Merge branch 'rename-info_cnt-to-nr_info'
[linux-2.6-block.git] / include / uapi / linux / bpf.h
CommitLineData
e2be04c7 1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
daedfb22
AS
2/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 */
8#ifndef _UAPI__LINUX_BPF_H__
9#define _UAPI__LINUX_BPF_H__
10
11#include <linux/types.h>
c15952dc 12#include <linux/bpf_common.h>
daedfb22
AS
13
14/* Extended instruction set based on top of classic BPF */
15
16/* instruction classes */
17#define BPF_ALU64 0x07 /* alu mode in double word width */
18
19/* ld/ldx fields */
cb5f7334 20#define BPF_DW 0x18 /* double word (64-bit) */
daedfb22
AS
21#define BPF_XADD 0xc0 /* exclusive add */
22
23/* alu/jmp fields */
24#define BPF_MOV 0xb0 /* mov reg to reg */
25#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
26
27/* change endianness of a register */
28#define BPF_END 0xd0 /* flags for endianness conversion: */
29#define BPF_TO_LE 0x00 /* convert to little-endian */
30#define BPF_TO_BE 0x08 /* convert to big-endian */
31#define BPF_FROM_LE BPF_TO_LE
32#define BPF_FROM_BE BPF_TO_BE
33
92b31a9a 34/* jmp encodings */
daedfb22 35#define BPF_JNE 0x50 /* jump != */
92b31a9a
DB
36#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
daedfb22
AS
38#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
92b31a9a
DB
40#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
daedfb22
AS
42#define BPF_CALL 0x80 /* function call */
43#define BPF_EXIT 0x90 /* function return */
44
45/* Register numbers */
46enum {
47 BPF_REG_0 = 0,
48 BPF_REG_1,
49 BPF_REG_2,
50 BPF_REG_3,
51 BPF_REG_4,
52 BPF_REG_5,
53 BPF_REG_6,
54 BPF_REG_7,
55 BPF_REG_8,
56 BPF_REG_9,
57 BPF_REG_10,
58 __MAX_BPF_REG,
59};
60
61/* BPF has 10 general purpose 64-bit registers and stack frame. */
62#define MAX_BPF_REG __MAX_BPF_REG
63
64struct bpf_insn {
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
70};
71
b95a5c4d
DM
72/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
76};
77
de9cbbaa
RG
78struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
81};
82
b2197755 83/* BPF syscall commands, see bpf(2) man-page for details. */
99c55f7d 84enum bpf_cmd {
99c55f7d 85 BPF_MAP_CREATE,
db20fd2b 86 BPF_MAP_LOOKUP_ELEM,
db20fd2b 87 BPF_MAP_UPDATE_ELEM,
db20fd2b 88 BPF_MAP_DELETE_ELEM,
db20fd2b 89 BPF_MAP_GET_NEXT_KEY,
09756af4 90 BPF_PROG_LOAD,
b2197755
DB
91 BPF_OBJ_PIN,
92 BPF_OBJ_GET,
f4324551
DM
93 BPF_PROG_ATTACH,
94 BPF_PROG_DETACH,
1cf1cae9 95 BPF_PROG_TEST_RUN,
34ad5580
MKL
96 BPF_PROG_GET_NEXT_ID,
97 BPF_MAP_GET_NEXT_ID,
b16d9aa4 98 BPF_PROG_GET_FD_BY_ID,
bd5f5f4e 99 BPF_MAP_GET_FD_BY_ID,
1e270976 100 BPF_OBJ_GET_INFO_BY_FD,
468e2f64 101 BPF_PROG_QUERY,
c4f6699d 102 BPF_RAW_TRACEPOINT_OPEN,
f56a653c 103 BPF_BTF_LOAD,
78958fca 104 BPF_BTF_GET_FD_BY_ID,
41bdc4b4 105 BPF_TASK_FD_QUERY,
bd513cd0 106 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
99c55f7d
AS
107};
108
109enum bpf_map_type {
110 BPF_MAP_TYPE_UNSPEC,
0f8e4bd8 111 BPF_MAP_TYPE_HASH,
28fbcfa0 112 BPF_MAP_TYPE_ARRAY,
04fd61ab 113 BPF_MAP_TYPE_PROG_ARRAY,
ea317b26 114 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
824bd0ce 115 BPF_MAP_TYPE_PERCPU_HASH,
a10423b8 116 BPF_MAP_TYPE_PERCPU_ARRAY,
d5a3b1f6 117 BPF_MAP_TYPE_STACK_TRACE,
4ed8ec52 118 BPF_MAP_TYPE_CGROUP_ARRAY,
29ba732a 119 BPF_MAP_TYPE_LRU_HASH,
8f844938 120 BPF_MAP_TYPE_LRU_PERCPU_HASH,
b95a5c4d 121 BPF_MAP_TYPE_LPM_TRIE,
56f668df 122 BPF_MAP_TYPE_ARRAY_OF_MAPS,
bcc6b1b7 123 BPF_MAP_TYPE_HASH_OF_MAPS,
546ac1ff 124 BPF_MAP_TYPE_DEVMAP,
174a79ff 125 BPF_MAP_TYPE_SOCKMAP,
6710e112 126 BPF_MAP_TYPE_CPUMAP,
fbfc504a 127 BPF_MAP_TYPE_XSKMAP,
81110384 128 BPF_MAP_TYPE_SOCKHASH,
de9cbbaa 129 BPF_MAP_TYPE_CGROUP_STORAGE,
5dc4c4b7 130 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
b741f163 131 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
f1a2e44a
MV
132 BPF_MAP_TYPE_QUEUE,
133 BPF_MAP_TYPE_STACK,
99c55f7d
AS
134};
135
09756af4
AS
136enum bpf_prog_type {
137 BPF_PROG_TYPE_UNSPEC,
ddd872bc 138 BPF_PROG_TYPE_SOCKET_FILTER,
2541517c 139 BPF_PROG_TYPE_KPROBE,
96be4325 140 BPF_PROG_TYPE_SCHED_CLS,
94caee8c 141 BPF_PROG_TYPE_SCHED_ACT,
98b5c2c6 142 BPF_PROG_TYPE_TRACEPOINT,
6a773a15 143 BPF_PROG_TYPE_XDP,
0515e599 144 BPF_PROG_TYPE_PERF_EVENT,
0e33661d 145 BPF_PROG_TYPE_CGROUP_SKB,
61023658 146 BPF_PROG_TYPE_CGROUP_SOCK,
3a0af8fd
TG
147 BPF_PROG_TYPE_LWT_IN,
148 BPF_PROG_TYPE_LWT_OUT,
149 BPF_PROG_TYPE_LWT_XMIT,
40304b2a 150 BPF_PROG_TYPE_SOCK_OPS,
b005fd18 151 BPF_PROG_TYPE_SK_SKB,
ebc614f6 152 BPF_PROG_TYPE_CGROUP_DEVICE,
4f738adb 153 BPF_PROG_TYPE_SK_MSG,
c4f6699d 154 BPF_PROG_TYPE_RAW_TRACEPOINT,
4fbac77d 155 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
004d4b27 156 BPF_PROG_TYPE_LWT_SEG6LOCAL,
f4364dcf 157 BPF_PROG_TYPE_LIRC_MODE2,
2dbb9b9e 158 BPF_PROG_TYPE_SK_REUSEPORT,
d58e468b 159 BPF_PROG_TYPE_FLOW_DISSECTOR,
09756af4
AS
160};
161
0e33661d
DM
162enum bpf_attach_type {
163 BPF_CGROUP_INET_INGRESS,
164 BPF_CGROUP_INET_EGRESS,
61023658 165 BPF_CGROUP_INET_SOCK_CREATE,
40304b2a 166 BPF_CGROUP_SOCK_OPS,
464bc0fd
JF
167 BPF_SK_SKB_STREAM_PARSER,
168 BPF_SK_SKB_STREAM_VERDICT,
ebc614f6 169 BPF_CGROUP_DEVICE,
4f738adb 170 BPF_SK_MSG_VERDICT,
4fbac77d
AI
171 BPF_CGROUP_INET4_BIND,
172 BPF_CGROUP_INET6_BIND,
d74bad4e
AI
173 BPF_CGROUP_INET4_CONNECT,
174 BPF_CGROUP_INET6_CONNECT,
aac3fc32
AI
175 BPF_CGROUP_INET4_POST_BIND,
176 BPF_CGROUP_INET6_POST_BIND,
1cedee13
AI
177 BPF_CGROUP_UDP4_SENDMSG,
178 BPF_CGROUP_UDP6_SENDMSG,
f4364dcf 179 BPF_LIRC_MODE2,
d58e468b 180 BPF_FLOW_DISSECTOR,
0e33661d
DM
181 __MAX_BPF_ATTACH_TYPE
182};
183
184#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
185
324bda9e
AS
186/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
187 *
188 * NONE(default): No further bpf programs allowed in the subtree.
189 *
190 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
191 * the program in this cgroup yields to sub-cgroup program.
192 *
193 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
194 * that cgroup program gets run in addition to the program in this cgroup.
195 *
196 * Only one program is allowed to be attached to a cgroup with
197 * NONE or BPF_F_ALLOW_OVERRIDE flag.
198 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
199 * release old program and attach the new one. Attach flags has to match.
200 *
201 * Multiple programs are allowed to be attached to a cgroup with
202 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
203 * (those that were attached first, run first)
204 * The programs of sub-cgroup are executed first, then programs of
205 * this cgroup and then programs of parent cgroup.
206 * When children program makes decision (like picking TCP CA or sock bind)
207 * parent program has a chance to override it.
208 *
209 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
210 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
211 * Ex1:
212 * cgrp1 (MULTI progs A, B) ->
213 * cgrp2 (OVERRIDE prog C) ->
214 * cgrp3 (MULTI prog D) ->
215 * cgrp4 (OVERRIDE prog E) ->
216 * cgrp5 (NONE prog F)
217 * the event in cgrp5 triggers execution of F,D,A,B in that order.
218 * if prog F is detached, the execution is E,D,A,B
219 * if prog F and D are detached, the execution is E,A,B
220 * if prog F, E and D are detached, the execution is C,A,B
221 *
222 * All eligible programs are executed regardless of return code from
223 * earlier programs.
7f677633
AS
224 */
225#define BPF_F_ALLOW_OVERRIDE (1U << 0)
324bda9e 226#define BPF_F_ALLOW_MULTI (1U << 1)
7f677633 227
e07b98d9
DM
228/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
229 * verifier will perform strict alignment checking as if the kernel
230 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
231 * and NET_IP_ALIGN defined to 2.
232 */
233#define BPF_F_STRICT_ALIGNMENT (1U << 0)
234
e9ee9efc
DM
235/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
236 * verifier will allow any alignment whatsoever. On platforms
237 * with strict alignment requirements for loads ands stores (such
238 * as sparc and mips) the verifier validates that all loads and
239 * stores provably follow this requirement. This flag turns that
240 * checking and enforcement off.
241 *
242 * It is mostly used for testing when we want to validate the
243 * context and memory access aspects of the verifier, but because
244 * of an unaligned access the alignment check would trigger before
245 * the one we are interested in.
246 */
247#define BPF_F_ANY_ALIGNMENT (1U << 1)
248
cc8b0b92 249/* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
f1a66f85
DB
250#define BPF_PSEUDO_MAP_FD 1
251
cc8b0b92
AS
252/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
253 * offset to another bpf function
254 */
255#define BPF_PSEUDO_CALL 1
256
3274f520
AS
257/* flags for BPF_MAP_UPDATE_ELEM command */
258#define BPF_ANY 0 /* create new element or update existing */
259#define BPF_NOEXIST 1 /* create new element if it didn't exist */
260#define BPF_EXIST 2 /* update existing element */
261
96eabe7a 262/* flags for BPF_MAP_CREATE command */
6c905981 263#define BPF_F_NO_PREALLOC (1U << 0)
29ba732a 264/* Instead of having one common LRU list in the
8f844938 265 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
29ba732a
MKL
266 * which can scale and perform better.
267 * Note, the LRU nodes (including free nodes) cannot be moved
268 * across different LRU lists.
269 */
270#define BPF_F_NO_COMMON_LRU (1U << 1)
96eabe7a
MKL
271/* Specify numa node during map creation */
272#define BPF_F_NUMA_NODE (1U << 2)
6c905981 273
cb4d2b3f
MKL
274#define BPF_OBJ_NAME_LEN 16U
275
6e71b04a
CF
276/* Flags for accessing BPF object */
277#define BPF_F_RDONLY (1U << 3)
278#define BPF_F_WRONLY (1U << 4)
279
615755a7
SL
280/* Flag for stack_map, store build_id+offset instead of pointer */
281#define BPF_F_STACK_BUILD_ID (1U << 5)
282
96b3b6c9
LB
283/* Zero-initialize hash function seed. This should only be used for testing. */
284#define BPF_F_ZERO_SEED (1U << 6)
285
2f183360
LB
286/* flags for BPF_PROG_QUERY */
287#define BPF_F_QUERY_EFFECTIVE (1U << 0)
288
615755a7
SL
289enum bpf_stack_build_id_status {
290 /* user space need an empty entry to identify end of a trace */
291 BPF_STACK_BUILD_ID_EMPTY = 0,
292 /* with valid build_id and offset */
293 BPF_STACK_BUILD_ID_VALID = 1,
294 /* couldn't get build_id, fallback to ip */
295 BPF_STACK_BUILD_ID_IP = 2,
296};
297
298#define BPF_BUILD_ID_SIZE 20
299struct bpf_stack_build_id {
300 __s32 status;
301 unsigned char build_id[BPF_BUILD_ID_SIZE];
302 union {
303 __u64 offset;
304 __u64 ip;
305 };
306};
307
99c55f7d
AS
308union bpf_attr {
309 struct { /* anonymous struct used by BPF_MAP_CREATE command */
310 __u32 map_type; /* one of enum bpf_map_type */
311 __u32 key_size; /* size of key in bytes */
312 __u32 value_size; /* size of value in bytes */
313 __u32 max_entries; /* max number of entries in a map */
96eabe7a
MKL
314 __u32 map_flags; /* BPF_MAP_CREATE related
315 * flags defined above.
316 */
56f668df 317 __u32 inner_map_fd; /* fd pointing to the inner map */
96eabe7a
MKL
318 __u32 numa_node; /* numa node (effective only if
319 * BPF_F_NUMA_NODE is set).
