **sendmsg**\ () or **sendfile**\ () system call, the first
2033 * *bytes* will be sent and the eBPF program will be re-run with
2034 * the pointer for start of data pointing to byte number *bytes*
2035 * **+ 1**. If *bytes* is larger than the current data being
2036 * processed, then the eBPF verdict will be applied to multiple
2037 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2038 * consumed.
2039 *
2040 * Note that if a socket closes with the internal counter holding
2041 * a non-zero value, this is not a problem because data is not
2042 * being buffered for *bytes* and is sent as it is received.
2043 * Return
2044 * 0
2045 *
bdb7b79b 2046 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
ab127040
QM
2047 * Description
2048 * For socket policies, prevent the execution of the verdict eBPF
2049 * program for message *msg* until *bytes* (byte number) have been
2050 * accumulated.
2051 *
2052 * This can be used when one needs a specific number of bytes
2053 * before a verdict can be assigned, even if the data spans
2054 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2055 * case would be a user calling **sendmsg**\ () repeatedly with
2056 * 1-byte long message segments. Obviously, this is bad for
2057 * performance, but it is still valid. If the eBPF program needs
2058 * *bytes* bytes to validate a header, this helper can be used to
2059 * prevent the eBPF program to be called again until *bytes* have
2060 * been accumulated.
2061 * Return
2062 * 0
2063 *
bdb7b79b 2064 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
ab127040
QM
2065 * Description
2066 * For socket policies, pull in non-linear data from user space
2067 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2068 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2069 * respectively.
2070 *
2071 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2072 * *msg* it can only parse data that the (**data**, **data_end**)
2073 * pointers have already consumed. For **sendmsg**\ () hooks this
2074 * is likely the first scatterlist element. But for calls relying
2075 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2076 * be the range (**0**, **0**) because the data is shared with
2077 * user space and by default the objective is to avoid allowing
2078 * user space to modify data while (or after) eBPF verdict is
2079 * being decided. This helper can be used to pull in data and to
2080 * set the start and end pointer to given values. Data will be
2081 * copied if necessary (i.e. if data was not linear and if start
2082 * and end pointers do not point to the same chunk).
2083 *
32e7dc28 2084 * A call to this helper is susceptible to change the underlying
ab127040
QM
2085 * packet buffer. Therefore, at load time, all checks on pointers
2086 * previously done by the verifier are invalidated and must be
2087 * performed again, if the helper is used in combination with
2088 * direct packet access.
2089 *
2090 * All values for *flags* are reserved for future usage, and must
2091 * be left at zero.
2092 * Return
2093 * 0 on success, or a negative error in case of failure.
2094 *
bdb7b79b 2095 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
7aa79a86
QM
2096 * Description
2097 * Bind the socket associated to *ctx* to the address pointed by
2098 * *addr*, of length *addr_len*. This allows for making outgoing
2099 * connection from the desired IP address, which can be useful for
2100 * example when all processes inside a cgroup should use one
2101 * single IP address on a host that has multiple IP configured.
2102 *
2103 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2104 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
8086fbaf
SF
2105 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2106 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2107 * behavior and lets the kernel efficiently pick up an unused
2108 * port as long as 4-tuple is unique. Passing non-zero port might
2109 * lead to degraded performance.
7aa79a86
QM
2110 * Return
2111 * 0 on success, or a negative error in case of failure.
2d020dd7 2112 *
bdb7b79b 2113 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2d020dd7
QM
2114 * Description
2115 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
c8741e2b
JDB
2116 * possible to both shrink and grow the packet tail.
2117 * Shrink done via *delta* being a negative integer.
2d020dd7 2118 *
32e7dc28 2119 * A call to this helper is susceptible to change the underlying
2d020dd7
QM
2120 * packet buffer. Therefore, at load time, all checks on pointers
2121 * previously done by the verifier are invalidated and must be
2122 * performed again, if the helper is used in combination with
2123 * direct packet access.
2124 * Return
2125 * 0 on success, or a negative error in case of failure.
2126 *
bdb7b79b 2127 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2d020dd7
QM
2128 * Description
2129 * Retrieve the XFRM state (IP transform framework, see also
2130 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2131 *
2132 * The retrieved value is stored in the **struct bpf_xfrm_state**
2133 * pointed by *xfrm_state* and of length *size*.
2134 *
2135 * All values for *flags* are reserved for future usage, and must
2136 * be left at zero.
2137 *
2138 * This helper is available only if the kernel was compiled with
2139 * **CONFIG_XFRM** configuration option.
2140 * Return
2141 * 0 on success, or a negative error in case of failure.
c195651e 2142 *
bdb7b79b 2143 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
c195651e 2144 * Description
79552fbc
QM
2145 * Return a user or a kernel stack in bpf program provided buffer.
2146 * To achieve this, the helper needs *ctx*, which is a pointer
2147 * to the context on which the tracing program is executed.
2148 * To store the stacktrace, the bpf program provides *buf* with
2149 * a nonnegative *size*.
2150 *
2151 * The last argument, *flags*, holds the number of stack frames to
2152 * skip (from 0 to 255), masked with
2153 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2154 * the following flags:
2155 *
2156 * **BPF_F_USER_STACK**
2157 * Collect a user space stack instead of a kernel stack.
2158 * **BPF_F_USER_BUILD_ID**
2159 * Collect buildid+offset instead of ips for user stack,
2160 * only valid if **BPF_F_USER_STACK** is also specified.
2161 *
2162 * **bpf_get_stack**\ () can collect up to
2163 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2164 * to sufficient large buffer size. Note that
2165 * this limit can be controlled with the **sysctl** program, and
2166 * that it should be manually increased in order to profile long
2167 * user stacks (such as stacks for Java programs). To do so, use:
2168 *
2169 * ::
2170 *
2171 * # sysctl kernel.perf_event_max_stack=<new value>
c195651e 2172 * Return
7a279e93
QM
2173 * A non-negative value equal to or less than *size* on success,
2174 * or a negative error in case of failure.
4e1ec56c 2175 *
bdb7b79b 2176 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
4e1ec56c
DB
2177 * Description
2178 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2179 * it provides an easy way to load *len* bytes from *offset*
2180 * from the packet associated to *skb*, into the buffer pointed
2181 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2182 * a fifth argument *start_header* exists in order to select a
2183 * base offset to start from. *start_header* can be one of:
2184 *
2185 * **BPF_HDR_START_MAC**
2186 * Base offset to load data from is *skb*'s mac header.
2187 * **BPF_HDR_START_NET**
2188 * Base offset to load data from is *skb*'s network header.
2189 *
2190 * In general, "direct packet access" is the preferred method to
2191 * access packet data, however, this helper is in particular useful
2192 * in socket filters where *skb*\ **->data** does not always point
2193 * to the start of the mac header and where "direct packet access"
2194 * is not available.
4e1ec56c
DB
2195 * Return
2196 * 0 on success, or a negative error in case of failure.
2197 *
bdb7b79b 2198 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
87f5fc7e
DA
2199 * Description
2200 * Do FIB lookup in kernel tables using parameters in *params*.
2201 * If lookup is successful and result shows packet is to be
2202 * forwarded, the neighbor tables are searched for the nexthop.
2203 * If successful (ie., FIB lookup shows forwarding and nexthop
fa898d76
DA
2204 * is resolved), the nexthop address is returned in ipv4_dst
2205 * or ipv6_dst based on family, smac is set to mac address of
2206 * egress device, dmac is set to nexthop mac address, rt_metric
4c79579b
DA
2207 * is set to metric from route (IPv4/IPv6 only), and ifindex
2208 * is set to the device index of the nexthop from the FIB lookup.
87f5fc7e 2209 *
90b1023f
QM
2210 * *plen* argument is the size of the passed in struct.
2211 * *flags* argument can be a combination of one or more of the
2212 * following values:
87f5fc7e 2213 *
7a279e93
QM
2214 * **BPF_FIB_LOOKUP_DIRECT**
2215 * Do a direct table lookup vs full lookup using FIB
2216 * rules.
2217 * **BPF_FIB_LOOKUP_OUTPUT**
2218 * Perform lookup from an egress perspective (default is
2219 * ingress).
87f5fc7e 2220 *
90b1023f
QM
2221 * *ctx* is either **struct xdp_md** for XDP programs or
2222 * **struct sk_buff** tc cls_act programs.
2223 * Return
4c79579b
DA
2224 * * < 0 if any input argument is invalid
2225 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2226 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2bae79d2 2227 * packet is not forwarded or needs assist from full stack
81110384 2228 *
bdb7b79b 2229 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
81110384
JF
2230 * Description
2231 * Add an entry to, or update a sockhash *map* referencing sockets.
2232 * The *skops* is used as a new value for the entry associated to
2233 * *key*. *flags* is one of:
2234 *
2235 * **BPF_NOEXIST**
2236 * The entry for *key* must not exist in the map.
2237 * **BPF_EXIST**
2238 * The entry for *key* must already exist in the map.
2239 * **BPF_ANY**
2240 * No condition on the existence of the entry for *key*.
2241 *
2242 * If the *map* has eBPF programs (parser and verdict), those will
2243 * be inherited by the socket being added. If the socket is
2244 * already attached to eBPF programs, this results in an error.
2245 * Return
2246 * 0 on success, or a negative error in case of failure.
2247 *
bdb7b79b 2248 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
81110384
JF
2249 * Description
2250 * This helper is used in programs implementing policies at the
2251 * socket level. If the message *msg* is allowed to pass (i.e. if
2252 * the verdict eBPF program returns **SK_PASS**), redirect it to
2253 * the socket referenced by *map* (of type
2254 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2255 * egress interfaces can be used for redirection. The
2256 * **BPF_F_INGRESS** value in *flags* is used to make the
2257 * distinction (ingress path is selected if the flag is present,
2258 * egress path otherwise). This is the only flag supported for now.
2259 * Return
2260 * **SK_PASS** on success, or **SK_DROP** on error.
2261 *
bdb7b79b 2262 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
81110384
JF
2263 * Description
2264 * This helper is used in programs implementing policies at the
2265 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
49f3d12b 2266 * if the verdict eBPF program returns **SK_PASS**), redirect it
81110384
JF
2267 * to the socket referenced by *map* (of type
2268 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2269 * egress interfaces can be used for redirection. The
2270 * **BPF_F_INGRESS** value in *flags* is used to make the
2271 * distinction (ingress path is selected if the flag is present,
2272 * egress otherwise). This is the only flag supported for now.
2273 * Return
2274 * **SK_PASS** on success, or **SK_DROP** on error.
fe94cc29 2275 *
bdb7b79b 2276 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
fe94cc29
MX
2277 * Description
2278 * Encapsulate the packet associated to *skb* within a Layer 3
2279 * protocol header. This header is provided in the buffer at
2280 * address *hdr*, with *len* its size in bytes. *type* indicates
2281 * the protocol of the header and can be one of:
2282 *
2283 * **BPF_LWT_ENCAP_SEG6**
2284 * IPv6 encapsulation with Segment Routing Header
2285 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2286 * the IPv6 header is computed by the kernel.
2287 * **BPF_LWT_ENCAP_SEG6_INLINE**
2288 * Only works if *skb* contains an IPv6 packet. Insert a
2289 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2290 * the IPv6 header.
3e0bd37c
PO
2291 * **BPF_LWT_ENCAP_IP**
2292 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2293 * must be IPv4 or IPv6, followed by zero or more
80867c5e
QM
2294 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2295 * total bytes in all prepended headers. Please note that
2296 * if **skb_is_gso**\ (*skb*) is true, no more than two
2297 * headers can be prepended, and the inner header, if
2298 * present, should be either GRE or UDP/GUE.
3e0bd37c 2299 *
80867c5e
QM
2300 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2301 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2302 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2303 * **BPF_PROG_TYPE_LWT_XMIT**.
fe94cc29 2304 *
32e7dc28 2305 * A call to this helper is susceptible to change the underlying
fe94cc29
MX
2306 * packet buffer. Therefore, at load time, all checks on pointers
2307 * previously done by the verifier are invalidated and must be
2308 * performed again, if the helper is used in combination with
2309 * direct packet access.
2310 * Return
2311 * 0 on success, or a negative error in case of failure.
2312 *
bdb7b79b 2313 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
fe94cc29
MX
2314 * Description
2315 * Store *len* bytes from address *from* into the packet
2316 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2317 * inside the outermost IPv6 Segment Routing Header can be
2318 * modified through this helper.
2319 *
32e7dc28 2320 * A call to this helper is susceptible to change the underlying
fe94cc29
MX
2321 * packet buffer. Therefore, at load time, all checks on pointers
2322 * previously done by the verifier are invalidated and must be
2323 * performed again, if the helper is used in combination with
2324 * direct packet access.
2325 * Return
2326 * 0 on success, or a negative error in case of failure.
2327 *
bdb7b79b 2328 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
fe94cc29
MX
2329 * Description
2330 * Adjust the size allocated to TLVs in the outermost IPv6
2331 * Segment Routing Header contained in the packet associated to
2332 * *skb*, at position *offset* by *delta* bytes. Only offsets
2333 * after the segments are accepted. *delta* can be as well
2334 * positive (growing) as negative (shrinking).
2335 *
32e7dc28 2336 * A call to this helper is susceptible to change the underlying
fe94cc29
MX
2337 * packet buffer. Therefore, at load time, all checks on pointers
2338 * previously done by the verifier are invalidated and must be
2339 * performed again, if the helper is used in combination with
2340 * direct packet access.
2341 * Return
2342 * 0 on success, or a negative error in case of failure.
2343 *
bdb7b79b 2344 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
fe94cc29
MX
2345 * Description
2346 * Apply an IPv6 Segment Routing action of type *action* to the
2347 * packet associated to *skb*. Each action takes a parameter
2348 * contained at address *param*, and of length *param_len* bytes.
2349 * *action* can be one of:
2350 *
2351 * **SEG6_LOCAL_ACTION_END_X**
2352 * End.X action: Endpoint with Layer-3 cross-connect.
2353 * Type of *param*: **struct in6_addr**.
2354 * **SEG6_LOCAL_ACTION_END_T**
2355 * End.T action: Endpoint with specific IPv6 table lookup.
2356 * Type of *param*: **int**.
2357 * **SEG6_LOCAL_ACTION_END_B6**
2358 * End.B6 action: Endpoint bound to an SRv6 policy.
80867c5e 2359 * Type of *param*: **struct ipv6_sr_hdr**.
fe94cc29
MX
2360 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2361 * End.B6.Encap action: Endpoint bound to an SRv6
2362 * encapsulation policy.
80867c5e 2363 * Type of *param*: **struct ipv6_sr_hdr**.
fe94cc29 2364 *
32e7dc28 2365 * A call to this helper is susceptible to change the underlying
fe94cc29
MX
2366 * packet buffer. Therefore, at load time, all checks on pointers
2367 * previously done by the verifier are invalidated and must be
2368 * performed again, if the helper is used in combination with
2369 * direct packet access.
2370 * Return
2371 * 0 on success, or a negative error in case of failure.
f4364dcf 2372 *
bdb7b79b 2373 * long bpf_rc_repeat(void *ctx)
f4364dcf
SY
2374 * Description
2375 * This helper is used in programs implementing IR decoding, to
62369db2
QM
2376 * report a successfully decoded repeat key message. This delays
2377 * the generation of a key up event for previously generated
2378 * key down event.
f4364dcf 2379 *
62369db2
QM
2380 * Some IR protocols like NEC have a special IR message for
2381 * repeating last button, for when a button is held down.
f4364dcf
SY
2382 *
2383 * The *ctx* should point to the lirc sample as passed into
2384 * the program.
