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