Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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_JMP32 0x06 /* jmp mode in word width */
18#define BPF_ALU64 0x07 /* alu mode in double word width */
19
20/* ld/ldx fields */
21#define BPF_DW 0x18 /* double word (64-bit) */
22#define BPF_MEMSX 0x80 /* load with sign extension */
23#define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
24#define BPF_XADD 0xc0 /* exclusive add - legacy name */
25
26/* alu/jmp fields */
27#define BPF_MOV 0xb0 /* mov reg to reg */
28#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
29
30/* change endianness of a register */
31#define BPF_END 0xd0 /* flags for endianness conversion: */
32#define BPF_TO_LE 0x00 /* convert to little-endian */
33#define BPF_TO_BE 0x08 /* convert to big-endian */
34#define BPF_FROM_LE BPF_TO_LE
35#define BPF_FROM_BE BPF_TO_BE
36
37/* jmp encodings */
38#define BPF_JNE 0x50 /* jump != */
39#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
40#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
41#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
42#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
43#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
44#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
45#define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */
46#define BPF_CALL 0x80 /* function call */
47#define BPF_EXIT 0x90 /* function return */
48
49/* atomic op type fields (stored in immediate) */
50#define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
51#define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
52#define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
53
54#define BPF_LOAD_ACQ 0x100 /* load-acquire */
55#define BPF_STORE_REL 0x110 /* store-release */
56
57enum bpf_cond_pseudo_jmp {
58 BPF_MAY_GOTO = 0,
59};
60
61/* Register numbers */
62enum {
63 BPF_REG_0 = 0,
64 BPF_REG_1,
65 BPF_REG_2,
66 BPF_REG_3,
67 BPF_REG_4,
68 BPF_REG_5,
69 BPF_REG_6,
70 BPF_REG_7,
71 BPF_REG_8,
72 BPF_REG_9,
73 BPF_REG_10,
74 __MAX_BPF_REG,
75};
76
77/* BPF has 10 general purpose 64-bit registers and stack frame. */
78#define MAX_BPF_REG __MAX_BPF_REG
79
80struct bpf_insn {
81 __u8 code; /* opcode */
82 __u8 dst_reg:4; /* dest register */
83 __u8 src_reg:4; /* source register */
84 __s16 off; /* signed offset */
85 __s32 imm; /* signed immediate constant */
86};
87
88/* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
89 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
90 * the trailing flexible array member) instead.
91 */
92struct bpf_lpm_trie_key {
93 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
94 __u8 data[0]; /* Arbitrary size */
95};
96
97/* Header for bpf_lpm_trie_key structs */
98struct bpf_lpm_trie_key_hdr {
99 __u32 prefixlen;
100};
101
102/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
103struct bpf_lpm_trie_key_u8 {
104 union {
105 struct bpf_lpm_trie_key_hdr hdr;
106 __u32 prefixlen;
107 };
108 __u8 data[]; /* Arbitrary size */
109};
110
111struct bpf_cgroup_storage_key {
112 __u64 cgroup_inode_id; /* cgroup inode id */
113 __u32 attach_type; /* program attach type (enum bpf_attach_type) */
114};
115
116enum bpf_cgroup_iter_order {
117 BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
118 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */
119 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */
120 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */
121 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */
122};
123
124union bpf_iter_link_info {
125 struct {
126 __u32 map_fd;
127 } map;
128 struct {
129 enum bpf_cgroup_iter_order order;
130
131 /* At most one of cgroup_fd and cgroup_id can be non-zero. If
132 * both are zero, the walk starts from the default cgroup v2
133 * root. For walking v1 hierarchy, one should always explicitly
134 * specify cgroup_fd.
135 */
136 __u32 cgroup_fd;
137 __u64 cgroup_id;
138 } cgroup;
139 /* Parameters of task iterators. */
140 struct {
141 __u32 tid;
142 __u32 pid;
143 __u32 pid_fd;
144 } task;
145};
146
147/* BPF syscall commands, see bpf(2) man-page for more details. */
148/**
149 * DOC: eBPF Syscall Preamble
150 *
151 * The operation to be performed by the **bpf**\ () system call is determined
152 * by the *cmd* argument. Each operation takes an accompanying argument,
153 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
154 * below). The size argument is the size of the union pointed to by *attr*.
155 */
156/**
157 * DOC: eBPF Syscall Commands
158 *
159 * BPF_MAP_CREATE
160 * Description
161 * Create a map and return a file descriptor that refers to the
162 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
163 * is automatically enabled for the new file descriptor.
164 *
165 * Applying **close**\ (2) to the file descriptor returned by
166 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
167 *
168 * Return
169 * A new file descriptor (a nonnegative integer), or -1 if an
170 * error occurred (in which case, *errno* is set appropriately).
171 *
172 * BPF_MAP_LOOKUP_ELEM
173 * Description
174 * Look up an element with a given *key* in the map referred to
175 * by the file descriptor *map_fd*.
