]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - tools/include/uapi/linux/bpf.h
bpf: Implement BPF ring buffer and verifier support for it
[mirror_ubuntu-hirsute-kernel.git] / tools / include / uapi / linux / bpf.h
CommitLineData
fb7df12d 1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
971e827b
ACM
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>
12#include <linux/bpf_common.h>
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 */
971e827b
ACM
18#define BPF_ALU64 0x07 /* alu mode in double word width */
19
20/* ld/ldx fields */
d6d4f60c 21#define BPF_DW 0x18 /* double word (64-bit) */
971e827b
ACM
22#define BPF_XADD 0xc0 /* exclusive add */
23
24/* alu/jmp fields */
25#define BPF_MOV 0xb0 /* mov reg to reg */
26#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27
28/* change endianness of a register */
29#define BPF_END 0xd0 /* flags for endianness conversion: */
30#define BPF_TO_LE 0x00 /* convert to little-endian */
31#define BPF_TO_BE 0x08 /* convert to big-endian */
32#define BPF_FROM_LE BPF_TO_LE
33#define BPF_FROM_BE BPF_TO_BE
34
92b31a9a 35/* jmp encodings */
971e827b 36#define BPF_JNE 0x50 /* jump != */
92b31a9a
DB
37#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
971e827b
ACM
39#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
92b31a9a
DB
41#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
971e827b
ACM
43#define BPF_CALL 0x80 /* function call */
44#define BPF_EXIT 0x90 /* function return */
45
46/* Register numbers */
47enum {
48 BPF_REG_0 = 0,
49 BPF_REG_1,
50 BPF_REG_2,
51 BPF_REG_3,
52 BPF_REG_4,
53 BPF_REG_5,
54 BPF_REG_6,
55 BPF_REG_7,
56 BPF_REG_8,
57 BPF_REG_9,
58 BPF_REG_10,
59 __MAX_BPF_REG,
60};
61
62/* BPF has 10 general purpose 64-bit registers and stack frame. */
63#define MAX_BPF_REG __MAX_BPF_REG
64
65struct bpf_insn {
66 __u8 code; /* opcode */
67 __u8 dst_reg:4; /* dest register */
68 __u8 src_reg:4; /* source register */
69 __s16 off; /* signed offset */
70 __s32 imm; /* signed immediate constant */
71};
72
9a738266
MS
73/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
74struct bpf_lpm_trie_key {
75 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
fb53d3b6 76 __u8 data[0]; /* Arbitrary size */
9a738266
MS
77};
78
c419cf52
RG
79struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
82};
83
971e827b
ACM
84/* BPF syscall commands, see bpf(2) man-page for details. */
85enum bpf_cmd {
86 BPF_MAP_CREATE,
87 BPF_MAP_LOOKUP_ELEM,
88 BPF_MAP_UPDATE_ELEM,
89 BPF_MAP_DELETE_ELEM,
90 BPF_MAP_GET_NEXT_KEY,
91 BPF_PROG_LOAD,
92 BPF_OBJ_PIN,
93 BPF_OBJ_GET,
0cb34dc2
JS
94 BPF_PROG_ATTACH,
95 BPF_PROG_DETACH,
30848873 96 BPF_PROG_TEST_RUN,
95b9afd3
MKL
97 BPF_PROG_GET_NEXT_ID,
98 BPF_MAP_GET_NEXT_ID,
99 BPF_PROG_GET_FD_BY_ID,
100 BPF_MAP_GET_FD_BY_ID,
101 BPF_OBJ_GET_INFO_BY_FD,
defd9c47 102 BPF_PROG_QUERY,
a0fe3e57 103 BPF_RAW_TRACEPOINT_OPEN,
3bd86a84 104 BPF_BTF_LOAD,
7a01f6a3 105 BPF_BTF_GET_FD_BY_ID,
30687ad9 106 BPF_TASK_FD_QUERY,
da4e1b15 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
c83fef6b 108 BPF_MAP_FREEZE,
d2648e1e 109 BPF_BTF_GET_NEXT_ID,
a1e3a3b8
YS
110 BPF_MAP_LOOKUP_BATCH,
111 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
112 BPF_MAP_UPDATE_BATCH,
113 BPF_MAP_DELETE_BATCH,
af6eea57 114 BPF_LINK_CREATE,
cc4f864b 115 BPF_LINK_UPDATE,
f2e10bff
AN
116 BPF_LINK_GET_FD_BY_ID,
117 BPF_LINK_GET_NEXT_ID,
d46edd67 118 BPF_ENABLE_STATS,
ac51d99b 119 BPF_ITER_CREATE,
971e827b
ACM
120};
121
122enum bpf_map_type {
123 BPF_MAP_TYPE_UNSPEC,
124 BPF_MAP_TYPE_HASH,
125 BPF_MAP_TYPE_ARRAY,
126 BPF_MAP_TYPE_PROG_ARRAY,
127 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
128 BPF_MAP_TYPE_PERCPU_HASH,
129 BPF_MAP_TYPE_PERCPU_ARRAY,
130 BPF_MAP_TYPE_STACK_TRACE,
791cceb8 131 BPF_MAP_TYPE_CGROUP_ARRAY,
0cb34dc2
JS
132 BPF_MAP_TYPE_LRU_HASH,
133 BPF_MAP_TYPE_LRU_PERCPU_HASH,
9a738266 134 BPF_MAP_TYPE_LPM_TRIE,
fb30d4b7
MKL
135 BPF_MAP_TYPE_ARRAY_OF_MAPS,
136 BPF_MAP_TYPE_HASH_OF_MAPS,
81f6bf81 137 BPF_MAP_TYPE_DEVMAP,
69e8cc13 138 BPF_MAP_TYPE_SOCKMAP,
6710e112 139 BPF_MAP_TYPE_CPUMAP,
cb9c28ef 140 BPF_MAP_TYPE_XSKMAP,
b8b394fa 141 BPF_MAP_TYPE_SOCKHASH,
c419cf52 142 BPF_MAP_TYPE_CGROUP_STORAGE,
3bd43a8c 143 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
25025e0a 144 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
da4e1b15
MV
145 BPF_MAP_TYPE_QUEUE,
146 BPF_MAP_TYPE_STACK,
948d930e 147 BPF_MAP_TYPE_SK_STORAGE,
10fbe211 148 BPF_MAP_TYPE_DEVMAP_HASH,
17328d61 149 BPF_MAP_TYPE_STRUCT_OPS,
457f4436 150 BPF_MAP_TYPE_RINGBUF,
971e827b
ACM
151};
152
6c4fc209
DB
153/* Note that tracing related programs such as
154 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
155 * are not subject to a stable API since kernel internal data
156 * structures can change from release to release and may
157 * therefore break existing tracing BPF programs. Tracing BPF
158 * programs correspond to /a/ specific kernel which is to be
159 * analyzed, and not /a/ specific kernel /and/ all future ones.
160 */
971e827b
ACM
161enum bpf_prog_type {
162 BPF_PROG_TYPE_UNSPEC,
163 BPF_PROG_TYPE_SOCKET_FILTER,
164 BPF_PROG_TYPE_KPROBE,
165 BPF_PROG_TYPE_SCHED_CLS,
166 BPF_PROG_TYPE_SCHED_ACT,
167 BPF_PROG_TYPE_TRACEPOINT,
791cceb8 168 BPF_PROG_TYPE_XDP,
0cb34dc2
JS
169 BPF_PROG_TYPE_PERF_EVENT,
170 BPF_PROG_TYPE_CGROUP_SKB,
171 BPF_PROG_TYPE_CGROUP_SOCK,
172 BPF_PROG_TYPE_LWT_IN,
173 BPF_PROG_TYPE_LWT_OUT,
174 BPF_PROG_TYPE_LWT_XMIT,
04df41e3 175 BPF_PROG_TYPE_SOCK_OPS,
69e8cc13 176 BPF_PROG_TYPE_SK_SKB,
ebc614f6 177 BPF_PROG_TYPE_CGROUP_DEVICE,
82a86168 178 BPF_PROG_TYPE_SK_MSG,
a0fe3e57 179 BPF_PROG_TYPE_RAW_TRACEPOINT,
e50b0a6f 180 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
c99a84ea 181 BPF_PROG_TYPE_LWT_SEG6LOCAL,
6bdd533c 182 BPF_PROG_TYPE_LIRC_MODE2,
3bd43a8c 183 BPF_PROG_TYPE_SK_REUSEPORT,
2f965e3f 184 BPF_PROG_TYPE_FLOW_DISSECTOR,
196398d4 185 BPF_PROG_TYPE_CGROUP_SYSCTL,
4635b0ae 186 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
aa6ab647 187 BPF_PROG_TYPE_CGROUP_SOCKOPT,
12a8654b 188 BPF_PROG_TYPE_TRACING,
17328d61 189 BPF_PROG_TYPE_STRUCT_OPS,
2db6eab1 190 BPF_PROG_TYPE_EXT,
fc611f47 191 BPF_PROG_TYPE_LSM,
971e827b
ACM
192};
193
0cb34dc2
JS
194enum bpf_attach_type {
195 BPF_CGROUP_INET_INGRESS,
196 BPF_CGROUP_INET_EGRESS,
197 BPF_CGROUP_INET_SOCK_CREATE,
04df41e3 198 BPF_CGROUP_SOCK_OPS,
464bc0fd
JF
199 BPF_SK_SKB_STREAM_PARSER,
200 BPF_SK_SKB_STREAM_VERDICT,
ebc614f6 201 BPF_CGROUP_DEVICE,
82a86168 202 BPF_SK_MSG_VERDICT,
e50b0a6f
AI
203 BPF_CGROUP_INET4_BIND,
204 BPF_CGROUP_INET6_BIND,
622adafb
AI
205 BPF_CGROUP_INET4_CONNECT,
206 BPF_CGROUP_INET6_CONNECT,
1d436885
AI
207 BPF_CGROUP_INET4_POST_BIND,
208 BPF_CGROUP_INET6_POST_BIND,
3024cf82
AI
209 BPF_CGROUP_UDP4_SENDMSG,
210 BPF_CGROUP_UDP6_SENDMSG,
6bdd533c 211 BPF_LIRC_MODE2,
2f965e3f 212 BPF_FLOW_DISSECTOR,
196398d4 213 BPF_CGROUP_SYSCTL,
3dbc6ada
DB
214 BPF_CGROUP_UDP4_RECVMSG,
215 BPF_CGROUP_UDP6_RECVMSG,
aa6ab647
SF
216 BPF_CGROUP_GETSOCKOPT,
217 BPF_CGROUP_SETSOCKOPT,
12a8654b 218 BPF_TRACE_RAW_TP,
b8c54ea4
AS
219 BPF_TRACE_FENTRY,
220 BPF_TRACE_FEXIT,
ae240823 221 BPF_MODIFY_RETURN,
fc611f47 222 BPF_LSM_MAC,
15d83c4d 223 BPF_TRACE_ITER,
1b66d253
DB
224 BPF_CGROUP_INET4_GETPEERNAME,
225 BPF_CGROUP_INET6_GETPEERNAME,
226 BPF_CGROUP_INET4_GETSOCKNAME,
227 BPF_CGROUP_INET6_GETSOCKNAME,
0cb34dc2
JS
228 __MAX_BPF_ATTACH_TYPE
229};
230
231#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
232
f2e10bff
AN
233enum bpf_link_type {
234 BPF_LINK_TYPE_UNSPEC = 0,
235 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
236 BPF_LINK_TYPE_TRACING = 2,
237 BPF_LINK_TYPE_CGROUP = 3,
de4e05ca 238 BPF_LINK_TYPE_ITER = 4,
f2e10bff
AN
239
240 MAX_BPF_LINK_TYPE,
241};
242
defd9c47
AS
243/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
244 *
245 * NONE(default): No further bpf programs allowed in the subtree.
246 *
247 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
248 * the program in this cgroup yields to sub-cgroup program.
249 *
250 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
251 * that cgroup program gets run in addition to the program in this cgroup.
252 *
253 * Only one program is allowed to be attached to a cgroup with
254 * NONE or BPF_F_ALLOW_OVERRIDE flag.
255 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
256 * release old program and attach the new one. Attach flags has to match.
257 *
258 * Multiple programs are allowed to be attached to a cgroup with
259 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
260 * (those that were attached first, run first)
261 * The programs of sub-cgroup are executed first, then programs of
262 * this cgroup and then programs of parent cgroup.
263 * When children program makes decision (like picking TCP CA or sock bind)
264 * parent program has a chance to override it.
265 *
7dd68b32
AI
266 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
267 * programs for a cgroup. Though it's possible to replace an old program at
268 * any position by also specifying BPF_F_REPLACE flag and position itself in
269 * replace_bpf_fd attribute. Old program at this position will be released.
270 *
defd9c47
AS
271 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
272 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
273 * Ex1:
274 * cgrp1 (MULTI progs A, B) ->
275 * cgrp2 (OVERRIDE prog C) ->
276 * cgrp3 (MULTI prog D) ->
277 * cgrp4 (OVERRIDE prog E) ->
278 * cgrp5 (NONE prog F)
279 * the event in cgrp5 triggers execution of F,D,A,B in that order.
280 * if prog F is detached, the execution is E,D,A,B
281 * if prog F and D are detached, the execution is E,A,B
282 * if prog F, E and D are detached, the execution is C,A,B
283 *
284 * All eligible programs are executed regardless of return code from
285 * earlier programs.
5463b3d0
SR
286 */
287#define BPF_F_ALLOW_OVERRIDE (1U << 0)
defd9c47 288#define BPF_F_ALLOW_MULTI (1U << 1)
7dd68b32 289#define BPF_F_REPLACE (1U << 2)
5463b3d0 290
e07b98d9
DM
291/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
292 * verifier will perform strict alignment checking as if the kernel
293 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
294 * and NET_IP_ALIGN defined to 2.
295 */
296#define BPF_F_STRICT_ALIGNMENT (1U << 0)
297
e9ee9efc
DM
298/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
299 * verifier will allow any alignment whatsoever. On platforms
300 * with strict alignment requirements for loads ands stores (such
301 * as sparc and mips) the verifier validates that all loads and
302 * stores provably follow this requirement. This flag turns that
303 * checking and enforcement off.
304 *
305 * It is mostly used for testing when we want to validate the
306 * context and memory access aspects of the verifier, but because
307 * of an unaligned access the alignment check would trigger before
308 * the one we are interested in.
309 */
310#define BPF_F_ANY_ALIGNMENT (1U << 1)
311
9ce33e33
JW
312/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
313 * Verifier does sub-register def/use analysis and identifies instructions whose
314 * def only matters for low 32-bit, high 32-bit is never referenced later
315 * through implicit zero extension. Therefore verifier notifies JIT back-ends
316 * that it is safe to ignore clearing high 32-bit for these instructions. This
317 * saves some back-ends a lot of code-gen. However such optimization is not
318 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
319 * hence hasn't used verifier's analysis result. But, we really want to have a
320 * way to be able to verify the correctness of the described optimization on
321 * x86_64 on which testsuites are frequently exercised.
322 *
323 * So, this flag is introduced. Once it is set, verifier will randomize high
324 * 32-bit for those instructions who has been identified as safe to ignore them.
325 * Then, if verifier is not doing correct analysis, such randomization will
326 * regress tests to expose bugs.
327 */
328#define BPF_F_TEST_RND_HI32 (1U << 2)
329
0fc2e0b8
AS
330/* The verifier internal test flag. Behavior is undefined */
331#define BPF_F_TEST_STATE_FREQ (1U << 3)
332
c83fef6b
DB
333/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
334 * two extensions:
335 *
336 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
337 * insn[0].imm: map fd map fd
338 * insn[1].imm: 0 offset into value
339 * insn[0].off: 0 0
340 * insn[1].off: 0 0
341 * ldimm64 rewrite: address of map address of map[0]+offset
342 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
343 */
971e827b 344#define BPF_PSEUDO_MAP_FD 1
c83fef6b 345#define BPF_PSEUDO_MAP_VALUE 2
971e827b 346
48cca7e4
AS
347/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
348 * offset to another bpf function
349 */
350#define BPF_PSEUDO_CALL 1
351
971e827b 352/* flags for BPF_MAP_UPDATE_ELEM command */
1aae4bdd
AN
353enum {
354 BPF_ANY = 0, /* create new element or update existing */
355 BPF_NOEXIST = 1, /* create new element if it didn't exist */
356 BPF_EXIST = 2, /* update existing element */
357 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
358};
971e827b 359
ad17d0e6 360/* flags for BPF_MAP_CREATE command */
1aae4bdd
AN
361enum {
362 BPF_F_NO_PREALLOC = (1U << 0),
0cb34dc2
JS
363/* Instead of having one common LRU list in the
364 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
365 * which can scale and perform better.
366 * Note, the LRU nodes (including free nodes) cannot be moved
367 * across different LRU lists.
368 */
1aae4bdd 369 BPF_F_NO_COMMON_LRU = (1U << 1),
ad17d0e6 370/* Specify numa node during map creation */
1aae4bdd 371 BPF_F_NUMA_NODE = (1U << 2),
88cda1c9 372
c83fef6b 373/* Flags for accessing BPF object from syscall side. */
1aae4bdd
AN
374 BPF_F_RDONLY = (1U << 3),
375 BPF_F_WRONLY = (1U << 4),
e27afb84 376
81f77fd0 377/* Flag for stack_map, store build_id+offset instead of pointer */
1aae4bdd 378 BPF_F_STACK_BUILD_ID = (1U << 5),
81f77fd0 379
608114e4 380/* Zero-initialize hash function seed. This should only be used for testing. */
1aae4bdd 381 BPF_F_ZERO_SEED = (1U << 6),
608114e4 382
c83fef6b 383/* Flags for accessing BPF object from program side. */
1aae4bdd
AN
384 BPF_F_RDONLY_PROG = (1U << 7),
385 BPF_F_WRONLY_PROG = (1U << 8),
c83fef6b 386
9e819ffc 387/* Clone map from listener for newly accepted socket */
1aae4bdd 388 BPF_F_CLONE = (1U << 9),
9e819ffc 389
fc970227 390/* Enable memory-mapping BPF map */
1aae4bdd
AN
391 BPF_F_MMAPABLE = (1U << 10),
392};
fc970227 393
f5bfcd95
AI
394/* Flags for BPF_PROG_QUERY. */
395
396/* Query effective (directly attached + inherited from ancestor cgroups)
397 * programs that will be executed for events within a cgroup.
398 * attach_flags with this flag are returned only for directly attached programs.
399 */
608114e4
LB
400#define BPF_F_QUERY_EFFECTIVE (1U << 0)
401
d46edd67
SL
402/* type for BPF_ENABLE_STATS */
403enum bpf_stats_type {
404 /* enabled run_time_ns and run_cnt */
405 BPF_STATS_RUN_TIME = 0,
406};
407
81f77fd0
SL
408enum bpf_stack_build_id_status {
409 /* user space need an empty entry to identify end of a trace */
410 BPF_STACK_BUILD_ID_EMPTY = 0,
411 /* with valid build_id and offset */
412 BPF_STACK_BUILD_ID_VALID = 1,
413 /* couldn't get build_id, fallback to ip */
414 BPF_STACK_BUILD_ID_IP = 2,
415};
416
417#define BPF_BUILD_ID_SIZE 20
418struct bpf_stack_build_id {
419 __s32 status;
420 unsigned char build_id[BPF_BUILD_ID_SIZE];
421 union {
422 __u64 offset;
423 __u64 ip;
424 };
425};
426
1aae4bdd
AN
427#define BPF_OBJ_NAME_LEN 16U
428
971e827b
ACM
429union bpf_attr {
430 struct { /* anonymous struct used by BPF_MAP_CREATE command */
431 __u32 map_type; /* one of enum bpf_map_type */
432 __u32 key_size; /* size of key in bytes */
433 __u32 value_size; /* size of value in bytes */
434 __u32 max_entries; /* max number of entries in a map */
ad17d0e6
MKL
435 __u32 map_flags; /* BPF_MAP_CREATE related
436 * flags defined above.
437 */
fb30d4b7 438 __u32 inner_map_fd; /* fd pointing to the inner map */
ad17d0e6
MKL
439 __u32 numa_node; /* numa node (effective only if
440 * BPF_F_NUMA_NODE is set).