320 */
067cae47 321 char map_name[BPF_OBJ_NAME_LEN];
a3884572 322 __u32 map_ifindex; /* ifindex of netdev to create on */
a26ca7c9 323 __u32 btf_fd; /* fd pointing to a BTF type data */
9b2cf328
MKL
324 __u32 btf_key_type_id; /* BTF type_id of the key */
325 __u32 btf_value_type_id; /* BTF type_id of the value */
99c55f7d 326 };
db20fd2b
AS
327
328 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
329 __u32 map_fd;
330 __aligned_u64 key;
331 union {
332 __aligned_u64 value;
333 __aligned_u64 next_key;
334 };
3274f520 335 __u64 flags;
db20fd2b 336 };
09756af4
AS
337
338 struct { /* anonymous struct used by BPF_PROG_LOAD command */
339 __u32 prog_type; /* one of enum bpf_prog_type */
340 __u32 insn_cnt;
341 __aligned_u64 insns;
342 __aligned_u64 license;
cbd35700
AS
343 __u32 log_level; /* verbosity level of verifier */
344 __u32 log_size; /* size of user buffer */
345 __aligned_u64 log_buf; /* user supplied buffer */
2541517c 346 __u32 kern_version; /* checked when prog_type=kprobe */
e07b98d9 347 __u32 prog_flags;
067cae47 348 char prog_name[BPF_OBJ_NAME_LEN];
1f6f4cb7 349 __u32 prog_ifindex; /* ifindex of netdev to prep for */
5e43f899
AI
350 /* For some prog types expected attach type must be known at
351 * load time to verify attach type specific parts of prog
352 * (context accesses, allowed helpers, etc).
353 */
354 __u32 expected_attach_type;
838e9690
YS
355 __u32 prog_btf_fd; /* fd pointing to BTF type data */
356 __u32 func_info_rec_size; /* userspace bpf_func_info size */
357 __aligned_u64 func_info; /* func info */
358 __u32 func_info_cnt; /* number of bpf_func_info records */
c454a46b
MKL
359 __u32 line_info_rec_size; /* userspace bpf_line_info size */
360 __aligned_u64 line_info; /* line info */
361 __u32 line_info_cnt; /* number of bpf_line_info records */
09756af4 362 };
b2197755
DB
363
364 struct { /* anonymous struct used by BPF_OBJ_* commands */
365 __aligned_u64 pathname;
366 __u32 bpf_fd;
6e71b04a 367 __u32 file_flags;
b2197755 368 };
f4324551
DM
369
370 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
371 __u32 target_fd; /* container object to attach to */
372 __u32 attach_bpf_fd; /* eBPF program to attach */
373 __u32 attach_type;
7f677633 374 __u32 attach_flags;
f4324551 375 };
1cf1cae9
AS
376
377 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
378 __u32 prog_fd;
379 __u32 retval;
b5a36b1e
LB
380 __u32 data_size_in; /* input: len of data_in */
381 __u32 data_size_out; /* input/output: len of data_out
382 * returns ENOSPC if data_out
383 * is too small.
384 */
1cf1cae9
AS
385 __aligned_u64 data_in;
386 __aligned_u64 data_out;
387 __u32 repeat;
388 __u32 duration;
389 } test;
34ad5580 390
b16d9aa4
MKL
391 struct { /* anonymous struct used by BPF_*_GET_*_ID */
392 union {
393 __u32 start_id;
394 __u32 prog_id;
bd5f5f4e 395 __u32 map_id;
78958fca 396 __u32 btf_id;
b16d9aa4 397 };
34ad5580 398 __u32 next_id;
6e71b04a 399 __u32 open_flags;
34ad5580 400 };
1e270976
MKL
401
402 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
403 __u32 bpf_fd;
404 __u32 info_len;
405 __aligned_u64 info;
406 } info;
468e2f64
AS
407
408 struct { /* anonymous struct used by BPF_PROG_QUERY command */
409 __u32 target_fd; /* container object to query */
410 __u32 attach_type;
411 __u32 query_flags;
412 __u32 attach_flags;
413 __aligned_u64 prog_ids;
414 __u32 prog_cnt;
415 } query;
c4f6699d
AS
416
417 struct {
418 __u64 name;
419 __u32 prog_fd;
420 } raw_tracepoint;
f56a653c
MKL
421
422 struct { /* anonymous struct for BPF_BTF_LOAD */
423 __aligned_u64 btf;
424 __aligned_u64 btf_log_buf;
425 __u32 btf_size;
426 __u32 btf_log_size;
427 __u32 btf_log_level;
428 };
41bdc4b4
YS
429
430 struct {
431 __u32 pid; /* input: pid */
432 __u32 fd; /* input: fd */
433 __u32 flags; /* input: flags */
434 __u32 buf_len; /* input/output: buf len */
435 __aligned_u64 buf; /* input/output:
436 * tp_name for tracepoint
437 * symbol for kprobe
438 * filename for uprobe
439 */
440 __u32 prog_id; /* output: prod_id */
441 __u32 fd_type; /* output: BPF_FD_TYPE_* */
442 __u64 probe_offset; /* output: probe_offset */
443 __u64 probe_addr; /* output: probe_addr */
444 } task_fd_query;
99c55f7d
AS
445} __attribute__((aligned(8)));
446
56a092c8
QM
447/* The description below is an attempt at providing documentation to eBPF
448 * developers about the multiple available eBPF helper functions. It can be
449 * parsed and used to produce a manual page. The workflow is the following,
450 * and requires the rst2man utility:
451 *
452 * $ ./scripts/bpf_helpers_doc.py \
453 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
454 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
455 * $ man /tmp/bpf-helpers.7
456 *
457 * Note that in order to produce this external documentation, some RST
458 * formatting is used in the descriptions to get "bold" and "italics" in
459 * manual pages. Also note that the few trailing white spaces are
460 * intentional, removing them would break paragraphs for rst2man.
461 *
462 * Start of BPF helper function descriptions:
ad4a5223
QM
463 *
464 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
465 * Description
466 * Perform a lookup in *map* for an entry associated to *key*.
467 * Return
468 * Map value associated to *key*, or **NULL** if no entry was
469 * found.
470 *
471 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
472 * Description
473 * Add or update the value of the entry associated to *key* in
474 * *map* with *value*. *flags* is one of:
475 *
476 * **BPF_NOEXIST**
477 * The entry for *key* must not exist in the map.
478 * **BPF_EXIST**
479 * The entry for *key* must already exist in the map.
480 * **BPF_ANY**
481 * No condition on the existence of the entry for *key*.
482 *
483 * Flag value **BPF_NOEXIST** cannot be used for maps of types
484 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
485 * elements always exist), the helper would return an error.
486 * Return
487 * 0 on success, or a negative error in case of failure.
488 *
489 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
490 * Description
491 * Delete entry with *key* from *map*.
492 * Return
493 * 0 on success, or a negative error in case of failure.
494 *
f1a2e44a
MV
495 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
496 * Description
497 * Push an element *value* in *map*. *flags* is one of:
498 *
499 * **BPF_EXIST**
500 * If the queue/stack is full, the oldest element is removed to
501 * make room for this.
502 * Return
503 * 0 on success, or a negative error in case of failure.
504 *
ad4a5223
QM
505 * int bpf_probe_read(void *dst, u32 size, const void *src)
506 * Description
507 * For tracing programs, safely attempt to read *size* bytes from
508 * address *src* and store the data in *dst*.
509 * Return
510 * 0 on success, or a negative error in case of failure.
511 *
512 * u64 bpf_ktime_get_ns(void)
513 * Description
514 * Return the time elapsed since system boot, in nanoseconds.
515 * Return
516 * Current *ktime*.
517 *
518 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
519 * Description
520 * This helper is a "printk()-like" facility for debugging. It
521 * prints a message defined by format *fmt* (of size *fmt_size*)
522 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
523 * available. It can take up to three additional **u64**
524 * arguments (as an eBPF helpers, the total number of arguments is
525 * limited to five).
526 *
527 * Each time the helper is called, it appends a line to the trace.
528 * The format of the trace is customizable, and the exact output
529 * one will get depends on the options set in
530 * *\/sys/kernel/debug/tracing/trace_options* (see also the
531 * *README* file under the same directory). However, it usually
532 * defaults to something like:
533 *
534 * ::
535 *
536 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
537 *
538 * In the above:
539 *
540 * * ``telnet`` is the name of the current task.
541 * * ``470`` is the PID of the current task.
542 * * ``001`` is the CPU number on which the task is
543 * running.
544 * * In ``.N..``, each character refers to a set of
545 * options (whether irqs are enabled, scheduling
546 * options, whether hard/softirqs are running, level of
547 * preempt_disabled respectively). **N** means that
548 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
549 * are set.
550 * * ``419421.045894`` is a timestamp.
551 * * ``0x00000001`` is a fake value used by BPF for the
552 * instruction pointer register.
553 * * ``<formatted msg>`` is the message formatted with
554 * *fmt*.
555 *
556 * The conversion specifiers supported by *fmt* are similar, but
557 * more limited than for printk(). They are **%d**, **%i**,
558 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
559 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
560 * of field, padding with zeroes, etc.) is available, and the
561 * helper will return **-EINVAL** (but print nothing) if it
562 * encounters an unknown specifier.
563 *
564 * Also, note that **bpf_trace_printk**\ () is slow, and should
565 * only be used for debugging purposes. For this reason, a notice
566 * bloc (spanning several lines) is printed to kernel logs and
567 * states that the helper should not be used "for production use"
568 * the first time this helper is used (or more precisely, when
569 * **trace_printk**\ () buffers are allocated). For passing values
570 * to user space, perf events should be preferred.
571 * Return
572 * The number of bytes written to the buffer, or a negative error
573 * in case of failure.
574 *
1fdd08be
QM
575 * u32 bpf_get_prandom_u32(void)
576 * Description
577 * Get a pseudo-random number.
578 *
579 * From a security point of view, this helper uses its own
580 * pseudo-random internal state, and cannot be used to infer the
581 * seed of other random functions in the kernel. However, it is
582 * essential to note that the generator used by the helper is not
583 * cryptographically secure.
584 * Return
585 * A random 32-bit unsigned value.
586 *
587 * u32 bpf_get_smp_processor_id(void)
588 * Description
589 * Get the SMP (symmetric multiprocessing) processor id. Note that
590 * all programs run with preemption disabled, which means that the
591 * SMP processor id is stable during all the execution of the
592 * program.
593 * Return
594 * The SMP id of the processor running the program.
595 *
ad4a5223
QM
596 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
597 * Description
598 * Store *len* bytes from address *from* into the packet
599 * associated to *skb*, at *offset*. *flags* are a combination of
600 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
601 * checksum for the packet after storing the bytes) and
602 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
603 * **->swhash** and *skb*\ **->l4hash** to 0).
604 *
605 * A call to this helper is susceptible to change the underlaying
606 * packet buffer. Therefore, at load time, all checks on pointers
607 * previously done by the verifier are invalidated and must be
608 * performed again, if the helper is used in combination with
609 * direct packet access.
610 * Return
611 * 0 on success, or a negative error in case of failure.
612 *
613 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
614 * Description
615 * Recompute the layer 3 (e.g. IP) checksum for the packet
616 * associated to *skb*. Computation is incremental, so the helper
617 * must know the former value of the header field that was
618 * modified (*from*), the new value of this field (*to*), and the
619 * number of bytes (2 or 4) for this field, stored in *size*.
620 * Alternatively, it is possible to store the difference between
621 * the previous and the new values of the header field in *to*, by
622 * setting *from* and *size* to 0. For both methods, *offset*
623 * indicates the location of the IP checksum within the packet.
624 *
625 * This helper works in combination with **bpf_csum_diff**\ (),
626 * which does not update the checksum in-place, but offers more
627 * flexibility and can handle sizes larger than 2 or 4 for the
628 * checksum to update.
629 *
630 * A call to this helper is susceptible to change the underlaying
631 * packet buffer. Therefore, at load time, all checks on pointers
632 * previously done by the verifier are invalidated and must be
633 * performed again, if the helper is used in combination with
634 * direct packet access.
635 * Return
636 * 0 on success, or a negative error in case of failure.
637 *
638 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
639 * Description
640 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
641 * packet associated to *skb*. Computation is incremental, so the
642 * helper must know the former value of the header field that was
643 * modified (*from*), the new value of this field (*to*), and the
644 * number of bytes (2 or 4) for this field, stored on the lowest
645 * four bits of *flags*. Alternatively, it is possible to store
646 * the difference between the previous and the new values of the
647 * header field in *to*, by setting *from* and the four lowest
648 * bits of *flags* to 0. For both methods, *offset* indicates the
649 * location of the IP checksum within the packet. In addition to
650 * the size of the field, *flags* can be added (bitwise OR) actual
651 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
652 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
653 * for updates resulting in a null checksum the value is set to
654 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
655 * the checksum is to be computed against a pseudo-header.
656 *
657 * This helper works in combination with **bpf_csum_diff**\ (),
658 * which does not update the checksum in-place, but offers more
659 * flexibility and can handle sizes larger than 2 or 4 for the
660 * checksum to update.
661 *
662 * A call to this helper is susceptible to change the underlaying
663 * packet buffer. Therefore, at load time, all checks on pointers
664 * previously done by the verifier are invalidated and must be
665 * performed again, if the helper is used in combination with
666 * direct packet access.
667 * Return
668 * 0 on success, or a negative error in case of failure.
669 *
670 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
671 * Description
672 * This special helper is used to trigger a "tail call", or in
673 * other words, to jump into another eBPF program. The same stack
674 * frame is used (but values on stack and in registers for the
675 * caller are not accessible to the callee). This mechanism allows
676 * for program chaining, either for raising the maximum number of
677 * available eBPF instructions, or to execute given programs in
678 * conditional blocks. For security reasons, there is an upper
679 * limit to the number of successive tail calls that can be
680 * performed.
681 *
682 * Upon call of this helper, the program attempts to jump into a
683 * program referenced at index *index* in *prog_array_map*, a
684 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
685 * *ctx*, a pointer to the context.
686 *
687 * If the call succeeds, the kernel immediately runs the first
688 * instruction of the new program. This is not a function call,
689 * and it never returns to the previous program. If the call
690 * fails, then the helper has no effect, and the caller continues
691 * to run its subsequent instructions. A call can fail if the
692 * destination program for the jump does not exist (i.e. *index*
693 * is superior to the number of entries in *prog_array_map*), or
694 * if the maximum number of tail calls has been reached for this
695 * chain of programs. This limit is defined in the kernel by the
696 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
697 * which is currently set to 32.
698 * Return
699 * 0 on success, or a negative error in case of failure.
700 *
701 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
702 * Description
703 * Clone and redirect the packet associated to *skb* to another
704 * net device of index *ifindex*. Both ingress and egress
705 * interfaces can be used for redirection. The **BPF_F_INGRESS**
706 * value in *flags* is used to make the distinction (ingress path
707 * is selected if the flag is present, egress path otherwise).
708 * This is the only flag supported for now.
709 *
710 * In comparison with **bpf_redirect**\ () helper,
711 * **bpf_clone_redirect**\ () has the associated cost of
712 * duplicating the packet buffer, but this can be executed out of
713 * the eBPF program. Conversely, **bpf_redirect**\ () is more
714 * efficient, but it is handled through an action code where the
715 * redirection happens only after the eBPF program has returned.
716 *
717 * A call to this helper is susceptible to change the underlaying
718 * packet buffer. Therefore, at load time, all checks on pointers
719 * previously done by the verifier are invalidated and must be
720 * performed again, if the helper is used in combination with
721 * direct packet access.
722 * Return
723 * 0 on success, or a negative error in case of failure.
c456dec4
QM
724 *
725 * u64 bpf_get_current_pid_tgid(void)
726 * Return
727 * A 64-bit integer containing the current tgid and pid, and
728 * created as such:
729 * *current_task*\ **->tgid << 32 \|**
730 * *current_task*\ **->pid**.
