]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/shared/linux/bpf.h
tree-wide: drop several missing_*.h and import relevant headers from kernel-5.0
[thirdparty/systemd.git] / src / shared / linux / bpf.h
1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 */
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
10
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
13
14 /* Extended instruction set based on top of classic BPF */
15
16 /* instruction classes */
17 #define BPF_ALU64 0x07 /* alu mode in double word width */
18
19 /* ld/ldx fields */
20 #define BPF_DW 0x18 /* double word (64-bit) */
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
34 /* jmp encodings */
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
44
45 /* Register numbers */
46 enum {
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
64 struct 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
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct 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
78 struct bpf_cgroup_storage_key {
79 __u64 cgroup_inode_id; /* cgroup inode id */
80 __u32 attach_type; /* program attach type */
81 };
82
83 /* BPF syscall commands, see bpf(2) man-page for details. */
84 enum bpf_cmd {
85 BPF_MAP_CREATE,
86 BPF_MAP_LOOKUP_ELEM,
87 BPF_MAP_UPDATE_ELEM,
88 BPF_MAP_DELETE_ELEM,
89 BPF_MAP_GET_NEXT_KEY,
90 BPF_PROG_LOAD,
91 BPF_OBJ_PIN,
92 BPF_OBJ_GET,
93 BPF_PROG_ATTACH,
94 BPF_PROG_DETACH,
95 BPF_PROG_TEST_RUN,
96 BPF_PROG_GET_NEXT_ID,
97 BPF_MAP_GET_NEXT_ID,
98 BPF_PROG_GET_FD_BY_ID,
99 BPF_MAP_GET_FD_BY_ID,
100 BPF_OBJ_GET_INFO_BY_FD,
101 BPF_PROG_QUERY,
102 BPF_RAW_TRACEPOINT_OPEN,
103 BPF_BTF_LOAD,
104 BPF_BTF_GET_FD_BY_ID,
105 BPF_TASK_FD_QUERY,
106 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
107 };
108
109 enum bpf_map_type {
110 BPF_MAP_TYPE_UNSPEC,
111 BPF_MAP_TYPE_HASH,
112 BPF_MAP_TYPE_ARRAY,
113 BPF_MAP_TYPE_PROG_ARRAY,
114 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
115 BPF_MAP_TYPE_PERCPU_HASH,
116 BPF_MAP_TYPE_PERCPU_ARRAY,
117 BPF_MAP_TYPE_STACK_TRACE,
118 BPF_MAP_TYPE_CGROUP_ARRAY,
119 BPF_MAP_TYPE_LRU_HASH,
120 BPF_MAP_TYPE_LRU_PERCPU_HASH,
121 BPF_MAP_TYPE_LPM_TRIE,
122 BPF_MAP_TYPE_ARRAY_OF_MAPS,
123 BPF_MAP_TYPE_HASH_OF_MAPS,
124 BPF_MAP_TYPE_DEVMAP,
125 BPF_MAP_TYPE_SOCKMAP,
126 BPF_MAP_TYPE_CPUMAP,
127 BPF_MAP_TYPE_XSKMAP,
128 BPF_MAP_TYPE_SOCKHASH,
129 BPF_MAP_TYPE_CGROUP_STORAGE,
130 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
131 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
132 BPF_MAP_TYPE_QUEUE,
133 BPF_MAP_TYPE_STACK,
134 };
135
136 /* Note that tracing related programs such as
137 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
138 * are not subject to a stable API since kernel internal data
139 * structures can change from release to release and may
140 * therefore break existing tracing BPF programs. Tracing BPF
141 * programs correspond to /a/ specific kernel which is to be
142 * analyzed, and not /a/ specific kernel /and/ all future ones.
143 */
144 enum bpf_prog_type {
145 BPF_PROG_TYPE_UNSPEC,
146 BPF_PROG_TYPE_SOCKET_FILTER,
147 BPF_PROG_TYPE_KPROBE,
148 BPF_PROG_TYPE_SCHED_CLS,
149 BPF_PROG_TYPE_SCHED_ACT,
150 BPF_PROG_TYPE_TRACEPOINT,
151 BPF_PROG_TYPE_XDP,
152 BPF_PROG_TYPE_PERF_EVENT,
153 BPF_PROG_TYPE_CGROUP_SKB,
154 BPF_PROG_TYPE_CGROUP_SOCK,
155 BPF_PROG_TYPE_LWT_IN,
156 BPF_PROG_TYPE_LWT_OUT,
157 BPF_PROG_TYPE_LWT_XMIT,
158 BPF_PROG_TYPE_SOCK_OPS,
159 BPF_PROG_TYPE_SK_SKB,
160 BPF_PROG_TYPE_CGROUP_DEVICE,
161 BPF_PROG_TYPE_SK_MSG,
162 BPF_PROG_TYPE_RAW_TRACEPOINT,
163 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
164 BPF_PROG_TYPE_LWT_SEG6LOCAL,
165 BPF_PROG_TYPE_LIRC_MODE2,
166 BPF_PROG_TYPE_SK_REUSEPORT,
167 BPF_PROG_TYPE_FLOW_DISSECTOR,
168 };
169
170 enum bpf_attach_type {
171 BPF_CGROUP_INET_INGRESS,
172 BPF_CGROUP_INET_EGRESS,
173 BPF_CGROUP_INET_SOCK_CREATE,
174 BPF_CGROUP_SOCK_OPS,
175 BPF_SK_SKB_STREAM_PARSER,
176 BPF_SK_SKB_STREAM_VERDICT,
177 BPF_CGROUP_DEVICE,
178 BPF_SK_MSG_VERDICT,
179 BPF_CGROUP_INET4_BIND,
180 BPF_CGROUP_INET6_BIND,
181 BPF_CGROUP_INET4_CONNECT,
182 BPF_CGROUP_INET6_CONNECT,
183 BPF_CGROUP_INET4_POST_BIND,
184 BPF_CGROUP_INET6_POST_BIND,
185 BPF_CGROUP_UDP4_SENDMSG,
186 BPF_CGROUP_UDP6_SENDMSG,
187 BPF_LIRC_MODE2,
188 BPF_FLOW_DISSECTOR,
189 __MAX_BPF_ATTACH_TYPE
190 };
191
192 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
193
194 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
195 *
196 * NONE(default): No further bpf programs allowed in the subtree.
197 *
198 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
199 * the program in this cgroup yields to sub-cgroup program.
200 *
201 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
202 * that cgroup program gets run in addition to the program in this cgroup.
203 *
204 * Only one program is allowed to be attached to a cgroup with
205 * NONE or BPF_F_ALLOW_OVERRIDE flag.
206 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
207 * release old program and attach the new one. Attach flags has to match.
208 *
209 * Multiple programs are allowed to be attached to a cgroup with
210 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
211 * (those that were attached first, run first)
212 * The programs of sub-cgroup are executed first, then programs of
213 * this cgroup and then programs of parent cgroup.
214 * When children program makes decision (like picking TCP CA or sock bind)
215 * parent program has a chance to override it.
216 *
217 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
218 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
219 * Ex1:
220 * cgrp1 (MULTI progs A, B) ->
221 * cgrp2 (OVERRIDE prog C) ->
222 * cgrp3 (MULTI prog D) ->
223 * cgrp4 (OVERRIDE prog E) ->
224 * cgrp5 (NONE prog F)
225 * the event in cgrp5 triggers execution of F,D,A,B in that order.
226 * if prog F is detached, the execution is E,D,A,B
227 * if prog F and D are detached, the execution is E,A,B
228 * if prog F, E and D are detached, the execution is C,A,B
229 *
230 * All eligible programs are executed regardless of return code from
231 * earlier programs.
232 */
233 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
234 #define BPF_F_ALLOW_MULTI (1U << 1)
235
236 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
237 * verifier will perform strict alignment checking as if the kernel
238 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
239 * and NET_IP_ALIGN defined to 2.
240 */
241 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
242
243 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
244 * verifier will allow any alignment whatsoever. On platforms
245 * with strict alignment requirements for loads ands stores (such
246 * as sparc and mips) the verifier validates that all loads and
247 * stores provably follow this requirement. This flag turns that
248 * checking and enforcement off.
249 *
250 * It is mostly used for testing when we want to validate the
251 * context and memory access aspects of the verifier, but because
252 * of an unaligned access the alignment check would trigger before
253 * the one we are interested in.
254 */
255 #define BPF_F_ANY_ALIGNMENT (1U << 1)
256
257 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
258 #define BPF_PSEUDO_MAP_FD 1
259
260 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
261 * offset to another bpf function
262 */
263 #define BPF_PSEUDO_CALL 1
264
265 /* flags for BPF_MAP_UPDATE_ELEM command */
266 #define BPF_ANY 0 /* create new element or update existing */
267 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
268 #define BPF_EXIST 2 /* update existing element */
269
270 /* flags for BPF_MAP_CREATE command */
271 #define BPF_F_NO_PREALLOC (1U << 0)
272 /* Instead of having one common LRU list in the
273 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
274 * which can scale and perform better.
275 * Note, the LRU nodes (including free nodes) cannot be moved
276 * across different LRU lists.
277 */
278 #define BPF_F_NO_COMMON_LRU (1U << 1)
279 /* Specify numa node during map creation */
280 #define BPF_F_NUMA_NODE (1U << 2)
281
282 #define BPF_OBJ_NAME_LEN 16U
283
284 /* Flags for accessing BPF object */
285 #define BPF_F_RDONLY (1U << 3)
286 #define BPF_F_WRONLY (1U << 4)
287
288 /* Flag for stack_map, store build_id+offset instead of pointer */
289 #define BPF_F_STACK_BUILD_ID (1U << 5)
290
291 /* Zero-initialize hash function seed. This should only be used for testing. */
292 #define BPF_F_ZERO_SEED (1U << 6)
293
294 /* flags for BPF_PROG_QUERY */
295 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
296
297 enum bpf_stack_build_id_status {
298 /* user space need an empty entry to identify end of a trace */
299 BPF_STACK_BUILD_ID_EMPTY = 0,
300 /* with valid build_id and offset */
301 BPF_STACK_BUILD_ID_VALID = 1,
302 /* couldn't get build_id, fallback to ip */
303 BPF_STACK_BUILD_ID_IP = 2,
304 };
305
306 #define BPF_BUILD_ID_SIZE 20
307 struct bpf_stack_build_id {
308 __s32 status;
309 unsigned char build_id[BPF_BUILD_ID_SIZE];
310 union {
311 __u64 offset;
312 __u64 ip;
313 };
314 };
315
316 union bpf_attr {
317 struct { /* anonymous struct used by BPF_MAP_CREATE command */
318 __u32 map_type; /* one of enum bpf_map_type */
319 __u32 key_size; /* size of key in bytes */
320 __u32 value_size; /* size of value in bytes */
321 __u32 max_entries; /* max number of entries in a map */
322 __u32 map_flags; /* BPF_MAP_CREATE related
323 * flags defined above.
324 */
325 __u32 inner_map_fd; /* fd pointing to the inner map */
326 __u32 numa_node; /* numa node (effective only if
327 * BPF_F_NUMA_NODE is set).
328 */
329 char map_name[BPF_OBJ_NAME_LEN];
330 __u32 map_ifindex; /* ifindex of netdev to create on */
331 __u32 btf_fd; /* fd pointing to a BTF type data */
332 __u32 btf_key_type_id; /* BTF type_id of the key */
333 __u32 btf_value_type_id; /* BTF type_id of the value */
334 };
335
336 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
337 __u32 map_fd;
338 __aligned_u64 key;
339 union {
340 __aligned_u64 value;
341 __aligned_u64 next_key;
342 };
343 __u64 flags;
344 };
345
346 struct { /* anonymous struct used by BPF_PROG_LOAD command */
347 __u32 prog_type; /* one of enum bpf_prog_type */
348 __u32 insn_cnt;
349 __aligned_u64 insns;
350 __aligned_u64 license;
351 __u32 log_level; /* verbosity level of verifier */
352 __u32 log_size; /* size of user buffer */
353 __aligned_u64 log_buf; /* user supplied buffer */
354 __u32 kern_version; /* not used */
355 __u32 prog_flags;
356 char prog_name[BPF_OBJ_NAME_LEN];
357 __u32 prog_ifindex; /* ifindex of netdev to prep for */
358 /* For some prog types expected attach type must be known at
359 * load time to verify attach type specific parts of prog
360 * (context accesses, allowed helpers, etc).
361 */
362 __u32 expected_attach_type;
363 __u32 prog_btf_fd; /* fd pointing to BTF type data */
364 __u32 func_info_rec_size; /* userspace bpf_func_info size */
365 __aligned_u64 func_info; /* func info */
366 __u32 func_info_cnt; /* number of bpf_func_info records */
367 __u32 line_info_rec_size; /* userspace bpf_line_info size */
368 __aligned_u64 line_info; /* line info */
369 __u32 line_info_cnt; /* number of bpf_line_info records */
370 };
371
372 struct { /* anonymous struct used by BPF_OBJ_* commands */
373 __aligned_u64 pathname;
374 __u32 bpf_fd;
375 __u32 file_flags;
376 };
377
378 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
379 __u32 target_fd; /* container object to attach to */
380 __u32 attach_bpf_fd; /* eBPF program to attach */
381 __u32 attach_type;
382 __u32 attach_flags;
383 };
384
385 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
386 __u32 prog_fd;
387 __u32 retval;
388 __u32 data_size_in; /* input: len of data_in */
389 __u32 data_size_out; /* input/output: len of data_out
390 * returns ENOSPC if data_out
391 * is too small.
