]> git.proxmox.com Git - mirror_iproute2.git/blame - include/uapi/linux/bpf.h
Merge branch 'master' into next
[mirror_iproute2.git] / include / uapi / linux / bpf.h
CommitLineData
ba914908 1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
b54ac87e
DB
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 __LINUX_BPF_H__
9#define __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 */
f79b7733 17#define BPF_JMP32 0x06 /* jmp mode in word width */
b54ac87e
DB
18#define BPF_ALU64 0x07 /* alu mode in double word width */
19
20/* ld/ldx fields */
07ed8df6 21#define BPF_DW 0x18 /* double word (64-bit) */
b54ac87e
DB
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
3af3d358 35/* jmp encodings */
b54ac87e 36#define BPF_JNE 0x50 /* jump != */
3af3d358
SH
37#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
38#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
b54ac87e
DB
39#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
40#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
3af3d358
SH
41#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
42#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
b54ac87e
DB
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
b479a7d7
SH
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 */
76 __u8 data[0]; /* Arbitrary size */
77};
78
74eb09ad
DA
79struct bpf_cgroup_storage_key {
80 __u64 cgroup_inode_id; /* cgroup inode id */
81 __u32 attach_type; /* program attach type */
82};
83
0198930b 84/* BPF syscall commands, see bpf(2) man-page for details. */
b54ac87e 85enum bpf_cmd {
b54ac87e 86 BPF_MAP_CREATE,
b54ac87e 87 BPF_MAP_LOOKUP_ELEM,
b54ac87e 88 BPF_MAP_UPDATE_ELEM,
b54ac87e 89 BPF_MAP_DELETE_ELEM,
b54ac87e 90 BPF_MAP_GET_NEXT_KEY,
b54ac87e 91 BPF_PROG_LOAD,
0198930b
SH
92 BPF_OBJ_PIN,
93 BPF_OBJ_GET,
1a97748b
SH
94 BPF_PROG_ATTACH,
95 BPF_PROG_DETACH,
45f78b4d 96 BPF_PROG_TEST_RUN,
410556ad
SH
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,
f53da99a 102 BPF_PROG_QUERY,
dcf7997b 103 BPF_RAW_TRACEPOINT_OPEN,
fd95ec0e 104 BPF_BTF_LOAD,
4276e652 105 BPF_BTF_GET_FD_BY_ID,
57ac202c 106 BPF_TASK_FD_QUERY,
39776a86 107 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
188c7fe6 108 BPF_MAP_FREEZE,
b54ac87e
DB
109};
110
111enum bpf_map_type {
112 BPF_MAP_TYPE_UNSPEC,
113 BPF_MAP_TYPE_HASH,
114 BPF_MAP_TYPE_ARRAY,
8f42ceaf 115 BPF_MAP_TYPE_PROG_ARRAY,
4f3489cd 116 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
8593b2ca
SH
117 BPF_MAP_TYPE_PERCPU_HASH,
118 BPF_MAP_TYPE_PERCPU_ARRAY,
2421ab75 119 BPF_MAP_TYPE_STACK_TRACE,
4824bb41 120 BPF_MAP_TYPE_CGROUP_ARRAY,
1a97748b
SH
121 BPF_MAP_TYPE_LRU_HASH,
122 BPF_MAP_TYPE_LRU_PERCPU_HASH,
b479a7d7 123 BPF_MAP_TYPE_LPM_TRIE,
45f78b4d
SH
124 BPF_MAP_TYPE_ARRAY_OF_MAPS,
125 BPF_MAP_TYPE_HASH_OF_MAPS,
f0b9b795 126 BPF_MAP_TYPE_DEVMAP,
18d7817c 127 BPF_MAP_TYPE_SOCKMAP,
fa19d6bc 128 BPF_MAP_TYPE_CPUMAP,
fd95ec0e 129 BPF_MAP_TYPE_XSKMAP,
4276e652 130 BPF_MAP_TYPE_SOCKHASH,
74eb09ad 131 BPF_MAP_TYPE_CGROUP_STORAGE,
0ebb4209 132 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
39776a86
SH
133 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
134 BPF_MAP_TYPE_QUEUE,
135 BPF_MAP_TYPE_STACK,
fd658097 136 BPF_MAP_TYPE_SK_STORAGE,
b54ac87e
DB
137};
138
fdce94d0
DA
139/* Note that tracing related programs such as
140 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
141 * are not subject to a stable API since kernel internal data
142 * structures can change from release to release and may
143 * therefore break existing tracing BPF programs. Tracing BPF
144 * programs correspond to /a/ specific kernel which is to be
145 * analyzed, and not /a/ specific kernel /and/ all future ones.
146 */
b54ac87e
DB
147enum bpf_prog_type {
148 BPF_PROG_TYPE_UNSPEC,
149 BPF_PROG_TYPE_SOCKET_FILTER,
aeedd8e1 150 BPF_PROG_TYPE_KPROBE,
b54ac87e 151 BPF_PROG_TYPE_SCHED_CLS,
cbdc3ed8 152 BPF_PROG_TYPE_SCHED_ACT,
bc9fb257 153 BPF_PROG_TYPE_TRACEPOINT,
ba91cd9d 154 BPF_PROG_TYPE_XDP,
e8a67bc4 155 BPF_PROG_TYPE_PERF_EVENT,
1a97748b 156 BPF_PROG_TYPE_CGROUP_SKB,
143a704b
SH
157 BPF_PROG_TYPE_CGROUP_SOCK,
158 BPF_PROG_TYPE_LWT_IN,
159 BPF_PROG_TYPE_LWT_OUT,
160 BPF_PROG_TYPE_LWT_XMIT,
f0b9b795 161 BPF_PROG_TYPE_SOCK_OPS,
18d7817c 162 BPF_PROG_TYPE_SK_SKB,
665ef5a5 163 BPF_PROG_TYPE_CGROUP_DEVICE,
9effc146 164 BPF_PROG_TYPE_SK_MSG,
dcf7997b
SH
165 BPF_PROG_TYPE_RAW_TRACEPOINT,
166 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
57ac202c 167 BPF_PROG_TYPE_LWT_SEG6LOCAL,
17678d30 168 BPF_PROG_TYPE_LIRC_MODE2,
0ebb4209 169 BPF_PROG_TYPE_SK_REUSEPORT,
d9c0be4e 170 BPF_PROG_TYPE_FLOW_DISSECTOR,
70de8a7f 171 BPF_PROG_TYPE_CGROUP_SYSCTL,
fd658097 172 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
b54ac87e
DB
173};
174
1a97748b
SH
175enum bpf_attach_type {
176 BPF_CGROUP_INET_INGRESS,
177 BPF_CGROUP_INET_EGRESS,
143a704b 178 BPF_CGROUP_INET_SOCK_CREATE,
f0b9b795 179 BPF_CGROUP_SOCK_OPS,
8707fd8c
SH
180 BPF_SK_SKB_STREAM_PARSER,
181 BPF_SK_SKB_STREAM_VERDICT,
665ef5a5 182 BPF_CGROUP_DEVICE,
9effc146 183 BPF_SK_MSG_VERDICT,
dcf7997b
SH
184 BPF_CGROUP_INET4_BIND,
185 BPF_CGROUP_INET6_BIND,
186 BPF_CGROUP_INET4_CONNECT,
187 BPF_CGROUP_INET6_CONNECT,
188 BPF_CGROUP_INET4_POST_BIND,
189 BPF_CGROUP_INET6_POST_BIND,
17678d30
SH
190 BPF_CGROUP_UDP4_SENDMSG,
191 BPF_CGROUP_UDP6_SENDMSG,
192 BPF_LIRC_MODE2,
d9c0be4e 193 BPF_FLOW_DISSECTOR,
70de8a7f 194 BPF_CGROUP_SYSCTL,
77a38037
SH
195 BPF_CGROUP_UDP4_RECVMSG,
196 BPF_CGROUP_UDP6_RECVMSG,
1a97748b
SH
197 __MAX_BPF_ATTACH_TYPE
198};
199
200#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
201
f53da99a
SH
202/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
203 *
204 * NONE(default): No further bpf programs allowed in the subtree.
205 *
206 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
207 * the program in this cgroup yields to sub-cgroup program.
208 *
209 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
210 * that cgroup program gets run in addition to the program in this cgroup.
211 *
212 * Only one program is allowed to be attached to a cgroup with
213 * NONE or BPF_F_ALLOW_OVERRIDE flag.
214 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
215 * release old program and attach the new one. Attach flags has to match.
216 *
217 * Multiple programs are allowed to be attached to a cgroup with
218 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
219 * (those that were attached first, run first)
220 * The programs of sub-cgroup are executed first, then programs of
221 * this cgroup and then programs of parent cgroup.
222 * When children program makes decision (like picking TCP CA or sock bind)
223 * parent program has a chance to override it.
224 *
225 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
226 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
227 * Ex1:
228 * cgrp1 (MULTI progs A, B) ->
229 * cgrp2 (OVERRIDE prog C) ->
230 * cgrp3 (MULTI prog D) ->
231 * cgrp4 (OVERRIDE prog E) ->
232 * cgrp5 (NONE prog F)
233 * the event in cgrp5 triggers execution of F,D,A,B in that order.
234 * if prog F is detached, the execution is E,D,A,B
235 * if prog F and D are detached, the execution is E,A,B
236 * if prog F, E and D are detached, the execution is C,A,B
237 *
238 * All eligible programs are executed regardless of return code from
239 * earlier programs.
cad54934
SH
240 */
241#define BPF_F_ALLOW_OVERRIDE (1U << 0)
f53da99a 242#define BPF_F_ALLOW_MULTI (1U << 1)
cad54934 243
a2325adf
SH
244/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
245 * verifier will perform strict alignment checking as if the kernel
246 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
247 * and NET_IP_ALIGN defined to 2.
248 */
249#define BPF_F_STRICT_ALIGNMENT (1U << 0)
250
17689d30
DA
251/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
252 * verifier will allow any alignment whatsoever. On platforms
253 * with strict alignment requirements for loads ands stores (such
254 * as sparc and mips) the verifier validates that all loads and
255 * stores provably follow this requirement. This flag turns that
256 * checking and enforcement off.
257 *
258 * It is mostly used for testing when we want to validate the
259 * context and memory access aspects of the verifier, but because
260 * of an unaligned access the alignment check would trigger before
261 * the one we are interested in.
262 */
263#define BPF_F_ANY_ALIGNMENT (1U << 1)
264
b2f8eb7f
DA
265/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
266 * Verifier does sub-register def/use analysis and identifies instructions whose
267 * def only matters for low 32-bit, high 32-bit is never referenced later
268 * through implicit zero extension. Therefore verifier notifies JIT back-ends
269 * that it is safe to ignore clearing high 32-bit for these instructions. This
270 * saves some back-ends a lot of code-gen. However such optimization is not
271 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
272 * hence hasn't used verifier's analysis result. But, we really want to have a
273 * way to be able to verify the correctness of the described optimization on
274 * x86_64 on which testsuites are frequently exercised.
275 *
276 * So, this flag is introduced. Once it is set, verifier will randomize high
277 * 32-bit for those instructions who has been identified as safe to ignore them.
278 * Then, if verifier is not doing correct analysis, such randomization will
279 * regress tests to expose bugs.
280 */
281#define BPF_F_TEST_RND_HI32 (1U << 2)
282
188c7fe6
DA
283/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
284 * two extensions:
285 *
286 * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE
287 * insn[0].imm: map fd map fd
288 * insn[1].imm: 0 offset into value
289 * insn[0].off: 0 0
290 * insn[1].off: 0 0
291 * ldimm64 rewrite: address of map address of map[0]+offset
292 * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE
293 */
b54ac87e 294#define BPF_PSEUDO_MAP_FD 1
188c7fe6 295#define BPF_PSEUDO_MAP_VALUE 2
b54ac87e 296
f82517f8
DA
297/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
298 * offset to another bpf function
299 */
300#define BPF_PSEUDO_CALL 1
301
b54ac87e
DB
302/* flags for BPF_MAP_UPDATE_ELEM command */
303#define BPF_ANY 0 /* create new element or update existing */
304#define BPF_NOEXIST 1 /* create new element if it didn't exist */
305#define BPF_EXIST 2 /* update existing element */
6b2d60bd 306#define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
b54ac87e 307
18d7817c 308/* flags for BPF_MAP_CREATE command */
b7e0091a 309#define BPF_F_NO_PREALLOC (1U << 0)
1a97748b
SH
310/* Instead of having one common LRU list in the
311 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
312 * which can scale and perform better.
313 * Note, the LRU nodes (including free nodes) cannot be moved
314 * across different LRU lists.
315 */
316#define BPF_F_NO_COMMON_LRU (1U << 1)
18d7817c
SH
317/* Specify numa node during map creation */
318#define BPF_F_NUMA_NODE (1U << 2)
b7e0091a 319
1db903de
SH
320#define BPF_OBJ_NAME_LEN 16U
321
188c7fe6 322/* Flags for accessing BPF object from syscall side. */
fa19d6bc
SH
323#define BPF_F_RDONLY (1U << 3)
324#define BPF_F_WRONLY (1U << 4)
325
9effc146
DA
326/* Flag for stack_map, store build_id+offset instead of pointer */
327#define BPF_F_STACK_BUILD_ID (1U << 5)
328
17689d30
DA
329/* Zero-initialize hash function seed. This should only be used for testing. */
330#define BPF_F_ZERO_SEED (1U << 6)
331
188c7fe6
DA
332/* Flags for accessing BPF object from program side. */
333#define BPF_F_RDONLY_PROG (1U << 7)
334#define BPF_F_WRONLY_PROG (1U << 8)
335
17689d30
DA
336/* flags for BPF_PROG_QUERY */
337#define BPF_F_QUERY_EFFECTIVE (1U << 0)
338
9effc146
DA
339enum bpf_stack_build_id_status {
340 /* user space need an empty entry to identify end of a trace */
341 BPF_STACK_BUILD_ID_EMPTY = 0,
342 /* with valid build_id and offset */
343 BPF_STACK_BUILD_ID_VALID = 1,
344 /* couldn't get build_id, fallback to ip */
345 BPF_STACK_BUILD_ID_IP = 2,
346};
347
348#define BPF_BUILD_ID_SIZE 20
349struct bpf_stack_build_id {
350 __s32 status;
351 unsigned char build_id[BPF_BUILD_ID_SIZE];
352 union {
353 __u64 offset;
354 __u64 ip;
355 };
356};
357
b54ac87e
DB
358union bpf_attr {
359 struct { /* anonymous struct used by BPF_MAP_CREATE command */
360 __u32 map_type; /* one of enum bpf_map_type */
361 __u32 key_size; /* size of key in bytes */
362 __u32 value_size; /* size of value in bytes */
363 __u32 max_entries; /* max number of entries in a map */
18d7817c
SH
364 __u32 map_flags; /* BPF_MAP_CREATE related
365 * flags defined above.
366 */
45f78b4d 367 __u32 inner_map_fd; /* fd pointing to the inner map */
18d7817c
SH
368 __u32 numa_node; /* numa node (effective only if
369 * BPF_F_NUMA_NODE is set).
370 */
f53da99a 371 char map_name[BPF_OBJ_NAME_LEN];
c0788a09 372 __u32 map_ifindex; /* ifindex of netdev to create on */
fd95ec0e 373 __u32 btf_fd; /* fd pointing to a BTF type data */
57ac202c
DA
374 __u32 btf_key_type_id; /* BTF type_id of the key */
375 __u32 btf_value_type_id; /* BTF type_id of the value */
b54ac87e
DB
376 };
377
378 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
379 __u32 map_fd;
380 __aligned_u64 key;
381 union {
382 __aligned_u64 value;
383 __aligned_u64 next_key;
384 };
385 __u64 flags;
386 };
387
388 struct { /* anonymous struct used by BPF_PROG_LOAD command */
389 __u32 prog_type; /* one of enum bpf_prog_type */
390 __u32 insn_cnt;
391 __aligned_u64 insns;
392 __aligned_u64 license;
393 __u32 log_level; /* verbosity level of verifier */
394 __u32 log_size; /* size of user buffer */
395 __aligned_u64 log_buf; /* user supplied buffer */
fdce94d0 396 __u32 kern_version; /* not used */
a2325adf 397 __u32 prog_flags;
f53da99a 398 char prog_name[BPF_OBJ_NAME_LEN];
b38778bb 399 __u32 prog_ifindex; /* ifindex of netdev to prep for */
dcf7997b
SH
400 /* For some prog types expected attach type must be known at
401 * load time to verify attach type specific parts of prog
402 * (context accesses, allowed helpers, etc).
403 */
404 __u32 expected_attach_type;
17689d30
DA
405 __u32 prog_btf_fd; /* fd pointing to BTF type data */
406 __u32 func_info_rec_size; /* userspace bpf_func_info size */
407 __aligned_u64 func_info; /* func info */
408 __u32 func_info_cnt; /* number of bpf_func_info records */
409 __u32 line_info_rec_size; /* userspace bpf_line_info size */
410 __aligned_u64 line_info; /* line info */
411 __u32 line_info_cnt; /* number of bpf_line_info records */
b54ac87e 412 };
0198930b
SH
413
414 struct { /* anonymous struct used by BPF_OBJ_* commands */
415 __aligned_u64 pathname;
416 __u32 bpf_fd;
fa19d6bc 417 __u32 file_flags;
0198930b 418 };
1a97748b
SH
419
420 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
421 __u32 target_fd; /* container object to attach to */
422 __u32 attach_bpf_fd; /* eBPF program to attach */
423 __u32 attach_type;
cad54934 424 __u32 attach_flags;
1a97748b 425 };
45f78b4d
SH
426
427 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
428 __u32 prog_fd;
429 __u32 retval;
17689d30
DA
430 __u32 data_size_in; /* input: len of data_in */
431 __u32 data_size_out; /* input/output: len of data_out
432 * returns ENOSPC if data_out
433 * is too small.