441 */
067cae47 442 char map_name[BPF_OBJ_NAME_LEN];
a3884572 443 __u32 map_ifindex; /* ifindex of netdev to create on */
3bd86a84 444 __u32 btf_fd; /* fd pointing to a BTF type data */
f03b15d3
MKL
445 __u32 btf_key_type_id; /* BTF type_id of the key */
446 __u32 btf_value_type_id; /* BTF type_id of the value */
17328d61
MKL
447 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
448 * struct stored as the
449 * map value
450 */
971e827b
ACM
451 };
452
453 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
454 __u32 map_fd;
455 __aligned_u64 key;
456 union {
457 __aligned_u64 value;
458 __aligned_u64 next_key;
459 };
460 __u64 flags;
461 };
462
a1e3a3b8
YS
463 struct { /* struct used by BPF_MAP_*_BATCH commands */
464 __aligned_u64 in_batch; /* start batch,
465 * NULL to start from beginning
466 */
467 __aligned_u64 out_batch; /* output: next start batch */
468 __aligned_u64 keys;
469 __aligned_u64 values;
470 __u32 count; /* input/output:
471 * input: # of key/value
472 * elements
473 * output: # of filled elements
474 */
475 __u32 map_fd;
476 __u64 elem_flags;
477 __u64 flags;
478 } batch;
479
971e827b
ACM
480 struct { /* anonymous struct used by BPF_PROG_LOAD command */
481 __u32 prog_type; /* one of enum bpf_prog_type */
482 __u32 insn_cnt;
483 __aligned_u64 insns;
484 __aligned_u64 license;
485 __u32 log_level; /* verbosity level of verifier */
486 __u32 log_size; /* size of user buffer */
487 __aligned_u64 log_buf; /* user supplied buffer */
6c4fc209 488 __u32 kern_version; /* not used */
e07b98d9 489 __u32 prog_flags;
067cae47 490 char prog_name[BPF_OBJ_NAME_LEN];
1f6f4cb7 491 __u32 prog_ifindex; /* ifindex of netdev to prep for */
d7be143b
AI
492 /* For some prog types expected attach type must be known at
493 * load time to verify attach type specific parts of prog
494 * (context accesses, allowed helpers, etc).
495 */
496 __u32 expected_attach_type;
cc19435c
YS
497 __u32 prog_btf_fd; /* fd pointing to BTF type data */
498 __u32 func_info_rec_size; /* userspace bpf_func_info size */
499 __aligned_u64 func_info; /* func info */
500 __u32 func_info_cnt; /* number of bpf_func_info records */
ee491d8d
MKL
501 __u32 line_info_rec_size; /* userspace bpf_line_info size */
502 __aligned_u64 line_info; /* line info */
503 __u32 line_info_cnt; /* number of bpf_line_info records */
ccfe29eb 504 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
e7bf94db 505 __u32 attach_prog_fd; /* 0 to attach to vmlinux */
971e827b
ACM
506 };
507
508 struct { /* anonymous struct used by BPF_OBJ_* commands */
509 __aligned_u64 pathname;
510 __u32 bpf_fd;
e27afb84 511 __u32 file_flags;
971e827b 512 };
0cb34dc2
JS
513
514 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
515 __u32 target_fd; /* container object to attach to */
516 __u32 attach_bpf_fd; /* eBPF program to attach */
517 __u32 attach_type;
5463b3d0 518 __u32 attach_flags;
7dd68b32
AI
519 __u32 replace_bpf_fd; /* previously attached eBPF
520 * program to replace if
521 * BPF_F_REPLACE is used
522 */
0cb34dc2 523 };
30848873
AS
524
525 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
526 __u32 prog_fd;
527 __u32 retval;
2587a974
LB
528 __u32 data_size_in; /* input: len of data_in */
529 __u32 data_size_out; /* input/output: len of data_out
530 * returns ENOSPC if data_out
531 * is too small.
532 */
30848873
AS
533 __aligned_u64 data_in;
534 __aligned_u64 data_out;
535 __u32 repeat;
536 __u32 duration;
5e903c65
SF
537 __u32 ctx_size_in; /* input: len of ctx_in */
538 __u32 ctx_size_out; /* input/output: len of ctx_out
539 * returns ENOSPC if ctx_out
540 * is too small.
541 */
542 __aligned_u64 ctx_in;
543 __aligned_u64 ctx_out;
30848873 544 } test;
95b9afd3
MKL
545
546 struct { /* anonymous struct used by BPF_*_GET_*_ID */
547 union {
548 __u32 start_id;
549 __u32 prog_id;
550 __u32 map_id;
7a01f6a3 551 __u32 btf_id;
f2e10bff 552 __u32 link_id;
95b9afd3
MKL
553 };
554 __u32 next_id;
e27afb84 555 __u32 open_flags;
95b9afd3
MKL
556 };
557
558 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
559 __u32 bpf_fd;
560 __u32 info_len;
561 __aligned_u64 info;
562 } info;
defd9c47
AS
563
564 struct { /* anonymous struct used by BPF_PROG_QUERY command */
565 __u32 target_fd; /* container object to query */
566 __u32 attach_type;
567 __u32 query_flags;
568 __u32 attach_flags;
569 __aligned_u64 prog_ids;
570 __u32 prog_cnt;
571 } query;
a0fe3e57 572
af6eea57 573 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
a0fe3e57
AS
574 __u64 name;
575 __u32 prog_fd;
576 } raw_tracepoint;
3bd86a84
MKL
577
578 struct { /* anonymous struct for BPF_BTF_LOAD */
579 __aligned_u64 btf;
580 __aligned_u64 btf_log_buf;
581 __u32 btf_size;
582 __u32 btf_log_size;
583 __u32 btf_log_level;
584 };
30687ad9
YS
585
586 struct {
587 __u32 pid; /* input: pid */
588 __u32 fd; /* input: fd */
589 __u32 flags; /* input: flags */
590 __u32 buf_len; /* input/output: buf len */
591 __aligned_u64 buf; /* input/output:
592 * tp_name for tracepoint
593 * symbol for kprobe
594 * filename for uprobe
595 */
596 __u32 prog_id; /* output: prod_id */
597 __u32 fd_type; /* output: BPF_FD_TYPE_* */
598 __u64 probe_offset; /* output: probe_offset */
599 __u64 probe_addr; /* output: probe_addr */
600 } task_fd_query;
af6eea57
AN
601
602 struct { /* struct used by BPF_LINK_CREATE command */
603 __u32 prog_fd; /* eBPF program to attach */
604 __u32 target_fd; /* object to attach to */
605 __u32 attach_type; /* attach type */
606 __u32 flags; /* extra flags */
607 } link_create;
cc4f864b
AN
608
609 struct { /* struct used by BPF_LINK_UPDATE command */
610 __u32 link_fd; /* link fd */
611 /* new program fd to update link with */
612 __u32 new_prog_fd;
613 __u32 flags; /* extra flags */
614 /* expected link's program fd; is specified only if
615 * BPF_F_REPLACE flag is set in flags */
616 __u32 old_prog_fd;
617 } link_update;
618
d46edd67
SL
619 struct { /* struct used by BPF_ENABLE_STATS command */
620 __u32 type;
621 } enable_stats;
622
ac51d99b
YS
623 struct { /* struct used by BPF_ITER_CREATE command */
624 __u32 link_fd;
625 __u32 flags;
626 } iter_create;
627
971e827b
ACM
628} __attribute__((aligned(8)));
629
9cde0c88
QM
630/* The description below is an attempt at providing documentation to eBPF
631 * developers about the multiple available eBPF helper functions. It can be
632 * parsed and used to produce a manual page. The workflow is the following,
633 * and requires the rst2man utility:
634 *
635 * $ ./scripts/bpf_helpers_doc.py \
636 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
637 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
638 * $ man /tmp/bpf-helpers.7
639 *
640 * Note that in order to produce this external documentation, some RST
641 * formatting is used in the descriptions to get "bold" and "italics" in
642 * manual pages. Also note that the few trailing white spaces are
643 * intentional, removing them would break paragraphs for rst2man.
644 *
645 * Start of BPF helper function descriptions:
646 *
647 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
648 * Description
649 * Perform a lookup in *map* for an entry associated to *key*.
650 * Return
651 * Map value associated to *key*, or **NULL** if no entry was
652 * found.
653 *
654 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
655 * Description
656 * Add or update the value of the entry associated to *key* in
657 * *map* with *value*. *flags* is one of:
658 *
659 * **BPF_NOEXIST**
660 * The entry for *key* must not exist in the map.
661 * **BPF_EXIST**
662 * The entry for *key* must already exist in the map.
663 * **BPF_ANY**
664 * No condition on the existence of the entry for *key*.
665 *
666 * Flag value **BPF_NOEXIST** cannot be used for maps of types
667 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
668 * elements always exist), the helper would return an error.
669 * Return
670 * 0 on success, or a negative error in case of failure.
671 *
672 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
673 * Description
674 * Delete entry with *key* from *map*.
675 * Return
676 * 0 on success, or a negative error in case of failure.
677 *
6ae08ae3 678 * int bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
9cde0c88
QM
679 * Description
680 * For tracing programs, safely attempt to read *size* bytes from
6ae08ae3
DB
681 * kernel space address *unsafe_ptr* and store the data in *dst*.
682 *
ff20460e
QM
683 * Generally, use **bpf_probe_read_user**\ () or
684 * **bpf_probe_read_kernel**\ () instead.
9cde0c88
QM
685 * Return
686 * 0 on success, or a negative error in case of failure.
0cb34dc2
JS
687 *
688 * u64 bpf_ktime_get_ns(void)
9cde0c88
QM
689 * Description
690 * Return the time elapsed since system boot, in nanoseconds.
71d19214 691 * Does not include time the system was suspended.
ff20460e 692 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
9cde0c88
QM
693 * Return
694 * Current *ktime*.
695 *
696 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
697 * Description
698 * This helper is a "printk()-like" facility for debugging. It
699 * prints a message defined by format *fmt* (of size *fmt_size*)
700 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
701 * available. It can take up to three additional **u64**
702 * arguments (as an eBPF helpers, the total number of arguments is
703 * limited to five).
704 *
705 * Each time the helper is called, it appends a line to the trace.
1f8919b1
PW
706 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
707 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
9cde0c88
QM
708 * The format of the trace is customizable, and the exact output
709 * one will get depends on the options set in
710 * *\/sys/kernel/debug/tracing/trace_options* (see also the
711 * *README* file under the same directory). However, it usually
712 * defaults to something like:
713 *
714 * ::
715 *
716 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
717 *
718 * In the above:
719 *
720 * * ``telnet`` is the name of the current task.
721 * * ``470`` is the PID of the current task.
722 * * ``001`` is the CPU number on which the task is
723 * running.
724 * * In ``.N..``, each character refers to a set of
725 * options (whether irqs are enabled, scheduling
726 * options, whether hard/softirqs are running, level of
727 * preempt_disabled respectively). **N** means that
728 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
729 * are set.
730 * * ``419421.045894`` is a timestamp.
731 * * ``0x00000001`` is a fake value used by BPF for the
732 * instruction pointer register.
733 * * ``<formatted msg>`` is the message formatted with
734 * *fmt*.
735 *
736 * The conversion specifiers supported by *fmt* are similar, but
737 * more limited than for printk(). They are **%d**, **%i**,
738 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
739 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
740 * of field, padding with zeroes, etc.) is available, and the
741 * helper will return **-EINVAL** (but print nothing) if it
742 * encounters an unknown specifier.
743 *
744 * Also, note that **bpf_trace_printk**\ () is slow, and should
745 * only be used for debugging purposes. For this reason, a notice
746 * bloc (spanning several lines) is printed to kernel logs and
747 * states that the helper should not be used "for production use"
748 * the first time this helper is used (or more precisely, when
749 * **trace_printk**\ () buffers are allocated). For passing values
750 * to user space, perf events should be preferred.
751 * Return
752 * The number of bytes written to the buffer, or a negative error
753 * in case of failure.
754 *
755 * u32 bpf_get_prandom_u32(void)
756 * Description
757 * Get a pseudo-random number.
758 *
759 * From a security point of view, this helper uses its own
760 * pseudo-random internal state, and cannot be used to infer the
761 * seed of other random functions in the kernel. However, it is
762 * essential to note that the generator used by the helper is not
763 * cryptographically secure.
764 * Return
765 * A random 32-bit unsigned value.
766 *
767 * u32 bpf_get_smp_processor_id(void)
768 * Description
769 * Get the SMP (symmetric multiprocessing) processor id. Note that
770 * all programs run with preemption disabled, which means that the
771 * SMP processor id is stable during all the execution of the
772 * program.
773 * Return
774 * The SMP id of the processor running the program.
775 *
776 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
777 * Description
778 * Store *len* bytes from address *from* into the packet
779 * associated to *skb*, at *offset*. *flags* are a combination of
780 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
781 * checksum for the packet after storing the bytes) and
782 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
783 * **->swhash** and *skb*\ **->l4hash** to 0).
784 *
c1fe1e70 785 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
786 * packet buffer. Therefore, at load time, all checks on pointers
787 * previously done by the verifier are invalidated and must be
788 * performed again, if the helper is used in combination with
789 * direct packet access.
790 * Return
791 * 0 on success, or a negative error in case of failure.
792 *
793 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
794 * Description
795 * Recompute the layer 3 (e.g. IP) checksum for the packet
796 * associated to *skb*. Computation is incremental, so the helper
797 * must know the former value of the header field that was
798 * modified (*from*), the new value of this field (*to*), and the
799 * number of bytes (2 or 4) for this field, stored in *size*.
800 * Alternatively, it is possible to store the difference between
801 * the previous and the new values of the header field in *to*, by
802 * setting *from* and *size* to 0. For both methods, *offset*
803 * indicates the location of the IP checksum within the packet.
804 *
805 * This helper works in combination with **bpf_csum_diff**\ (),
806 * which does not update the checksum in-place, but offers more
807 * flexibility and can handle sizes larger than 2 or 4 for the
808 * checksum to update.
809 *
c1fe1e70 810 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
811 * packet buffer. Therefore, at load time, all checks on pointers
812 * previously done by the verifier are invalidated and must be
813 * performed again, if the helper is used in combination with
814 * direct packet access.
815 * Return
816 * 0 on success, or a negative error in case of failure.
817 *
818 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
819 * Description
820 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
821 * packet associated to *skb*. Computation is incremental, so the
822 * helper must know the former value of the header field that was
823 * modified (*from*), the new value of this field (*to*), and the
824 * number of bytes (2 or 4) for this field, stored on the lowest
825 * four bits of *flags*. Alternatively, it is possible to store
826 * the difference between the previous and the new values of the
827 * header field in *to*, by setting *from* and the four lowest
828 * bits of *flags* to 0. For both methods, *offset* indicates the
829 * location of the IP checksum within the packet. In addition to
830 * the size of the field, *flags* can be added (bitwise OR) actual
831 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
832 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
833 * for updates resulting in a null checksum the value is set to
834 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
835 * the checksum is to be computed against a pseudo-header.
836 *
837 * This helper works in combination with **bpf_csum_diff**\ (),
838 * which does not update the checksum in-place, but offers more
839 * flexibility and can handle sizes larger than 2 or 4 for the
840 * checksum to update.
841 *
c1fe1e70 842 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
843 * packet buffer. Therefore, at load time, all checks on pointers
844 * previously done by the verifier are invalidated and must be
845 * performed again, if the helper is used in combination with
846 * direct packet access.
847 * Return
848 * 0 on success, or a negative error in case of failure.
849 *
850 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
851 * Description
852 * This special helper is used to trigger a "tail call", or in
853 * other words, to jump into another eBPF program. The same stack
854 * frame is used (but values on stack and in registers for the
855 * caller are not accessible to the callee). This mechanism allows
856 * for program chaining, either for raising the maximum number of
857 * available eBPF instructions, or to execute given programs in
858 * conditional blocks. For security reasons, there is an upper
859 * limit to the number of successive tail calls that can be
860 * performed.
861 *
862 * Upon call of this helper, the program attempts to jump into a
863 * program referenced at index *index* in *prog_array_map*, a
864 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
865 * *ctx*, a pointer to the context.
866 *
867 * If the call succeeds, the kernel immediately runs the first
868 * instruction of the new program. This is not a function call,
869 * and it never returns to the previous program. If the call
870 * fails, then the helper has no effect, and the caller continues
871 * to run its subsequent instructions. A call can fail if the
872 * destination program for the jump does not exist (i.e. *index*
873 * is superior to the number of entries in *prog_array_map*), or
874 * if the maximum number of tail calls has been reached for this
875 * chain of programs. This limit is defined in the kernel by the
876 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
877 * which is currently set to 32.
878 * Return
879 * 0 on success, or a negative error in case of failure.
880 *
881 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
882 * Description
883 * Clone and redirect the packet associated to *skb* to another
884 * net device of index *ifindex*. Both ingress and egress
885 * interfaces can be used for redirection. The **BPF_F_INGRESS**
886 * value in *flags* is used to make the distinction (ingress path
887 * is selected if the flag is present, egress path otherwise).
888 * This is the only flag supported for now.
889 *
890 * In comparison with **bpf_redirect**\ () helper,
891 * **bpf_clone_redirect**\ () has the associated cost of
892 * duplicating the packet buffer, but this can be executed out of
893 * the eBPF program. Conversely, **bpf_redirect**\ () is more
894 * efficient, but it is handled through an action code where the
895 * redirection happens only after the eBPF program has returned.
896 *
c1fe1e70 897 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
898 * packet buffer. Therefore, at load time, all checks on pointers
899 * previously done by the verifier are invalidated and must be
900 * performed again, if the helper is used in combination with
901 * direct packet access.
902 * Return
903 * 0 on success, or a negative error in case of failure.
0cb34dc2
JS
904 *
905 * u64 bpf_get_current_pid_tgid(void)
9cde0c88
QM
906 * Return
907 * A 64-bit integer containing the current tgid and pid, and
908 * created as such:
909 * *current_task*\ **->tgid << 32 \|**
910 * *current_task*\ **->pid**.
0cb34dc2
JS
911 *
912 * u64 bpf_get_current_uid_gid(void)
9cde0c88
QM
913 * Return
914 * A 64-bit integer containing the current GID and UID, and
915 * created as such: *current_gid* **<< 32 \|** *current_uid*.
916 *
5f0e5412 917 * int bpf_get_current_comm(void *buf, u32 size_of_buf)
9cde0c88
QM
918 * Description
919 * Copy the **comm** attribute of the current task into *buf* of
920 * *size_of_buf*. The **comm** attribute contains the name of
921 * the executable (excluding the path) for the current task. The
922 * *size_of_buf* must be strictly positive. On success, the
923 * helper makes sure that the *buf* is NUL-terminated. On failure,
924 * it is filled with zeroes.
925 * Return
926 * 0 on success, or a negative error in case of failure.
927 *
928 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
929 * Description
930 * Retrieve the classid for the current task, i.e. for the net_cls
931 * cgroup to which *skb* belongs.
932 *
933 * This helper can be used on TC egress path, but not on ingress.
934 *
935 * The net_cls cgroup provides an interface to tag network packets
936 * based on a user-provided identifier for all traffic coming from
937 * the tasks belonging to the related cgroup. See also the related
938 * kernel documentation, available from the Linux sources in file
da82c92f 939 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
9cde0c88
QM
940 *
941 * The Linux kernel has two versions for cgroups: there are
942 * cgroups v1 and cgroups v2. Both are available to users, who can
943 * use a mixture of them, but note that the net_cls cgroup is for
944 * cgroup v1 only. This makes it incompatible with BPF programs
945 * run on cgroups, which is a cgroup-v2-only feature (a socket can
946 * only hold data for one version of cgroups at a time).
947 *
948 * This helper is only available is the kernel was compiled with
949 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
950 * "**y**" or to "**m**".
951 * Return
952 * The classid, or 0 for the default unconfigured classid.
953 *
954 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
955 * Description
956 * Push a *vlan_tci* (VLAN tag control information) of protocol
957 * *vlan_proto* to the packet associated to *skb*, then update
958 * the checksum. Note that if *vlan_proto* is different from
959 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
960 * be **ETH_P_8021Q**.
961 *
c1fe1e70 962 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
963 * packet buffer. Therefore, at load time, all checks on pointers
964 * previously done by the verifier are invalidated and must be
965 * performed again, if the helper is used in combination with
966 * direct packet access.
967 * Return
968 * 0 on success, or a negative error in case of failure.
969 *
970 * int bpf_skb_vlan_pop(struct sk_buff *skb)
971 * Description
972 * Pop a VLAN header from the packet associated to *skb*.
973 *
c1fe1e70 974 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
975 * packet buffer. Therefore, at load time, all checks on pointers
976 * previously done by the verifier are invalidated and must be
977 * performed again, if the helper is used in combination with
978 * direct packet access.
979 * Return
980 * 0 on success, or a negative error in case of failure.
981 *
982 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
983 * Description
984 * Get tunnel metadata. This helper takes a pointer *key* to an
985 * empty **struct bpf_tunnel_key** of **size**, that will be
986 * filled with tunnel metadata for the packet associated to *skb*.
987 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
988 * indicates that the tunnel is based on IPv6 protocol instead of
989 * IPv4.