392 */
393 __aligned_u64 data_in;
394 __aligned_u64 data_out;
395 __u32 repeat;
396 __u32 duration;
397 } test;
398
399 struct { /* anonymous struct used by BPF_*_GET_*_ID */
400 union {
401 __u32 start_id;
402 __u32 prog_id;
403 __u32 map_id;
404 __u32 btf_id;
405 };
406 __u32 next_id;
407 __u32 open_flags;
408 };
409
410 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
411 __u32 bpf_fd;
412 __u32 info_len;
413 __aligned_u64 info;
414 } info;
415
416 struct { /* anonymous struct used by BPF_PROG_QUERY command */
417 __u32 target_fd; /* container object to query */
418 __u32 attach_type;
419 __u32 query_flags;
420 __u32 attach_flags;
421 __aligned_u64 prog_ids;
422 __u32 prog_cnt;
423 } query;
424
425 struct {
426 __u64 name;
427 __u32 prog_fd;
428 } raw_tracepoint;
429
430 struct { /* anonymous struct for BPF_BTF_LOAD */
431 __aligned_u64 btf;
432 __aligned_u64 btf_log_buf;
433 __u32 btf_size;
434 __u32 btf_log_size;
435 __u32 btf_log_level;
436 };
437
438 struct {
439 __u32 pid; /* input: pid */
440 __u32 fd; /* input: fd */
441 __u32 flags; /* input: flags */
442 __u32 buf_len; /* input/output: buf len */
443 __aligned_u64 buf; /* input/output:
444 * tp_name for tracepoint
445 * symbol for kprobe
446 * filename for uprobe
447 */
448 __u32 prog_id; /* output: prod_id */
449 __u32 fd_type; /* output: BPF_FD_TYPE_* */
450 __u64 probe_offset; /* output: probe_offset */
451 __u64 probe_addr; /* output: probe_addr */
452 } task_fd_query;
453 } __attribute__((aligned(8)));
454
455 /* The description below is an attempt at providing documentation to eBPF
456 * developers about the multiple available eBPF helper functions. It can be
457 * parsed and used to produce a manual page. The workflow is the following,
458 * and requires the rst2man utility:
459 *
460 * $ ./scripts/bpf_helpers_doc.py \
461 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
462 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
463 * $ man /tmp/bpf-helpers.7
464 *
465 * Note that in order to produce this external documentation, some RST
466 * formatting is used in the descriptions to get "bold" and "italics" in
467 * manual pages. Also note that the few trailing white spaces are
468 * intentional, removing them would break paragraphs for rst2man.
469 *
470 * Start of BPF helper function descriptions:
471 *
472 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
473 * Description
474 * Perform a lookup in *map* for an entry associated to *key*.
475 * Return
476 * Map value associated to *key*, or **NULL** if no entry was
477 * found.
478 *
479 * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
480 * Description
481 * Add or update the value of the entry associated to *key* in
482 * *map* with *value*. *flags* is one of:
483 *
484 * **BPF_NOEXIST**
485 * The entry for *key* must not exist in the map.
486 * **BPF_EXIST**
487 * The entry for *key* must already exist in the map.
488 * **BPF_ANY**
489 * No condition on the existence of the entry for *key*.
490 *
491 * Flag value **BPF_NOEXIST** cannot be used for maps of types
492 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
493 * elements always exist), the helper would return an error.
494 * Return
495 * 0 on success, or a negative error in case of failure.
496 *
497 * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
498 * Description
499 * Delete entry with *key* from *map*.
500 * Return
501 * 0 on success, or a negative error in case of failure.
502 *
503 * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
504 * Description
505 * Push an element *value* in *map*. *flags* is one of:
506 *
507 * **BPF_EXIST**
508 * If the queue/stack is full, the oldest element is removed to
509 * make room for this.
510 * Return
511 * 0 on success, or a negative error in case of failure.
512 *
513 * int bpf_probe_read(void *dst, u32 size, const void *src)
514 * Description
515 * For tracing programs, safely attempt to read *size* bytes from
516 * address *src* and store the data in *dst*.
517 * Return
518 * 0 on success, or a negative error in case of failure.
519 *
520 * u64 bpf_ktime_get_ns(void)
521 * Description
522 * Return the time elapsed since system boot, in nanoseconds.
523 * Return
524 * Current *ktime*.
525 *
526 * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
527 * Description
528 * This helper is a "printk()-like" facility for debugging. It
529 * prints a message defined by format *fmt* (of size *fmt_size*)
530 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
531 * available. It can take up to three additional **u64**
532 * arguments (as an eBPF helpers, the total number of arguments is
533 * limited to five).
534 *
535 * Each time the helper is called, it appends a line to the trace.
536 * The format of the trace is customizable, and the exact output
537 * one will get depends on the options set in
538 * *\/sys/kernel/debug/tracing/trace_options* (see also the
539 * *README* file under the same directory). However, it usually
540 * defaults to something like:
541 *
542 * ::
543 *
544 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
545 *
546 * In the above:
547 *
548 * * ``telnet`` is the name of the current task.
549 * * ``470`` is the PID of the current task.
550 * * ``001`` is the CPU number on which the task is
551 * running.
552 * * In ``.N..``, each character refers to a set of
553 * options (whether irqs are enabled, scheduling
554 * options, whether hard/softirqs are running, level of
555 * preempt_disabled respectively). **N** means that
556 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
557 * are set.
558 * * ``419421.045894`` is a timestamp.
559 * * ``0x00000001`` is a fake value used by BPF for the
560 * instruction pointer register.
561 * * ``<formatted msg>`` is the message formatted with
562 * *fmt*.
563 *
564 * The conversion specifiers supported by *fmt* are similar, but
565 * more limited than for printk(). They are **%d**, **%i**,
566 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
567 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
568 * of field, padding with zeroes, etc.) is available, and the
569 * helper will return **-EINVAL** (but print nothing) if it
570 * encounters an unknown specifier.
571 *
572 * Also, note that **bpf_trace_printk**\ () is slow, and should
573 * only be used for debugging purposes. For this reason, a notice
574 * bloc (spanning several lines) is printed to kernel logs and
575 * states that the helper should not be used "for production use"
576 * the first time this helper is used (or more precisely, when
577 * **trace_printk**\ () buffers are allocated). For passing values
578 * to user space, perf events should be preferred.
579 * Return
580 * The number of bytes written to the buffer, or a negative error
581 * in case of failure.
582 *
583 * u32 bpf_get_prandom_u32(void)
584 * Description
585 * Get a pseudo-random number.
586 *
587 * From a security point of view, this helper uses its own
588 * pseudo-random internal state, and cannot be used to infer the
589 * seed of other random functions in the kernel. However, it is
590 * essential to note that the generator used by the helper is not
591 * cryptographically secure.
592 * Return
593 * A random 32-bit unsigned value.
594 *
595 * u32 bpf_get_smp_processor_id(void)
596 * Description
597 * Get the SMP (symmetric multiprocessing) processor id. Note that
598 * all programs run with preemption disabled, which means that the
599 * SMP processor id is stable during all the execution of the
600 * program.
601 * Return
602 * The SMP id of the processor running the program.
603 *
604 * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
605 * Description
606 * Store *len* bytes from address *from* into the packet
607 * associated to *skb*, at *offset*. *flags* are a combination of
608 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
609 * checksum for the packet after storing the bytes) and
610 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
611 * **->swhash** and *skb*\ **->l4hash** to 0).
612 *
613 * A call to this helper is susceptible to change the underlaying
614 * packet buffer. Therefore, at load time, all checks on pointers
615 * previously done by the verifier are invalidated and must be
616 * performed again, if the helper is used in combination with
617 * direct packet access.
618 * Return
619 * 0 on success, or a negative error in case of failure.
620 *
621 * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
622 * Description
623 * Recompute the layer 3 (e.g. IP) checksum for the packet
624 * associated to *skb*. Computation is incremental, so the helper
625 * must know the former value of the header field that was
626 * modified (*from*), the new value of this field (*to*), and the
627 * number of bytes (2 or 4) for this field, stored in *size*.
628 * Alternatively, it is possible to store the difference between
629 * the previous and the new values of the header field in *to*, by
630 * setting *from* and *size* to 0. For both methods, *offset*
631 * indicates the location of the IP checksum within the packet.
632 *
633 * This helper works in combination with **bpf_csum_diff**\ (),
634 * which does not update the checksum in-place, but offers more
635 * flexibility and can handle sizes larger than 2 or 4 for the
636 * checksum to update.
637 *
638 * A call to this helper is susceptible to change the underlaying
639 * packet buffer. Therefore, at load time, all checks on pointers
640 * previously done by the verifier are invalidated and must be
641 * performed again, if the helper is used in combination with
642 * direct packet access.
643 * Return
644 * 0 on success, or a negative error in case of failure.
645 *
646 * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
647 * Description
648 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
649 * packet associated to *skb*. Computation is incremental, so the
650 * helper must know the former value of the header field that was
651 * modified (*from*), the new value of this field (*to*), and the
652 * number of bytes (2 or 4) for this field, stored on the lowest
653 * four bits of *flags*. Alternatively, it is possible to store
654 * the difference between the previous and the new values of the
655 * header field in *to*, by setting *from* and the four lowest
656 * bits of *flags* to 0. For both methods, *offset* indicates the
657 * location of the IP checksum within the packet. In addition to
658 * the size of the field, *flags* can be added (bitwise OR) actual
659 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
660 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
661 * for updates resulting in a null checksum the value is set to
662 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
663 * the checksum is to be computed against a pseudo-header.
664 *
665 * This helper works in combination with **bpf_csum_diff**\ (),
666 * which does not update the checksum in-place, but offers more
667 * flexibility and can handle sizes larger than 2 or 4 for the
668 * checksum to update.
669 *
670 * A call to this helper is susceptible to change the underlaying
671 * packet buffer. Therefore, at load time, all checks on pointers
672 * previously done by the verifier are invalidated and must be
673 * performed again, if the helper is used in combination with
674 * direct packet access.
675 * Return
676 * 0 on success, or a negative error in case of failure.
677 *
678 * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
679 * Description
680 * This special helper is used to trigger a "tail call", or in
681 * other words, to jump into another eBPF program. The same stack
682 * frame is used (but values on stack and in registers for the
683 * caller are not accessible to the callee). This mechanism allows
684 * for program chaining, either for raising the maximum number of
685 * available eBPF instructions, or to execute given programs in
686 * conditional blocks. For security reasons, there is an upper
687 * limit to the number of successive tail calls that can be
688 * performed.
689 *
690 * Upon call of this helper, the program attempts to jump into a
691 * program referenced at index *index* in *prog_array_map*, a
692 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
693 * *ctx*, a pointer to the context.
694 *
695 * If the call succeeds, the kernel immediately runs the first
696 * instruction of the new program. This is not a function call,
697 * and it never returns to the previous program. If the call
698 * fails, then the helper has no effect, and the caller continues
699 * to run its subsequent instructions. A call can fail if the
700 * destination program for the jump does not exist (i.e. *index*
701 * is superior to the number of entries in *prog_array_map*), or
702 * if the maximum number of tail calls has been reached for this
703 * chain of programs. This limit is defined in the kernel by the
704 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
705 * which is currently set to 32.
706 * Return
707 * 0 on success, or a negative error in case of failure.
708 *
709 * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
710 * Description
711 * Clone and redirect the packet associated to *skb* to another
712 * net device of index *ifindex*. Both ingress and egress
713 * interfaces can be used for redirection. The **BPF_F_INGRESS**
714 * value in *flags* is used to make the distinction (ingress path
715 * is selected if the flag is present, egress path otherwise).
716 * This is the only flag supported for now.
717 *
718 * In comparison with **bpf_redirect**\ () helper,
719 * **bpf_clone_redirect**\ () has the associated cost of
720 * duplicating the packet buffer, but this can be executed out of
721 * the eBPF program. Conversely, **bpf_redirect**\ () is more
722 * efficient, but it is handled through an action code where the
723 * redirection happens only after the eBPF program has returned.
724 *
725 * A call to this helper is susceptible to change the underlaying
726 * packet buffer. Therefore, at load time, all checks on pointers
727 * previously done by the verifier are invalidated and must be
728 * performed again, if the helper is used in combination with
729 * direct packet access.
730 * Return
731 * 0 on success, or a negative error in case of failure.
732 *
733 * u64 bpf_get_current_pid_tgid(void)
734 * Return
735 * A 64-bit integer containing the current tgid and pid, and
736 * created as such:
737 * *current_task*\ **->tgid << 32 \|**
738 * *current_task*\ **->pid**.
739 *
740 * u64 bpf_get_current_uid_gid(void)
741 * Return
742 * A 64-bit integer containing the current GID and UID, and
743 * created as such: *current_gid* **<< 32 \|** *current_uid*.
744 *
745 * int bpf_get_current_comm(char *buf, u32 size_of_buf)
746 * Description
747 * Copy the **comm** attribute of the current task into *buf* of
748 * *size_of_buf*. The **comm** attribute contains the name of
749 * the executable (excluding the path) for the current task. The
750 * *size_of_buf* must be strictly positive. On success, the
751 * helper makes sure that the *buf* is NUL-terminated. On failure,
752 * it is filled with zeroes.
753 * Return
754 * 0 on success, or a negative error in case of failure.
755 *
756 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
757 * Description
758 * Retrieve the classid for the current task, i.e. for the net_cls
759 * cgroup to which *skb* belongs.
760 *
761 * This helper can be used on TC egress path, but not on ingress.
762 *
763 * The net_cls cgroup provides an interface to tag network packets
764 * based on a user-provided identifier for all traffic coming from
765 * the tasks belonging to the related cgroup. See also the related
766 * kernel documentation, available from the Linux sources in file
767 * *Documentation/cgroup-v1/net_cls.txt*.
768 *
769 * The Linux kernel has two versions for cgroups: there are
770 * cgroups v1 and cgroups v2. Both are available to users, who can
771 * use a mixture of them, but note that the net_cls cgroup is for
772 * cgroup v1 only. This makes it incompatible with BPF programs
773 * run on cgroups, which is a cgroup-v2-only feature (a socket can
774 * only hold data for one version of cgroups at a time).
775 *
776 * This helper is only available is the kernel was compiled with
777 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
778 * "**y**" or to "**m**".
779 * Return
780 * The classid, or 0 for the default unconfigured classid.
781 *
782 * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
783 * Description
784 * Push a *vlan_tci* (VLAN tag control information) of protocol
785 * *vlan_proto* to the packet associated to *skb*, then update
786 * the checksum. Note that if *vlan_proto* is different from
787 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
788 * be **ETH_P_8021Q**.