731 *
732 * u64 bpf_get_current_uid_gid(void)
733 * Return
734 * A 64-bit integer containing the current GID and UID, and
735 * created as such: *current_gid* **<< 32 \|** *current_uid*.
736 *
737 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
738 * Description
739 * Copy the **comm** attribute of the current task into *buf* of
740 * *size_of_buf*. The **comm** attribute contains the name of
741 * the executable (excluding the path) for the current task. The
742 * *size_of_buf* must be strictly positive. On success, the
743 * helper makes sure that the *buf* is NUL-terminated. On failure,
744 * it is filled with zeroes.
745 * Return
746 * 0 on success, or a negative error in case of failure.
747 *
1fdd08be
QM
748 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
749 * Description
750 * Retrieve the classid for the current task, i.e. for the net_cls
751 * cgroup to which *skb* belongs.
752 *
753 * This helper can be used on TC egress path, but not on ingress.
754 *
755 * The net_cls cgroup provides an interface to tag network packets
756 * based on a user-provided identifier for all traffic coming from
757 * the tasks belonging to the related cgroup. See also the related
758 * kernel documentation, available from the Linux sources in file
759 * *Documentation/cgroup-v1/net_cls.txt*.
760 *
761 * The Linux kernel has two versions for cgroups: there are
762 * cgroups v1 and cgroups v2. Both are available to users, who can
763 * use a mixture of them, but note that the net_cls cgroup is for
764 * cgroup v1 only. This makes it incompatible with BPF programs
765 * run on cgroups, which is a cgroup-v2-only feature (a socket can
766 * only hold data for one version of cgroups at a time).
767 *
768 * This helper is only available is the kernel was compiled with
769 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
770 * "**y**" or to "**m**".
771 * Return
772 * The classid, or 0 for the default unconfigured classid.
773 *
c456dec4
QM
774 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
775 * Description
776 * Push a *vlan_tci* (VLAN tag control information) of protocol
777 * *vlan_proto* to the packet associated to *skb*, then update
778 * the checksum. Note that if *vlan_proto* is different from
779 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
780 * be **ETH_P_8021Q**.
781 *
782 * A call to this helper is susceptible to change the underlaying
783 * packet buffer. Therefore, at load time, all checks on pointers
784 * previously done by the verifier are invalidated and must be
785 * performed again, if the helper is used in combination with
786 * direct packet access.
787 * Return
788 * 0 on success, or a negative error in case of failure.
789 *
790 * int bpf_skb_vlan_pop(struct sk_buff *skb)
791 * Description
792 * Pop a VLAN header from the packet associated to *skb*.
793 *
794 * A call to this helper is susceptible to change the underlaying
795 * packet buffer. Therefore, at load time, all checks on pointers
796 * previously done by the verifier are invalidated and must be
797 * performed again, if the helper is used in combination with
798 * direct packet access.
799 * Return
800 * 0 on success, or a negative error in case of failure.
801 *
802 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
803 * Description
804 * Get tunnel metadata. This helper takes a pointer *key* to an
805 * empty **struct bpf_tunnel_key** of **size**, that will be
806 * filled with tunnel metadata for the packet associated to *skb*.
807 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
808 * indicates that the tunnel is based on IPv6 protocol instead of
809 * IPv4.
810 *
811 * The **struct bpf_tunnel_key** is an object that generalizes the
812 * principal parameters used by various tunneling protocols into a
813 * single struct. This way, it can be used to easily make a
814 * decision based on the contents of the encapsulation header,
815 * "summarized" in this struct. In particular, it holds the IP
816 * address of the remote end (IPv4 or IPv6, depending on the case)
817 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
818 * this struct exposes the *key*\ **->tunnel_id**, which is
819 * generally mapped to a VNI (Virtual Network Identifier), making
820 * it programmable together with the **bpf_skb_set_tunnel_key**\
821 * () helper.
822 *
823 * Let's imagine that the following code is part of a program
824 * attached to the TC ingress interface, on one end of a GRE
825 * tunnel, and is supposed to filter out all messages coming from
826 * remote ends with IPv4 address other than 10.0.0.1:
827 *
828 * ::
829 *
830 * int ret;
831 * struct bpf_tunnel_key key = {};
832 *
833 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
834 * if (ret < 0)
835 * return TC_ACT_SHOT; // drop packet
836 *
837 * if (key.remote_ipv4 != 0x0a000001)
838 * return TC_ACT_SHOT; // drop packet
839 *
840 * return TC_ACT_OK; // accept packet
841 *
842 * This interface can also be used with all encapsulation devices
843 * that can operate in "collect metadata" mode: instead of having
844 * one network device per specific configuration, the "collect
845 * metadata" mode only requires a single device where the
846 * configuration can be extracted from this helper.
847 *
848 * This can be used together with various tunnels such as VXLan,
849 * Geneve, GRE or IP in IP (IPIP).
850 * Return
851 * 0 on success, or a negative error in case of failure.
852 *
853 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
854 * Description
855 * Populate tunnel metadata for packet associated to *skb.* The
856 * tunnel metadata is set to the contents of *key*, of *size*. The
857 * *flags* can be set to a combination of the following values:
858 *
859 * **BPF_F_TUNINFO_IPV6**
860 * Indicate that the tunnel is based on IPv6 protocol
861 * instead of IPv4.
862 * **BPF_F_ZERO_CSUM_TX**
863 * For IPv4 packets, add a flag to tunnel metadata
864 * indicating that checksum computation should be skipped
865 * and checksum set to zeroes.
866 * **BPF_F_DONT_FRAGMENT**
867 * Add a flag to tunnel metadata indicating that the
868 * packet should not be fragmented.
869 * **BPF_F_SEQ_NUMBER**
870 * Add a flag to tunnel metadata indicating that a
871 * sequence number should be added to tunnel header before
872 * sending the packet. This flag was added for GRE
873 * encapsulation, but might be used with other protocols
874 * as well in the future.
875 *
876 * Here is a typical usage on the transmit path:
877 *
878 * ::
879 *
880 * struct bpf_tunnel_key key;
881 * populate key ...
882 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
883 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
884 *
885 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
886 * helper for additional information.
887 * Return
888 * 0 on success, or a negative error in case of failure.
889 *
c6b5fb86
QM
890 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
891 * Description
892 * Read the value of a perf event counter. This helper relies on a
893 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
894 * the perf event counter is selected when *map* is updated with
895 * perf event file descriptors. The *map* is an array whose size
896 * is the number of available CPUs, and each cell contains a value
897 * relative to one CPU. The value to retrieve is indicated by
898 * *flags*, that contains the index of the CPU to look up, masked
899 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
900 * **BPF_F_CURRENT_CPU** to indicate that the value for the
901 * current CPU should be retrieved.
902 *
903 * Note that before Linux 4.13, only hardware perf event can be
904 * retrieved.
905 *
906 * Also, be aware that the newer helper
907 * **bpf_perf_event_read_value**\ () is recommended over
3bd5a09b 908 * **bpf_perf_event_read**\ () in general. The latter has some ABI
c6b5fb86
QM
909 * quirks where error and counter value are used as a return code
910 * (which is wrong to do since ranges may overlap). This issue is
3bd5a09b
QM
911 * fixed with **bpf_perf_event_read_value**\ (), which at the same
912 * time provides more features over the **bpf_perf_event_read**\
913 * () interface. Please refer to the description of
c6b5fb86
QM
914 * **bpf_perf_event_read_value**\ () for details.
915 * Return
916 * The value of the perf event counter read from the map, or a
917 * negative error code in case of failure.
918 *
c456dec4
QM
919 * int bpf_redirect(u32 ifindex, u64 flags)
920 * Description
921 * Redirect the packet to another net device of index *ifindex*.
922 * This helper is somewhat similar to **bpf_clone_redirect**\
923 * (), except that the packet is not cloned, which provides
924 * increased performance.
925 *
926 * Except for XDP, both ingress and egress interfaces can be used
927 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
928 * to make the distinction (ingress path is selected if the flag
929 * is present, egress path otherwise). Currently, XDP only
930 * supports redirection to the egress interface, and accepts no
931 * flag at all.
932 *
933 * The same effect can be attained with the more generic
934 * **bpf_redirect_map**\ (), which requires specific maps to be
935 * used but offers better performance.
936 * Return
937 * For XDP, the helper returns **XDP_REDIRECT** on success or
938 * **XDP_ABORTED** on error. For other program types, the values
939 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
940 * error.
941 *
1fdd08be
QM
942 * u32 bpf_get_route_realm(struct sk_buff *skb)
943 * Description
944 * Retrieve the realm or the route, that is to say the
945 * **tclassid** field of the destination for the *skb*. The
946 * indentifier retrieved is a user-provided tag, similar to the
947 * one used with the net_cls cgroup (see description for
948 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
949 * held by a route (a destination entry), not by a task.
950 *
951 * Retrieving this identifier works with the clsact TC egress hook
952 * (see also **tc-bpf(8)**), or alternatively on conventional
953 * classful egress qdiscs, but not on TC ingress path. In case of
954 * clsact TC egress hook, this has the advantage that, internally,
955 * the destination entry has not been dropped yet in the transmit
956 * path. Therefore, the destination entry does not need to be
957 * artificially held via **netif_keep_dst**\ () for a classful
958 * qdisc until the *skb* is freed.
959 *
960 * This helper is available only if the kernel was compiled with
961 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
962 * Return
963 * The realm of the route for the packet associated to *skb*, or 0
964 * if none was found.
965 *
c456dec4
QM
966 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
967 * Description
968 * Write raw *data* blob into a special BPF perf event held by
969 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
970 * event must have the following attributes: **PERF_SAMPLE_RAW**
971 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
972 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
973 *
974 * The *flags* are used to indicate the index in *map* for which
975 * the value must be put, masked with **BPF_F_INDEX_MASK**.
976 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
977 * to indicate that the index of the current CPU core should be
978 * used.
979 *
980 * The value to write, of *size*, is passed through eBPF stack and
981 * pointed by *data*.
982 *
983 * The context of the program *ctx* needs also be passed to the
984 * helper.
985 *
986 * On user space, a program willing to read the values needs to
987 * call **perf_event_open**\ () on the perf event (either for
988 * one or for all CPUs) and to store the file descriptor into the
989 * *map*. This must be done before the eBPF program can send data
990 * into it. An example is available in file
991 * *samples/bpf/trace_output_user.c* in the Linux kernel source
992 * tree (the eBPF program counterpart is in
993 * *samples/bpf/trace_output_kern.c*).
994 *
995 * **bpf_perf_event_output**\ () achieves better performance
996 * than **bpf_trace_printk**\ () for sharing data with user
997 * space, and is much better suitable for streaming data from eBPF
998 * programs.
999 *
1000 * Note that this helper is not restricted to tracing use cases
1001 * and can be used with programs attached to TC or XDP as well,
1002 * where it allows for passing data to user space listeners. Data
1003 * can be:
1004 *
1005 * * Only custom structs,
1006 * * Only the packet payload, or
1007 * * A combination of both.
1008 * Return
1009 * 0 on success, or a negative error in case of failure.
1010 *
1fdd08be
QM
1011 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1012 * Description
1013 * This helper was provided as an easy way to load data from a
1014 * packet. It can be used to load *len* bytes from *offset* from
1015 * the packet associated to *skb*, into the buffer pointed by
1016 * *to*.
1017 *
1018 * Since Linux 4.7, usage of this helper has mostly been replaced
1019 * by "direct packet access", enabling packet data to be
1020 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1021 * pointing respectively to the first byte of packet data and to
1022 * the byte after the last byte of packet data. However, it
1023 * remains useful if one wishes to read large quantities of data
1024 * at once from a packet into the eBPF stack.
1025 * Return
1026 * 0 on success, or a negative error in case of failure.
1027 *
c456dec4
QM
1028 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1029 * Description
1030 * Walk a user or a kernel stack and return its id. To achieve
1031 * this, the helper needs *ctx*, which is a pointer to the context
1032 * on which the tracing program is executed, and a pointer to a
1033 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1034 *
1035 * The last argument, *flags*, holds the number of stack frames to
1036 * skip (from 0 to 255), masked with
1037 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1038 * a combination of the following flags:
1039 *
1040 * **BPF_F_USER_STACK**
1041 * Collect a user space stack instead of a kernel stack.
1042 * **BPF_F_FAST_STACK_CMP**
1043 * Compare stacks by hash only.
1044 * **BPF_F_REUSE_STACKID**
1045 * If two different stacks hash into the same *stackid*,
1046 * discard the old one.
1047 *
1048 * The stack id retrieved is a 32 bit long integer handle which
1049 * can be further combined with other data (including other stack
1050 * ids) and used as a key into maps. This can be useful for
1051 * generating a variety of graphs (such as flame graphs or off-cpu
1052 * graphs).
1053 *
1054 * For walking a stack, this helper is an improvement over
1055 * **bpf_probe_read**\ (), which can be used with unrolled loops
1056 * but is not efficient and consumes a lot of eBPF instructions.
1057 * Instead, **bpf_get_stackid**\ () can collect up to
1058 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1059 * this limit can be controlled with the **sysctl** program, and
1060 * that it should be manually increased in order to profile long
1061 * user stacks (such as stacks for Java programs). To do so, use:
1062 *
1063 * ::
1064 *
1065 * # sysctl kernel.perf_event_max_stack=<new value>
c456dec4
QM
1066 * Return
1067 * The positive or null stack id on success, or a negative error
1068 * in case of failure.
1069 *
1fdd08be
QM
1070 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1071 * Description
1072 * Compute a checksum difference, from the raw buffer pointed by
1073 * *from*, of length *from_size* (that must be a multiple of 4),
1074 * towards the raw buffer pointed by *to*, of size *to_size*
1075 * (same remark). An optional *seed* can be added to the value
1076 * (this can be cascaded, the seed may come from a previous call
1077 * to the helper).
1078 *
1079 * This is flexible enough to be used in several ways:
1080 *
1081 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1082 * checksum, it can be used when pushing new data.
1083 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1084 * checksum, it can be used when removing data from a packet.
1085 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1086 * can be used to compute a diff. Note that *from_size* and
1087 * *to_size* do not need to be equal.
1088 *
1089 * This helper can be used in combination with
1090 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1091 * which one can feed in the difference computed with
1092 * **bpf_csum_diff**\ ().
1093 * Return
1094 * The checksum result, or a negative error code in case of
1095 * failure.
1096 *
1097 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1098 * Description
1099 * Retrieve tunnel options metadata for the packet associated to
1100 * *skb*, and store the raw tunnel option data to the buffer *opt*
1101 * of *size*.
1102 *
1103 * This helper can be used with encapsulation devices that can
1104 * operate in "collect metadata" mode (please refer to the related
1105 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1106 * more details). A particular example where this can be used is
1107 * in combination with the Geneve encapsulation protocol, where it
1108 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1109 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1110 * the eBPF program. This allows for full customization of these
1111 * headers.
1112 * Return
1113 * The size of the option data retrieved.
