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