434 */
45f78b4d
SH
435 __aligned_u64 data_in;
436 __aligned_u64 data_out;
437 __u32 repeat;
438 __u32 duration;
188c7fe6
DA
439 __u32 ctx_size_in; /* input: len of ctx_in */
440 __u32 ctx_size_out; /* input/output: len of ctx_out
441 * returns ENOSPC if ctx_out
442 * is too small.
443 */
444 __aligned_u64 ctx_in;
445 __aligned_u64 ctx_out;
45f78b4d 446 } test;
410556ad
SH
447
448 struct { /* anonymous struct used by BPF_*_GET_*_ID */
449 union {
450 __u32 start_id;
451 __u32 prog_id;
452 __u32 map_id;
4276e652 453 __u32 btf_id;
410556ad
SH
454 };
455 __u32 next_id;
fa19d6bc 456 __u32 open_flags;
410556ad
SH
457 };
458
459 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
460 __u32 bpf_fd;
461 __u32 info_len;
462 __aligned_u64 info;
463 } info;
f53da99a
SH
464
465 struct { /* anonymous struct used by BPF_PROG_QUERY command */
466 __u32 target_fd; /* container object to query */
467 __u32 attach_type;
468 __u32 query_flags;
469 __u32 attach_flags;
470 __aligned_u64 prog_ids;
471 __u32 prog_cnt;
472 } query;
dcf7997b
SH
473
474 struct {
475 __u64 name;
476 __u32 prog_fd;
477 } raw_tracepoint;
fd95ec0e
DA
478
479 struct { /* anonymous struct for BPF_BTF_LOAD */
480 __aligned_u64 btf;
481 __aligned_u64 btf_log_buf;
482 __u32 btf_size;
483 __u32 btf_log_size;
484 __u32 btf_log_level;
485 };
57ac202c
DA
486
487 struct {
488 __u32 pid; /* input: pid */
489 __u32 fd; /* input: fd */
490 __u32 flags; /* input: flags */
491 __u32 buf_len; /* input/output: buf len */
492 __aligned_u64 buf; /* input/output:
493 * tp_name for tracepoint
494 * symbol for kprobe
495 * filename for uprobe
496 */
497 __u32 prog_id; /* output: prod_id */
498 __u32 fd_type; /* output: BPF_FD_TYPE_* */
499 __u64 probe_offset; /* output: probe_offset */
500 __u64 probe_addr; /* output: probe_addr */
501 } task_fd_query;
b54ac87e
DB
502} __attribute__((aligned(8)));
503
fd95ec0e
DA
504/* The description below is an attempt at providing documentation to eBPF
505 * developers about the multiple available eBPF helper functions. It can be
506 * parsed and used to produce a manual page. The workflow is the following,
507 * and requires the rst2man utility:
508 *
509 * $ ./scripts/bpf_helpers_doc.py \
510 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
511 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
512 * $ man /tmp/bpf-helpers.7
513 *
514 * Note that in order to produce this external documentation, some RST
515 * formatting is used in the descriptions to get "bold" and "italics" in
516 * manual pages. Also note that the few trailing white spaces are
517 * intentional, removing them would break paragraphs for rst2man.
518 *
519 * Start of BPF helper function descriptions:
520 *
521 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
522 * Description
523 * Perform a lookup in *map* for an entry associated to *key*.
524 * Return
525 * Map value associated to *key*, or **NULL** if no entry was
526 * found.
527 *
528 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
529 * Description
530 * Add or update the value of the entry associated to *key* in
531 * *map* with *value*. *flags* is one of:
532 *
533 * **BPF_NOEXIST**
534 * The entry for *key* must not exist in the map.
535 * **BPF_EXIST**
536 * The entry for *key* must already exist in the map.
537 * **BPF_ANY**
538 * No condition on the existence of the entry for *key*.
539 *
540 * Flag value **BPF_NOEXIST** cannot be used for maps of types
541 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
542 * elements always exist), the helper would return an error.
543 * Return
544 * 0 on success, or a negative error in case of failure.
545 *
546 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
547 * Description
548 * Delete entry with *key* from *map*.
549 * Return
550 * 0 on success, or a negative error in case of failure.
551 *
552 * int bpf_probe_read(void *dst, u32 size, const void *src)
553 * Description
554 * For tracing programs, safely attempt to read *size* bytes from
555 * address *src* and store the data in *dst*.
556 * Return
557 * 0 on success, or a negative error in case of failure.
1a97748b
SH
558 *
559 * u64 bpf_ktime_get_ns(void)
fd95ec0e
DA
560 * Description
561 * Return the time elapsed since system boot, in nanoseconds.
562 * Return
563 * Current *ktime*.
564 *
565 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
566 * Description
567 * This helper is a "printk()-like" facility for debugging. It
568 * prints a message defined by format *fmt* (of size *fmt_size*)
569 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
570 * available. It can take up to three additional **u64**
571 * arguments (as an eBPF helpers, the total number of arguments is
572 * limited to five).
573 *
574 * Each time the helper is called, it appends a line to the trace.
575 * The format of the trace is customizable, and the exact output
576 * one will get depends on the options set in
577 * *\/sys/kernel/debug/tracing/trace_options* (see also the
578 * *README* file under the same directory). However, it usually
579 * defaults to something like:
580 *
581 * ::
582 *
583 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
584 *
585 * In the above:
586 *
587 * * ``telnet`` is the name of the current task.
588 * * ``470`` is the PID of the current task.
589 * * ``001`` is the CPU number on which the task is
590 * running.
591 * * In ``.N..``, each character refers to a set of
592 * options (whether irqs are enabled, scheduling
593 * options, whether hard/softirqs are running, level of
594 * preempt_disabled respectively). **N** means that
595 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
596 * are set.
597 * * ``419421.045894`` is a timestamp.
598 * * ``0x00000001`` is a fake value used by BPF for the
599 * instruction pointer register.
600 * * ``<formatted msg>`` is the message formatted with
601 * *fmt*.
602 *
603 * The conversion specifiers supported by *fmt* are similar, but
604 * more limited than for printk(). They are **%d**, **%i**,
605 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
606 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
607 * of field, padding with zeroes, etc.) is available, and the
608 * helper will return **-EINVAL** (but print nothing) if it
609 * encounters an unknown specifier.
610 *
611 * Also, note that **bpf_trace_printk**\ () is slow, and should
612 * only be used for debugging purposes. For this reason, a notice
613 * bloc (spanning several lines) is printed to kernel logs and
614 * states that the helper should not be used "for production use"
615 * the first time this helper is used (or more precisely, when
616 * **trace_printk**\ () buffers are allocated). For passing values
617 * to user space, perf events should be preferred.
618 * Return
619 * The number of bytes written to the buffer, or a negative error
620 * in case of failure.
621 *
622 * u32 bpf_get_prandom_u32(void)
623 * Description
624 * Get a pseudo-random number.
625 *
626 * From a security point of view, this helper uses its own
627 * pseudo-random internal state, and cannot be used to infer the
628 * seed of other random functions in the kernel. However, it is
629 * essential to note that the generator used by the helper is not
630 * cryptographically secure.
631 * Return
632 * A random 32-bit unsigned value.
633 *
634 * u32 bpf_get_smp_processor_id(void)
635 * Description
636 * Get the SMP (symmetric multiprocessing) processor id. Note that
637 * all programs run with preemption disabled, which means that the
638 * SMP processor id is stable during all the execution of the
639 * program.
640 * Return
641 * The SMP id of the processor running the program.
642 *
643 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
644 * Description
645 * Store *len* bytes from address *from* into the packet
646 * associated to *skb*, at *offset*. *flags* are a combination of
647 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
648 * checksum for the packet after storing the bytes) and
649 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
650 * **->swhash** and *skb*\ **->l4hash** to 0).
651 *
b60ed9a3 652 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
653 * packet buffer. Therefore, at load time, all checks on pointers
654 * previously done by the verifier are invalidated and must be
655 * performed again, if the helper is used in combination with
656 * direct packet access.
657 * Return
658 * 0 on success, or a negative error in case of failure.
659 *
660 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
661 * Description
662 * Recompute the layer 3 (e.g. IP) checksum for the packet
663 * associated to *skb*. Computation is incremental, so the helper
664 * must know the former value of the header field that was
665 * modified (*from*), the new value of this field (*to*), and the
666 * number of bytes (2 or 4) for this field, stored in *size*.
667 * Alternatively, it is possible to store the difference between
668 * the previous and the new values of the header field in *to*, by
669 * setting *from* and *size* to 0. For both methods, *offset*
670 * indicates the location of the IP checksum within the packet.
671 *
672 * This helper works in combination with **bpf_csum_diff**\ (),
673 * which does not update the checksum in-place, but offers more
674 * flexibility and can handle sizes larger than 2 or 4 for the
675 * checksum to update.
676 *
b60ed9a3 677 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
678 * packet buffer. Therefore, at load time, all checks on pointers
679 * previously done by the verifier are invalidated and must be
680 * performed again, if the helper is used in combination with
681 * direct packet access.
682 * Return
683 * 0 on success, or a negative error in case of failure.
684 *
685 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
686 * Description
687 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
688 * packet associated to *skb*. Computation is incremental, so the
689 * helper must know the former value of the header field that was
690 * modified (*from*), the new value of this field (*to*), and the
691 * number of bytes (2 or 4) for this field, stored on the lowest
692 * four bits of *flags*. Alternatively, it is possible to store
693 * the difference between the previous and the new values of the
694 * header field in *to*, by setting *from* and the four lowest
695 * bits of *flags* to 0. For both methods, *offset* indicates the
696 * location of the IP checksum within the packet. In addition to
697 * the size of the field, *flags* can be added (bitwise OR) actual
698 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
699 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
700 * for updates resulting in a null checksum the value is set to
701 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
702 * the checksum is to be computed against a pseudo-header.
703 *
704 * This helper works in combination with **bpf_csum_diff**\ (),
705 * which does not update the checksum in-place, but offers more
706 * flexibility and can handle sizes larger than 2 or 4 for the
707 * checksum to update.
708 *
b60ed9a3 709 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
710 * packet buffer. Therefore, at load time, all checks on pointers
711 * previously done by the verifier are invalidated and must be
712 * performed again, if the helper is used in combination with
713 * direct packet access.
714 * Return
715 * 0 on success, or a negative error in case of failure.
716 *
717 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
718 * Description
719 * This special helper is used to trigger a "tail call", or in
720 * other words, to jump into another eBPF program. The same stack
721 * frame is used (but values on stack and in registers for the
722 * caller are not accessible to the callee). This mechanism allows
723 * for program chaining, either for raising the maximum number of
724 * available eBPF instructions, or to execute given programs in
725 * conditional blocks. For security reasons, there is an upper
726 * limit to the number of successive tail calls that can be
727 * performed.
728 *
729 * Upon call of this helper, the program attempts to jump into a
730 * program referenced at index *index* in *prog_array_map*, a
731 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
732 * *ctx*, a pointer to the context.
733 *
734 * If the call succeeds, the kernel immediately runs the first
735 * instruction of the new program. This is not a function call,
736 * and it never returns to the previous program. If the call
737 * fails, then the helper has no effect, and the caller continues
738 * to run its subsequent instructions. A call can fail if the
739 * destination program for the jump does not exist (i.e. *index*
740 * is superior to the number of entries in *prog_array_map*), or
741 * if the maximum number of tail calls has been reached for this
742 * chain of programs. This limit is defined in the kernel by the
743 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
744 * which is currently set to 32.
745 * Return
746 * 0 on success, or a negative error in case of failure.
747 *
748 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
749 * Description
750 * Clone and redirect the packet associated to *skb* to another
751 * net device of index *ifindex*. Both ingress and egress
752 * interfaces can be used for redirection. The **BPF_F_INGRESS**
753 * value in *flags* is used to make the distinction (ingress path
754 * is selected if the flag is present, egress path otherwise).
755 * This is the only flag supported for now.
756 *
757 * In comparison with **bpf_redirect**\ () helper,
758 * **bpf_clone_redirect**\ () has the associated cost of
759 * duplicating the packet buffer, but this can be executed out of
760 * the eBPF program. Conversely, **bpf_redirect**\ () is more
761 * efficient, but it is handled through an action code where the
762 * redirection happens only after the eBPF program has returned.
763 *
b60ed9a3 764 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
765 * packet buffer. Therefore, at load time, all checks on pointers
766 * previously done by the verifier are invalidated and must be
767 * performed again, if the helper is used in combination with
768 * direct packet access.
769 * Return
770 * 0 on success, or a negative error in case of failure.
1a97748b
SH
771 *
772 * u64 bpf_get_current_pid_tgid(void)
fd95ec0e
DA
773 * Return
774 * A 64-bit integer containing the current tgid and pid, and
775 * created as such:
776 * *current_task*\ **->tgid << 32 \|**
777 * *current_task*\ **->pid**.
1a97748b
SH
778 *
779 * u64 bpf_get_current_uid_gid(void)
fd95ec0e
DA
780 * Return
781 * A 64-bit integer containing the current GID and UID, and
782 * created as such: *current_gid* **<< 32 \|** *current_uid*.
783 *
784 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
785 * Description
786 * Copy the **comm** attribute of the current task into *buf* of
787 * *size_of_buf*. The **comm** attribute contains the name of
788 * the executable (excluding the path) for the current task. The
789 * *size_of_buf* must be strictly positive. On success, the
790 * helper makes sure that the *buf* is NUL-terminated. On failure,
791 * it is filled with zeroes.
792 * Return
793 * 0 on success, or a negative error in case of failure.
794 *
795 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
796 * Description
797 * Retrieve the classid for the current task, i.e. for the net_cls
798 * cgroup to which *skb* belongs.
799 *
800 * This helper can be used on TC egress path, but not on ingress.
801 *
802 * The net_cls cgroup provides an interface to tag network packets
803 * based on a user-provided identifier for all traffic coming from
804 * the tasks belonging to the related cgroup. See also the related
805 * kernel documentation, available from the Linux sources in file
806 * *Documentation/cgroup-v1/net_cls.txt*.
807 *
808 * The Linux kernel has two versions for cgroups: there are
809 * cgroups v1 and cgroups v2. Both are available to users, who can
810 * use a mixture of them, but note that the net_cls cgroup is for
811 * cgroup v1 only. This makes it incompatible with BPF programs
812 * run on cgroups, which is a cgroup-v2-only feature (a socket can
813 * only hold data for one version of cgroups at a time).
814 *
815 * This helper is only available is the kernel was compiled with
816 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
817 * "**y**" or to "**m**".
818 * Return
819 * The classid, or 0 for the default unconfigured classid.
820 *
821 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
822 * Description
823 * Push a *vlan_tci* (VLAN tag control information) of protocol
824 * *vlan_proto* to the packet associated to *skb*, then update
825 * the checksum. Note that if *vlan_proto* is different from
826 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
827 * be **ETH_P_8021Q**.
828 *
b60ed9a3 829 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
830 * packet buffer. Therefore, at load time, all checks on pointers
831 * previously done by the verifier are invalidated and must be
832 * performed again, if the helper is used in combination with
833 * direct packet access.
834 * Return
835 * 0 on success, or a negative error in case of failure.
836 *
837 * int bpf_skb_vlan_pop(struct sk_buff *skb)
838 * Description
839 * Pop a VLAN header from the packet associated to *skb*.
840 *
b60ed9a3 841 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
842 * packet buffer. Therefore, at load time, all checks on pointers
843 * previously done by the verifier are invalidated and must be
844 * performed again, if the helper is used in combination with
845 * direct packet access.
846 * Return
847 * 0 on success, or a negative error in case of failure.
848 *
849 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
850 * Description
851 * Get tunnel metadata. This helper takes a pointer *key* to an
852 * empty **struct bpf_tunnel_key** of **size**, that will be
853 * filled with tunnel metadata for the packet associated to *skb*.
854 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
855 * indicates that the tunnel is based on IPv6 protocol instead of
856 * IPv4.
857 *
858 * The **struct bpf_tunnel_key** is an object that generalizes the
859 * principal parameters used by various tunneling protocols into a
860 * single struct. This way, it can be used to easily make a
861 * decision based on the contents of the encapsulation header,
862 * "summarized" in this struct. In particular, it holds the IP
863 * address of the remote end (IPv4 or IPv6, depending on the case)
864 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
865 * this struct exposes the *key*\ **->tunnel_id**, which is
866 * generally mapped to a VNI (Virtual Network Identifier), making
867 * it programmable together with the **bpf_skb_set_tunnel_key**\
868 * () helper.