1114 *
1115 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1116 * Description
1117 * Set tunnel options metadata for the packet associated to *skb*
1118 * to the option data contained in the raw buffer *opt* of *size*.
1119 *
1120 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1121 * helper for additional information.
1122 * Return
1123 * 0 on success, or a negative error in case of failure.
1124 *
1125 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1126 * Description
1127 * Change the protocol of the *skb* to *proto*. Currently
1128 * supported are transition from IPv4 to IPv6, and from IPv6 to
1129 * IPv4. The helper takes care of the groundwork for the
1130 * transition, including resizing the socket buffer. The eBPF
1131 * program is expected to fill the new headers, if any, via
1132 * **skb_store_bytes**\ () and to recompute the checksums with
1133 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1134 * (). The main case for this helper is to perform NAT64
1135 * operations out of an eBPF program.
1136 *
1137 * Internally, the GSO type is marked as dodgy so that headers are
1138 * checked and segments are recalculated by the GSO/GRO engine.
1139 * The size for GSO target is adapted as well.
1140 *
1141 * All values for *flags* are reserved for future usage, and must
1142 * be left at zero.
1143 *
1144 * A call to this helper is susceptible to change the underlaying
1145 * packet buffer. Therefore, at load time, all checks on pointers
1146 * previously done by the verifier are invalidated and must be
1147 * performed again, if the helper is used in combination with
1148 * direct packet access.
1149 * Return
1150 * 0 on success, or a negative error in case of failure.
1151 *
1152 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1153 * Description
1154 * Change the packet type for the packet associated to *skb*. This
1155 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1156 * the eBPF program does not have a write access to *skb*\
1157 * **->pkt_type** beside this helper. Using a helper here allows
1158 * for graceful handling of errors.
1159 *
1160 * The major use case is to change incoming *skb*s to
1161 * **PACKET_HOST** in a programmatic way instead of having to
1162 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1163 * example.
1164 *
1165 * Note that *type* only allows certain values. At this time, they
1166 * are:
1167 *
1168 * **PACKET_HOST**
1169 * Packet is for us.
1170 * **PACKET_BROADCAST**
1171 * Send packet to all.
1172 * **PACKET_MULTICAST**
1173 * Send packet to group.
1174 * **PACKET_OTHERHOST**
1175 * Send packet to someone else.
1176 * Return
1177 * 0 on success, or a negative error in case of failure.
1178 *
c6b5fb86
QM
1179 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1180 * Description
1181 * Check whether *skb* is a descendant of the cgroup2 held by
1182 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1183 * Return
1184 * The return value depends on the result of the test, and can be:
1185 *
1186 * * 0, if the *skb* failed the cgroup2 descendant test.
1187 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1188 * * A negative error code, if an error occurred.
1189 *
fa15601a
QM
1190 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1191 * Description
1192 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1193 * not set, in particular if the hash was cleared due to mangling,
1194 * recompute this hash. Later accesses to the hash can be done
1195 * directly with *skb*\ **->hash**.
1196 *
1197 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1198 * prototype with **bpf_skb_change_proto**\ (), or calling
1199 * **bpf_skb_store_bytes**\ () with the
1200 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1201 * the hash and to trigger a new computation for the next call to
1202 * **bpf_get_hash_recalc**\ ().
1203 * Return
1204 * The 32-bit hash.
1205 *
c456dec4
QM
1206 * u64 bpf_get_current_task(void)
1207 * Return
1208 * A pointer to the current task struct.
fa15601a 1209 *
c6b5fb86
QM
1210 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1211 * Description
1212 * Attempt in a safe way to write *len* bytes from the buffer
1213 * *src* to *dst* in memory. It only works for threads that are in
1214 * user context, and *dst* must be a valid user space address.
1215 *
1216 * This helper should not be used to implement any kind of
1217 * security mechanism because of TOC-TOU attacks, but rather to
1218 * debug, divert, and manipulate execution of semi-cooperative
1219 * processes.
1220 *
1221 * Keep in mind that this feature is meant for experiments, and it
1222 * has a risk of crashing the system and running programs.
1223 * Therefore, when an eBPF program using this helper is attached,
1224 * a warning including PID and process name is printed to kernel
1225 * logs.
1226 * Return
1227 * 0 on success, or a negative error in case of failure.
1228 *
1229 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1230 * Description
1231 * Check whether the probe is being run is the context of a given
1232 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1233 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1234 * Return
1235 * The return value depends on the result of the test, and can be:
1236 *
1237 * * 0, if the *skb* task belongs to the cgroup2.
1238 * * 1, if the *skb* task does not belong to the cgroup2.
1239 * * A negative error code, if an error occurred.
1240 *
fa15601a
QM
1241 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1242 * Description
1243 * Resize (trim or grow) the packet associated to *skb* to the
1244 * new *len*. The *flags* are reserved for future usage, and must
1245 * be left at zero.
1246 *
1247 * The basic idea is that the helper performs the needed work to
1248 * change the size of the packet, then the eBPF program rewrites
1249 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1250 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1251 * and others. This helper is a slow path utility intended for
1252 * replies with control messages. And because it is targeted for
1253 * slow path, the helper itself can afford to be slow: it
1254 * implicitly linearizes, unclones and drops offloads from the
1255 * *skb*.
1256 *
1257 * A call to this helper is susceptible to change the underlaying
1258 * packet buffer. Therefore, at load time, all checks on pointers
1259 * previously done by the verifier are invalidated and must be
1260 * performed again, if the helper is used in combination with
1261 * direct packet access.
1262 * Return
1263 * 0 on success, or a negative error in case of failure.
1264 *
1265 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1266 * Description
1267 * Pull in non-linear data in case the *skb* is non-linear and not
1268 * all of *len* are part of the linear section. Make *len* bytes
1269 * from *skb* readable and writable. If a zero value is passed for
1270 * *len*, then the whole length of the *skb* is pulled.
1271 *
1272 * This helper is only needed for reading and writing with direct
1273 * packet access.
1274 *
1275 * For direct packet access, testing that offsets to access
1276 * are within packet boundaries (test on *skb*\ **->data_end**) is
1277 * susceptible to fail if offsets are invalid, or if the requested
1278 * data is in non-linear parts of the *skb*. On failure the
1279 * program can just bail out, or in the case of a non-linear
1280 * buffer, use a helper to make the data available. The
1281 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1282 * the data. Another one consists in using **bpf_skb_pull_data**
1283 * to pull in once the non-linear parts, then retesting and
1284 * eventually access the data.
1285 *
1286 * At the same time, this also makes sure the *skb* is uncloned,
1287 * which is a necessary condition for direct write. As this needs
1288 * to be an invariant for the write part only, the verifier
1289 * detects writes and adds a prologue that is calling
1290 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1291 * the very beginning in case it is indeed cloned.
1292 *
1293 * A call to this helper is susceptible to change the underlaying
1294 * packet buffer. Therefore, at load time, all checks on pointers
1295 * previously done by the verifier are invalidated and must be
1296 * performed again, if the helper is used in combination with
1297 * direct packet access.
1298 * Return
1299 * 0 on success, or a negative error in case of failure.
1300 *
1301 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1302 * Description
1303 * Add the checksum *csum* into *skb*\ **->csum** in case the
1304 * driver has supplied a checksum for the entire packet into that
1305 * field. Return an error otherwise. This helper is intended to be
1306 * used in combination with **bpf_csum_diff**\ (), in particular
1307 * when the checksum needs to be updated after data has been
1308 * written into the packet through direct packet access.
1309 * Return
1310 * The checksum on success, or a negative error code in case of
1311 * failure.
1312 *
1313 * void bpf_set_hash_invalid(struct sk_buff *skb)
1314 * Description
1315 * Invalidate the current *skb*\ **->hash**. It can be used after
1316 * mangling on headers through direct packet access, in order to
1317 * indicate that the hash is outdated and to trigger a
1318 * recalculation the next time the kernel tries to access this
1319 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1320 *
1321 * int bpf_get_numa_node_id(void)
1322 * Description
1323 * Return the id of the current NUMA node. The primary use case
1324 * for this helper is the selection of sockets for the local NUMA
1325 * node, when the program is attached to sockets using the
1326 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1327 * but the helper is also available to other eBPF program types,
1328 * similarly to **bpf_get_smp_processor_id**\ ().
1329 * Return
1330 * The id of current NUMA node.
1331 *
c6b5fb86
QM
1332 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1333 * Description
1334 * Grows headroom of packet associated to *skb* and adjusts the
1335 * offset of the MAC header accordingly, adding *len* bytes of
1336 * space. It automatically extends and reallocates memory as
1337 * required.
1338 *
1339 * This helper can be used on a layer 3 *skb* to push a MAC header
1340 * for redirection into a layer 2 device.
1341 *
1342 * All values for *flags* are reserved for future usage, and must
1343 * be left at zero.
1344 *
1345 * A call to this helper is susceptible to change the underlaying
1346 * packet buffer. Therefore, at load time, all checks on pointers
1347 * previously done by the verifier are invalidated and must be
1348 * performed again, if the helper is used in combination with
1349 * direct packet access.
1350 * Return
1351 * 0 on success, or a negative error in case of failure.
1352 *
1353 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1354 * Description
1355 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1356 * it is possible to use a negative value for *delta*. This helper
1357 * can be used to prepare the packet for pushing or popping
1358 * headers.
1359 *
1360 * A call to this helper is susceptible to change the underlaying
1361 * packet buffer. Therefore, at load time, all checks on pointers
1362 * previously done by the verifier are invalidated and must be
1363 * performed again, if the helper is used in combination with
1364 * direct packet access.
1365 * Return
1366 * 0 on success, or a negative error in case of failure.
1367 *
1368 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1369 * Description
1370 * Copy a NUL terminated string from an unsafe address
1371 * *unsafe_ptr* to *dst*. The *size* should include the
1372 * terminating NUL byte. In case the string length is smaller than
1373 * *size*, the target is not padded with further NUL bytes. If the
1374 * string length is larger than *size*, just *size*-1 bytes are
1375 * copied and the last byte is set to NUL.
1376 *
1377 * On success, the length of the copied string is returned. This
1378 * makes this helper useful in tracing programs for reading
1379 * strings, and more importantly to get its length at runtime. See
1380 * the following snippet:
1381 *
1382 * ::
1383 *
1384 * SEC("kprobe/sys_open")
1385 * void bpf_sys_open(struct pt_regs *ctx)
1386 * {
1387 * char buf[PATHLEN]; // PATHLEN is defined to 256
1388 * int res = bpf_probe_read_str(buf, sizeof(buf),
1389 * ctx->di);
1390 *
1391 * // Consume buf, for example push it to
1392 * // userspace via bpf_perf_event_output(); we
1393 * // can use res (the string length) as event
1394 * // size, after checking its boundaries.
1395 * }
1396 *
1397 * In comparison, using **bpf_probe_read()** helper here instead
1398 * to read the string would require to estimate the length at
1399 * compile time, and would often result in copying more memory
1400 * than necessary.
1401 *
1402 * Another useful use case is when parsing individual process
1403 * arguments or individual environment variables navigating
1404 * *current*\ **->mm->arg_start** and *current*\
1405 * **->mm->env_start**: using this helper and the return value,
1406 * one can quickly iterate at the right offset of the memory area.
1407 * Return
1408 * On success, the strictly positive length of the string,
1409 * including the trailing NUL character. On error, a negative
1410 * value.
1411 *
1412 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1413 * Description
1414 * If the **struct sk_buff** pointed by *skb* has a known socket,
1415 * retrieve the cookie (generated by the kernel) of this socket.
1416 * If no cookie has been set yet, generate a new cookie. Once
1417 * generated, the socket cookie remains stable for the life of the
1418 * socket. This helper can be useful for monitoring per socket
1419 * networking traffic statistics as it provides a unique socket
1420 * identifier per namespace.
1421 * Return
1422 * A 8-byte long non-decreasing number on success, or 0 if the
1423 * socket field is missing inside *skb*.
1424 *
d692f113
AI
1425 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1426 * Description
1427 * Equivalent to bpf_get_socket_cookie() helper that accepts
1428 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1429 * Return
1430 * A 8-byte long non-decreasing number.
1431 *
1432 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1433 * Description
1434 * Equivalent to bpf_get_socket_cookie() helper that accepts
1435 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1436 * Return
1437 * A 8-byte long non-decreasing number.
1438 *
c6b5fb86
QM
1439 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1440 * Return
1441 * The owner UID of the socket associated to *skb*. If the socket
1442 * is **NULL**, or if it is not a full socket (i.e. if it is a
1443 * time-wait or a request socket instead), **overflowuid** value
1444 * is returned (note that **overflowuid** might also be the actual
1445 * UID value for the socket).
1446 *
fa15601a
QM
1447 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1448 * Description
1449 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1450 * to value *hash*.
1451 * Return
1452 * 0
1453 *
a3ef8e9a 1454 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
7aa79a86
QM
1455 * Description
1456 * Emulate a call to **setsockopt()** on the socket associated to
1457 * *bpf_socket*, which must be a full socket. The *level* at
1458 * which the option resides and the name *optname* of the option
1459 * must be specified, see **setsockopt(2)** for more information.
1460 * The option value of length *optlen* is pointed by *optval*.
1461 *
1462 * This helper actually implements a subset of **setsockopt()**.
1463 * It supports the following *level*\ s:
1464 *
1465 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1466 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1467 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1468 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1469 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1470 * **TCP_BPF_SNDCWND_CLAMP**.
1471 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1472 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1473 * Return
1474 * 0 on success, or a negative error in case of failure.
1475 *
b55cbc8d 1476 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
fa15601a
QM
1477 * Description
1478 * Grow or shrink the room for data in the packet associated to
1479 * *skb* by *len_diff*, and according to the selected *mode*.
1480 *
1481 * There is a single supported mode at this time:
1482 *
1483 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1484 * (room space is added or removed below the layer 3 header).
1485 *
1486 * All values for *flags* are reserved for future usage, and must
1487 * be left at zero.
1488 *
1489 * A call to this helper is susceptible to change the underlaying
1490 * packet buffer. Therefore, at load time, all checks on pointers
1491 * previously done by the verifier are invalidated and must be
1492 * performed again, if the helper is used in combination with
1493 * direct packet access.
1494 * Return
1495 * 0 on success, or a negative error in case of failure.
1496 *
ab127040
QM
1497 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1498 * Description
1499 * Redirect the packet to the endpoint referenced by *map* at
1500 * index *key*. Depending on its type, this *map* can contain
1501 * references to net devices (for forwarding packets through other
1502 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1503 * but this is only implemented for native XDP (with driver
1504 * support) as of this writing).
1505 *
1506 * All values for *flags* are reserved for future usage, and must
1507 * be left at zero.
1508 *
1509 * When used to redirect packets to net devices, this helper
1510 * provides a high performance increase over **bpf_redirect**\ ().
1511 * This is due to various implementation details of the underlying
1512 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1513 * () tries to send packet as a "bulk" to the device.