990 *
991 * The **struct bpf_tunnel_key** is an object that generalizes the
992 * principal parameters used by various tunneling protocols into a
993 * single struct. This way, it can be used to easily make a
994 * decision based on the contents of the encapsulation header,
995 * "summarized" in this struct. In particular, it holds the IP
996 * address of the remote end (IPv4 or IPv6, depending on the case)
997 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
998 * this struct exposes the *key*\ **->tunnel_id**, which is
999 * generally mapped to a VNI (Virtual Network Identifier), making
1000 * it programmable together with the **bpf_skb_set_tunnel_key**\
1001 * () helper.
1002 *
1003 * Let's imagine that the following code is part of a program
1004 * attached to the TC ingress interface, on one end of a GRE
1005 * tunnel, and is supposed to filter out all messages coming from
1006 * remote ends with IPv4 address other than 10.0.0.1:
1007 *
1008 * ::
1009 *
1010 * int ret;
1011 * struct bpf_tunnel_key key = {};
1012 *
1013 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1014 * if (ret < 0)
1015 * return TC_ACT_SHOT; // drop packet
1016 *
1017 * if (key.remote_ipv4 != 0x0a000001)
1018 * return TC_ACT_SHOT; // drop packet
1019 *
1020 * return TC_ACT_OK; // accept packet
1021 *
1022 * This interface can also be used with all encapsulation devices
1023 * that can operate in "collect metadata" mode: instead of having
1024 * one network device per specific configuration, the "collect
1025 * metadata" mode only requires a single device where the
1026 * configuration can be extracted from this helper.
1027 *
1028 * This can be used together with various tunnels such as VXLan,
1029 * Geneve, GRE or IP in IP (IPIP).
1030 * Return
1031 * 0 on success, or a negative error in case of failure.
1032 *
1033 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1034 * Description
1035 * Populate tunnel metadata for packet associated to *skb.* The
1036 * tunnel metadata is set to the contents of *key*, of *size*. The
1037 * *flags* can be set to a combination of the following values:
1038 *
1039 * **BPF_F_TUNINFO_IPV6**
1040 * Indicate that the tunnel is based on IPv6 protocol
1041 * instead of IPv4.
1042 * **BPF_F_ZERO_CSUM_TX**
1043 * For IPv4 packets, add a flag to tunnel metadata
1044 * indicating that checksum computation should be skipped
1045 * and checksum set to zeroes.
1046 * **BPF_F_DONT_FRAGMENT**
1047 * Add a flag to tunnel metadata indicating that the
1048 * packet should not be fragmented.
1049 * **BPF_F_SEQ_NUMBER**
1050 * Add a flag to tunnel metadata indicating that a
1051 * sequence number should be added to tunnel header before
1052 * sending the packet. This flag was added for GRE
1053 * encapsulation, but might be used with other protocols
1054 * as well in the future.
1055 *
1056 * Here is a typical usage on the transmit path:
1057 *
1058 * ::
1059 *
1060 * struct bpf_tunnel_key key;
1061 * populate key ...
1062 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1063 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1064 *
1065 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1066 * helper for additional information.
1067 * Return
1068 * 0 on success, or a negative error in case of failure.
1069 *
1070 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1071 * Description
1072 * Read the value of a perf event counter. This helper relies on a
1073 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1074 * the perf event counter is selected when *map* is updated with
1075 * perf event file descriptors. The *map* is an array whose size
1076 * is the number of available CPUs, and each cell contains a value
1077 * relative to one CPU. The value to retrieve is indicated by
1078 * *flags*, that contains the index of the CPU to look up, masked
1079 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1080 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1081 * current CPU should be retrieved.
1082 *
1083 * Note that before Linux 4.13, only hardware perf event can be
1084 * retrieved.
1085 *
1086 * Also, be aware that the newer helper
1087 * **bpf_perf_event_read_value**\ () is recommended over
a56497d3 1088 * **bpf_perf_event_read**\ () in general. The latter has some ABI
9cde0c88
QM
1089 * quirks where error and counter value are used as a return code
1090 * (which is wrong to do since ranges may overlap). This issue is
a56497d3
QM
1091 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1092 * time provides more features over the **bpf_perf_event_read**\
1093 * () interface. Please refer to the description of
9cde0c88
QM
1094 * **bpf_perf_event_read_value**\ () for details.
1095 * Return
1096 * The value of the perf event counter read from the map, or a
1097 * negative error code in case of failure.
1098 *
1099 * int bpf_redirect(u32 ifindex, u64 flags)
1100 * Description
1101 * Redirect the packet to another net device of index *ifindex*.
1102 * This helper is somewhat similar to **bpf_clone_redirect**\
1103 * (), except that the packet is not cloned, which provides
1104 * increased performance.
1105 *
1106 * Except for XDP, both ingress and egress interfaces can be used
1107 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1108 * to make the distinction (ingress path is selected if the flag
1109 * is present, egress path otherwise). Currently, XDP only
1110 * supports redirection to the egress interface, and accepts no
1111 * flag at all.
1112 *
f25975f4
THJ
1113 * The same effect can also be attained with the more generic
1114 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1115 * redirect target instead of providing it directly to the helper.
9cde0c88
QM
1116 * Return
1117 * For XDP, the helper returns **XDP_REDIRECT** on success or
1118 * **XDP_ABORTED** on error. For other program types, the values
1119 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1120 * error.
1121 *
1122 * u32 bpf_get_route_realm(struct sk_buff *skb)
1123 * Description
1124 * Retrieve the realm or the route, that is to say the
1125 * **tclassid** field of the destination for the *skb*. The
1126 * indentifier retrieved is a user-provided tag, similar to the
1127 * one used with the net_cls cgroup (see description for
1128 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1129 * held by a route (a destination entry), not by a task.
1130 *
1131 * Retrieving this identifier works with the clsact TC egress hook
1132 * (see also **tc-bpf(8)**), or alternatively on conventional
1133 * classful egress qdiscs, but not on TC ingress path. In case of
1134 * clsact TC egress hook, this has the advantage that, internally,
1135 * the destination entry has not been dropped yet in the transmit
1136 * path. Therefore, the destination entry does not need to be
1137 * artificially held via **netif_keep_dst**\ () for a classful
1138 * qdisc until the *skb* is freed.
1139 *
1140 * This helper is available only if the kernel was compiled with
1141 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1142 * Return
1143 * The realm of the route for the packet associated to *skb*, or 0
1144 * if none was found.
1145 *
5f0e5412 1146 * int bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
9cde0c88
QM
1147 * Description
1148 * Write raw *data* blob into a special BPF perf event held by
1149 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1150 * event must have the following attributes: **PERF_SAMPLE_RAW**
1151 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1152 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1153 *
1154 * The *flags* are used to indicate the index in *map* for which
1155 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1156 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1157 * to indicate that the index of the current CPU core should be
1158 * used.
1159 *
1160 * The value to write, of *size*, is passed through eBPF stack and
1161 * pointed by *data*.
1162 *
1163 * The context of the program *ctx* needs also be passed to the
1164 * helper.
1165 *
1166 * On user space, a program willing to read the values needs to
1167 * call **perf_event_open**\ () on the perf event (either for
1168 * one or for all CPUs) and to store the file descriptor into the
1169 * *map*. This must be done before the eBPF program can send data
1170 * into it. An example is available in file
1171 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1172 * tree (the eBPF program counterpart is in
1173 * *samples/bpf/trace_output_kern.c*).
1174 *
1175 * **bpf_perf_event_output**\ () achieves better performance
1176 * than **bpf_trace_printk**\ () for sharing data with user
1177 * space, and is much better suitable for streaming data from eBPF
1178 * programs.
1179 *
1180 * Note that this helper is not restricted to tracing use cases
1181 * and can be used with programs attached to TC or XDP as well,
1182 * where it allows for passing data to user space listeners. Data
1183 * can be:
1184 *
1185 * * Only custom structs,
1186 * * Only the packet payload, or
1187 * * A combination of both.
1188 * Return
1189 * 0 on success, or a negative error in case of failure.
1190 *
5f0e5412 1191 * int bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
9cde0c88
QM
1192 * Description
1193 * This helper was provided as an easy way to load data from a
1194 * packet. It can be used to load *len* bytes from *offset* from
1195 * the packet associated to *skb*, into the buffer pointed by
1196 * *to*.
1197 *
1198 * Since Linux 4.7, usage of this helper has mostly been replaced
1199 * by "direct packet access", enabling packet data to be
1200 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1201 * pointing respectively to the first byte of packet data and to
1202 * the byte after the last byte of packet data. However, it
1203 * remains useful if one wishes to read large quantities of data
1204 * at once from a packet into the eBPF stack.
1205 * Return
1206 * 0 on success, or a negative error in case of failure.
1207 *
5f0e5412 1208 * int bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
9cde0c88
QM
1209 * Description
1210 * Walk a user or a kernel stack and return its id. To achieve
1211 * this, the helper needs *ctx*, which is a pointer to the context
1212 * on which the tracing program is executed, and a pointer to a
1213 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1214 *
1215 * The last argument, *flags*, holds the number of stack frames to
1216 * skip (from 0 to 255), masked with
1217 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1218 * a combination of the following flags:
1219 *
1220 * **BPF_F_USER_STACK**
1221 * Collect a user space stack instead of a kernel stack.
1222 * **BPF_F_FAST_STACK_CMP**
1223 * Compare stacks by hash only.
1224 * **BPF_F_REUSE_STACKID**
1225 * If two different stacks hash into the same *stackid*,
1226 * discard the old one.
1227 *
1228 * The stack id retrieved is a 32 bit long integer handle which
1229 * can be further combined with other data (including other stack
1230 * ids) and used as a key into maps. This can be useful for
1231 * generating a variety of graphs (such as flame graphs or off-cpu
1232 * graphs).
1233 *
1234 * For walking a stack, this helper is an improvement over
1235 * **bpf_probe_read**\ (), which can be used with unrolled loops
1236 * but is not efficient and consumes a lot of eBPF instructions.
1237 * Instead, **bpf_get_stackid**\ () can collect up to
1238 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1239 * this limit can be controlled with the **sysctl** program, and
1240 * that it should be manually increased in order to profile long
1241 * user stacks (such as stacks for Java programs). To do so, use:
1242 *
1243 * ::
1244 *
1245 * # sysctl kernel.perf_event_max_stack=<new value>
9cde0c88
QM
1246 * Return
1247 * The positive or null stack id on success, or a negative error
1248 * in case of failure.
1249 *
1250 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1251 * Description
1252 * Compute a checksum difference, from the raw buffer pointed by
1253 * *from*, of length *from_size* (that must be a multiple of 4),
1254 * towards the raw buffer pointed by *to*, of size *to_size*
1255 * (same remark). An optional *seed* can be added to the value
1256 * (this can be cascaded, the seed may come from a previous call
1257 * to the helper).
1258 *
1259 * This is flexible enough to be used in several ways:
1260 *
1261 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1262 * checksum, it can be used when pushing new data.
1263 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1264 * checksum, it can be used when removing data from a packet.
1265 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1266 * can be used to compute a diff. Note that *from_size* and
1267 * *to_size* do not need to be equal.
1268 *
1269 * This helper can be used in combination with
1270 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1271 * which one can feed in the difference computed with
1272 * **bpf_csum_diff**\ ().
1273 * Return
1274 * The checksum result, or a negative error code in case of
1275 * failure.
1276 *
5f0e5412 1277 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
9cde0c88
QM
1278 * Description
1279 * Retrieve tunnel options metadata for the packet associated to
1280 * *skb*, and store the raw tunnel option data to the buffer *opt*
1281 * of *size*.
1282 *
1283 * This helper can be used with encapsulation devices that can
1284 * operate in "collect metadata" mode (please refer to the related
1285 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1286 * more details). A particular example where this can be used is
1287 * in combination with the Geneve encapsulation protocol, where it
1288 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1289 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1290 * the eBPF program. This allows for full customization of these
1291 * headers.
1292 * Return
1293 * The size of the option data retrieved.
1294 *
5f0e5412 1295 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
9cde0c88
QM
1296 * Description
1297 * Set tunnel options metadata for the packet associated to *skb*
1298 * to the option data contained in the raw buffer *opt* of *size*.
1299 *
1300 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1301 * helper for additional information.
1302 * Return
1303 * 0 on success, or a negative error in case of failure.
1304 *
1305 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1306 * Description
1307 * Change the protocol of the *skb* to *proto*. Currently
1308 * supported are transition from IPv4 to IPv6, and from IPv6 to
1309 * IPv4. The helper takes care of the groundwork for the
1310 * transition, including resizing the socket buffer. The eBPF
1311 * program is expected to fill the new headers, if any, via
1312 * **skb_store_bytes**\ () and to recompute the checksums with
1313 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1314 * (). The main case for this helper is to perform NAT64
1315 * operations out of an eBPF program.
1316 *
1317 * Internally, the GSO type is marked as dodgy so that headers are
1318 * checked and segments are recalculated by the GSO/GRO engine.
1319 * The size for GSO target is adapted as well.
1320 *
1321 * All values for *flags* are reserved for future usage, and must
1322 * be left at zero.
1323 *
c1fe1e70 1324 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1325 * packet buffer. Therefore, at load time, all checks on pointers
1326 * previously done by the verifier are invalidated and must be
1327 * performed again, if the helper is used in combination with
1328 * direct packet access.
1329 * Return
1330 * 0 on success, or a negative error in case of failure.
1331 *
1332 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1333 * Description
1334 * Change the packet type for the packet associated to *skb*. This
1335 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1336 * the eBPF program does not have a write access to *skb*\
1337 * **->pkt_type** beside this helper. Using a helper here allows
1338 * for graceful handling of errors.
1339 *
1340 * The major use case is to change incoming *skb*s to
1341 * **PACKET_HOST** in a programmatic way instead of having to
1342 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1343 * example.
1344 *
1345 * Note that *type* only allows certain values. At this time, they
1346 * are:
1347 *
1348 * **PACKET_HOST**
1349 * Packet is for us.
1350 * **PACKET_BROADCAST**
1351 * Send packet to all.
1352 * **PACKET_MULTICAST**
1353 * Send packet to group.
1354 * **PACKET_OTHERHOST**
1355 * Send packet to someone else.
1356 * Return
1357 * 0 on success, or a negative error in case of failure.
1358 *
1359 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1360 * Description
1361 * Check whether *skb* is a descendant of the cgroup2 held by
1362 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1363 * Return
1364 * The return value depends on the result of the test, and can be:
1365 *
1366 * * 0, if the *skb* failed the cgroup2 descendant test.
1367 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1368 * * A negative error code, if an error occurred.
1369 *
1370 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1371 * Description
1372 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1373 * not set, in particular if the hash was cleared due to mangling,
1374 * recompute this hash. Later accesses to the hash can be done
1375 * directly with *skb*\ **->hash**.
1376 *
1377 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1378 * prototype with **bpf_skb_change_proto**\ (), or calling
1379 * **bpf_skb_store_bytes**\ () with the
1380 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1381 * the hash and to trigger a new computation for the next call to
1382 * **bpf_get_hash_recalc**\ ().
1383 * Return
1384 * The 32-bit hash.
0cb34dc2
JS
1385 *
1386 * u64 bpf_get_current_task(void)
9cde0c88
QM
1387 * Return
1388 * A pointer to the current task struct.
1389 *
1390 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1391 * Description
1392 * Attempt in a safe way to write *len* bytes from the buffer
1393 * *src* to *dst* in memory. It only works for threads that are in
1394 * user context, and *dst* must be a valid user space address.
1395 *
1396 * This helper should not be used to implement any kind of
1397 * security mechanism because of TOC-TOU attacks, but rather to
1398 * debug, divert, and manipulate execution of semi-cooperative
1399 * processes.
1400 *
1401 * Keep in mind that this feature is meant for experiments, and it
1402 * has a risk of crashing the system and running programs.
1403 * Therefore, when an eBPF program using this helper is attached,
1404 * a warning including PID and process name is printed to kernel
1405 * logs.
1406 * Return
1407 * 0 on success, or a negative error in case of failure.
1408 *
1409 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1410 * Description
1411 * Check whether the probe is being run is the context of a given
1412 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1413 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1414 * Return
1415 * The return value depends on the result of the test, and can be:
1416 *
1417 * * 0, if the *skb* task belongs to the cgroup2.
1418 * * 1, if the *skb* task does not belong to the cgroup2.
1419 * * A negative error code, if an error occurred.
1420 *
1421 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1422 * Description
1423 * Resize (trim or grow) the packet associated to *skb* to the
1424 * new *len*. The *flags* are reserved for future usage, and must
1425 * be left at zero.
1426 *
1427 * The basic idea is that the helper performs the needed work to
1428 * change the size of the packet, then the eBPF program rewrites
1429 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1430 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1431 * and others. This helper is a slow path utility intended for
1432 * replies with control messages. And because it is targeted for
1433 * slow path, the helper itself can afford to be slow: it
1434 * implicitly linearizes, unclones and drops offloads from the
1435 * *skb*.
1436 *
c1fe1e70 1437 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1438 * packet buffer. Therefore, at load time, all checks on pointers
1439 * previously done by the verifier are invalidated and must be
1440 * performed again, if the helper is used in combination with
1441 * direct packet access.
1442 * Return
1443 * 0 on success, or a negative error in case of failure.
1444 *
1445 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1446 * Description
1447 * Pull in non-linear data in case the *skb* is non-linear and not
1448 * all of *len* are part of the linear section. Make *len* bytes
1449 * from *skb* readable and writable. If a zero value is passed for
1450 * *len*, then the whole length of the *skb* is pulled.
1451 *
1452 * This helper is only needed for reading and writing with direct
1453 * packet access.
1454 *
1455 * For direct packet access, testing that offsets to access
1456 * are within packet boundaries (test on *skb*\ **->data_end**) is
1457 * susceptible to fail if offsets are invalid, or if the requested
1458 * data is in non-linear parts of the *skb*. On failure the
1459 * program can just bail out, or in the case of a non-linear
1460 * buffer, use a helper to make the data available. The
1461 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1462 * the data. Another one consists in using **bpf_skb_pull_data**
1463 * to pull in once the non-linear parts, then retesting and
1464 * eventually access the data.
1465 *
1466 * At the same time, this also makes sure the *skb* is uncloned,
1467 * which is a necessary condition for direct write. As this needs
1468 * to be an invariant for the write part only, the verifier
1469 * detects writes and adds a prologue that is calling
1470 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1471 * the very beginning in case it is indeed cloned.
1472 *
c1fe1e70 1473 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1474 * packet buffer. Therefore, at load time, all checks on pointers
1475 * previously done by the verifier are invalidated and must be
1476 * performed again, if the helper is used in combination with
1477 * direct packet access.
1478 * Return
1479 * 0 on success, or a negative error in case of failure.
1480 *
1481 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1482 * Description
1483 * Add the checksum *csum* into *skb*\ **->csum** in case the
1484 * driver has supplied a checksum for the entire packet into that
1485 * field. Return an error otherwise. This helper is intended to be
1486 * used in combination with **bpf_csum_diff**\ (), in particular
1487 * when the checksum needs to be updated after data has been
1488 * written into the packet through direct packet access.
1489 * Return
1490 * The checksum on success, or a negative error code in case of
1491 * failure.
1492 *
1493 * void bpf_set_hash_invalid(struct sk_buff *skb)
1494 * Description
1495 * Invalidate the current *skb*\ **->hash**. It can be used after
1496 * mangling on headers through direct packet access, in order to
1497 * indicate that the hash is outdated and to trigger a
1498 * recalculation the next time the kernel tries to access this
1499 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1500 *
1501 * int bpf_get_numa_node_id(void)
1502 * Description
1503 * Return the id of the current NUMA node. The primary use case
1504 * for this helper is the selection of sockets for the local NUMA
1505 * node, when the program is attached to sockets using the
1506 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1507 * but the helper is also available to other eBPF program types,
1508 * similarly to **bpf_get_smp_processor_id**\ ().
1509 * Return
1510 * The id of current NUMA node.
1511 *
1512 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1513 * Description
1514 * Grows headroom of packet associated to *skb* and adjusts the
1515 * offset of the MAC header accordingly, adding *len* bytes of
1516 * space. It automatically extends and reallocates memory as
1517 * required.
1518 *
1519 * This helper can be used on a layer 3 *skb* to push a MAC header
1520 * for redirection into a layer 2 device.
1521 *
1522 * All values for *flags* are reserved for future usage, and must
1523 * be left at zero.
1524 *
c1fe1e70 1525 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1526 * packet buffer. Therefore, at load time, all checks on pointers
1527 * previously done by the verifier are invalidated and must be
1528 * performed again, if the helper is used in combination with
1529 * direct packet access.
1530 * Return
1531 * 0 on success, or a negative error in case of failure.
1532 *
1533 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1534 * Description
1535 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1536 * it is possible to use a negative value for *delta*. This helper
1537 * can be used to prepare the packet for pushing or popping
1538 * headers.
1539 *
c1fe1e70 1540 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1541 * packet buffer. Therefore, at load time, all checks on pointers
1542 * previously done by the verifier are invalidated and must be
1543 * performed again, if the helper is used in combination with
1544 * direct packet access.