2385 *
f4364dcf
SY
2386 * This helper is only available is the kernel was compiled with
2387 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2388 * "**y**".
f4364dcf
SY
2389 * Return
2390 * 0
2391 *
bdb7b79b 2392 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
f4364dcf
SY
2393 * Description
2394 * This helper is used in programs implementing IR decoding, to
62369db2
QM
2395 * report a successfully decoded key press with *scancode*,
2396 * *toggle* value in the given *protocol*. The scancode will be
2397 * translated to a keycode using the rc keymap, and reported as
2398 * an input key down event. After a period a key up event is
2399 * generated. This period can be extended by calling either
2400 * **bpf_rc_keydown**\ () again with the same values, or calling
2401 * **bpf_rc_repeat**\ ().
f4364dcf 2402 *
ab8d7809 2403 * Some protocols include a toggle bit, in case the button was
62369db2 2404 * released and pressed again between consecutive scancodes.
f4364dcf
SY
2405 *
2406 * The *ctx* should point to the lirc sample as passed into
2407 * the program.
2408 *
62369db2
QM
2409 * The *protocol* is the decoded protocol number (see
2410 * **enum rc_proto** for some predefined values).
2411 *
f4364dcf
SY
2412 * This helper is only available is the kernel was compiled with
2413 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2414 * "**y**".
f4364dcf
SY
2415 * Return
2416 * 0
cb20b08e 2417 *
62369db2 2418 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
cb20b08e
DB
2419 * Description
2420 * Return the cgroup v2 id of the socket associated with the *skb*.
2421 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2422 * helper for cgroup v1 by providing a tag resp. identifier that
2423 * can be matched on or used for map lookups e.g. to implement
2424 * policy. The cgroup v2 id of a given path in the hierarchy is
2425 * exposed in user space through the f_handle API in order to get
2426 * to the same 64-bit id.
2427 *
2428 * This helper can be used on TC egress path, but not on ingress,
2429 * and is available only if the kernel was compiled with the
2430 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2431 * Return
2432 * The id is returned or 0 in case the id could not be retrieved.
bf6fa2c8
YS
2433 *
2434 * u64 bpf_get_current_cgroup_id(void)
2435 * Return
2436 * A 64-bit integer containing the current cgroup id based
2437 * on the cgroup within which the current task is running.
cd339431 2438 *
62369db2 2439 * void *bpf_get_local_storage(void *map, u64 flags)
cd339431
RG
2440 * Description
2441 * Get the pointer to the local storage area.
2442 * The type and the size of the local storage is defined
2443 * by the *map* argument.
2444 * The *flags* meaning is specific for each map type,
2445 * and has to be 0 for cgroup local storage.
2446 *
90b1023f
QM
2447 * Depending on the BPF program type, a local storage area
2448 * can be shared between multiple instances of the BPF program,
cd339431
RG
2449 * running simultaneously.
2450 *
2451 * A user should care about the synchronization by himself.
91c960b0 2452 * For example, by using the **BPF_ATOMIC** instructions to alter
cd339431
RG
2453 * the shared data.
2454 * Return
90b1023f 2455 * A pointer to the local storage area.
2dbb9b9e 2456 *
bdb7b79b 2457 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2dbb9b9e 2458 * Description
90b1023f
QM
2459 * Select a **SO_REUSEPORT** socket from a
2460 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2461 * It checks the selected socket is matching the incoming
2462 * request in the socket buffer.
2dbb9b9e
MKL
2463 * Return
2464 * 0 on success, or a negative error in case of failure.
6acc9b43 2465 *
62369db2
QM
2466 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2467 * Description
2468 * Return id of cgroup v2 that is ancestor of cgroup associated
2469 * with the *skb* at the *ancestor_level*. The root cgroup is at
2470 * *ancestor_level* zero and each step down the hierarchy
2471 * increments the level. If *ancestor_level* == level of cgroup
2472 * associated with *skb*, then return value will be same as that
2473 * of **bpf_skb_cgroup_id**\ ().
2474 *
2475 * The helper is useful to implement policies based on cgroups
2476 * that are upper in hierarchy than immediate cgroup associated
2477 * with *skb*.
2478 *
2479 * The format of returned id and helper limitations are same as in
2480 * **bpf_skb_cgroup_id**\ ().
2481 * Return
2482 * The id is returned or 0 in case the id could not be retrieved.
2483 *
f71c6143 2484 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
6acc9b43
JS
2485 * Description
2486 * Look for TCP socket matching *tuple*, optionally in a child
2487 * network namespace *netns*. The return value must be checked,
90b1023f 2488 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2489 *
2490 * The *ctx* should point to the context of the program, such as
2491 * the skb or socket (depending on the hook in use). This is used
2492 * to determine the base network namespace for the lookup.
2493 *
2494 * *tuple_size* must be one of:
2495 *
2496 * **sizeof**\ (*tuple*\ **->ipv4**)
2497 * Look for an IPv4 socket.
2498 * **sizeof**\ (*tuple*\ **->ipv6**)
2499 * Look for an IPv6 socket.
2500 *
f71c6143 2501 * If the *netns* is a negative signed 32-bit integer, then the
bfdfa517 2502 * socket lookup table in the netns associated with the *ctx*
f71c6143
JS
2503 * will be used. For the TC hooks, this is the netns of the device
2504 * in the skb. For socket hooks, this is the netns of the socket.
2505 * If *netns* is any other signed 32-bit value greater than or
2506 * equal to zero then it specifies the ID of the netns relative to
2507 * the netns associated with the *ctx*. *netns* values beyond the
2508 * range of 32-bit integers are reserved for future use.
6acc9b43
JS
2509 *
2510 * All values for *flags* are reserved for future usage, and must
2511 * be left at zero.
2512 *
2513 * This helper is available only if the kernel was compiled with
2514 * **CONFIG_NET** configuration option.
2515 * Return
0bd72117
DB
2516 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2517 * For sockets with reuseport option, the **struct bpf_sock**
80867c5e
QM
2518 * result is from *reuse*\ **->socks**\ [] using the hash of the
2519 * tuple.
6acc9b43 2520 *
f71c6143 2521 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
6acc9b43
JS
2522 * Description
2523 * Look for UDP socket matching *tuple*, optionally in a child
2524 * network namespace *netns*. The return value must be checked,
90b1023f 2525 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2526 *
2527 * The *ctx* should point to the context of the program, such as
2528 * the skb or socket (depending on the hook in use). This is used
2529 * to determine the base network namespace for the lookup.
2530 *
2531 * *tuple_size* must be one of:
2532 *
2533 * **sizeof**\ (*tuple*\ **->ipv4**)
2534 * Look for an IPv4 socket.
2535 * **sizeof**\ (*tuple*\ **->ipv6**)
2536 * Look for an IPv6 socket.
2537 *
f71c6143 2538 * If the *netns* is a negative signed 32-bit integer, then the
bfdfa517 2539 * socket lookup table in the netns associated with the *ctx*
f71c6143
JS
2540 * will be used. For the TC hooks, this is the netns of the device
2541 * in the skb. For socket hooks, this is the netns of the socket.
2542 * If *netns* is any other signed 32-bit value greater than or
2543 * equal to zero then it specifies the ID of the netns relative to
2544 * the netns associated with the *ctx*. *netns* values beyond the
2545 * range of 32-bit integers are reserved for future use.
6acc9b43
JS
2546 *
2547 * All values for *flags* are reserved for future usage, and must
2548 * be left at zero.
2549 *
2550 * This helper is available only if the kernel was compiled with
2551 * **CONFIG_NET** configuration option.
2552 * Return
0bd72117
DB
2553 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2554 * For sockets with reuseport option, the **struct bpf_sock**
80867c5e
QM
2555 * result is from *reuse*\ **->socks**\ [] using the hash of the
2556 * tuple.
6acc9b43 2557 *
a5fa25ad 2558 * long bpf_sk_release(void *sock)
6acc9b43 2559 * Description
90b1023f
QM
2560 * Release the reference held by *sock*. *sock* must be a
2561 * non-**NULL** pointer that was returned from
2562 * **bpf_sk_lookup_xxx**\ ().
6acc9b43
JS
2563 * Return
2564 * 0 on success, or a negative error in case of failure.
6fff607e 2565 *
bdb7b79b 2566 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
62369db2
QM
2567 * Description
2568 * Push an element *value* in *map*. *flags* is one of:
2569 *
2570 * **BPF_EXIST**
2571 * If the queue/stack is full, the oldest element is
2572 * removed to make room for this.
2573 * Return
2574 * 0 on success, or a negative error in case of failure.
2575 *
bdb7b79b 2576 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
90b1023f
QM
2577 * Description
2578 * Pop an element from *map*.
2579 * Return
2580 * 0 on success, or a negative error in case of failure.
2581 *
bdb7b79b 2582 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
90b1023f
QM
2583 * Description
2584 * Get an element from *map* without removing it.
2585 * Return
2586 * 0 on success, or a negative error in case of failure.
2587 *
bdb7b79b 2588 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
6fff607e 2589 * Description
90b1023f 2590 * For socket policies, insert *len* bytes into *msg* at offset
6fff607e
JF
2591 * *start*.
2592 *
2593 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
90b1023f 2594 * *msg* it may want to insert metadata or options into the *msg*.
6fff607e
JF
2595 * This can later be read and used by any of the lower layer BPF
2596 * hooks.
2597 *
2598 * This helper may fail if under memory pressure (a malloc
2599 * fails) in these cases BPF programs will get an appropriate
2600 * error and BPF programs will need to handle them.
6fff607e
JF
2601 * Return
2602 * 0 on success, or a negative error in case of failure.
7246d8ed 2603 *
bdb7b79b 2604 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
90b1023f 2605 * Description
5f0e5412 2606 * Will remove *len* bytes from a *msg* starting at byte *start*.
7246d8ed
JF
2607 * This may result in **ENOMEM** errors under certain situations if
2608 * an allocation and copy are required due to a full ring buffer.
2609 * However, the helper will try to avoid doing the allocation
2610 * if possible. Other errors can occur if input parameters are
90b1023f 2611 * invalid either due to *start* byte not being valid part of *msg*
7246d8ed 2612 * payload and/or *pop* value being to large.
7246d8ed 2613 * Return
90b1023f 2614 * 0 on success, or a negative error in case of failure.
01d3240a 2615 *
bdb7b79b 2616 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
01d3240a
SY
2617 * Description
2618 * This helper is used in programs implementing IR decoding, to
2619 * report a successfully decoded pointer movement.
7246d8ed 2620 *
01d3240a
SY
2621 * The *ctx* should point to the lirc sample as passed into
2622 * the program.
2623 *
2624 * This helper is only available is the kernel was compiled with
2625 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2626 * "**y**".
7246d8ed 2627 * Return
01d3240a 2628 * 0
46f8bc92 2629 *
bdb7b79b 2630 * long bpf_spin_lock(struct bpf_spin_lock *lock)
0eb09785
QM
2631 * Description
2632 * Acquire a spinlock represented by the pointer *lock*, which is
2633 * stored as part of a value of a map. Taking the lock allows to
2634 * safely update the rest of the fields in that value. The
2635 * spinlock can (and must) later be released with a call to
2636 * **bpf_spin_unlock**\ (\ *lock*\ ).
2637 *
2638 * Spinlocks in BPF programs come with a number of restrictions
2639 * and constraints:
2640 *
2641 * * **bpf_spin_lock** objects are only allowed inside maps of
2642 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2643 * list could be extended in the future).
2644 * * BTF description of the map is mandatory.
2645 * * The BPF program can take ONE lock at a time, since taking two
2646 * or more could cause dead locks.
2647 * * Only one **struct bpf_spin_lock** is allowed per map element.
2648 * * When the lock is taken, calls (either BPF to BPF or helpers)
2649 * are not allowed.
2650 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2651 * allowed inside a spinlock-ed region.
2652 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2653 * the lock, on all execution paths, before it returns.
2654 * * The BPF program can access **struct bpf_spin_lock** only via
2655 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2656 * helpers. Loading or storing data into the **struct
2657 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2658 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2659 * of the map value must be a struct and have **struct
2660 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2661 * Nested lock inside another struct is not allowed.
2662 * * The **struct bpf_spin_lock** *lock* field in a map value must
2663 * be aligned on a multiple of 4 bytes in that value.
2664 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2665 * the **bpf_spin_lock** field to user space.
2666 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2667 * a BPF program, do not update the **bpf_spin_lock** field.
2668 * * **bpf_spin_lock** cannot be on the stack or inside a
2669 * networking packet (it can only be inside of a map values).
2670 * * **bpf_spin_lock** is available to root only.
2671 * * Tracing programs and socket filter programs cannot use
2672 * **bpf_spin_lock**\ () due to insufficient preemption checks
2673 * (but this may change in the future).
2674 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2675 * Return
2676 * 0
2677 *
bdb7b79b 2678 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
0eb09785
QM
2679 * Description
2680 * Release the *lock* previously locked by a call to
2681 * **bpf_spin_lock**\ (\ *lock*\ ).
2682 * Return
2683 * 0
2684 *
46f8bc92
MKL
2685 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2686 * Description
2687 * This helper gets a **struct bpf_sock** pointer such
62369db2 2688 * that all the fields in this **bpf_sock** can be accessed.
46f8bc92 2689 * Return
62369db2 2690 * A **struct bpf_sock** pointer on success, or **NULL** in
46f8bc92 2691 * case of failure.
655a51e5
MKL
2692 *
2693 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2694 * Description
2695 * This helper gets a **struct bpf_tcp_sock** pointer from a
2696 * **struct bpf_sock** pointer.
655a51e5 2697 * Return
62369db2 2698 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
655a51e5 2699 * case of failure.
f7c917ba 2700 *
bdb7b79b 2701 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
62369db2
QM
2702 * Description
2703 * Set ECN (Explicit Congestion Notification) field of IP header
2704 * to **CE** (Congestion Encountered) if current value is **ECT**
2705 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2706 * and IPv4.
2707 * Return
2708 * 1 if the **CE** flag is set (either by the current helper call
2709 * or because it was already present), 0 if it is not set.
dbafd7dd
MKL
2710 *
2711 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2712 * Description
62369db2
QM
2713 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2714 * **bpf_sk_release**\ () is unnecessary and not allowed.
dbafd7dd 2715 * Return
62369db2 2716 * A **struct bpf_sock** pointer on success, or **NULL** in
dbafd7dd 2717 * case of failure.
edbf8c01
LB
2718 *
2719 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2720 * Description
2721 * Look for TCP socket matching *tuple*, optionally in a child
2722 * network namespace *netns*. The return value must be checked,
2723 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2724 *
80867c5e
QM
2725 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2726 * that it also returns timewait or request sockets. Use
2727 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2728 * full structure.
edbf8c01
LB
2729 *
2730 * This helper is available only if the kernel was compiled with
2731 * **CONFIG_NET** configuration option.
2732 * Return
2733 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2734 * For sockets with reuseport option, the **struct bpf_sock**
80867c5e
QM
2735 * result is from *reuse*\ **->socks**\ [] using the hash of the
2736 * tuple.