1514 * Return
1515 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1516 *
1517 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1518 * Description
1519 * Redirect the packet to the socket referenced by *map* (of type
1520 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1521 * egress interfaces can be used for redirection. The
1522 * **BPF_F_INGRESS** value in *flags* is used to make the
1523 * distinction (ingress path is selected if the flag is present,
1524 * egress path otherwise). This is the only flag supported for now.
1525 * Return
1526 * **SK_PASS** on success, or **SK_DROP** on error.
1527 *
a3ef8e9a 1528 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
ab127040
QM
1529 * Description
1530 * Add an entry to, or update a *map* referencing sockets. The
1531 * *skops* is used as a new value for the entry associated to
1532 * *key*. *flags* is one of:
1533 *
1534 * **BPF_NOEXIST**
1535 * The entry for *key* must not exist in the map.
1536 * **BPF_EXIST**
1537 * The entry for *key* must already exist in the map.
1538 * **BPF_ANY**
1539 * No condition on the existence of the entry for *key*.
1540 *
1541 * If the *map* has eBPF programs (parser and verdict), those will
1542 * be inherited by the socket being added. If the socket is
1543 * already attached to eBPF programs, this results in an error.
1544 * Return
1545 * 0 on success, or a negative error in case of failure.
1546 *
fa15601a
QM
1547 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1548 * Description
1549 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1550 * *delta* (which can be positive or negative). Note that this
1551 * operation modifies the address stored in *xdp_md*\ **->data**,
1552 * so the latter must be loaded only after the helper has been
1553 * called.
1554 *
1555 * The use of *xdp_md*\ **->data_meta** is optional and programs
1556 * are not required to use it. The rationale is that when the
1557 * packet is processed with XDP (e.g. as DoS filter), it is
1558 * possible to push further meta data along with it before passing
1559 * to the stack, and to give the guarantee that an ingress eBPF
1560 * program attached as a TC classifier on the same device can pick
1561 * this up for further post-processing. Since TC works with socket
1562 * buffers, it remains possible to set from XDP the **mark** or
1563 * **priority** pointers, or other pointers for the socket buffer.
1564 * Having this scratch space generic and programmable allows for
1565 * more flexibility as the user is free to store whatever meta
1566 * data they need.
1567 *
1568 * A call to this helper is susceptible to change the underlaying
1569 * packet buffer. Therefore, at load time, all checks on pointers
1570 * previously done by the verifier are invalidated and must be
1571 * performed again, if the helper is used in combination with
1572 * direct packet access.
1573 * Return
1574 * 0 on success, or a negative error in case of failure.
7aa79a86
QM
1575 *
1576 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1577 * Description
1578 * Read the value of a perf event counter, and store it into *buf*
1579 * of size *buf_size*. This helper relies on a *map* of type
1580 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1581 * counter is selected when *map* is updated with perf event file
1582 * descriptors. The *map* is an array whose size is the number of
1583 * available CPUs, and each cell contains a value relative to one
1584 * CPU. The value to retrieve is indicated by *flags*, that
1585 * contains the index of the CPU to look up, masked with
1586 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1587 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1588 * current CPU should be retrieved.
1589 *
1590 * This helper behaves in a way close to
1591 * **bpf_perf_event_read**\ () helper, save that instead of
1592 * just returning the value observed, it fills the *buf*
1593 * structure. This allows for additional data to be retrieved: in
1594 * particular, the enabled and running times (in *buf*\
1595 * **->enabled** and *buf*\ **->running**, respectively) are
1596 * copied. In general, **bpf_perf_event_read_value**\ () is
1597 * recommended over **bpf_perf_event_read**\ (), which has some
1598 * ABI issues and provides fewer functionalities.
1599 *
1600 * These values are interesting, because hardware PMU (Performance
1601 * Monitoring Unit) counters are limited resources. When there are
1602 * more PMU based perf events opened than available counters,
1603 * kernel will multiplex these events so each event gets certain
1604 * percentage (but not all) of the PMU time. In case that
1605 * multiplexing happens, the number of samples or counter value
1606 * will not reflect the case compared to when no multiplexing
1607 * occurs. This makes comparison between different runs difficult.
1608 * Typically, the counter value should be normalized before
1609 * comparing to other experiments. The usual normalization is done
1610 * as follows.
1611 *
1612 * ::
1613 *
1614 * normalized_counter = counter * t_enabled / t_running
1615 *
1616 * Where t_enabled is the time enabled for event and t_running is
1617 * the time running for event since last normalization. The
1618 * enabled and running times are accumulated since the perf event
1619 * open. To achieve scaling factor between two invocations of an
1620 * eBPF program, users can can use CPU id as the key (which is
1621 * typical for perf array usage model) to remember the previous
1622 * value and do the calculation inside the eBPF program.
1623 * Return
1624 * 0 on success, or a negative error in case of failure.
1625 *
a3ef8e9a 1626 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
7aa79a86
QM
1627 * Description
1628 * For en eBPF program attached to a perf event, retrieve the
1629 * value of the event counter associated to *ctx* and store it in
1630 * the structure pointed by *buf* and of size *buf_size*. Enabled
1631 * and running times are also stored in the structure (see
1632 * description of helper **bpf_perf_event_read_value**\ () for
1633 * more details).
1634 * Return
1635 * 0 on success, or a negative error in case of failure.
1636 *
a3ef8e9a 1637 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
7aa79a86
QM
1638 * Description
1639 * Emulate a call to **getsockopt()** on the socket associated to
1640 * *bpf_socket*, which must be a full socket. The *level* at
1641 * which the option resides and the name *optname* of the option
1642 * must be specified, see **getsockopt(2)** for more information.
1643 * The retrieved value is stored in the structure pointed by
1644 * *opval* and of length *optlen*.
1645 *
1646 * This helper actually implements a subset of **getsockopt()**.
1647 * It supports the following *level*\ s:
1648 *
1649 * * **IPPROTO_TCP**, which supports *optname*
1650 * **TCP_CONGESTION**.
1651 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1652 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1653 * Return
1654 * 0 on success, or a negative error in case of failure.
1655 *
1656 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1657 * Description
1658 * Used for error injection, this helper uses kprobes to override
1659 * the return value of the probed function, and to set it to *rc*.
1660 * The first argument is the context *regs* on which the kprobe
1661 * works.
1662 *
1663 * This helper works by setting setting the PC (program counter)
1664 * to an override function which is run in place of the original
1665 * probed function. This means the probed function is not run at
1666 * all. The replacement function just returns with the required
1667 * value.
1668 *
1669 * This helper has security implications, and thus is subject to
1670 * restrictions. It is only available if the kernel was compiled
1671 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1672 * option, and in this case it only works on functions tagged with
1673 * **ALLOW_ERROR_INJECTION** in the kernel code.
1674 *
1675 * Also, the helper is only available for the architectures having
1676 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1677 * x86 architecture is the only one to support this feature.
1678 * Return
1679 * 0
1680 *
a3ef8e9a 1681 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
7aa79a86
QM
1682 * Description
1683 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1684 * for the full TCP socket associated to *bpf_sock_ops* to
1685 * *argval*.
1686 *
1687 * The primary use of this field is to determine if there should
1688 * be calls to eBPF programs of type
1689 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1690 * code. A program of the same type can change its value, per
1691 * connection and as necessary, when the connection is
1692 * established. This field is directly accessible for reading, but
1693 * this helper must be used for updates in order to return an
1694 * error if an eBPF program tries to set a callback that is not
1695 * supported in the current kernel.
1696 *
1697 * The supported callback values that *argval* can combine are:
1698 *
1699 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1700 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1701 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1702 *
1703 * Here are some examples of where one could call such eBPF
1704 * program:
1705 *
1706 * * When RTO fires.
1707 * * When a packet is retransmitted.
1708 * * When the connection terminates.
1709 * * When a packet is sent.
1710 * * When a packet is received.
1711 * Return
1712 * Code **-EINVAL** if the socket is not a full TCP socket;
1713 * otherwise, a positive number containing the bits that could not
1714 * be set is returned (which comes down to 0 if all bits were set
1715 * as required).
1716 *
ab127040
QM
1717 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1718 * Description
1719 * This helper is used in programs implementing policies at the
1720 * socket level. If the message *msg* is allowed to pass (i.e. if
1721 * the verdict eBPF program returns **SK_PASS**), redirect it to
1722 * the socket referenced by *map* (of type
1723 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1724 * egress interfaces can be used for redirection. The
1725 * **BPF_F_INGRESS** value in *flags* is used to make the
1726 * distinction (ingress path is selected if the flag is present,
1727 * egress path otherwise). This is the only flag supported for now.
1728 * Return
1729 * **SK_PASS** on success, or **SK_DROP** on error.
1730 *
1731 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1732 * Description
1733 * For socket policies, apply the verdict of the eBPF program to
1734 * the next *bytes* (number of bytes) of message *msg*.
1735 *
1736 * For example, this helper can be used in the following cases:
1737 *
1738 * * A single **sendmsg**\ () or **sendfile**\ () system call
1739 * contains multiple logical messages that the eBPF program is
1740 * supposed to read and for which it should apply a verdict.
1741 * * An eBPF program only cares to read the first *bytes* of a
1742 * *msg*. If the message has a large payload, then setting up
1743 * and calling the eBPF program repeatedly for all bytes, even
1744 * though the verdict is already known, would create unnecessary
1745 * overhead.
1746 *
1747 * When called from within an eBPF program, the helper sets a
1748 * counter internal to the BPF infrastructure, that is used to
1749 * apply the last verdict to the next *bytes*. If *bytes* is
1750 * smaller than the current data being processed from a
1751 * **sendmsg**\ () or **sendfile**\ () system call, the first
1752 * *bytes* will be sent and the eBPF program will be re-run with
1753 * the pointer for start of data pointing to byte number *bytes*
1754 * **+ 1**. If *bytes* is larger than the current data being
1755 * processed, then the eBPF verdict will be applied to multiple
1756 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1757 * consumed.
1758 *
1759 * Note that if a socket closes with the internal counter holding
1760 * a non-zero value, this is not a problem because data is not
1761 * being buffered for *bytes* and is sent as it is received.
1762 * Return
1763 * 0
1764 *
1765 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1766 * Description
1767 * For socket policies, prevent the execution of the verdict eBPF
1768 * program for message *msg* until *bytes* (byte number) have been
1769 * accumulated.
1770 *
1771 * This can be used when one needs a specific number of bytes
1772 * before a verdict can be assigned, even if the data spans
1773 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1774 * case would be a user calling **sendmsg**\ () repeatedly with
1775 * 1-byte long message segments. Obviously, this is bad for
1776 * performance, but it is still valid. If the eBPF program needs
1777 * *bytes* bytes to validate a header, this helper can be used to
1778 * prevent the eBPF program to be called again until *bytes* have
1779 * been accumulated.
1780 * Return
1781 * 0
1782 *
1783 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1784 * Description
1785 * For socket policies, pull in non-linear data from user space
1786 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1787 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1788 * respectively.
1789 *
1790 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1791 * *msg* it can only parse data that the (**data**, **data_end**)
1792 * pointers have already consumed. For **sendmsg**\ () hooks this
1793 * is likely the first scatterlist element. But for calls relying
1794 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1795 * be the range (**0**, **0**) because the data is shared with
1796 * user space and by default the objective is to avoid allowing
1797 * user space to modify data while (or after) eBPF verdict is
1798 * being decided. This helper can be used to pull in data and to
1799 * set the start and end pointer to given values. Data will be
1800 * copied if necessary (i.e. if data was not linear and if start
1801 * and end pointers do not point to the same chunk).
1802 *
1803 * A call to this helper is susceptible to change the underlaying
1804 * packet buffer. Therefore, at load time, all checks on pointers
1805 * previously done by the verifier are invalidated and must be
1806 * performed again, if the helper is used in combination with
1807 * direct packet access.
1808 *
1809 * All values for *flags* are reserved for future usage, and must
1810 * be left at zero.
1811 * Return
1812 * 0 on success, or a negative error in case of failure.
1813 *
a3ef8e9a 1814 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
7aa79a86
QM
1815 * Description
1816 * Bind the socket associated to *ctx* to the address pointed by
1817 * *addr*, of length *addr_len*. This allows for making outgoing
1818 * connection from the desired IP address, which can be useful for
1819 * example when all processes inside a cgroup should use one
1820 * single IP address on a host that has multiple IP configured.
1821 *
1822 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1823 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1824 * **AF_INET6**). Looking for a free port to bind to can be
1825 * expensive, therefore binding to port is not permitted by the
1826 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1827 * must be set to zero.
1828 * Return
1829 * 0 on success, or a negative error in case of failure.
2d020dd7
QM
1830 *
1831 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1832 * Description
1833 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1834 * only possible to shrink the packet as of this writing,
1835 * therefore *delta* must be a negative integer.
1836 *
1837 * A call to this helper is susceptible to change the underlaying
1838 * packet buffer. Therefore, at load time, all checks on pointers
1839 * previously done by the verifier are invalidated and must be
1840 * performed again, if the helper is used in combination with
1841 * direct packet access.
1842 * Return
1843 * 0 on success, or a negative error in case of failure.
1844 *
1845 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1846 * Description
1847 * Retrieve the XFRM state (IP transform framework, see also
1848 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1849 *
1850 * The retrieved value is stored in the **struct bpf_xfrm_state**
1851 * pointed by *xfrm_state* and of length *size*.
1852 *
1853 * All values for *flags* are reserved for future usage, and must
1854 * be left at zero.
1855 *
1856 * This helper is available only if the kernel was compiled with
1857 * **CONFIG_XFRM** configuration option.
1858 * Return
1859 * 0 on success, or a negative error in case of failure.
c195651e
YS
1860 *
1861 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1862 * Description
79552fbc
QM
1863 * Return a user or a kernel stack in bpf program provided buffer.
1864 * To achieve this, the helper needs *ctx*, which is a pointer
1865 * to the context on which the tracing program is executed.
1866 * To store the stacktrace, the bpf program provides *buf* with
1867 * a nonnegative *size*.
1868 *
1869 * The last argument, *flags*, holds the number of stack frames to
1870 * skip (from 0 to 255), masked with
1871 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1872 * the following flags:
1873 *
1874 * **BPF_F_USER_STACK**
1875 * Collect a user space stack instead of a kernel stack.
1876 * **BPF_F_USER_BUILD_ID**
1877 * Collect buildid+offset instead of ips for user stack,
1878 * only valid if **BPF_F_USER_STACK** is also specified.
1879 *
1880 * **bpf_get_stack**\ () can collect up to
1881 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1882 * to sufficient large buffer size. Note that
1883 * this limit can be controlled with the **sysctl** program, and
1884 * that it should be manually increased in order to profile long
1885 * user stacks (such as stacks for Java programs). To do so, use:
1886 *
1887 * ::
1888 *
1889 * # sysctl kernel.perf_event_max_stack=<new value>
c195651e 1890 * Return
7a279e93
QM
1891 * A non-negative value equal to or less than *size* on success,
1892 * or a negative error in case of failure.