789 *
790 * A call to this helper is susceptible to change the underlaying
791 * packet buffer. Therefore, at load time, all checks on pointers
792 * previously done by the verifier are invalidated and must be
793 * performed again, if the helper is used in combination with
794 * direct packet access.
795 * Return
796 * 0 on success, or a negative error in case of failure.
797 *
798 * int bpf_skb_vlan_pop(struct sk_buff *skb)
799 * Description
800 * Pop a VLAN header from the packet associated to *skb*.
801 *
802 * A call to this helper is susceptible to change the underlaying
803 * packet buffer. Therefore, at load time, all checks on pointers
804 * previously done by the verifier are invalidated and must be
805 * performed again, if the helper is used in combination with
806 * direct packet access.
807 * Return
808 * 0 on success, or a negative error in case of failure.
809 *
810 * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
811 * Description
812 * Get tunnel metadata. This helper takes a pointer *key* to an
813 * empty **struct bpf_tunnel_key** of **size**, that will be
814 * filled with tunnel metadata for the packet associated to *skb*.
815 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
816 * indicates that the tunnel is based on IPv6 protocol instead of
817 * IPv4.
818 *
819 * The **struct bpf_tunnel_key** is an object that generalizes the
820 * principal parameters used by various tunneling protocols into a
821 * single struct. This way, it can be used to easily make a
822 * decision based on the contents of the encapsulation header,
823 * "summarized" in this struct. In particular, it holds the IP
824 * address of the remote end (IPv4 or IPv6, depending on the case)
825 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
826 * this struct exposes the *key*\ **->tunnel_id**, which is
827 * generally mapped to a VNI (Virtual Network Identifier), making
828 * it programmable together with the **bpf_skb_set_tunnel_key**\
829 * () helper.
830 *
831 * Let's imagine that the following code is part of a program
832 * attached to the TC ingress interface, on one end of a GRE
833 * tunnel, and is supposed to filter out all messages coming from
834 * remote ends with IPv4 address other than 10.0.0.1:
835 *
836 * ::
837 *
838 * int ret;
839 * struct bpf_tunnel_key key = {};
840 *
841 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
842 * if (ret < 0)
843 * return TC_ACT_SHOT; // drop packet
844 *
845 * if (key.remote_ipv4 != 0x0a000001)
846 * return TC_ACT_SHOT; // drop packet
847 *
848 * return TC_ACT_OK; // accept packet
849 *
850 * This interface can also be used with all encapsulation devices
851 * that can operate in "collect metadata" mode: instead of having
852 * one network device per specific configuration, the "collect
853 * metadata" mode only requires a single device where the
854 * configuration can be extracted from this helper.
855 *
856 * This can be used together with various tunnels such as VXLan,
857 * Geneve, GRE or IP in IP (IPIP).
858 * Return
859 * 0 on success, or a negative error in case of failure.
860 *
861 * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
862 * Description
863 * Populate tunnel metadata for packet associated to *skb.* The
864 * tunnel metadata is set to the contents of *key*, of *size*. The
865 * *flags* can be set to a combination of the following values:
866 *
867 * **BPF_F_TUNINFO_IPV6**
868 * Indicate that the tunnel is based on IPv6 protocol
869 * instead of IPv4.
870 * **BPF_F_ZERO_CSUM_TX**
871 * For IPv4 packets, add a flag to tunnel metadata
872 * indicating that checksum computation should be skipped
873 * and checksum set to zeroes.
874 * **BPF_F_DONT_FRAGMENT**
875 * Add a flag to tunnel metadata indicating that the
876 * packet should not be fragmented.
877 * **BPF_F_SEQ_NUMBER**
878 * Add a flag to tunnel metadata indicating that a
879 * sequence number should be added to tunnel header before
880 * sending the packet. This flag was added for GRE
881 * encapsulation, but might be used with other protocols
882 * as well in the future.
883 *
884 * Here is a typical usage on the transmit path:
885 *
886 * ::
887 *
888 * struct bpf_tunnel_key key;
889 * populate key ...
890 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
891 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
892 *
893 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
894 * helper for additional information.
895 * Return
896 * 0 on success, or a negative error in case of failure.
897 *
898 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
899 * Description
900 * Read the value of a perf event counter. This helper relies on a
901 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
902 * the perf event counter is selected when *map* is updated with
903 * perf event file descriptors. The *map* is an array whose size
904 * is the number of available CPUs, and each cell contains a value
905 * relative to one CPU. The value to retrieve is indicated by
906 * *flags*, that contains the index of the CPU to look up, masked
907 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
908 * **BPF_F_CURRENT_CPU** to indicate that the value for the
909 * current CPU should be retrieved.
910 *
911 * Note that before Linux 4.13, only hardware perf event can be
912 * retrieved.
913 *
914 * Also, be aware that the newer helper
915 * **bpf_perf_event_read_value**\ () is recommended over
916 * **bpf_perf_event_read**\ () in general. The latter has some ABI
917 * quirks where error and counter value are used as a return code
918 * (which is wrong to do since ranges may overlap). This issue is
919 * fixed with **bpf_perf_event_read_value**\ (), which at the same
920 * time provides more features over the **bpf_perf_event_read**\
921 * () interface. Please refer to the description of
922 * **bpf_perf_event_read_value**\ () for details.
923 * Return
924 * The value of the perf event counter read from the map, or a
925 * negative error code in case of failure.
926 *
927 * int bpf_redirect(u32 ifindex, u64 flags)
928 * Description
929 * Redirect the packet to another net device of index *ifindex*.
930 * This helper is somewhat similar to **bpf_clone_redirect**\
931 * (), except that the packet is not cloned, which provides
932 * increased performance.
933 *
934 * Except for XDP, both ingress and egress interfaces can be used
935 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
936 * to make the distinction (ingress path is selected if the flag
937 * is present, egress path otherwise). Currently, XDP only
938 * supports redirection to the egress interface, and accepts no
939 * flag at all.
940 *
941 * The same effect can be attained with the more generic
942 * **bpf_redirect_map**\ (), which requires specific maps to be
943 * used but offers better performance.
944 * Return
945 * For XDP, the helper returns **XDP_REDIRECT** on success or
946 * **XDP_ABORTED** on error. For other program types, the values
947 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
948 * error.
949 *
950 * u32 bpf_get_route_realm(struct sk_buff *skb)
951 * Description
952 * Retrieve the realm or the route, that is to say the
953 * **tclassid** field of the destination for the *skb*. The
954 * indentifier retrieved is a user-provided tag, similar to the
955 * one used with the net_cls cgroup (see description for
956 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
957 * held by a route (a destination entry), not by a task.
958 *
959 * Retrieving this identifier works with the clsact TC egress hook
960 * (see also **tc-bpf(8)**), or alternatively on conventional
961 * classful egress qdiscs, but not on TC ingress path. In case of
962 * clsact TC egress hook, this has the advantage that, internally,
963 * the destination entry has not been dropped yet in the transmit
964 * path. Therefore, the destination entry does not need to be
965 * artificially held via **netif_keep_dst**\ () for a classful
966 * qdisc until the *skb* is freed.
967 *
968 * This helper is available only if the kernel was compiled with
969 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
970 * Return
971 * The realm of the route for the packet associated to *skb*, or 0
972 * if none was found.
973 *
974 * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
975 * Description
976 * Write raw *data* blob into a special BPF perf event held by
977 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
978 * event must have the following attributes: **PERF_SAMPLE_RAW**
979 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
980 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
981 *
982 * The *flags* are used to indicate the index in *map* for which
983 * the value must be put, masked with **BPF_F_INDEX_MASK**.
984 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
985 * to indicate that the index of the current CPU core should be
986 * used.
987 *
988 * The value to write, of *size*, is passed through eBPF stack and
989 * pointed by *data*.
990 *
991 * The context of the program *ctx* needs also be passed to the
992 * helper.
993 *
994 * On user space, a program willing to read the values needs to
995 * call **perf_event_open**\ () on the perf event (either for
996 * one or for all CPUs) and to store the file descriptor into the
997 * *map*. This must be done before the eBPF program can send data
998 * into it. An example is available in file
999 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1000 * tree (the eBPF program counterpart is in
1001 * *samples/bpf/trace_output_kern.c*).
1002 *
1003 * **bpf_perf_event_output**\ () achieves better performance
1004 * than **bpf_trace_printk**\ () for sharing data with user
1005 * space, and is much better suitable for streaming data from eBPF
1006 * programs.
1007 *
1008 * Note that this helper is not restricted to tracing use cases
1009 * and can be used with programs attached to TC or XDP as well,
1010 * where it allows for passing data to user space listeners. Data
1011 * can be:
1012 *
1013 * * Only custom structs,
1014 * * Only the packet payload, or
1015 * * A combination of both.
1016 * Return
1017 * 0 on success, or a negative error in case of failure.
1018 *
1019 * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
1020 * Description
1021 * This helper was provided as an easy way to load data from a
1022 * packet. It can be used to load *len* bytes from *offset* from
1023 * the packet associated to *skb*, into the buffer pointed by
1024 * *to*.
1025 *
1026 * Since Linux 4.7, usage of this helper has mostly been replaced
1027 * by "direct packet access", enabling packet data to be
1028 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
1029 * pointing respectively to the first byte of packet data and to
1030 * the byte after the last byte of packet data. However, it
1031 * remains useful if one wishes to read large quantities of data
1032 * at once from a packet into the eBPF stack.
1033 * Return
1034 * 0 on success, or a negative error in case of failure.
1035 *
1036 * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
1037 * Description
1038 * Walk a user or a kernel stack and return its id. To achieve
1039 * this, the helper needs *ctx*, which is a pointer to the context
1040 * on which the tracing program is executed, and a pointer to a
1041 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
1042 *
1043 * The last argument, *flags*, holds the number of stack frames to
1044 * skip (from 0 to 255), masked with
1045 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1046 * a combination of the following flags:
1047 *
1048 * **BPF_F_USER_STACK**
1049 * Collect a user space stack instead of a kernel stack.
1050 * **BPF_F_FAST_STACK_CMP**
1051 * Compare stacks by hash only.
1052 * **BPF_F_REUSE_STACKID**
1053 * If two different stacks hash into the same *stackid*,
1054 * discard the old one.
1055 *
1056 * The stack id retrieved is a 32 bit long integer handle which
1057 * can be further combined with other data (including other stack
1058 * ids) and used as a key into maps. This can be useful for
1059 * generating a variety of graphs (such as flame graphs or off-cpu
1060 * graphs).
1061 *
1062 * For walking a stack, this helper is an improvement over
1063 * **bpf_probe_read**\ (), which can be used with unrolled loops
1064 * but is not efficient and consumes a lot of eBPF instructions.
1065 * Instead, **bpf_get_stackid**\ () can collect up to
1066 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
1067 * this limit can be controlled with the **sysctl** program, and
1068 * that it should be manually increased in order to profile long
1069 * user stacks (such as stacks for Java programs). To do so, use:
1070 *
1071 * ::
1072 *
1073 * # sysctl kernel.perf_event_max_stack=<new value>
1074 * Return
1075 * The positive or null stack id on success, or a negative error
1076 * in case of failure.
1077 *
1078 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
1079 * Description
1080 * Compute a checksum difference, from the raw buffer pointed by
1081 * *from*, of length *from_size* (that must be a multiple of 4),
1082 * towards the raw buffer pointed by *to*, of size *to_size*
1083 * (same remark). An optional *seed* can be added to the value
1084 * (this can be cascaded, the seed may come from a previous call
1085 * to the helper).
1086 *
1087 * This is flexible enough to be used in several ways:
1088 *
1089 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
1090 * checksum, it can be used when pushing new data.
1091 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
1092 * checksum, it can be used when removing data from a packet.
1093 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
1094 * can be used to compute a diff. Note that *from_size* and
1095 * *to_size* do not need to be equal.
1096 *
1097 * This helper can be used in combination with
1098 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
1099 * which one can feed in the difference computed with
1100 * **bpf_csum_diff**\ ().
1101 * Return
1102 * The checksum result, or a negative error code in case of
1103 * failure.
1104 *
1105 * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1106 * Description
1107 * Retrieve tunnel options metadata for the packet associated to
1108 * *skb*, and store the raw tunnel option data to the buffer *opt*
1109 * of *size*.
1110 *
1111 * This helper can be used with encapsulation devices that can
1112 * operate in "collect metadata" mode (please refer to the related
1113 * note in the description of **bpf_skb_get_tunnel_key**\ () for
1114 * more details). A particular example where this can be used is
1115 * in combination with the Geneve encapsulation protocol, where it
1116 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
1117 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
1118 * the eBPF program. This allows for full customization of these
1119 * headers.
1120 * Return
1121 * The size of the option data retrieved.
1122 *
1123 * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
1124 * Description
1125 * Set tunnel options metadata for the packet associated to *skb*
1126 * to the option data contained in the raw buffer *opt* of *size*.
1127 *
1128 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
1129 * helper for additional information.
1130 * Return
1131 * 0 on success, or a negative error in case of failure.
1132 *
1133 * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
1134 * Description
1135 * Change the protocol of the *skb* to *proto*. Currently
1136 * supported are transition from IPv4 to IPv6, and from IPv6 to
1137 * IPv4. The helper takes care of the groundwork for the
1138 * transition, including resizing the socket buffer. The eBPF
1139 * program is expected to fill the new headers, if any, via
1140 * **skb_store_bytes**\ () and to recompute the checksums with
1141 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
1142 * (). The main case for this helper is to perform NAT64
1143 * operations out of an eBPF program.
1144 *
1145 * Internally, the GSO type is marked as dodgy so that headers are
1146 * checked and segments are recalculated by the GSO/GRO engine.
1147 * The size for GSO target is adapted as well.
1148 *
1149 * All values for *flags* are reserved for future usage, and must
1150 * be left at zero.
1151 *
1152 * A call to this helper is susceptible to change the underlaying
1153 * packet buffer. Therefore, at load time, all checks on pointers
1154 * previously done by the verifier are invalidated and must be
1155 * performed again, if the helper is used in combination with
1156 * direct packet access.
1157 * Return
1158 * 0 on success, or a negative error in case of failure.