1545 * Return
1546 * 0 on success, or a negative error in case of failure.
9a738266 1547 *
6ae08ae3 1548 * int bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
9cde0c88 1549 * Description
6ae08ae3 1550 * Copy a NUL terminated string from an unsafe kernel address
ff20460e 1551 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
6ae08ae3 1552 * more details.
9cde0c88 1553 *
ff20460e
QM
1554 * Generally, use **bpf_probe_read_user_str**\ () or
1555 * **bpf_probe_read_kernel_str**\ () instead.
9cde0c88
QM
1556 * Return
1557 * On success, the strictly positive length of the string,
1558 * including the trailing NUL character. On error, a negative
1559 * value.
1560 *
1561 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1562 * Description
1563 * If the **struct sk_buff** pointed by *skb* has a known socket,
1564 * retrieve the cookie (generated by the kernel) of this socket.
1565 * If no cookie has been set yet, generate a new cookie. Once
1566 * generated, the socket cookie remains stable for the life of the
1567 * socket. This helper can be useful for monitoring per socket
609a2ca5
DB
1568 * networking traffic statistics as it provides a global socket
1569 * identifier that can be assumed unique.
9cde0c88
QM
1570 * Return
1571 * A 8-byte long non-decreasing number on success, or 0 if the
1572 * socket field is missing inside *skb*.
1573 *
a40b712e
AI
1574 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1575 * Description
1576 * Equivalent to bpf_get_socket_cookie() helper that accepts
ea6eced0 1577 * *skb*, but gets socket from **struct bpf_sock_addr** context.
a40b712e
AI
1578 * Return
1579 * A 8-byte long non-decreasing number.
1580 *
1581 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1582 * Description
ff20460e 1583 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
ea6eced0 1584 * *skb*, but gets socket from **struct bpf_sock_ops** context.
a40b712e
AI
1585 * Return
1586 * A 8-byte long non-decreasing number.
1587 *
9cde0c88
QM
1588 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1589 * Return
1590 * The owner UID of the socket associated to *skb*. If the socket
1591 * is **NULL**, or if it is not a full socket (i.e. if it is a
1592 * time-wait or a request socket instead), **overflowuid** value
1593 * is returned (note that **overflowuid** might also be the actual
1594 * UID value for the socket).
1595 *
1596 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1597 * Description
1598 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1599 * to value *hash*.
1600 * Return
1601 * 0
1602 *
beecf11b 1603 * int bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
9cde0c88
QM
1604 * Description
1605 * Emulate a call to **setsockopt()** on the socket associated to
1606 * *bpf_socket*, which must be a full socket. The *level* at
1607 * which the option resides and the name *optname* of the option
1608 * must be specified, see **setsockopt(2)** for more information.
1609 * The option value of length *optlen* is pointed by *optval*.
1610 *
beecf11b 1611 * *bpf_socket* should be one of the following:
ff20460e 1612 *
beecf11b
SF
1613 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1614 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1615 * and **BPF_CGROUP_INET6_CONNECT**.
1616 *
9cde0c88
QM
1617 * This helper actually implements a subset of **setsockopt()**.
1618 * It supports the following *level*\ s:
1619 *
1620 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1621 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1622 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1623 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1624 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1625 * **TCP_BPF_SNDCWND_CLAMP**.
1626 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1627 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1628 * Return
1629 * 0 on success, or a negative error in case of failure.
1630 *
b55cbc8d 1631 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
9cde0c88
QM
1632 * Description
1633 * Grow or shrink the room for data in the packet associated to
1634 * *skb* by *len_diff*, and according to the selected *mode*.
1635 *
6c408dec
WB
1636 * There are two supported modes at this time:
1637 *
1638 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1639 * (room space is added or removed below the layer 2 header).
9cde0c88
QM
1640 *
1641 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1642 * (room space is added or removed below the layer 3 header).
1643 *
6c408dec
WB
1644 * The following flags are supported at this time:
1645 *
1646 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1647 * Adjusting mss in this way is not allowed for datagrams.
1648 *
c1fe1e70
QM
1649 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1650 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
6c408dec
WB
1651 * Any new space is reserved to hold a tunnel header.
1652 * Configure skb offsets and other fields accordingly.
1653 *
c1fe1e70
QM
1654 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1655 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
6c408dec 1656 * Use with ENCAP_L3 flags to further specify the tunnel type.
9cde0c88 1657 *
c1fe1e70 1658 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
1db04c30 1659 * Use with ENCAP_L3/L4 flags to further specify the tunnel
c1fe1e70 1660 * type; *len* is the length of the inner MAC header.
1db04c30 1661 *
c1fe1e70 1662 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1663 * packet buffer. Therefore, at load time, all checks on pointers
1664 * previously done by the verifier are invalidated and must be
1665 * performed again, if the helper is used in combination with
1666 * direct packet access.
1667 * Return
1668 * 0 on success, or a negative error in case of failure.
1669 *
1670 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1671 * Description
1672 * Redirect the packet to the endpoint referenced by *map* at
1673 * index *key*. Depending on its type, this *map* can contain
1674 * references to net devices (for forwarding packets through other
1675 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1676 * but this is only implemented for native XDP (with driver
1677 * support) as of this writing).
1678 *
3745ee18
PP
1679 * The lower two bits of *flags* are used as the return code if
1680 * the map lookup fails. This is so that the return value can be
ff20460e
QM
1681 * one of the XDP program return codes up to **XDP_TX**, as chosen
1682 * by the caller. Any higher bits in the *flags* argument must be
3745ee18 1683 * unset.
9cde0c88 1684 *
ff20460e
QM
1685 * See also **bpf_redirect**\ (), which only supports redirecting
1686 * to an ifindex, but doesn't require a map to do so.
9cde0c88 1687 * Return
f25975f4 1688 * **XDP_REDIRECT** on success, or the value of the two lower bits
a33d3147 1689 * of the *flags* argument on error.
9cde0c88 1690 *
5f0e5412 1691 * int bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
9cde0c88
QM
1692 * Description
1693 * Redirect the packet to the socket referenced by *map* (of type
1694 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1695 * egress interfaces can be used for redirection. The
1696 * **BPF_F_INGRESS** value in *flags* is used to make the
1697 * distinction (ingress path is selected if the flag is present,
1698 * egress path otherwise). This is the only flag supported for now.
1699 * Return
1700 * **SK_PASS** on success, or **SK_DROP** on error.
1701 *
96871b9f 1702 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
9cde0c88
QM
1703 * Description
1704 * Add an entry to, or update a *map* referencing sockets. The
1705 * *skops* is used as a new value for the entry associated to
1706 * *key*. *flags* is one of:
1707 *
1708 * **BPF_NOEXIST**
1709 * The entry for *key* must not exist in the map.
1710 * **BPF_EXIST**
1711 * The entry for *key* must already exist in the map.
1712 * **BPF_ANY**
1713 * No condition on the existence of the entry for *key*.
1714 *
1715 * If the *map* has eBPF programs (parser and verdict), those will
1716 * be inherited by the socket being added. If the socket is
1717 * already attached to eBPF programs, this results in an error.
1718 * Return
1719 * 0 on success, or a negative error in case of failure.
1720 *
1721 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1722 * Description
1723 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1724 * *delta* (which can be positive or negative). Note that this
1725 * operation modifies the address stored in *xdp_md*\ **->data**,
1726 * so the latter must be loaded only after the helper has been
1727 * called.
1728 *
1729 * The use of *xdp_md*\ **->data_meta** is optional and programs
1730 * are not required to use it. The rationale is that when the
1731 * packet is processed with XDP (e.g. as DoS filter), it is
1732 * possible to push further meta data along with it before passing
1733 * to the stack, and to give the guarantee that an ingress eBPF
1734 * program attached as a TC classifier on the same device can pick
1735 * this up for further post-processing. Since TC works with socket
1736 * buffers, it remains possible to set from XDP the **mark** or
1737 * **priority** pointers, or other pointers for the socket buffer.
1738 * Having this scratch space generic and programmable allows for
1739 * more flexibility as the user is free to store whatever meta
1740 * data they need.
1741 *
c1fe1e70 1742 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1743 * packet buffer. Therefore, at load time, all checks on pointers
1744 * previously done by the verifier are invalidated and must be
1745 * performed again, if the helper is used in combination with
1746 * direct packet access.
1747 * Return
1748 * 0 on success, or a negative error in case of failure.
1749 *
1750 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1751 * Description
1752 * Read the value of a perf event counter, and store it into *buf*
1753 * of size *buf_size*. This helper relies on a *map* of type
1754 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1755 * counter is selected when *map* is updated with perf event file
1756 * descriptors. The *map* is an array whose size is the number of
1757 * available CPUs, and each cell contains a value relative to one
1758 * CPU. The value to retrieve is indicated by *flags*, that
1759 * contains the index of the CPU to look up, masked with
1760 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1761 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1762 * current CPU should be retrieved.
1763 *
1764 * This helper behaves in a way close to
1765 * **bpf_perf_event_read**\ () helper, save that instead of
1766 * just returning the value observed, it fills the *buf*
1767 * structure. This allows for additional data to be retrieved: in
1768 * particular, the enabled and running times (in *buf*\
1769 * **->enabled** and *buf*\ **->running**, respectively) are
1770 * copied. In general, **bpf_perf_event_read_value**\ () is
1771 * recommended over **bpf_perf_event_read**\ (), which has some
1772 * ABI issues and provides fewer functionalities.
1773 *
1774 * These values are interesting, because hardware PMU (Performance
1775 * Monitoring Unit) counters are limited resources. When there are
1776 * more PMU based perf events opened than available counters,
1777 * kernel will multiplex these events so each event gets certain
1778 * percentage (but not all) of the PMU time. In case that
1779 * multiplexing happens, the number of samples or counter value
1780 * will not reflect the case compared to when no multiplexing
1781 * occurs. This makes comparison between different runs difficult.
1782 * Typically, the counter value should be normalized before
1783 * comparing to other experiments. The usual normalization is done
1784 * as follows.
1785 *
1786 * ::
1787 *
1788 * normalized_counter = counter * t_enabled / t_running
1789 *
1790 * Where t_enabled is the time enabled for event and t_running is
1791 * the time running for event since last normalization. The
1792 * enabled and running times are accumulated since the perf event
1793 * open. To achieve scaling factor between two invocations of an
ff20460e 1794 * eBPF program, users can use CPU id as the key (which is
9cde0c88
QM
1795 * typical for perf array usage model) to remember the previous
1796 * value and do the calculation inside the eBPF program.
1797 * Return
1798 * 0 on success, or a negative error in case of failure.
1799 *
96871b9f 1800 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
9cde0c88
QM
1801 * Description
1802 * For en eBPF program attached to a perf event, retrieve the
1803 * value of the event counter associated to *ctx* and store it in
1804 * the structure pointed by *buf* and of size *buf_size*. Enabled
1805 * and running times are also stored in the structure (see
1806 * description of helper **bpf_perf_event_read_value**\ () for
1807 * more details).
1808 * Return
1809 * 0 on success, or a negative error in case of failure.
1810 *
beecf11b 1811 * int bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
9cde0c88
QM
1812 * Description
1813 * Emulate a call to **getsockopt()** on the socket associated to
1814 * *bpf_socket*, which must be a full socket. The *level* at
1815 * which the option resides and the name *optname* of the option
1816 * must be specified, see **getsockopt(2)** for more information.
1817 * The retrieved value is stored in the structure pointed by
1818 * *opval* and of length *optlen*.
1819 *
beecf11b 1820 * *bpf_socket* should be one of the following:
ff20460e 1821 *
beecf11b
SF
1822 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
1823 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
1824 * and **BPF_CGROUP_INET6_CONNECT**.
1825 *
9cde0c88
QM
1826 * This helper actually implements a subset of **getsockopt()**.
1827 * It supports the following *level*\ s:
1828 *
1829 * * **IPPROTO_TCP**, which supports *optname*
1830 * **TCP_CONGESTION**.
1831 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1832 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1833 * Return
1834 * 0 on success, or a negative error in case of failure.
1835 *
1f8919b1 1836 * int bpf_override_return(struct pt_regs *regs, u64 rc)
9cde0c88
QM
1837 * Description
1838 * Used for error injection, this helper uses kprobes to override
1839 * the return value of the probed function, and to set it to *rc*.
1840 * The first argument is the context *regs* on which the kprobe
1841 * works.
1842 *
ff20460e 1843 * This helper works by setting the PC (program counter)
9cde0c88
QM
1844 * to an override function which is run in place of the original
1845 * probed function. This means the probed function is not run at
1846 * all. The replacement function just returns with the required
1847 * value.
1848 *
1849 * This helper has security implications, and thus is subject to
1850 * restrictions. It is only available if the kernel was compiled
1851 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1852 * option, and in this case it only works on functions tagged with
1853 * **ALLOW_ERROR_INJECTION** in the kernel code.
1854 *
1855 * Also, the helper is only available for the architectures having
1856 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1857 * x86 architecture is the only one to support this feature.
1858 * Return
1859 * 0
1860 *
96871b9f 1861 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
9cde0c88
QM
1862 * Description
1863 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1864 * for the full TCP socket associated to *bpf_sock_ops* to
1865 * *argval*.
1866 *
1867 * The primary use of this field is to determine if there should
1868 * be calls to eBPF programs of type
1869 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1870 * code. A program of the same type can change its value, per
1871 * connection and as necessary, when the connection is
1872 * established. This field is directly accessible for reading, but
1873 * this helper must be used for updates in order to return an
1874 * error if an eBPF program tries to set a callback that is not
1875 * supported in the current kernel.
1876 *
4635b0ae 1877 * *argval* is a flag array which can combine these flags:
9cde0c88
QM
1878 *
1879 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1880 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1881 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
692cbaa9 1882 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
9cde0c88 1883 *
4635b0ae
MM
1884 * Therefore, this function can be used to clear a callback flag by
1885 * setting the appropriate bit to zero. e.g. to disable the RTO
1886 * callback:
1887 *
1888 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1889 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1890 *
9cde0c88
QM
1891 * Here are some examples of where one could call such eBPF
1892 * program:
1893 *
1894 * * When RTO fires.
1895 * * When a packet is retransmitted.
1896 * * When the connection terminates.
1897 * * When a packet is sent.
1898 * * When a packet is received.
1899 * Return
1900 * Code **-EINVAL** if the socket is not a full TCP socket;
1901 * otherwise, a positive number containing the bits that could not
1902 * be set is returned (which comes down to 0 if all bits were set
1903 * as required).
1904 *
1905 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1906 * Description
1907 * This helper is used in programs implementing policies at the
1908 * socket level. If the message *msg* is allowed to pass (i.e. if
1909 * the verdict eBPF program returns **SK_PASS**), redirect it to
1910 * the socket referenced by *map* (of type
1911 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1912 * egress interfaces can be used for redirection. The
1913 * **BPF_F_INGRESS** value in *flags* is used to make the
1914 * distinction (ingress path is selected if the flag is present,
1915 * egress path otherwise). This is the only flag supported for now.
1916 * Return
1917 * **SK_PASS** on success, or **SK_DROP** on error.
1918 *
1919 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1920 * Description
1921 * For socket policies, apply the verdict of the eBPF program to
1922 * the next *bytes* (number of bytes) of message *msg*.
1923 *
1924 * For example, this helper can be used in the following cases:
1925 *
1926 * * A single **sendmsg**\ () or **sendfile**\ () system call
1927 * contains multiple logical messages that the eBPF program is
1928 * supposed to read and for which it should apply a verdict.
1929 * * An eBPF program only cares to read the first *bytes* of a
1930 * *msg*. If the message has a large payload, then setting up
1931 * and calling the eBPF program repeatedly for all bytes, even
1932 * though the verdict is already known, would create unnecessary
1933 * overhead.
1934 *
1935 * When called from within an eBPF program, the helper sets a
1936 * counter internal to the BPF infrastructure, that is used to
1937 * apply the last verdict to the next *bytes*. If *bytes* is
1938 * smaller than the current data being processed from a
1939 * **sendmsg**\ () or **sendfile**\ () system call, the first
1940 * *bytes* will be sent and the eBPF program will be re-run with
1941 * the pointer for start of data pointing to byte number *bytes*
1942 * **+ 1**. If *bytes* is larger than the current data being
1943 * processed, then the eBPF verdict will be applied to multiple
1944 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1945 * consumed.
1946 *
1947 * Note that if a socket closes with the internal counter holding
1948 * a non-zero value, this is not a problem because data is not
1949 * being buffered for *bytes* and is sent as it is received.
1950 * Return
1951 * 0
1952 *
1953 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1954 * Description
1955 * For socket policies, prevent the execution of the verdict eBPF
1956 * program for message *msg* until *bytes* (byte number) have been
1957 * accumulated.
1958 *
1959 * This can be used when one needs a specific number of bytes
1960 * before a verdict can be assigned, even if the data spans
1961 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1962 * case would be a user calling **sendmsg**\ () repeatedly with
1963 * 1-byte long message segments. Obviously, this is bad for
1964 * performance, but it is still valid. If the eBPF program needs
1965 * *bytes* bytes to validate a header, this helper can be used to
1966 * prevent the eBPF program to be called again until *bytes* have
1967 * been accumulated.
1968 * Return
1969 * 0
1970 *
1971 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1972 * Description
1973 * For socket policies, pull in non-linear data from user space
1974 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1975 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1976 * respectively.
1977 *
1978 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1979 * *msg* it can only parse data that the (**data**, **data_end**)
1980 * pointers have already consumed. For **sendmsg**\ () hooks this
1981 * is likely the first scatterlist element. But for calls relying
1982 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1983 * be the range (**0**, **0**) because the data is shared with
1984 * user space and by default the objective is to avoid allowing
1985 * user space to modify data while (or after) eBPF verdict is
1986 * being decided. This helper can be used to pull in data and to
1987 * set the start and end pointer to given values. Data will be
1988 * copied if necessary (i.e. if data was not linear and if start
1989 * and end pointers do not point to the same chunk).
1990 *
c1fe1e70 1991 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
1992 * packet buffer. Therefore, at load time, all checks on pointers
1993 * previously done by the verifier are invalidated and must be
1994 * performed again, if the helper is used in combination with
1995 * direct packet access.
1996 *
1997 * All values for *flags* are reserved for future usage, and must
1998 * be left at zero.
1999 * Return
2000 * 0 on success, or a negative error in case of failure.
2001 *
96871b9f 2002 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
9cde0c88
QM
2003 * Description
2004 * Bind the socket associated to *ctx* to the address pointed by
2005 * *addr*, of length *addr_len*. This allows for making outgoing
2006 * connection from the desired IP address, which can be useful for
2007 * example when all processes inside a cgroup should use one
2008 * single IP address on a host that has multiple IP configured.
2009 *
2010 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2011 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
8086fbaf
SF
2012 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2013 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2014 * behavior and lets the kernel efficiently pick up an unused
2015 * port as long as 4-tuple is unique. Passing non-zero port might
2016 * lead to degraded performance.
9cde0c88
QM
2017 * Return
2018 * 0 on success, or a negative error in case of failure.
2019 *
2020 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2021 * Description
2022 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
fb53d3b6
AS
2023 * possible to both shrink and grow the packet tail.
2024 * Shrink done via *delta* being a negative integer.
9cde0c88 2025 *
c1fe1e70 2026 * A call to this helper is susceptible to change the underlying
9cde0c88
QM
2027 * packet buffer. Therefore, at load time, all checks on pointers
2028 * previously done by the verifier are invalidated and must be
2029 * performed again, if the helper is used in combination with
2030 * direct packet access.
2031 * Return
2032 * 0 on success, or a negative error in case of failure.
2033 *
2034 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2035 * Description
2036 * Retrieve the XFRM state (IP transform framework, see also
2037 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2038 *
2039 * The retrieved value is stored in the **struct bpf_xfrm_state**
2040 * pointed by *xfrm_state* and of length *size*.
2041 *
2042 * All values for *flags* are reserved for future usage, and must
2043 * be left at zero.
2044 *
2045 * This helper is available only if the kernel was compiled with
2046 * **CONFIG_XFRM** configuration option.
2047 * Return
2048 * 0 on success, or a negative error in case of failure.
de2ff05f 2049 *
5f0e5412 2050 * int bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
de2ff05f 2051 * Description
a56497d3
QM
2052 * Return a user or a kernel stack in bpf program provided buffer.
2053 * To achieve this, the helper needs *ctx*, which is a pointer
2054 * to the context on which the tracing program is executed.
2055 * To store the stacktrace, the bpf program provides *buf* with
2056 * a nonnegative *size*.
2057 *
2058 * The last argument, *flags*, holds the number of stack frames to
2059 * skip (from 0 to 255), masked with
2060 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2061 * the following flags:
2062 *
2063 * **BPF_F_USER_STACK**
2064 * Collect a user space stack instead of a kernel stack.