176 *
177 * The *flags* argument may be specified as one of the
178 * following:
179 *
180 * **BPF_F_LOCK**
181 * Look up the value of a spin-locked map without
182 * returning the lock. This must be specified if the
183 * elements contain a spinlock.
184 *
185 * Return
186 * Returns zero on success. On error, -1 is returned and *errno*
187 * is set appropriately.
188 *
189 * BPF_MAP_UPDATE_ELEM
190 * Description
191 * Create or update an element (key/value pair) in a specified map.
192 *
193 * The *flags* argument should be specified as one of the
194 * following:
195 *
196 * **BPF_ANY**
197 * Create a new element or update an existing element.
198 * **BPF_NOEXIST**
199 * Create a new element only if it did not exist.
200 * **BPF_EXIST**
201 * Update an existing element.
202 * **BPF_F_LOCK**
203 * Update a spin_lock-ed map element.
204 *
205 * Return
206 * Returns zero on success. On error, -1 is returned and *errno*
207 * is set appropriately.
208 *
209 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
210 * **E2BIG**, **EEXIST**, or **ENOENT**.
211 *
212 * **E2BIG**
213 * The number of elements in the map reached the
214 * *max_entries* limit specified at map creation time.
215 * **EEXIST**
216 * If *flags* specifies **BPF_NOEXIST** and the element
217 * with *key* already exists in the map.
218 * **ENOENT**
219 * If *flags* specifies **BPF_EXIST** and the element with
220 * *key* does not exist in the map.
221 *
222 * BPF_MAP_DELETE_ELEM
223 * Description
224 * Look up and delete an element by key in a specified map.
225 *
226 * Return
227 * Returns zero on success. On error, -1 is returned and *errno*
228 * is set appropriately.
229 *
230 * BPF_MAP_GET_NEXT_KEY
231 * Description
232 * Look up an element by key in a specified map and return the key
233 * of the next element. Can be used to iterate over all elements
234 * in the map.
235 *
236 * Return
237 * Returns zero on success. On error, -1 is returned and *errno*
238 * is set appropriately.
239 *
240 * The following cases can be used to iterate over all elements of
241 * the map:
242 *
243 * * If *key* is not found, the operation returns zero and sets
244 * the *next_key* pointer to the key of the first element.
245 * * If *key* is found, the operation returns zero and sets the
246 * *next_key* pointer to the key of the next element.
247 * * If *key* is the last element, returns -1 and *errno* is set
248 * to **ENOENT**.
249 *
250 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
251 * **EINVAL** on error.
252 *
253 * BPF_PROG_LOAD
254 * Description
255 * Verify and load an eBPF program, returning a new file
256 * descriptor associated with the program.
257 *
258 * Applying **close**\ (2) to the file descriptor returned by
259 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
260 *
261 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
262 * automatically enabled for the new file descriptor.
263 *
264 * Return
265 * A new file descriptor (a nonnegative integer), or -1 if an
266 * error occurred (in which case, *errno* is set appropriately).
267 *
268 * BPF_OBJ_PIN
269 * Description
270 * Pin an eBPF program or map referred by the specified *bpf_fd*
271 * to the provided *pathname* on the filesystem.
272 *
273 * The *pathname* argument must not contain a dot (".").
274 *
275 * On success, *pathname* retains a reference to the eBPF object,
276 * preventing deallocation of the object when the original
277 * *bpf_fd* is closed. This allow the eBPF object to live beyond
278 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
279 * process.
280 *
281 * Applying **unlink**\ (2) or similar calls to the *pathname*
282 * unpins the object from the filesystem, removing the reference.
283 * If no other file descriptors or filesystem nodes refer to the
284 * same object, it will be deallocated (see NOTES).
285 *
286 * The filesystem type for the parent directory of *pathname* must
287 * be **BPF_FS_MAGIC**.
288 *
289 * Return
290 * Returns zero on success. On error, -1 is returned and *errno*
291 * is set appropriately.
292 *
293 * BPF_OBJ_GET
294 * Description
295 * Open a file descriptor for the eBPF object pinned to the
296 * specified *pathname*.
297 *
298 * Return
299 * A new file descriptor (a nonnegative integer), or -1 if an
300 * error occurred (in which case, *errno* is set appropriately).
301 *
302 * BPF_PROG_ATTACH
303 * Description
304 * Attach an eBPF program to a *target_fd* at the specified
305 * *attach_type* hook.
306 *
307 * The *attach_type* specifies the eBPF attachment point to
308 * attach the program to, and must be one of *bpf_attach_type*
309 * (see below).
310 *
311 * The *attach_bpf_fd* must be a valid file descriptor for a
312 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
313 * or sock_ops type corresponding to the specified *attach_type*.