869 *
870 * Let's imagine that the following code is part of a program
871 * attached to the TC ingress interface, on one end of a GRE
872 * tunnel, and is supposed to filter out all messages coming from
873 * remote ends with IPv4 address other than 10.0.0.1:
874 *
875 * ::
876 *
877 * int ret;
878 * struct bpf_tunnel_key key = {};
879 *
880 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
881 * if (ret < 0)
882 * return TC_ACT_SHOT; // drop packet
883 *
884 * if (key.remote_ipv4 != 0x0a000001)
885 * return TC_ACT_SHOT; // drop packet
886 *
887 * return TC_ACT_OK; // accept packet
888 *
889 * This interface can also be used with all encapsulation devices
890 * that can operate in "collect metadata" mode: instead of having
891 * one network device per specific configuration, the "collect
892 * metadata" mode only requires a single device where the
893 * configuration can be extracted from this helper.
894 *
895 * This can be used together with various tunnels such as VXLan,
896 * Geneve, GRE or IP in IP (IPIP).
897 * Return
898 * 0 on success, or a negative error in case of failure.
899 *
900 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
901 * Description
902 * Populate tunnel metadata for packet associated to *skb.* The
903 * tunnel metadata is set to the contents of *key*, of *size*. The
904 * *flags* can be set to a combination of the following values:
905 *
906 * **BPF_F_TUNINFO_IPV6**
907 * Indicate that the tunnel is based on IPv6 protocol
908 * instead of IPv4.
909 * **BPF_F_ZERO_CSUM_TX**
910 * For IPv4 packets, add a flag to tunnel metadata
911 * indicating that checksum computation should be skipped
912 * and checksum set to zeroes.
913 * **BPF_F_DONT_FRAGMENT**
914 * Add a flag to tunnel metadata indicating that the
915 * packet should not be fragmented.
916 * **BPF_F_SEQ_NUMBER**
917 * Add a flag to tunnel metadata indicating that a
918 * sequence number should be added to tunnel header before
919 * sending the packet. This flag was added for GRE
920 * encapsulation, but might be used with other protocols
921 * as well in the future.
922 *
923 * Here is a typical usage on the transmit path:
924 *
925 * ::
926 *
927 * struct bpf_tunnel_key key;
928 * populate key ...
929 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
930 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
931 *
932 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
933 * helper for additional information.
934 * Return
935 * 0 on success, or a negative error in case of failure.
936 *
937 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
938 * Description
939 * Read the value of a perf event counter. This helper relies on a
940 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
941 * the perf event counter is selected when *map* is updated with
942 * perf event file descriptors. The *map* is an array whose size
943 * is the number of available CPUs, and each cell contains a value
944 * relative to one CPU. The value to retrieve is indicated by
945 * *flags*, that contains the index of the CPU to look up, masked
946 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
947 * **BPF_F_CURRENT_CPU** to indicate that the value for the
948 * current CPU should be retrieved.
949 *
950 * Note that before Linux 4.13, only hardware perf event can be
951 * retrieved.
952 *
953 * Also, be aware that the newer helper
954 * **bpf_perf_event_read_value**\ () is recommended over
955 * **bpf_perf_event_read**\ () in general. The latter has some ABI
956 * quirks where error and counter value are used as a return code
957 * (which is wrong to do since ranges may overlap). This issue is
958 * fixed with **bpf_perf_event_read_value**\ (), which at the same
959 * time provides more features over the **bpf_perf_event_read**\
960 * () interface. Please refer to the description of
961 * **bpf_perf_event_read_value**\ () for details.
962 * Return
963 * The value of the perf event counter read from the map, or a
964 * negative error code in case of failure.
965 *
966 * int bpf_redirect(u32 ifindex, u64 flags)
967 * Description
968 * Redirect the packet to another net device of index *ifindex*.
969 * This helper is somewhat similar to **bpf_clone_redirect**\
970 * (), except that the packet is not cloned, which provides
971 * increased performance.
972 *
973 * Except for XDP, both ingress and egress interfaces can be used
974 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
975 * to make the distinction (ingress path is selected if the flag
976 * is present, egress path otherwise). Currently, XDP only
977 * supports redirection to the egress interface, and accepts no
978 * flag at all.
979 *
980 * The same effect can be attained with the more generic
981 * **bpf_redirect_map**\ (), which requires specific maps to be
982 * used but offers better performance.
983 * Return
984 * For XDP, the helper returns **XDP_REDIRECT** on success or
985 * **XDP_ABORTED** on error. For other program types, the values
986 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
987 * error.
988 *
989 * u32 bpf_get_route_realm(struct sk_buff *skb)
990 * Description
991 * Retrieve the realm or the route, that is to say the
992 * **tclassid** field of the destination for the *skb*. The
993 * indentifier retrieved is a user-provided tag, similar to the
994 * one used with the net_cls cgroup (see description for
995 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
996 * held by a route (a destination entry), not by a task.
997 *
998 * Retrieving this identifier works with the clsact TC egress hook
999 * (see also **tc-bpf(8)**), or alternatively on conventional
1000 * classful egress qdiscs, but not on TC ingress path. In case of
1001 * clsact TC egress hook, this has the advantage that, internally,
1002 * the destination entry has not been dropped yet in the transmit
1003 * path. Therefore, the destination entry does not need to be
1004 * artificially held via **netif_keep_dst**\ () for a classful
1005 * qdisc until the *skb* is freed.
1006 *
1007 * This helper is available only if the kernel was compiled with
1008 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1009 * Return
1010 * The realm of the route for the packet associated to *skb*, or 0
1011 * if none was found.
1012 *
1013 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1014 * Description
1015 * Write raw *data* blob into a special BPF perf event held by
1016 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1017 * event must have the following attributes: **PERF_SAMPLE_RAW**
1018 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1019 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1020 *
1021 * The *flags* are used to indicate the index in *map* for which
1022 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1023 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1024 * to indicate that the index of the current CPU core should be
1025 * used.
1026 *
1027 * The value to write, of *size*, is passed through eBPF stack and
1028 * pointed by *data*.
1029 *
1030 * The context of the program *ctx* needs also be passed to the
1031 * helper.
1032 *
1033 * On user space, a program willing to read the values needs to
1034 * call **perf_event_open**\ () on the perf event (either for
1035 * one or for all CPUs) and to store the file descriptor into the
1036 * *map*. This must be done before the eBPF program can send data
1037 * into it. An example is available in file
1038 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1039 * tree (the eBPF program counterpart is in
1040 * *samples/bpf/trace_output_kern.c*).
1041 *
1042 * **bpf_perf_event_output**\ () achieves better performance
1043 * than **bpf_trace_printk**\ () for sharing data with user
1044 * space, and is much better suitable for streaming data from eBPF
1045 * programs.
1046 *
1047 * Note that this helper is not restricted to tracing use cases
1048 * and can be used with programs attached to TC or XDP as well,
1049 * where it allows for passing data to user space listeners. Data
1050 * can be:
1051 *
1052 * * Only custom structs,
1053 * * Only the packet payload, or
1054 * * A combination of both.
1055 * Return
1056 * 0 on success, or a negative error in case of failure.
1057 *
1058 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1059 * Description
1060 * This helper was provided as an easy way to load data from a
1061 * packet. It can be used to load *len* bytes from *offset* from
1062 * the packet associated to *skb*, into the buffer pointed by
1063 * *to*.
1064 *
1065 * Since Linux 4.7, usage of this helper has mostly been replaced
1066 * by "direct packet access", enabling packet data to be
1067 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1068 * pointing respectively to the first byte of packet data and to
1069 * the byte after the last byte of packet data. However, it
1070 * remains useful if one wishes to read large quantities of data
1071 * at once from a packet into the eBPF stack.
1072 * Return
1073 * 0 on success, or a negative error in case of failure.
1074 *
1075 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1076 * Description
1077 * Walk a user or a kernel stack and return its id. To achieve
1078 * this, the helper needs *ctx*, which is a pointer to the context
1079 * on which the tracing program is executed, and a pointer to a
1080 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1081 *
1082 * The last argument, *flags*, holds the number of stack frames to
1083 * skip (from 0 to 255), masked with
1084 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1085 * a combination of the following flags:
1086 *
1087 * **BPF_F_USER_STACK**
1088 * Collect a user space stack instead of a kernel stack.
1089 * **BPF_F_FAST_STACK_CMP**
1090 * Compare stacks by hash only.
1091 * **BPF_F_REUSE_STACKID**
1092 * If two different stacks hash into the same *stackid*,
1093 * discard the old one.
1094 *
1095 * The stack id retrieved is a 32 bit long integer handle which
1096 * can be further combined with other data (including other stack
1097 * ids) and used as a key into maps. This can be useful for
1098 * generating a variety of graphs (such as flame graphs or off-cpu
1099 * graphs).
1100 *
1101 * For walking a stack, this helper is an improvement over
1102 * **bpf_probe_read**\ (), which can be used with unrolled loops
1103 * but is not efficient and consumes a lot of eBPF instructions.
1104 * Instead, **bpf_get_stackid**\ () can collect up to
1105 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1106 * this limit can be controlled with the **sysctl** program, and
1107 * that it should be manually increased in order to profile long
1108 * user stacks (such as stacks for Java programs). To do so, use:
1109 *
1110 * ::
1111 *
1112 * # sysctl kernel.perf_event_max_stack=<new value>
fd95ec0e
DA
1113 * Return
1114 * The positive or null stack id on success, or a negative error
1115 * in case of failure.
1116 *
1117 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1118 * Description
1119 * Compute a checksum difference, from the raw buffer pointed by
1120 * *from*, of length *from_size* (that must be a multiple of 4),
1121 * towards the raw buffer pointed by *to*, of size *to_size*
1122 * (same remark). An optional *seed* can be added to the value
1123 * (this can be cascaded, the seed may come from a previous call
1124 * to the helper).
1125 *
1126 * This is flexible enough to be used in several ways:
1127 *
1128 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1129 * checksum, it can be used when pushing new data.
1130 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1131 * checksum, it can be used when removing data from a packet.
1132 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1133 * can be used to compute a diff. Note that *from_size* and
1134 * *to_size* do not need to be equal.
1135 *
1136 * This helper can be used in combination with
1137 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1138 * which one can feed in the difference computed with
1139 * **bpf_csum_diff**\ ().
1140 * Return
1141 * The checksum result, or a negative error code in case of
1142 * failure.
1143 *
1144 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1145 * Description
1146 * Retrieve tunnel options metadata for the packet associated to
1147 * *skb*, and store the raw tunnel option data to the buffer *opt*
1148 * of *size*.
1149 *
1150 * This helper can be used with encapsulation devices that can
1151 * operate in "collect metadata" mode (please refer to the related
1152 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1153 * more details). A particular example where this can be used is
1154 * in combination with the Geneve encapsulation protocol, where it
1155 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1156 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1157 * the eBPF program. This allows for full customization of these
1158 * headers.
1159 * Return
1160 * The size of the option data retrieved.
1161 *
1162 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1163 * Description
1164 * Set tunnel options metadata for the packet associated to *skb*
1165 * to the option data contained in the raw buffer *opt* of *size*.
1166 *
1167 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1168 * helper for additional information.
1169 * Return
1170 * 0 on success, or a negative error in case of failure.
1171 *
1172 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1173 * Description
1174 * Change the protocol of the *skb* to *proto*. Currently
1175 * supported are transition from IPv4 to IPv6, and from IPv6 to
1176 * IPv4. The helper takes care of the groundwork for the
1177 * transition, including resizing the socket buffer. The eBPF
1178 * program is expected to fill the new headers, if any, via
1179 * **skb_store_bytes**\ () and to recompute the checksums with
1180 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1181 * (). The main case for this helper is to perform NAT64
1182 * operations out of an eBPF program.
1183 *
1184 * Internally, the GSO type is marked as dodgy so that headers are
1185 * checked and segments are recalculated by the GSO/GRO engine.
1186 * The size for GSO target is adapted as well.
1187 *
1188 * All values for *flags* are reserved for future usage, and must
1189 * be left at zero.
1190 *
b60ed9a3 1191 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1192 * packet buffer. Therefore, at load time, all checks on pointers
1193 * previously done by the verifier are invalidated and must be
1194 * performed again, if the helper is used in combination with
1195 * direct packet access.
1196 * Return
1197 * 0 on success, or a negative error in case of failure.
1198 *
1199 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1200 * Description
1201 * Change the packet type for the packet associated to *skb*. This
1202 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1203 * the eBPF program does not have a write access to *skb*\
1204 * **->pkt_type** beside this helper. Using a helper here allows
1205 * for graceful handling of errors.
1206 *
1207 * The major use case is to change incoming *skb*s to
1208 * **PACKET_HOST** in a programmatic way instead of having to
1209 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1210 * example.
1211 *
1212 * Note that *type* only allows certain values. At this time, they
1213 * are:
1214 *
1215 * **PACKET_HOST**
1216 * Packet is for us.
1217 * **PACKET_BROADCAST**
1218 * Send packet to all.
1219 * **PACKET_MULTICAST**
1220 * Send packet to group.
1221 * **PACKET_OTHERHOST**
1222 * Send packet to someone else.
1223 * Return
1224 * 0 on success, or a negative error in case of failure.
1225 *
1226 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1227 * Description
1228 * Check whether *skb* is a descendant of the cgroup2 held by
1229 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1230 * Return
1231 * The return value depends on the result of the test, and can be:
1232 *
1233 * * 0, if the *skb* failed the cgroup2 descendant test.
1234 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1235 * * A negative error code, if an error occurred.
1236 *
1237 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1238 * Description
1239 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1240 * not set, in particular if the hash was cleared due to mangling,
1241 * recompute this hash. Later accesses to the hash can be done
1242 * directly with *skb*\ **->hash**.
1243 *
1244 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1245 * prototype with **bpf_skb_change_proto**\ (), or calling
1246 * **bpf_skb_store_bytes**\ () with the
1247 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1248 * the hash and to trigger a new computation for the next call to
1249 * **bpf_get_hash_recalc**\ ().
1250 * Return
1251 * The 32-bit hash.
1a97748b
SH
1252 *
1253 * u64 bpf_get_current_task(void)
fd95ec0e
DA
1254 * Return
1255 * A pointer to the current task struct.
1256 *
1257 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1258 * Description
1259 * Attempt in a safe way to write *len* bytes from the buffer
1260 * *src* to *dst* in memory. It only works for threads that are in
1261 * user context, and *dst* must be a valid user space address.
1262 *
1263 * This helper should not be used to implement any kind of
1264 * security mechanism because of TOC-TOU attacks, but rather to
1265 * debug, divert, and manipulate execution of semi-cooperative
1266 * processes.
1267 *
1268 * Keep in mind that this feature is meant for experiments, and it
1269 * has a risk of crashing the system and running programs.
1270 * Therefore, when an eBPF program using this helper is attached,
1271 * a warning including PID and process name is printed to kernel
1272 * logs.
1273 * Return
1274 * 0 on success, or a negative error in case of failure.
1275 *
1276 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1277 * Description
1278 * Check whether the probe is being run is the context of a given
1279 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1280 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1281 * Return
1282 * The return value depends on the result of the test, and can be:
1283 *
1284 * * 0, if the *skb* task belongs to the cgroup2.
1285 * * 1, if the *skb* task does not belong to the cgroup2.
1286 * * A negative error code, if an error occurred.
1287 *
1288 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1289 * Description
1290 * Resize (trim or grow) the packet associated to *skb* to the
1291 * new *len*. The *flags* are reserved for future usage, and must
1292 * be left at zero.
1293 *
1294 * The basic idea is that the helper performs the needed work to
1295 * change the size of the packet, then the eBPF program rewrites
1296 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1297 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1298 * and others. This helper is a slow path utility intended for
1299 * replies with control messages. And because it is targeted for
1300 * slow path, the helper itself can afford to be slow: it
1301 * implicitly linearizes, unclones and drops offloads from the
1302 * *skb*.
1303 *
b60ed9a3 1304 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1305 * packet buffer. Therefore, at load time, all checks on pointers
1306 * previously done by the verifier are invalidated and must be
1307 * performed again, if the helper is used in combination with
1308 * direct packet access.
1309 * Return
1310 * 0 on success, or a negative error in case of failure.
1311 *
1312 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1313 * Description
1314 * Pull in non-linear data in case the *skb* is non-linear and not
1315 * all of *len* are part of the linear section. Make *len* bytes
1316 * from *skb* readable and writable. If a zero value is passed for
1317 * *len*, then the whole length of the *skb* is pulled.
1318 *
1319 * This helper is only needed for reading and writing with direct
1320 * packet access.
1321 *
1322 * For direct packet access, testing that offsets to access
1323 * are within packet boundaries (test on *skb*\ **->data_end**) is
1324 * susceptible to fail if offsets are invalid, or if the requested
1325 * data is in non-linear parts of the *skb*. On failure the
1326 * program can just bail out, or in the case of a non-linear
1327 * buffer, use a helper to make the data available. The
1328 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1329 * the data. Another one consists in using **bpf_skb_pull_data**
1330 * to pull in once the non-linear parts, then retesting and
1331 * eventually access the data.
1332 *
1333 * At the same time, this also makes sure the *skb* is uncloned,
1334 * which is a necessary condition for direct write. As this needs
1335 * to be an invariant for the write part only, the verifier
1336 * detects writes and adds a prologue that is calling
1337 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1338 * the very beginning in case it is indeed cloned.
1339 *
b60ed9a3 1340 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1341 * packet buffer. Therefore, at load time, all checks on pointers
1342 * previously done by the verifier are invalidated and must be
1343 * performed again, if the helper is used in combination with
1344 * direct packet access.