4e1ec56c 1893 *
2bae79d2 1894 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
4e1ec56c
DB
1895 * Description
1896 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1897 * it provides an easy way to load *len* bytes from *offset*
1898 * from the packet associated to *skb*, into the buffer pointed
1899 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1900 * a fifth argument *start_header* exists in order to select a
1901 * base offset to start from. *start_header* can be one of:
1902 *
1903 * **BPF_HDR_START_MAC**
1904 * Base offset to load data from is *skb*'s mac header.
1905 * **BPF_HDR_START_NET**
1906 * Base offset to load data from is *skb*'s network header.
1907 *
1908 * In general, "direct packet access" is the preferred method to
1909 * access packet data, however, this helper is in particular useful
1910 * in socket filters where *skb*\ **->data** does not always point
1911 * to the start of the mac header and where "direct packet access"
1912 * is not available.
4e1ec56c
DB
1913 * Return
1914 * 0 on success, or a negative error in case of failure.
1915 *
87f5fc7e
DA
1916 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1917 * Description
1918 * Do FIB lookup in kernel tables using parameters in *params*.
1919 * If lookup is successful and result shows packet is to be
1920 * forwarded, the neighbor tables are searched for the nexthop.
1921 * If successful (ie., FIB lookup shows forwarding and nexthop
fa898d76
DA
1922 * is resolved), the nexthop address is returned in ipv4_dst
1923 * or ipv6_dst based on family, smac is set to mac address of
1924 * egress device, dmac is set to nexthop mac address, rt_metric
4c79579b
DA
1925 * is set to metric from route (IPv4/IPv6 only), and ifindex
1926 * is set to the device index of the nexthop from the FIB lookup.
87f5fc7e 1927 *
90b1023f
QM
1928 * *plen* argument is the size of the passed in struct.
1929 * *flags* argument can be a combination of one or more of the
1930 * following values:
87f5fc7e 1931 *
7a279e93
QM
1932 * **BPF_FIB_LOOKUP_DIRECT**
1933 * Do a direct table lookup vs full lookup using FIB
1934 * rules.
1935 * **BPF_FIB_LOOKUP_OUTPUT**
1936 * Perform lookup from an egress perspective (default is
1937 * ingress).
87f5fc7e 1938 *
90b1023f
QM
1939 * *ctx* is either **struct xdp_md** for XDP programs or
1940 * **struct sk_buff** tc cls_act programs.
1941 * Return
4c79579b
DA
1942 * * < 0 if any input argument is invalid
1943 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1944 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2bae79d2 1945 * packet is not forwarded or needs assist from full stack
81110384
JF
1946 *
1947 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1948 * Description
1949 * Add an entry to, or update a sockhash *map* referencing sockets.
1950 * The *skops* is used as a new value for the entry associated to
1951 * *key*. *flags* is one of:
1952 *
1953 * **BPF_NOEXIST**
1954 * The entry for *key* must not exist in the map.
1955 * **BPF_EXIST**
1956 * The entry for *key* must already exist in the map.
1957 * **BPF_ANY**
1958 * No condition on the existence of the entry for *key*.
1959 *
1960 * If the *map* has eBPF programs (parser and verdict), those will
1961 * be inherited by the socket being added. If the socket is
1962 * already attached to eBPF programs, this results in an error.
1963 * Return
1964 * 0 on success, or a negative error in case of failure.
1965 *
1966 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1967 * Description
1968 * This helper is used in programs implementing policies at the
1969 * socket level. If the message *msg* is allowed to pass (i.e. if
1970 * the verdict eBPF program returns **SK_PASS**), redirect it to
1971 * the socket referenced by *map* (of type
1972 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1973 * egress interfaces can be used for redirection. The
1974 * **BPF_F_INGRESS** value in *flags* is used to make the
1975 * distinction (ingress path is selected if the flag is present,
1976 * egress path otherwise). This is the only flag supported for now.
1977 * Return
1978 * **SK_PASS** on success, or **SK_DROP** on error.
1979 *
1980 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1981 * Description
1982 * This helper is used in programs implementing policies at the
1983 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1984 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1985 * to the socket referenced by *map* (of type
1986 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1987 * egress interfaces can be used for redirection. The
1988 * **BPF_F_INGRESS** value in *flags* is used to make the
1989 * distinction (ingress path is selected if the flag is present,
1990 * egress otherwise). This is the only flag supported for now.
1991 * Return
1992 * **SK_PASS** on success, or **SK_DROP** on error.
fe94cc29
MX
1993 *
1994 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
1995 * Description
1996 * Encapsulate the packet associated to *skb* within a Layer 3
1997 * protocol header. This header is provided in the buffer at
1998 * address *hdr*, with *len* its size in bytes. *type* indicates
1999 * the protocol of the header and can be one of:
2000 *
2001 * **BPF_LWT_ENCAP_SEG6**
2002 * IPv6 encapsulation with Segment Routing Header
2003 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2004 * the IPv6 header is computed by the kernel.
2005 * **BPF_LWT_ENCAP_SEG6_INLINE**
2006 * Only works if *skb* contains an IPv6 packet. Insert a
2007 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2008 * the IPv6 header.
2009 *
2010 * A call to this helper is susceptible to change the underlaying
2011 * packet buffer. Therefore, at load time, all checks on pointers
2012 * previously done by the verifier are invalidated and must be
2013 * performed again, if the helper is used in combination with
2014 * direct packet access.
2015 * Return
2016 * 0 on success, or a negative error in case of failure.
2017 *
2018 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2019 * Description
2020 * Store *len* bytes from address *from* into the packet
2021 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2022 * inside the outermost IPv6 Segment Routing Header can be
2023 * modified through this helper.
2024 *
2025 * A call to this helper is susceptible to change the underlaying
2026 * packet buffer. Therefore, at load time, all checks on pointers
2027 * previously done by the verifier are invalidated and must be
2028 * performed again, if the helper is used in combination with
2029 * direct packet access.
2030 * Return
2031 * 0 on success, or a negative error in case of failure.
2032 *
2033 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2034 * Description
2035 * Adjust the size allocated to TLVs in the outermost IPv6
2036 * Segment Routing Header contained in the packet associated to
2037 * *skb*, at position *offset* by *delta* bytes. Only offsets
2038 * after the segments are accepted. *delta* can be as well
2039 * positive (growing) as negative (shrinking).
2040 *
2041 * A call to this helper is susceptible to change the underlaying
2042 * packet buffer. Therefore, at load time, all checks on pointers
2043 * previously done by the verifier are invalidated and must be
2044 * performed again, if the helper is used in combination with
2045 * direct packet access.
2046 * Return
2047 * 0 on success, or a negative error in case of failure.
2048 *
2049 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2050 * Description
2051 * Apply an IPv6 Segment Routing action of type *action* to the
2052 * packet associated to *skb*. Each action takes a parameter
2053 * contained at address *param*, and of length *param_len* bytes.
2054 * *action* can be one of:
2055 *
2056 * **SEG6_LOCAL_ACTION_END_X**
2057 * End.X action: Endpoint with Layer-3 cross-connect.
2058 * Type of *param*: **struct in6_addr**.
2059 * **SEG6_LOCAL_ACTION_END_T**
2060 * End.T action: Endpoint with specific IPv6 table lookup.
2061 * Type of *param*: **int**.
2062 * **SEG6_LOCAL_ACTION_END_B6**
2063 * End.B6 action: Endpoint bound to an SRv6 policy.
2064 * Type of param: **struct ipv6_sr_hdr**.
2065 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2066 * End.B6.Encap action: Endpoint bound to an SRv6
2067 * encapsulation policy.
2068 * Type of param: **struct ipv6_sr_hdr**.
2069 *
2070 * A call to this helper is susceptible to change the underlaying
2071 * packet buffer. Therefore, at load time, all checks on pointers
2072 * previously done by the verifier are invalidated and must be
2073 * performed again, if the helper is used in combination with
2074 * direct packet access.
2075 * Return
2076 * 0 on success, or a negative error in case of failure.
f4364dcf
SY
2077 *
2078 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2079 * Description
2080 * This helper is used in programs implementing IR decoding, to
2081 * report a successfully decoded key press with *scancode*,
2082 * *toggle* value in the given *protocol*. The scancode will be
2083 * translated to a keycode using the rc keymap, and reported as
2084 * an input key down event. After a period a key up event is
2085 * generated. This period can be extended by calling either
90b1023f
QM
2086 * **bpf_rc_keydown**\ () again with the same values, or calling
2087 * **bpf_rc_repeat**\ ().
f4364dcf
SY
2088 *
2089 * Some protocols include a toggle bit, in case the button was
2090 * released and pressed again between consecutive scancodes.
2091 *
2092 * The *ctx* should point to the lirc sample as passed into
2093 * the program.
2094 *
2095 * The *protocol* is the decoded protocol number (see
2096 * **enum rc_proto** for some predefined values).
2097 *
2098 * This helper is only available is the kernel was compiled with
2099 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2100 * "**y**".
f4364dcf
SY
2101 * Return
2102 * 0
2103 *
2104 * int bpf_rc_repeat(void *ctx)
2105 * Description
2106 * This helper is used in programs implementing IR decoding, to
2107 * report a successfully decoded repeat key message. This delays
2108 * the generation of a key up event for previously generated
2109 * key down event.
2110 *
2111 * Some IR protocols like NEC have a special IR message for
2112 * repeating last button, for when a button is held down.
2113 *
2114 * The *ctx* should point to the lirc sample as passed into
2115 * the program.
2116 *
2117 * This helper is only available is the kernel was compiled with
2118 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2119 * "**y**".
f4364dcf
SY
2120 * Return
2121 * 0
cb20b08e
DB
2122 *
2123 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2124 * Description
2125 * Return the cgroup v2 id of the socket associated with the *skb*.
2126 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2127 * helper for cgroup v1 by providing a tag resp. identifier that
2128 * can be matched on or used for map lookups e.g. to implement
2129 * policy. The cgroup v2 id of a given path in the hierarchy is
2130 * exposed in user space through the f_handle API in order to get
2131 * to the same 64-bit id.
2132 *
2133 * This helper can be used on TC egress path, but not on ingress,
2134 * and is available only if the kernel was compiled with the
2135 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2136 * Return
2137 * The id is returned or 0 in case the id could not be retrieved.
bf6fa2c8 2138 *
77236281
AI
2139 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2140 * Description
2141 * Return id of cgroup v2 that is ancestor of cgroup associated
2142 * with the *skb* at the *ancestor_level*. The root cgroup is at
2143 * *ancestor_level* zero and each step down the hierarchy
2144 * increments the level. If *ancestor_level* == level of cgroup
2145 * associated with *skb*, then return value will be same as that
2146 * of **bpf_skb_cgroup_id**\ ().
2147 *
2148 * The helper is useful to implement policies based on cgroups
2149 * that are upper in hierarchy than immediate cgroup associated
2150 * with *skb*.
2151 *
2152 * The format of returned id and helper limitations are same as in
2153 * **bpf_skb_cgroup_id**\ ().
2154 * Return
2155 * The id is returned or 0 in case the id could not be retrieved.
2156 *
bf6fa2c8
YS
2157 * u64 bpf_get_current_cgroup_id(void)
2158 * Return
2159 * A 64-bit integer containing the current cgroup id based
2160 * on the cgroup within which the current task is running.
cd339431
RG
2161 *
2162 * void* get_local_storage(void *map, u64 flags)
2163 * Description
2164 * Get the pointer to the local storage area.
2165 * The type and the size of the local storage is defined
2166 * by the *map* argument.
2167 * The *flags* meaning is specific for each map type,
2168 * and has to be 0 for cgroup local storage.
2169 *
90b1023f
QM
2170 * Depending on the BPF program type, a local storage area
2171 * can be shared between multiple instances of the BPF program,
cd339431
RG
2172 * running simultaneously.
2173 *
2174 * A user should care about the synchronization by himself.
90b1023f 2175 * For example, by using the **BPF_STX_XADD** instruction to alter
cd339431
RG
2176 * the shared data.
2177 * Return
90b1023f 2178 * A pointer to the local storage area.
2dbb9b9e
MKL
2179 *
2180 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2181 * Description
90b1023f
QM
2182 * Select a **SO_REUSEPORT** socket from a
2183 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2184 * It checks the selected socket is matching the incoming
2185 * request in the socket buffer.
2dbb9b9e
MKL
2186 * Return
2187 * 0 on success, or a negative error in case of failure.
6acc9b43
JS
2188 *
2189 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2190 * Description
2191 * Look for TCP socket matching *tuple*, optionally in a child
2192 * network namespace *netns*. The return value must be checked,
90b1023f 2193 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2194 *
2195 * The *ctx* should point to the context of the program, such as
2196 * the skb or socket (depending on the hook in use). This is used
2197 * to determine the base network namespace for the lookup.
2198 *
2199 * *tuple_size* must be one of:
2200 *
2201 * **sizeof**\ (*tuple*\ **->ipv4**)
2202 * Look for an IPv4 socket.
2203 * **sizeof**\ (*tuple*\ **->ipv6**)
2204 * Look for an IPv6 socket.
2205 *
2206 * If the *netns* is zero, then the socket lookup table in the
2207 * netns associated with the *ctx* will be used. For the TC hooks,
2208 * this in the netns of the device in the skb. For socket hooks,
2209 * this in the netns of the socket. If *netns* is non-zero, then
2210 * it specifies the ID of the netns relative to the netns
2211 * associated with the *ctx*.
2212 *
2213 * All values for *flags* are reserved for future usage, and must
2214 * be left at zero.
2215 *
2216 * This helper is available only if the kernel was compiled with
2217 * **CONFIG_NET** configuration option.
2218 * Return
90b1023f
QM
2219 * A pointer to *struct bpf_sock*, or **NULL** in case of failure.
2220 * For sockets with reuseport option, **struct bpf_sock**
2221 * return is from **reuse->socks**\ [] using hash of the packet.
6acc9b43
JS
2222 *
2223 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u32 netns, u64 flags)
2224 * Description
2225 * Look for UDP socket matching *tuple*, optionally in a child
2226 * network namespace *netns*. The return value must be checked,
90b1023f 2227 * and if non-**NULL**, released via **bpf_sk_release**\ ().
6acc9b43
JS
2228 *
2229 * The *ctx* should point to the context of the program, such as
2230 * the skb or socket (depending on the hook in use). This is used
2231 * to determine the base network namespace for the lookup.
2232 *
2233 * *tuple_size* must be one of:
2234 *
2235 * **sizeof**\ (*tuple*\ **->ipv4**)
2236 * Look for an IPv4 socket.
2237 * **sizeof**\ (*tuple*\ **->ipv6**)
2238 * Look for an IPv6 socket.
2239 *
2240 * If the *netns* is zero, then the socket lookup table in the
2241 * netns associated with the *ctx* will be used. For the TC hooks,
2242 * this in the netns of the device in the skb. For socket hooks,
2243 * this in the netns of the socket. If *netns* is non-zero, then
2244 * it specifies the ID of the netns relative to the netns
2245 * associated with the *ctx*.
2246 *
2247 * All values for *flags* are reserved for future usage, and must
2248 * be left at zero.
2249 *
2250 * This helper is available only if the kernel was compiled with
2251 * **CONFIG_NET** configuration option.