2065 * **BPF_F_USER_BUILD_ID**
2066 * Collect buildid+offset instead of ips for user stack,
2067 * only valid if **BPF_F_USER_STACK** is also specified.
2068 *
2069 * **bpf_get_stack**\ () can collect up to
2070 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2071 * to sufficient large buffer size. Note that
2072 * this limit can be controlled with the **sysctl** program, and
2073 * that it should be manually increased in order to profile long
2074 * user stacks (such as stacks for Java programs). To do so, use:
2075 *
2076 * ::
2077 *
2078 * # sysctl kernel.perf_event_max_stack=<new value>
de2ff05f 2079 * Return
7a279e93
QM
2080 * A non-negative value equal to or less than *size* on success,
2081 * or a negative error in case of failure.
32b3652c 2082 *
5f0e5412 2083 * int bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
32b3652c
DB
2084 * Description
2085 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2086 * it provides an easy way to load *len* bytes from *offset*
2087 * from the packet associated to *skb*, into the buffer pointed
2088 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2089 * a fifth argument *start_header* exists in order to select a
2090 * base offset to start from. *start_header* can be one of:
2091 *
2092 * **BPF_HDR_START_MAC**
2093 * Base offset to load data from is *skb*'s mac header.
2094 * **BPF_HDR_START_NET**
2095 * Base offset to load data from is *skb*'s network header.
2096 *
2097 * In general, "direct packet access" is the preferred method to
2098 * access packet data, however, this helper is in particular useful
2099 * in socket filters where *skb*\ **->data** does not always point
2100 * to the start of the mac header and where "direct packet access"
2101 * is not available.
32b3652c
DB
2102 * Return
2103 * 0 on success, or a negative error in case of failure.
2104 *
cb9c28ef
PB
2105 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2106 * Description
2107 * Do FIB lookup in kernel tables using parameters in *params*.
2108 * If lookup is successful and result shows packet is to be
2109 * forwarded, the neighbor tables are searched for the nexthop.
2110 * If successful (ie., FIB lookup shows forwarding and nexthop
6bdd533c
SY
2111 * is resolved), the nexthop address is returned in ipv4_dst
2112 * or ipv6_dst based on family, smac is set to mac address of
2113 * egress device, dmac is set to nexthop mac address, rt_metric
9b8ca379
QM
2114 * is set to metric from route (IPv4/IPv6 only), and ifindex
2115 * is set to the device index of the nexthop from the FIB lookup.
cb9c28ef 2116 *
0bd72117
DB
2117 * *plen* argument is the size of the passed in struct.
2118 * *flags* argument can be a combination of one or more of the
2119 * following values:
cb9c28ef 2120 *
7a279e93
QM
2121 * **BPF_FIB_LOOKUP_DIRECT**
2122 * Do a direct table lookup vs full lookup using FIB
2123 * rules.
2124 * **BPF_FIB_LOOKUP_OUTPUT**
2125 * Perform lookup from an egress perspective (default is
2126 * ingress).
cb9c28ef 2127 *
0bd72117
DB
2128 * *ctx* is either **struct xdp_md** for XDP programs or
2129 * **struct sk_buff** tc cls_act programs.
2130 * Return
9b8ca379
QM
2131 * * < 0 if any input argument is invalid
2132 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2133 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2134 * packet is not forwarded or needs assist from full stack
b8b394fa 2135 *
5f0e5412 2136 * int bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
b8b394fa
JF
2137 * Description
2138 * Add an entry to, or update a sockhash *map* referencing sockets.
2139 * The *skops* is used as a new value for the entry associated to
2140 * *key*. *flags* is one of:
2141 *
2142 * **BPF_NOEXIST**
2143 * The entry for *key* must not exist in the map.
2144 * **BPF_EXIST**
2145 * The entry for *key* must already exist in the map.
2146 * **BPF_ANY**
2147 * No condition on the existence of the entry for *key*.
2148 *
2149 * If the *map* has eBPF programs (parser and verdict), those will
2150 * be inherited by the socket being added. If the socket is
2151 * already attached to eBPF programs, this results in an error.
2152 * Return
2153 * 0 on success, or a negative error in case of failure.
2154 *
2155 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2156 * Description
2157 * This helper is used in programs implementing policies at the
2158 * socket level. If the message *msg* is allowed to pass (i.e. if
2159 * the verdict eBPF program returns **SK_PASS**), redirect it to
2160 * the socket referenced by *map* (of type
2161 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2162 * egress interfaces can be used for redirection. The
2163 * **BPF_F_INGRESS** value in *flags* is used to make the
2164 * distinction (ingress path is selected if the flag is present,
2165 * egress path otherwise). This is the only flag supported for now.
2166 * Return
2167 * **SK_PASS** on success, or **SK_DROP** on error.
2168 *
2169 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2170 * Description
2171 * This helper is used in programs implementing policies at the
2172 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2173 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2174 * to the socket referenced by *map* (of type
2175 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2176 * egress interfaces can be used for redirection. The
2177 * **BPF_F_INGRESS** value in *flags* is used to make the
2178 * distinction (ingress path is selected if the flag is present,
2179 * egress otherwise). This is the only flag supported for now.
2180 * Return
2181 * **SK_PASS** on success, or **SK_DROP** on error.
c99a84ea
MX
2182 *
2183 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2184 * Description
2185 * Encapsulate the packet associated to *skb* within a Layer 3
2186 * protocol header. This header is provided in the buffer at
2187 * address *hdr*, with *len* its size in bytes. *type* indicates
2188 * the protocol of the header and can be one of:
2189 *
2190 * **BPF_LWT_ENCAP_SEG6**
2191 * IPv6 encapsulation with Segment Routing Header
2192 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2193 * the IPv6 header is computed by the kernel.
2194 * **BPF_LWT_ENCAP_SEG6_INLINE**
2195 * Only works if *skb* contains an IPv6 packet. Insert a
2196 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2197 * the IPv6 header.
755db477
PO
2198 * **BPF_LWT_ENCAP_IP**
2199 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2200 * must be IPv4 or IPv6, followed by zero or more
c1fe1e70
QM
2201 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2202 * total bytes in all prepended headers. Please note that
2203 * if **skb_is_gso**\ (*skb*) is true, no more than two
2204 * headers can be prepended, and the inner header, if
2205 * present, should be either GRE or UDP/GUE.
2206 *
2207 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2208 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2209 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2210 * **BPF_PROG_TYPE_LWT_XMIT**.
2211 *
2212 * A call to this helper is susceptible to change the underlying
c99a84ea
MX
2213 * packet buffer. Therefore, at load time, all checks on pointers
2214 * previously done by the verifier are invalidated and must be
2215 * performed again, if the helper is used in combination with
2216 * direct packet access.
2217 * Return
2218 * 0 on success, or a negative error in case of failure.
2219 *
2220 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2221 * Description
2222 * Store *len* bytes from address *from* into the packet
2223 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2224 * inside the outermost IPv6 Segment Routing Header can be
2225 * modified through this helper.
2226 *
c1fe1e70 2227 * A call to this helper is susceptible to change the underlying
c99a84ea
MX
2228 * packet buffer. Therefore, at load time, all checks on pointers
2229 * previously done by the verifier are invalidated and must be
2230 * performed again, if the helper is used in combination with
2231 * direct packet access.
2232 * Return
2233 * 0 on success, or a negative error in case of failure.
2234 *
2235 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2236 * Description
2237 * Adjust the size allocated to TLVs in the outermost IPv6
2238 * Segment Routing Header contained in the packet associated to
2239 * *skb*, at position *offset* by *delta* bytes. Only offsets
2240 * after the segments are accepted. *delta* can be as well
2241 * positive (growing) as negative (shrinking).
2242 *
c1fe1e70 2243 * A call to this helper is susceptible to change the underlying
c99a84ea
MX
2244 * packet buffer. Therefore, at load time, all checks on pointers
2245 * previously done by the verifier are invalidated and must be
2246 * performed again, if the helper is used in combination with
2247 * direct packet access.
2248 * Return
2249 * 0 on success, or a negative error in case of failure.
2250 *
2251 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2252 * Description
2253 * Apply an IPv6 Segment Routing action of type *action* to the
2254 * packet associated to *skb*. Each action takes a parameter
2255 * contained at address *param*, and of length *param_len* bytes.
2256 * *action* can be one of:
2257 *
2258 * **SEG6_LOCAL_ACTION_END_X**
2259 * End.X action: Endpoint with Layer-3 cross-connect.
2260 * Type of *param*: **struct in6_addr**.
2261 * **SEG6_LOCAL_ACTION_END_T**
2262 * End.T action: Endpoint with specific IPv6 table lookup.
2263 * Type of *param*: **int**.
2264 * **SEG6_LOCAL_ACTION_END_B6**
2265 * End.B6 action: Endpoint bound to an SRv6 policy.
c1fe1e70 2266 * Type of *param*: **struct ipv6_sr_hdr**.
c99a84ea
MX
2267 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2268 * End.B6.Encap action: Endpoint bound to an SRv6
2269 * encapsulation policy.
c1fe1e70 2270 * Type of *param*: **struct ipv6_sr_hdr**.
c99a84ea 2271 *
c1fe1e70 2272 * A call to this helper is susceptible to change the underlying
c99a84ea
MX
2273 * packet buffer. Therefore, at load time, all checks on pointers
2274 * previously done by the verifier are invalidated and must be
2275 * performed again, if the helper is used in combination with
2276 * direct packet access.
2277 * Return
2278 * 0 on success, or a negative error in case of failure.
6bdd533c 2279 *
ea6eced0 2280 * int bpf_rc_repeat(void *ctx)
6bdd533c
SY
2281 * Description
2282 * This helper is used in programs implementing IR decoding, to
ea6eced0
QM
2283 * report a successfully decoded repeat key message. This delays
2284 * the generation of a key up event for previously generated
2285 * key down event.
6bdd533c 2286 *
ea6eced0
QM
2287 * Some IR protocols like NEC have a special IR message for
2288 * repeating last button, for when a button is held down.
6bdd533c
SY
2289 *
2290 * The *ctx* should point to the lirc sample as passed into
2291 * the program.
2292 *
6bdd533c
SY
2293 * This helper is only available is the kernel was compiled with
2294 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2295 * "**y**".
6bdd533c
SY
2296 * Return
2297 * 0
2298 *
ea6eced0 2299 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
6bdd533c
SY
2300 * Description
2301 * This helper is used in programs implementing IR decoding, to
ea6eced0
QM
2302 * report a successfully decoded key press with *scancode*,
2303 * *toggle* value in the given *protocol*. The scancode will be
2304 * translated to a keycode using the rc keymap, and reported as
2305 * an input key down event. After a period a key up event is
2306 * generated. This period can be extended by calling either
2307 * **bpf_rc_keydown**\ () again with the same values, or calling
2308 * **bpf_rc_repeat**\ ().
6bdd533c 2309 *
ff20460e 2310 * Some protocols include a toggle bit, in case the button was
ea6eced0 2311 * released and pressed again between consecutive scancodes.
6bdd533c
SY
2312 *
2313 * The *ctx* should point to the lirc sample as passed into
2314 * the program.
2315 *
ea6eced0
QM
2316 * The *protocol* is the decoded protocol number (see
2317 * **enum rc_proto** for some predefined values).
2318 *
6bdd533c
SY
2319 * This helper is only available is the kernel was compiled with
2320 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2321 * "**y**".
6bdd533c
SY
2322 * Return
2323 * 0
6b6a1925 2324 *
ea6eced0 2325 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
6b6a1925
DB
2326 * Description
2327 * Return the cgroup v2 id of the socket associated with the *skb*.
2328 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2329 * helper for cgroup v1 by providing a tag resp. identifier that
2330 * can be matched on or used for map lookups e.g. to implement
2331 * policy. The cgroup v2 id of a given path in the hierarchy is
2332 * exposed in user space through the f_handle API in order to get
2333 * to the same 64-bit id.
2334 *
2335 * This helper can be used on TC egress path, but not on ingress,
2336 * and is available only if the kernel was compiled with the
2337 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2338 * Return
2339 * The id is returned or 0 in case the id could not be retrieved.
c7ddbbaf
YS
2340 *
2341 * u64 bpf_get_current_cgroup_id(void)
2342 * Return
2343 * A 64-bit integer containing the current cgroup id based
2344 * on the cgroup within which the current task is running.
c419cf52 2345 *
ea6eced0 2346 * void *bpf_get_local_storage(void *map, u64 flags)
c419cf52
RG
2347 * Description
2348 * Get the pointer to the local storage area.
2349 * The type and the size of the local storage is defined
2350 * by the *map* argument.
2351 * The *flags* meaning is specific for each map type,
2352 * and has to be 0 for cgroup local storage.
2353 *
0bd72117
DB
2354 * Depending on the BPF program type, a local storage area
2355 * can be shared between multiple instances of the BPF program,
c419cf52
RG
2356 * running simultaneously.
2357 *
2358 * A user should care about the synchronization by himself.
0bd72117 2359 * For example, by using the **BPF_STX_XADD** instruction to alter
c419cf52
RG
2360 * the shared data.
2361 * Return
0bd72117 2362 * A pointer to the local storage area.
3bd43a8c
MKL
2363 *
2364 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2365 * Description
0bd72117
DB
2366 * Select a **SO_REUSEPORT** socket from a
2367 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2368 * It checks the selected socket is matching the incoming
2369 * request in the socket buffer.
3bd43a8c
MKL
2370 * Return
2371 * 0 on success, or a negative error in case of failure.
6acc9b43 2372 *
ea6eced0
QM
2373 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2374 * Description
2375 * Return id of cgroup v2 that is ancestor of cgroup associated
2376 * with the *skb* at the *ancestor_level*. The root cgroup is at
2377 * *ancestor_level* zero and each step down the hierarchy
2378 * increments the level. If *ancestor_level* == level of cgroup
2379 * associated with *skb*, then return value will be same as that
2380 * of **bpf_skb_cgroup_id**\ ().
2381 *
2382 * The helper is useful to implement policies based on cgroups
2383 * that are upper in hierarchy than immediate cgroup associated
2384 * with *skb*.
2385 *
2386 * The format of returned id and helper limitations are same as in
2387 * **bpf_skb_cgroup_id**\ ().
2388 * Return
2389 * The id is returned or 0 in case the id could not be retrieved.
2390 *
f71c6143 2391 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
6acc9b43
JS
2392 * Description
2393 * Look for TCP socket matching *tuple*, optionally in a child
2394 * network namespace *netns*. The return value must be checked,
0bd72117 2395 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2396 *
2397 * The *ctx* should point to the context of the program, such as
2398 * the skb or socket (depending on the hook in use). This is used
2399 * to determine the base network namespace for the lookup.
2400 *
2401 * *tuple_size* must be one of:
2402 *
2403 * **sizeof**\ (*tuple*\ **->ipv4**)
2404 * Look for an IPv4 socket.
2405 * **sizeof**\ (*tuple*\ **->ipv6**)
2406 * Look for an IPv6 socket.
2407 *
f71c6143
JS
2408 * If the *netns* is a negative signed 32-bit integer, then the
2409 * socket lookup table in the netns associated with the *ctx* will
2410 * will be used. For the TC hooks, this is the netns of the device
2411 * in the skb. For socket hooks, this is the netns of the socket.
2412 * If *netns* is any other signed 32-bit value greater than or
2413 * equal to zero then it specifies the ID of the netns relative to
2414 * the netns associated with the *ctx*. *netns* values beyond the
2415 * range of 32-bit integers are reserved for future use.
6acc9b43
JS
2416 *
2417 * All values for *flags* are reserved for future usage, and must
2418 * be left at zero.
2419 *
2420 * This helper is available only if the kernel was compiled with
2421 * **CONFIG_NET** configuration option.
2422 * Return
0bd72117
DB
2423 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2424 * For sockets with reuseport option, the **struct bpf_sock**
c1fe1e70
QM
2425 * result is from *reuse*\ **->socks**\ [] using the hash of the
2426 * tuple.
6acc9b43 2427 *
f71c6143 2428 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
6acc9b43
JS
2429 * Description
2430 * Look for UDP socket matching *tuple*, optionally in a child
2431 * network namespace *netns*. The return value must be checked,
0bd72117 2432 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2433 *
2434 * The *ctx* should point to the context of the program, such as
2435 * the skb or socket (depending on the hook in use). This is used
2436 * to determine the base network namespace for the lookup.
2437 *
2438 * *tuple_size* must be one of:
2439 *
2440 * **sizeof**\ (*tuple*\ **->ipv4**)
2441 * Look for an IPv4 socket.
2442 * **sizeof**\ (*tuple*\ **->ipv6**)
2443 * Look for an IPv6 socket.
2444 *
f71c6143
JS
2445 * If the *netns* is a negative signed 32-bit integer, then the
2446 * socket lookup table in the netns associated with the *ctx* will
2447 * will be used. For the TC hooks, this is the netns of the device
2448 * in the skb. For socket hooks, this is the netns of the socket.
2449 * If *netns* is any other signed 32-bit value greater than or
2450 * equal to zero then it specifies the ID of the netns relative to
2451 * the netns associated with the *ctx*. *netns* values beyond the
2452 * range of 32-bit integers are reserved for future use.
6acc9b43
JS
2453 *
2454 * All values for *flags* are reserved for future usage, and must
2455 * be left at zero.
2456 *
2457 * This helper is available only if the kernel was compiled with
2458 * **CONFIG_NET** configuration option.
2459 * Return
0bd72117
DB
2460 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2461 * For sockets with reuseport option, the **struct bpf_sock**
c1fe1e70
QM
2462 * result is from *reuse*\ **->socks**\ [] using the hash of the
2463 * tuple.
6acc9b43 2464 *
0bd72117 2465 * int bpf_sk_release(struct bpf_sock *sock)
6acc9b43 2466 * Description
0bd72117
DB
2467 * Release the reference held by *sock*. *sock* must be a
2468 * non-**NULL** pointer that was returned from
2469 * **bpf_sk_lookup_xxx**\ ().
6acc9b43
JS
2470 * Return
2471 * 0 on success, or a negative error in case of failure.
f908d26b 2472 *
ea6eced0
QM
2473 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2474 * Description
2475 * Push an element *value* in *map*. *flags* is one of:
2476 *
2477 * **BPF_EXIST**
2478 * If the queue/stack is full, the oldest element is
2479 * removed to make room for this.
2480 * Return
2481 * 0 on success, or a negative error in case of failure.
2482 *
0bd72117
DB
2483 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2484 * Description
2485 * Pop an element from *map*.
2486 * Return
2487 * 0 on success, or a negative error in case of failure.
2488 *
2489 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2490 * Description
2491 * Get an element from *map* without removing it.
2492 * Return
2493 * 0 on success, or a negative error in case of failure.
2494 *
5f0e5412 2495 * int bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
f908d26b 2496 * Description
0bd72117 2497 * For socket policies, insert *len* bytes into *msg* at offset
f908d26b
JF
2498 * *start*.
2499 *
2500 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
0bd72117 2501 * *msg* it may want to insert metadata or options into the *msg*.
f908d26b
JF
2502 * This can later be read and used by any of the lower layer BPF
2503 * hooks.
2504 *
2505 * This helper may fail if under memory pressure (a malloc
2506 * fails) in these cases BPF programs will get an appropriate
2507 * error and BPF programs will need to handle them.
f908d26b
JF
2508 * Return
2509 * 0 on success, or a negative error in case of failure.
d913a227 2510 *
5f0e5412 2511 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
0bd72117 2512 * Description
5f0e5412 2513 * Will remove *len* bytes from a *msg* starting at byte *start*.
d913a227
JF
2514 * This may result in **ENOMEM** errors under certain situations if
2515 * an allocation and copy are required due to a full ring buffer.
2516 * However, the helper will try to avoid doing the allocation
2517 * if possible. Other errors can occur if input parameters are
0bd72117 2518 * invalid either due to *start* byte not being valid part of *msg*
d913a227 2519 * payload and/or *pop* value being to large.
01d3240a
SY
2520 * Return
2521 * 0 on success, or a negative error in case of failure.
2522 *
2523 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2524 * Description
2525 * This helper is used in programs implementing IR decoding, to
2526 * report a successfully decoded pointer movement.
2527 *
2528 * The *ctx* should point to the lirc sample as passed into
2529 * the program.
d913a227 2530 *
01d3240a
SY
2531 * This helper is only available is the kernel was compiled with
2532 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2533 * "**y**".
d913a227 2534 * Return
01d3240a 2535 * 0
281f9e75 2536 *
ea6eced0
QM
2537 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2538 * Description
2539 * Acquire a spinlock represented by the pointer *lock*, which is
2540 * stored as part of a value of a map. Taking the lock allows to
2541 * safely update the rest of the fields in that value. The
2542 * spinlock can (and must) later be released with a call to
2543 * **bpf_spin_unlock**\ (\ *lock*\ ).