1159 *
1160 * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
1161 * Description
1162 * Change the packet type for the packet associated to *skb*. This
1163 * comes down to setting *skb*\ **->pkt_type** to *type*, except
1164 * the eBPF program does not have a write access to *skb*\
1165 * **->pkt_type** beside this helper. Using a helper here allows
1166 * for graceful handling of errors.
1167 *
1168 * The major use case is to change incoming *skb*s to
1169 * **PACKET_HOST** in a programmatic way instead of having to
1170 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
1171 * example.
1172 *
1173 * Note that *type* only allows certain values. At this time, they
1174 * are:
1175 *
1176 * **PACKET_HOST**
1177 * Packet is for us.
1178 * **PACKET_BROADCAST**
1179 * Send packet to all.
1180 * **PACKET_MULTICAST**
1181 * Send packet to group.
1182 * **PACKET_OTHERHOST**
1183 * Send packet to someone else.
1184 * Return
1185 * 0 on success, or a negative error in case of failure.
1186 *
1187 * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
1188 * Description
1189 * Check whether *skb* is a descendant of the cgroup2 held by
1190 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1191 * Return
1192 * The return value depends on the result of the test, and can be:
1193 *
1194 * * 0, if the *skb* failed the cgroup2 descendant test.
1195 * * 1, if the *skb* succeeded the cgroup2 descendant test.
1196 * * A negative error code, if an error occurred.
1197 *
1198 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
1199 * Description
1200 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
1201 * not set, in particular if the hash was cleared due to mangling,
1202 * recompute this hash. Later accesses to the hash can be done
1203 * directly with *skb*\ **->hash**.
1204 *
1205 * Calling **bpf_set_hash_invalid**\ (), changing a packet
1206 * prototype with **bpf_skb_change_proto**\ (), or calling
1207 * **bpf_skb_store_bytes**\ () with the
1208 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
1209 * the hash and to trigger a new computation for the next call to
1210 * **bpf_get_hash_recalc**\ ().
1211 * Return
1212 * The 32-bit hash.
1213 *
1214 * u64 bpf_get_current_task(void)
1215 * Return
1216 * A pointer to the current task struct.
1217 *
1218 * int bpf_probe_write_user(void *dst, const void *src, u32 len)
1219 * Description
1220 * Attempt in a safe way to write *len* bytes from the buffer
1221 * *src* to *dst* in memory. It only works for threads that are in
1222 * user context, and *dst* must be a valid user space address.
1223 *
1224 * This helper should not be used to implement any kind of
1225 * security mechanism because of TOC-TOU attacks, but rather to
1226 * debug, divert, and manipulate execution of semi-cooperative
1227 * processes.
1228 *
1229 * Keep in mind that this feature is meant for experiments, and it
1230 * has a risk of crashing the system and running programs.
1231 * Therefore, when an eBPF program using this helper is attached,
1232 * a warning including PID and process name is printed to kernel
1233 * logs.
1234 * Return
1235 * 0 on success, or a negative error in case of failure.
1236 *
1237 * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
1238 * Description
1239 * Check whether the probe is being run is the context of a given
1240 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
1241 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
1242 * Return
1243 * The return value depends on the result of the test, and can be:
1244 *
1245 * * 0, if the *skb* task belongs to the cgroup2.
1246 * * 1, if the *skb* task does not belong to the cgroup2.
1247 * * A negative error code, if an error occurred.
1248 *
1249 * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
1250 * Description
1251 * Resize (trim or grow) the packet associated to *skb* to the
1252 * new *len*. The *flags* are reserved for future usage, and must
1253 * be left at zero.
1254 *
1255 * The basic idea is that the helper performs the needed work to
1256 * change the size of the packet, then the eBPF program rewrites
1257 * the rest via helpers like **bpf_skb_store_bytes**\ (),
1258 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
1259 * and others. This helper is a slow path utility intended for
1260 * replies with control messages. And because it is targeted for
1261 * slow path, the helper itself can afford to be slow: it
1262 * implicitly linearizes, unclones and drops offloads from the
1263 * *skb*.
1264 *
1265 * A call to this helper is susceptible to change the underlaying
1266 * packet buffer. Therefore, at load time, all checks on pointers
1267 * previously done by the verifier are invalidated and must be
1268 * performed again, if the helper is used in combination with
1269 * direct packet access.
1270 * Return
1271 * 0 on success, or a negative error in case of failure.
1272 *
1273 * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
1274 * Description
1275 * Pull in non-linear data in case the *skb* is non-linear and not
1276 * all of *len* are part of the linear section. Make *len* bytes
1277 * from *skb* readable and writable. If a zero value is passed for
1278 * *len*, then the whole length of the *skb* is pulled.
1279 *
1280 * This helper is only needed for reading and writing with direct
1281 * packet access.
1282 *
1283 * For direct packet access, testing that offsets to access
1284 * are within packet boundaries (test on *skb*\ **->data_end**) is
1285 * susceptible to fail if offsets are invalid, or if the requested
1286 * data is in non-linear parts of the *skb*. On failure the
1287 * program can just bail out, or in the case of a non-linear
1288 * buffer, use a helper to make the data available. The
1289 * **bpf_skb_load_bytes**\ () helper is a first solution to access
1290 * the data. Another one consists in using **bpf_skb_pull_data**
1291 * to pull in once the non-linear parts, then retesting and
1292 * eventually access the data.
1293 *
1294 * At the same time, this also makes sure the *skb* is uncloned,
1295 * which is a necessary condition for direct write. As this needs
1296 * to be an invariant for the write part only, the verifier
1297 * detects writes and adds a prologue that is calling
1298 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
1299 * the very beginning in case it is indeed cloned.
1300 *
1301 * A call to this helper is susceptible to change the underlaying
1302 * packet buffer. Therefore, at load time, all checks on pointers
1303 * previously done by the verifier are invalidated and must be
1304 * performed again, if the helper is used in combination with
1305 * direct packet access.
1306 * Return
1307 * 0 on success, or a negative error in case of failure.
1308 *
1309 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
1310 * Description
1311 * Add the checksum *csum* into *skb*\ **->csum** in case the
1312 * driver has supplied a checksum for the entire packet into that
1313 * field. Return an error otherwise. This helper is intended to be
1314 * used in combination with **bpf_csum_diff**\ (), in particular
1315 * when the checksum needs to be updated after data has been
1316 * written into the packet through direct packet access.
1317 * Return
1318 * The checksum on success, or a negative error code in case of
1319 * failure.
1320 *
1321 * void bpf_set_hash_invalid(struct sk_buff *skb)
1322 * Description
1323 * Invalidate the current *skb*\ **->hash**. It can be used after
1324 * mangling on headers through direct packet access, in order to
1325 * indicate that the hash is outdated and to trigger a
1326 * recalculation the next time the kernel tries to access this
1327 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
1328 *
1329 * int bpf_get_numa_node_id(void)
1330 * Description
1331 * Return the id of the current NUMA node. The primary use case
1332 * for this helper is the selection of sockets for the local NUMA
1333 * node, when the program is attached to sockets using the
1334 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
1335 * but the helper is also available to other eBPF program types,
1336 * similarly to **bpf_get_smp_processor_id**\ ().
1337 * Return
1338 * The id of current NUMA node.
1339 *
1340 * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
1341 * Description
1342 * Grows headroom of packet associated to *skb* and adjusts the
1343 * offset of the MAC header accordingly, adding *len* bytes of
1344 * space. It automatically extends and reallocates memory as
1345 * required.
1346 *
1347 * This helper can be used on a layer 3 *skb* to push a MAC header
1348 * for redirection into a layer 2 device.
1349 *
1350 * All values for *flags* are reserved for future usage, and must
1351 * be left at zero.
1352 *
1353 * A call to this helper is susceptible to change the underlaying
1354 * packet buffer. Therefore, at load time, all checks on pointers
1355 * previously done by the verifier are invalidated and must be
1356 * performed again, if the helper is used in combination with
1357 * direct packet access.
1358 * Return
1359 * 0 on success, or a negative error in case of failure.
1360 *
1361 * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
1362 * Description
1363 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
1364 * it is possible to use a negative value for *delta*. This helper
1365 * can be used to prepare the packet for pushing or popping
1366 * headers.
1367 *
1368 * A call to this helper is susceptible to change the underlaying
1369 * packet buffer. Therefore, at load time, all checks on pointers
1370 * previously done by the verifier are invalidated and must be
1371 * performed again, if the helper is used in combination with
1372 * direct packet access.
1373 * Return
1374 * 0 on success, or a negative error in case of failure.
1375 *
1376 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
1377 * Description
1378 * Copy a NUL terminated string from an unsafe address
1379 * *unsafe_ptr* to *dst*. The *size* should include the
1380 * terminating NUL byte. In case the string length is smaller than
1381 * *size*, the target is not padded with further NUL bytes. If the
1382 * string length is larger than *size*, just *size*-1 bytes are
1383 * copied and the last byte is set to NUL.
1384 *
1385 * On success, the length of the copied string is returned. This
1386 * makes this helper useful in tracing programs for reading
1387 * strings, and more importantly to get its length at runtime. See
1388 * the following snippet:
1389 *
1390 * ::
1391 *
1392 * SEC("kprobe/sys_open")
1393 * void bpf_sys_open(struct pt_regs *ctx)
1394 * {
1395 * char buf[PATHLEN]; // PATHLEN is defined to 256
1396 * int res = bpf_probe_read_str(buf, sizeof(buf),
1397 * ctx->di);
1398 *
1399 * // Consume buf, for example push it to
1400 * // userspace via bpf_perf_event_output(); we
1401 * // can use res (the string length) as event
1402 * // size, after checking its boundaries.
1403 * }
1404 *
1405 * In comparison, using **bpf_probe_read()** helper here instead
1406 * to read the string would require to estimate the length at
1407 * compile time, and would often result in copying more memory
1408 * than necessary.
1409 *
1410 * Another useful use case is when parsing individual process
1411 * arguments or individual environment variables navigating
1412 * *current*\ **->mm->arg_start** and *current*\
1413 * **->mm->env_start**: using this helper and the return value,
1414 * one can quickly iterate at the right offset of the memory area.
1415 * Return
1416 * On success, the strictly positive length of the string,
1417 * including the trailing NUL character. On error, a negative
1418 * value.
1419 *
1420 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
1421 * Description
1422 * If the **struct sk_buff** pointed by *skb* has a known socket,
1423 * retrieve the cookie (generated by the kernel) of this socket.
1424 * If no cookie has been set yet, generate a new cookie. Once
1425 * generated, the socket cookie remains stable for the life of the
1426 * socket. This helper can be useful for monitoring per socket
1427 * networking traffic statistics as it provides a unique socket
1428 * identifier per namespace.
1429 * Return
1430 * A 8-byte long non-decreasing number on success, or 0 if the
1431 * socket field is missing inside *skb*.
1432 *
1433 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
1434 * Description
1435 * Equivalent to bpf_get_socket_cookie() helper that accepts
1436 * *skb*, but gets socket from **struct bpf_sock_addr** contex.
1437 * Return
1438 * A 8-byte long non-decreasing number.
1439 *
1440 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
1441 * Description
1442 * Equivalent to bpf_get_socket_cookie() helper that accepts
1443 * *skb*, but gets socket from **struct bpf_sock_ops** contex.
1444 * Return
1445 * A 8-byte long non-decreasing number.
1446 *
1447 * u32 bpf_get_socket_uid(struct sk_buff *skb)
1448 * Return
1449 * The owner UID of the socket associated to *skb*. If the socket
1450 * is **NULL**, or if it is not a full socket (i.e. if it is a
1451 * time-wait or a request socket instead), **overflowuid** value
1452 * is returned (note that **overflowuid** might also be the actual
1453 * UID value for the socket).
1454 *
1455 * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
1456 * Description
1457 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
1458 * to value *hash*.
1459 * Return
1460 * 0
1461 *
1462 * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1463 * Description
1464 * Emulate a call to **setsockopt()** on the socket associated to
1465 * *bpf_socket*, which must be a full socket. The *level* at
1466 * which the option resides and the name *optname* of the option
1467 * must be specified, see **setsockopt(2)** for more information.
1468 * The option value of length *optlen* is pointed by *optval*.
1469 *
1470 * This helper actually implements a subset of **setsockopt()**.
1471 * It supports the following *level*\ s:
1472 *
1473 * * **SOL_SOCKET**, which supports the following *optname*\ s:
1474 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
1475 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
1476 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
1477 * **TCP_CONGESTION**, **TCP_BPF_IW**,
1478 * **TCP_BPF_SNDCWND_CLAMP**.
1479 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1480 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1481 * Return
1482 * 0 on success, or a negative error in case of failure.
1483 *
1484 * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
1485 * Description
1486 * Grow or shrink the room for data in the packet associated to
1487 * *skb* by *len_diff*, and according to the selected *mode*.
1488 *
1489 * There is a single supported mode at this time:
1490 *
1491 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
1492 * (room space is added or removed below the layer 3 header).
1493 *
1494 * All values for *flags* are reserved for future usage, and must
1495 * be left at zero.
1496 *
1497 * A call to this helper is susceptible to change the underlaying
1498 * packet buffer. Therefore, at load time, all checks on pointers
1499 * previously done by the verifier are invalidated and must be
1500 * performed again, if the helper is used in combination with
1501 * direct packet access.
1502 * Return
1503 * 0 on success, or a negative error in case of failure.
1504 *
1505 * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1506 * Description
1507 * Redirect the packet to the endpoint referenced by *map* at
1508 * index *key*. Depending on its type, this *map* can contain
1509 * references to net devices (for forwarding packets through other
1510 * ports), or to CPUs (for redirecting XDP frames to another CPU;
1511 * but this is only implemented for native XDP (with driver
1512 * support) as of this writing).
1513 *
1514 * All values for *flags* are reserved for future usage, and must
1515 * be left at zero.
1516 *
1517 * When used to redirect packets to net devices, this helper
1518 * provides a high performance increase over **bpf_redirect**\ ().