314 *
315 * The *target_fd* must be a valid file descriptor for a kernel
316 * object which depends on the attach type of *attach_bpf_fd*:
317 *
318 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
319 * **BPF_PROG_TYPE_CGROUP_SKB**,
320 * **BPF_PROG_TYPE_CGROUP_SOCK**,
321 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
322 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
323 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
324 * **BPF_PROG_TYPE_SOCK_OPS**
325 *
326 * Control Group v2 hierarchy with the eBPF controller
327 * enabled. Requires the kernel to be compiled with
328 * **CONFIG_CGROUP_BPF**.
329 *
330 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
331 *
332 * Network namespace (eg /proc/self/ns/net).
333 *
334 * **BPF_PROG_TYPE_LIRC_MODE2**
335 *
336 * LIRC device path (eg /dev/lircN). Requires the kernel
337 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
338 *
339 * **BPF_PROG_TYPE_SK_SKB**,
340 * **BPF_PROG_TYPE_SK_MSG**
341 *
342 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
343 *
344 * Return
345 * Returns zero on success. On error, -1 is returned and *errno*
346 * is set appropriately.
347 *
348 * BPF_PROG_DETACH
349 * Description
350 * Detach the eBPF program associated with the *target_fd* at the
351 * hook specified by *attach_type*. The program must have been
352 * previously attached using **BPF_PROG_ATTACH**.
353 *
354 * Return
355 * Returns zero on success. On error, -1 is returned and *errno*
356 * is set appropriately.
357 *
358 * BPF_PROG_TEST_RUN
359 * Description
360 * Run the eBPF program associated with the *prog_fd* a *repeat*
361 * number of times against a provided program context *ctx_in* and
362 * data *data_in*, and return the modified program context
363 * *ctx_out*, *data_out* (for example, packet data), result of the
364 * execution *retval*, and *duration* of the test run.
365 *
366 * The sizes of the buffers provided as input and output
367 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
368 * be provided in the corresponding variables *ctx_size_in*,
369 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
370 * of these parameters are not provided (ie set to NULL), the
371 * corresponding size field must be zero.
372 *
373 * Some program types have particular requirements:
374 *
375 * **BPF_PROG_TYPE_SK_LOOKUP**
376 * *data_in* and *data_out* must be NULL.
377 *
378 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
379 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
380 *
381 * *ctx_out*, *data_in* and *data_out* must be NULL.
382 * *repeat* must be zero.
383 *
384 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
385 *
386 * Return
387 * Returns zero on success. On error, -1 is returned and *errno*
388 * is set appropriately.
389 *
390 * **ENOSPC**
391 * Either *data_size_out* or *ctx_size_out* is too small.
392 * **ENOTSUPP**
393 * This command is not supported by the program type of
394 * the program referred to by *prog_fd*.
395 *
396 * BPF_PROG_GET_NEXT_ID
397 * Description
398 * Fetch the next eBPF program currently loaded into the kernel.
399 *
400 * Looks for the eBPF program with an id greater than *start_id*
401 * and updates *next_id* on success. If no other eBPF programs
402 * remain with ids higher than *start_id*, returns -1 and sets
403 * *errno* to **ENOENT**.
404 *
405 * Return
406 * Returns zero on success. On error, or when no id remains, -1
407 * is returned and *errno* is set appropriately.
408 *
409 * BPF_MAP_GET_NEXT_ID
410 * Description
411 * Fetch the next eBPF map currently loaded into the kernel.
412 *
413 * Looks for the eBPF map with an id greater than *start_id*
414 * and updates *next_id* on success. If no other eBPF maps
415 * remain with ids higher than *start_id*, returns -1 and sets
416 * *errno* to **ENOENT**.
417 *
418 * Return
419 * Returns zero on success. On error, or when no id remains, -1
420 * is returned and *errno* is set appropriately.
421 *
422 * BPF_PROG_GET_FD_BY_ID
423 * Description
424 * Open a file descriptor for the eBPF program corresponding to
425 * *prog_id*.
426 *
427 * Return
428 * A new file descriptor (a nonnegative integer), or -1 if an
429 * error occurred (in which case, *errno* is set appropriately).
430 *
431 * BPF_MAP_GET_FD_BY_ID
432 * Description
433 * Open a file descriptor for the eBPF map corresponding to
434 * *map_id*.
435 *
436 * Return
437 * A new file descriptor (a nonnegative integer), or -1 if an
438 * error occurred (in which case, *errno* is set appropriately).
439 *
440 * BPF_OBJ_GET_INFO_BY_FD
441 * Description
442 * Obtain information about the eBPF object corresponding to
443 * *bpf_fd*.
444 *
445 * Populates up to *info_len* bytes of *info*, which will be in
446 * one of the following formats depending on the eBPF object type
447 * of *bpf_fd*:
448 *
449 * * **struct bpf_prog_info**
450 * * **struct bpf_map_info**
451 * * **struct bpf_btf_info**
452 * * **struct bpf_link_info**
453 * * **struct bpf_token_info**
454 *
455 * Return
456 * Returns zero on success. On error, -1 is returned and *errno*
457 * is set appropriately.