2544 *
2545 * Spinlocks in BPF programs come with a number of restrictions
2546 * and constraints:
2547 *
2548 * * **bpf_spin_lock** objects are only allowed inside maps of
2549 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2550 * list could be extended in the future).
2551 * * BTF description of the map is mandatory.
2552 * * The BPF program can take ONE lock at a time, since taking two
2553 * or more could cause dead locks.
2554 * * Only one **struct bpf_spin_lock** is allowed per map element.
2555 * * When the lock is taken, calls (either BPF to BPF or helpers)
2556 * are not allowed.
2557 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2558 * allowed inside a spinlock-ed region.
2559 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2560 * the lock, on all execution paths, before it returns.
2561 * * The BPF program can access **struct bpf_spin_lock** only via
2562 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2563 * helpers. Loading or storing data into the **struct
2564 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2565 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2566 * of the map value must be a struct and have **struct
2567 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2568 * Nested lock inside another struct is not allowed.
2569 * * The **struct bpf_spin_lock** *lock* field in a map value must
2570 * be aligned on a multiple of 4 bytes in that value.
2571 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2572 * the **bpf_spin_lock** field to user space.
2573 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2574 * a BPF program, do not update the **bpf_spin_lock** field.
2575 * * **bpf_spin_lock** cannot be on the stack or inside a
2576 * networking packet (it can only be inside of a map values).
2577 * * **bpf_spin_lock** is available to root only.
2578 * * Tracing programs and socket filter programs cannot use
2579 * **bpf_spin_lock**\ () due to insufficient preemption checks
2580 * (but this may change in the future).
2581 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2582 * Return
2583 * 0
2584 *
2585 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2586 * Description
2587 * Release the *lock* previously locked by a call to
2588 * **bpf_spin_lock**\ (\ *lock*\ ).
2589 * Return
2590 * 0
2591 *
281f9e75
MKL
2592 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2593 * Description
2594 * This helper gets a **struct bpf_sock** pointer such
ea6eced0 2595 * that all the fields in this **bpf_sock** can be accessed.
281f9e75 2596 * Return
ea6eced0 2597 * A **struct bpf_sock** pointer on success, or **NULL** in
281f9e75
MKL
2598 * case of failure.
2599 *
2600 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2601 * Description
2602 * This helper gets a **struct bpf_tcp_sock** pointer from a
2603 * **struct bpf_sock** pointer.
281f9e75 2604 * Return
ea6eced0 2605 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
281f9e75 2606 * case of failure.
5cce85c6 2607 *
5f0e5412 2608 * int bpf_skb_ecn_set_ce(struct sk_buff *skb)
ea6eced0
QM
2609 * Description
2610 * Set ECN (Explicit Congestion Notification) field of IP header
2611 * to **CE** (Congestion Encountered) if current value is **ECT**
2612 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2613 * and IPv4.
2614 * Return
2615 * 1 if the **CE** flag is set (either by the current helper call
2616 * or because it was already present), 0 if it is not set.
ef776a27
MKL
2617 *
2618 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2619 * Description
ea6eced0
QM
2620 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2621 * **bpf_sk_release**\ () is unnecessary and not allowed.
ef776a27 2622 * Return
ea6eced0 2623 * A **struct bpf_sock** pointer on success, or **NULL** in
ef776a27 2624 * case of failure.
253c8dde
LB
2625 *
2626 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2627 * Description
2628 * Look for TCP socket matching *tuple*, optionally in a child
2629 * network namespace *netns*. The return value must be checked,
2630 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2631 *
c1fe1e70
QM
2632 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2633 * that it also returns timewait or request sockets. Use
2634 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2635 * full structure.
253c8dde
LB
2636 *
2637 * This helper is available only if the kernel was compiled with
2638 * **CONFIG_NET** configuration option.
2639 * Return
2640 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2641 * For sockets with reuseport option, the **struct bpf_sock**
c1fe1e70
QM
2642 * result is from *reuse*\ **->socks**\ [] using the hash of the
2643 * tuple.
253c8dde
LB
2644 *
2645 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2646 * Description
c1fe1e70
QM
2647 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2648 * the listening socket in *sk*.
253c8dde 2649 *
c1fe1e70
QM
2650 * *iph* points to the start of the IPv4 or IPv6 header, while
2651 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2652 * **sizeof**\ (**struct ip6hdr**).
253c8dde 2653 *
c1fe1e70
QM
2654 * *th* points to the start of the TCP header, while *th_len*
2655 * contains **sizeof**\ (**struct tcphdr**).
253c8dde 2656 * Return
c1fe1e70
QM
2657 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2658 * error otherwise.
196398d4
AI
2659 *
2660 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2661 * Description
2662 * Get name of sysctl in /proc/sys/ and copy it into provided by
2663 * program buffer *buf* of size *buf_len*.
2664 *
2665 * The buffer is always NUL terminated, unless it's zero-sized.
2666 *
2667 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2668 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2669 * only (e.g. "tcp_mem").
2670 * Return
2671 * Number of character copied (not including the trailing NUL).
2672 *
2673 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2674 * truncated name in this case).
2675 *
2676 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2677 * Description
2678 * Get current value of sysctl as it is presented in /proc/sys
2679 * (incl. newline, etc), and copy it as a string into provided
2680 * by program buffer *buf* of size *buf_len*.
2681 *
2682 * The whole value is copied, no matter what file position user
2683 * space issued e.g. sys_read at.
2684 *
2685 * The buffer is always NUL terminated, unless it's zero-sized.
2686 * Return
2687 * Number of character copied (not including the trailing NUL).
2688 *
2689 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2690 * truncated name in this case).
2691 *
2692 * **-EINVAL** if current value was unavailable, e.g. because
2693 * sysctl is uninitialized and read returns -EIO for it.
2694 *
2695 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2696 * Description
2697 * Get new value being written by user space to sysctl (before
2698 * the actual write happens) and copy it as a string into
2699 * provided by program buffer *buf* of size *buf_len*.
2700 *
2701 * User space may write new value at file position > 0.
2702 *
2703 * The buffer is always NUL terminated, unless it's zero-sized.
2704 * Return
2705 * Number of character copied (not including the trailing NUL).
2706 *
2707 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2708 * truncated name in this case).
2709 *
2710 * **-EINVAL** if sysctl is being read.
2711 *
2712 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2713 * Description
2714 * Override new value being written by user space to sysctl with
2715 * value provided by program in buffer *buf* of size *buf_len*.
2716 *
2717 * *buf* should contain a string in same form as provided by user
2718 * space on sysctl write.
2719 *
2720 * User space may write new value at file position > 0. To override
2721 * the whole sysctl value file position should be set to zero.
2722 * Return
2723 * 0 on success.
2724 *
2725 * **-E2BIG** if the *buf_len* is too big.
2726 *
2727 * **-EINVAL** if sysctl is being read.
b457e553
AI
2728 *
2729 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2730 * Description
2731 * Convert the initial part of the string from buffer *buf* of
2732 * size *buf_len* to a long integer according to the given base
2733 * and save the result in *res*.
2734 *
2735 * The string may begin with an arbitrary amount of white space
c1fe1e70
QM
2736 * (as determined by **isspace**\ (3)) followed by a single
2737 * optional '**-**' sign.
b457e553
AI
2738 *
2739 * Five least significant bits of *flags* encode base, other bits
2740 * are currently unused.
2741 *
2742 * Base must be either 8, 10, 16 or 0 to detect it automatically
c1fe1e70 2743 * similar to user space **strtol**\ (3).
b457e553
AI
2744 * Return
2745 * Number of characters consumed on success. Must be positive but
c1fe1e70 2746 * no more than *buf_len*.
b457e553
AI
2747 *
2748 * **-EINVAL** if no valid digits were found or unsupported base
2749 * was provided.
2750 *
2751 * **-ERANGE** if resulting value was out of range.
2752 *
2753 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2754 * Description
2755 * Convert the initial part of the string from buffer *buf* of
2756 * size *buf_len* to an unsigned long integer according to the
2757 * given base and save the result in *res*.
2758 *
2759 * The string may begin with an arbitrary amount of white space
c1fe1e70 2760 * (as determined by **isspace**\ (3)).
b457e553
AI
2761 *
2762 * Five least significant bits of *flags* encode base, other bits
2763 * are currently unused.
2764 *
2765 * Base must be either 8, 10, 16 or 0 to detect it automatically
c1fe1e70 2766 * similar to user space **strtoul**\ (3).
b457e553
AI
2767 * Return
2768 * Number of characters consumed on success. Must be positive but
c1fe1e70 2769 * no more than *buf_len*.
b457e553
AI
2770 *
2771 * **-EINVAL** if no valid digits were found or unsupported base
2772 * was provided.
2773 *
2774 * **-ERANGE** if resulting value was out of range.
948d930e
MKL
2775 *
2776 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2777 * Description
c1fe1e70 2778 * Get a bpf-local-storage from a *sk*.
948d930e
MKL
2779 *
2780 * Logically, it could be thought of getting the value from
2781 * a *map* with *sk* as the **key**. From this
2782 * perspective, the usage is not much different from
c1fe1e70
QM
2783 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2784 * helper enforces the key must be a full socket and the map must
2785 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
948d930e
MKL
2786 *
2787 * Underneath, the value is stored locally at *sk* instead of
c1fe1e70
QM
2788 * the *map*. The *map* is used as the bpf-local-storage
2789 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2790 * searched against all bpf-local-storages residing at *sk*.
948d930e 2791 *
c1fe1e70 2792 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
948d930e
MKL
2793 * used such that a new bpf-local-storage will be
2794 * created if one does not exist. *value* can be used
c1fe1e70 2795 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
948d930e 2796 * the initial value of a bpf-local-storage. If *value* is
c1fe1e70 2797 * **NULL**, the new bpf-local-storage will be zero initialized.
948d930e
MKL
2798 * Return
2799 * A bpf-local-storage pointer is returned on success.
2800 *
2801 * **NULL** if not found or there was an error in adding
2802 * a new bpf-local-storage.
2803 *
2804 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2805 * Description
c1fe1e70 2806 * Delete a bpf-local-storage from a *sk*.
948d930e
MKL
2807 * Return
2808 * 0 on success.
2809 *
2810 * **-ENOENT** if the bpf-local-storage cannot be found.
edaccf89
YS
2811 *
2812 * int bpf_send_signal(u32 sig)
2813 * Description
8482941f
YS
2814 * Send signal *sig* to the process of the current task.
2815 * The signal may be delivered to any of this process's threads.
edaccf89
YS
2816 * Return
2817 * 0 on success or successfully queued.
2818 *
2819 * **-EBUSY** if work queue under nmi is full.
2820 *
2821 * **-EINVAL** if *sig* is invalid.
2822 *
2823 * **-EPERM** if no permission to send the *sig*.
2824 *
2825 * **-EAGAIN** if bpf program can try again.
3745ee18
PP
2826 *
2827 * s64 bpf_tcp_gen_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2828 * Description
2829 * Try to issue a SYN cookie for the packet with corresponding
2830 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
2831 *
2832 * *iph* points to the start of the IPv4 or IPv6 header, while
2833 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2834 * **sizeof**\ (**struct ip6hdr**).
2835 *
2836 * *th* points to the start of the TCP header, while *th_len*
2837 * contains the length of the TCP header.
3745ee18
PP
2838 * Return
2839 * On success, lower 32 bits hold the generated SYN cookie in
2840 * followed by 16 bits which hold the MSS value for that cookie,
2841 * and the top 16 bits are unused.
2842 *
2843 * On failure, the returned value is one of the following:
2844 *
2845 * **-EINVAL** SYN cookie cannot be issued due to error
2846 *
2847 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
2848 *
2849 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
2850 *
2851 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
a7658e1a
AS
2852 *
2853 * int bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2854 * Description
2855 * Write raw *data* blob into a special BPF perf event held by
2856 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2857 * event must have the following attributes: **PERF_SAMPLE_RAW**
2858 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2859 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2860 *
2861 * The *flags* are used to indicate the index in *map* for which
2862 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2863 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2864 * to indicate that the index of the current CPU core should be
2865 * used.
2866 *
2867 * The value to write, of *size*, is passed through eBPF stack and
2868 * pointed by *data*.
2869 *
2870 * *ctx* is a pointer to in-kernel struct sk_buff.
2871 *
2872 * This helper is similar to **bpf_perf_event_output**\ () but
2873 * restricted to raw_tracepoint bpf programs.
2874 * Return
2875 * 0 on success, or a negative error in case of failure.
6ae08ae3
DB
2876 *
2877 * int bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
2878 * Description
2879 * Safely attempt to read *size* bytes from user space address
2880 * *unsafe_ptr* and store the data in *dst*.
2881 * Return
2882 * 0 on success, or a negative error in case of failure.
2883 *
2884 * int bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
2885 * Description
2886 * Safely attempt to read *size* bytes from kernel space address
2887 * *unsafe_ptr* and store the data in *dst*.
2888 * Return
2889 * 0 on success, or a negative error in case of failure.
2890 *
2891 * int bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
2892 * Description
2893 * Copy a NUL terminated string from an unsafe user address
2894 * *unsafe_ptr* to *dst*. The *size* should include the
2895 * terminating NUL byte. In case the string length is smaller than
2896 * *size*, the target is not padded with further NUL bytes. If the
2897 * string length is larger than *size*, just *size*-1 bytes are
2898 * copied and the last byte is set to NUL.
2899 *
2900 * On success, the length of the copied string is returned. This
2901 * makes this helper useful in tracing programs for reading
2902 * strings, and more importantly to get its length at runtime. See
2903 * the following snippet:
2904 *
2905 * ::
2906 *
2907 * SEC("kprobe/sys_open")
2908 * void bpf_sys_open(struct pt_regs *ctx)
2909 * {
2910 * char buf[PATHLEN]; // PATHLEN is defined to 256
2911 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
2912 * ctx->di);
2913 *
2914 * // Consume buf, for example push it to
2915 * // userspace via bpf_perf_event_output(); we
2916 * // can use res (the string length) as event
2917 * // size, after checking its boundaries.
2918 * }
2919 *
ff20460e 2920 * In comparison, using **bpf_probe_read_user**\ () helper here
6ae08ae3
DB
2921 * instead to read the string would require to estimate the length
2922 * at compile time, and would often result in copying more memory
2923 * than necessary.
2924 *
2925 * Another useful use case is when parsing individual process
2926 * arguments or individual environment variables navigating
2927 * *current*\ **->mm->arg_start** and *current*\
2928 * **->mm->env_start**: using this helper and the return value,
2929 * one can quickly iterate at the right offset of the memory area.
2930 * Return
2931 * On success, the strictly positive length of the string,
2932 * including the trailing NUL character. On error, a negative
2933 * value.
2934 *
2935 * int bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
2936 * Description
2937 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
ff20460e 2938 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
6ae08ae3 2939 * Return
ff20460e 2940 * On success, the strictly positive length of the string, including
6ae08ae3 2941 * the trailing NUL character. On error, a negative value.
17328d61
MKL
2942 *
2943 * int bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
2944 * Description
ff20460e 2945 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
17328d61
MKL
2946 * *rcv_nxt* is the ack_seq to be sent out.
2947 * Return
2948 * 0 on success, or a negative error in case of failure.
2949 *
8482941f
YS
2950 * int bpf_send_signal_thread(u32 sig)
2951 * Description
2952 * Send signal *sig* to the thread corresponding to the current task.
2953 * Return
2954 * 0 on success or successfully queued.
2955 *
2956 * **-EBUSY** if work queue under nmi is full.
2957 *
2958 * **-EINVAL** if *sig* is invalid.
2959 *
2960 * **-EPERM** if no permission to send the *sig*.
2961 *
2962 * **-EAGAIN** if bpf program can try again.
0a49c1a8
MKL
2963 *
2964 * u64 bpf_jiffies64(void)
2965 * Description
2966 * Obtain the 64bit jiffies
2967 * Return
2968 * The 64 bit jiffies
67306f84
DX
2969 *
2970 * int bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
2971 * Description
2972 * For an eBPF program attached to a perf event, retrieve the
ff20460e
QM
2973 * branch records (**struct perf_branch_entry**) associated to *ctx*
2974 * and store it in the buffer pointed by *buf* up to size
67306f84
DX
2975 * *size* bytes.
2976 * Return
2977 * On success, number of bytes written to *buf*. On error, a
2978 * negative value.
2979 *
2980 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
ff20460e 2981 * instead return the number of bytes required to store all the
67306f84
DX
2982 * branch entries. If this flag is set, *buf* may be NULL.
2983 *
2984 * **-EINVAL** if arguments invalid or **size** not a multiple
ff20460e 2985 * of **sizeof**\ (**struct perf_branch_entry**\ ).
67306f84
DX
2986 *
2987 * **-ENOENT** if architecture does not support branch records.
b4490c5c
CN
2988 *
2989 * int bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
2990 * Description
2991 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
2992 * *namespace* will be returned in *nsdata*.
ff20460e
QM
2993 * Return
2994 * 0 on success, or one of the following in case of failure:
b4490c5c
CN
2995 *
2996 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
2997 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
2998 *
2999 * **-ENOENT** if pidns does not exists for the current task.
3000 *
d831ee84
EC
3001 * int bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3002 * Description
3003 * Write raw *data* blob into a special BPF perf event held by
3004 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3005 * event must have the following attributes: **PERF_SAMPLE_RAW**
3006 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3007 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3008 *
3009 * The *flags* are used to indicate the index in *map* for which
3010 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3011 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3012 * to indicate that the index of the current CPU core should be
3013 * used.
3014 *
3015 * The value to write, of *size*, is passed through eBPF stack and
3016 * pointed by *data*.
3017 *
3018 * *ctx* is a pointer to in-kernel struct xdp_buff.
3019 *
3020 * This helper is similar to **bpf_perf_eventoutput**\ () but
3021 * restricted to raw_tracepoint bpf programs.
3022 * Return
3023 * 0 on success, or a negative error in case of failure.
f318903c
DB
3024 *
3025 * u64 bpf_get_netns_cookie(void *ctx)
3026 * Description
3027 * Retrieve the cookie (generated by the kernel) of the network
3028 * namespace the input *ctx* is associated with. The network
3029 * namespace cookie remains stable for its lifetime and provides
3030 * a global identifier that can be assumed unique. If *ctx* is
3031 * NULL, then the helper returns the cookie for the initial
3032 * network namespace. The cookie itself is very similar to that
ff20460e
QM
3033 * of **bpf_get_socket_cookie**\ () helper, but for network
3034 * namespaces instead of sockets.
f318903c
DB
3035 * Return
3036 * A 8-byte long opaque number.
0f09abd1
DB
3037 *
3038 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3039 * Description
3040 * Return id of cgroup v2 that is ancestor of the cgroup associated
3041 * with the current task at the *ancestor_level*. The root cgroup
3042 * is at *ancestor_level* zero and each step down the hierarchy
3043 * increments the level. If *ancestor_level* == level of cgroup
3044 * associated with the current task, then return value will be the
3045 * same as that of **bpf_get_current_cgroup_id**\ ().
3046 *
3047 * The helper is useful to implement policies based on cgroups
3048 * that are upper in hierarchy than immediate cgroup associated
3049 * with the current task.
3050 *
3051 * The format of returned id and helper limitations are same as in
3052 * **bpf_get_current_cgroup_id**\ ().
3053 * Return
3054 * The id is returned or 0 in case the id could not be retrieved.
cf7fbe66
JS
3055 *
3056 * int bpf_sk_assign(struct sk_buff *skb, struct bpf_sock *sk, u64 flags)
3057 * Description
3058 * Assign the *sk* to the *skb*. When combined with appropriate
3059 * routing configuration to receive the packet towards the socket,
3060 * will cause *skb* to be delivered to the specified socket.
3061 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3062 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3063 * interfere with successful delivery to the socket.
3064 *
3065 * This operation is only valid from TC ingress path.
3066 *
3067 * The *flags* argument must be zero.
3068 * Return
ff20460e 3069 * 0 on success, or a negative error in case of failure:
cf7fbe66 3070 *
ff20460e
QM
3071 * **-EINVAL** if specified *flags* are not supported.
3072 *
3073 * **-ENOENT** if the socket is unavailable for assignment.
3074 *
3075 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3076 *
3077 * **-EOPNOTSUPP** if the operation is not supported, for example
3078 * a call from outside of TC ingress.
3079 *
3080 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3081 * (reuseport).
71d19214
3082 *
3083 * u64 bpf_ktime_get_boot_ns(void)
3084 * Description
3085 * Return the time elapsed since system boot, in nanoseconds.
3086 * Does include the time the system was suspended.
ff20460e 3087 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
71d19214
3088 * Return
3089 * Current *ktime*.
492e639f
YS
3090 *
3091 * int bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3092 * Description
ff20460e
QM
3093 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3094 * out the format string.