39904084 2737 *
c0df236e 2738 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
39904084 2739 * Description
80867c5e
QM
2740 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2741 * the listening socket in *sk*.
39904084 2742 *
80867c5e
QM
2743 * *iph* points to the start of the IPv4 or IPv6 header, while
2744 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2745 * **sizeof**\ (**struct ip6hdr**).
39904084 2746 *
80867c5e
QM
2747 * *th* points to the start of the TCP header, while *th_len*
2748 * contains **sizeof**\ (**struct tcphdr**).
39904084 2749 * Return
80867c5e
QM
2750 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2751 * error otherwise.
808649fb 2752 *
bdb7b79b 2753 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
808649fb
AI
2754 * Description
2755 * Get name of sysctl in /proc/sys/ and copy it into provided by
2756 * program buffer *buf* of size *buf_len*.
2757 *
2758 * The buffer is always NUL terminated, unless it's zero-sized.
2759 *
2760 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2761 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2762 * only (e.g. "tcp_mem").
2763 * Return
2764 * Number of character copied (not including the trailing NUL).
2765 *
2766 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2767 * truncated name in this case).
1d11b301 2768 *
bdb7b79b 2769 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
1d11b301
AI
2770 * Description
2771 * Get current value of sysctl as it is presented in /proc/sys
2772 * (incl. newline, etc), and copy it as a string into provided
2773 * by program buffer *buf* of size *buf_len*.
2774 *
2775 * The whole value is copied, no matter what file position user
2776 * space issued e.g. sys_read at.
2777 *
2778 * The buffer is always NUL terminated, unless it's zero-sized.
2779 * Return
2780 * Number of character copied (not including the trailing NUL).
2781 *
2782 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2783 * truncated name in this case).
2784 *
2785 * **-EINVAL** if current value was unavailable, e.g. because
2786 * sysctl is uninitialized and read returns -EIO for it.
4e63acdf 2787 *
bdb7b79b 2788 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4e63acdf
AI
2789 * Description
2790 * Get new value being written by user space to sysctl (before
2791 * the actual write happens) and copy it as a string into
2792 * provided by program buffer *buf* of size *buf_len*.
2793 *
2794 * User space may write new value at file position > 0.
2795 *
2796 * The buffer is always NUL terminated, unless it's zero-sized.
2797 * Return
2798 * Number of character copied (not including the trailing NUL).
2799 *
2800 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2801 * truncated name in this case).
2802 *
2803 * **-EINVAL** if sysctl is being read.
2804 *
bdb7b79b 2805 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
4e63acdf
AI
2806 * Description
2807 * Override new value being written by user space to sysctl with
2808 * value provided by program in buffer *buf* of size *buf_len*.
2809 *
2810 * *buf* should contain a string in same form as provided by user
2811 * space on sysctl write.
2812 *
2813 * User space may write new value at file position > 0. To override
2814 * the whole sysctl value file position should be set to zero.
2815 * Return
2816 * 0 on success.
2817 *
2818 * **-E2BIG** if the *buf_len* is too big.
2819 *
2820 * **-EINVAL** if sysctl is being read.
d7a4cb9b 2821 *
bdb7b79b 2822 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
d7a4cb9b
AI
2823 * Description
2824 * Convert the initial part of the string from buffer *buf* of
2825 * size *buf_len* to a long integer according to the given base
2826 * and save the result in *res*.
2827 *
2828 * The string may begin with an arbitrary amount of white space
80867c5e
QM
2829 * (as determined by **isspace**\ (3)) followed by a single
2830 * optional '**-**' sign.
d7a4cb9b
AI
2831 *
2832 * Five least significant bits of *flags* encode base, other bits
2833 * are currently unused.
2834 *
2835 * Base must be either 8, 10, 16 or 0 to detect it automatically
80867c5e 2836 * similar to user space **strtol**\ (3).
d7a4cb9b
AI
2837 * Return
2838 * Number of characters consumed on success. Must be positive but
80867c5e 2839 * no more than *buf_len*.
d7a4cb9b
AI
2840 *
2841 * **-EINVAL** if no valid digits were found or unsupported base
2842 * was provided.
2843 *
2844 * **-ERANGE** if resulting value was out of range.
2845 *
bdb7b79b 2846 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
d7a4cb9b
AI
2847 * Description
2848 * Convert the initial part of the string from buffer *buf* of
2849 * size *buf_len* to an unsigned long integer according to the
2850 * given base and save the result in *res*.
2851 *
2852 * The string may begin with an arbitrary amount of white space
80867c5e 2853 * (as determined by **isspace**\ (3)).
d7a4cb9b
AI
2854 *
2855 * Five least significant bits of *flags* encode base, other bits
2856 * are currently unused.
2857 *
2858 * Base must be either 8, 10, 16 or 0 to detect it automatically
80867c5e 2859 * similar to user space **strtoul**\ (3).
d7a4cb9b
AI
2860 * Return
2861 * Number of characters consumed on success. Must be positive but
80867c5e 2862 * no more than *buf_len*.
d7a4cb9b
AI
2863 *
2864 * **-EINVAL** if no valid digits were found or unsupported base
2865 * was provided.
2866 *
2867 * **-ERANGE** if resulting value was out of range.
6ac99e8f 2868 *
30897832 2869 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
6ac99e8f 2870 * Description
80867c5e 2871 * Get a bpf-local-storage from a *sk*.
6ac99e8f
MKL
2872 *
2873 * Logically, it could be thought of getting the value from
2874 * a *map* with *sk* as the **key**. From this
2875 * perspective, the usage is not much different from
80867c5e
QM
2876 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2877 * helper enforces the key must be a full socket and the map must
2878 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
6ac99e8f
MKL
2879 *
2880 * Underneath, the value is stored locally at *sk* instead of
80867c5e
QM
2881 * the *map*. The *map* is used as the bpf-local-storage
2882 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2883 * searched against all bpf-local-storages residing at *sk*.
6ac99e8f 2884 *
30897832
KS
2885 * *sk* is a kernel **struct sock** pointer for LSM program.
2886 * *sk* is a **struct bpf_sock** pointer for other program types.
2887 *
80867c5e 2888 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
6ac99e8f
MKL
2889 * used such that a new bpf-local-storage will be
2890 * created if one does not exist. *value* can be used
80867c5e 2891 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
6ac99e8f 2892 * the initial value of a bpf-local-storage. If *value* is
80867c5e 2893 * **NULL**, the new bpf-local-storage will be zero initialized.
6ac99e8f
MKL
2894 * Return
2895 * A bpf-local-storage pointer is returned on success.
2896 *
2897 * **NULL** if not found or there was an error in adding
2898 * a new bpf-local-storage.
2899 *
30897832 2900 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
6ac99e8f 2901 * Description
80867c5e 2902 * Delete a bpf-local-storage from a *sk*.
6ac99e8f
MKL
2903 * Return
2904 * 0 on success.
2905 *
2906 * **-ENOENT** if the bpf-local-storage cannot be found.
592a3498 2907 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
8b401f9e 2908 *
bdb7b79b 2909 * long bpf_send_signal(u32 sig)
8b401f9e 2910 * Description
8482941f
YS
2911 * Send signal *sig* to the process of the current task.
2912 * The signal may be delivered to any of this process's threads.
8b401f9e
YS
2913 * Return
2914 * 0 on success or successfully queued.
2915 *
2916 * **-EBUSY** if work queue under nmi is full.
2917 *
2918 * **-EINVAL** if *sig* is invalid.
2919 *
2920 * **-EPERM** if no permission to send the *sig*.
2921 *
2922 * **-EAGAIN** if bpf program can try again.
70d66244 2923 *
c0df236e 2924 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
70d66244
PP
2925 * Description
2926 * Try to issue a SYN cookie for the packet with corresponding
2927 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2928 *
2929 * *iph* points to the start of the IPv4 or IPv6 header, while
2930 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2931 * **sizeof**\ (**struct ip6hdr**).
2932 *
2933 * *th* points to the start of the TCP header, while *th_len*
2934 * contains the length of the TCP header.
70d66244
PP
2935 * Return
2936 * On success, lower 32 bits hold the generated SYN cookie in
2937 * followed by 16 bits which hold the MSS value for that cookie,
2938 * and the top 16 bits are unused.
2939 *
2940 * On failure, the returned value is one of the following:
2941 *
2942 * **-EINVAL** SYN cookie cannot be issued due to error
2943 *
2944 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2945 *
2946 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2947 *
2948 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
a7658e1a 2949 *
bdb7b79b 2950 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
a7658e1a
AS
2951 * Description
2952 * Write raw *data* blob into a special BPF perf event held by
2953 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2954 * event must have the following attributes: **PERF_SAMPLE_RAW**
2955 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2956 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2957 *
2958 * The *flags* are used to indicate the index in *map* for which
2959 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2960 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2961 * to indicate that the index of the current CPU core should be
2962 * used.
2963 *
2964 * The value to write, of *size*, is passed through eBPF stack and
2965 * pointed by *data*.
2966 *
2967 * *ctx* is a pointer to in-kernel struct sk_buff.
2968 *
2969 * This helper is similar to **bpf_perf_event_output**\ () but
2970 * restricted to raw_tracepoint bpf programs.
2971 * Return
2972 * 0 on success, or a negative error in case of failure.
6ae08ae3 2973 *
bdb7b79b 2974 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
6ae08ae3
DB
2975 * Description
2976 * Safely attempt to read *size* bytes from user space address
2977 * *unsafe_ptr* and store the data in *dst*.
2978 * Return
2979 * 0 on success, or a negative error in case of failure.
2980 *
bdb7b79b 2981 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
6ae08ae3
DB
2982 * Description
2983 * Safely attempt to read *size* bytes from kernel space address
2984 * *unsafe_ptr* and store the data in *dst*.
2985 * Return
2986 * 0 on success, or a negative error in case of failure.
2987 *
bdb7b79b 2988 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
6ae08ae3
DB
2989 * Description
2990 * Copy a NUL terminated string from an unsafe user address
2991 * *unsafe_ptr* to *dst*. The *size* should include the
2992 * terminating NUL byte. In case the string length is smaller than
2993 * *size*, the target is not padded with further NUL bytes. If the
2994 * string length is larger than *size*, just *size*-1 bytes are
2995 * copied and the last byte is set to NUL.
2996 *
c6458e72
BJ
2997 * On success, returns the number of bytes that were written,
2998 * including the terminal NUL. This makes this helper useful in
2999 * tracing programs for reading strings, and more importantly to
3000 * get its length at runtime. See the following snippet:
6ae08ae3
DB
3001 *
3002 * ::
3003 *
3004 * SEC("kprobe/sys_open")
3005 * void bpf_sys_open(struct pt_regs *ctx)
3006 * {
3007 * char buf[PATHLEN]; // PATHLEN is defined to 256
3008 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3009 * ctx->di);
3010 *
3011 * // Consume buf, for example push it to
3012 * // userspace via bpf_perf_event_output(); we
3013 * // can use res (the string length) as event
3014 * // size, after checking its boundaries.
3015 * }
3016 *
ab8d7809 3017 * In comparison, using **bpf_probe_read_user**\ () helper here
6ae08ae3
DB
3018 * instead to read the string would require to estimate the length
3019 * at compile time, and would often result in copying more memory
3020 * than necessary.
3021 *
3022 * Another useful use case is when parsing individual process
3023 * arguments or individual environment variables navigating
3024 * *current*\ **->mm->arg_start** and *current*\
3025 * **->mm->env_start**: using this helper and the return value,
3026 * one can quickly iterate at the right offset of the memory area.
3027 * Return
c6458e72 3028 * On success, the strictly positive length of the output string,
6ae08ae3
DB
3029 * including the trailing NUL character. On error, a negative
3030 * value.
3031 *
bdb7b79b 3032 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
6ae08ae3
DB
3033 * Description
3034 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
ab8d7809 3035 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
6ae08ae3 3036 * Return
ab8d7809 3037 * On success, the strictly positive length of the string, including
6ae08ae3 3038 * the trailing NUL character. On error, a negative value.
206057fe 3039 *
bdb7b79b 3040 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
206057fe 3041 * Description
ab8d7809 3042 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
206057fe
MKL
3043 * *rcv_nxt* is the ack_seq to be sent out.
3044 * Return
3045 * 0 on success, or a negative error in case of failure.
3046 *
bdb7b79b 3047 * long bpf_send_signal_thread(u32 sig)
8482941f
YS
3048 * Description
3049 * Send signal *sig* to the thread corresponding to the current task.
3050 * Return
3051 * 0 on success or successfully queued.
3052 *
3053 * **-EBUSY** if work queue under nmi is full.
3054 *
3055 * **-EINVAL** if *sig* is invalid.
3056 *
3057 * **-EPERM** if no permission to send the *sig*.
3058 *
3059 * **-EAGAIN** if bpf program can try again.
5576b991
MKL
3060 *
3061 * u64 bpf_jiffies64(void)
3062 * Description
3063 * Obtain the 64bit jiffies
3064 * Return
3065 * The 64 bit jiffies
fff7b643 3066 *
bdb7b79b 3067 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
fff7b643
DX
3068 * Description
3069 * For an eBPF program attached to a perf event, retrieve the
ab8d7809
QM
3070 * branch records (**struct perf_branch_entry**) associated to *ctx*
3071 * and store it in the buffer pointed by *buf* up to size
fff7b643
DX
3072 * *size* bytes.
3073 * Return
3074 * On success, number of bytes written to *buf*. On error, a
3075 * negative value.
3076 *
3077 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
ab8d7809 3078 * instead return the number of bytes required to store all the
fff7b643
DX
3079 * branch entries. If this flag is set, *buf* may be NULL.
3080 *
3081 * **-EINVAL** if arguments invalid or **size** not a multiple
ab8d7809 3082 * of **sizeof**\ (**struct perf_branch_entry**\ ).
fff7b643
DX
3083 *
3084 * **-ENOENT** if architecture does not support branch records.
b4490c5c 3085 *
bdb7b79b 3086 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
b4490c5c
CN
3087 * Description
3088 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3089 * *namespace* will be returned in *nsdata*.
ab8d7809
QM
3090 * Return
3091 * 0 on success, or one of the following in case of failure:
b4490c5c
CN
3092 *
3093 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3094 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3095 *
3096 * **-ENOENT** if pidns does not exists for the current task.
3097 *
bdb7b79b 3098 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
d831ee84
EC
3099 * Description
3100 * Write raw *data* blob into a special BPF perf event held by
3101 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3102 * event must have the following attributes: **PERF_SAMPLE_RAW**
3103 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3104 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3105 *
3106 * The *flags* are used to indicate the index in *map* for which
3107 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3108 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3109 * to indicate that the index of the current CPU core should be
3110 * used.
3111 *
3112 * The value to write, of *size*, is passed through eBPF stack and
3113 * pointed by *data*.
3114 *
3115 * *ctx* is a pointer to in-kernel struct xdp_buff.
3116 *
3117 * This helper is similar to **bpf_perf_eventoutput**\ () but
3118 * restricted to raw_tracepoint bpf programs.