2252 * Return
90b1023f
QM
2253 * A pointer to **struct bpf_sock**, or **NULL** in case of
2254 * failure. For sockets with reuseport option, **struct bpf_sock**
2255 * return is from **reuse->socks**\ [] using hash of the packet.
6acc9b43 2256 *
90b1023f 2257 * int bpf_sk_release(struct bpf_sock *sock)
6acc9b43 2258 * Description
90b1023f
QM
2259 * Release the reference held by *sock*. *sock* must be a
2260 * non-**NULL** pointer that was returned from
2261 * **bpf_sk_lookup_xxx**\ ().
6acc9b43
JS
2262 * Return
2263 * 0 on success, or a negative error in case of failure.
6fff607e 2264 *
90b1023f
QM
2265 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2266 * Description
2267 * Pop an element from *map*.
2268 * Return
2269 * 0 on success, or a negative error in case of failure.
2270 *
2271 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2272 * Description
2273 * Get an element from *map* without removing it.
2274 * Return
2275 * 0 on success, or a negative error in case of failure.
2276 *
6fff607e
JF
2277 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2278 * Description
90b1023f 2279 * For socket policies, insert *len* bytes into *msg* at offset
6fff607e
JF
2280 * *start*.
2281 *
2282 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
90b1023f 2283 * *msg* it may want to insert metadata or options into the *msg*.
6fff607e
JF
2284 * This can later be read and used by any of the lower layer BPF
2285 * hooks.
2286 *
2287 * This helper may fail if under memory pressure (a malloc
2288 * fails) in these cases BPF programs will get an appropriate
2289 * error and BPF programs will need to handle them.
6fff607e
JF
2290 * Return
2291 * 0 on success, or a negative error in case of failure.
7246d8ed
JF
2292 *
2293 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
90b1023f 2294 * Description
7246d8ed
JF
2295 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2296 * This may result in **ENOMEM** errors under certain situations if
2297 * an allocation and copy are required due to a full ring buffer.
2298 * However, the helper will try to avoid doing the allocation
2299 * if possible. Other errors can occur if input parameters are
90b1023f 2300 * invalid either due to *start* byte not being valid part of *msg*
7246d8ed 2301 * payload and/or *pop* value being to large.
7246d8ed 2302 * Return
90b1023f 2303 * 0 on success, or a negative error in case of failure.
01d3240a
SY
2304 *
2305 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2306 * Description
2307 * This helper is used in programs implementing IR decoding, to
2308 * report a successfully decoded pointer movement.
2309 *
2310 * The *ctx* should point to the lirc sample as passed into
2311 * the program.
2312 *
2313 * This helper is only available is the kernel was compiled with
2314 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2315 * "**y**".
2316 * Return
2317 * 0
ebb676da
TG
2318 */
2319#define __BPF_FUNC_MAPPER(FN) \
2320 FN(unspec), \
2321 FN(map_lookup_elem), \
2322 FN(map_update_elem), \
2323 FN(map_delete_elem), \
2324 FN(probe_read), \
2325 FN(ktime_get_ns), \
2326 FN(trace_printk), \
2327 FN(get_prandom_u32), \
2328 FN(get_smp_processor_id), \
2329 FN(skb_store_bytes), \
2330 FN(l3_csum_replace), \
2331 FN(l4_csum_replace), \
2332 FN(tail_call), \
2333 FN(clone_redirect), \
2334 FN(get_current_pid_tgid), \
2335 FN(get_current_uid_gid), \
2336 FN(get_current_comm), \
2337 FN(get_cgroup_classid), \
2338 FN(skb_vlan_push), \
2339 FN(skb_vlan_pop), \
2340 FN(skb_get_tunnel_key), \
2341 FN(skb_set_tunnel_key), \
2342 FN(perf_event_read), \
2343 FN(redirect), \
2344 FN(get_route_realm), \
2345 FN(perf_event_output), \
2346 FN(skb_load_bytes), \
2347 FN(get_stackid), \
2348 FN(csum_diff), \
2349 FN(skb_get_tunnel_opt), \
2350 FN(skb_set_tunnel_opt), \
2351 FN(skb_change_proto), \
2352 FN(skb_change_type), \
2353 FN(skb_under_cgroup), \
2354 FN(get_hash_recalc), \
2355 FN(get_current_task), \
2356 FN(probe_write_user), \
2357 FN(current_task_under_cgroup), \
2358 FN(skb_change_tail), \
2359 FN(skb_pull_data), \
2360 FN(csum_update), \
2361 FN(set_hash_invalid), \
3a0af8fd 2362 FN(get_numa_node_id), \
17bedab2 2363 FN(skb_change_head), \
a5e8c070 2364 FN(xdp_adjust_head), \
91b8270f 2365 FN(probe_read_str), \
6acc5c29 2366 FN(get_socket_cookie), \
ded092cd 2367 FN(get_socket_uid), \
8c4b4c7e 2368 FN(set_hash), \
2be7e212 2369 FN(setsockopt), \
97f91a7c 2370 FN(skb_adjust_room), \
174a79ff
JF
2371 FN(redirect_map), \
2372 FN(sk_redirect_map), \
2373 FN(sock_map_update), \
908432ca 2374 FN(xdp_adjust_meta), \
4bebdc7a 2375 FN(perf_event_read_value), \
cd86d1fd 2376 FN(perf_prog_read_value), \
9802d865 2377 FN(getsockopt), \
b13d8807 2378 FN(override_return), \
4f738adb 2379 FN(sock_ops_cb_flags_set), \
2a100317 2380 FN(msg_redirect_map), \
91843d54 2381 FN(msg_apply_bytes), \
015632bb 2382 FN(msg_cork_bytes), \
d74bad4e 2383 FN(msg_pull_data), \
b32cc5b9 2384 FN(bind), \
12bed760 2385 FN(xdp_adjust_tail), \
c195651e 2386 FN(skb_get_xfrm_state), \
4e1ec56c 2387 FN(get_stack), \
87f5fc7e 2388 FN(skb_load_bytes_relative), \
81110384
JF
2389 FN(fib_lookup), \
2390 FN(sock_hash_update), \
2391 FN(msg_redirect_hash), \
fe94cc29
MX
2392 FN(sk_redirect_hash), \
2393 FN(lwt_push_encap), \
2394 FN(lwt_seg6_store_bytes), \
2395 FN(lwt_seg6_adjust_srh), \
f4364dcf
SY
2396 FN(lwt_seg6_action), \
2397 FN(rc_repeat), \
cb20b08e 2398 FN(rc_keydown), \
bf6fa2c8 2399 FN(skb_cgroup_id), \
cd339431 2400 FN(get_current_cgroup_id), \
2dbb9b9e 2401 FN(get_local_storage), \
77236281 2402 FN(sk_select_reuseport), \
6acc9b43
JS
2403 FN(skb_ancestor_cgroup_id), \
2404 FN(sk_lookup_tcp), \
2405 FN(sk_lookup_udp), \
f1a2e44a
MV
2406 FN(sk_release), \
2407 FN(map_push_elem), \
2408 FN(map_pop_elem), \
6fff607e 2409 FN(map_peek_elem), \
7246d8ed 2410 FN(msg_push_data), \
01d3240a
SY
2411 FN(msg_pop_data), \
2412 FN(rc_pointer_rel),
ebb676da 2413
09756af4
AS
2414/* integer value in 'imm' field of BPF_CALL instruction selects which helper
2415 * function eBPF program intends to call
2416 */
ebb676da 2417#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
09756af4 2418enum bpf_func_id {
ebb676da 2419 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
09756af4
AS
2420 __BPF_FUNC_MAX_ID,
2421};
ebb676da 2422#undef __BPF_ENUM_FN
09756af4 2423
781c53bc
DB
2424/* All flags used by eBPF helper functions, placed here. */
2425
2426/* BPF_FUNC_skb_store_bytes flags. */
2427#define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
8afd54c8 2428#define BPF_F_INVALIDATE_HASH (1ULL << 1)
781c53bc
DB
2429
2430/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2431 * First 4 bits are for passing the header field size.
2432 */
2433#define BPF_F_HDR_FIELD_MASK 0xfULL
2434
2435/* BPF_FUNC_l4_csum_replace flags. */
2436#define BPF_F_PSEUDO_HDR (1ULL << 4)
2f72959a 2437#define BPF_F_MARK_MANGLED_0 (1ULL << 5)
d1b662ad 2438#define BPF_F_MARK_ENFORCE (1ULL << 6)
781c53bc
DB
2439
2440/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2441#define BPF_F_INGRESS (1ULL << 0)
2442
c6c33454
DB
2443/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2444#define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2445
c195651e 2446/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
d5a3b1f6
AS
2447#define BPF_F_SKIP_FIELD_MASK 0xffULL
2448#define BPF_F_USER_STACK (1ULL << 8)
c195651e 2449/* flags used by BPF_FUNC_get_stackid only. */
d5a3b1f6
AS
2450#define BPF_F_FAST_STACK_CMP (1ULL << 9)
2451#define BPF_F_REUSE_STACKID (1ULL << 10)
c195651e
YS
2452/* flags used by BPF_FUNC_get_stack only. */
2453#define BPF_F_USER_BUILD_ID (1ULL << 11)
d5a3b1f6 2454
2da897e5
DB
2455/* BPF_FUNC_skb_set_tunnel_key flags. */
2456#define BPF_F_ZERO_CSUM_TX (1ULL << 1)
22080870 2457#define BPF_F_DONT_FRAGMENT (1ULL << 2)
77a5196a 2458#define BPF_F_SEQ_NUMBER (1ULL << 3)
2da897e5 2459
908432ca
YS
2460/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2461 * BPF_FUNC_perf_event_read_value flags.
2462 */
1e33759c
DB
2463#define BPF_F_INDEX_MASK 0xffffffffULL
2464#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
555c8a86
DB
2465/* BPF_FUNC_perf_event_output for sk_buff input context. */
2466#define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
1e33759c 2467
2be7e212
DB
2468/* Mode for BPF_FUNC_skb_adjust_room helper. */
2469enum bpf_adj_room_mode {
2470 BPF_ADJ_ROOM_NET,
2471};
2472
4e1ec56c
DB
2473/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2474enum bpf_hdr_start_off {
2475 BPF_HDR_START_MAC,
2476 BPF_HDR_START_NET,
2477};
2478
fe94cc29
MX
2479/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2480enum bpf_lwt_encap_mode {
2481 BPF_LWT_ENCAP_SEG6,
2482 BPF_LWT_ENCAP_SEG6_INLINE
2483};
2484
9bac3d6d
AS
2485/* user accessible mirror of in-kernel sk_buff.
2486 * new fields can only be added to the end of this structure
2487 */
2488struct __sk_buff {
2489 __u32 len;
2490 __u32 pkt_type;
2491 __u32 mark;
2492 __u32 queue_mapping;
c2497395
AS
2493 __u32 protocol;
2494 __u32 vlan_present;
2495 __u32 vlan_tci;
27cd5452 2496 __u32 vlan_proto;
bcad5718 2497 __u32 priority;
37e82c2f
AS
2498 __u32 ingress_ifindex;
2499 __u32 ifindex;
d691f9e8
AS
2500 __u32 tc_index;
2501 __u32 cb[5];
ba7591d8 2502 __u32 hash;
045efa82 2503 __u32 tc_classid;
969bf05e
AS
2504 __u32 data;
2505 __u32 data_end;
b1d9fc41 2506 __u32 napi_id;
8a31db56 2507
de8f3a83 2508 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
8a31db56
JF
2509 __u32 family;
2510 __u32 remote_ip4; /* Stored in network byte order */
2511 __u32 local_ip4; /* Stored in network byte order */
2512 __u32 remote_ip6[4]; /* Stored in network byte order */
2513 __u32 local_ip6[4]; /* Stored in network byte order */
2514 __u32 remote_port; /* Stored in network byte order */
2515 __u32 local_port; /* stored in host byte order */
de8f3a83
DB
2516 /* ... here. */
2517
2518 __u32 data_meta;
d58e468b 2519 struct bpf_flow_keys *flow_keys;
f11216b2 2520 __u64 tstamp;
e3da08d0 2521 __u32 wire_len;
9bac3d6d
AS
2522};
2523
d3aa45ce
AS
2524struct bpf_tunnel_key {
2525 __u32 tunnel_id;
c6c33454
DB
2526 union {
2527 __u32 remote_ipv4;
2528 __u32 remote_ipv6[4];
2529 };
2530 __u8 tunnel_tos;
2531 __u8 tunnel_ttl;
1fbc2e0c 2532 __u16 tunnel_ext; /* Padding, future use. */
4018ab18 2533 __u32 tunnel_label;
d3aa45ce
AS
2534};
2535
12bed760
EB
2536/* user accessible mirror of in-kernel xfrm_state.
2537 * new fields can only be added to the end of this structure
2538 */
2539struct bpf_xfrm_state {
2540 __u32 reqid;
2541 __u32 spi; /* Stored in network byte order */
2542 __u16 family;
1fbc2e0c 2543 __u16 ext; /* Padding, future use. */
12bed760
EB
2544 union {
2545 __u32 remote_ipv4; /* Stored in network byte order */
2546 __u32 remote_ipv6[4]; /* Stored in network byte order */
2547 };
2548};
2549
3a0af8fd
TG
2550/* Generic BPF return codes which all BPF program types may support.
2551 * The values are binary compatible with their TC_ACT_* counter-part to
2552 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2553 * programs.
2554 *
2555 * XDP is handled seprately, see XDP_*.
2556 */
2557enum bpf_ret_code {
2558 BPF_OK = 0,
2559 /* 1 reserved */
2560 BPF_DROP = 2,
2561 /* 3-6 reserved */
2562 BPF_REDIRECT = 7,
2563 /* >127 are reserved for prog type specific return codes */
2564};
2565
61023658
DA
2566struct bpf_sock {
2567 __u32 bound_dev_if;
aa4c1037
DA
2568 __u32 family;
2569 __u32 type;
2570 __u32 protocol;
482dca93
DA
2571 __u32 mark;
2572 __u32 priority;
aac3fc32
AI
2573 __u32 src_ip4; /* Allows 1,2,4-byte read.
2574 * Stored in network byte order.
2575 */
2576 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2577 * Stored in network byte order.
2578 */
2579 __u32 src_port; /* Allows 4-byte read.
2580 * Stored in host byte order
2581 */
61023658
DA
2582};
2583
6acc9b43
JS
2584struct bpf_sock_tuple {
2585 union {
2586 struct {
2587 __be32 saddr;
2588 __be32 daddr;
2589 __be16 sport;
2590 __be16 dport;
2591 } ipv4;
2592 struct {
2593 __be32 saddr[4];
2594 __be32 daddr[4];
2595 __be16 sport;
2596 __be16 dport;
2597 } ipv6;
2598 };
2599};
2600
17bedab2
MKL
2601#define XDP_PACKET_HEADROOM 256
2602
6a773a15
BB
2603/* User return codes for XDP prog type.