458 *
459 * BPF_PROG_QUERY
460 * Description
461 * Obtain information about eBPF programs associated with the
462 * specified *attach_type* hook.
463 *
464 * The *target_fd* must be a valid file descriptor for a kernel
465 * object which depends on the attach type of *attach_bpf_fd*:
466 *
467 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
468 * **BPF_PROG_TYPE_CGROUP_SKB**,
469 * **BPF_PROG_TYPE_CGROUP_SOCK**,
470 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
471 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
472 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
473 * **BPF_PROG_TYPE_SOCK_OPS**
474 *
475 * Control Group v2 hierarchy with the eBPF controller
476 * enabled. Requires the kernel to be compiled with
477 * **CONFIG_CGROUP_BPF**.
478 *
479 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
480 *
481 * Network namespace (eg /proc/self/ns/net).
482 *
483 * **BPF_PROG_TYPE_LIRC_MODE2**
484 *
485 * LIRC device path (eg /dev/lircN). Requires the kernel
486 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
487 *
488 * **BPF_PROG_QUERY** always fetches the number of programs
489 * attached and the *attach_flags* which were used to attach those
490 * programs. Additionally, if *prog_ids* is nonzero and the number
491 * of attached programs is less than *prog_cnt*, populates
492 * *prog_ids* with the eBPF program ids of the programs attached
493 * at *target_fd*.
494 *
495 * The following flags may alter the result:
496 *
497 * **BPF_F_QUERY_EFFECTIVE**
498 * Only return information regarding programs which are
499 * currently effective at the specified *target_fd*.
500 *
501 * Return
502 * Returns zero on success. On error, -1 is returned and *errno*
503 * is set appropriately.
504 *
505 * BPF_RAW_TRACEPOINT_OPEN
506 * Description
507 * Attach an eBPF program to a tracepoint *name* to access kernel
508 * internal arguments of the tracepoint in their raw form.
509 *
510 * The *prog_fd* must be a valid file descriptor associated with
511 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
512 *
513 * No ABI guarantees are made about the content of tracepoint
514 * arguments exposed to the corresponding eBPF program.
515 *
516 * Applying **close**\ (2) to the file descriptor returned by
517 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
518 *
519 * Return
520 * A new file descriptor (a nonnegative integer), or -1 if an
521 * error occurred (in which case, *errno* is set appropriately).
522 *
523 * BPF_BTF_LOAD
524 * Description
525 * Verify and load BPF Type Format (BTF) metadata into the kernel,
526 * returning a new file descriptor associated with the metadata.
527 * BTF is described in more detail at
528 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
529 *
530 * The *btf* parameter must point to valid memory providing
531 * *btf_size* bytes of BTF binary metadata.
532 *
533 * The returned file descriptor can be passed to other **bpf**\ ()
534 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
535 * associate the BTF with those objects.
536 *
537 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
538 * parameters to specify a *btf_log_buf*, *btf_log_size* and
539 * *btf_log_level* which allow the kernel to return freeform log
540 * output regarding the BTF verification process.
541 *
542 * Return
543 * A new file descriptor (a nonnegative integer), or -1 if an
544 * error occurred (in which case, *errno* is set appropriately).
545 *
546 * BPF_BTF_GET_FD_BY_ID
547 * Description
548 * Open a file descriptor for the BPF Type Format (BTF)
549 * corresponding to *btf_id*.
550 *
551 * Return
552 * A new file descriptor (a nonnegative integer), or -1 if an
553 * error occurred (in which case, *errno* is set appropriately).
554 *
555 * BPF_TASK_FD_QUERY
556 * Description
557 * Obtain information about eBPF programs associated with the
558 * target process identified by *pid* and *fd*.
559 *
560 * If the *pid* and *fd* are associated with a tracepoint, kprobe
561 * or uprobe perf event, then the *prog_id* and *fd_type* will
562 * be populated with the eBPF program id and file descriptor type
563 * of type **bpf_task_fd_type**. If associated with a kprobe or
564 * uprobe, the *probe_offset* and *probe_addr* will also be
565 * populated. Optionally, if *buf* is provided, then up to
566 * *buf_len* bytes of *buf* will be populated with the name of
567 * the tracepoint, kprobe or uprobe.
568 *
569 * The resulting *prog_id* may be introspected in deeper detail
570 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
571 *
572 * Return
573 * Returns zero on success. On error, -1 is returned and *errno*
574 * is set appropriately.
575 *
576 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
577 * Description
578 * Look up an element with the given *key* in the map referred to
579 * by the file descriptor *fd*, and if found, delete the element.
580 *
581 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
582 * types, the *flags* argument needs to be set to 0, but for other
583 * map types, it may be specified as:
584 *
585 * **BPF_F_LOCK**
586 * Look up and delete the value of a spin-locked map
587 * without returning the lock. This must be specified if
588 * the elements contain a spinlock.