1519 * This is due to various implementation details of the underlying
1520 * mechanisms, one of which is the fact that **bpf_redirect_map**\
1521 * () tries to send packet as a "bulk" to the device.
1522 * Return
1523 * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
1524 *
1525 * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
1526 * Description
1527 * Redirect the packet to the socket referenced by *map* (of type
1528 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1529 * egress interfaces can be used for redirection. The
1530 * **BPF_F_INGRESS** value in *flags* is used to make the
1531 * distinction (ingress path is selected if the flag is present,
1532 * egress path otherwise). This is the only flag supported for now.
1533 * Return
1534 * **SK_PASS** on success, or **SK_DROP** on error.
1535 *
1536 * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
1537 * Description
1538 * Add an entry to, or update a *map* referencing sockets. The
1539 * *skops* is used as a new value for the entry associated to
1540 * *key*. *flags* is one of:
1541 *
1542 * **BPF_NOEXIST**
1543 * The entry for *key* must not exist in the map.
1544 * **BPF_EXIST**
1545 * The entry for *key* must already exist in the map.
1546 * **BPF_ANY**
1547 * No condition on the existence of the entry for *key*.
1548 *
1549 * If the *map* has eBPF programs (parser and verdict), those will
1550 * be inherited by the socket being added. If the socket is
1551 * already attached to eBPF programs, this results in an error.
1552 * Return
1553 * 0 on success, or a negative error in case of failure.
1554 *
1555 * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
1556 * Description
1557 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
1558 * *delta* (which can be positive or negative). Note that this
1559 * operation modifies the address stored in *xdp_md*\ **->data**,
1560 * so the latter must be loaded only after the helper has been
1561 * called.
1562 *
1563 * The use of *xdp_md*\ **->data_meta** is optional and programs
1564 * are not required to use it. The rationale is that when the
1565 * packet is processed with XDP (e.g. as DoS filter), it is
1566 * possible to push further meta data along with it before passing
1567 * to the stack, and to give the guarantee that an ingress eBPF
1568 * program attached as a TC classifier on the same device can pick
1569 * this up for further post-processing. Since TC works with socket
1570 * buffers, it remains possible to set from XDP the **mark** or
1571 * **priority** pointers, or other pointers for the socket buffer.
1572 * Having this scratch space generic and programmable allows for
1573 * more flexibility as the user is free to store whatever meta
1574 * data they need.
1575 *
1576 * A call to this helper is susceptible to change the underlaying
1577 * packet buffer. Therefore, at load time, all checks on pointers
1578 * previously done by the verifier are invalidated and must be
1579 * performed again, if the helper is used in combination with
1580 * direct packet access.
1581 * Return
1582 * 0 on success, or a negative error in case of failure.
1583 *
1584 * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
1585 * Description
1586 * Read the value of a perf event counter, and store it into *buf*
1587 * of size *buf_size*. This helper relies on a *map* of type
1588 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
1589 * counter is selected when *map* is updated with perf event file
1590 * descriptors. The *map* is an array whose size is the number of
1591 * available CPUs, and each cell contains a value relative to one
1592 * CPU. The value to retrieve is indicated by *flags*, that
1593 * contains the index of the CPU to look up, masked with
1594 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1595 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1596 * current CPU should be retrieved.
1597 *
1598 * This helper behaves in a way close to
1599 * **bpf_perf_event_read**\ () helper, save that instead of
1600 * just returning the value observed, it fills the *buf*
1601 * structure. This allows for additional data to be retrieved: in
1602 * particular, the enabled and running times (in *buf*\
1603 * **->enabled** and *buf*\ **->running**, respectively) are
1604 * copied. In general, **bpf_perf_event_read_value**\ () is
1605 * recommended over **bpf_perf_event_read**\ (), which has some
1606 * ABI issues and provides fewer functionalities.
1607 *
1608 * These values are interesting, because hardware PMU (Performance
1609 * Monitoring Unit) counters are limited resources. When there are
1610 * more PMU based perf events opened than available counters,
1611 * kernel will multiplex these events so each event gets certain
1612 * percentage (but not all) of the PMU time. In case that
1613 * multiplexing happens, the number of samples or counter value
1614 * will not reflect the case compared to when no multiplexing
1615 * occurs. This makes comparison between different runs difficult.
1616 * Typically, the counter value should be normalized before
1617 * comparing to other experiments. The usual normalization is done
1618 * as follows.
1619 *
1620 * ::
1621 *
1622 * normalized_counter = counter * t_enabled / t_running
1623 *
1624 * Where t_enabled is the time enabled for event and t_running is
1625 * the time running for event since last normalization. The
1626 * enabled and running times are accumulated since the perf event
1627 * open. To achieve scaling factor between two invocations of an
1628 * eBPF program, users can can use CPU id as the key (which is
1629 * typical for perf array usage model) to remember the previous
1630 * value and do the calculation inside the eBPF program.
1631 * Return
1632 * 0 on success, or a negative error in case of failure.
1633 *
1634 * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
1635 * Description
1636 * For en eBPF program attached to a perf event, retrieve the
1637 * value of the event counter associated to *ctx* and store it in
1638 * the structure pointed by *buf* and of size *buf_size*. Enabled
1639 * and running times are also stored in the structure (see
1640 * description of helper **bpf_perf_event_read_value**\ () for
1641 * more details).
1642 * Return
1643 * 0 on success, or a negative error in case of failure.
1644 *
1645 * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
1646 * Description
1647 * Emulate a call to **getsockopt()** on the socket associated to
1648 * *bpf_socket*, which must be a full socket. The *level* at
1649 * which the option resides and the name *optname* of the option
1650 * must be specified, see **getsockopt(2)** for more information.
1651 * The retrieved value is stored in the structure pointed by
1652 * *opval* and of length *optlen*.
1653 *
1654 * This helper actually implements a subset of **getsockopt()**.
1655 * It supports the following *level*\ s:
1656 *
1657 * * **IPPROTO_TCP**, which supports *optname*
1658 * **TCP_CONGESTION**.
1659 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
1660 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
1661 * Return
1662 * 0 on success, or a negative error in case of failure.
1663 *
1664 * int bpf_override_return(struct pt_reg *regs, u64 rc)
1665 * Description
1666 * Used for error injection, this helper uses kprobes to override
1667 * the return value of the probed function, and to set it to *rc*.
1668 * The first argument is the context *regs* on which the kprobe
1669 * works.
1670 *
1671 * This helper works by setting setting the PC (program counter)
1672 * to an override function which is run in place of the original
1673 * probed function. This means the probed function is not run at
1674 * all. The replacement function just returns with the required
1675 * value.
1676 *
1677 * This helper has security implications, and thus is subject to
1678 * restrictions. It is only available if the kernel was compiled
1679 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
1680 * option, and in this case it only works on functions tagged with
1681 * **ALLOW_ERROR_INJECTION** in the kernel code.
1682 *
1683 * Also, the helper is only available for the architectures having
1684 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
1685 * x86 architecture is the only one to support this feature.
1686 * Return
1687 * 0
1688 *
1689 * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
1690 * Description
1691 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
1692 * for the full TCP socket associated to *bpf_sock_ops* to
1693 * *argval*.
1694 *
1695 * The primary use of this field is to determine if there should
1696 * be calls to eBPF programs of type
1697 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
1698 * code. A program of the same type can change its value, per
1699 * connection and as necessary, when the connection is
1700 * established. This field is directly accessible for reading, but
1701 * this helper must be used for updates in order to return an
1702 * error if an eBPF program tries to set a callback that is not
1703 * supported in the current kernel.
1704 *
1705 * The supported callback values that *argval* can combine are:
1706 *
1707 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
1708 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
1709 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
1710 *
1711 * Here are some examples of where one could call such eBPF
1712 * program:
1713 *
1714 * * When RTO fires.
1715 * * When a packet is retransmitted.
1716 * * When the connection terminates.
1717 * * When a packet is sent.
1718 * * When a packet is received.
1719 * Return
1720 * Code **-EINVAL** if the socket is not a full TCP socket;
1721 * otherwise, a positive number containing the bits that could not
1722 * be set is returned (which comes down to 0 if all bits were set
1723 * as required).
1724 *
1725 * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
1726 * Description
1727 * This helper is used in programs implementing policies at the
1728 * socket level. If the message *msg* is allowed to pass (i.e. if
1729 * the verdict eBPF program returns **SK_PASS**), redirect it to
1730 * the socket referenced by *map* (of type
1731 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
1732 * egress interfaces can be used for redirection. The
1733 * **BPF_F_INGRESS** value in *flags* is used to make the
1734 * distinction (ingress path is selected if the flag is present,
1735 * egress path otherwise). This is the only flag supported for now.
1736 * Return
1737 * **SK_PASS** on success, or **SK_DROP** on error.
1738 *
1739 * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
1740 * Description
1741 * For socket policies, apply the verdict of the eBPF program to
1742 * the next *bytes* (number of bytes) of message *msg*.
1743 *
1744 * For example, this helper can be used in the following cases:
1745 *
1746 * * A single **sendmsg**\ () or **sendfile**\ () system call
1747 * contains multiple logical messages that the eBPF program is
1748 * supposed to read and for which it should apply a verdict.
1749 * * An eBPF program only cares to read the first *bytes* of a
1750 * *msg*. If the message has a large payload, then setting up
1751 * and calling the eBPF program repeatedly for all bytes, even
1752 * though the verdict is already known, would create unnecessary
1753 * overhead.
1754 *
1755 * When called from within an eBPF program, the helper sets a
1756 * counter internal to the BPF infrastructure, that is used to
1757 * apply the last verdict to the next *bytes*. If *bytes* is
1758 * smaller than the current data being processed from a
1759 * **sendmsg**\ () or **sendfile**\ () system call, the first
1760 * *bytes* will be sent and the eBPF program will be re-run with
1761 * the pointer for start of data pointing to byte number *bytes*
1762 * **+ 1**. If *bytes* is larger than the current data being
1763 * processed, then the eBPF verdict will be applied to multiple
1764 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
1765 * consumed.
1766 *
1767 * Note that if a socket closes with the internal counter holding
1768 * a non-zero value, this is not a problem because data is not
1769 * being buffered for *bytes* and is sent as it is received.
1770 * Return
1771 * 0
1772 *
1773 * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
1774 * Description
1775 * For socket policies, prevent the execution of the verdict eBPF
1776 * program for message *msg* until *bytes* (byte number) have been
1777 * accumulated.
1778 *
1779 * This can be used when one needs a specific number of bytes
1780 * before a verdict can be assigned, even if the data spans
1781 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
1782 * case would be a user calling **sendmsg**\ () repeatedly with
1783 * 1-byte long message segments. Obviously, this is bad for
1784 * performance, but it is still valid. If the eBPF program needs
1785 * *bytes* bytes to validate a header, this helper can be used to
1786 * prevent the eBPF program to be called again until *bytes* have
1787 * been accumulated.
1788 * Return
1789 * 0
1790 *
1791 * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
1792 * Description
1793 * For socket policies, pull in non-linear data from user space
1794 * for *msg* and set pointers *msg*\ **->data** and *msg*\
1795 * **->data_end** to *start* and *end* bytes offsets into *msg*,
1796 * respectively.
1797 *
1798 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
1799 * *msg* it can only parse data that the (**data**, **data_end**)
1800 * pointers have already consumed. For **sendmsg**\ () hooks this
1801 * is likely the first scatterlist element. But for calls relying
1802 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
1803 * be the range (**0**, **0**) because the data is shared with
1804 * user space and by default the objective is to avoid allowing
1805 * user space to modify data while (or after) eBPF verdict is
1806 * being decided. This helper can be used to pull in data and to
1807 * set the start and end pointer to given values. Data will be
1808 * copied if necessary (i.e. if data was not linear and if start
1809 * and end pointers do not point to the same chunk).
1810 *
1811 * A call to this helper is susceptible to change the underlaying
1812 * packet buffer. Therefore, at load time, all checks on pointers
1813 * previously done by the verifier are invalidated and must be
1814 * performed again, if the helper is used in combination with
1815 * direct packet access.
1816 *
1817 * All values for *flags* are reserved for future usage, and must
1818 * be left at zero.
1819 * Return
1820 * 0 on success, or a negative error in case of failure.
1821 *
1822 * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
1823 * Description
1824 * Bind the socket associated to *ctx* to the address pointed by
1825 * *addr*, of length *addr_len*. This allows for making outgoing
1826 * connection from the desired IP address, which can be useful for
1827 * example when all processes inside a cgroup should use one
1828 * single IP address on a host that has multiple IP configured.
1829 *
1830 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
1831 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
1832 * **AF_INET6**). Looking for a free port to bind to can be
1833 * expensive, therefore binding to port is not permitted by the
1834 * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
1835 * must be set to zero.
1836 * Return
1837 * 0 on success, or a negative error in case of failure.
1838 *
1839 * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
1840 * Description
1841 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
1842 * only possible to shrink the packet as of this writing,
1843 * therefore *delta* must be a negative integer.
1844 *
1845 * A call to this helper is susceptible to change the underlaying
1846 * packet buffer. Therefore, at load time, all checks on pointers
1847 * previously done by the verifier are invalidated and must be
1848 * performed again, if the helper is used in combination with
1849 * direct packet access.
1850 * Return
1851 * 0 on success, or a negative error in case of failure.
1852 *
1853 * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
1854 * Description
1855 * Retrieve the XFRM state (IP transform framework, see also
1856 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
1857 *
1858 * The retrieved value is stored in the **struct bpf_xfrm_state**
1859 * pointed by *xfrm_state* and of length *size*.
1860 *
1861 * All values for *flags* are reserved for future usage, and must
1862 * be left at zero.
1863 *
1864 * This helper is available only if the kernel was compiled with
1865 * **CONFIG_XFRM** configuration option.
1866 * Return
1867 * 0 on success, or a negative error in case of failure.
1868 *
1869 * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
1870 * Description
1871 * Return a user or a kernel stack in bpf program provided buffer.
1872 * To achieve this, the helper needs *ctx*, which is a pointer
1873 * to the context on which the tracing program is executed.
1874 * To store the stacktrace, the bpf program provides *buf* with
1875 * a nonnegative *size*.