3119 * Return
3120 * 0 on success, or a negative error in case of failure.
f318903c
DB
3121 *
3122 * u64 bpf_get_netns_cookie(void *ctx)
3123 * Description
3124 * Retrieve the cookie (generated by the kernel) of the network
3125 * namespace the input *ctx* is associated with. The network
3126 * namespace cookie remains stable for its lifetime and provides
3127 * a global identifier that can be assumed unique. If *ctx* is
3128 * NULL, then the helper returns the cookie for the initial
3129 * network namespace. The cookie itself is very similar to that
ab8d7809
QM
3130 * of **bpf_get_socket_cookie**\ () helper, but for network
3131 * namespaces instead of sockets.
f318903c
DB
3132 * Return
3133 * A 8-byte long opaque number.
0f09abd1
DB
3134 *
3135 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3136 * Description
3137 * Return id of cgroup v2 that is ancestor of the cgroup associated
3138 * with the current task at the *ancestor_level*. The root cgroup
3139 * is at *ancestor_level* zero and each step down the hierarchy
3140 * increments the level. If *ancestor_level* == level of cgroup
3141 * associated with the current task, then return value will be the
3142 * same as that of **bpf_get_current_cgroup_id**\ ().
3143 *
3144 * The helper is useful to implement policies based on cgroups
3145 * that are upper in hierarchy than immediate cgroup associated
3146 * with the current task.
3147 *
3148 * The format of returned id and helper limitations are same as in
3149 * **bpf_get_current_cgroup_id**\ ().
3150 * Return
3151 * The id is returned or 0 in case the id could not be retrieved.
cf7fbe66 3152 *
27e5203b 3153 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
cf7fbe66 3154 * Description
e9ddbb77
JS
3155 * Helper is overloaded depending on BPF program type. This
3156 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3157 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3158 *
cf7fbe66
JS
3159 * Assign the *sk* to the *skb*. When combined with appropriate
3160 * routing configuration to receive the packet towards the socket,
3161 * will cause *skb* to be delivered to the specified socket.
3162 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3163 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3164 * interfere with successful delivery to the socket.
3165 *
3166 * This operation is only valid from TC ingress path.
3167 *
3168 * The *flags* argument must be zero.
3169 * Return
ab8d7809 3170 * 0 on success, or a negative error in case of failure:
cf7fbe66 3171 *
ab8d7809
QM
3172 * **-EINVAL** if specified *flags* are not supported.
3173 *
3174 * **-ENOENT** if the socket is unavailable for assignment.
cf7fbe66 3175 *
ab8d7809
QM
3176 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3177 *
3178 * **-EOPNOTSUPP** if the operation is not supported, for example
3179 * a call from outside of TC ingress.
3180 *
3181 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3182 * (reuseport).
71d19214 3183 *
e9ddbb77
JS
3184 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3185 * Description
3186 * Helper is overloaded depending on BPF program type. This
3187 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3188 *
3189 * Select the *sk* as a result of a socket lookup.
3190 *
3191 * For the operation to succeed passed socket must be compatible
3192 * with the packet description provided by the *ctx* object.
3193 *
3194 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3195 * be an exact match. While IP family (**AF_INET** or
3196 * **AF_INET6**) must be compatible, that is IPv6 sockets
3197 * that are not v6-only can be selected for IPv4 packets.
3198 *
3199 * Only TCP listeners and UDP unconnected sockets can be
3200 * selected. *sk* can also be NULL to reset any previous
3201 * selection.
3202 *
3203 * *flags* argument can combination of following values:
3204 *
3205 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3206 * socket selection, potentially done by a BPF program
3207 * that ran before us.
3208 *
3209 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3210 * load-balancing within reuseport group for the socket
3211 * being selected.
3212 *
3213 * On success *ctx->sk* will point to the selected socket.
3214 *
3215 * Return
3216 * 0 on success, or a negative errno in case of failure.
3217 *
3218 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3219 * not compatible with packet family (*ctx->family*).
3220 *
3221 * * **-EEXIST** if socket has been already selected,
3222 * potentially by another program, and
3223 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3224 *
3225 * * **-EINVAL** if unsupported flags were specified.
3226 *
3227 * * **-EPROTOTYPE** if socket L4 protocol
3228 * (*sk->protocol*) doesn't match packet protocol
3229 * (*ctx->protocol*).
3230 *
3231 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3232 * state (TCP listening or UDP unconnected).
3233 *
71d19214
3234 * u64 bpf_ktime_get_boot_ns(void)
3235 * Description
3236 * Return the time elapsed since system boot, in nanoseconds.
3237 * Does include the time the system was suspended.
ab8d7809 3238 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
71d19214
3239 * Return
3240 * Current *ktime*.
492e639f 3241 *
bdb7b79b 3242 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
492e639f 3243 * Description
ab8d7809
QM
3244 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3245 * out the format string.
492e639f
YS
3246 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3247 * the format string itself. The *data* and *data_len* are format string
ab8d7809 3248 * arguments. The *data* are a **u64** array and corresponding format string
492e639f
YS
3249 * values are stored in the array. For strings and pointers where pointees
3250 * are accessed, only the pointer values are stored in the *data* array.
ab8d7809 3251 * The *data_len* is the size of *data* in bytes.
492e639f
YS
3252 *
3253 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3254 * Reading kernel memory may fail due to either invalid address or
3255 * valid address but requiring a major memory fault. If reading kernel memory
3256 * fails, the string for **%s** will be an empty string, and the ip
3257 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
ab8d7809 3258 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
492e639f 3259 * Return
ab8d7809
QM
3260 * 0 on success, or a negative error in case of failure:
3261 *
3262 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3263 * by returning 1 from bpf program.
3264 *
3265 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3266 *
3267 * **-E2BIG** if *fmt* contains too many format specifiers.
492e639f 3268 *
ab8d7809 3269 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
492e639f 3270 *
bdb7b79b 3271 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
492e639f 3272 * Description
ab8d7809 3273 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
492e639f 3274 * The *m* represents the seq_file. The *data* and *len* represent the
ab8d7809 3275 * data to write in bytes.
492e639f 3276 * Return
ab8d7809 3277 * 0 on success, or a negative error in case of failure:
492e639f 3278 *
ab8d7809 3279 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
f307fa2c 3280 *
a5fa25ad 3281 * u64 bpf_sk_cgroup_id(void *sk)
f307fa2c
AI
3282 * Description
3283 * Return the cgroup v2 id of the socket *sk*.
3284 *
a5fa25ad 3285 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
f307fa2c
AI
3286 * returned from **bpf_sk_lookup_xxx**\ (),
3287 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3288 * same as in **bpf_skb_cgroup_id**\ ().
3289 *
3290 * This helper is available only if the kernel was compiled with
3291 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3292 * Return
3293 * The id is returned or 0 in case the id could not be retrieved.
3294 *
a5fa25ad 3295 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
f307fa2c
AI
3296 * Description
3297 * Return id of cgroup v2 that is ancestor of cgroup associated
3298 * with the *sk* at the *ancestor_level*. The root cgroup is at
3299 * *ancestor_level* zero and each step down the hierarchy
3300 * increments the level. If *ancestor_level* == level of cgroup
3301 * associated with *sk*, then return value will be same as that
3302 * of **bpf_sk_cgroup_id**\ ().
3303 *
3304 * The helper is useful to implement policies based on cgroups
3305 * that are upper in hierarchy than immediate cgroup associated
3306 * with *sk*.
3307 *
3308 * The format of returned id and helper limitations are same as in
3309 * **bpf_sk_cgroup_id**\ ().
3310 * Return
3311 * The id is returned or 0 in case the id could not be retrieved.
457f4436 3312 *
e1613b57 3313 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
457f4436
AN
3314 * Description
3315 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
bcc7f554
QM
3316 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3317 * of new data availability is sent.
3318 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3319 * of new data availability is sent unconditionally.
457f4436 3320 * Return
bcc7f554 3321 * 0 on success, or a negative error in case of failure.
457f4436
AN
3322 *
3323 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3324 * Description
3325 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3326 * Return
3327 * Valid pointer with *size* bytes of memory available; NULL,
3328 * otherwise.
3329 *
3330 * void bpf_ringbuf_submit(void *data, u64 flags)
3331 * Description
3332 * Submit reserved ring buffer sample, pointed to by *data*.
bcc7f554
QM
3333 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3334 * of new data availability is sent.
3335 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3336 * of new data availability is sent unconditionally.
457f4436
AN
3337 * Return
3338 * Nothing. Always succeeds.
3339 *
3340 * void bpf_ringbuf_discard(void *data, u64 flags)
3341 * Description
3342 * Discard reserved ring buffer sample, pointed to by *data*.
bcc7f554
QM
3343 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
3344 * of new data availability is sent.
3345 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
3346 * of new data availability is sent unconditionally.
457f4436
AN
3347 * Return
3348 * Nothing. Always succeeds.
3349 *
3350 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3351 * Description
3352 * Query various characteristics of provided ring buffer. What
3353 * exactly is queries is determined by *flags*:
bcc7f554
QM
3354 *
3355 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
3356 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
3357 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
3358 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
3359 *
3360 * Data returned is just a momentary snapshot of actual values
457f4436
AN
3361 * and could be inaccurate, so this facility should be used to
3362 * power heuristics and for reporting, not to make 100% correct
3363 * calculation.
3364 * Return
bcc7f554 3365 * Requested value, or 0, if *flags* are not recognized.
7cdec54f 3366 *
bdb7b79b 3367 * long bpf_csum_level(struct sk_buff *skb, u64 level)
7cdec54f
DB
3368 * Description
3369 * Change the skbs checksum level by one layer up or down, or
3370 * reset it entirely to none in order to have the stack perform
3371 * checksum validation. The level is applicable to the following
3372 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
3373 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
3374 * through **bpf_skb_adjust_room**\ () helper with passing in
3375 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
3376 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
3377 * the UDP header is removed. Similarly, an encap of the latter
3378 * into the former could be accompanied by a helper call to
3379 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
3380 * skb is still intended to be processed in higher layers of the
3381 * stack instead of just egressing at tc.
3382 *
3383 * There are three supported level settings at this time:
3384 *
3385 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
3386 * with CHECKSUM_UNNECESSARY.
3387 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
3388 * with CHECKSUM_UNNECESSARY.
3389 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
3390 * sets CHECKSUM_NONE to force checksum validation by the stack.
3391 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
3392 * skb->csum_level.
3393 * Return
3394 * 0 on success, or a negative error in case of failure. In the
3395 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
3396 * is returned or the error code -EACCES in case the skb is not
3397 * subject to CHECKSUM_UNNECESSARY.
af7ec138
YS
3398 *
3399 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
3400 * Description
3401 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
3402 * Return
938c3efd 3403 * *sk* if casting is valid, or **NULL** otherwise.
478cfbdf
YS
3404 *
3405 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
3406 * Description
3407 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
3408 * Return
938c3efd 3409 * *sk* if casting is valid, or **NULL** otherwise.
478cfbdf
YS
3410 *
3411 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
3412 * Description
3413 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
3414 * Return
938c3efd 3415 * *sk* if casting is valid, or **NULL** otherwise.
478cfbdf
YS
3416 *
3417 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
3418 * Description
3419 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
3420 * Return
938c3efd 3421 * *sk* if casting is valid, or **NULL** otherwise.
0d4fad3e
YS
3422 *
3423 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
3424 * Description
3425 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
3426 * Return
938c3efd 3427 * *sk* if casting is valid, or **NULL** otherwise.
fa28dcb8
SL
3428 *
3429 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
3430 * Description
3431 * Return a user or a kernel stack in bpf program provided buffer.
3432 * To achieve this, the helper needs *task*, which is a valid
938c3efd
QM
3433 * pointer to **struct task_struct**. To store the stacktrace, the
3434 * bpf program provides *buf* with a nonnegative *size*.
fa28dcb8
SL
3435 *
3436 * The last argument, *flags*, holds the number of stack frames to
3437 * skip (from 0 to 255), masked with
3438 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3439 * the following flags:
3440 *
3441 * **BPF_F_USER_STACK**
3442 * Collect a user space stack instead of a kernel stack.
3443 * **BPF_F_USER_BUILD_ID**
3444 * Collect buildid+offset instead of ips for user stack,
3445 * only valid if **BPF_F_USER_STACK** is also specified.
3446 *
3447 * **bpf_get_task_stack**\ () can collect up to
3448 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3449 * to sufficient large buffer size. Note that
3450 * this limit can be controlled with the **sysctl** program, and
3451 * that it should be manually increased in order to profile long
3452 * user stacks (such as stacks for Java programs). To do so, use:
3453 *
3454 * ::
3455 *
3456 * # sysctl kernel.perf_event_max_stack=<new value>
3457 * Return
3458 * A non-negative value equal to or less than *size* on success,
3459 * or a negative error in case of failure.
3460 *
0813a841
MKL
3461 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
3462 * Description
3463 * Load header option. Support reading a particular TCP header
938c3efd 3464 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
0813a841
MKL
3465 *
3466 * If *flags* is 0, it will search the option from the
938c3efd 3467 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
0813a841 3468 * has details on what skb_data contains under different
938c3efd 3469 * *skops*\ **->op**.
0813a841
MKL
3470 *
3471 * The first byte of the *searchby_res* specifies the
3472 * kind that it wants to search.
3473 *
3474 * If the searching kind is an experimental kind
3475 * (i.e. 253 or 254 according to RFC6994). It also
3476 * needs to specify the "magic" which is either
3477 * 2 bytes or 4 bytes. It then also needs to
3478 * specify the size of the magic by using
3479 * the 2nd byte which is "kind-length" of a TCP
3480 * header option and the "kind-length" also
3481 * includes the first 2 bytes "kind" and "kind-length"
3482 * itself as a normal TCP header option also does.
3483 *
3484 * For example, to search experimental kind 254 with
3485 * 2 byte magic 0xeB9F, the searchby_res should be
3486 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
3487 *
3488 * To search for the standard window scale option (3),
938c3efd 3489 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
0813a841
MKL
3490 * Note, kind-length must be 0 for regular option.
3491 *
3492 * Searching for No-Op (0) and End-of-Option-List (1) are
3493 * not supported.
3494 *
3495 * *len* must be at least 2 bytes which is the minimal size
3496 * of a header option.
3497 *
3498 * Supported flags:
938c3efd 3499 *
0813a841
MKL
3500 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
3501 * saved_syn packet or the just-received syn packet.
3502 *
3503 * Return
938c3efd
QM
3504 * > 0 when found, the header option is copied to *searchby_res*.
3505 * The return value is the total length copied. On failure, a
3506 * negative error code is returned:
0813a841 3507 *
938c3efd 3508 * **-EINVAL** if a parameter is invalid.
0813a841 3509 *
938c3efd 3510 * **-ENOMSG** if the option is not found.
0813a841 3511 *
938c3efd
QM
3512 * **-ENOENT** if no syn packet is available when
3513 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
0813a841 3514 *
938c3efd
QM
3515 * **-ENOSPC** if there is not enough space. Only *len* number of
3516 * bytes are copied.
0813a841 3517 *
938c3efd
QM
3518 * **-EFAULT** on failure to parse the header options in the
3519 * packet.
0813a841 3520 *
938c3efd
QM
3521 * **-EPERM** if the helper cannot be used under the current
3522 * *skops*\ **->op**.
0813a841
MKL
3523 *
3524 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
3525 * Description
3526 * Store header option. The data will be copied
3527 * from buffer *from* with length *len* to the TCP header.