2604 * A valid XDP program must return one of these defined values. All other
9beb8bed
DB
2605 * return codes are reserved for future use. Unknown return codes will
2606 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6a773a15
BB
2607 */
2608enum xdp_action {
2609 XDP_ABORTED = 0,
2610 XDP_DROP,
2611 XDP_PASS,
6ce96ca3 2612 XDP_TX,
814abfab 2613 XDP_REDIRECT,
6a773a15
BB
2614};
2615
2616/* user accessible metadata for XDP packet hook
2617 * new fields must be added to the end of this structure
2618 */
2619struct xdp_md {
2620 __u32 data;
2621 __u32 data_end;
de8f3a83 2622 __u32 data_meta;
daaf24c6 2623 /* Below access go through struct xdp_rxq_info */
02dd3291
JDB
2624 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2625 __u32 rx_queue_index; /* rxq->queue_index */
6a773a15
BB
2626};
2627
174a79ff 2628enum sk_action {
bfa64075
JF
2629 SK_DROP = 0,
2630 SK_PASS,
174a79ff
JF
2631};
2632
4f738adb
JF
2633/* user accessible metadata for SK_MSG packet hook, new fields must
2634 * be added to the end of this structure
2635 */
2636struct sk_msg_md {
2637 void *data;
2638 void *data_end;
303def35
JF
2639
2640 __u32 family;
2641 __u32 remote_ip4; /* Stored in network byte order */
2642 __u32 local_ip4; /* Stored in network byte order */
2643 __u32 remote_ip6[4]; /* Stored in network byte order */
2644 __u32 local_ip6[4]; /* Stored in network byte order */
2645 __u32 remote_port; /* Stored in network byte order */
2646 __u32 local_port; /* stored in host byte order */
4f738adb
JF
2647};
2648
2dbb9b9e
MKL
2649struct sk_reuseport_md {
2650 /*
2651 * Start of directly accessible data. It begins from
2652 * the tcp/udp header.
2653 */
2654 void *data;
2655 void *data_end; /* End of directly accessible data */
2656 /*
2657 * Total length of packet (starting from the tcp/udp header).
2658 * Note that the directly accessible bytes (data_end - data)
2659 * could be less than this "len". Those bytes could be
2660 * indirectly read by a helper "bpf_skb_load_bytes()".
2661 */
2662 __u32 len;
2663 /*
2664 * Eth protocol in the mac header (network byte order). e.g.
2665 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2666 */
2667 __u32 eth_protocol;
2668 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2669 __u32 bind_inany; /* Is sock bound to an INANY address? */
2670 __u32 hash; /* A hash of the packet 4 tuples */
2671};
2672
1e270976
MKL
2673#define BPF_TAG_SIZE 8
2674
2675struct bpf_prog_info {
2676 __u32 type;
2677 __u32 id;
2678 __u8 tag[BPF_TAG_SIZE];
2679 __u32 jited_prog_len;
2680 __u32 xlated_prog_len;
2681 __aligned_u64 jited_prog_insns;
2682 __aligned_u64 xlated_prog_insns;
cb4d2b3f
MKL
2683 __u64 load_time; /* ns since boottime */
2684 __u32 created_by_uid;
2685 __u32 nr_map_ids;
2686 __aligned_u64 map_ids;
067cae47 2687 char name[BPF_OBJ_NAME_LEN];
675fc275 2688 __u32 ifindex;
b85fab0e 2689 __u32 gpl_compatible:1;
675fc275
JK
2690 __u64 netns_dev;
2691 __u64 netns_ino;
dbecd738 2692 __u32 nr_jited_ksyms;
815581c1 2693 __u32 nr_jited_func_lens;
dbecd738 2694 __aligned_u64 jited_ksyms;
815581c1 2695 __aligned_u64 jited_func_lens;
838e9690
YS
2696 __u32 btf_id;
2697 __u32 func_info_rec_size;
2698 __aligned_u64 func_info;
11d8b82d
YS
2699 __u32 nr_func_info;
2700 __u32 nr_line_info;
c454a46b
MKL
2701 __aligned_u64 line_info;
2702 __aligned_u64 jited_line_info;
11d8b82d 2703 __u32 nr_jited_line_info;
c454a46b
MKL
2704 __u32 line_info_rec_size;
2705 __u32 jited_line_info_rec_size;
1e270976
MKL
2706} __attribute__((aligned(8)));
2707
2708struct bpf_map_info {
2709 __u32 type;
2710 __u32 id;
2711 __u32 key_size;
2712 __u32 value_size;
2713 __u32 max_entries;
2714 __u32 map_flags;
067cae47 2715 char name[BPF_OBJ_NAME_LEN];
52775b33 2716 __u32 ifindex;
36f9814a 2717 __u32 :32;
52775b33
JK
2718 __u64 netns_dev;
2719 __u64 netns_ino;
78958fca 2720 __u32 btf_id;
9b2cf328
MKL
2721 __u32 btf_key_type_id;
2722 __u32 btf_value_type_id;
1e270976
MKL
2723} __attribute__((aligned(8)));
2724
62dab84c
MKL
2725struct bpf_btf_info {
2726 __aligned_u64 btf;
2727 __u32 btf_size;
2728 __u32 id;
2729} __attribute__((aligned(8)));
2730
4fbac77d
AI
2731/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2732 * by user and intended to be used by socket (e.g. to bind to, depends on
2733 * attach attach type).
2734 */
2735struct bpf_sock_addr {
2736 __u32 user_family; /* Allows 4-byte read, but no write. */
2737 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2738 * Stored in network byte order.
2739 */
2740 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2741 * Stored in network byte order.
2742 */
2743 __u32 user_port; /* Allows 4-byte read and write.
2744 * Stored in network byte order
2745 */
2746 __u32 family; /* Allows 4-byte read, but no write */
2747 __u32 type; /* Allows 4-byte read, but no write */
2748 __u32 protocol; /* Allows 4-byte read, but no write */
1cedee13
AI
2749 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2750 * Stored in network byte order.
2751 */
2752 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2753 * Stored in network byte order.
2754 */
4fbac77d
AI
2755};
2756
40304b2a
LB
2757/* User bpf_sock_ops struct to access socket values and specify request ops
2758 * and their replies.
2759 * Some of this fields are in network (bigendian) byte order and may need
2760 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2761 * New fields can only be added at the end of this structure
2762 */
2763struct bpf_sock_ops {
2764 __u32 op;
2765 union {
de525be2
LB
2766 __u32 args[4]; /* Optionally passed to bpf program */
2767 __u32 reply; /* Returned by bpf program */
2768 __u32 replylong[4]; /* Optionally returned by bpf prog */
40304b2a
LB
2769 };
2770 __u32 family;
2771 __u32 remote_ip4; /* Stored in network byte order */
2772 __u32 local_ip4; /* Stored in network byte order */
2773 __u32 remote_ip6[4]; /* Stored in network byte order */
2774 __u32 local_ip6[4]; /* Stored in network byte order */
2775 __u32 remote_port; /* Stored in network byte order */
2776 __u32 local_port; /* stored in host byte order */
f19397a5
LB
2777 __u32 is_fullsock; /* Some TCP fields are only valid if
2778 * there is a full socket. If not, the
2779 * fields read as zero.
2780 */
2781 __u32 snd_cwnd;
2782 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
b13d8807 2783 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
44f0e430
LB
2784 __u32 state;
2785 __u32 rtt_min;
2786 __u32 snd_ssthresh;
2787 __u32 rcv_nxt;
2788 __u32 snd_nxt;
2789 __u32 snd_una;
2790 __u32 mss_cache;
2791 __u32 ecn_flags;
2792 __u32 rate_delivered;
2793 __u32 rate_interval_us;
2794 __u32 packets_out;
2795 __u32 retrans_out;
2796 __u32 total_retrans;
2797 __u32 segs_in;
2798 __u32 data_segs_in;
2799 __u32 segs_out;
2800 __u32 data_segs_out;
2801 __u32 lost_out;
2802 __u32 sacked_out;
2803 __u32 sk_txhash;
2804 __u64 bytes_received;
2805 __u64 bytes_acked;
40304b2a
LB
2806};
2807
b13d8807 2808/* Definitions for bpf_sock_ops_cb_flags */
f89013f6 2809#define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
a31ad29e 2810#define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
d4487491
LB
2811#define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2812#define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
b13d8807
LB
2813 * supported cb flags
2814 */
2815
40304b2a
LB
2816/* List of known BPF sock_ops operators.
2817 * New entries can only be added at the end
2818 */
2819enum {
2820 BPF_SOCK_OPS_VOID,
8550f328
LB
2821 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2822 * -1 if default value should be used
2823 */
13d3b1eb
LB
2824 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2825 * window (in packets) or -1 if default
2826 * value should be used
2827 */
9872a4bd
LB
2828 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2829 * active connection is initialized
2830 */
2831 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2832 * active connection is
2833 * established
2834 */
2835 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2836 * passive connection is
2837 * established
2838 */
91b5b21c
LB
2839 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2840 * needs ECN
2841 */
e6546ef6
LB
2842 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2843 * based on the path and may be
2844 * dependent on the congestion control
2845 * algorithm. In general it indicates
2846 * a congestion threshold. RTTs above
2847 * this indicate congestion
2848 */
f89013f6
LB
2849 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2850 * Arg1: value of icsk_retransmits
2851 * Arg2: value of icsk_rto
2852 * Arg3: whether RTO has expired
2853 */
a31ad29e
LB
2854 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2855 * Arg1: sequence number of 1st byte
2856 * Arg2: # segments
2857 * Arg3: return value of
2858 * tcp_transmit_skb (0 => success)
2859 */
d4487491
LB
2860 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2861 * Arg1: old_state
2862 * Arg2: new_state
2863 */
f333ee0c
AI
2864 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2865 * socket transition to LISTEN state.
2866 */
d4487491
LB
2867};
2868
2869/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2870 * changes between the TCP and BPF versions. Ideally this should never happen.
2871 * If it does, we need to add code to convert them before calling
2872 * the BPF sock_ops function.
2873 */
2874enum {
2875 BPF_TCP_ESTABLISHED = 1,
2876 BPF_TCP_SYN_SENT,
2877 BPF_TCP_SYN_RECV,
2878 BPF_TCP_FIN_WAIT1,
2879 BPF_TCP_FIN_WAIT2,
2880 BPF_TCP_TIME_WAIT,
2881 BPF_TCP_CLOSE,
2882 BPF_TCP_CLOSE_WAIT,
2883 BPF_TCP_LAST_ACK,
2884 BPF_TCP_LISTEN,
2885 BPF_TCP_CLOSING, /* Now a valid state */
2886 BPF_TCP_NEW_SYN_RECV,
2887
2888 BPF_TCP_MAX_STATES /* Leave at the end! */
40304b2a
LB
2889};
2890
fc747810 2891#define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
13bf9641 2892#define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
fc747810 2893
908432ca
YS
2894struct bpf_perf_event_value {
2895 __u64 counter;
2896 __u64 enabled;
2897 __u64 running;
2898};
2899
ebc614f6
RG
2900#define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2901#define BPF_DEVCG_ACC_READ (1ULL << 1)
2902#define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2903
2904#define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2905#define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2906
2907struct bpf_cgroup_dev_ctx {
06ef0ccb
YS
2908 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2909 __u32 access_type;
ebc614f6
RG
2910 __u32 major;
2911 __u32 minor;
2912};
2913
c4f6699d
AS
2914struct bpf_raw_tracepoint_args {
2915 __u64 args[0];
2916};
2917
87f5fc7e
DA
2918/* DIRECT: Skip the FIB rules and go to FIB table associated with device
2919 * OUTPUT: Do lookup from egress perspective; default is ingress
2920 */
2921#define BPF_FIB_LOOKUP_DIRECT BIT(0)
2922#define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2923
4c79579b
DA
2924enum {
2925 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2926 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2927 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2928 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2929 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2930 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2931 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2932 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2933 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2934};
2935
87f5fc7e 2936struct bpf_fib_lookup {
fa898d76
DA
2937 /* input: network family for lookup (AF_INET, AF_INET6)
2938 * output: network family of egress nexthop
2939 */
2940 __u8 family;
87f5fc7e
DA
2941
2942 /* set if lookup is to consider L4 data - e.g., FIB rules */
2943 __u8 l4_protocol;
2944 __be16 sport;
2945 __be16 dport;
2946
2947 /* total length of packet from network header - used for MTU check */
2948 __u16 tot_len;
4c79579b
DA
2949
2950 /* input: L3 device index for lookup
2951 * output: device index from FIB lookup
2952 */
2953 __u32 ifindex;
87f5fc7e
DA
2954
2955 union {
2956 /* inputs to lookup */
2957 __u8 tos; /* AF_INET */
bd3a08aa 2958 __be32 flowinfo; /* AF_INET6, flow_label + priority */
87f5fc7e 2959
fa898d76
DA
2960 /* output: metric of fib result (IPv4/IPv6 only) */
2961 __u32 rt_metric;
87f5fc7e
DA
2962 };
2963
2964 union {
87f5fc7e
DA
2965 __be32 ipv4_src;
2966 __u32 ipv6_src[4]; /* in6_addr; network order */
2967 };
2968
fa898d76
DA
2969 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2970 * network header. output: bpf_fib_lookup sets to gateway address
2971 * if FIB lookup returns gateway route
87f5fc7e
DA
2972 */
2973 union {
87f5fc7e
DA
2974 __be32 ipv4_dst;
2975 __u32 ipv6_dst[4]; /* in6_addr; network order */
2976 };
2977
2978 /* output */
2979 __be16 h_vlan_proto;
2980 __be16 h_vlan_TCI;
2981 __u8 smac[6]; /* ETH_ALEN */
2982 __u8 dmac[6]; /* ETH_ALEN */
2983};
2984
41bdc4b4
YS
2985enum bpf_task_fd_type {
2986 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
2987 BPF_FD_TYPE_TRACEPOINT, /* tp name */
2988 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
2989 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
2990 BPF_FD_TYPE_UPROBE, /* filename + offset */
2991 BPF_FD_TYPE_URETPROBE, /* filename + offset */
2992};
2993
d58e468b
PP
2994struct bpf_flow_keys {
2995 __u16 nhoff;
2996 __u16 thoff;
2997 __u16 addr_proto; /* ETH_P_* of valid addrs */
2998 __u8 is_frag;
2999 __u8 is_first_frag;
3000 __u8 is_encap;
3001 __u8 ip_proto;
3002 __be16 n_proto;
3003 __be16 sport;
3004 __be16 dport;
3005 union {
3006 struct {
3007 __be32 ipv4_src;
3008 __be32 ipv4_dst;
3009 };
3010 struct {
3011 __u32 ipv6_src[4]; /* in6_addr; network order */
3012 __u32 ipv6_dst[4]; /* in6_addr; network order */
3013 };
3014 };
3015};
3016
838e9690 3017struct bpf_func_info {
d30d42e0 3018 __u32 insn_off;
838e9690
YS
3019 __u32 type_id;
3020};
3021
c454a46b
MKL
3022#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3023#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3024
3025struct bpf_line_info {
3026 __u32 insn_off;
3027 __u32 file_name_off;
3028 __u32 line_off;
3029 __u32 line_col;
3030};
3031
daedfb22 3032#endif /* _UAPI__LINUX_BPF_H__ */