1876 *
1877 * The last argument, *flags*, holds the number of stack frames to
1878 * skip (from 0 to 255), masked with
1879 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
1880 * the following flags:
1881 *
1882 * **BPF_F_USER_STACK**
1883 * Collect a user space stack instead of a kernel stack.
1884 * **BPF_F_USER_BUILD_ID**
1885 * Collect buildid+offset instead of ips for user stack,
1886 * only valid if **BPF_F_USER_STACK** is also specified.
1887 *
1888 * **bpf_get_stack**\ () can collect up to
1889 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
1890 * to sufficient large buffer size. Note that
1891 * this limit can be controlled with the **sysctl** program, and
1892 * that it should be manually increased in order to profile long
1893 * user stacks (such as stacks for Java programs). To do so, use:
1894 *
1895 * ::
1896 *
1897 * # sysctl kernel.perf_event_max_stack=<new value>
1898 * Return
1899 * A non-negative value equal to or less than *size* on success,
1900 * or a negative error in case of failure.
1901 *
1902 * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
1903 * Description
1904 * This helper is similar to **bpf_skb_load_bytes**\ () in that
1905 * it provides an easy way to load *len* bytes from *offset*
1906 * from the packet associated to *skb*, into the buffer pointed
1907 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
1908 * a fifth argument *start_header* exists in order to select a
1909 * base offset to start from. *start_header* can be one of:
1910 *
1911 * **BPF_HDR_START_MAC**
1912 * Base offset to load data from is *skb*'s mac header.
1913 * **BPF_HDR_START_NET**
1914 * Base offset to load data from is *skb*'s network header.
1915 *
1916 * In general, "direct packet access" is the preferred method to
1917 * access packet data, however, this helper is in particular useful
1918 * in socket filters where *skb*\ **->data** does not always point
1919 * to the start of the mac header and where "direct packet access"
1920 * is not available.
1921 * Return
1922 * 0 on success, or a negative error in case of failure.
1923 *
1924 * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
1925 * Description
1926 * Do FIB lookup in kernel tables using parameters in *params*.
1927 * If lookup is successful and result shows packet is to be
1928 * forwarded, the neighbor tables are searched for the nexthop.
1929 * If successful (ie., FIB lookup shows forwarding and nexthop
1930 * is resolved), the nexthop address is returned in ipv4_dst
1931 * or ipv6_dst based on family, smac is set to mac address of
1932 * egress device, dmac is set to nexthop mac address, rt_metric
1933 * is set to metric from route (IPv4/IPv6 only), and ifindex
1934 * is set to the device index of the nexthop from the FIB lookup.
1935 *
1936 * *plen* argument is the size of the passed in struct.
1937 * *flags* argument can be a combination of one or more of the
1938 * following values:
1939 *
1940 * **BPF_FIB_LOOKUP_DIRECT**
1941 * Do a direct table lookup vs full lookup using FIB
1942 * rules.
1943 * **BPF_FIB_LOOKUP_OUTPUT**
1944 * Perform lookup from an egress perspective (default is
1945 * ingress).
1946 *
1947 * *ctx* is either **struct xdp_md** for XDP programs or
1948 * **struct sk_buff** tc cls_act programs.
1949 * Return
1950 * * < 0 if any input argument is invalid
1951 * * 0 on success (packet is forwarded, nexthop neighbor exists)
1952 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
1953 * packet is not forwarded or needs assist from full stack
1954 *
1955 * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
1956 * Description
1957 * Add an entry to, or update a sockhash *map* referencing sockets.
1958 * The *skops* is used as a new value for the entry associated to
1959 * *key*. *flags* is one of:
1960 *
1961 * **BPF_NOEXIST**
1962 * The entry for *key* must not exist in the map.
1963 * **BPF_EXIST**
1964 * The entry for *key* must already exist in the map.
1965 * **BPF_ANY**
1966 * No condition on the existence of the entry for *key*.
1967 *
1968 * If the *map* has eBPF programs (parser and verdict), those will
1969 * be inherited by the socket being added. If the socket is
1970 * already attached to eBPF programs, this results in an error.
1971 * Return
1972 * 0 on success, or a negative error in case of failure.
1973 *
1974 * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
1975 * Description
1976 * This helper is used in programs implementing policies at the
1977 * socket level. If the message *msg* is allowed to pass (i.e. if
1978 * the verdict eBPF program returns **SK_PASS**), redirect it to
1979 * the socket referenced by *map* (of type
1980 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1981 * egress interfaces can be used for redirection. The
1982 * **BPF_F_INGRESS** value in *flags* is used to make the
1983 * distinction (ingress path is selected if the flag is present,
1984 * egress path otherwise). This is the only flag supported for now.
1985 * Return
1986 * **SK_PASS** on success, or **SK_DROP** on error.
1987 *
1988 * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
1989 * Description
1990 * This helper is used in programs implementing policies at the
1991 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
1992 * if the verdeict eBPF program returns **SK_PASS**), redirect it
1993 * to the socket referenced by *map* (of type
1994 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
1995 * egress interfaces can be used for redirection. The
1996 * **BPF_F_INGRESS** value in *flags* is used to make the
1997 * distinction (ingress path is selected if the flag is present,
1998 * egress otherwise). This is the only flag supported for now.
1999 * Return
2000 * **SK_PASS** on success, or **SK_DROP** on error.
2001 *
2002 * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
2003 * Description
2004 * Encapsulate the packet associated to *skb* within a Layer 3
2005 * protocol header. This header is provided in the buffer at
2006 * address *hdr*, with *len* its size in bytes. *type* indicates
2007 * the protocol of the header and can be one of:
2008 *
2009 * **BPF_LWT_ENCAP_SEG6**
2010 * IPv6 encapsulation with Segment Routing Header
2011 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
2012 * the IPv6 header is computed by the kernel.
2013 * **BPF_LWT_ENCAP_SEG6_INLINE**
2014 * Only works if *skb* contains an IPv6 packet. Insert a
2015 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
2016 * the IPv6 header.
2017 *
2018 * A call to this helper is susceptible to change the underlaying
2019 * packet buffer. Therefore, at load time, all checks on pointers
2020 * previously done by the verifier are invalidated and must be
2021 * performed again, if the helper is used in combination with
2022 * direct packet access.
2023 * Return
2024 * 0 on success, or a negative error in case of failure.
2025 *
2026 * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
2027 * Description
2028 * Store *len* bytes from address *from* into the packet
2029 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
2030 * inside the outermost IPv6 Segment Routing Header can be
2031 * modified through this helper.
2032 *
2033 * A call to this helper is susceptible to change the underlaying
2034 * packet buffer. Therefore, at load time, all checks on pointers
2035 * previously done by the verifier are invalidated and must be
2036 * performed again, if the helper is used in combination with
2037 * direct packet access.
2038 * Return
2039 * 0 on success, or a negative error in case of failure.
2040 *
2041 * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
2042 * Description
2043 * Adjust the size allocated to TLVs in the outermost IPv6
2044 * Segment Routing Header contained in the packet associated to
2045 * *skb*, at position *offset* by *delta* bytes. Only offsets
2046 * after the segments are accepted. *delta* can be as well
2047 * positive (growing) as negative (shrinking).
2048 *
2049 * A call to this helper is susceptible to change the underlaying
2050 * packet buffer. Therefore, at load time, all checks on pointers
2051 * previously done by the verifier are invalidated and must be
2052 * performed again, if the helper is used in combination with
2053 * direct packet access.
2054 * Return
2055 * 0 on success, or a negative error in case of failure.
2056 *
2057 * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
2058 * Description
2059 * Apply an IPv6 Segment Routing action of type *action* to the
2060 * packet associated to *skb*. Each action takes a parameter
2061 * contained at address *param*, and of length *param_len* bytes.
2062 * *action* can be one of:
2063 *
2064 * **SEG6_LOCAL_ACTION_END_X**
2065 * End.X action: Endpoint with Layer-3 cross-connect.
2066 * Type of *param*: **struct in6_addr**.
2067 * **SEG6_LOCAL_ACTION_END_T**
2068 * End.T action: Endpoint with specific IPv6 table lookup.
2069 * Type of *param*: **int**.
2070 * **SEG6_LOCAL_ACTION_END_B6**
2071 * End.B6 action: Endpoint bound to an SRv6 policy.
2072 * Type of param: **struct ipv6_sr_hdr**.
2073 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
2074 * End.B6.Encap action: Endpoint bound to an SRv6
2075 * encapsulation policy.
2076 * Type of param: **struct ipv6_sr_hdr**.
2077 *
2078 * A call to this helper is susceptible to change the underlaying
2079 * packet buffer. Therefore, at load time, all checks on pointers
2080 * previously done by the verifier are invalidated and must be
2081 * performed again, if the helper is used in combination with
2082 * direct packet access.
2083 * Return
2084 * 0 on success, or a negative error in case of failure.
2085 *
2086 * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
2087 * Description
2088 * This helper is used in programs implementing IR decoding, to
2089 * report a successfully decoded key press with *scancode*,
2090 * *toggle* value in the given *protocol*. The scancode will be
2091 * translated to a keycode using the rc keymap, and reported as
2092 * an input key down event. After a period a key up event is
2093 * generated. This period can be extended by calling either
2094 * **bpf_rc_keydown**\ () again with the same values, or calling
2095 * **bpf_rc_repeat**\ ().
2096 *
2097 * Some protocols include a toggle bit, in case the button was
2098 * released and pressed again between consecutive scancodes.
2099 *
2100 * The *ctx* should point to the lirc sample as passed into
2101 * the program.
2102 *
2103 * The *protocol* is the decoded protocol number (see
2104 * **enum rc_proto** for some predefined values).
2105 *
2106 * This helper is only available is the kernel was compiled with
2107 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2108 * "**y**".
2109 * Return
2110 * 0
2111 *
2112 * int bpf_rc_repeat(void *ctx)
2113 * Description
2114 * This helper is used in programs implementing IR decoding, to
2115 * report a successfully decoded repeat key message. This delays
2116 * the generation of a key up event for previously generated
2117 * key down event.
2118 *
2119 * Some IR protocols like NEC have a special IR message for
2120 * repeating last button, for when a button is held down.
2121 *
2122 * The *ctx* should point to the lirc sample as passed into
2123 * the program.
2124 *
2125 * This helper is only available is the kernel was compiled with
2126 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2127 * "**y**".
2128 * Return
2129 * 0
2130 *
2131 * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
2132 * Description
2133 * Return the cgroup v2 id of the socket associated with the *skb*.
2134 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
2135 * helper for cgroup v1 by providing a tag resp. identifier that
2136 * can be matched on or used for map lookups e.g. to implement
2137 * policy. The cgroup v2 id of a given path in the hierarchy is
2138 * exposed in user space through the f_handle API in order to get
2139 * to the same 64-bit id.
2140 *
2141 * This helper can be used on TC egress path, but not on ingress,
2142 * and is available only if the kernel was compiled with the
2143 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
2144 * Return
2145 * The id is returned or 0 in case the id could not be retrieved.
2146 *
2147 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
2148 * Description
2149 * Return id of cgroup v2 that is ancestor of cgroup associated
2150 * with the *skb* at the *ancestor_level*. The root cgroup is at
2151 * *ancestor_level* zero and each step down the hierarchy
2152 * increments the level. If *ancestor_level* == level of cgroup
2153 * associated with *skb*, then return value will be same as that
2154 * of **bpf_skb_cgroup_id**\ ().
2155 *
2156 * The helper is useful to implement policies based on cgroups
2157 * that are upper in hierarchy than immediate cgroup associated
2158 * with *skb*.
2159 *
2160 * The format of returned id and helper limitations are same as in
2161 * **bpf_skb_cgroup_id**\ ().
2162 * Return
2163 * The id is returned or 0 in case the id could not be retrieved.
2164 *
2165 * u64 bpf_get_current_cgroup_id(void)
2166 * Return
2167 * A 64-bit integer containing the current cgroup id based
2168 * on the cgroup within which the current task is running.
2169 *
2170 * void* get_local_storage(void *map, u64 flags)
2171 * Description
2172 * Get the pointer to the local storage area.
2173 * The type and the size of the local storage is defined
2174 * by the *map* argument.
2175 * The *flags* meaning is specific for each map type,
2176 * and has to be 0 for cgroup local storage.
2177 *
2178 * Depending on the BPF program type, a local storage area
2179 * can be shared between multiple instances of the BPF program,
2180 * running simultaneously.
2181 *
2182 * A user should care about the synchronization by himself.
2183 * For example, by using the **BPF_STX_XADD** instruction to alter
2184 * the shared data.
2185 * Return
2186 * A pointer to the local storage area.
2187 *
2188 * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
2189 * Description
2190 * Select a **SO_REUSEPORT** socket from a
2191 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
2192 * It checks the selected socket is matching the incoming
2193 * request in the socket buffer.
2194 * Return
2195 * 0 on success, or a negative error in case of failure.
2196 *
2197 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2198 * Description
2199 * Look for TCP socket matching *tuple*, optionally in a child
2200 * network namespace *netns*. The return value must be checked,
2201 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2202 *
2203 * The *ctx* should point to the context of the program, such as
2204 * the skb or socket (depending on the hook in use). This is used
2205 * to determine the base network namespace for the lookup.
2206 *
2207 * *tuple_size* must be one of:
2208 *
2209 * **sizeof**\ (*tuple*\ **->ipv4**)
2210 * Look for an IPv4 socket.
2211 * **sizeof**\ (*tuple*\ **->ipv6**)
2212 * Look for an IPv6 socket.
2213 *
2214 * If the *netns* is a negative signed 32-bit integer, then the
2215 * socket lookup table in the netns associated with the *ctx* will
2216 * will be used. For the TC hooks, this is the netns of the device
2217 * in the skb. For socket hooks, this is the netns of the socket.
2218 * If *netns* is any other signed 32-bit value greater than or
2219 * equal to zero then it specifies the ID of the netns relative to
2220 * the netns associated with the *ctx*. *netns* values beyond the
2221 * range of 32-bit integers are reserved for future use.
2222 *
2223 * All values for *flags* are reserved for future usage, and must
2224 * be left at zero.