1345 * Return
1346 * 0 on success, or a negative error in case of failure.
1347 *
1348 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1349 * Description
1350 * Add the checksum *csum* into *skb*\ **->csum** in case the
1351 * driver has supplied a checksum for the entire packet into that
1352 * field. Return an error otherwise. This helper is intended to be
1353 * used in combination with **bpf_csum_diff**\ (), in particular
1354 * when the checksum needs to be updated after data has been
1355 * written into the packet through direct packet access.
1356 * Return
1357 * The checksum on success, or a negative error code in case of
1358 * failure.
1359 *
1360 * void bpf_set_hash_invalid(struct sk_buff *skb)
1361 * Description
1362 * Invalidate the current *skb*\ **->hash**. It can be used after
1363 * mangling on headers through direct packet access, in order to
1364 * indicate that the hash is outdated and to trigger a
1365 * recalculation the next time the kernel tries to access this
1366 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1367 *
1368 * int bpf_get_numa_node_id(void)
1369 * Description
1370 * Return the id of the current NUMA node. The primary use case
1371 * for this helper is the selection of sockets for the local NUMA
1372 * node, when the program is attached to sockets using the
1373 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1374 * but the helper is also available to other eBPF program types,
1375 * similarly to **bpf_get_smp_processor_id**\ ().
1376 * Return
1377 * The id of current NUMA node.
1378 *
1379 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1380 * Description
1381 * Grows headroom of packet associated to *skb* and adjusts the
1382 * offset of the MAC header accordingly, adding *len* bytes of
1383 * space. It automatically extends and reallocates memory as
1384 * required.
1385 *
1386 * This helper can be used on a layer 3 *skb* to push a MAC header
1387 * for redirection into a layer 2 device.
1388 *
1389 * All values for *flags* are reserved for future usage, and must
1390 * be left at zero.
1391 *
b60ed9a3 1392 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1393 * packet buffer. Therefore, at load time, all checks on pointers
1394 * previously done by the verifier are invalidated and must be
1395 * performed again, if the helper is used in combination with
1396 * direct packet access.
1397 * Return
1398 * 0 on success, or a negative error in case of failure.
1399 *
1400 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1401 * Description
1402 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1403 * it is possible to use a negative value for *delta*. This helper
1404 * can be used to prepare the packet for pushing or popping
1405 * headers.
1406 *
b60ed9a3 1407 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1408 * packet buffer. Therefore, at load time, all checks on pointers
1409 * previously done by the verifier are invalidated and must be
1410 * performed again, if the helper is used in combination with
1411 * direct packet access.
1412 * Return
1413 * 0 on success, or a negative error in case of failure.
a044b36a
SH
1414 *
1415 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
fd95ec0e
DA
1416 * Description
1417 * Copy a NUL terminated string from an unsafe address
1418 * *unsafe_ptr* to *dst*. The *size* should include the
1419 * terminating NUL byte. In case the string length is smaller than
1420 * *size*, the target is not padded with further NUL bytes. If the
1421 * string length is larger than *size*, just *size*-1 bytes are
1422 * copied and the last byte is set to NUL.
1423 *
1424 * On success, the length of the copied string is returned. This
1425 * makes this helper useful in tracing programs for reading
1426 * strings, and more importantly to get its length at runtime. See
1427 * the following snippet:
1428 *
1429 * ::
1430 *
1431 * SEC("kprobe/sys_open")
1432 * void bpf_sys_open(struct pt_regs *ctx)
1433 * {
1434 * char buf[PATHLEN]; // PATHLEN is defined to 256
1435 * int res = bpf_probe_read_str(buf, sizeof(buf),
1436 * ctx->di);
1437 *
1438 * // Consume buf, for example push it to
1439 * // userspace via bpf_perf_event_output(); we
1440 * // can use res (the string length) as event
1441 * // size, after checking its boundaries.
1442 * }
1443 *
1444 * In comparison, using **bpf_probe_read()** helper here instead
1445 * to read the string would require to estimate the length at
1446 * compile time, and would often result in copying more memory
1447 * than necessary.
1448 *
1449 * Another useful use case is when parsing individual process
1450 * arguments or individual environment variables navigating
1451 * *current*\ **->mm->arg_start** and *current*\
1452 * **->mm->env_start**: using this helper and the return value,
1453 * one can quickly iterate at the right offset of the memory area.
1454 * Return
1455 * On success, the strictly positive length of the string,
1456 * including the trailing NUL character. On error, a negative
1457 * value.
1458 *
1459 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1460 * Description
1461 * If the **struct sk_buff** pointed by *skb* has a known socket,
1462 * retrieve the cookie (generated by the kernel) of this socket.
1463 * If no cookie has been set yet, generate a new cookie. Once
1464 * generated, the socket cookie remains stable for the life of the
1465 * socket. This helper can be useful for monitoring per socket
1466 * networking traffic statistics as it provides a unique socket
1467 * identifier per namespace.
1468 * Return
1469 * A 8-byte long non-decreasing number on success, or 0 if the
1470 * socket field is missing inside *skb*.
1471 *
74eb09ad
DA
1472 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1473 * Description
1474 * Equivalent to bpf_get_socket_cookie() helper that accepts
cdeb2674 1475 * *skb*, but gets socket from **struct bpf_sock_addr** context.
74eb09ad
DA
1476 * Return
1477 * A 8-byte long non-decreasing number.
1478 *
1479 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1480 * Description
1481 * Equivalent to bpf_get_socket_cookie() helper that accepts
cdeb2674 1482 * *skb*, but gets socket from **struct bpf_sock_ops** context.
74eb09ad
DA
1483 * Return
1484 * A 8-byte long non-decreasing number.
1485 *
fd95ec0e
DA
1486 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1487 * Return
1488 * The owner UID of the socket associated to *skb*. If the socket
1489 * is **NULL**, or if it is not a full socket (i.e. if it is a
1490 * time-wait or a request socket instead), **overflowuid** value
1491 * is returned (note that **overflowuid** might also be the actual
1492 * UID value for the socket).
1493 *
1494 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1495 * Description
1496 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1497 * to value *hash*.
1498 * Return
1499 * 0
1500 *
1501 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1502 * Description
1503 * Emulate a call to **setsockopt()** on the socket associated to
1504 * *bpf_socket*, which must be a full socket. The *level* at
1505 * which the option resides and the name *optname* of the option
1506 * must be specified, see **setsockopt(2)** for more information.
1507 * The option value of length *optlen* is pointed by *optval*.
1508 *
1509 * This helper actually implements a subset of **setsockopt()**.
1510 * It supports the following *level*\ s:
1511 *
1512 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1513 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1514 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1515 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1516 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1517 * **TCP_BPF_SNDCWND_CLAMP**.
1518 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1519 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1520 * Return
1521 * 0 on success, or a negative error in case of failure.
1522 *
39776a86 1523 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
fd95ec0e
DA
1524 * Description
1525 * Grow or shrink the room for data in the packet associated to
1526 * *skb* by *len_diff*, and according to the selected *mode*.
1527 *
188c7fe6
DA
1528 * There are two supported modes at this time:
1529 *
1530 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
1531 * (room space is added or removed below the layer 2 header).
fd95ec0e
DA
1532 *
1533 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1534 * (room space is added or removed below the layer 3 header).
1535 *
188c7fe6
DA
1536 * The following flags are supported at this time:
1537 *
1538 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
1539 * Adjusting mss in this way is not allowed for datagrams.
1540 *
b60ed9a3
SH
1541 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
1542 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
188c7fe6
DA
1543 * Any new space is reserved to hold a tunnel header.
1544 * Configure skb offsets and other fields accordingly.
1545 *
b60ed9a3
SH
1546 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
1547 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
188c7fe6
DA
1548 * Use with ENCAP_L3 flags to further specify the tunnel type.
1549 *
b60ed9a3 1550 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
188c7fe6 1551 * Use with ENCAP_L3/L4 flags to further specify the tunnel
b60ed9a3 1552 * type; *len* is the length of the inner MAC header.
fd95ec0e 1553 *
b60ed9a3 1554 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1555 * packet buffer. Therefore, at load time, all checks on pointers
1556 * previously done by the verifier are invalidated and must be
1557 * performed again, if the helper is used in combination with
1558 * direct packet access.
1559 * Return
1560 * 0 on success, or a negative error in case of failure.
1561 *
1562 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1563 * Description
1564 * Redirect the packet to the endpoint referenced by *map* at
1565 * index *key*. Depending on its type, this *map* can contain
1566 * references to net devices (for forwarding packets through other
1567 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1568 * but this is only implemented for native XDP (with driver
1569 * support) as of this writing).
1570 *
1571 * All values for *flags* are reserved for future usage, and must
1572 * be left at zero.
1573 *
1574 * When used to redirect packets to net devices, this helper
1575 * provides a high performance increase over **bpf_redirect**\ ().
1576 * This is due to various implementation details of the underlying
1577 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1578 * () tries to send packet as a "bulk" to the device.
1579 * Return
1580 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1581 *
1582 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1583 * Description
1584 * Redirect the packet to the socket referenced by *map* (of type
1585 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1586 * egress interfaces can be used for redirection. The
1587 * **BPF_F_INGRESS** value in *flags* is used to make the
1588 * distinction (ingress path is selected if the flag is present,
1589 * egress path otherwise). This is the only flag supported for now.
1590 * Return
1591 * **SK_PASS** on success, or **SK_DROP** on error.
1592 *
1593 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1594 * Description
1595 * Add an entry to, or update a *map* referencing sockets. The
1596 * *skops* is used as a new value for the entry associated to
1597 * *key*. *flags* is one of:
1598 *
1599 * **BPF_NOEXIST**
1600 * The entry for *key* must not exist in the map.
1601 * **BPF_EXIST**
1602 * The entry for *key* must already exist in the map.
1603 * **BPF_ANY**
1604 * No condition on the existence of the entry for *key*.
1605 *
1606 * If the *map* has eBPF programs (parser and verdict), those will
1607 * be inherited by the socket being added. If the socket is
1608 * already attached to eBPF programs, this results in an error.
1609 * Return
1610 * 0 on success, or a negative error in case of failure.
1611 *
1612 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1613 * Description
1614 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1615 * *delta* (which can be positive or negative). Note that this
1616 * operation modifies the address stored in *xdp_md*\ **->data**,
1617 * so the latter must be loaded only after the helper has been
1618 * called.
1619 *
1620 * The use of *xdp_md*\ **->data_meta** is optional and programs
1621 * are not required to use it. The rationale is that when the
1622 * packet is processed with XDP (e.g. as DoS filter), it is
1623 * possible to push further meta data along with it before passing
1624 * to the stack, and to give the guarantee that an ingress eBPF
1625 * program attached as a TC classifier on the same device can pick
1626 * this up for further post-processing. Since TC works with socket
1627 * buffers, it remains possible to set from XDP the **mark** or
1628 * **priority** pointers, or other pointers for the socket buffer.
1629 * Having this scratch space generic and programmable allows for
1630 * more flexibility as the user is free to store whatever meta
1631 * data they need.
1632 *
b60ed9a3 1633 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1634 * packet buffer. Therefore, at load time, all checks on pointers
1635 * previously done by the verifier are invalidated and must be
1636 * performed again, if the helper is used in combination with
1637 * direct packet access.
1638 * Return
1639 * 0 on success, or a negative error in case of failure.
1640 *
1641 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1642 * Description
1643 * Read the value of a perf event counter, and store it into *buf*
1644 * of size *buf_size*. This helper relies on a *map* of type
1645 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1646 * counter is selected when *map* is updated with perf event file
1647 * descriptors. The *map* is an array whose size is the number of
1648 * available CPUs, and each cell contains a value relative to one
1649 * CPU. The value to retrieve is indicated by *flags*, that
1650 * contains the index of the CPU to look up, masked with
1651 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1652 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1653 * current CPU should be retrieved.
1654 *
1655 * This helper behaves in a way close to
1656 * **bpf_perf_event_read**\ () helper, save that instead of
1657 * just returning the value observed, it fills the *buf*
1658 * structure. This allows for additional data to be retrieved: in
1659 * particular, the enabled and running times (in *buf*\
1660 * **->enabled** and *buf*\ **->running**, respectively) are
1661 * copied. In general, **bpf_perf_event_read_value**\ () is
1662 * recommended over **bpf_perf_event_read**\ (), which has some
1663 * ABI issues and provides fewer functionalities.
1664 *
1665 * These values are interesting, because hardware PMU (Performance
1666 * Monitoring Unit) counters are limited resources. When there are
1667 * more PMU based perf events opened than available counters,
1668 * kernel will multiplex these events so each event gets certain
1669 * percentage (but not all) of the PMU time. In case that
1670 * multiplexing happens, the number of samples or counter value
1671 * will not reflect the case compared to when no multiplexing
1672 * occurs. This makes comparison between different runs difficult.
1673 * Typically, the counter value should be normalized before
1674 * comparing to other experiments. The usual normalization is done
1675 * as follows.
1676 *
1677 * ::
1678 *
1679 * normalized_counter = counter * t_enabled / t_running
1680 *
1681 * Where t_enabled is the time enabled for event and t_running is
1682 * the time running for event since last normalization. The
1683 * enabled and running times are accumulated since the perf event
1684 * open. To achieve scaling factor between two invocations of an
1685 * eBPF program, users can can use CPU id as the key (which is
1686 * typical for perf array usage model) to remember the previous
1687 * value and do the calculation inside the eBPF program.
1688 * Return
1689 * 0 on success, or a negative error in case of failure.
1690 *
1691 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1692 * Description
1693 * For en eBPF program attached to a perf event, retrieve the
1694 * value of the event counter associated to *ctx* and store it in
1695 * the structure pointed by *buf* and of size *buf_size*. Enabled
1696 * and running times are also stored in the structure (see
1697 * description of helper **bpf_perf_event_read_value**\ () for
1698 * more details).
1699 * Return
1700 * 0 on success, or a negative error in case of failure.
1701 *
1702 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1703 * Description
1704 * Emulate a call to **getsockopt()** on the socket associated to
1705 * *bpf_socket*, which must be a full socket. The *level* at
1706 * which the option resides and the name *optname* of the option
1707 * must be specified, see **getsockopt(2)** for more information.
1708 * The retrieved value is stored in the structure pointed by
1709 * *opval* and of length *optlen*.
1710 *
1711 * This helper actually implements a subset of **getsockopt()**.
1712 * It supports the following *level*\ s:
1713 *
1714 * * **IPPROTO_TCP**, which supports *optname*
1715 * **TCP_CONGESTION**.
1716 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1717 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1718 * Return
1719 * 0 on success, or a negative error in case of failure.
1720 *
1721 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1722 * Description
1723 * Used for error injection, this helper uses kprobes to override
1724 * the return value of the probed function, and to set it to *rc*.
1725 * The first argument is the context *regs* on which the kprobe
1726 * works.
1727 *
1728 * This helper works by setting setting the PC (program counter)
1729 * to an override function which is run in place of the original
1730 * probed function. This means the probed function is not run at
1731 * all. The replacement function just returns with the required
1732 * value.
1733 *
1734 * This helper has security implications, and thus is subject to
1735 * restrictions. It is only available if the kernel was compiled
1736 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1737 * option, and in this case it only works on functions tagged with
1738 * **ALLOW_ERROR_INJECTION** in the kernel code.
1739 *
1740 * Also, the helper is only available for the architectures having
1741 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1742 * x86 architecture is the only one to support this feature.
1743 * Return
1744 * 0
1745 *
1746 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1747 * Description
1748 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1749 * for the full TCP socket associated to *bpf_sock_ops* to
1750 * *argval*.
1751 *
1752 * The primary use of this field is to determine if there should
1753 * be calls to eBPF programs of type
1754 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1755 * code. A program of the same type can change its value, per
1756 * connection and as necessary, when the connection is
1757 * established. This field is directly accessible for reading, but
1758 * this helper must be used for updates in order to return an
1759 * error if an eBPF program tries to set a callback that is not
1760 * supported in the current kernel.
1761 *
70de8a7f 1762 * *argval* is a flag array which can combine these flags:
fd95ec0e
DA
1763 *
1764 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1765 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1766 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1767 *
70de8a7f
DA
1768 * Therefore, this function can be used to clear a callback flag by
1769 * setting the appropriate bit to zero. e.g. to disable the RTO
1770 * callback:
1771 *
1772 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
1773 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
1774 *
fd95ec0e
DA
1775 * Here are some examples of where one could call such eBPF
1776 * program:
1777 *
1778 * * When RTO fires.
1779 * * When a packet is retransmitted.
1780 * * When the connection terminates.
1781 * * When a packet is sent.
1782 * * When a packet is received.
1783 * Return
1784 * Code **-EINVAL** if the socket is not a full TCP socket;
1785 * otherwise, a positive number containing the bits that could not
1786 * be set is returned (which comes down to 0 if all bits were set
1787 * as required).
1788 *
1789 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1790 * Description
1791 * This helper is used in programs implementing policies at the
1792 * socket level. If the message *msg* is allowed to pass (i.e. if
1793 * the verdict eBPF program returns **SK_PASS**), redirect it to
1794 * the socket referenced by *map* (of type
1795 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1796 * egress interfaces can be used for redirection. The
1797 * **BPF_F_INGRESS** value in *flags* is used to make the
1798 * distinction (ingress path is selected if the flag is present,
1799 * egress path otherwise). This is the only flag supported for now.