3528 *
3529 * The buffer *from* should have the whole option that
3530 * includes the kind, kind-length, and the actual
3531 * option data. The *len* must be at least kind-length
3532 * long. The kind-length does not have to be 4 byte
3533 * aligned. The kernel will take care of the padding
3534 * and setting the 4 bytes aligned value to th->doff.
3535 *
3536 * This helper will check for duplicated option
3537 * by searching the same option in the outgoing skb.
3538 *
3539 * This helper can only be called during
938c3efd 3540 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
0813a841
MKL
3541 *
3542 * Return
3543 * 0 on success, or negative error in case of failure:
3544 *
938c3efd 3545 * **-EINVAL** If param is invalid.
0813a841 3546 *
938c3efd
QM
3547 * **-ENOSPC** if there is not enough space in the header.
3548 * Nothing has been written
0813a841 3549 *
938c3efd 3550 * **-EEXIST** if the option already exists.
0813a841 3551 *
938c3efd 3552 * **-EFAULT** on failrue to parse the existing header options.
0813a841 3553 *
938c3efd
QM
3554 * **-EPERM** if the helper cannot be used under the current
3555 * *skops*\ **->op**.
0813a841
MKL
3556 *
3557 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
3558 * Description
3559 * Reserve *len* bytes for the bpf header option. The
938c3efd
QM
3560 * space will be used by **bpf_store_hdr_opt**\ () later in
3561 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
0813a841 3562 *
938c3efd 3563 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
0813a841
MKL
3564 * the total number of bytes will be reserved.
3565 *
3566 * This helper can only be called during
938c3efd 3567 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
0813a841
MKL
3568 *
3569 * Return
3570 * 0 on success, or negative error in case of failure:
3571 *
938c3efd 3572 * **-EINVAL** if a parameter is invalid.
0813a841 3573 *
938c3efd 3574 * **-ENOSPC** if there is not enough space in the header.
0813a841 3575 *
938c3efd
QM
3576 * **-EPERM** if the helper cannot be used under the current
3577 * *skops*\ **->op**.
6e22ab9d 3578 *
8ea63684
KS
3579 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
3580 * Description
3581 * Get a bpf_local_storage from an *inode*.
3582 *
3583 * Logically, it could be thought of as getting the value from
3584 * a *map* with *inode* as the **key**. From this
3585 * perspective, the usage is not much different from
3586 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
3587 * helper enforces the key must be an inode and the map must also
3588 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
3589 *
3590 * Underneath, the value is stored locally at *inode* instead of
3591 * the *map*. The *map* is used as the bpf-local-storage
3592 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3593 * searched against all bpf_local_storage residing at *inode*.
3594 *
3595 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3596 * used such that a new bpf_local_storage will be
3597 * created if one does not exist. *value* can be used
3598 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3599 * the initial value of a bpf_local_storage. If *value* is
3600 * **NULL**, the new bpf_local_storage will be zero initialized.
3601 * Return
3602 * A bpf_local_storage pointer is returned on success.
3603 *
3604 * **NULL** if not found or there was an error in adding
3605 * a new bpf_local_storage.
3606 *
3607 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
3608 * Description
3609 * Delete a bpf_local_storage from an *inode*.
3610 * Return
3611 * 0 on success.
3612 *
3613 * **-ENOENT** if the bpf_local_storage cannot be found.
6e22ab9d
JO
3614 *
3615 * long bpf_d_path(struct path *path, char *buf, u32 sz)
3616 * Description
938c3efd
QM
3617 * Return full path for given **struct path** object, which
3618 * needs to be the kernel BTF *path* object. The path is
3619 * returned in the provided buffer *buf* of size *sz* and
6e22ab9d
JO
3620 * is zero terminated.
3621 *
3622 * Return
3623 * On success, the strictly positive length of the string,
3624 * including the trailing NUL character. On error, a negative
3625 * value.
07be4c4a
AS
3626 *
3627 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
3628 * Description
3629 * Read *size* bytes from user space address *user_ptr* and store
938c3efd 3630 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
07be4c4a
AS
3631 * Return
3632 * 0 on success, or a negative error in case of failure.
c4d0bfb4
AM
3633 *
3634 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
3635 * Description
3636 * Use BTF to store a string representation of *ptr*->ptr in *str*,
3637 * using *ptr*->type_id. This value should specify the type
3638 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
3639 * can be used to look up vmlinux BTF type ids. Traversing the
3640 * data structure using BTF, the type information and values are
3641 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
3642 * the pointer data is carried out to avoid kernel crashes during
3643 * operation. Smaller types can use string space on the stack;
3644 * larger programs can use map data to store the string
3645 * representation.
3646 *
3647 * The string can be subsequently shared with userspace via
3648 * bpf_perf_event_output() or ring buffer interfaces.
3649 * bpf_trace_printk() is to be avoided as it places too small
3650 * a limit on string size to be useful.
3651 *
3652 * *flags* is a combination of
3653 *
3654 * **BTF_F_COMPACT**
3655 * no formatting around type information
3656 * **BTF_F_NONAME**
3657 * no struct/union member names/types
3658 * **BTF_F_PTR_RAW**
3659 * show raw (unobfuscated) pointer values;
3660 * equivalent to printk specifier %px.
3661 * **BTF_F_ZERO**
3662 * show zero-valued struct/union members; they
3663 * are not displayed by default
3664 *
3665 * Return
3666 * The number of bytes that were written (or would have been
3667 * written if output had to be truncated due to string size),
3668 * or a negative error in cases of failure.
eb411377
AM
3669 *
3670 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
3671 * Description
3672 * Use BTF to write to seq_write a string representation of
3673 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
3674 * *flags* are identical to those used for bpf_snprintf_btf.
3675 * Return
3676 * 0 on success or a negative error in case of failure.
b426ce83
DB
3677 *
3678 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
3679 * Description
3680 * See **bpf_get_cgroup_classid**\ () for the main description.
3681 * This helper differs from **bpf_get_cgroup_classid**\ () in that
3682 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
3683 * associated socket instead of the current process.
3684 * Return
3685 * The id is returned or 0 in case the id could not be retrieved.
b4ab3141 3686 *
ba452c9e 3687 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
b4ab3141
DB
3688 * Description
3689 * Redirect the packet to another net device of index *ifindex*
3690 * and fill in L2 addresses from neighboring subsystem. This helper
3691 * is somewhat similar to **bpf_redirect**\ (), except that it
dd2ce6a5 3692 * populates L2 addresses as well, meaning, internally, the helper
ba452c9e
THJ
3693 * relies on the neighbor lookup for the L2 address of the nexthop.
3694 *
3695 * The helper will perform a FIB lookup based on the skb's
3696 * networking header to get the address of the next hop, unless
3697 * this is supplied by the caller in the *params* argument. The
3698 * *plen* argument indicates the len of *params* and should be set
3699 * to 0 if *params* is NULL.
dd2ce6a5 3700 *
b4ab3141 3701 * The *flags* argument is reserved and must be 0. The helper is
dd2ce6a5
DB
3702 * currently only supported for tc BPF program types, and enabled
3703 * for IPv4 and IPv6 protocols.
b4ab3141
DB
3704 * Return
3705 * The helper returns **TC_ACT_REDIRECT** on success or
3706 * **TC_ACT_SHOT** on error.
eaa6bcb7
HL
3707 *
3708 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
3709 * Description
3710 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3711 * pointer to the percpu kernel variable on *cpu*. A ksym is an
3712 * extern variable decorated with '__ksym'. For ksym, there is a
3713 * global var (either static or global) defined of the same name
3714 * in the kernel. The ksym is percpu if the global var is percpu.
3715 * The returned pointer points to the global percpu var on *cpu*.
3716 *
3717 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
3718 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
3719 * happens if *cpu* is larger than nr_cpu_ids. The caller of
3720 * bpf_per_cpu_ptr() must check the returned value.
3721 * Return
3722 * A pointer pointing to the kernel percpu variable on *cpu*, or
3723 * NULL, if *cpu* is invalid.
63d9b80d
HL
3724 *
3725 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
3726 * Description
3727 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
3728 * pointer to the percpu kernel variable on this cpu. See the
3729 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
3730 *
3731 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
3732 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
3733 * never return NULL.
3734 * Return
3735 * A pointer pointing to the kernel percpu variable on this cpu.
9aa1206e
DB
3736 *
3737 * long bpf_redirect_peer(u32 ifindex, u64 flags)
3738 * Description
3739 * Redirect the packet to another net device of index *ifindex*.
3740 * This helper is somewhat similar to **bpf_redirect**\ (), except
3741 * that the redirection happens to the *ifindex*' peer device and
3742 * the netns switch takes place from ingress to ingress without
3743 * going through the CPU's backlog queue.
3744 *
3745 * The *flags* argument is reserved and must be 0. The helper is
3746 * currently only supported for tc BPF program types at the ingress
3747 * hook and for veth device types. The peer device must reside in a
3748 * different network namespace.
3749 * Return
3750 * The helper returns **TC_ACT_REDIRECT** on success or
3751 * **TC_ACT_SHOT** on error.
4cf1bc1f
KS
3752 *
3753 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
3754 * Description
3755 * Get a bpf_local_storage from the *task*.
3756 *
3757 * Logically, it could be thought of as getting the value from
3758 * a *map* with *task* as the **key**. From this
3759 * perspective, the usage is not much different from
3760 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
3761 * helper enforces the key must be an task_struct and the map must also
3762 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
3763 *
3764 * Underneath, the value is stored locally at *task* instead of
3765 * the *map*. The *map* is used as the bpf-local-storage
3766 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3767 * searched against all bpf_local_storage residing at *task*.
3768 *
3769 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
3770 * used such that a new bpf_local_storage will be
3771 * created if one does not exist. *value* can be used
3772 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
3773 * the initial value of a bpf_local_storage. If *value* is
3774 * **NULL**, the new bpf_local_storage will be zero initialized.
3775 * Return
3776 * A bpf_local_storage pointer is returned on success.
3777 *
3778 * **NULL** if not found or there was an error in adding
3779 * a new bpf_local_storage.
3780 *
3781 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
3782 * Description
3783 * Delete a bpf_local_storage from a *task*.
3784 * Return
3785 * 0 on success.
3786 *
3787 * **-ENOENT** if the bpf_local_storage cannot be found.
3ca1032a
KS
3788 *
3789 * struct task_struct *bpf_get_current_task_btf(void)
3790 * Description
3791 * Return a BTF pointer to the "current" task.
3792 * This pointer can also be used in helpers that accept an
3793 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
3794 * Return
3795 * Pointer to the current task.
3f6719c7
KS
3796 *
3797 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
3798 * Description
3799 * Set or clear certain options on *bprm*:
3800 *
3801 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
3802 * which sets the **AT_SECURE** auxv for glibc. The bit
3803 * is cleared if the flag is not specified.
3804 * Return
3805 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
d0551261
DB
3806 *
3807 * u64 bpf_ktime_get_coarse_ns(void)
3808 * Description
3809 * Return a coarse-grained version of the time elapsed since
3810 * system boot, in nanoseconds. Does not include time the system
3811 * was suspended.
3812 *
3813 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
3814 * Return
3815 * Current *ktime*.
27672f0d
KS
3816 *
3817 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
3818 * Description
3819 * Returns the stored IMA hash of the *inode* (if it's avaialable).
3820 * If the hash is larger than *size*, then only *size*
3821 * bytes will be copied to *dst*
3822 * Return
3823 * The **hash_algo** is returned on success,
3824 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
3825 * invalid arguments are passed.
4f19cab7
FR
3826 *
3827 * struct socket *bpf_sock_from_file(struct file *file)
3828 * Description
3829 * If the given file represents a socket, returns the associated
3830 * socket.
3831 * Return
3832 * A pointer to a struct socket on success or NULL if the file is
3833 * not a socket.