2225 *
2226 * This helper is available only if the kernel was compiled with
2227 * **CONFIG_NET** configuration option.
2228 * Return
2229 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2230 * For sockets with reuseport option, the **struct bpf_sock**
2231 * result is from **reuse->socks**\ [] using the hash of the tuple.
2232 *
2233 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
2234 * Description
2235 * Look for UDP socket matching *tuple*, optionally in a child
2236 * network namespace *netns*. The return value must be checked,
2237 * and if non-**NULL**, released via **bpf_sk_release**\ ().
2238 *
2239 * The *ctx* should point to the context of the program, such as
2240 * the skb or socket (depending on the hook in use). This is used
2241 * to determine the base network namespace for the lookup.
2242 *
2243 * *tuple_size* must be one of:
2244 *
2245 * **sizeof**\ (*tuple*\ **->ipv4**)
2246 * Look for an IPv4 socket.
2247 * **sizeof**\ (*tuple*\ **->ipv6**)
2248 * Look for an IPv6 socket.
2249 *
2250 * If the *netns* is a negative signed 32-bit integer, then the
2251 * socket lookup table in the netns associated with the *ctx* will
2252 * will be used. For the TC hooks, this is the netns of the device
2253 * in the skb. For socket hooks, this is the netns of the socket.
2254 * If *netns* is any other signed 32-bit value greater than or
2255 * equal to zero then it specifies the ID of the netns relative to
2256 * the netns associated with the *ctx*. *netns* values beyond the
2257 * range of 32-bit integers are reserved for future use.
2258 *
2259 * All values for *flags* are reserved for future usage, and must
2260 * be left at zero.
2261 *
2262 * This helper is available only if the kernel was compiled with
2263 * **CONFIG_NET** configuration option.
2264 * Return
2265 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
2266 * For sockets with reuseport option, the **struct bpf_sock**
2267 * result is from **reuse->socks**\ [] using the hash of the tuple.
2268 *
2269 * int bpf_sk_release(struct bpf_sock *sock)
2270 * Description
2271 * Release the reference held by *sock*. *sock* must be a
2272 * non-**NULL** pointer that was returned from
2273 * **bpf_sk_lookup_xxx**\ ().
2274 * Return
2275 * 0 on success, or a negative error in case of failure.
2276 *
2277 * int bpf_map_pop_elem(struct bpf_map *map, void *value)
2278 * Description
2279 * Pop an element from *map*.
2280 * Return
2281 * 0 on success, or a negative error in case of failure.
2282 *
2283 * int bpf_map_peek_elem(struct bpf_map *map, void *value)
2284 * Description
2285 * Get an element from *map* without removing it.
2286 * Return
2287 * 0 on success, or a negative error in case of failure.
2288 *
2289 * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
2290 * Description
2291 * For socket policies, insert *len* bytes into *msg* at offset
2292 * *start*.
2293 *
2294 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2295 * *msg* it may want to insert metadata or options into the *msg*.
2296 * This can later be read and used by any of the lower layer BPF
2297 * hooks.
2298 *
2299 * This helper may fail if under memory pressure (a malloc
2300 * fails) in these cases BPF programs will get an appropriate
2301 * error and BPF programs will need to handle them.
2302 * Return
2303 * 0 on success, or a negative error in case of failure.
2304 *
2305 * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags)
2306 * Description
2307 * Will remove *pop* bytes from a *msg* starting at byte *start*.
2308 * This may result in **ENOMEM** errors under certain situations if
2309 * an allocation and copy are required due to a full ring buffer.
2310 * However, the helper will try to avoid doing the allocation
2311 * if possible. Other errors can occur if input parameters are
2312 * invalid either due to *start* byte not being valid part of *msg*
2313 * payload and/or *pop* value being to large.
2314 * Return
2315 * 0 on success, or a negative error in case of failure.
2316 *
2317 * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
2318 * Description
2319 * This helper is used in programs implementing IR decoding, to
2320 * report a successfully decoded pointer movement.
2321 *
2322 * The *ctx* should point to the lirc sample as passed into
2323 * the program.
2324 *
2325 * This helper is only available is the kernel was compiled with
2326 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
2327 * "**y**".
2328 * Return
2329 * 0
2330 */
2331 #define __BPF_FUNC_MAPPER(FN) \
2332 FN(unspec), \
2333 FN(map_lookup_elem), \
2334 FN(map_update_elem), \
2335 FN(map_delete_elem), \
2336 FN(probe_read), \
2337 FN(ktime_get_ns), \
2338 FN(trace_printk), \
2339 FN(get_prandom_u32), \
2340 FN(get_smp_processor_id), \
2341 FN(skb_store_bytes), \
2342 FN(l3_csum_replace), \
2343 FN(l4_csum_replace), \
2344 FN(tail_call), \
2345 FN(clone_redirect), \
2346 FN(get_current_pid_tgid), \
2347 FN(get_current_uid_gid), \
2348 FN(get_current_comm), \
2349 FN(get_cgroup_classid), \
2350 FN(skb_vlan_push), \
2351 FN(skb_vlan_pop), \
2352 FN(skb_get_tunnel_key), \
2353 FN(skb_set_tunnel_key), \
2354 FN(perf_event_read), \
2355 FN(redirect), \
2356 FN(get_route_realm), \
2357 FN(perf_event_output), \
2358 FN(skb_load_bytes), \
2359 FN(get_stackid), \
2360 FN(csum_diff), \
2361 FN(skb_get_tunnel_opt), \
2362 FN(skb_set_tunnel_opt), \
2363 FN(skb_change_proto), \
2364 FN(skb_change_type), \
2365 FN(skb_under_cgroup), \
2366 FN(get_hash_recalc), \
2367 FN(get_current_task), \
2368 FN(probe_write_user), \
2369 FN(current_task_under_cgroup), \
2370 FN(skb_change_tail), \
2371 FN(skb_pull_data), \
2372 FN(csum_update), \
2373 FN(set_hash_invalid), \
2374 FN(get_numa_node_id), \
2375 FN(skb_change_head), \
2376 FN(xdp_adjust_head), \
2377 FN(probe_read_str), \
2378 FN(get_socket_cookie), \
2379 FN(get_socket_uid), \
2380 FN(set_hash), \
2381 FN(setsockopt), \
2382 FN(skb_adjust_room), \
2383 FN(redirect_map), \
2384 FN(sk_redirect_map), \
2385 FN(sock_map_update), \
2386 FN(xdp_adjust_meta), \
2387 FN(perf_event_read_value), \
2388 FN(perf_prog_read_value), \
2389 FN(getsockopt), \
2390 FN(override_return), \
2391 FN(sock_ops_cb_flags_set), \
2392 FN(msg_redirect_map), \
2393 FN(msg_apply_bytes), \
2394 FN(msg_cork_bytes), \
2395 FN(msg_pull_data), \
2396 FN(bind), \
2397 FN(xdp_adjust_tail), \
2398 FN(skb_get_xfrm_state), \
2399 FN(get_stack), \
2400 FN(skb_load_bytes_relative), \
2401 FN(fib_lookup), \
2402 FN(sock_hash_update), \
2403 FN(msg_redirect_hash), \
2404 FN(sk_redirect_hash), \
2405 FN(lwt_push_encap), \
2406 FN(lwt_seg6_store_bytes), \
2407 FN(lwt_seg6_adjust_srh), \
2408 FN(lwt_seg6_action), \
2409 FN(rc_repeat), \
2410 FN(rc_keydown), \
2411 FN(skb_cgroup_id), \
2412 FN(get_current_cgroup_id), \
2413 FN(get_local_storage), \
2414 FN(sk_select_reuseport), \
2415 FN(skb_ancestor_cgroup_id), \
2416 FN(sk_lookup_tcp), \
2417 FN(sk_lookup_udp), \
2418 FN(sk_release), \
2419 FN(map_push_elem), \
2420 FN(map_pop_elem), \
2421 FN(map_peek_elem), \
2422 FN(msg_push_data), \
2423 FN(msg_pop_data), \
2424 FN(rc_pointer_rel),
2425
2426 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
2427 * function eBPF program intends to call
2428 */
2429 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
2430 enum bpf_func_id {
2431 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
2432 __BPF_FUNC_MAX_ID,
2433 };
2434 #undef __BPF_ENUM_FN
2435
2436 /* All flags used by eBPF helper functions, placed here. */
2437
2438 /* BPF_FUNC_skb_store_bytes flags. */
2439 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
2440 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
2441
2442 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
2443 * First 4 bits are for passing the header field size.
2444 */
2445 #define BPF_F_HDR_FIELD_MASK 0xfULL
2446
2447 /* BPF_FUNC_l4_csum_replace flags. */
2448 #define BPF_F_PSEUDO_HDR (1ULL << 4)
2449 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
2450 #define BPF_F_MARK_ENFORCE (1ULL << 6)
2451
2452 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
2453 #define BPF_F_INGRESS (1ULL << 0)
2454
2455 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
2456 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
2457
2458 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
2459 #define BPF_F_SKIP_FIELD_MASK 0xffULL
2460 #define BPF_F_USER_STACK (1ULL << 8)
2461 /* flags used by BPF_FUNC_get_stackid only. */
2462 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
2463 #define BPF_F_REUSE_STACKID (1ULL << 10)
2464 /* flags used by BPF_FUNC_get_stack only. */
2465 #define BPF_F_USER_BUILD_ID (1ULL << 11)
2466
2467 /* BPF_FUNC_skb_set_tunnel_key flags. */
2468 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
2469 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
2470 #define BPF_F_SEQ_NUMBER (1ULL << 3)
2471
2472 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
2473 * BPF_FUNC_perf_event_read_value flags.
2474 */
2475 #define BPF_F_INDEX_MASK 0xffffffffULL
2476 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
2477 /* BPF_FUNC_perf_event_output for sk_buff input context. */
2478 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
2479
2480 /* Current network namespace */
2481 #define BPF_F_CURRENT_NETNS (-1L)
2482
2483 /* Mode for BPF_FUNC_skb_adjust_room helper. */
2484 enum bpf_adj_room_mode {
2485 BPF_ADJ_ROOM_NET,
2486 };
2487
2488 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
2489 enum bpf_hdr_start_off {
2490 BPF_HDR_START_MAC,
2491 BPF_HDR_START_NET,
2492 };
2493
2494 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
2495 enum bpf_lwt_encap_mode {
2496 BPF_LWT_ENCAP_SEG6,
2497 BPF_LWT_ENCAP_SEG6_INLINE
2498 };
2499
2500 #define __bpf_md_ptr(type, name) \
2501 union { \
2502 type name; \
2503 __u64 :64; \
2504 } __attribute__((aligned(8)))
2505
2506 /* user accessible mirror of in-kernel sk_buff.
2507 * new fields can only be added to the end of this structure
2508 */
2509 struct __sk_buff {
2510 __u32 len;
2511 __u32 pkt_type;
2512 __u32 mark;
2513 __u32 queue_mapping;
2514 __u32 protocol;
2515 __u32 vlan_present;
2516 __u32 vlan_tci;
2517 __u32 vlan_proto;
2518 __u32 priority;
2519 __u32 ingress_ifindex;
2520 __u32 ifindex;
2521 __u32 tc_index;
2522 __u32 cb[5];
2523 __u32 hash;
2524 __u32 tc_classid;
2525 __u32 data;
2526 __u32 data_end;
2527 __u32 napi_id;
2528
2529 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
2530 __u32 family;
2531 __u32 remote_ip4; /* Stored in network byte order */
2532 __u32 local_ip4; /* Stored in network byte order */
2533 __u32 remote_ip6[4]; /* Stored in network byte order */
2534 __u32 local_ip6[4]; /* Stored in network byte order */
2535 __u32 remote_port; /* Stored in network byte order */
2536 __u32 local_port; /* stored in host byte order */
2537 /* ... here. */
2538
2539 __u32 data_meta;
2540 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
2541 __u64 tstamp;
2542 __u32 wire_len;
2543 };
2544
2545 struct bpf_tunnel_key {
2546 __u32 tunnel_id;
2547 union {
2548 __u32 remote_ipv4;
2549 __u32 remote_ipv6[4];
2550 };
2551 __u8 tunnel_tos;
2552 __u8 tunnel_ttl;
2553 __u16 tunnel_ext; /* Padding, future use. */
2554 __u32 tunnel_label;
2555 };
2556
2557 /* user accessible mirror of in-kernel xfrm_state.
2558 * new fields can only be added to the end of this structure
2559 */
2560 struct bpf_xfrm_state {
2561 __u32 reqid;
2562 __u32 spi; /* Stored in network byte order */
2563 __u16 family;
2564 __u16 ext; /* Padding, future use. */
2565 union {
2566 __u32 remote_ipv4; /* Stored in network byte order */
2567 __u32 remote_ipv6[4]; /* Stored in network byte order */
2568 };
2569 };
2570
2571 /* Generic BPF return codes which all BPF program types may support.
2572 * The values are binary compatible with their TC_ACT_* counter-part to
2573 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
2574 * programs.
2575 *
2576 * XDP is handled seprately, see XDP_*.
2577 */
2578 enum bpf_ret_code {
2579 BPF_OK = 0,
2580 /* 1 reserved */
2581 BPF_DROP = 2,
2582 /* 3-6 reserved */
2583 BPF_REDIRECT = 7,
2584 /* >127 are reserved for prog type specific return codes */
2585 };
2586
2587 struct bpf_sock {
2588 __u32 bound_dev_if;
2589 __u32 family;
2590 __u32 type;
2591 __u32 protocol;
2592 __u32 mark;
2593 __u32 priority;
2594 __u32 src_ip4; /* Allows 1,2,4-byte read.
2595 * Stored in network byte order.
2596 */
2597 __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
2598 * Stored in network byte order.
2599 */
2600 __u32 src_port; /* Allows 4-byte read.