589 *
590 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
591 * implement this command as a "pop" operation, deleting the top
592 * element rather than one corresponding to *key*.
593 * The *key* and *key_len* parameters should be zeroed when
594 * issuing this operation for these map types.
595 *
596 * This command is only valid for the following map types:
597 * * **BPF_MAP_TYPE_QUEUE**
598 * * **BPF_MAP_TYPE_STACK**
599 * * **BPF_MAP_TYPE_HASH**
600 * * **BPF_MAP_TYPE_PERCPU_HASH**
601 * * **BPF_MAP_TYPE_LRU_HASH**
602 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
603 *
604 * Return
605 * Returns zero on success. On error, -1 is returned and *errno*
606 * is set appropriately.
607 *
608 * BPF_MAP_FREEZE
609 * Description
610 * Freeze the permissions of the specified map.
611 *
612 * Write permissions may be frozen by passing zero *flags*.
613 * Upon success, no future syscall invocations may alter the
614 * map state of *map_fd*. Write operations from eBPF programs
615 * are still possible for a frozen map.
616 *
617 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
618 *
619 * Return
620 * Returns zero on success. On error, -1 is returned and *errno*
621 * is set appropriately.
622 *
623 * BPF_BTF_GET_NEXT_ID
624 * Description
625 * Fetch the next BPF Type Format (BTF) object currently loaded
626 * into the kernel.
627 *
628 * Looks for the BTF object with an id greater than *start_id*
629 * and updates *next_id* on success. If no other BTF objects
630 * remain with ids higher than *start_id*, returns -1 and sets
631 * *errno* to **ENOENT**.
632 *
633 * Return
634 * Returns zero on success. On error, or when no id remains, -1
635 * is returned and *errno* is set appropriately.
636 *
637 * BPF_MAP_LOOKUP_BATCH
638 * Description
639 * Iterate and fetch multiple elements in a map.
640 *
641 * Two opaque values are used to manage batch operations,
642 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
643 * to NULL to begin the batched operation. After each subsequent
644 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
645 * *out_batch* as the *in_batch* for the next operation to
646 * continue iteration from the current point. Both *in_batch* and
647 * *out_batch* must point to memory large enough to hold a key,
648 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,
649 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters
650 * must be at least 4 bytes wide regardless of key size.
651 *
652 * The *keys* and *values* are output parameters which must point
653 * to memory large enough to hold *count* items based on the key
654 * and value size of the map *map_fd*. The *keys* buffer must be
655 * of *key_size* * *count*. The *values* buffer must be of
656 * *value_size* * *count*.
657 *
658 * The *elem_flags* argument may be specified as one of the
659 * following:
660 *
661 * **BPF_F_LOCK**
662 * Look up the value of a spin-locked map without
663 * returning the lock. This must be specified if the
664 * elements contain a spinlock.
665 *
666 * On success, *count* elements from the map are copied into the
667 * user buffer, with the keys copied into *keys* and the values
668 * copied into the corresponding indices in *values*.
669 *
670 * If an error is returned and *errno* is not **EFAULT**, *count*
671 * is set to the number of successfully processed elements.
672 *
673 * Return
674 * Returns zero on success. On error, -1 is returned and *errno*
675 * is set appropriately.
676 *
677 * May set *errno* to **ENOSPC** to indicate that *keys* or
678 * *values* is too small to dump an entire bucket during
679 * iteration of a hash-based map type.
680 *
681 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
682 * Description
683 * Iterate and delete all elements in a map.
684 *
685 * This operation has the same behavior as
686 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
687 *
688 * * Every element that is successfully returned is also deleted
689 * from the map. This is at least *count* elements. Note that
690 * *count* is both an input and an output parameter.
691 * * Upon returning with *errno* set to **EFAULT**, up to
692 * *count* elements may be deleted without returning the keys
693 * and values of the deleted elements.
694 *
695 * Return
696 * Returns zero on success. On error, -1 is returned and *errno*
697 * is set appropriately.
698 *
699 * BPF_MAP_UPDATE_BATCH
700 * Description
701 * Update multiple elements in a map by *key*.
702 *
703 * The *keys* and *values* are input parameters which must point
704 * to memory large enough to hold *count* items based on the key
705 * and value size of the map *map_fd*. The *keys* buffer must be
706 * of *key_size* * *count*. The *values* buffer must be of
707 * *value_size* * *count*.
708 *
709 * Each element specified in *keys* is sequentially updated to the
710 * value in the corresponding index in *values*. The *in_batch*
711 * and *out_batch* parameters are ignored and should be zeroed.
712 *
713 * The *elem_flags* argument should be specified as one of the
714 * following:
715 *
716 * **BPF_ANY**
717 * Create new elements or update a existing elements.
718 * **BPF_NOEXIST**
719 * Create new elements only if they do not exist.
720 * **BPF_EXIST**
721 * Update existing elements.
