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