492e639f
YS
3095 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3096 * the format string itself. The *data* and *data_len* are format string
ff20460e 3097 * arguments. The *data* are a **u64** array and corresponding format string
492e639f
YS
3098 * values are stored in the array. For strings and pointers where pointees
3099 * are accessed, only the pointer values are stored in the *data* array.
ff20460e 3100 * The *data_len* is the size of *data* in bytes.
492e639f
YS
3101 *
3102 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3103 * Reading kernel memory may fail due to either invalid address or
3104 * valid address but requiring a major memory fault. If reading kernel memory
3105 * fails, the string for **%s** will be an empty string, and the ip
3106 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
ff20460e 3107 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
492e639f 3108 * Return
ff20460e
QM
3109 * 0 on success, or a negative error in case of failure:
3110 *
3111 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
3112 * by returning 1 from bpf program.
3113 *
3114 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
3115 *
3116 * **-E2BIG** if *fmt* contains too many format specifiers.
492e639f 3117 *
ff20460e 3118 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
492e639f
YS
3119 *
3120 * int bpf_seq_write(struct seq_file *m, const void *data, u32 len)
3121 * Description
ff20460e 3122 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
492e639f 3123 * The *m* represents the seq_file. The *data* and *len* represent the
ff20460e 3124 * data to write in bytes.
492e639f 3125 * Return
ff20460e 3126 * 0 on success, or a negative error in case of failure:
492e639f 3127 *
ff20460e 3128 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
f307fa2c
AI
3129 *
3130 * u64 bpf_sk_cgroup_id(struct bpf_sock *sk)
3131 * Description
3132 * Return the cgroup v2 id of the socket *sk*.
3133 *
3134 * *sk* must be a non-**NULL** pointer to a full socket, e.g. one
3135 * returned from **bpf_sk_lookup_xxx**\ (),
3136 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
3137 * same as in **bpf_skb_cgroup_id**\ ().
3138 *
3139 * This helper is available only if the kernel was compiled with
3140 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
3141 * Return
3142 * The id is returned or 0 in case the id could not be retrieved.
3143 *
3144 * u64 bpf_sk_ancestor_cgroup_id(struct bpf_sock *sk, int ancestor_level)
3145 * Description
3146 * Return id of cgroup v2 that is ancestor of cgroup associated
3147 * with the *sk* at the *ancestor_level*. The root cgroup is at
3148 * *ancestor_level* zero and each step down the hierarchy
3149 * increments the level. If *ancestor_level* == level of cgroup
3150 * associated with *sk*, then return value will be same as that
3151 * of **bpf_sk_cgroup_id**\ ().
3152 *
3153 * The helper is useful to implement policies based on cgroups
3154 * that are upper in hierarchy than immediate cgroup associated
3155 * with *sk*.
3156 *
3157 * The format of returned id and helper limitations are same as in
3158 * **bpf_sk_cgroup_id**\ ().
3159 * Return
3160 * The id is returned or 0 in case the id could not be retrieved.
457f4436
AN
3161 *
3162 * void *bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
3163 * Description
3164 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
3165 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3166 * new data availability is sent.
3167 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3168 * new data availability is sent unconditionally.
3169 * Return
3170 * 0, on success;
3171 * < 0, on error.
3172 *
3173 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
3174 * Description
3175 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
3176 * Return
3177 * Valid pointer with *size* bytes of memory available; NULL,
3178 * otherwise.
3179 *
3180 * void bpf_ringbuf_submit(void *data, u64 flags)
3181 * Description
3182 * Submit reserved ring buffer sample, pointed to by *data*.
3183 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3184 * new data availability is sent.
3185 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3186 * new data availability is sent unconditionally.
3187 * Return
3188 * Nothing. Always succeeds.
3189 *
3190 * void bpf_ringbuf_discard(void *data, u64 flags)
3191 * Description
3192 * Discard reserved ring buffer sample, pointed to by *data*.
3193 * If BPF_RB_NO_WAKEUP is specified in *flags*, no notification of
3194 * new data availability is sent.
3195 * IF BPF_RB_FORCE_WAKEUP is specified in *flags*, notification of
3196 * new data availability is sent unconditionally.
3197 * Return
3198 * Nothing. Always succeeds.
3199 *
3200 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
3201 * Description
3202 * Query various characteristics of provided ring buffer. What
3203 * exactly is queries is determined by *flags*:
3204 * - BPF_RB_AVAIL_DATA - amount of data not yet consumed;
3205 * - BPF_RB_RING_SIZE - the size of ring buffer;
3206 * - BPF_RB_CONS_POS - consumer position (can wrap around);
3207 * - BPF_RB_PROD_POS - producer(s) position (can wrap around);
3208 * Data returned is just a momentary snapshots of actual values
3209 * and could be inaccurate, so this facility should be used to
3210 * power heuristics and for reporting, not to make 100% correct
3211 * calculation.
3212 * Return
3213 * Requested value, or 0, if flags are not recognized.
0cb34dc2
JS
3214 */
3215#define __BPF_FUNC_MAPPER(FN) \
3216 FN(unspec), \
3217 FN(map_lookup_elem), \
3218 FN(map_update_elem), \
3219 FN(map_delete_elem), \
3220 FN(probe_read), \
3221 FN(ktime_get_ns), \
3222 FN(trace_printk), \
3223 FN(get_prandom_u32), \
3224 FN(get_smp_processor_id), \
3225 FN(skb_store_bytes), \
3226 FN(l3_csum_replace), \
3227 FN(l4_csum_replace), \
3228 FN(tail_call), \
3229 FN(clone_redirect), \
3230 FN(get_current_pid_tgid), \
3231 FN(get_current_uid_gid), \
3232 FN(get_current_comm), \
3233 FN(get_cgroup_classid), \
3234 FN(skb_vlan_push), \
3235 FN(skb_vlan_pop), \
3236 FN(skb_get_tunnel_key), \
3237 FN(skb_set_tunnel_key), \
3238 FN(perf_event_read), \
3239 FN(redirect), \
3240 FN(get_route_realm), \
3241 FN(perf_event_output), \
3242 FN(skb_load_bytes), \
3243 FN(get_stackid), \
3244 FN(csum_diff), \
3245 FN(skb_get_tunnel_opt), \
3246 FN(skb_set_tunnel_opt), \
3247 FN(skb_change_proto), \
3248 FN(skb_change_type), \
3249 FN(skb_under_cgroup), \
3250 FN(get_hash_recalc), \
3251 FN(get_current_task), \
3252 FN(probe_write_user), \
3253 FN(current_task_under_cgroup), \
3254 FN(skb_change_tail), \
3255 FN(skb_pull_data), \
3256 FN(csum_update), \
3257 FN(set_hash_invalid), \
3258 FN(get_numa_node_id), \
3259 FN(skb_change_head), \
9a738266 3260 FN(xdp_adjust_head), \
91b8270f 3261 FN(probe_read_str), \
6acc5c29 3262 FN(get_socket_cookie), \
ded092cd 3263 FN(get_socket_uid), \
04df41e3 3264 FN(set_hash), \
2be7e212 3265 FN(setsockopt), \
996139e8 3266 FN(skb_adjust_room), \
69e8cc13
JF
3267 FN(redirect_map), \
3268 FN(sk_redirect_map), \
ac29991b 3269 FN(sock_map_update), \
020a32d9 3270 FN(xdp_adjust_meta), \
81b9cf80 3271 FN(perf_event_read_value), \
e27afb84 3272 FN(perf_prog_read_value), \
965de87e 3273 FN(getsockopt), \
d6d4f60c 3274 FN(override_return), \
4c4c3c27
JF
3275 FN(sock_ops_cb_flags_set), \
3276 FN(msg_redirect_map), \
468b3fde 3277 FN(msg_apply_bytes), \
0dcbbf67 3278 FN(msg_cork_bytes), \
622adafb 3279 FN(msg_pull_data), \
0367d0a2 3280 FN(bind), \
29a36f9e 3281 FN(xdp_adjust_tail), \
de2ff05f 3282 FN(skb_get_xfrm_state), \
32b3652c 3283 FN(get_stack), \
cb9c28ef 3284 FN(skb_load_bytes_relative), \
b8b394fa
JF
3285 FN(fib_lookup), \
3286 FN(sock_hash_update), \
3287 FN(msg_redirect_hash), \
c99a84ea
MX
3288 FN(sk_redirect_hash), \
3289 FN(lwt_push_encap), \
3290 FN(lwt_seg6_store_bytes), \
3291 FN(lwt_seg6_adjust_srh), \
6bdd533c
SY
3292 FN(lwt_seg6_action), \
3293 FN(rc_repeat), \
6b6a1925 3294 FN(rc_keydown), \
c7ddbbaf 3295 FN(skb_cgroup_id), \
c419cf52 3296 FN(get_current_cgroup_id), \
3bd43a8c 3297 FN(get_local_storage), \
539764d0 3298 FN(sk_select_reuseport), \
6acc9b43
JS
3299 FN(skb_ancestor_cgroup_id), \
3300 FN(sk_lookup_tcp), \
3301 FN(sk_lookup_udp), \
da4e1b15
MV
3302 FN(sk_release), \
3303 FN(map_push_elem), \
3304 FN(map_pop_elem), \
f908d26b 3305 FN(map_peek_elem), \
d913a227 3306 FN(msg_push_data), \
01d3240a 3307 FN(msg_pop_data), \
7dac3ae4
AS
3308 FN(rc_pointer_rel), \
3309 FN(spin_lock), \
281f9e75
MKL
3310 FN(spin_unlock), \
3311 FN(sk_fullsock), \
5cce85c6 3312 FN(tcp_sock), \
ef776a27 3313 FN(skb_ecn_set_ce), \
253c8dde
LB
3314 FN(get_listener_sock), \
3315 FN(skc_lookup_tcp), \
196398d4
AI
3316 FN(tcp_check_syncookie), \
3317 FN(sysctl_get_name), \
3318 FN(sysctl_get_current_value), \
3319 FN(sysctl_get_new_value), \
b457e553
AI
3320 FN(sysctl_set_new_value), \
3321 FN(strtol), \
948d930e
MKL
3322 FN(strtoul), \
3323 FN(sk_storage_get), \
edaccf89 3324 FN(sk_storage_delete), \
3745ee18 3325 FN(send_signal), \
a7658e1a 3326 FN(tcp_gen_syncookie), \
6ae08ae3
DB
3327 FN(skb_output), \
3328 FN(probe_read_user), \
3329 FN(probe_read_kernel), \
3330 FN(probe_read_user_str), \
17328d61 3331 FN(probe_read_kernel_str), \
8482941f 3332 FN(tcp_send_ack), \
0a49c1a8 3333 FN(send_signal_thread), \
67306f84 3334 FN(jiffies64), \
b4490c5c 3335 FN(read_branch_records), \
d831ee84 3336 FN(get_ns_current_pid_tgid), \
f318903c 3337 FN(xdp_output), \
0f09abd1 3338 FN(get_netns_cookie), \
cf7fbe66 3339 FN(get_current_ancestor_cgroup_id), \
71d19214 3340 FN(sk_assign), \
492e639f
YS
3341 FN(ktime_get_boot_ns), \
3342 FN(seq_printf), \
f307fa2c
AI
3343 FN(seq_write), \
3344 FN(sk_cgroup_id), \
457f4436
AN
3345 FN(sk_ancestor_cgroup_id), \
3346 FN(ringbuf_output), \
3347 FN(ringbuf_reserve), \
3348 FN(ringbuf_submit), \
3349 FN(ringbuf_discard), \
3350 FN(ringbuf_query),
0cb34dc2 3351
971e827b
ACM
3352/* integer value in 'imm' field of BPF_CALL instruction selects which helper
3353 * function eBPF program intends to call
3354 */
0cb34dc2 3355#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
971e827b 3356enum bpf_func_id {
0cb34dc2 3357 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
971e827b
ACM
3358 __BPF_FUNC_MAX_ID,
3359};
0cb34dc2 3360#undef __BPF_ENUM_FN
971e827b
ACM
3361
3362/* All flags used by eBPF helper functions, placed here. */
3363
3364/* BPF_FUNC_skb_store_bytes flags. */
1aae4bdd
AN
3365enum {
3366 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
3367 BPF_F_INVALIDATE_HASH = (1ULL << 1),
3368};
971e827b
ACM
3369
3370/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
3371 * First 4 bits are for passing the header field size.
3372 */
1aae4bdd
AN
3373enum {
3374 BPF_F_HDR_FIELD_MASK = 0xfULL,
3375};
971e827b
ACM
3376
3377/* BPF_FUNC_l4_csum_replace flags. */
1aae4bdd
AN
3378enum {
3379 BPF_F_PSEUDO_HDR = (1ULL << 4),
3380 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
3381 BPF_F_MARK_ENFORCE = (1ULL << 6),
3382};
971e827b
ACM
3383
3384/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
1aae4bdd
AN
3385enum {
3386 BPF_F_INGRESS = (1ULL << 0),
3387};
971e827b
ACM
3388
3389/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
1aae4bdd
AN
3390enum {
3391 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
3392};
971e827b 3393
de2ff05f 3394/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
1aae4bdd
AN
3395enum {
3396 BPF_F_SKIP_FIELD_MASK = 0xffULL,
3397 BPF_F_USER_STACK = (1ULL << 8),
de2ff05f 3398/* flags used by BPF_FUNC_get_stackid only. */
1aae4bdd
AN
3399 BPF_F_FAST_STACK_CMP = (1ULL << 9),
3400 BPF_F_REUSE_STACKID = (1ULL << 10),
de2ff05f 3401/* flags used by BPF_FUNC_get_stack only. */
1aae4bdd
AN
3402 BPF_F_USER_BUILD_ID = (1ULL << 11),
3403};
971e827b
ACM
3404
3405/* BPF_FUNC_skb_set_tunnel_key flags. */
1aae4bdd
AN
3406enum {
3407 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
3408 BPF_F_DONT_FRAGMENT = (1ULL << 2),
3409 BPF_F_SEQ_NUMBER = (1ULL << 3),
3410};
971e827b 3411
e27afb84
AS
3412/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
3413 * BPF_FUNC_perf_event_read_value flags.
3414 */
1aae4bdd
AN
3415enum {
3416 BPF_F_INDEX_MASK = 0xffffffffULL,
3417 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
791cceb8 3418/* BPF_FUNC_perf_event_output for sk_buff input context. */
1aae4bdd
AN
3419 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
3420};
971e827b 3421
f71c6143 3422/* Current network namespace */
1aae4bdd
AN
3423enum {
3424 BPF_F_CURRENT_NETNS = (-1L),
3425};
f71c6143 3426
6c408dec 3427/* BPF_FUNC_skb_adjust_room flags. */
1aae4bdd
AN
3428enum {
3429 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
3430 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
3431 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
3432 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
3433 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
3434};
6c408dec 3435
1aae4bdd
AN
3436enum {
3437 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
3438 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
3439};
1db04c30 3440
bfb35c27 3441#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
1db04c30
AM
3442 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
3443 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6c408dec 3444
196398d4 3445/* BPF_FUNC_sysctl_get_name flags. */
1aae4bdd
AN
3446enum {
3447 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
3448};
196398d4 3449
948d930e 3450/* BPF_FUNC_sk_storage_get flags */
1aae4bdd
AN
3451enum {
3452 BPF_SK_STORAGE_GET_F_CREATE = (1ULL << 0),
3453};
948d930e 3454
67306f84 3455/* BPF_FUNC_read_branch_records flags. */
1aae4bdd
AN
3456enum {
3457 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
3458};
67306f84 3459
457f4436
AN
3460/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
3461 * BPF_FUNC_bpf_ringbuf_output flags.
3462 */
3463enum {
3464 BPF_RB_NO_WAKEUP = (1ULL << 0),
3465 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
3466};
3467
3468/* BPF_FUNC_bpf_ringbuf_query flags */
3469enum {
3470 BPF_RB_AVAIL_DATA = 0,
3471 BPF_RB_RING_SIZE = 1,
3472 BPF_RB_CONS_POS = 2,
3473 BPF_RB_PROD_POS = 3,
3474};
3475
3476/* BPF ring buffer constants */
3477enum {
3478 BPF_RINGBUF_BUSY_BIT = (1U << 31),
3479 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
3480 BPF_RINGBUF_HDR_SZ = 8,
3481};
3482
2be7e212
DB
3483/* Mode for BPF_FUNC_skb_adjust_room helper. */
3484enum bpf_adj_room_mode {
d62c1d72 3485 BPF_ADJ_ROOM_NET,
6c408dec 3486 BPF_ADJ_ROOM_MAC,
2be7e212
DB
3487};
3488
32b3652c
DB
3489/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
3490enum bpf_hdr_start_off {
3491 BPF_HDR_START_MAC,
3492 BPF_HDR_START_NET,
3493};
3494
c99a84ea
MX
3495/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
3496enum bpf_lwt_encap_mode {
3497 BPF_LWT_ENCAP_SEG6,
755db477
PO
3498 BPF_LWT_ENCAP_SEG6_INLINE,
3499 BPF_LWT_ENCAP_IP,
c99a84ea
MX
3500};
3501
b7df9ada
DB
3502#define __bpf_md_ptr(type, name) \
3503union { \
3504 type name; \
3505 __u64 :64; \
3506} __attribute__((aligned(8)))
3507
971e827b
ACM
3508/* user accessible mirror of in-kernel sk_buff.
3509 * new fields can only be added to the end of this structure
3510 */
3511struct __sk_buff {
3512 __u32 len;
3513 __u32 pkt_type;
3514 __u32 mark;
3515 __u32 queue_mapping;
3516 __u32 protocol;
3517 __u32 vlan_present;
3518 __u32 vlan_tci;
3519 __u32 vlan_proto;
3520 __u32 priority;
3521 __u32 ingress_ifindex;
3522 __u32 ifindex;
3523 __u32 tc_index;
3524 __u32 cb[5];
3525 __u32 hash;
3526 __u32 tc_classid;
3527 __u32 data;
3528 __u32 data_end;
b1d9fc41 3529 __u32 napi_id;
69e8cc13 3530
ac29991b 3531 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
69e8cc13
JF
3532 __u32 family;
3533 __u32 remote_ip4; /* Stored in network byte order */
3534 __u32 local_ip4; /* Stored in network byte order */
3535 __u32 remote_ip6[4]; /* Stored in network byte order */
3536 __u32 local_ip6[4]; /* Stored in network byte order */
3537 __u32 remote_port; /* Stored in network byte order */
3538 __u32 local_port; /* stored in host byte order */
ac29991b
DB
3539 /* ... here. */
3540
3541 __u32 data_meta;
b7df9ada 3542 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
f11216b2 3543 __u64 tstamp;
e3da08d0 3544 __u32 wire_len;
d9ff286a 3545 __u32 gso_segs;
281f9e75 3546 __bpf_md_ptr(struct bpf_sock *, sk);
b0ac4941 3547 __u32 gso_size;
971e827b
ACM
3548};
3549
3550struct bpf_tunnel_key {
3551 __u32 tunnel_id;
3552 union {
3553 __u32 remote_ipv4;
3554 __u32 remote_ipv6[4];
3555 };
3556 __u8 tunnel_tos;
3557 __u8 tunnel_ttl;
6b6a1925 3558 __u16 tunnel_ext; /* Padding, future use. */
971e827b
ACM
3559 __u32 tunnel_label;
3560};
3561
29a36f9e
EB
3562/* user accessible mirror of in-kernel xfrm_state.
3563 * new fields can only be added to the end of this structure
3564 */
3565struct bpf_xfrm_state {
3566 __u32 reqid;
3567 __u32 spi; /* Stored in network byte order */
3568 __u16 family;
6b6a1925 3569 __u16 ext; /* Padding, future use. */
29a36f9e
EB
3570 union {
3571 __u32 remote_ipv4; /* Stored in network byte order */
3572 __u32 remote_ipv6[4]; /* Stored in network byte order */
3573 };
3574};
3575
0cb34dc2
JS
3576/* Generic BPF return codes which all BPF program types may support.
3577 * The values are binary compatible with their TC_ACT_* counter-part to
3578 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
3579 * programs.
3580 *
3581 * XDP is handled seprately, see XDP_*.
3582 */
3583enum bpf_ret_code {
3584 BPF_OK = 0,
3585 /* 1 reserved */
3586 BPF_DROP = 2,
3587 /* 3-6 reserved */
3588 BPF_REDIRECT = 7,
755db477
PO
3589 /* >127 are reserved for prog type specific return codes.
3590 *
3591 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3592 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3593 * changed and should be routed based on its new L3 header.
3594 * (This is an L3 redirect, as opposed to L2 redirect
3595 * represented by BPF_REDIRECT above).