2601 * Stored in host byte order
2602 */
2603 };
2604
2605 struct bpf_sock_tuple {
2606 union {
2607 struct {
2608 __be32 saddr;
2609 __be32 daddr;
2610 __be16 sport;
2611 __be16 dport;
2612 } ipv4;
2613 struct {
2614 __be32 saddr[4];
2615 __be32 daddr[4];
2616 __be16 sport;
2617 __be16 dport;
2618 } ipv6;
2619 };
2620 };
2621
2622 #define XDP_PACKET_HEADROOM 256
2623
2624 /* User return codes for XDP prog type.
2625 * A valid XDP program must return one of these defined values. All other
2626 * return codes are reserved for future use. Unknown return codes will
2627 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
2628 */
2629 enum xdp_action {
2630 XDP_ABORTED = 0,
2631 XDP_DROP,
2632 XDP_PASS,
2633 XDP_TX,
2634 XDP_REDIRECT,
2635 };
2636
2637 /* user accessible metadata for XDP packet hook
2638 * new fields must be added to the end of this structure
2639 */
2640 struct xdp_md {
2641 __u32 data;
2642 __u32 data_end;
2643 __u32 data_meta;
2644 /* Below access go through struct xdp_rxq_info */
2645 __u32 ingress_ifindex; /* rxq->dev->ifindex */
2646 __u32 rx_queue_index; /* rxq->queue_index */
2647 };
2648
2649 enum sk_action {
2650 SK_DROP = 0,
2651 SK_PASS,
2652 };
2653
2654 /* user accessible metadata for SK_MSG packet hook, new fields must
2655 * be added to the end of this structure
2656 */
2657 struct sk_msg_md {
2658 __bpf_md_ptr(void *, data);
2659 __bpf_md_ptr(void *, data_end);
2660
2661 __u32 family;
2662 __u32 remote_ip4; /* Stored in network byte order */
2663 __u32 local_ip4; /* Stored in network byte order */
2664 __u32 remote_ip6[4]; /* Stored in network byte order */
2665 __u32 local_ip6[4]; /* Stored in network byte order */
2666 __u32 remote_port; /* Stored in network byte order */
2667 __u32 local_port; /* stored in host byte order */
2668 __u32 size; /* Total size of sk_msg */
2669 };
2670
2671 struct sk_reuseport_md {
2672 /*
2673 * Start of directly accessible data. It begins from
2674 * the tcp/udp header.
2675 */
2676 __bpf_md_ptr(void *, data);
2677 /* End of directly accessible data */
2678 __bpf_md_ptr(void *, data_end);
2679 /*
2680 * Total length of packet (starting from the tcp/udp header).
2681 * Note that the directly accessible bytes (data_end - data)
2682 * could be less than this "len". Those bytes could be
2683 * indirectly read by a helper "bpf_skb_load_bytes()".
2684 */
2685 __u32 len;
2686 /*
2687 * Eth protocol in the mac header (network byte order). e.g.
2688 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
2689 */
2690 __u32 eth_protocol;
2691 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
2692 __u32 bind_inany; /* Is sock bound to an INANY address? */
2693 __u32 hash; /* A hash of the packet 4 tuples */
2694 };
2695
2696 #define BPF_TAG_SIZE 8
2697
2698 struct bpf_prog_info {
2699 __u32 type;
2700 __u32 id;
2701 __u8 tag[BPF_TAG_SIZE];
2702 __u32 jited_prog_len;
2703 __u32 xlated_prog_len;
2704 __aligned_u64 jited_prog_insns;
2705 __aligned_u64 xlated_prog_insns;
2706 __u64 load_time; /* ns since boottime */
2707 __u32 created_by_uid;
2708 __u32 nr_map_ids;
2709 __aligned_u64 map_ids;
2710 char name[BPF_OBJ_NAME_LEN];
2711 __u32 ifindex;
2712 __u32 gpl_compatible:1;
2713 __u64 netns_dev;
2714 __u64 netns_ino;
2715 __u32 nr_jited_ksyms;
2716 __u32 nr_jited_func_lens;
2717 __aligned_u64 jited_ksyms;
2718 __aligned_u64 jited_func_lens;
2719 __u32 btf_id;
2720 __u32 func_info_rec_size;
2721 __aligned_u64 func_info;
2722 __u32 nr_func_info;
2723 __u32 nr_line_info;
2724 __aligned_u64 line_info;
2725 __aligned_u64 jited_line_info;
2726 __u32 nr_jited_line_info;
2727 __u32 line_info_rec_size;
2728 __u32 jited_line_info_rec_size;
2729 __u32 nr_prog_tags;
2730 __aligned_u64 prog_tags;
2731 } __attribute__((aligned(8)));
2732
2733 struct bpf_map_info {
2734 __u32 type;
2735 __u32 id;
2736 __u32 key_size;
2737 __u32 value_size;
2738 __u32 max_entries;
2739 __u32 map_flags;
2740 char name[BPF_OBJ_NAME_LEN];
2741 __u32 ifindex;
2742 __u32 :32;
2743 __u64 netns_dev;
2744 __u64 netns_ino;
2745 __u32 btf_id;
2746 __u32 btf_key_type_id;
2747 __u32 btf_value_type_id;
2748 } __attribute__((aligned(8)));
2749
2750 struct bpf_btf_info {
2751 __aligned_u64 btf;
2752 __u32 btf_size;
2753 __u32 id;
2754 } __attribute__((aligned(8)));
2755
2756 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
2757 * by user and intended to be used by socket (e.g. to bind to, depends on
2758 * attach attach type).
2759 */
2760 struct bpf_sock_addr {
2761 __u32 user_family; /* Allows 4-byte read, but no write. */
2762 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
2763 * Stored in network byte order.
2764 */
2765 __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2766 * Stored in network byte order.
2767 */
2768 __u32 user_port; /* Allows 4-byte read and write.
2769 * Stored in network byte order
2770 */
2771 __u32 family; /* Allows 4-byte read, but no write */
2772 __u32 type; /* Allows 4-byte read, but no write */
2773 __u32 protocol; /* Allows 4-byte read, but no write */
2774 __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
2775 * Stored in network byte order.
2776 */
2777 __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
2778 * Stored in network byte order.
2779 */
2780 };
2781
2782 /* User bpf_sock_ops struct to access socket values and specify request ops
2783 * and their replies.
2784 * Some of this fields are in network (bigendian) byte order and may need
2785 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
2786 * New fields can only be added at the end of this structure
2787 */
2788 struct bpf_sock_ops {
2789 __u32 op;
2790 union {
2791 __u32 args[4]; /* Optionally passed to bpf program */
2792 __u32 reply; /* Returned by bpf program */
2793 __u32 replylong[4]; /* Optionally returned by bpf prog */
2794 };
2795 __u32 family;
2796 __u32 remote_ip4; /* Stored in network byte order */
2797 __u32 local_ip4; /* Stored in network byte order */
2798 __u32 remote_ip6[4]; /* Stored in network byte order */
2799 __u32 local_ip6[4]; /* Stored in network byte order */
2800 __u32 remote_port; /* Stored in network byte order */
2801 __u32 local_port; /* stored in host byte order */
2802 __u32 is_fullsock; /* Some TCP fields are only valid if
2803 * there is a full socket. If not, the
2804 * fields read as zero.
2805 */
2806 __u32 snd_cwnd;
2807 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
2808 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
2809 __u32 state;
2810 __u32 rtt_min;
2811 __u32 snd_ssthresh;
2812 __u32 rcv_nxt;
2813 __u32 snd_nxt;
2814 __u32 snd_una;
2815 __u32 mss_cache;
2816 __u32 ecn_flags;
2817 __u32 rate_delivered;
2818 __u32 rate_interval_us;
2819 __u32 packets_out;
2820 __u32 retrans_out;
2821 __u32 total_retrans;
2822 __u32 segs_in;
2823 __u32 data_segs_in;
2824 __u32 segs_out;
2825 __u32 data_segs_out;
2826 __u32 lost_out;
2827 __u32 sacked_out;
2828 __u32 sk_txhash;
2829 __u64 bytes_received;
2830 __u64 bytes_acked;
2831 };
2832
2833 /* Definitions for bpf_sock_ops_cb_flags */
2834 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
2835 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
2836 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
2837 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
2838 * supported cb flags
2839 */
2840
2841 /* List of known BPF sock_ops operators.
2842 * New entries can only be added at the end
2843 */
2844 enum {
2845 BPF_SOCK_OPS_VOID,
2846 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
2847 * -1 if default value should be used
2848 */
2849 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
2850 * window (in packets) or -1 if default
2851 * value should be used
2852 */
2853 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
2854 * active connection is initialized
2855 */
2856 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
2857 * active connection is
2858 * established
2859 */
2860 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
2861 * passive connection is
2862 * established
2863 */
2864 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
2865 * needs ECN
2866 */
2867 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
2868 * based on the path and may be
2869 * dependent on the congestion control
2870 * algorithm. In general it indicates
2871 * a congestion threshold. RTTs above
2872 * this indicate congestion
2873 */
2874 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
2875 * Arg1: value of icsk_retransmits
2876 * Arg2: value of icsk_rto
2877 * Arg3: whether RTO has expired
2878 */
2879 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
2880 * Arg1: sequence number of 1st byte
2881 * Arg2: # segments
2882 * Arg3: return value of
2883 * tcp_transmit_skb (0 => success)
2884 */
2885 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
2886 * Arg1: old_state
2887 * Arg2: new_state
2888 */
2889 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
2890 * socket transition to LISTEN state.
2891 */
2892 };
2893
2894 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
2895 * changes between the TCP and BPF versions. Ideally this should never happen.
2896 * If it does, we need to add code to convert them before calling
2897 * the BPF sock_ops function.
2898 */
2899 enum {
2900 BPF_TCP_ESTABLISHED = 1,
2901 BPF_TCP_SYN_SENT,
2902 BPF_TCP_SYN_RECV,
2903 BPF_TCP_FIN_WAIT1,
2904 BPF_TCP_FIN_WAIT2,
2905 BPF_TCP_TIME_WAIT,
2906 BPF_TCP_CLOSE,
2907 BPF_TCP_CLOSE_WAIT,
2908 BPF_TCP_LAST_ACK,
2909 BPF_TCP_LISTEN,
2910 BPF_TCP_CLOSING, /* Now a valid state */
2911 BPF_TCP_NEW_SYN_RECV,
2912
2913 BPF_TCP_MAX_STATES /* Leave at the end! */
2914 };
2915
2916 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
2917 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
2918
2919 struct bpf_perf_event_value {
2920 __u64 counter;
2921 __u64 enabled;
2922 __u64 running;
2923 };
2924
2925 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
2926 #define BPF_DEVCG_ACC_READ (1ULL << 1)
2927 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
2928
2929 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
2930 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
2931
2932 struct bpf_cgroup_dev_ctx {
2933 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
2934 __u32 access_type;
2935 __u32 major;
2936 __u32 minor;
2937 };
2938
2939 struct bpf_raw_tracepoint_args {
2940 __u64 args[0];
2941 };
2942
2943 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
2944 * OUTPUT: Do lookup from egress perspective; default is ingress
2945 */
2946 #define BPF_FIB_LOOKUP_DIRECT BIT(0)
2947 #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
2948
2949 enum {
2950 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
2951 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
2952 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
2953 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
2954 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
2955 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
2956 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
2957 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
2958 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
2959 };
2960
2961 struct bpf_fib_lookup {
2962 /* input: network family for lookup (AF_INET, AF_INET6)
2963 * output: network family of egress nexthop
2964 */
2965 __u8 family;
2966
2967 /* set if lookup is to consider L4 data - e.g., FIB rules */
2968 __u8 l4_protocol;
2969 __be16 sport;
2970 __be16 dport;
2971
2972 /* total length of packet from network header - used for MTU check */
2973 __u16 tot_len;
2974
2975 /* input: L3 device index for lookup
2976 * output: device index from FIB lookup
2977 */
2978 __u32 ifindex;
2979
2980 union {
2981 /* inputs to lookup */
2982 __u8 tos; /* AF_INET */
2983 __be32 flowinfo; /* AF_INET6, flow_label + priority */
2984
2985 /* output: metric of fib result (IPv4/IPv6 only) */
2986 __u32 rt_metric;
2987 };
2988
2989 union {
2990 __be32 ipv4_src;
2991 __u32 ipv6_src[4]; /* in6_addr; network order */
2992 };
2993
2994 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
2995 * network header. output: bpf_fib_lookup sets to gateway address
2996 * if FIB lookup returns gateway route
2997 */
2998 union {
2999 __be32 ipv4_dst;
3000 __u32 ipv6_dst[4]; /* in6_addr; network order */
3001 };
3002
3003 /* output */
3004 __be16 h_vlan_proto;
3005 __be16 h_vlan_TCI;
3006 __u8 smac[6]; /* ETH_ALEN */
3007 __u8 dmac[6]; /* ETH_ALEN */
3008 };
3009
3010 enum bpf_task_fd_type {
3011 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
3012 BPF_FD_TYPE_TRACEPOINT, /* tp name */
3013 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
3014 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
3015 BPF_FD_TYPE_UPROBE, /* filename + offset */
3016 BPF_FD_TYPE_URETPROBE, /* filename + offset */
3017 };
3018
3019 struct bpf_flow_keys {
3020 __u16 nhoff;
3021 __u16 thoff;
3022 __u16 addr_proto; /* ETH_P_* of valid addrs */
3023 __u8 is_frag;
3024 __u8 is_first_frag;
3025 __u8 is_encap;
3026 __u8 ip_proto;
3027 __be16 n_proto;
3028 __be16 sport;
3029 __be16 dport;
3030 union {
3031 struct {
3032 __be32 ipv4_src;
3033 __be32 ipv4_dst;
3034 };
3035 struct {
3036 __u32 ipv6_src[4]; /* in6_addr; network order */
3037 __u32 ipv6_dst[4]; /* in6_addr; network order */
3038 };
3039 };
3040 };
3041
3042 struct bpf_func_info {
3043 __u32 insn_off;
3044 __u32 type_id;
3045 };
3046
3047 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
3048 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
3049
3050 struct bpf_line_info {
3051 __u32 insn_off;
3052 __u32 file_name_off;
3053 __u32 line_off;
3054 __u32 line_col;
3055 };
3056
3057 #endif /* _UAPI__LINUX_BPF_H__ */