1800 * Return
1801 * **SK_PASS** on success, or **SK_DROP** on error.
1802 *
1803 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1804 * Description
1805 * For socket policies, apply the verdict of the eBPF program to
1806 * the next *bytes* (number of bytes) of message *msg*.
1807 *
1808 * For example, this helper can be used in the following cases:
1809 *
1810 * * A single **sendmsg**\ () or **sendfile**\ () system call
1811 * contains multiple logical messages that the eBPF program is
1812 * supposed to read and for which it should apply a verdict.
1813 * * An eBPF program only cares to read the first *bytes* of a
1814 * *msg*. If the message has a large payload, then setting up
1815 * and calling the eBPF program repeatedly for all bytes, even
1816 * though the verdict is already known, would create unnecessary
1817 * overhead.
1818 *
1819 * When called from within an eBPF program, the helper sets a
1820 * counter internal to the BPF infrastructure, that is used to
1821 * apply the last verdict to the next *bytes*. If *bytes* is
1822 * smaller than the current data being processed from a
1823 * **sendmsg**\ () or **sendfile**\ () system call, the first
1824 * *bytes* will be sent and the eBPF program will be re-run with
1825 * the pointer for start of data pointing to byte number *bytes*
1826 * **+ 1**. If *bytes* is larger than the current data being
1827 * processed, then the eBPF verdict will be applied to multiple
1828 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1829 * consumed.
1830 *
1831 * Note that if a socket closes with the internal counter holding
1832 * a non-zero value, this is not a problem because data is not
1833 * being buffered for *bytes* and is sent as it is received.
1834 * Return
1835 * 0
1836 *
1837 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1838 * Description
1839 * For socket policies, prevent the execution of the verdict eBPF
1840 * program for message *msg* until *bytes* (byte number) have been
1841 * accumulated.
1842 *
1843 * This can be used when one needs a specific number of bytes
1844 * before a verdict can be assigned, even if the data spans
1845 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1846 * case would be a user calling **sendmsg**\ () repeatedly with
1847 * 1-byte long message segments. Obviously, this is bad for
1848 * performance, but it is still valid. If the eBPF program needs
1849 * *bytes* bytes to validate a header, this helper can be used to
1850 * prevent the eBPF program to be called again until *bytes* have
1851 * been accumulated.
1852 * Return
1853 * 0
1854 *
1855 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1856 * Description
1857 * For socket policies, pull in non-linear data from user space
1858 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1859 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1860 * respectively.
1861 *
1862 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1863 * *msg* it can only parse data that the (**data**, **data_end**)
1864 * pointers have already consumed. For **sendmsg**\ () hooks this
1865 * is likely the first scatterlist element. But for calls relying
1866 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1867 * be the range (**0**, **0**) because the data is shared with
1868 * user space and by default the objective is to avoid allowing
1869 * user space to modify data while (or after) eBPF verdict is
1870 * being decided. This helper can be used to pull in data and to
1871 * set the start and end pointer to given values. Data will be
1872 * copied if necessary (i.e. if data was not linear and if start
1873 * and end pointers do not point to the same chunk).
1874 *
b60ed9a3 1875 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1876 * packet buffer. Therefore, at load time, all checks on pointers
1877 * previously done by the verifier are invalidated and must be
1878 * performed again, if the helper is used in combination with
1879 * direct packet access.
1880 *
1881 * All values for *flags* are reserved for future usage, and must
1882 * be left at zero.
1883 * Return
1884 * 0 on success, or a negative error in case of failure.
1885 *
1886 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1887 * Description
1888 * Bind the socket associated to *ctx* to the address pointed by
1889 * *addr*, of length *addr_len*. This allows for making outgoing
1890 * connection from the desired IP address, which can be useful for
1891 * example when all processes inside a cgroup should use one
1892 * single IP address on a host that has multiple IP configured.
1893 *
1894 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1895 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1896 * **AF_INET6**). Looking for a free port to bind to can be
1897 * expensive, therefore binding to port is not permitted by the
1898 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1899 * must be set to zero.
1900 * Return
1901 * 0 on success, or a negative error in case of failure.
1902 *
1903 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1904 * Description
1905 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1906 * only possible to shrink the packet as of this writing,
1907 * therefore *delta* must be a negative integer.
1908 *
b60ed9a3 1909 * A call to this helper is susceptible to change the underlying
fd95ec0e
DA
1910 * packet buffer. Therefore, at load time, all checks on pointers
1911 * previously done by the verifier are invalidated and must be
1912 * performed again, if the helper is used in combination with
1913 * direct packet access.
1914 * Return
1915 * 0 on success, or a negative error in case of failure.
1916 *
1917 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1918 * Description
1919 * Retrieve the XFRM state (IP transform framework, see also
1920 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1921 *
1922 * The retrieved value is stored in the **struct bpf_xfrm_state**
1923 * pointed by *xfrm_state* and of length *size*.
1924 *
1925 * All values for *flags* are reserved for future usage, and must
1926 * be left at zero.
1927 *
1928 * This helper is available only if the kernel was compiled with
1929 * **CONFIG_XFRM** configuration option.
1930 * Return
1931 * 0 on success, or a negative error in case of failure.
1932 *
1933 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1934 * Description
1935 * Return a user or a kernel stack in bpf program provided buffer.
1936 * To achieve this, the helper needs *ctx*, which is a pointer
1937 * to the context on which the tracing program is executed.
1938 * To store the stacktrace, the bpf program provides *buf* with
1939 * a nonnegative *size*.
1940 *
1941 * The last argument, *flags*, holds the number of stack frames to
1942 * skip (from 0 to 255), masked with
1943 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1944 * the following flags:
1945 *
1946 * **BPF_F_USER_STACK**
1947 * Collect a user space stack instead of a kernel stack.
1948 * **BPF_F_USER_BUILD_ID**
1949 * Collect buildid+offset instead of ips for user stack,
1950 * only valid if **BPF_F_USER_STACK** is also specified.
1951 *
1952 * **bpf_get_stack**\ () can collect up to
1953 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1954 * to sufficient large buffer size. Note that
1955 * this limit can be controlled with the **sysctl** program, and
1956 * that it should be manually increased in order to profile long
1957 * user stacks (such as stacks for Java programs). To do so, use:
1958 *
1959 * ::
1960 *
1961 * # sysctl kernel.perf_event_max_stack=<new value>
fd95ec0e 1962 * Return
17678d30
SH
1963 * A non-negative value equal to or less than *size* on success,
1964 * or a negative error in case of failure.
fd95ec0e 1965 *
3eebc1d4 1966 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
fd95ec0e
DA
1967 * Description
1968 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1969 * it provides an easy way to load *len* bytes from *offset*
1970 * from the packet associated to *skb*, into the buffer pointed
1971 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1972 * a fifth argument *start_header* exists in order to select a
1973 * base offset to start from. *start_header* can be one of:
1974 *
1975 * **BPF_HDR_START_MAC**
1976 * Base offset to load data from is *skb*'s mac header.
1977 * **BPF_HDR_START_NET**
1978 * Base offset to load data from is *skb*'s network header.
1979 *
1980 * In general, "direct packet access" is the preferred method to
1981 * access packet data, however, this helper is in particular useful
1982 * in socket filters where *skb*\ **->data** does not always point
1983 * to the start of the mac header and where "direct packet access"
1984 * is not available.
fd95ec0e
DA
1985 * Return
1986 * 0 on success, or a negative error in case of failure.
1987 *
4276e652
DA
1988 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1989 * Description
1990 * Do FIB lookup in kernel tables using parameters in *params*.
1991 * If lookup is successful and result shows packet is to be
1992 * forwarded, the neighbor tables are searched for the nexthop.
1993 * If successful (ie., FIB lookup shows forwarding and nexthop
17678d30
SH
1994 * is resolved), the nexthop address is returned in ipv4_dst
1995 * or ipv6_dst based on family, smac is set to mac address of
1996 * egress device, dmac is set to nexthop mac address, rt_metric
22ddbd82
DA
1997 * is set to metric from route (IPv4/IPv6 only), and ifindex
1998 * is set to the device index of the nexthop from the FIB lookup.
4276e652 1999 *
17689d30
DA
2000 * *plen* argument is the size of the passed in struct.
2001 * *flags* argument can be a combination of one or more of the
2002 * following values:
4276e652 2003 *
17678d30
SH
2004 * **BPF_FIB_LOOKUP_DIRECT**
2005 * Do a direct table lookup vs full lookup using FIB
2006 * rules.
2007 * **BPF_FIB_LOOKUP_OUTPUT**
2008 * Perform lookup from an egress perspective (default is
2009 * ingress).
4276e652 2010 *
17689d30
DA
2011 * *ctx* is either **struct xdp_md** for XDP programs or
2012 * **struct sk_buff** tc cls_act programs.
2013 * Return
22ddbd82
DA
2014 * * < 0 if any input argument is invalid
2015 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2016 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3eebc1d4 2017 * packet is not forwarded or needs assist from full stack
4276e652
DA
2018 *
2019 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
2020 * Description
2021 * Add an entry to, or update a sockhash *map* referencing sockets.
2022 * The *skops* is used as a new value for the entry associated to
2023 * *key*. *flags* is one of:
2024 *
2025 * **BPF_NOEXIST**
2026 * The entry for *key* must not exist in the map.
2027 * **BPF_EXIST**
2028 * The entry for *key* must already exist in the map.
2029 * **BPF_ANY**
2030 * No condition on the existence of the entry for *key*.
2031 *
2032 * If the *map* has eBPF programs (parser and verdict), those will
2033 * be inherited by the socket being added. If the socket is
2034 * already attached to eBPF programs, this results in an error.
2035 * Return
2036 * 0 on success, or a negative error in case of failure.
2037 *
2038 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2039 * Description
2040 * This helper is used in programs implementing policies at the
2041 * socket level. If the message *msg* is allowed to pass (i.e. if
2042 * the verdict eBPF program returns **SK_PASS**), redirect it to
2043 * the socket referenced by *map* (of type
2044 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2045 * egress interfaces can be used for redirection. The
2046 * **BPF_F_INGRESS** value in *flags* is used to make the
2047 * distinction (ingress path is selected if the flag is present,
2048 * egress path otherwise). This is the only flag supported for now.
2049 * Return
2050 * **SK_PASS** on success, or **SK_DROP** on error.
2051 *
2052 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
2053 * Description
2054 * This helper is used in programs implementing policies at the
2055 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
2056 * if the verdeict eBPF program returns **SK_PASS**), redirect it
2057 * to the socket referenced by *map* (of type
2058 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2059 * egress interfaces can be used for redirection. The
2060 * **BPF_F_INGRESS** value in *flags* is used to make the
2061 * distinction (ingress path is selected if the flag is present,
2062 * egress otherwise). This is the only flag supported for now.
2063 * Return
2064 * **SK_PASS** on success, or **SK_DROP** on error.
57ac202c
DA
2065 *
2066 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2067 * Description
2068 * Encapsulate the packet associated to *skb* within a Layer 3
2069 * protocol header. This header is provided in the buffer at
2070 * address *hdr*, with *len* its size in bytes. *type* indicates
2071 * the protocol of the header and can be one of:
2072 *
2073 * **BPF_LWT_ENCAP_SEG6**
2074 * IPv6 encapsulation with Segment Routing Header
2075 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2076 * the IPv6 header is computed by the kernel.
2077 * **BPF_LWT_ENCAP_SEG6_INLINE**
2078 * Only works if *skb* contains an IPv6 packet. Insert a
2079 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2080 * the IPv6 header.
41fda879
DA
2081 * **BPF_LWT_ENCAP_IP**
2082 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
2083 * must be IPv4 or IPv6, followed by zero or more
b60ed9a3
SH
2084 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
2085 * total bytes in all prepended headers. Please note that
2086 * if **skb_is_gso**\ (*skb*) is true, no more than two
2087 * headers can be prepended, and the inner header, if
2088 * present, should be either GRE or UDP/GUE.
2089 *
2090 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
2091 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
2092 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
2093 * **BPF_PROG_TYPE_LWT_XMIT**.
2094 *
2095 * A call to this helper is susceptible to change the underlying
57ac202c
DA
2096 * packet buffer. Therefore, at load time, all checks on pointers
2097 * previously done by the verifier are invalidated and must be
2098 * performed again, if the helper is used in combination with
2099 * direct packet access.
2100 * Return
2101 * 0 on success, or a negative error in case of failure.
2102 *
2103 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2104 * Description
2105 * Store *len* bytes from address *from* into the packet
2106 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2107 * inside the outermost IPv6 Segment Routing Header can be
2108 * modified through this helper.
2109 *
b60ed9a3 2110 * A call to this helper is susceptible to change the underlying
57ac202c
DA
2111 * packet buffer. Therefore, at load time, all checks on pointers
2112 * previously done by the verifier are invalidated and must be
2113 * performed again, if the helper is used in combination with
2114 * direct packet access.
2115 * Return
2116 * 0 on success, or a negative error in case of failure.
2117 *
2118 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2119 * Description
2120 * Adjust the size allocated to TLVs in the outermost IPv6
2121 * Segment Routing Header contained in the packet associated to
2122 * *skb*, at position *offset* by *delta* bytes. Only offsets
2123 * after the segments are accepted. *delta* can be as well
2124 * positive (growing) as negative (shrinking).
2125 *
b60ed9a3 2126 * A call to this helper is susceptible to change the underlying
57ac202c
DA
2127 * packet buffer. Therefore, at load time, all checks on pointers
2128 * previously done by the verifier are invalidated and must be
2129 * performed again, if the helper is used in combination with
2130 * direct packet access.
2131 * Return
2132 * 0 on success, or a negative error in case of failure.
2133 *
2134 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2135 * Description
2136 * Apply an IPv6 Segment Routing action of type *action* to the
2137 * packet associated to *skb*. Each action takes a parameter
2138 * contained at address *param*, and of length *param_len* bytes.
2139 * *action* can be one of:
2140 *
2141 * **SEG6_LOCAL_ACTION_END_X**
2142 * End.X action: Endpoint with Layer-3 cross-connect.
2143 * Type of *param*: **struct in6_addr**.
2144 * **SEG6_LOCAL_ACTION_END_T**
2145 * End.T action: Endpoint with specific IPv6 table lookup.
2146 * Type of *param*: **int**.
2147 * **SEG6_LOCAL_ACTION_END_B6**
2148 * End.B6 action: Endpoint bound to an SRv6 policy.
b60ed9a3 2149 * Type of *param*: **struct ipv6_sr_hdr**.
57ac202c
DA
2150 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2151 * End.B6.Encap action: Endpoint bound to an SRv6
2152 * encapsulation policy.
b60ed9a3 2153 * Type of *param*: **struct ipv6_sr_hdr**.
57ac202c 2154 *
b60ed9a3 2155 * A call to this helper is susceptible to change the underlying
57ac202c
DA
2156 * packet buffer. Therefore, at load time, all checks on pointers
2157 * previously done by the verifier are invalidated and must be
2158 * performed again, if the helper is used in combination with
2159 * direct packet access.
2160 * Return
2161 * 0 on success, or a negative error in case of failure.
17678d30 2162 *
cdeb2674 2163 * int bpf_rc_repeat(void *ctx)
17678d30
SH
2164 * Description
2165 * This helper is used in programs implementing IR decoding, to
cdeb2674
DA
2166 * report a successfully decoded repeat key message. This delays
2167 * the generation of a key up event for previously generated
2168 * key down event.
17678d30 2169 *
cdeb2674
DA
2170 * Some IR protocols like NEC have a special IR message for
2171 * repeating last button, for when a button is held down.
17678d30
SH
2172 *
2173 * The *ctx* should point to the lirc sample as passed into
2174 * the program.
2175 *
17678d30
SH
2176 * This helper is only available is the kernel was compiled with
2177 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2178 * "**y**".
17678d30
SH
2179 * Return
2180 * 0
2181 *
cdeb2674 2182 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
17678d30
SH
2183 * Description
2184 * This helper is used in programs implementing IR decoding, to
cdeb2674
DA
2185 * report a successfully decoded key press with *scancode*,
2186 * *toggle* value in the given *protocol*. The scancode will be
2187 * translated to a keycode using the rc keymap, and reported as
2188 * an input key down event. After a period a key up event is
2189 * generated. This period can be extended by calling either
2190 * **bpf_rc_keydown**\ () again with the same values, or calling
2191 * **bpf_rc_repeat**\ ().
17678d30 2192 *
cdeb2674
DA
2193 * Some protocols include a toggle bit, in case the button was
2194 * released and pressed again between consecutive scancodes.
17678d30
SH
2195 *
2196 * The *ctx* should point to the lirc sample as passed into
2197 * the program.
2198 *
cdeb2674
DA
2199 * The *protocol* is the decoded protocol number (see
2200 * **enum rc_proto** for some predefined values).
2201 *
17678d30
SH
2202 * This helper is only available is the kernel was compiled with
2203 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2204 * "**y**".
17678d30
SH
2205 * Return
2206 * 0
2207 *
cdeb2674 2208 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
17678d30
SH
2209 * Description
2210 * Return the cgroup v2 id of the socket associated with the *skb*.