3596 */
3597 BPF_LWT_REROUTE = 128,
0cb34dc2
JS
3598};
3599
3600struct bpf_sock {
3601 __u32 bound_dev_if;
3602 __u32 family;
3603 __u32 type;
3604 __u32 protocol;
ac29991b
DB
3605 __u32 mark;
3606 __u32 priority;
281f9e75
MKL
3607 /* IP address also allows 1 and 2 bytes access */
3608 __u32 src_ip4;
3609 __u32 src_ip6[4];
3610 __u32 src_port; /* host byte order */
3611 __u32 dst_port; /* network byte order */
3612 __u32 dst_ip4;
3613 __u32 dst_ip6[4];
3614 __u32 state;
3615};
3616
3617struct bpf_tcp_sock {
3618 __u32 snd_cwnd; /* Sending congestion window */
3619 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3620 __u32 rtt_min;
3621 __u32 snd_ssthresh; /* Slow start size threshold */
3622 __u32 rcv_nxt; /* What we want to receive next */
3623 __u32 snd_nxt; /* Next sequence we send */
3624 __u32 snd_una; /* First byte we want an ack for */
3625 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3626 __u32 ecn_flags; /* ECN status bits. */
3627 __u32 rate_delivered; /* saved rate sample: packets delivered */
3628 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3629 __u32 packets_out; /* Packets which are "in flight" */
3630 __u32 retrans_out; /* Retransmitted packets out */
3631 __u32 total_retrans; /* Total retransmits for entire connection */
3632 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3633 * total number of segments in.
3634 */
3635 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3636 * total number of data segments in.
3637 */
3638 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3639 * The total number of segments sent.
3640 */
3641 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3642 * total number of data segments sent.
1d436885 3643 */
281f9e75
MKL
3644 __u32 lost_out; /* Lost packets */
3645 __u32 sacked_out; /* SACK'd packets */
3646 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3647 * sum(delta(rcv_nxt)), or how many bytes
3648 * were acked.
1d436885 3649 */
281f9e75
MKL
3650 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3651 * sum(delta(snd_una)), or how many bytes
3652 * were acked.
1d436885 3653 */
692cbaa9
SF
3654 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
3655 * total number of DSACK blocks received
3656 */
3657 __u32 delivered; /* Total data packets delivered incl. rexmits */
3658 __u32 delivered_ce; /* Like the above but only ECE marked packets */
3659 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
0cb34dc2
JS
3660};
3661
6acc9b43
JS
3662struct bpf_sock_tuple {
3663 union {
3664 struct {
3665 __be32 saddr;
3666 __be32 daddr;
3667 __be16 sport;
3668 __be16 dport;
3669 } ipv4;
3670 struct {
3671 __be32 saddr[4];
3672 __be32 daddr[4];
3673 __be16 sport;
3674 __be16 dport;
3675 } ipv6;
3676 };
3677};
3678
91eda599
JL
3679struct bpf_xdp_sock {
3680 __u32 queue_id;
3681};
3682
0cb34dc2
JS
3683#define XDP_PACKET_HEADROOM 256
3684
791cceb8
ACM
3685/* User return codes for XDP prog type.
3686 * A valid XDP program must return one of these defined values. All other
ac29991b
DB
3687 * return codes are reserved for future use. Unknown return codes will
3688 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
791cceb8
ACM
3689 */
3690enum xdp_action {
3691 XDP_ABORTED = 0,
3692 XDP_DROP,
3693 XDP_PASS,
3694 XDP_TX,
ac29991b 3695 XDP_REDIRECT,
791cceb8
ACM
3696};
3697
3698/* user accessible metadata for XDP packet hook
3699 * new fields must be added to the end of this structure
3700 */
3701struct xdp_md {
3702 __u32 data;
3703 __u32 data_end;
ac29991b 3704 __u32 data_meta;
e7b2823a
JDB
3705 /* Below access go through struct xdp_rxq_info */
3706 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3707 __u32 rx_queue_index; /* rxq->queue_index */
791cceb8
ACM
3708};
3709
69e8cc13 3710enum sk_action {
bfa64075
JF
3711 SK_DROP = 0,
3712 SK_PASS,
69e8cc13
JF
3713};
3714
82a86168
JF
3715/* user accessible metadata for SK_MSG packet hook, new fields must
3716 * be added to the end of this structure
3717 */
3718struct sk_msg_md {
b7df9ada
DB
3719 __bpf_md_ptr(void *, data);
3720 __bpf_md_ptr(void *, data_end);
4da0dcab
JF
3721
3722 __u32 family;
3723 __u32 remote_ip4; /* Stored in network byte order */
3724 __u32 local_ip4; /* Stored in network byte order */
3725 __u32 remote_ip6[4]; /* Stored in network byte order */
3726 __u32 local_ip6[4]; /* Stored in network byte order */
3727 __u32 remote_port; /* Stored in network byte order */
3728 __u32 local_port; /* stored in host byte order */
584e4681 3729 __u32 size; /* Total size of sk_msg */
13d70f5a
JF
3730
3731 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
82a86168
JF
3732};
3733
3bd43a8c
MKL
3734struct sk_reuseport_md {
3735 /*
3736 * Start of directly accessible data. It begins from
3737 * the tcp/udp header.
3738 */
b7df9ada
DB
3739 __bpf_md_ptr(void *, data);
3740 /* End of directly accessible data */
3741 __bpf_md_ptr(void *, data_end);
3bd43a8c
MKL
3742 /*
3743 * Total length of packet (starting from the tcp/udp header).
3744 * Note that the directly accessible bytes (data_end - data)
3745 * could be less than this "len". Those bytes could be
3746 * indirectly read by a helper "bpf_skb_load_bytes()".
3747 */
3748 __u32 len;
3749 /*
3750 * Eth protocol in the mac header (network byte order). e.g.
3751 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3752 */
3753 __u32 eth_protocol;
3754 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3755 __u32 bind_inany; /* Is sock bound to an INANY address? */
3756 __u32 hash; /* A hash of the packet 4 tuples */
3757};
3758
95b9afd3
MKL
3759#define BPF_TAG_SIZE 8
3760
3761struct bpf_prog_info {
3762 __u32 type;
3763 __u32 id;
3764 __u8 tag[BPF_TAG_SIZE];
3765 __u32 jited_prog_len;
3766 __u32 xlated_prog_len;
3767 __aligned_u64 jited_prog_insns;
3768 __aligned_u64 xlated_prog_insns;
88cda1c9
MKL
3769 __u64 load_time; /* ns since boottime */
3770 __u32 created_by_uid;
3771 __u32 nr_map_ids;
3772 __aligned_u64 map_ids;
e27afb84 3773 char name[BPF_OBJ_NAME_LEN];
675fc275 3774 __u32 ifindex;
fb6ef42b 3775 __u32 gpl_compatible:1;
0472301a 3776 __u32 :31; /* alignment pad */
675fc275
JK
3777 __u64 netns_dev;
3778 __u64 netns_ino;
dd0c5f07 3779 __u32 nr_jited_ksyms;
bd980d43 3780 __u32 nr_jited_func_lens;
dd0c5f07 3781 __aligned_u64 jited_ksyms;
bd980d43 3782 __aligned_u64 jited_func_lens;
cc19435c
YS
3783 __u32 btf_id;
3784 __u32 func_info_rec_size;
3785 __aligned_u64 func_info;
b4f8623c
YS
3786 __u32 nr_func_info;
3787 __u32 nr_line_info;
ee491d8d
MKL
3788 __aligned_u64 line_info;
3789 __aligned_u64 jited_line_info;
b4f8623c 3790 __u32 nr_jited_line_info;
ee491d8d
MKL
3791 __u32 line_info_rec_size;
3792 __u32 jited_line_info_rec_size;
eb896a69
SL
3793 __u32 nr_prog_tags;
3794 __aligned_u64 prog_tags;
b1eca86d
AS
3795 __u64 run_time_ns;
3796 __u64 run_cnt;
95b9afd3
MKL
3797} __attribute__((aligned(8)));
3798
3799struct bpf_map_info {
3800 __u32 type;
3801 __u32 id;
3802 __u32 key_size;
3803 __u32 value_size;
3804 __u32 max_entries;
3805 __u32 map_flags;
067cae47 3806 char name[BPF_OBJ_NAME_LEN];
52775b33 3807 __u32 ifindex;
17328d61 3808 __u32 btf_vmlinux_value_type_id;
52775b33
JK
3809 __u64 netns_dev;
3810 __u64 netns_ino;
7a01f6a3 3811 __u32 btf_id;
f03b15d3
MKL
3812 __u32 btf_key_type_id;
3813 __u32 btf_value_type_id;
7a01f6a3
MKL
3814} __attribute__((aligned(8)));
3815
3816struct bpf_btf_info {
3817 __aligned_u64 btf;
3818 __u32 btf_size;
3819 __u32 id;
95b9afd3
MKL
3820} __attribute__((aligned(8)));
3821
f2e10bff
AN
3822struct bpf_link_info {
3823 __u32 type;
3824 __u32 id;
3825 __u32 prog_id;
3826 union {
3827 struct {
3828 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
3829 __u32 tp_name_len; /* in/out: tp_name buffer len */
3830 } raw_tracepoint;
3831 struct {
3832 __u32 attach_type;
3833 } tracing;
3834 struct {
3835 __u64 cgroup_id;
3836 __u32 attach_type;
3837 } cgroup;
3838 };
3839} __attribute__((aligned(8)));
3840
e50b0a6f
AI
3841/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3842 * by user and intended to be used by socket (e.g. to bind to, depends on
3843 * attach attach type).
3844 */
3845struct bpf_sock_addr {
3846 __u32 user_family; /* Allows 4-byte read, but no write. */
3847 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3848 * Stored in network byte order.
3849 */
073a4834 3850 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
e50b0a6f
AI
3851 * Stored in network byte order.
3852 */
7aebfa1b 3853 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
e50b0a6f
AI
3854 * Stored in network byte order
3855 */
3856 __u32 family; /* Allows 4-byte read, but no write */
3857 __u32 type; /* Allows 4-byte read, but no write */
3858 __u32 protocol; /* Allows 4-byte read, but no write */
4cfacbe6 3859 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3024cf82
AI
3860 * Stored in network byte order.
3861 */
073a4834 3862 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
3024cf82
AI
3863 * Stored in network byte order.
3864 */
cd17d777 3865 __bpf_md_ptr(struct bpf_sock *, sk);
e50b0a6f
AI
3866};
3867
04df41e3
LB
3868/* User bpf_sock_ops struct to access socket values and specify request ops
3869 * and their replies.
f1d6cb2d
ACM
3870 * Some of this fields are in network (bigendian) byte order and may need
3871 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
04df41e3
LB
3872 * New fields can only be added at the end of this structure
3873 */
3874struct bpf_sock_ops {
3875 __u32 op;
3876 union {
d6d4f60c
LB
3877 __u32 args[4]; /* Optionally passed to bpf program */
3878 __u32 reply; /* Returned by bpf program */
3879 __u32 replylong[4]; /* Optionally returned by bpf prog */
04df41e3
LB
3880 };
3881 __u32 family;
f1d6cb2d
ACM
3882 __u32 remote_ip4; /* Stored in network byte order */
3883 __u32 local_ip4; /* Stored in network byte order */
3884 __u32 remote_ip6[4]; /* Stored in network byte order */
3885 __u32 local_ip6[4]; /* Stored in network byte order */
3886 __u32 remote_port; /* Stored in network byte order */
3887 __u32 local_port; /* stored in host byte order */
e7b2823a
JDB
3888 __u32 is_fullsock; /* Some TCP fields are only valid if
3889 * there is a full socket. If not, the
3890 * fields read as zero.
3891 */
3892 __u32 snd_cwnd;
3893 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
d6d4f60c
LB
3894 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3895 __u32 state;
3896 __u32 rtt_min;
3897 __u32 snd_ssthresh;
3898 __u32 rcv_nxt;
3899 __u32 snd_nxt;
3900 __u32 snd_una;
3901 __u32 mss_cache;
3902 __u32 ecn_flags;
3903 __u32 rate_delivered;
3904 __u32 rate_interval_us;
3905 __u32 packets_out;
3906 __u32 retrans_out;
3907 __u32 total_retrans;
3908 __u32 segs_in;
3909 __u32 data_segs_in;
3910 __u32 segs_out;
3911 __u32 data_segs_out;
3912 __u32 lost_out;
3913 __u32 sacked_out;
3914 __u32 sk_txhash;
3915 __u64 bytes_received;
3916 __u64 bytes_acked;
cd17d777 3917 __bpf_md_ptr(struct bpf_sock *, sk);
04df41e3
LB
3918};
3919
d6d4f60c 3920/* Definitions for bpf_sock_ops_cb_flags */
1aae4bdd
AN
3921enum {
3922 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
3923 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
3924 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
3925 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
3926/* Mask of all currently supported cb flags */
3927 BPF_SOCK_OPS_ALL_CB_FLAGS = 0xF,
3928};
d6d4f60c 3929
04df41e3
LB
3930/* List of known BPF sock_ops operators.
3931 * New entries can only be added at the end
3932 */
3933enum {
3934 BPF_SOCK_OPS_VOID,
3935 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3936 * -1 if default value should be used
3937 */
3938 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3939 * window (in packets) or -1 if default
3940 * value should be used
3941 */
3942 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3943 * active connection is initialized
3944 */
3945 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3946 * active connection is
3947 * established
3948 */
3949 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3950 * passive connection is
3951 * established
3952 */
3953 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3954 * needs ECN
3955 */
e27afb84
AS
3956 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3957 * based on the path and may be
3958 * dependent on the congestion control
3959 * algorithm. In general it indicates
3960 * a congestion threshold. RTTs above
3961 * this indicate congestion
3962 */
d6d4f60c
LB
3963 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3964 * Arg1: value of icsk_retransmits
3965 * Arg2: value of icsk_rto
3966 * Arg3: whether RTO has expired
3967 */
3968 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3969 * Arg1: sequence number of 1st byte
3970 * Arg2: # segments
3971 * Arg3: return value of
3972 * tcp_transmit_skb (0 => success)
3973 */
3974 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3975 * Arg1: old_state
3976 * Arg2: new_state
3977 */
060a7fcc
AI
3978 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3979 * socket transition to LISTEN state.
3980 */
692cbaa9
SF
3981 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
3982 */
d6d4f60c
LB
3983};
3984
3985/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3986 * changes between the TCP and BPF versions. Ideally this should never happen.
3987 * If it does, we need to add code to convert them before calling
3988 * the BPF sock_ops function.
3989 */
3990enum {
3991 BPF_TCP_ESTABLISHED = 1,
3992 BPF_TCP_SYN_SENT,
3993 BPF_TCP_SYN_RECV,
3994 BPF_TCP_FIN_WAIT1,
3995 BPF_TCP_FIN_WAIT2,
3996 BPF_TCP_TIME_WAIT,
3997 BPF_TCP_CLOSE,
3998 BPF_TCP_CLOSE_WAIT,
3999 BPF_TCP_LAST_ACK,
4000 BPF_TCP_LISTEN,
4001 BPF_TCP_CLOSING, /* Now a valid state */
4002 BPF_TCP_NEW_SYN_RECV,
4003
4004 BPF_TCP_MAX_STATES /* Leave at the end! */
04df41e3
LB
4005};
4006
1aae4bdd
AN
4007enum {
4008 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
4009 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
4010};
04df41e3 4011
e27afb84
AS
4012struct bpf_perf_event_value {
4013 __u64 counter;
4014 __u64 enabled;
4015 __u64 running;
4016};
4017
1aae4bdd
AN
4018enum {
4019 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
4020 BPF_DEVCG_ACC_READ = (1ULL << 1),
4021 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
4022};
ebc614f6 4023
1aae4bdd
AN
4024enum {
4025 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
4026 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
4027};
ebc614f6
RG
4028
4029struct bpf_cgroup_dev_ctx {
e7b2823a
JDB
4030 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
4031 __u32 access_type;
ebc614f6
RG
4032 __u32 major;
4033 __u32 minor;
4034};
4035
a0fe3e57
AS
4036struct bpf_raw_tracepoint_args {
4037 __u64 args[0];
4038};
4039
cb9c28ef
PB
4040/* DIRECT: Skip the FIB rules and go to FIB table associated with device
4041 * OUTPUT: Do lookup from egress perspective; default is ingress
4042 */
1aae4bdd
AN
4043enum {
4044 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
4045 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
4046};
cb9c28ef 4047
9b8ca379
QM
4048enum {
4049 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
4050 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
4051 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
4052 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
4053 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
4054 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
4055 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
4056 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
4057 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
4058};
4059
cb9c28ef 4060struct bpf_fib_lookup {
6bdd533c
SY
4061 /* input: network family for lookup (AF_INET, AF_INET6)
4062 * output: network family of egress nexthop
4063 */
4064 __u8 family;
cb9c28ef
PB
4065
4066 /* set if lookup is to consider L4 data - e.g., FIB rules */
4067 __u8 l4_protocol;
4068 __be16 sport;
4069 __be16 dport;
4070
4071 /* total length of packet from network header - used for MTU check */
4072 __u16 tot_len;
9b8ca379
QM
4073
4074 /* input: L3 device index for lookup
4075 * output: device index from FIB lookup
4076 */
4077 __u32 ifindex;
cb9c28ef
PB
4078
4079 union {
4080 /* inputs to lookup */
4081 __u8 tos; /* AF_INET */
f568b472 4082 __be32 flowinfo; /* AF_INET6, flow_label + priority */
cb9c28ef 4083
6bdd533c
SY
4084 /* output: metric of fib result (IPv4/IPv6 only) */
4085 __u32 rt_metric;
cb9c28ef
PB
4086 };
4087
4088 union {
cb9c28ef
PB
4089 __be32 ipv4_src;
4090 __u32 ipv6_src[4]; /* in6_addr; network order */
4091 };
4092
6bdd533c
SY
4093 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
4094 * network header. output: bpf_fib_lookup sets to gateway address
4095 * if FIB lookup returns gateway route
cb9c28ef
PB
4096 */
4097 union {
cb9c28ef
PB
4098 __be32 ipv4_dst;
4099 __u32 ipv6_dst[4]; /* in6_addr; network order */
4100 };
4101
4102 /* output */
4103 __be16 h_vlan_proto;
4104 __be16 h_vlan_TCI;
4105 __u8 smac[6]; /* ETH_ALEN */
4106 __u8 dmac[6]; /* ETH_ALEN */
4107};
4108
30687ad9
YS
4109enum bpf_task_fd_type {
4110 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
4111 BPF_FD_TYPE_TRACEPOINT, /* tp name */
4112 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
4113 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
4114 BPF_FD_TYPE_UPROBE, /* filename + offset */
4115 BPF_FD_TYPE_URETPROBE, /* filename + offset */
4116};
4117
1aae4bdd
AN
4118enum {
4119 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
4120 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
4121 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
4122};
57debff2 4123
2f965e3f
PP
4124struct bpf_flow_keys {
4125 __u16 nhoff;
4126 __u16 thoff;
4127 __u16 addr_proto; /* ETH_P_* of valid addrs */
4128 __u8 is_frag;
4129 __u8 is_first_frag;
4130 __u8 is_encap;
4131 __u8 ip_proto;
4132 __be16 n_proto;
4133 __be16 sport;
4134 __be16 dport;
4135 union {
4136 struct {
4137 __be32 ipv4_src;
4138 __be32 ipv4_dst;
4139 };
4140 struct {
4141 __u32 ipv6_src[4]; /* in6_addr; network order */
4142 __u32 ipv6_dst[4]; /* in6_addr; network order */
4143 };
4144 };
57debff2 4145 __u32 flags;
71c99e32 4146 __be32 flow_label;
2f965e3f
PP
4147};
4148
cc19435c 4149struct bpf_func_info {
555249df 4150 __u32 insn_off;
cc19435c
YS
4151 __u32 type_id;
4152};
4153
ee491d8d
MKL
4154#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
4155#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
4156
4157struct bpf_line_info {
4158 __u32 insn_off;
4159 __u32 file_name_off;
4160 __u32 line_off;
4161 __u32 line_col;
4162};
4163
7dac3ae4
AS
4164struct bpf_spin_lock {
4165 __u32 val;
4166};
196398d4
AI
4167
4168struct bpf_sysctl {
4169 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
4170 * Allows 1,2,4-byte read, but no write.
4171 */
4172 __u32 file_pos; /* Sysctl file position to read from, write to.
4173 * Allows 1,2,4-byte read an 4-byte write.
4174 */
4175};
4176
aa6ab647
SF
4177struct bpf_sockopt {
4178 __bpf_md_ptr(struct bpf_sock *, sk);
4179 __bpf_md_ptr(void *, optval);
4180 __bpf_md_ptr(void *, optval_end);
4181
4182 __s32 level;
4183 __s32 optname;
4184 __s32 optlen;
4185 __s32 retval;
4186};
4187
b4490c5c
CN
4188struct bpf_pidns_info {
4189 __u32 pid;
4190 __u32 tgid;
4191};
971e827b 4192#endif /* _UAPI__LINUX_BPF_H__ */