ebb676da
TG
3834 */
3835#define __BPF_FUNC_MAPPER(FN) \
3836 FN(unspec), \
3837 FN(map_lookup_elem), \
3838 FN(map_update_elem), \
3839 FN(map_delete_elem), \
3840 FN(probe_read), \
3841 FN(ktime_get_ns), \
3842 FN(trace_printk), \
3843 FN(get_prandom_u32), \
3844 FN(get_smp_processor_id), \
3845 FN(skb_store_bytes), \
3846 FN(l3_csum_replace), \
3847 FN(l4_csum_replace), \
3848 FN(tail_call), \
3849 FN(clone_redirect), \
3850 FN(get_current_pid_tgid), \
3851 FN(get_current_uid_gid), \
3852 FN(get_current_comm), \
3853 FN(get_cgroup_classid), \
3854 FN(skb_vlan_push), \
3855 FN(skb_vlan_pop), \
3856 FN(skb_get_tunnel_key), \
3857 FN(skb_set_tunnel_key), \
3858 FN(perf_event_read), \
3859 FN(redirect), \
3860 FN(get_route_realm), \
3861 FN(perf_event_output), \
3862 FN(skb_load_bytes), \
3863 FN(get_stackid), \
3864 FN(csum_diff), \
3865 FN(skb_get_tunnel_opt), \
3866 FN(skb_set_tunnel_opt), \
3867 FN(skb_change_proto), \
3868 FN(skb_change_type), \
3869 FN(skb_under_cgroup), \
3870 FN(get_hash_recalc), \
3871 FN(get_current_task), \
3872 FN(probe_write_user), \
3873 FN(current_task_under_cgroup), \
3874 FN(skb_change_tail), \
3875 FN(skb_pull_data), \
3876 FN(csum_update), \
3877 FN(set_hash_invalid), \
3a0af8fd 3878 FN(get_numa_node_id), \
17bedab2 3879 FN(skb_change_head), \
a5e8c070 3880 FN(xdp_adjust_head), \
91b8270f 3881 FN(probe_read_str), \
6acc5c29 3882 FN(get_socket_cookie), \
ded092cd 3883 FN(get_socket_uid), \
8c4b4c7e 3884 FN(set_hash), \
2be7e212 3885 FN(setsockopt), \
97f91a7c 3886 FN(skb_adjust_room), \
174a79ff
JF
3887 FN(redirect_map), \
3888 FN(sk_redirect_map), \
3889 FN(sock_map_update), \
908432ca 3890 FN(xdp_adjust_meta), \
4bebdc7a 3891 FN(perf_event_read_value), \
cd86d1fd 3892 FN(perf_prog_read_value), \
9802d865 3893 FN(getsockopt), \
b13d8807 3894 FN(override_return), \
4f738adb 3895 FN(sock_ops_cb_flags_set), \
2a100317 3896 FN(msg_redirect_map), \
91843d54 3897 FN(msg_apply_bytes), \
015632bb 3898 FN(msg_cork_bytes), \
d74bad4e 3899 FN(msg_pull_data), \
b32cc5b9 3900 FN(bind), \
12bed760 3901 FN(xdp_adjust_tail), \
c195651e 3902 FN(skb_get_xfrm_state), \
4e1ec56c 3903 FN(get_stack), \
87f5fc7e 3904 FN(skb_load_bytes_relative), \
81110384
JF
3905 FN(fib_lookup), \
3906 FN(sock_hash_update), \
3907 FN(msg_redirect_hash), \
fe94cc29
MX
3908 FN(sk_redirect_hash), \
3909 FN(lwt_push_encap), \
3910 FN(lwt_seg6_store_bytes), \
3911 FN(lwt_seg6_adjust_srh), \
f4364dcf
SY
3912 FN(lwt_seg6_action), \
3913 FN(rc_repeat), \
cb20b08e 3914 FN(rc_keydown), \
bf6fa2c8 3915 FN(skb_cgroup_id), \
cd339431 3916 FN(get_current_cgroup_id), \
2dbb9b9e 3917 FN(get_local_storage), \
77236281 3918 FN(sk_select_reuseport), \
6acc9b43
JS
3919 FN(skb_ancestor_cgroup_id), \
3920 FN(sk_lookup_tcp), \
3921 FN(sk_lookup_udp), \
f1a2e44a
MV
3922 FN(sk_release), \
3923 FN(map_push_elem), \
3924 FN(map_pop_elem), \
6fff607e 3925 FN(map_peek_elem), \
7246d8ed 3926 FN(msg_push_data), \
01d3240a 3927 FN(msg_pop_data), \
d83525ca
AS
3928 FN(rc_pointer_rel), \
3929 FN(spin_lock), \
46f8bc92 3930 FN(spin_unlock), \
655a51e5 3931 FN(sk_fullsock), \
f7c917ba 3932 FN(tcp_sock), \
dbafd7dd 3933 FN(skb_ecn_set_ce), \
edbf8c01 3934 FN(get_listener_sock), \
39904084 3935 FN(skc_lookup_tcp), \
808649fb 3936 FN(tcp_check_syncookie), \
1d11b301 3937 FN(sysctl_get_name), \
4e63acdf
AI
3938 FN(sysctl_get_current_value), \
3939 FN(sysctl_get_new_value), \
d7a4cb9b
AI
3940 FN(sysctl_set_new_value), \
3941 FN(strtol), \
6ac99e8f
MKL
3942 FN(strtoul), \
3943 FN(sk_storage_get), \
8b401f9e 3944 FN(sk_storage_delete), \
70d66244 3945 FN(send_signal), \
a7658e1a 3946 FN(tcp_gen_syncookie), \
6ae08ae3
DB
3947 FN(skb_output), \
3948 FN(probe_read_user), \
3949 FN(probe_read_kernel), \
3950 FN(probe_read_user_str), \
206057fe 3951 FN(probe_read_kernel_str), \
8482941f 3952 FN(tcp_send_ack), \
5576b991 3953 FN(send_signal_thread), \
fff7b643 3954 FN(jiffies64), \
b4490c5c 3955 FN(read_branch_records), \
d831ee84 3956 FN(get_ns_current_pid_tgid), \
f318903c 3957 FN(xdp_output), \
0f09abd1 3958 FN(get_netns_cookie), \
cf7fbe66 3959 FN(get_current_ancestor_cgroup_id), \
71d19214 3960 FN(sk_assign), \
492e639f
YS
3961 FN(ktime_get_boot_ns), \
3962 FN(seq_printf), \
f307fa2c
AI
3963 FN(seq_write), \
3964 FN(sk_cgroup_id), \
457f4436
AN
3965 FN(sk_ancestor_cgroup_id), \
3966 FN(ringbuf_output), \
3967 FN(ringbuf_reserve), \
3968 FN(ringbuf_submit), \
3969 FN(ringbuf_discard), \
7cdec54f 3970 FN(ringbuf_query), \
af7ec138 3971 FN(csum_level), \
478cfbdf
YS
3972 FN(skc_to_tcp6_sock), \
3973 FN(skc_to_tcp_sock), \
3974 FN(skc_to_tcp_timewait_sock), \
0d4fad3e 3975 FN(skc_to_tcp_request_sock), \
fa28dcb8
SL
3976 FN(skc_to_udp6_sock), \
3977 FN(get_task_stack), \
0813a841
MKL
3978 FN(load_hdr_opt), \
3979 FN(store_hdr_opt), \
8ea63684
KS
3980 FN(reserve_hdr_opt), \
3981 FN(inode_storage_get), \
3982 FN(inode_storage_delete), \
6e22ab9d 3983 FN(d_path), \
07be4c4a 3984 FN(copy_from_user), \
c4d0bfb4 3985 FN(snprintf_btf), \
eb411377 3986 FN(seq_printf_btf), \
b426ce83 3987 FN(skb_cgroup_classid), \
b4ab3141 3988 FN(redirect_neigh), \
b7906b70
AN
3989 FN(per_cpu_ptr), \
3990 FN(this_cpu_ptr), \
9aa1206e 3991 FN(redirect_peer), \
4cf1bc1f
KS
3992 FN(task_storage_get), \
3993 FN(task_storage_delete), \
3ca1032a 3994 FN(get_current_task_btf), \
3f6719c7 3995 FN(bprm_opts_set), \
d0551261 3996 FN(ktime_get_coarse_ns), \
27672f0d 3997 FN(ima_inode_hash), \
4f19cab7 3998 FN(sock_from_file), \
fa28dcb8 3999 /* */
ebb676da 4000
09756af4
AS
4001/* integer value in 'imm' field of BPF_CALL instruction selects which helper
4002 * function eBPF program intends to call
4003 */
ebb676da 4004#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
09756af4 4005enum bpf_func_id {
ebb676da 4006 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
09756af4
AS
4007 __BPF_FUNC_MAX_ID,
4008};
ebb676da 4009#undef __BPF_ENUM_FN
09756af4 4010
781c53bc
DB
4011/* All flags used by eBPF helper functions, placed here. */
4012
4013/* BPF_FUNC_skb_store_bytes flags. */
1aae4bdd
AN
4014enum {
4015 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
4016 BPF_F_INVALIDATE_HASH = (1ULL << 1),
4017};
781c53bc
DB
4018
4019/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
4020 * First 4 bits are for passing the header field size.
4021 */
1aae4bdd
AN
4022enum {
4023 BPF_F_HDR_FIELD_MASK = 0xfULL,
4024};
781c53bc
DB
4025
4026/* BPF_FUNC_l4_csum_replace flags. */
1aae4bdd
AN
4027enum {
4028 BPF_F_PSEUDO_HDR = (1ULL << 4),
4029 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
4030 BPF_F_MARK_ENFORCE = (1ULL << 6),
4031};
781c53bc
DB
4032
4033/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
1aae4bdd
AN
4034enum {
4035 BPF_F_INGRESS = (1ULL << 0),
4036};
781c53bc 4037
c6c33454 4038/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
1aae4bdd
AN
4039enum {
4040 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
4041};
c6c33454 4042
c195651e 4043/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
1aae4bdd
AN
4044enum {
4045 BPF_F_SKIP_FIELD_MASK = 0xffULL,
4046 BPF_F_USER_STACK = (1ULL << 8),
c195651e 4047/* flags used by BPF_FUNC_get_stackid only. */
1aae4bdd
AN
4048 BPF_F_FAST_STACK_CMP = (1ULL << 9),
4049 BPF_F_REUSE_STACKID = (1ULL << 10),
c195651e 4050/* flags used by BPF_FUNC_get_stack only. */
1aae4bdd
AN
4051 BPF_F_USER_BUILD_ID = (1ULL << 11),
4052};
d5a3b1f6 4053
2da897e5 4054/* BPF_FUNC_skb_set_tunnel_key flags. */
1aae4bdd
AN
4055enum {
4056 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
4057 BPF_F_DONT_FRAGMENT = (1ULL << 2),
4058 BPF_F_SEQ_NUMBER = (1ULL << 3),
4059};
2da897e5 4060
908432ca
YS
4061/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
4062 * BPF_FUNC_perf_event_read_value flags.
4063 */
1aae4bdd
AN
4064enum {
4065 BPF_F_INDEX_MASK = 0xffffffffULL,
4066 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
555c8a86 4067/* BPF_FUNC_perf_event_output for sk_buff input context. */
1aae4bdd
AN
4068 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
4069};
1e33759c 4070
f71c6143 4071/* Current network namespace */
1aae4bdd
AN
4072enum {
4073 BPF_F_CURRENT_NETNS = (-1L),
4074};
f71c6143 4075
7cdec54f
DB
4076/* BPF_FUNC_csum_level level values. */
4077enum {
4078 BPF_CSUM_LEVEL_QUERY,
4079 BPF_CSUM_LEVEL_INC,
4080 BPF_CSUM_LEVEL_DEC,
4081 BPF_CSUM_LEVEL_RESET,
4082};
4083
2278f6cc 4084/* BPF_FUNC_skb_adjust_room flags. */
1aae4bdd
AN
4085enum {
4086 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
4087 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
4088 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
4089 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
4090 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
836e66c2 4091 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
1aae4bdd 4092};
2278f6cc 4093
1aae4bdd
AN
4094enum {
4095 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
4096 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
4097};
58dfc900 4098
bfb35c27 4099#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
58dfc900
AM
4100 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
4101 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
868d5235 4102
808649fb 4103/* BPF_FUNC_sysctl_get_name flags. */
1aae4bdd
AN
4104enum {
4105 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
4106};
808649fb 4107
f836a56e 4108/* BPF_FUNC_<kernel_obj>_storage_get flags */
1aae4bdd 4109enum {
f836a56e
KS
4110 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
4111 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
4112 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
4113 */
4114 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
1aae4bdd 4115};
6ac99e8f 4116
fff7b643 4117/* BPF_FUNC_read_branch_records flags. */
1aae4bdd
AN
4118enum {
4119 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
4120};
fff7b643 4121
457f4436
AN
4122/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
4123 * BPF_FUNC_bpf_ringbuf_output flags.
4124 */
4125enum {
4126 BPF_RB_NO_WAKEUP = (1ULL << 0),
4127 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
4128};
4129
4130/* BPF_FUNC_bpf_ringbuf_query flags */
4131enum {
4132 BPF_RB_AVAIL_DATA = 0,
4133 BPF_RB_RING_SIZE = 1,
4134 BPF_RB_CONS_POS = 2,
4135 BPF_RB_PROD_POS = 3,
4136};
4137
4138/* BPF ring buffer constants */
4139enum {
4140 BPF_RINGBUF_BUSY_BIT = (1U << 31),
4141 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
4142 BPF_RINGBUF_HDR_SZ = 8,
4143};
4144
e9ddbb77
JS
4145/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
4146enum {
4147 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
4148 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
4149};
4150
2be7e212
DB
4151/* Mode for BPF_FUNC_skb_adjust_room helper. */
4152enum bpf_adj_room_mode {
4153 BPF_ADJ_ROOM_NET,
14aa3192 4154 BPF_ADJ_ROOM_MAC,
2be7e212
DB
4155};
4156
4e1ec56c
DB
4157/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
4158enum bpf_hdr_start_off {
4159 BPF_HDR_START_MAC,
4160 BPF_HDR_START_NET,
4161};
4162
fe94cc29
MX
4163/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
4164enum bpf_lwt_encap_mode {
4165 BPF_LWT_ENCAP_SEG6,
3e0bd37c
PO
4166 BPF_LWT_ENCAP_SEG6_INLINE,
4167 BPF_LWT_ENCAP_IP,
fe94cc29
MX
4168};
4169
3f6719c7
KS
4170/* Flags for bpf_bprm_opts_set helper */
4171enum {
4172 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
4173};
4174
b7df9ada
DB
4175#define __bpf_md_ptr(type, name) \
4176union { \
4177 type name; \
4178 __u64 :64; \
4179} __attribute__((aligned(8)))
4180
9bac3d6d
AS
4181/* user accessible mirror of in-kernel sk_buff.
4182 * new fields can only be added to the end of this structure
4183 */
4184struct __sk_buff {
4185 __u32 len;
4186 __u32 pkt_type;
4187 __u32 mark;
4188 __u32 queue_mapping;
c2497395
AS
4189 __u32 protocol;
4190 __u32 vlan_present;
4191 __u32 vlan_tci;
27cd5452 4192 __u32 vlan_proto;
bcad5718 4193 __u32 priority;
37e82c2f
AS
4194 __u32 ingress_ifindex;
4195 __u32 ifindex;
d691f9e8
AS
4196 __u32 tc_index;
4197 __u32 cb[5];
ba7591d8 4198 __u32 hash;
045efa82 4199 __u32 tc_classid;
969bf05e
AS
4200 __u32 data;
4201 __u32 data_end;
b1d9fc41 4202 __u32 napi_id;
8a31db56 4203
de8f3a83 4204 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
8a31db56
JF
4205 __u32 family;
4206 __u32 remote_ip4; /* Stored in network byte order */
4207 __u32 local_ip4; /* Stored in network byte order */
4208 __u32 remote_ip6[4]; /* Stored in network byte order */
4209 __u32 local_ip6[4]; /* Stored in network byte order */
4210 __u32 remote_port; /* Stored in network byte order */
4211 __u32 local_port; /* stored in host byte order */
de8f3a83
DB
4212 /* ... here. */
4213
4214 __u32 data_meta;
b7df9ada 4215 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
f11216b2 4216 __u64 tstamp;
e3da08d0 4217 __u32 wire_len;
d9ff286a 4218 __u32 gso_segs;
46f8bc92 4219 __bpf_md_ptr(struct bpf_sock *, sk);
cf62089b 4220 __u32 gso_size;
9bac3d6d
AS
4221};
4222
d3aa45ce
AS
4223struct bpf_tunnel_key {
4224 __u32 tunnel_id;
c6c33454
DB
4225 union {
4226 __u32 remote_ipv4;
4227 __u32 remote_ipv6[4];
4228 };
4229 __u8 tunnel_tos;
4230 __u8 tunnel_ttl;
1fbc2e0c 4231 __u16 tunnel_ext; /* Padding, future use. */
4018ab18 4232 __u32 tunnel_label;
d3aa45ce
AS
4233};
4234
12bed760
EB
4235/* user accessible mirror of in-kernel xfrm_state.
4236 * new fields can only be added to the end of this structure
4237 */
4238struct bpf_xfrm_state {
4239 __u32 reqid;
4240 __u32 spi; /* Stored in network byte order */
4241 __u16 family;
1fbc2e0c 4242 __u16 ext; /* Padding, future use. */
12bed760
EB
4243 union {
4244 __u32 remote_ipv4; /* Stored in network byte order */
4245 __u32 remote_ipv6[4]; /* Stored in network byte order */
4246 };
4247};
4248
3a0af8fd
TG
4249/* Generic BPF return codes which all BPF program types may support.
4250 * The values are binary compatible with their TC_ACT_* counter-part to
4251 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
4252 * programs.
4253 *
4254 * XDP is handled seprately, see XDP_*.
4255 */
4256enum bpf_ret_code {
4257 BPF_OK = 0,
4258 /* 1 reserved */
4259 BPF_DROP = 2,
4260 /* 3-6 reserved */
4261 BPF_REDIRECT = 7,
3e0bd37c
PO
4262 /* >127 are reserved for prog type specific return codes.
4263 *
4264 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
4265 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
4266 * changed and should be routed based on its new L3 header.
4267 * (This is an L3 redirect, as opposed to L2 redirect
4268 * represented by BPF_REDIRECT above).