722 * **BPF_F_LOCK**
723 * Update spin_lock-ed map elements. This must be
724 * specified if the map value contains a spinlock.
725 *
726 * On success, *count* elements from the map are updated.
727 *
728 * If an error is returned and *errno* is not **EFAULT**, *count*
729 * is set to the number of successfully processed elements.
730 *
731 * Return
732 * Returns zero on success. On error, -1 is returned and *errno*
733 * is set appropriately.
734 *
735 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
736 * **E2BIG**. **E2BIG** indicates that the number of elements in
737 * the map reached the *max_entries* limit specified at map
738 * creation time.
739 *
740 * May set *errno* to one of the following error codes under
741 * specific circumstances:
742 *
743 * **EEXIST**
744 * If *flags* specifies **BPF_NOEXIST** and the element
745 * with *key* already exists in the map.
746 * **ENOENT**
747 * If *flags* specifies **BPF_EXIST** and the element with
748 * *key* does not exist in the map.
749 *
750 * BPF_MAP_DELETE_BATCH
751 * Description
752 * Delete multiple elements in a map by *key*.
753 *
754 * The *keys* parameter is an input parameter which must point
755 * to memory large enough to hold *count* items based on the key
756 * size of the map *map_fd*, that is, *key_size* * *count*.
757 *
758 * Each element specified in *keys* is sequentially deleted. The
759 * *in_batch*, *out_batch*, and *values* parameters are ignored
760 * and should be zeroed.
761 *
762 * The *elem_flags* argument may be specified as one of the
763 * following:
764 *
765 * **BPF_F_LOCK**
766 * Look up the value of a spin-locked map without
767 * returning the lock. This must be specified if the
768 * elements contain a spinlock.
769 *
770 * On success, *count* elements from the map are updated.
771 *
772 * If an error is returned and *errno* is not **EFAULT**, *count*
773 * is set to the number of successfully processed elements. If
774 * *errno* is **EFAULT**, up to *count* elements may be been
775 * deleted.
776 *
777 * Return
778 * Returns zero on success. On error, -1 is returned and *errno*
779 * is set appropriately.
780 *
781 * BPF_LINK_CREATE
782 * Description
783 * Attach an eBPF program to a *target_fd* at the specified
784 * *attach_type* hook and return a file descriptor handle for
785 * managing the link.
786 *
787 * Return
788 * A new file descriptor (a nonnegative integer), or -1 if an
789 * error occurred (in which case, *errno* is set appropriately).
790 *
791 * BPF_LINK_UPDATE
792 * Description
793 * Update the eBPF program in the specified *link_fd* to
794 * *new_prog_fd*.
795 *
796 * Return
797 * Returns zero on success. On error, -1 is returned and *errno*
798 * is set appropriately.
799 *
800 * BPF_LINK_GET_FD_BY_ID
801 * Description
802 * Open a file descriptor for the eBPF Link corresponding to
803 * *link_id*.
804 *
805 * Return
806 * A new file descriptor (a nonnegative integer), or -1 if an
807 * error occurred (in which case, *errno* is set appropriately).
808 *
809 * BPF_LINK_GET_NEXT_ID
810 * Description
811 * Fetch the next eBPF link currently loaded into the kernel.
812 *
813 * Looks for the eBPF link with an id greater than *start_id*
814 * and updates *next_id* on success. If no other eBPF links
815 * remain with ids higher than *start_id*, returns -1 and sets
816 * *errno* to **ENOENT**.
817 *
818 * Return
819 * Returns zero on success. On error, or when no id remains, -1
820 * is returned and *errno* is set appropriately.
821 *
822 * BPF_ENABLE_STATS
823 * Description
824 * Enable eBPF runtime statistics gathering.
825 *
826 * Runtime statistics gathering for the eBPF runtime is disabled
827 * by default to minimize the corresponding performance overhead.
828 * This command enables statistics globally.
829 *
830 * Multiple programs may independently enable statistics.
831 * After gathering the desired statistics, eBPF runtime statistics
832 * may be disabled again by calling **close**\ (2) for the file
833 * descriptor returned by this function. Statistics will only be
834 * disabled system-wide when all outstanding file descriptors
835 * returned by prior calls for this subcommand are closed.
836 *
837 * Return
838 * A new file descriptor (a nonnegative integer), or -1 if an
839 * error occurred (in which case, *errno* is set appropriately).
840 *
841 * BPF_ITER_CREATE
842 * Description
843 * Create an iterator on top of the specified *link_fd* (as
844 * previously created using **BPF_LINK_CREATE**) and return a
845 * file descriptor that can be used to trigger the iteration.
846 *
847 * If the resulting file descriptor is pinned to the filesystem
848 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
849 * for that path will trigger the iterator to read kernel state
850 * using the eBPF program attached to *link_fd*.
851 *
852 * Return
853 * A new file descriptor (a nonnegative integer), or -1 if an
854 * error occurred (in which case, *errno* is set appropriately).