2211 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2212 * helper for cgroup v1 by providing a tag resp. identifier that
2213 * can be matched on or used for map lookups e.g. to implement
2214 * policy. The cgroup v2 id of a given path in the hierarchy is
2215 * exposed in user space through the f_handle API in order to get
2216 * to the same 64-bit id.
2217 *
2218 * This helper can be used on TC egress path, but not on ingress,
2219 * and is available only if the kernel was compiled with the
2220 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2221 * Return
2222 * The id is returned or 0 in case the id could not be retrieved.
2223 *
2224 * u64 bpf_get_current_cgroup_id(void)
2225 * Return
2226 * A 64-bit integer containing the current cgroup id based
2227 * on the cgroup within which the current task is running.
74eb09ad 2228 *
cdeb2674 2229 * void *bpf_get_local_storage(void *map, u64 flags)
74eb09ad
DA
2230 * Description
2231 * Get the pointer to the local storage area.
2232 * The type and the size of the local storage is defined
2233 * by the *map* argument.
2234 * The *flags* meaning is specific for each map type,
2235 * and has to be 0 for cgroup local storage.
2236 *
17689d30
DA
2237 * Depending on the BPF program type, a local storage area
2238 * can be shared between multiple instances of the BPF program,
74eb09ad
DA
2239 * running simultaneously.
2240 *
2241 * A user should care about the synchronization by himself.
17689d30 2242 * For example, by using the **BPF_STX_XADD** instruction to alter
74eb09ad
DA
2243 * the shared data.
2244 * Return
17689d30 2245 * A pointer to the local storage area.
0ebb4209
SH
2246 *
2247 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2248 * Description
17689d30
DA
2249 * Select a **SO_REUSEPORT** socket from a
2250 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2251 * It checks the selected socket is matching the incoming
2252 * request in the socket buffer.
0ebb4209
SH
2253 * Return
2254 * 0 on success, or a negative error in case of failure.
39776a86 2255 *
cdeb2674
DA
2256 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2257 * Description
2258 * Return id of cgroup v2 that is ancestor of cgroup associated
2259 * with the *skb* at the *ancestor_level*. The root cgroup is at
2260 * *ancestor_level* zero and each step down the hierarchy
2261 * increments the level. If *ancestor_level* == level of cgroup
2262 * associated with *skb*, then return value will be same as that
2263 * of **bpf_skb_cgroup_id**\ ().
2264 *
2265 * The helper is useful to implement policies based on cgroups
2266 * that are upper in hierarchy than immediate cgroup associated
2267 * with *skb*.
2268 *
2269 * The format of returned id and helper limitations are same as in
2270 * **bpf_skb_cgroup_id**\ ().
2271 * Return
2272 * The id is returned or 0 in case the id could not be retrieved.
2273 *
028766ae 2274 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
39776a86
SH
2275 * Description
2276 * Look for TCP socket matching *tuple*, optionally in a child
2277 * network namespace *netns*. The return value must be checked,
17689d30 2278 * and if non-**NULL**, released via **bpf_sk_release**\ ().
39776a86
SH
2279 *
2280 * The *ctx* should point to the context of the program, such as
2281 * the skb or socket (depending on the hook in use). This is used
2282 * to determine the base network namespace for the lookup.
2283 *
2284 * *tuple_size* must be one of:
2285 *
2286 * **sizeof**\ (*tuple*\ **->ipv4**)
2287 * Look for an IPv4 socket.
2288 * **sizeof**\ (*tuple*\ **->ipv6**)
2289 * Look for an IPv6 socket.
2290 *
028766ae
SH
2291 * If the *netns* is a negative signed 32-bit integer, then the
2292 * socket lookup table in the netns associated with the *ctx* will
2293 * will be used. For the TC hooks, this is the netns of the device
2294 * in the skb. For socket hooks, this is the netns of the socket.
2295 * If *netns* is any other signed 32-bit value greater than or
2296 * equal to zero then it specifies the ID of the netns relative to
2297 * the netns associated with the *ctx*. *netns* values beyond the
2298 * range of 32-bit integers are reserved for future use.
39776a86
SH
2299 *
2300 * All values for *flags* are reserved for future usage, and must
2301 * be left at zero.
2302 *
2303 * This helper is available only if the kernel was compiled with
2304 * **CONFIG_NET** configuration option.
2305 * Return
17689d30
DA
2306 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2307 * For sockets with reuseport option, the **struct bpf_sock**
b60ed9a3
SH
2308 * result is from *reuse*\ **->socks**\ [] using the hash of the
2309 * tuple.
39776a86 2310 *
028766ae 2311 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
39776a86
SH
2312 * Description
2313 * Look for UDP socket matching *tuple*, optionally in a child
2314 * network namespace *netns*. The return value must be checked,
17689d30 2315 * and if non-**NULL**, released via **bpf_sk_release**\ ().
39776a86
SH
2316 *
2317 * The *ctx* should point to the context of the program, such as
2318 * the skb or socket (depending on the hook in use). This is used
2319 * to determine the base network namespace for the lookup.
2320 *
2321 * *tuple_size* must be one of:
2322 *
2323 * **sizeof**\ (*tuple*\ **->ipv4**)
2324 * Look for an IPv4 socket.
2325 * **sizeof**\ (*tuple*\ **->ipv6**)
2326 * Look for an IPv6 socket.
2327 *
028766ae
SH
2328 * If the *netns* is a negative signed 32-bit integer, then the
2329 * socket lookup table in the netns associated with the *ctx* will
2330 * will be used. For the TC hooks, this is the netns of the device
2331 * in the skb. For socket hooks, this is the netns of the socket.
2332 * If *netns* is any other signed 32-bit value greater than or
2333 * equal to zero then it specifies the ID of the netns relative to
2334 * the netns associated with the *ctx*. *netns* values beyond the
2335 * range of 32-bit integers are reserved for future use.
39776a86
SH
2336 *
2337 * All values for *flags* are reserved for future usage, and must
2338 * be left at zero.
2339 *
2340 * This helper is available only if the kernel was compiled with
2341 * **CONFIG_NET** configuration option.
2342 * Return
17689d30
DA
2343 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2344 * For sockets with reuseport option, the **struct bpf_sock**
b60ed9a3
SH
2345 * result is from *reuse*\ **->socks**\ [] using the hash of the
2346 * tuple.
39776a86 2347 *
17689d30 2348 * int bpf_sk_release(struct bpf_sock *sock)
39776a86 2349 * Description
17689d30
DA
2350 * Release the reference held by *sock*. *sock* must be a
2351 * non-**NULL** pointer that was returned from
2352 * **bpf_sk_lookup_xxx**\ ().
39776a86
SH
2353 * Return
2354 * 0 on success, or a negative error in case of failure.
2355 *
cdeb2674
DA
2356 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
2357 * Description
2358 * Push an element *value* in *map*. *flags* is one of:
2359 *
2360 * **BPF_EXIST**
2361 * If the queue/stack is full, the oldest element is
2362 * removed to make room for this.
2363 * Return
2364 * 0 on success, or a negative error in case of failure.
2365 *
17689d30
DA
2366 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2367 * Description
2368 * Pop an element from *map*.
2369 * Return
2370 * 0 on success, or a negative error in case of failure.
2371 *
2372 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2373 * Description
2374 * Get an element from *map* without removing it.
2375 * Return
2376 * 0 on success, or a negative error in case of failure.
2377 *
39776a86
SH
2378 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2379 * Description
17689d30 2380 * For socket policies, insert *len* bytes into *msg* at offset
39776a86
SH
2381 * *start*.
2382 *
2383 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
17689d30 2384 * *msg* it may want to insert metadata or options into the *msg*.
39776a86
SH
2385 * This can later be read and used by any of the lower layer BPF
2386 * hooks.
2387 *
2388 * This helper may fail if under memory pressure (a malloc
2389 * fails) in these cases BPF programs will get an appropriate
2390 * error and BPF programs will need to handle them.
17689d30
DA
2391 * Return
2392 * 0 on success, or a negative error in case of failure.
39776a86 2393 *
17689d30
DA
2394 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2395 * Description
2396 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2397 * This may result in **ENOMEM** errors under certain situations if
2398 * an allocation and copy are required due to a full ring buffer.
2399 * However, the helper will try to avoid doing the allocation
2400 * if possible. Other errors can occur if input parameters are
2401 * invalid either due to *start* byte not being valid part of *msg*
2402 * payload and/or *pop* value being to large.
39776a86
SH
2403 * Return
2404 * 0 on success, or a negative error in case of failure.
17689d30
DA
2405 *
2406 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2407 * Description
2408 * This helper is used in programs implementing IR decoding, to
2409 * report a successfully decoded pointer movement.
2410 *
2411 * The *ctx* should point to the lirc sample as passed into
2412 * the program.
2413 *
2414 * This helper is only available is the kernel was compiled with
2415 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2416 * "**y**".
2417 * Return
2418 * 0
41fda879 2419 *
cdeb2674
DA
2420 * int bpf_spin_lock(struct bpf_spin_lock *lock)
2421 * Description
2422 * Acquire a spinlock represented by the pointer *lock*, which is
2423 * stored as part of a value of a map. Taking the lock allows to
2424 * safely update the rest of the fields in that value. The
2425 * spinlock can (and must) later be released with a call to
2426 * **bpf_spin_unlock**\ (\ *lock*\ ).
2427 *
2428 * Spinlocks in BPF programs come with a number of restrictions
2429 * and constraints:
2430 *
2431 * * **bpf_spin_lock** objects are only allowed inside maps of
2432 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
2433 * list could be extended in the future).
2434 * * BTF description of the map is mandatory.
2435 * * The BPF program can take ONE lock at a time, since taking two
2436 * or more could cause dead locks.
2437 * * Only one **struct bpf_spin_lock** is allowed per map element.
2438 * * When the lock is taken, calls (either BPF to BPF or helpers)
2439 * are not allowed.
2440 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
2441 * allowed inside a spinlock-ed region.
2442 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
2443 * the lock, on all execution paths, before it returns.
2444 * * The BPF program can access **struct bpf_spin_lock** only via
2445 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
2446 * helpers. Loading or storing data into the **struct
2447 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
2448 * * To use the **bpf_spin_lock**\ () helper, the BTF description
2449 * of the map value must be a struct and have **struct
2450 * bpf_spin_lock** *anyname*\ **;** field at the top level.
2451 * Nested lock inside another struct is not allowed.
2452 * * The **struct bpf_spin_lock** *lock* field in a map value must
2453 * be aligned on a multiple of 4 bytes in that value.
2454 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
2455 * the **bpf_spin_lock** field to user space.
2456 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
2457 * a BPF program, do not update the **bpf_spin_lock** field.
2458 * * **bpf_spin_lock** cannot be on the stack or inside a
2459 * networking packet (it can only be inside of a map values).
2460 * * **bpf_spin_lock** is available to root only.
2461 * * Tracing programs and socket filter programs cannot use
2462 * **bpf_spin_lock**\ () due to insufficient preemption checks
2463 * (but this may change in the future).
2464 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
2465 * Return
2466 * 0
2467 *
2468 * int bpf_spin_unlock(struct bpf_spin_lock *lock)
2469 * Description
2470 * Release the *lock* previously locked by a call to
2471 * **bpf_spin_lock**\ (\ *lock*\ ).
2472 * Return
2473 * 0
2474 *
41fda879
DA
2475 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
2476 * Description
2477 * This helper gets a **struct bpf_sock** pointer such
cdeb2674 2478 * that all the fields in this **bpf_sock** can be accessed.
41fda879 2479 * Return
cdeb2674 2480 * A **struct bpf_sock** pointer on success, or **NULL** in
41fda879
DA
2481 * case of failure.
2482 *
2483 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
2484 * Description
2485 * This helper gets a **struct bpf_tcp_sock** pointer from a
2486 * **struct bpf_sock** pointer.
41fda879 2487 * Return
cdeb2674 2488 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
41fda879 2489 * case of failure.
dd4a2b68
SH
2490 *
2491 * int bpf_skb_ecn_set_ce(struct sk_buf *skb)
cdeb2674
DA
2492 * Description
2493 * Set ECN (Explicit Congestion Notification) field of IP header
2494 * to **CE** (Congestion Encountered) if current value is **ECT**
2495 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
2496 * and IPv4.
2497 * Return
2498 * 1 if the **CE** flag is set (either by the current helper call
2499 * or because it was already present), 0 if it is not set.
2500 *
2501 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
2502 * Description
2503 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
2504 * **bpf_sk_release**\ () is unnecessary and not allowed.
2505 * Return
2506 * A **struct bpf_sock** pointer on success, or **NULL** in
2507 * case of failure.
188c7fe6
DA
2508 *
2509 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2510 * Description
2511 * Look for TCP socket matching *tuple*, optionally in a child
2512 * network namespace *netns*. The return value must be checked,
2513 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2514 *
b60ed9a3
SH
2515 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
2516 * that it also returns timewait or request sockets. Use
2517 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
2518 * full structure.
188c7fe6
DA
2519 *
2520 * This helper is available only if the kernel was compiled with
2521 * **CONFIG_NET** configuration option.
2522 * Return
2523 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2524 * For sockets with reuseport option, the **struct bpf_sock**
b60ed9a3
SH
2525 * result is from *reuse*\ **->socks**\ [] using the hash of the
2526 * tuple.
188c7fe6
DA
2527 *
2528 * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
2529 * Description
b60ed9a3
SH
2530 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
2531 * the listening socket in *sk*.
188c7fe6 2532 *
b60ed9a3
SH
2533 * *iph* points to the start of the IPv4 or IPv6 header, while
2534 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
2535 * **sizeof**\ (**struct ip6hdr**).
188c7fe6 2536 *
b60ed9a3
SH
2537 * *th* points to the start of the TCP header, while *th_len*
2538 * contains **sizeof**\ (**struct tcphdr**).
188c7fe6
DA
2539 *
2540 * Return
b60ed9a3
SH
2541 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
2542 * error otherwise.
70de8a7f
DA
2543 *
2544 * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
2545 * Description
2546 * Get name of sysctl in /proc/sys/ and copy it into provided by
2547 * program buffer *buf* of size *buf_len*.
2548 *
2549 * The buffer is always NUL terminated, unless it's zero-sized.
2550 *
2551 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
2552 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
2553 * only (e.g. "tcp_mem").
2554 * Return
2555 * Number of character copied (not including the trailing NUL).
2556 *
2557 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2558 * truncated name in this case).
2559 *
2560 * int bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2561 * Description
2562 * Get current value of sysctl as it is presented in /proc/sys
2563 * (incl. newline, etc), and copy it as a string into provided
2564 * by program buffer *buf* of size *buf_len*.
2565 *
2566 * The whole value is copied, no matter what file position user
2567 * space issued e.g. sys_read at.
2568 *
2569 * The buffer is always NUL terminated, unless it's zero-sized.
2570 * Return
2571 * Number of character copied (not including the trailing NUL).
2572 *
2573 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2574 * truncated name in this case).
2575 *
2576 * **-EINVAL** if current value was unavailable, e.g. because
2577 * sysctl is uninitialized and read returns -EIO for it.
2578 *
2579 * int bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
2580 * Description
2581 * Get new value being written by user space to sysctl (before
2582 * the actual write happens) and copy it as a string into
2583 * provided by program buffer *buf* of size *buf_len*.
2584 *
2585 * User space may write new value at file position > 0.
2586 *
2587 * The buffer is always NUL terminated, unless it's zero-sized.
2588 * Return
2589 * Number of character copied (not including the trailing NUL).
2590 *
2591 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
2592 * truncated name in this case).
2593 *
2594 * **-EINVAL** if sysctl is being read.
2595 *
2596 * int bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
2597 * Description
2598 * Override new value being written by user space to sysctl with
2599 * value provided by program in buffer *buf* of size *buf_len*.
2600 *
2601 * *buf* should contain a string in same form as provided by user
2602 * space on sysctl write.
2603 *
2604 * User space may write new value at file position > 0. To override
2605 * the whole sysctl value file position should be set to zero.
2606 * Return
2607 * 0 on success.
2608 *
2609 * **-E2BIG** if the *buf_len* is too big.
2610 *
2611 * **-EINVAL** if sysctl is being read.
2612 *
2613 * int bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
2614 * Description
2615 * Convert the initial part of the string from buffer *buf* of
2616 * size *buf_len* to a long integer according to the given base
2617 * and save the result in *res*.
2618 *
2619 * The string may begin with an arbitrary amount of white space
b60ed9a3
SH
2620 * (as determined by **isspace**\ (3)) followed by a single
2621 * optional '**-**' sign.
70de8a7f
DA
2622 *
2623 * Five least significant bits of *flags* encode base, other bits
2624 * are currently unused.
2625 *
2626 * Base must be either 8, 10, 16 or 0 to detect it automatically
b60ed9a3 2627 * similar to user space **strtol**\ (3).
70de8a7f
DA
2628 * Return
2629 * Number of characters consumed on success. Must be positive but
b60ed9a3 2630 * no more than *buf_len*.
70de8a7f
DA
2631 *
2632 * **-EINVAL** if no valid digits were found or unsupported base
2633 * was provided.
2634 *
2635 * **-ERANGE** if resulting value was out of range.
2636 *
2637 * int bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
2638 * Description
2639 * Convert the initial part of the string from buffer *buf* of
2640 * size *buf_len* to an unsigned long integer according to the
2641 * given base and save the result in *res*.