4269 */
4270 BPF_LWT_REROUTE = 128,
3a0af8fd
TG
4271};
4272
61023658
DA
4273struct bpf_sock {
4274 __u32 bound_dev_if;
aa4c1037
DA
4275 __u32 family;
4276 __u32 type;
4277 __u32 protocol;
482dca93
DA
4278 __u32 mark;
4279 __u32 priority;
aa65d696
MKL
4280 /* IP address also allows 1 and 2 bytes access */
4281 __u32 src_ip4;
4282 __u32 src_ip6[4];
4283 __u32 src_port; /* host byte order */
4284 __u32 dst_port; /* network byte order */
4285 __u32 dst_ip4;
4286 __u32 dst_ip6[4];
4287 __u32 state;
c3c16f2e 4288 __s32 rx_queue_mapping;
61023658
DA
4289};
4290
655a51e5
MKL
4291struct bpf_tcp_sock {
4292 __u32 snd_cwnd; /* Sending congestion window */
4293 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
4294 __u32 rtt_min;
4295 __u32 snd_ssthresh; /* Slow start size threshold */
4296 __u32 rcv_nxt; /* What we want to receive next */
4297 __u32 snd_nxt; /* Next sequence we send */
4298 __u32 snd_una; /* First byte we want an ack for */
4299 __u32 mss_cache; /* Cached effective mss, not including SACKS */
4300 __u32 ecn_flags; /* ECN status bits. */
4301 __u32 rate_delivered; /* saved rate sample: packets delivered */
4302 __u32 rate_interval_us; /* saved rate sample: time elapsed */
4303 __u32 packets_out; /* Packets which are "in flight" */
4304 __u32 retrans_out; /* Retransmitted packets out */
4305 __u32 total_retrans; /* Total retransmits for entire connection */
4306 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
4307 * total number of segments in.
4308 */
4309 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
4310 * total number of data segments in.
4311 */
4312 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
4313 * The total number of segments sent.
4314 */
4315 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
4316 * total number of data segments sent.
4317 */
4318 __u32 lost_out; /* Lost packets */
4319 __u32 sacked_out; /* SACK'd packets */
4320 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
4321 * sum(delta(rcv_nxt)), or how many bytes
4322 * were acked.
4323 */
4324 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
4325 * sum(delta(snd_una)), or how many bytes
4326 * were acked.
4327 */
0357746d
SF
4328 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
4329 * total number of DSACK blocks received
4330 */
4331 __u32 delivered; /* Total data packets delivered incl. rexmits */
4332 __u32 delivered_ce; /* Like the above but only ECE marked packets */
c2cb5e82 4333 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
655a51e5
MKL
4334};
4335
6acc9b43
JS
4336struct bpf_sock_tuple {
4337 union {
4338 struct {
4339 __be32 saddr;
4340 __be32 daddr;
4341 __be16 sport;
4342 __be16 dport;
4343 } ipv4;
4344 struct {
4345 __be32 saddr[4];
4346 __be32 daddr[4];
4347 __be16 sport;
4348 __be16 dport;
4349 } ipv6;
4350 };
4351};
4352
fada7fdc
JL
4353struct bpf_xdp_sock {
4354 __u32 queue_id;
4355};
4356
17bedab2
MKL
4357#define XDP_PACKET_HEADROOM 256
4358
6a773a15
BB
4359/* User return codes for XDP prog type.
4360 * A valid XDP program must return one of these defined values. All other
9beb8bed
DB
4361 * return codes are reserved for future use. Unknown return codes will
4362 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6a773a15
BB
4363 */
4364enum xdp_action {
4365 XDP_ABORTED = 0,
4366 XDP_DROP,
4367 XDP_PASS,
6ce96ca3 4368 XDP_TX,
814abfab 4369 XDP_REDIRECT,
6a773a15
BB
4370};
4371
4372/* user accessible metadata for XDP packet hook
4373 * new fields must be added to the end of this structure
4374 */
4375struct xdp_md {
4376 __u32 data;
4377 __u32 data_end;
de8f3a83 4378 __u32 data_meta;
daaf24c6 4379 /* Below access go through struct xdp_rxq_info */
02dd3291
JDB
4380 __u32 ingress_ifindex; /* rxq->dev->ifindex */
4381 __u32 rx_queue_index; /* rxq->queue_index */
64b59025
DA
4382
4383 __u32 egress_ifindex; /* txq->dev->ifindex */
6a773a15
BB
4384};
4385
281920b7
JDB
4386/* DEVMAP map-value layout
4387 *
4388 * The struct data-layout of map-value is a configuration interface.
4389 * New members can only be added to the end of this structure.
4390 */
4391struct bpf_devmap_val {
4392 __u32 ifindex; /* device index */
4393 union {
4394 int fd; /* prog fd on map write */
4395 __u32 id; /* prog id on map read */
4396 } bpf_prog;
4397};
4398
644bfe51
LB
4399/* CPUMAP map-value layout
4400 *
4401 * The struct data-layout of map-value is a configuration interface.
4402 * New members can only be added to the end of this structure.
4403 */
4404struct bpf_cpumap_val {
4405 __u32 qsize; /* queue size to remote target CPU */
92164774
LB
4406 union {
4407 int fd; /* prog fd on map write */
4408 __u32 id; /* prog id on map read */
4409 } bpf_prog;
644bfe51
LB
4410};
4411
174a79ff 4412enum sk_action {
bfa64075
JF
4413 SK_DROP = 0,
4414 SK_PASS,
174a79ff
JF
4415};
4416
4f738adb
JF
4417/* user accessible metadata for SK_MSG packet hook, new fields must
4418 * be added to the end of this structure
4419 */
4420struct sk_msg_md {
b7df9ada
DB
4421 __bpf_md_ptr(void *, data);
4422 __bpf_md_ptr(void *, data_end);
303def35
JF
4423
4424 __u32 family;
4425 __u32 remote_ip4; /* Stored in network byte order */
4426 __u32 local_ip4; /* Stored in network byte order */
4427 __u32 remote_ip6[4]; /* Stored in network byte order */
4428 __u32 local_ip6[4]; /* Stored in network byte order */
4429 __u32 remote_port; /* Stored in network byte order */
4430 __u32 local_port; /* stored in host byte order */
3bdbd022 4431 __u32 size; /* Total size of sk_msg */
13d70f5a
JF
4432
4433 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
4f738adb
JF
4434};
4435
2dbb9b9e
MKL
4436struct sk_reuseport_md {
4437 /*
4438 * Start of directly accessible data. It begins from
4439 * the tcp/udp header.
4440 */
b7df9ada
DB
4441 __bpf_md_ptr(void *, data);
4442 /* End of directly accessible data */
4443 __bpf_md_ptr(void *, data_end);
2dbb9b9e
MKL
4444 /*
4445 * Total length of packet (starting from the tcp/udp header).
4446 * Note that the directly accessible bytes (data_end - data)
4447 * could be less than this "len". Those bytes could be
4448 * indirectly read by a helper "bpf_skb_load_bytes()".
4449 */
4450 __u32 len;
4451 /*
4452 * Eth protocol in the mac header (network byte order). e.g.
4453 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
4454 */
4455 __u32 eth_protocol;
4456 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
4457 __u32 bind_inany; /* Is sock bound to an INANY address? */
4458 __u32 hash; /* A hash of the packet 4 tuples */
4459};
4460
1e270976
MKL
4461#define BPF_TAG_SIZE 8
4462
4463struct bpf_prog_info {
4464 __u32 type;
4465 __u32 id;
4466 __u8 tag[BPF_TAG_SIZE];
4467 __u32 jited_prog_len;
4468 __u32 xlated_prog_len;
4469 __aligned_u64 jited_prog_insns;
4470 __aligned_u64 xlated_prog_insns;
cb4d2b3f
MKL
4471 __u64 load_time; /* ns since boottime */
4472 __u32 created_by_uid;
4473 __u32 nr_map_ids;
4474 __aligned_u64 map_ids;
067cae47 4475 char name[BPF_OBJ_NAME_LEN];
675fc275 4476 __u32 ifindex;
b85fab0e 4477 __u32 gpl_compatible:1;
0472301a 4478 __u32 :31; /* alignment pad */
675fc275
JK
4479 __u64 netns_dev;
4480 __u64 netns_ino;
dbecd738 4481 __u32 nr_jited_ksyms;
815581c1 4482 __u32 nr_jited_func_lens;
dbecd738 4483 __aligned_u64 jited_ksyms;
815581c1 4484 __aligned_u64 jited_func_lens;
838e9690
YS
4485 __u32 btf_id;
4486 __u32 func_info_rec_size;
4487 __aligned_u64 func_info;
11d8b82d
YS
4488 __u32 nr_func_info;
4489 __u32 nr_line_info;
c454a46b
MKL
4490 __aligned_u64 line_info;
4491 __aligned_u64 jited_line_info;
11d8b82d 4492 __u32 nr_jited_line_info;
c454a46b
MKL
4493 __u32 line_info_rec_size;
4494 __u32 jited_line_info_rec_size;
c872bdb3
SL
4495 __u32 nr_prog_tags;
4496 __aligned_u64 prog_tags;
5f8f8b93
AS
4497 __u64 run_time_ns;
4498 __u64 run_cnt;
1e270976
MKL
4499} __attribute__((aligned(8)));
4500
4501struct bpf_map_info {
4502 __u32 type;
4503 __u32 id;
4504 __u32 key_size;
4505 __u32 value_size;
4506 __u32 max_entries;
4507 __u32 map_flags;
067cae47 4508 char name[BPF_OBJ_NAME_LEN];
52775b33 4509 __u32 ifindex;
85d33df3 4510 __u32 btf_vmlinux_value_type_id;
52775b33
JK
4511 __u64 netns_dev;
4512 __u64 netns_ino;
78958fca 4513 __u32 btf_id;
9b2cf328
MKL
4514 __u32 btf_key_type_id;
4515 __u32 btf_value_type_id;
1e270976
MKL
4516} __attribute__((aligned(8)));
4517
62dab84c
MKL
4518struct bpf_btf_info {
4519 __aligned_u64 btf;
4520 __u32 btf_size;
4521 __u32 id;
53297220
AN
4522 __aligned_u64 name;
4523 __u32 name_len;
4524 __u32 kernel_btf;
62dab84c
MKL
4525} __attribute__((aligned(8)));
4526
f2e10bff
AN
4527struct bpf_link_info {
4528 __u32 type;
4529 __u32 id;
4530 __u32 prog_id;
4531 union {
4532 struct {
4533 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
4534 __u32 tp_name_len; /* in/out: tp_name buffer len */
4535 } raw_tracepoint;
4536 struct {
4537 __u32 attach_type;
4538 } tracing;
4539 struct {
4540 __u64 cgroup_id;
4541 __u32 attach_type;
4542 } cgroup;
6b0a249a
YS
4543 struct {
4544 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
4545 __u32 target_name_len; /* in/out: target_name buffer len */
4546 union {
b0c9eb37
YS
4547 struct {
4548 __u32 map_id;
4549 } map;
4550 };
6b0a249a 4551 } iter;
7f045a49
JS
4552 struct {
4553 __u32 netns_ino;
4554 __u32 attach_type;
4555 } netns;
c1931c97
AN
4556 struct {
4557 __u32 ifindex;
4558 } xdp;
f2e10bff
AN
4559 };
4560} __attribute__((aligned(8)));
4561
4fbac77d
AI
4562/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
4563 * by user and intended to be used by socket (e.g. to bind to, depends on
bfdfa517 4564 * attach type).
4fbac77d
AI
4565 */
4566struct bpf_sock_addr {
4567 __u32 user_family; /* Allows 4-byte read, but no write. */
4568 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
4569 * Stored in network byte order.
4570 */
d4ecfeb1 4571 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
4fbac77d
AI
4572 * Stored in network byte order.
4573 */
7aebfa1b 4574 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
4fbac77d
AI
4575 * Stored in network byte order
4576 */
4577 __u32 family; /* Allows 4-byte read, but no write */
4578 __u32 type; /* Allows 4-byte read, but no write */
4579 __u32 protocol; /* Allows 4-byte read, but no write */
600c70ba 4580 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
1cedee13
AI
4581 * Stored in network byte order.
4582 */
d4ecfeb1 4583 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
1cedee13
AI
4584 * Stored in network byte order.
4585 */
fb85c4a7 4586 __bpf_md_ptr(struct bpf_sock *, sk);
4fbac77d
AI
4587};
4588
40304b2a
LB
4589/* User bpf_sock_ops struct to access socket values and specify request ops
4590 * and their replies.
4591 * Some of this fields are in network (bigendian) byte order and may need
4592 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
4593 * New fields can only be added at the end of this structure
4594 */
4595struct bpf_sock_ops {
4596 __u32 op;
4597 union {
de525be2
LB
4598 __u32 args[4]; /* Optionally passed to bpf program */
4599 __u32 reply; /* Returned by bpf program */
4600 __u32 replylong[4]; /* Optionally returned by bpf prog */
40304b2a
LB
4601 };
4602 __u32 family;
4603 __u32 remote_ip4; /* Stored in network byte order */
4604 __u32 local_ip4; /* Stored in network byte order */
4605 __u32 remote_ip6[4]; /* Stored in network byte order */
4606 __u32 local_ip6[4]; /* Stored in network byte order */
4607 __u32 remote_port; /* Stored in network byte order */
4608 __u32 local_port; /* stored in host byte order */
f19397a5
LB
4609 __u32 is_fullsock; /* Some TCP fields are only valid if
4610 * there is a full socket. If not, the
4611 * fields read as zero.
4612 */
4613 __u32 snd_cwnd;
4614 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
b13d8807 4615 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
44f0e430
LB
4616 __u32 state;
4617 __u32 rtt_min;
4618 __u32 snd_ssthresh;
4619 __u32 rcv_nxt;
4620 __u32 snd_nxt;
4621 __u32 snd_una;
4622 __u32 mss_cache;
4623 __u32 ecn_flags;
4624 __u32 rate_delivered;
4625 __u32 rate_interval_us;
4626 __u32 packets_out;
4627 __u32 retrans_out;
4628 __u32 total_retrans;
4629 __u32 segs_in;
4630 __u32 data_segs_in;
4631 __u32 segs_out;
4632 __u32 data_segs_out;
4633 __u32 lost_out;
4634 __u32 sacked_out;
4635 __u32 sk_txhash;
4636 __u64 bytes_received;
4637 __u64 bytes_acked;
1314ef56 4638 __bpf_md_ptr(struct bpf_sock *, sk);
0813a841
MKL
4639 /* [skb_data, skb_data_end) covers the whole TCP header.
4640 *
4641 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
4642 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
4643 * header has not been written.
4644 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
4645 * been written so far.
4646 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
4647 * the 3WHS.
4648 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
4649 * the 3WHS.
4650 *
4651 * bpf_load_hdr_opt() can also be used to read a particular option.
4652 */
4653 __bpf_md_ptr(void *, skb_data);
4654 __bpf_md_ptr(void *, skb_data_end);
4655 __u32 skb_len; /* The total length of a packet.
4656 * It includes the header, options,
4657 * and payload.
4658 */
4659 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
4660 * an easy way to check for tcp_flags
4661 * without parsing skb_data.
4662 *
4663 * In particular, the skb_tcp_flags
4664 * will still be available in
4665 * BPF_SOCK_OPS_HDR_OPT_LEN even though
4666 * the outgoing header has not
4667 * been written yet.
4668 */
40304b2a
LB
4669};
4670
b13d8807 4671/* Definitions for bpf_sock_ops_cb_flags */
1aae4bdd
AN
4672enum {
4673 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
4674 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
4675 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
4676 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
0813a841
MKL
4677 /* Call bpf for all received TCP headers. The bpf prog will be
4678 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4679 *
4680 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4681 * for the header option related helpers that will be useful
4682 * to the bpf programs.