855 *
856 * BPF_LINK_DETACH
857 * Description
858 * Forcefully detach the specified *link_fd* from its
859 * corresponding attachment point.
860 *
861 * Return
862 * Returns zero on success. On error, -1 is returned and *errno*
863 * is set appropriately.
864 *
865 * BPF_PROG_BIND_MAP
866 * Description
867 * Bind a map to the lifetime of an eBPF program.
868 *
869 * The map identified by *map_fd* is bound to the program
870 * identified by *prog_fd* and only released when *prog_fd* is
871 * released. This may be used in cases where metadata should be
872 * associated with a program which otherwise does not contain any
873 * references to the map (for example, embedded in the eBPF
874 * program instructions).
875 *
876 * Return
877 * Returns zero on success. On error, -1 is returned and *errno*
878 * is set appropriately.
879 *
880 * BPF_TOKEN_CREATE
881 * Description
882 * Create BPF token with embedded information about what
883 * BPF-related functionality it allows:
884 * - a set of allowed bpf() syscall commands;
885 * - a set of allowed BPF map types to be created with
886 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;
887 * - a set of allowed BPF program types and BPF program attach
888 * types to be loaded with BPF_PROG_LOAD command, if
889 * BPF_PROG_LOAD itself is allowed.
890 *
891 * BPF token is created (derived) from an instance of BPF FS,
892 * assuming it has necessary delegation mount options specified.
893 * This BPF token can be passed as an extra parameter to various
894 * bpf() syscall commands to grant BPF subsystem functionality to
895 * unprivileged processes.
896 *
897 * When created, BPF token is "associated" with the owning
898 * user namespace of BPF FS instance (super block) that it was
899 * derived from, and subsequent BPF operations performed with
900 * BPF token would be performing capabilities checks (i.e.,
901 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within
902 * that user namespace. Without BPF token, such capabilities
903 * have to be granted in init user namespace, making bpf()
904 * syscall incompatible with user namespace, for the most part.
905 *
906 * Return
907 * A new file descriptor (a nonnegative integer), or -1 if an
908 * error occurred (in which case, *errno* is set appropriately).
909 *
910 * BPF_PROG_STREAM_READ_BY_FD
911 * Description
912 * Read data of a program's BPF stream. The program is identified
913 * by *prog_fd*, and the stream is identified by the *stream_id*.
914 * The data is copied to a buffer pointed to by *stream_buf*, and
915 * filled less than or equal to *stream_buf_len* bytes.
916 *
917 * Return
918 * Number of bytes read from the stream on success, or -1 if an
919 * error occurred (in which case, *errno* is set appropriately).
920 *
921 * NOTES
922 * eBPF objects (maps and programs) can be shared between processes.
923 *
924 * * After **fork**\ (2), the child inherits file descriptors
925 * referring to the same eBPF objects.
926 * * File descriptors referring to eBPF objects can be transferred over
927 * **unix**\ (7) domain sockets.
928 * * File descriptors referring to eBPF objects can be duplicated in the
929 * usual way, using **dup**\ (2) and similar calls.
930 * * File descriptors referring to eBPF objects can be pinned to the
931 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
932 *
933 * An eBPF object is deallocated only after all file descriptors referring
934 * to the object have been closed and no references remain pinned to the
935 * filesystem or attached (for example, bound to a program or device).
936 */
937enum bpf_cmd {
938 BPF_MAP_CREATE,
939 BPF_MAP_LOOKUP_ELEM,
940 BPF_MAP_UPDATE_ELEM,
941 BPF_MAP_DELETE_ELEM,
942 BPF_MAP_GET_NEXT_KEY,
943 BPF_PROG_LOAD,
944 BPF_OBJ_PIN,
945 BPF_OBJ_GET,
946 BPF_PROG_ATTACH,
947 BPF_PROG_DETACH,
948 BPF_PROG_TEST_RUN,
949 BPF_PROG_RUN = BPF_PROG_TEST_RUN,
950 BPF_PROG_GET_NEXT_ID,
951 BPF_MAP_GET_NEXT_ID,
952 BPF_PROG_GET_FD_BY_ID,