2642 *
2643 * The string may begin with an arbitrary amount of white space
b60ed9a3 2644 * (as determined by **isspace**\ (3)).
70de8a7f
DA
2645 *
2646 * Five least significant bits of *flags* encode base, other bits
2647 * are currently unused.
2648 *
2649 * Base must be either 8, 10, 16 or 0 to detect it automatically
b60ed9a3 2650 * similar to user space **strtoul**\ (3).
70de8a7f
DA
2651 * Return
2652 * Number of characters consumed on success. Must be positive but
b60ed9a3 2653 * no more than *buf_len*.
70de8a7f
DA
2654 *
2655 * **-EINVAL** if no valid digits were found or unsupported base
2656 * was provided.
2657 *
2658 * **-ERANGE** if resulting value was out of range.
fd658097
DA
2659 *
2660 * void *bpf_sk_storage_get(struct bpf_map *map, struct bpf_sock *sk, void *value, u64 flags)
2661 * Description
b60ed9a3 2662 * Get a bpf-local-storage from a *sk*.
fd658097
DA
2663 *
2664 * Logically, it could be thought of getting the value from
2665 * a *map* with *sk* as the **key**. From this
2666 * perspective, the usage is not much different from
b60ed9a3
SH
2667 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
2668 * helper enforces the key must be a full socket and the map must
2669 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
fd658097
DA
2670 *
2671 * Underneath, the value is stored locally at *sk* instead of
b60ed9a3
SH
2672 * the *map*. The *map* is used as the bpf-local-storage
2673 * "type". The bpf-local-storage "type" (i.e. the *map*) is
2674 * searched against all bpf-local-storages residing at *sk*.
fd658097 2675 *
b60ed9a3 2676 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
fd658097
DA
2677 * used such that a new bpf-local-storage will be
2678 * created if one does not exist. *value* can be used
b60ed9a3 2679 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
fd658097 2680 * the initial value of a bpf-local-storage. If *value* is
b60ed9a3 2681 * **NULL**, the new bpf-local-storage will be zero initialized.
fd658097
DA
2682 * Return
2683 * A bpf-local-storage pointer is returned on success.
2684 *
2685 * **NULL** if not found or there was an error in adding
2686 * a new bpf-local-storage.
2687 *
2688 * int bpf_sk_storage_delete(struct bpf_map *map, struct bpf_sock *sk)
2689 * Description
b60ed9a3 2690 * Delete a bpf-local-storage from a *sk*.
fd658097
DA
2691 * Return
2692 * 0 on success.
2693 *
2694 * **-ENOENT** if the bpf-local-storage cannot be found.
b2f8eb7f
DA
2695 *
2696 * int bpf_send_signal(u32 sig)
2697 * Description
2698 * Send signal *sig* to the current task.
2699 * Return
2700 * 0 on success or successfully queued.
2701 *
2702 * **-EBUSY** if work queue under nmi is full.
2703 *
2704 * **-EINVAL** if *sig* is invalid.
2705 *
2706 * **-EPERM** if no permission to send the *sig*.
2707 *
2708 * **-EAGAIN** if bpf program can try again.
1a97748b
SH
2709 */
2710#define __BPF_FUNC_MAPPER(FN) \
2711 FN(unspec), \
2712 FN(map_lookup_elem), \
2713 FN(map_update_elem), \
2714 FN(map_delete_elem), \
2715 FN(probe_read), \
2716 FN(ktime_get_ns), \
2717 FN(trace_printk), \
2718 FN(get_prandom_u32), \
2719 FN(get_smp_processor_id), \
2720 FN(skb_store_bytes), \
2721 FN(l3_csum_replace), \
2722 FN(l4_csum_replace), \
2723 FN(tail_call), \
2724 FN(clone_redirect), \
2725 FN(get_current_pid_tgid), \
2726 FN(get_current_uid_gid), \
2727 FN(get_current_comm), \
2728 FN(get_cgroup_classid), \
2729 FN(skb_vlan_push), \
2730 FN(skb_vlan_pop), \
2731 FN(skb_get_tunnel_key), \
2732 FN(skb_set_tunnel_key), \
2733 FN(perf_event_read), \
2734 FN(redirect), \
2735 FN(get_route_realm), \
2736 FN(perf_event_output), \
2737 FN(skb_load_bytes), \
2738 FN(get_stackid), \
2739 FN(csum_diff), \
2740 FN(skb_get_tunnel_opt), \
2741 FN(skb_set_tunnel_opt), \
2742 FN(skb_change_proto), \
2743 FN(skb_change_type), \
2744 FN(skb_under_cgroup), \
2745 FN(get_hash_recalc), \
2746 FN(get_current_task), \
2747 FN(probe_write_user), \
2748 FN(current_task_under_cgroup), \
2749 FN(skb_change_tail), \
2750 FN(skb_pull_data), \
2751 FN(csum_update), \
2752 FN(set_hash_invalid), \
143a704b 2753 FN(get_numa_node_id), \
3dd0bb51 2754 FN(skb_change_head), \
a044b36a 2755 FN(xdp_adjust_head), \
45f78b4d
SH
2756 FN(probe_read_str), \
2757 FN(get_socket_cookie), \
b68581d4 2758 FN(get_socket_uid), \
f0b9b795
SH
2759 FN(set_hash), \
2760 FN(setsockopt), \
2761 FN(skb_adjust_room), \
18d7817c
SH
2762 FN(redirect_map), \
2763 FN(sk_redirect_map), \
2764 FN(sock_map_update), \
f53da99a
SH
2765 FN(xdp_adjust_meta), \
2766 FN(perf_event_read_value), \
fa19d6bc 2767 FN(perf_prog_read_value), \
f82517f8 2768 FN(getsockopt), \
07ed8df6 2769 FN(override_return), \
9effc146
DA
2770 FN(sock_ops_cb_flags_set), \
2771 FN(msg_redirect_map), \
2772 FN(msg_apply_bytes), \
2773 FN(msg_cork_bytes), \
dcf7997b 2774 FN(msg_pull_data), \
fd95ec0e
DA
2775 FN(bind), \
2776 FN(xdp_adjust_tail), \
2777 FN(skb_get_xfrm_state), \
2778 FN(get_stack), \
4276e652
DA
2779 FN(skb_load_bytes_relative), \
2780 FN(fib_lookup), \
2781 FN(sock_hash_update), \
2782 FN(msg_redirect_hash), \
57ac202c
DA
2783 FN(sk_redirect_hash), \
2784 FN(lwt_push_encap), \
2785 FN(lwt_seg6_store_bytes), \
2786 FN(lwt_seg6_adjust_srh), \
17678d30
SH
2787 FN(lwt_seg6_action), \
2788 FN(rc_repeat), \
2789 FN(rc_keydown), \
2790 FN(skb_cgroup_id), \
74eb09ad 2791 FN(get_current_cgroup_id), \
0ebb4209
SH
2792 FN(get_local_storage), \
2793 FN(sk_select_reuseport), \
39776a86
SH
2794 FN(skb_ancestor_cgroup_id), \
2795 FN(sk_lookup_tcp), \
2796 FN(sk_lookup_udp), \
2797 FN(sk_release), \
2798 FN(map_push_elem), \
2799 FN(map_pop_elem), \
2800 FN(map_peek_elem), \
17689d30
DA
2801 FN(msg_push_data), \
2802 FN(msg_pop_data), \
6b2d60bd
DA
2803 FN(rc_pointer_rel), \
2804 FN(spin_lock), \
41fda879
DA
2805 FN(spin_unlock), \
2806 FN(sk_fullsock), \
dd4a2b68 2807 FN(tcp_sock), \
cdeb2674 2808 FN(skb_ecn_set_ce), \
188c7fe6
DA
2809 FN(get_listener_sock), \
2810 FN(skc_lookup_tcp), \
70de8a7f
DA
2811 FN(tcp_check_syncookie), \
2812 FN(sysctl_get_name), \
2813 FN(sysctl_get_current_value), \
2814 FN(sysctl_get_new_value), \
2815 FN(sysctl_set_new_value), \
2816 FN(strtol), \
fd658097
DA
2817 FN(strtoul), \
2818 FN(sk_storage_get), \
b2f8eb7f
DA
2819 FN(sk_storage_delete), \
2820 FN(send_signal),
1a97748b 2821
b54ac87e
DB
2822/* integer value in 'imm' field of BPF_CALL instruction selects which helper
2823 * function eBPF program intends to call
2824 */
1a97748b 2825#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
b54ac87e 2826enum bpf_func_id {
1a97748b 2827 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
b54ac87e
DB
2828 __BPF_FUNC_MAX_ID,
2829};
1a97748b 2830#undef __BPF_ENUM_FN
b54ac87e 2831
7321b7db
SH
2832/* All flags used by eBPF helper functions, placed here. */
2833
2834/* BPF_FUNC_skb_store_bytes flags. */
2835#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
b7e0091a 2836#define BPF_F_INVALIDATE_HASH (1ULL << 1)
7321b7db
SH
2837
2838/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2839 * First 4 bits are for passing the header field size.
2840 */
2841#define BPF_F_HDR_FIELD_MASK 0xfULL
2842
2843/* BPF_FUNC_l4_csum_replace flags. */
2844#define BPF_F_PSEUDO_HDR (1ULL << 4)
2421ab75 2845#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
b479a7d7 2846#define BPF_F_MARK_ENFORCE (1ULL << 6)
7321b7db
SH
2847
2848/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2849#define BPF_F_INGRESS (1ULL << 0)
2850
2851/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2852#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2853
fd95ec0e 2854/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2421ab75
SH
2855#define BPF_F_SKIP_FIELD_MASK 0xffULL
2856#define BPF_F_USER_STACK (1ULL << 8)
fd95ec0e 2857/* flags used by BPF_FUNC_get_stackid only. */
2421ab75
SH
2858#define BPF_F_FAST_STACK_CMP (1ULL << 9)
2859#define BPF_F_REUSE_STACKID (1ULL << 10)
fd95ec0e
DA
2860/* flags used by BPF_FUNC_get_stack only. */
2861#define BPF_F_USER_BUILD_ID (1ULL << 11)
2421ab75 2862
b7e0091a
SH
2863/* BPF_FUNC_skb_set_tunnel_key flags. */
2864#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2865#define BPF_F_DONT_FRAGMENT (1ULL << 2)
8c278eca 2866#define BPF_F_SEQ_NUMBER (1ULL << 3)
b7e0091a 2867
f53da99a
SH
2868/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2869 * BPF_FUNC_perf_event_read_value flags.
2870 */
32c0b9b7
SH
2871#define BPF_F_INDEX_MASK 0xffffffffULL
2872#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
ba91cd9d
SH
2873/* BPF_FUNC_perf_event_output for sk_buff input context. */
2874#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
32c0b9b7 2875
028766ae
SH
2876/* Current network namespace */
2877#define BPF_F_CURRENT_NETNS (-1L)
2878
188c7fe6
DA
2879/* BPF_FUNC_skb_adjust_room flags. */
2880#define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
2881
70de8a7f
DA
2882#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff
2883#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
188c7fe6
DA
2884
2885#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1)
2886#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2)
2887#define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3)
2888#define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4)
70de8a7f 2889#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
188c7fe6
DA
2890 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
2891 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
2892
70de8a7f
DA
2893/* BPF_FUNC_sysctl_get_name flags. */
2894#define BPF_F_SYSCTL_BASE_NAME (1ULL << 0)
2895
fd658097
DA
2896/* BPF_FUNC_sk_storage_get flags */
2897#define BPF_SK_STORAGE_GET_F_CREATE (1ULL << 0)
2898
f0b9b795
SH
2899/* Mode for BPF_FUNC_skb_adjust_room helper. */
2900enum bpf_adj_room_mode {
2901 BPF_ADJ_ROOM_NET,
188c7fe6 2902 BPF_ADJ_ROOM_MAC,
f0b9b795
SH
2903};
2904
fd95ec0e
DA
2905/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2906enum bpf_hdr_start_off {
2907 BPF_HDR_START_MAC,
2908 BPF_HDR_START_NET,
2909};
2910
57ac202c
DA
2911/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2912enum bpf_lwt_encap_mode {
2913 BPF_LWT_ENCAP_SEG6,
41fda879
DA
2914 BPF_LWT_ENCAP_SEG6_INLINE,
2915 BPF_LWT_ENCAP_IP,
57ac202c
DA
2916};
2917
028766ae
SH
2918#define __bpf_md_ptr(type, name) \
2919union { \
2920 type name; \
2921 __u64 :64; \
2922} __attribute__((aligned(8)))
2923
b54ac87e
DB
2924/* user accessible mirror of in-kernel sk_buff.
2925 * new fields can only be added to the end of this structure
2926 */
2927struct __sk_buff {
2928 __u32 len;
2929 __u32 pkt_type;
2930 __u32 mark;
2931 __u32 queue_mapping;
cbdc3ed8
SH
2932 __u32 protocol;
2933 __u32 vlan_present;
2934 __u32 vlan_tci;
2935 __u32 vlan_proto;
e46efaed 2936 __u32 priority;
ff631c3a
SH
2937 __u32 ingress_ifindex;
2938 __u32 ifindex;
2939 __u32 tc_index;
2940 __u32 cb[5];
4f3489cd 2941 __u32 hash;
4e39bfb9 2942 __u32 tc_classid;
31ce6e01
SH
2943 __u32 data;
2944 __u32 data_end;
5b0aa887 2945 __u32 napi_id;
18d7817c 2946
1db903de 2947 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
18d7817c
SH
2948 __u32 family;
2949 __u32 remote_ip4; /* Stored in network byte order */
2950 __u32 local_ip4; /* Stored in network byte order */
2951 __u32 remote_ip6[4]; /* Stored in network byte order */
2952 __u32 local_ip6[4]; /* Stored in network byte order */
2953 __u32 remote_port; /* Stored in network byte order */
2954 __u32 local_port; /* stored in host byte order */
1db903de
SH
2955 /* ... here. */
2956
2957 __u32 data_meta;
028766ae 2958 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
17689d30
DA
2959 __u64 tstamp;
2960 __u32 wire_len;
f79b7733 2961 __u32 gso_segs;
41fda879 2962 __bpf_md_ptr(struct bpf_sock *, sk);
b54ac87e
DB
2963};
2964
089d93d6
SH
2965struct bpf_tunnel_key {
2966 __u32 tunnel_id;
7321b7db
SH
2967 union {
2968 __u32 remote_ipv4;
2969 __u32 remote_ipv6[4];
2970 };
2971 __u8 tunnel_tos;
2972 __u8 tunnel_ttl;
17678d30 2973 __u16 tunnel_ext; /* Padding, future use. */
b7e0091a 2974 __u32 tunnel_label;
089d93d6
SH
2975};
2976
fd95ec0e
DA
2977/* user accessible mirror of in-kernel xfrm_state.
2978 * new fields can only be added to the end of this structure
2979 */
2980struct bpf_xfrm_state {
2981 __u32 reqid;
2982 __u32 spi; /* Stored in network byte order */
2983 __u16 family;
17678d30 2984 __u16 ext; /* Padding, future use. */
fd95ec0e
DA
2985 union {
2986 __u32 remote_ipv4; /* Stored in network byte order */
2987 __u32 remote_ipv6[4]; /* Stored in network byte order */
2988 };
2989};
2990
143a704b
SH
2991/* Generic BPF return codes which all BPF program types may support.
2992 * The values are binary compatible with their TC_ACT_* counter-part to
2993 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2994 * programs.
2995 *
2996 * XDP is handled seprately, see XDP_*.
2997 */
2998enum bpf_ret_code {
2999 BPF_OK = 0,
3000 /* 1 reserved */
3001 BPF_DROP = 2,
3002 /* 3-6 reserved */
3003 BPF_REDIRECT = 7,
41fda879
DA
3004 /* >127 are reserved for prog type specific return codes.
3005 *
3006 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
3007 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
3008 * changed and should be routed based on its new L3 header.
3009 * (This is an L3 redirect, as opposed to L2 redirect
3010 * represented by BPF_REDIRECT above).
3011 */
3012 BPF_LWT_REROUTE = 128,
143a704b
SH
3013};
3014
3015struct bpf_sock {
3016 __u32 bound_dev_if;
3017 __u32 family;
3018 __u32 type;
3019 __u32 protocol;
4a5b3035
SH
3020 __u32 mark;
3021 __u32 priority;
41fda879
DA
3022 /* IP address also allows 1 and 2 bytes access */
3023 __u32 src_ip4;
3024 __u32 src_ip6[4];
3025 __u32 src_port; /* host byte order */
3026 __u32 dst_port; /* network byte order */
3027 __u32 dst_ip4;
3028 __u32 dst_ip6[4];
3029 __u32 state;
3030};
3031
3032struct bpf_tcp_sock {
3033 __u32 snd_cwnd; /* Sending congestion window */
3034 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
3035 __u32 rtt_min;
3036 __u32 snd_ssthresh; /* Slow start size threshold */
3037 __u32 rcv_nxt; /* What we want to receive next */
3038 __u32 snd_nxt; /* Next sequence we send */
3039 __u32 snd_una; /* First byte we want an ack for */
3040 __u32 mss_cache; /* Cached effective mss, not including SACKS */
3041 __u32 ecn_flags; /* ECN status bits. */
3042 __u32 rate_delivered; /* saved rate sample: packets delivered */
3043 __u32 rate_interval_us; /* saved rate sample: time elapsed */
3044 __u32 packets_out; /* Packets which are "in flight" */
3045 __u32 retrans_out; /* Retransmitted packets out */
3046 __u32 total_retrans; /* Total retransmits for entire connection */
3047 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
3048 * total number of segments in.