4683 *
4684 * It could be used at the client/active side (i.e. connect() side)
4685 * when the server told it that the server was in syncookie
4686 * mode and required the active side to resend the bpf-written
4687 * options. The active side can keep writing the bpf-options until
4688 * it received a valid packet from the server side to confirm
4689 * the earlier packet (and options) has been received. The later
4690 * example patch is using it like this at the active side when the
4691 * server is in syncookie mode.
4692 *
4693 * The bpf prog will usually turn this off in the common cases.
4694 */
4695 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
4696 /* Call bpf when kernel has received a header option that
4697 * the kernel cannot handle. The bpf prog will be called under
4698 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
4699 *
4700 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
4701 * for the header option related helpers that will be useful
4702 * to the bpf programs.
4703 */
00d211a4 4704 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
0813a841
MKL
4705 /* Call bpf when the kernel is writing header options for the
4706 * outgoing packet. The bpf prog will first be called
4707 * to reserve space in a skb under
4708 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
4709 * the bpf prog will be called to write the header option(s)
4710 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4711 *
4712 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
4713 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
4714 * related helpers that will be useful to the bpf programs.
4715 *
4716 * The kernel gets its chance to reserve space and write
4717 * options first before the BPF program does.
4718 */
331fca43 4719 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
1aae4bdd 4720/* Mask of all currently supported cb flags */
331fca43 4721 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
1aae4bdd 4722};
b13d8807 4723
40304b2a
LB
4724/* List of known BPF sock_ops operators.
4725 * New entries can only be added at the end
4726 */
4727enum {
4728 BPF_SOCK_OPS_VOID,
8550f328
LB
4729 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
4730 * -1 if default value should be used
4731 */
13d3b1eb
LB
4732 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
4733 * window (in packets) or -1 if default
4734 * value should be used
4735 */
9872a4bd
LB
4736 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
4737 * active connection is initialized
4738 */
4739 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
4740 * active connection is
4741 * established
4742 */
4743 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
4744 * passive connection is
4745 * established
4746 */
91b5b21c
LB
4747 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
4748 * needs ECN
4749 */
e6546ef6
LB
4750 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
4751 * based on the path and may be
4752 * dependent on the congestion control
4753 * algorithm. In general it indicates
4754 * a congestion threshold. RTTs above
4755 * this indicate congestion
4756 */
f89013f6
LB
4757 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
4758 * Arg1: value of icsk_retransmits
4759 * Arg2: value of icsk_rto
4760 * Arg3: whether RTO has expired
4761 */
a31ad29e
LB
4762 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
4763 * Arg1: sequence number of 1st byte
4764 * Arg2: # segments
4765 * Arg3: return value of
4766 * tcp_transmit_skb (0 => success)
4767 */
d4487491
LB
4768 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
4769 * Arg1: old_state
4770 * Arg2: new_state
4771 */
f333ee0c
AI
4772 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
4773 * socket transition to LISTEN state.
4774 */
23729ff2
SF
4775 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
4776 */
0813a841
MKL
4777 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
4778 * It will be called to handle
4779 * the packets received at
4780 * an already established
4781 * connection.
4782 *
4783 * sock_ops->skb_data:
4784 * Referring to the received skb.
4785 * It covers the TCP header only.
4786 *
4787 * bpf_load_hdr_opt() can also
4788 * be used to search for a
4789 * particular option.
4790 */
4791 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
4792 * header option later in
4793 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4794 * Arg1: bool want_cookie. (in
4795 * writing SYNACK only)
4796 *
4797 * sock_ops->skb_data:
4798 * Not available because no header has
4799 * been written yet.
4800 *
4801 * sock_ops->skb_tcp_flags:
4802 * The tcp_flags of the
4803 * outgoing skb. (e.g. SYN, ACK, FIN).
4804 *
4805 * bpf_reserve_hdr_opt() should
4806 * be used to reserve space.
4807 */
4808 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
4809 * Arg1: bool want_cookie. (in
4810 * writing SYNACK only)
4811 *
4812 * sock_ops->skb_data:
4813 * Referring to the outgoing skb.
4814 * It covers the TCP header
4815 * that has already been written
4816 * by the kernel and the
4817 * earlier bpf-progs.
4818 *
4819 * sock_ops->skb_tcp_flags:
4820 * The tcp_flags of the outgoing
4821 * skb. (e.g. SYN, ACK, FIN).
4822 *
4823 * bpf_store_hdr_opt() should
4824 * be used to write the
4825 * option.
4826 *
4827 * bpf_load_hdr_opt() can also
4828 * be used to search for a
4829 * particular option that
4830 * has already been written
4831 * by the kernel or the
4832 * earlier bpf-progs.
4833 */
d4487491
LB
4834};
4835
4836/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
4837 * changes between the TCP and BPF versions. Ideally this should never happen.
4838 * If it does, we need to add code to convert them before calling
4839 * the BPF sock_ops function.
4840 */
4841enum {
4842 BPF_TCP_ESTABLISHED = 1,
4843 BPF_TCP_SYN_SENT,
4844 BPF_TCP_SYN_RECV,
4845 BPF_TCP_FIN_WAIT1,
4846 BPF_TCP_FIN_WAIT2,
4847 BPF_TCP_TIME_WAIT,
4848 BPF_TCP_CLOSE,
4849 BPF_TCP_CLOSE_WAIT,
4850 BPF_TCP_LAST_ACK,
4851 BPF_TCP_LISTEN,
4852 BPF_TCP_CLOSING, /* Now a valid state */
4853 BPF_TCP_NEW_SYN_RECV,
4854
4855 BPF_TCP_MAX_STATES /* Leave at the end! */
40304b2a
LB
4856};
4857
1aae4bdd
AN
4858enum {
4859 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4860 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
2b8ee4f0 4861 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
ca584ba0 4862 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
0813a841
MKL
4863 /* Copy the SYN pkt to optval
4864 *
4865 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
4866 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
4867 * to only getting from the saved_syn. It can either get the
4868 * syn packet from:
4869 *
4870 * 1. the just-received SYN packet (only available when writing the
4871 * SYNACK). It will be useful when it is not necessary to
4872 * save the SYN packet for latter use. It is also the only way
4873 * to get the SYN during syncookie mode because the syn
4874 * packet cannot be saved during syncookie.
4875 *
4876 * OR
4877 *
4878 * 2. the earlier saved syn which was done by
4879 * bpf_setsockopt(TCP_SAVE_SYN).
4880 *
4881 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
4882 * SYN packet is obtained.
4883 *
4884 * If the bpf-prog does not need the IP[46] header, the
4885 * bpf-prog can avoid parsing the IP header by using
4886 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
4887 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
4888 *
4889 * >0: Total number of bytes copied
4890 * -ENOSPC: Not enough space in optval. Only optlen number of
4891 * bytes is copied.
4892 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
4893 * is not saved by setsockopt(TCP_SAVE_SYN).
4894 */
4895 TCP_BPF_SYN = 1005, /* Copy the TCP header */
4896 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
267cf9fa 4897 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
0813a841
MKL
4898};
4899
4900enum {
4901 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
4902};
4903
4904/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
4905 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
4906 */
4907enum {
4908 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
4909 * total option spaces
4910 * required for an established
4911 * sk in order to calculate the
4912 * MSS. No skb is actually
4913 * sent.
4914 */
4915 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
4916 * when sending a SYN.
4917 */
1aae4bdd 4918};
fc747810 4919
908432ca
YS
4920struct bpf_perf_event_value {
4921 __u64 counter;
4922 __u64 enabled;
4923 __u64 running;
4924};
4925
1aae4bdd
AN
4926enum {
4927 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4928 BPF_DEVCG_ACC_READ = (1ULL << 1),
4929 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4930};
ebc614f6 4931
1aae4bdd
AN
4932enum {
4933 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4934 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4935};
ebc614f6
RG
4936
4937struct bpf_cgroup_dev_ctx {
06ef0ccb
YS
4938 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4939 __u32 access_type;
ebc614f6
RG
4940 __u32 major;
4941 __u32 minor;
4942};
4943
c4f6699d
AS
4944struct bpf_raw_tracepoint_args {
4945 __u64 args[0];
4946};
4947
87f5fc7e
DA
4948/* DIRECT: Skip the FIB rules and go to FIB table associated with device
4949 * OUTPUT: Do lookup from egress perspective; default is ingress
4950 */
1aae4bdd
AN
4951enum {
4952 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4953 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4954};
87f5fc7e 4955
4c79579b
DA
4956enum {
4957 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4958 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4959 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4960 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4961 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4962 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4963 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4964 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4965 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4966};
4967
87f5fc7e 4968struct bpf_fib_lookup {
fa898d76
DA
4969 /* input: network family for lookup (AF_INET, AF_INET6)
4970 * output: network family of egress nexthop
4971 */
4972 __u8 family;
87f5fc7e
DA
4973
4974 /* set if lookup is to consider L4 data - e.g., FIB rules */
4975 __u8 l4_protocol;
4976 __be16 sport;
4977 __be16 dport;
4978
4979 /* total length of packet from network header - used for MTU check */
4980 __u16 tot_len;
4c79579b
DA
4981
4982 /* input: L3 device index for lookup
4983 * output: device index from FIB lookup
4984 */
4985 __u32 ifindex;
87f5fc7e
DA
4986
4987 union {
4988 /* inputs to lookup */
4989 __u8 tos; /* AF_INET */
bd3a08aa 4990 __be32 flowinfo; /* AF_INET6, flow_label + priority */
87f5fc7e 4991
fa898d76
DA
4992 /* output: metric of fib result (IPv4/IPv6 only) */
4993 __u32 rt_metric;
87f5fc7e
DA
4994 };
4995
4996 union {
87f5fc7e
DA
4997 __be32 ipv4_src;
4998 __u32 ipv6_src[4]; /* in6_addr; network order */
4999 };
5000
fa898d76
DA
5001 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
5002 * network header. output: bpf_fib_lookup sets to gateway address
5003 * if FIB lookup returns gateway route
87f5fc7e
DA
5004 */
5005 union {
87f5fc7e
DA
5006 __be32 ipv4_dst;
5007 __u32 ipv6_dst[4]; /* in6_addr; network order */
5008 };
5009
5010 /* output */
5011 __be16 h_vlan_proto;
5012 __be16 h_vlan_TCI;
5013 __u8 smac[6]; /* ETH_ALEN */
5014 __u8 dmac[6]; /* ETH_ALEN */
5015};
5016
ba452c9e
THJ
5017struct bpf_redir_neigh {
5018 /* network family for lookup (AF_INET, AF_INET6) */
5019 __u32 nh_family;
5020 /* network address of nexthop; skips fib lookup to find gateway */
5021 union {
5022 __be32 ipv4_nh;
5023 __u32 ipv6_nh[4]; /* in6_addr; network order */
5024 };
5025};
5026
41bdc4b4
YS
5027enum bpf_task_fd_type {
5028 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
5029 BPF_FD_TYPE_TRACEPOINT, /* tp name */
5030 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
5031 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
5032 BPF_FD_TYPE_UPROBE, /* filename + offset */
5033 BPF_FD_TYPE_URETPROBE, /* filename + offset */
5034};
5035
1aae4bdd
AN
5036enum {
5037 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
5038 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
5039 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
5040};
086f9568 5041
d58e468b
PP
5042struct bpf_flow_keys {
5043 __u16 nhoff;
5044 __u16 thoff;
5045 __u16 addr_proto; /* ETH_P_* of valid addrs */
5046 __u8 is_frag;
5047 __u8 is_first_frag;
5048 __u8 is_encap;
5049 __u8 ip_proto;
5050 __be16 n_proto;
5051 __be16 sport;
5052 __be16 dport;
5053 union {
5054 struct {
5055 __be32 ipv4_src;
5056 __be32 ipv4_dst;
5057 };
5058 struct {
5059 __u32 ipv6_src[4]; /* in6_addr; network order */
5060 __u32 ipv6_dst[4]; /* in6_addr; network order */
5061 };
5062 };
086f9568 5063 __u32 flags;
71c99e32 5064 __be32 flow_label;
d58e468b
PP
5065};
5066
838e9690 5067struct bpf_func_info {
d30d42e0 5068 __u32 insn_off;
838e9690
YS
5069 __u32 type_id;
5070};
5071
c454a46b
MKL
5072#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
5073#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
5074
5075struct bpf_line_info {
5076 __u32 insn_off;
5077 __u32 file_name_off;
5078 __u32 line_off;
5079 __u32 line_col;
5080};
5081
d83525ca
AS
5082struct bpf_spin_lock {
5083 __u32 val;
5084};
7b146ceb
AI
5085
5086struct bpf_sysctl {
5087 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
5088 * Allows 1,2,4-byte read, but no write.
5089 */
e1550bfe
AI
5090 __u32 file_pos; /* Sysctl file position to read from, write to.
5091 * Allows 1,2,4-byte read an 4-byte write.
5092 */
7b146ceb
AI
5093};
5094
0d01da6a
SF
5095struct bpf_sockopt {
5096 __bpf_md_ptr(struct bpf_sock *, sk);
5097 __bpf_md_ptr(void *, optval);
5098 __bpf_md_ptr(void *, optval_end);
5099
5100 __s32 level;
5101 __s32 optname;
5102 __s32 optlen;
5103 __s32 retval;
5104};
5105
b4490c5c
CN
5106struct bpf_pidns_info {
5107 __u32 pid;
5108 __u32 tgid;
5109};
e9ddbb77
JS
5110
5111/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
5112struct bpf_sk_lookup {
5113 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
5114
5115 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
5116 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
5117 __u32 remote_ip4; /* Network byte order */
5118 __u32 remote_ip6[4]; /* Network byte order */
5119 __u32 remote_port; /* Network byte order */
5120 __u32 local_ip4; /* Network byte order */
5121 __u32 local_ip6[4]; /* Network byte order */
5122 __u32 local_port; /* Host byte order */
5123};
5124
c4d0bfb4
AM
5125/*
5126 * struct btf_ptr is used for typed pointer representation; the
5127 * type id is used to render the pointer data as the appropriate type
5128 * via the bpf_snprintf_btf() helper described above. A flags field -
5129 * potentially to specify additional details about the BTF pointer
5130 * (rather than its mode of display) - is included for future use.
5131 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
5132 */
5133struct btf_ptr {
5134 void *ptr;
5135 __u32 type_id;
5136 __u32 flags; /* BTF ptr flags; unused at present. */
5137};
5138
5139/*
5140 * Flags to control bpf_snprintf_btf() behaviour.
5141 * - BTF_F_COMPACT: no formatting around type information
5142 * - BTF_F_NONAME: no struct/union member names/types
5143 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
5144 * equivalent to %px.
5145 * - BTF_F_ZERO: show zero-valued struct/union members; they
5146 * are not displayed by default
5147 */
5148enum {
5149 BTF_F_COMPACT = (1ULL << 0),
5150 BTF_F_NONAME = (1ULL << 1),
5151 BTF_F_PTR_RAW = (1ULL << 2),
5152 BTF_F_ZERO = (1ULL << 3),
5153};
5154
daedfb22 5155#endif /* _UAPI__LINUX_BPF_H__ */