3049 */
3050 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
3051 * total number of data segments in.
3052 */
3053 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
3054 * The total number of segments sent.
3055 */
3056 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
3057 * total number of data segments sent.
dcf7997b 3058 */
41fda879
DA
3059 __u32 lost_out; /* Lost packets */
3060 __u32 sacked_out; /* SACK'd packets */
3061 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
3062 * sum(delta(rcv_nxt)), or how many bytes
3063 * were acked.
dcf7997b 3064 */
41fda879
DA
3065 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
3066 * sum(delta(snd_una)), or how many bytes
3067 * were acked.
dcf7997b 3068 */
143a704b
SH
3069};
3070
39776a86
SH
3071struct bpf_sock_tuple {
3072 union {
3073 struct {
3074 __be32 saddr;
3075 __be32 daddr;
3076 __be16 sport;
3077 __be16 dport;
3078 } ipv4;
3079 struct {
3080 __be32 saddr[4];
3081 __be32 daddr[4];
3082 __be16 sport;
3083 __be16 dport;
3084 } ipv6;
3085 };
3086};
3087
3dd0bb51
SH
3088#define XDP_PACKET_HEADROOM 256
3089
ba91cd9d
SH
3090/* User return codes for XDP prog type.
3091 * A valid XDP program must return one of these defined values. All other
44cf8415
SH
3092 * return codes are reserved for future use. Unknown return codes will
3093 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
ba91cd9d
SH
3094 */
3095enum xdp_action {
3096 XDP_ABORTED = 0,
3097 XDP_DROP,
3098 XDP_PASS,
3099 XDP_TX,
f0b9b795 3100 XDP_REDIRECT,
ba91cd9d
SH
3101};
3102
3103/* user accessible metadata for XDP packet hook
3104 * new fields must be added to the end of this structure
3105 */
3106struct xdp_md {
3107 __u32 data;
3108 __u32 data_end;
1db903de 3109 __u32 data_meta;
c0788a09
DA
3110 /* Below access go through struct xdp_rxq_info */
3111 __u32 ingress_ifindex; /* rxq->dev->ifindex */
3112 __u32 rx_queue_index; /* rxq->queue_index */
ba91cd9d
SH
3113};
3114
18d7817c 3115enum sk_action {
fe388b9e
SH
3116 SK_DROP = 0,
3117 SK_PASS,
18d7817c
SH
3118};
3119
9effc146
DA
3120/* user accessible metadata for SK_MSG packet hook, new fields must
3121 * be added to the end of this structure
3122 */
3123struct sk_msg_md {
028766ae
SH
3124 __bpf_md_ptr(void *, data);
3125 __bpf_md_ptr(void *, data_end);
57ac202c
DA
3126
3127 __u32 family;
3128 __u32 remote_ip4; /* Stored in network byte order */
3129 __u32 local_ip4; /* Stored in network byte order */
3130 __u32 remote_ip6[4]; /* Stored in network byte order */
3131 __u32 local_ip6[4]; /* Stored in network byte order */
3132 __u32 remote_port; /* Stored in network byte order */
3133 __u32 local_port; /* stored in host byte order */
fdce94d0 3134 __u32 size; /* Total size of sk_msg */
9effc146
DA
3135};
3136
0ebb4209
SH
3137struct sk_reuseport_md {
3138 /*
3139 * Start of directly accessible data. It begins from
3140 * the tcp/udp header.
3141 */
028766ae
SH
3142 __bpf_md_ptr(void *, data);
3143 /* End of directly accessible data */
3144 __bpf_md_ptr(void *, data_end);
0ebb4209
SH
3145 /*
3146 * Total length of packet (starting from the tcp/udp header).
3147 * Note that the directly accessible bytes (data_end - data)
3148 * could be less than this "len". Those bytes could be
3149 * indirectly read by a helper "bpf_skb_load_bytes()".
3150 */
3151 __u32 len;
3152 /*
3153 * Eth protocol in the mac header (network byte order). e.g.
3154 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
3155 */
3156 __u32 eth_protocol;
3157 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
3158 __u32 bind_inany; /* Is sock bound to an INANY address? */
3159 __u32 hash; /* A hash of the packet 4 tuples */
3160};
3161
410556ad
SH
3162#define BPF_TAG_SIZE 8
3163
3164struct bpf_prog_info {
3165 __u32 type;
3166 __u32 id;
3167 __u8 tag[BPF_TAG_SIZE];
3168 __u32 jited_prog_len;
3169 __u32 xlated_prog_len;
3170 __aligned_u64 jited_prog_insns;
3171 __aligned_u64 xlated_prog_insns;
1db903de
SH
3172 __u64 load_time; /* ns since boottime */
3173 __u32 created_by_uid;
3174 __u32 nr_map_ids;
3175 __aligned_u64 map_ids;
f53da99a 3176 char name[BPF_OBJ_NAME_LEN];
c0788a09 3177 __u32 ifindex;
fd95ec0e 3178 __u32 gpl_compatible:1;
c0788a09
DA
3179 __u64 netns_dev;
3180 __u64 netns_ino;
57ac202c
DA
3181 __u32 nr_jited_ksyms;
3182 __u32 nr_jited_func_lens;
3183 __aligned_u64 jited_ksyms;
3184 __aligned_u64 jited_func_lens;
17689d30
DA
3185 __u32 btf_id;
3186 __u32 func_info_rec_size;
3187 __aligned_u64 func_info;
3188 __u32 nr_func_info;
3189 __u32 nr_line_info;
3190 __aligned_u64 line_info;
3191 __aligned_u64 jited_line_info;
3192 __u32 nr_jited_line_info;
3193 __u32 line_info_rec_size;
3194 __u32 jited_line_info_rec_size;
fdce94d0
DA
3195 __u32 nr_prog_tags;
3196 __aligned_u64 prog_tags;
dd4a2b68
SH
3197 __u64 run_time_ns;
3198 __u64 run_cnt;
410556ad
SH
3199} __attribute__((aligned(8)));
3200
3201struct bpf_map_info {
3202 __u32 type;
3203 __u32 id;
3204 __u32 key_size;
3205 __u32 value_size;
3206 __u32 max_entries;
3207 __u32 map_flags;
f53da99a 3208 char name[BPF_OBJ_NAME_LEN];
07ed8df6 3209 __u32 ifindex;
4be85d57 3210 __u32 :32;
07ed8df6
DA
3211 __u64 netns_dev;
3212 __u64 netns_ino;
4276e652 3213 __u32 btf_id;
57ac202c
DA
3214 __u32 btf_key_type_id;
3215 __u32 btf_value_type_id;
4276e652
DA
3216} __attribute__((aligned(8)));
3217
3218struct bpf_btf_info {
3219 __aligned_u64 btf;
3220 __u32 btf_size;
3221 __u32 id;
410556ad
SH
3222} __attribute__((aligned(8)));
3223
dcf7997b
SH
3224/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
3225 * by user and intended to be used by socket (e.g. to bind to, depends on
3226 * attach attach type).
3227 */
3228struct bpf_sock_addr {
3229 __u32 user_family; /* Allows 4-byte read, but no write. */
3230 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
3231 * Stored in network byte order.
3232 */
3233 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3234 * Stored in network byte order.
3235 */
3236 __u32 user_port; /* Allows 4-byte read and write.
3237 * Stored in network byte order
3238 */
3239 __u32 family; /* Allows 4-byte read, but no write */
3240 __u32 type; /* Allows 4-byte read, but no write */
3241 __u32 protocol; /* Allows 4-byte read, but no write */
17678d30
SH
3242 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
3243 * Stored in network byte order.
3244 */
3245 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
3246 * Stored in network byte order.
3247 */
dcf7997b
SH
3248};
3249
f0b9b795
SH
3250/* User bpf_sock_ops struct to access socket values and specify request ops
3251 * and their replies.
3252 * Some of this fields are in network (bigendian) byte order and may need
3253 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
3254 * New fields can only be added at the end of this structure
3255 */
3256struct bpf_sock_ops {
3257 __u32 op;
3258 union {
07ed8df6
DA
3259 __u32 args[4]; /* Optionally passed to bpf program */
3260 __u32 reply; /* Returned by bpf program */
3261 __u32 replylong[4]; /* Optionally returned by bpf prog */
f0b9b795
SH
3262 };
3263 __u32 family;
3264 __u32 remote_ip4; /* Stored in network byte order */
3265 __u32 local_ip4; /* Stored in network byte order */
3266 __u32 remote_ip6[4]; /* Stored in network byte order */
3267 __u32 local_ip6[4]; /* Stored in network byte order */
3268 __u32 remote_port; /* Stored in network byte order */
3269 __u32 local_port; /* stored in host byte order */
b7f5fd36
SH
3270 __u32 is_fullsock; /* Some TCP fields are only valid if
3271 * there is a full socket. If not, the
3272 * fields read as zero.
3273 */
3274 __u32 snd_cwnd;
3275 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
07ed8df6
DA
3276 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
3277 __u32 state;
3278 __u32 rtt_min;
3279 __u32 snd_ssthresh;
3280 __u32 rcv_nxt;
3281 __u32 snd_nxt;
3282 __u32 snd_una;
3283 __u32 mss_cache;
3284 __u32 ecn_flags;
3285 __u32 rate_delivered;
3286 __u32 rate_interval_us;
3287 __u32 packets_out;
3288 __u32 retrans_out;
3289 __u32 total_retrans;
3290 __u32 segs_in;
3291 __u32 data_segs_in;
3292 __u32 segs_out;
3293 __u32 data_segs_out;
3294 __u32 lost_out;
3295 __u32 sacked_out;
3296 __u32 sk_txhash;
3297 __u64 bytes_received;
3298 __u64 bytes_acked;
f0b9b795
SH
3299};
3300
07ed8df6
DA
3301/* Definitions for bpf_sock_ops_cb_flags */
3302#define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
3303#define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
3304#define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
3305#define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
3306 * supported cb flags
3307 */
3308
f0b9b795
SH
3309/* List of known BPF sock_ops operators.
3310 * New entries can only be added at the end
3311 */
3312enum {
3313 BPF_SOCK_OPS_VOID,
3314 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
3315 * -1 if default value should be used
3316 */
3317 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
3318 * window (in packets) or -1 if default
3319 * value should be used
3320 */
3321 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
3322 * active connection is initialized
3323 */
3324 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
3325 * active connection is
3326 * established
3327 */
3328 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
3329 * passive connection is
3330 * established
3331 */
3332 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
3333 * needs ECN
3334 */
fa19d6bc
SH
3335 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
3336 * based on the path and may be
3337 * dependent on the congestion control
3338 * algorithm. In general it indicates
3339 * a congestion threshold. RTTs above
3340 * this indicate congestion
3341 */
07ed8df6
DA
3342 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
3343 * Arg1: value of icsk_retransmits
3344 * Arg2: value of icsk_rto
3345 * Arg3: whether RTO has expired
3346 */
3347 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
3348 * Arg1: sequence number of 1st byte
3349 * Arg2: # segments
3350 * Arg3: return value of
3351 * tcp_transmit_skb (0 => success)
3352 */
3353 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
3354 * Arg1: old_state
3355 * Arg2: new_state
3356 */
3eebc1d4
DA
3357 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
3358 * socket transition to LISTEN state.
3359 */
07ed8df6
DA
3360};
3361
3362/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
3363 * changes between the TCP and BPF versions. Ideally this should never happen.
3364 * If it does, we need to add code to convert them before calling
3365 * the BPF sock_ops function.
3366 */
3367enum {
3368 BPF_TCP_ESTABLISHED = 1,
3369 BPF_TCP_SYN_SENT,
3370 BPF_TCP_SYN_RECV,
3371 BPF_TCP_FIN_WAIT1,
3372 BPF_TCP_FIN_WAIT2,
3373 BPF_TCP_TIME_WAIT,
3374 BPF_TCP_CLOSE,
3375 BPF_TCP_CLOSE_WAIT,
3376 BPF_TCP_LAST_ACK,
3377 BPF_TCP_LISTEN,
3378 BPF_TCP_CLOSING, /* Now a valid state */
3379 BPF_TCP_NEW_SYN_RECV,
3380
3381 BPF_TCP_MAX_STATES /* Leave at the end! */
f0b9b795
SH
3382};
3383
3384#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
3385#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
3386
f53da99a
SH
3387struct bpf_perf_event_value {
3388 __u64 counter;
3389 __u64 enabled;
3390 __u64 running;
3391};
3392
665ef5a5
SH
3393#define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
3394#define BPF_DEVCG_ACC_READ (1ULL << 1)
3395#define BPF_DEVCG_ACC_WRITE (1ULL << 2)
3396
3397#define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
3398#define BPF_DEVCG_DEV_CHAR (1ULL << 1)
3399
3400struct bpf_cgroup_dev_ctx {
c0788a09
DA
3401 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
3402 __u32 access_type;
665ef5a5
SH
3403 __u32 major;
3404 __u32 minor;
3405};
3406
dcf7997b
SH
3407struct bpf_raw_tracepoint_args {
3408 __u64 args[0];
3409};
3410
4276e652
DA
3411/* DIRECT: Skip the FIB rules and go to FIB table associated with device
3412 * OUTPUT: Do lookup from egress perspective; default is ingress
3413 */
77a38037
SH
3414#define BPF_FIB_LOOKUP_DIRECT (1U << 0)
3415#define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
4276e652 3416
22ddbd82
DA
3417enum {
3418 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
3419 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
3420 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
3421 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
3422 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
3423 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
3424 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
3425 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
3426 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
3427};
3428
4276e652 3429struct bpf_fib_lookup {
17678d30
SH
3430 /* input: network family for lookup (AF_INET, AF_INET6)
3431 * output: network family of egress nexthop
3432 */
3433 __u8 family;
4276e652
DA
3434
3435 /* set if lookup is to consider L4 data - e.g., FIB rules */
3436 __u8 l4_protocol;
3437 __be16 sport;
3438 __be16 dport;
3439
3440 /* total length of packet from network header - used for MTU check */
3441 __u16 tot_len;
22ddbd82
DA
3442
3443 /* input: L3 device index for lookup
3444 * output: device index from FIB lookup
3445 */
3446 __u32 ifindex;
4276e652
DA
3447
3448 union {
3449 /* inputs to lookup */
3450 __u8 tos; /* AF_INET */
17678d30 3451 __be32 flowinfo; /* AF_INET6, flow_label + priority */
4276e652 3452
17678d30
SH
3453 /* output: metric of fib result (IPv4/IPv6 only) */
3454 __u32 rt_metric;
4276e652
DA
3455 };
3456
3457 union {
4276e652
DA
3458 __be32 ipv4_src;
3459 __u32 ipv6_src[4]; /* in6_addr; network order */
3460 };
3461
17678d30
SH
3462 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
3463 * network header. output: bpf_fib_lookup sets to gateway address
3464 * if FIB lookup returns gateway route
4276e652
DA
3465 */
3466 union {
4276e652
DA
3467 __be32 ipv4_dst;
3468 __u32 ipv6_dst[4]; /* in6_addr; network order */
3469 };
3470
3471 /* output */
3472 __be16 h_vlan_proto;
3473 __be16 h_vlan_TCI;
3474 __u8 smac[6]; /* ETH_ALEN */
3475 __u8 dmac[6]; /* ETH_ALEN */
3476};
3477
57ac202c
DA
3478enum bpf_task_fd_type {
3479 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3480 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3481 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3482 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3483 BPF_FD_TYPE_UPROBE, /* filename + offset */
3484 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3485};
3486
d9c0be4e
DA
3487struct bpf_flow_keys {
3488 __u16 nhoff;
3489 __u16 thoff;
3490 __u16 addr_proto; /* ETH_P_* of valid addrs */
3491 __u8 is_frag;
3492 __u8 is_first_frag;
3493 __u8 is_encap;
3494 __u8 ip_proto;
3495 __be16 n_proto;
3496 __be16 sport;
3497 __be16 dport;
3498 union {
3499 struct {
3500 __be32 ipv4_src;
3501 __be32 ipv4_dst;
3502 };
3503 struct {
3504 __u32 ipv6_src[4]; /* in6_addr; network order */
3505 __u32 ipv6_dst[4]; /* in6_addr; network order */
3506 };
3507 };
3508};
3509
17689d30
DA
3510struct bpf_func_info {
3511 __u32 insn_off;
3512 __u32 type_id;
3513};
3514
3515#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3516#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3517
3518struct bpf_line_info {
3519 __u32 insn_off;
3520 __u32 file_name_off;
3521 __u32 line_off;
3522 __u32 line_col;
3523};
3524
6b2d60bd
DA
3525struct bpf_spin_lock {
3526 __u32 val;
3527};
70de8a7f
DA
3528
3529struct bpf_sysctl {
3530 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
3531 * Allows 1,2,4-byte read, but no write.
3532 */
3533 __u32 file_pos; /* Sysctl file position to read from, write to.
3534 * Allows 1,2,4-byte read an 4-byte write.
3535 */
3536};
3537
b54ac87e 3538#endif /* __LINUX_BPF_H__ */