953 BPF_MAP_GET_FD_BY_ID,
954 BPF_OBJ_GET_INFO_BY_FD,
955 BPF_PROG_QUERY,
956 BPF_RAW_TRACEPOINT_OPEN,
957 BPF_BTF_LOAD,
958 BPF_BTF_GET_FD_BY_ID,
959 BPF_TASK_FD_QUERY,
960 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
961 BPF_MAP_FREEZE,
962 BPF_BTF_GET_NEXT_ID,
963 BPF_MAP_LOOKUP_BATCH,
964 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
965 BPF_MAP_UPDATE_BATCH,
966 BPF_MAP_DELETE_BATCH,
967 BPF_LINK_CREATE,
968 BPF_LINK_UPDATE,
969 BPF_LINK_GET_FD_BY_ID,
970 BPF_LINK_GET_NEXT_ID,
971 BPF_ENABLE_STATS,
972 BPF_ITER_CREATE,
973 BPF_LINK_DETACH,
974 BPF_PROG_BIND_MAP,
975 BPF_TOKEN_CREATE,
976 BPF_PROG_STREAM_READ_BY_FD,
977 __MAX_BPF_CMD,
978};
979
980enum bpf_map_type {
981 BPF_MAP_TYPE_UNSPEC,
982 BPF_MAP_TYPE_HASH,
983 BPF_MAP_TYPE_ARRAY,
984 BPF_MAP_TYPE_PROG_ARRAY,
985 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
986 BPF_MAP_TYPE_PERCPU_HASH,
987 BPF_MAP_TYPE_PERCPU_ARRAY,
988 BPF_MAP_TYPE_STACK_TRACE,
989 BPF_MAP_TYPE_CGROUP_ARRAY,
990 BPF_MAP_TYPE_LRU_HASH,
991 BPF_MAP_TYPE_LRU_PERCPU_HASH,
992 BPF_MAP_TYPE_LPM_TRIE,
993 BPF_MAP_TYPE_ARRAY_OF_MAPS,
994 BPF_MAP_TYPE_HASH_OF_MAPS,
995 BPF_MAP_TYPE_DEVMAP,
996 BPF_MAP_TYPE_SOCKMAP,
997 BPF_MAP_TYPE_CPUMAP,
998 BPF_MAP_TYPE_XSKMAP,
999 BPF_MAP_TYPE_SOCKHASH,
1000 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1001 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
1002 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
1003 * both cgroup-attached and other progs and supports all functionality
1004 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
1005 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
1006 */
1007 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1008 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
1009 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1010 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
1011 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
1012 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1013 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1014 * deprecated.
1015 */
1016 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1017 BPF_MAP_TYPE_QUEUE,
1018 BPF_MAP_TYPE_STACK,
1019 BPF_MAP_TYPE_SK_STORAGE,
1020 BPF_MAP_TYPE_DEVMAP_HASH,
1021 BPF_MAP_TYPE_STRUCT_OPS,
1022 BPF_MAP_TYPE_RINGBUF,
1023 BPF_MAP_TYPE_INODE_STORAGE,
1024 BPF_MAP_TYPE_TASK_STORAGE,
1025 BPF_MAP_TYPE_BLOOM_FILTER,
1026 BPF_MAP_TYPE_USER_RINGBUF,
1027 BPF_MAP_TYPE_CGRP_STORAGE,
1028 BPF_MAP_TYPE_ARENA,
1029 BPF_MAP_TYPE_INSN_ARRAY,
1030 __MAX_BPF_MAP_TYPE
1031};
1032
1033/* Note that tracing related programs such as
1034 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
1035 * are not subject to a stable API since kernel internal data
1036 * structures can change from release to release and may
1037 * therefore break existing tracing BPF programs. Tracing BPF
1038 * programs correspond to /a/ specific kernel which is to be
1039 * analyzed, and not /a/ specific kernel /and/ all future ones.
1040 */
1041enum bpf_prog_type {
1042 BPF_PROG_TYPE_UNSPEC,
1043 BPF_PROG_TYPE_SOCKET_FILTER,
1044 BPF_PROG_TYPE_KPROBE,
1045 BPF_PROG_TYPE_SCHED_CLS,
1046 BPF_PROG_TYPE_SCHED_ACT,
1047 BPF_PROG_TYPE_TRACEPOINT,
1048 BPF_PROG_TYPE_XDP,
1049 BPF_PROG_TYPE_PERF_EVENT,
1050 BPF_PROG_TYPE_CGROUP_SKB,
1051 BPF_PROG_TYPE_CGROUP_SOCK,
1052 BPF_PROG_TYPE_LWT_IN,
1053 BPF_PROG_TYPE_LWT_OUT,
1054 BPF_PROG_TYPE_LWT_XMIT,
1055 BPF_PROG_TYPE_SOCK_OPS,
1056 BPF_PROG_TYPE_SK_SKB,
1057 BPF_PROG_TYPE_CGROUP_DEVICE,
1058 BPF_PROG_TYPE_SK_MSG,
1059 BPF_PROG_TYPE_RAW_TRACEPOINT,
1060 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
1061 BPF_PROG_TYPE_LWT_SEG6LOCAL,
1062 BPF_PROG_TYPE_LIRC_MODE2,
1063 BPF_PROG_TYPE_SK_REUSEPORT,
1064 BPF_PROG_TYPE_FLOW_DISSECTOR,
1065 BPF_PROG_TYPE_CGROUP_SYSCTL,
1066 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
1067 BPF_PROG_TYPE_CGROUP_SOCKOPT,
1068 BPF_PROG_TYPE_TRACING,
1069 BPF_PROG_TYPE_STRUCT_OPS,
1070 BPF_PROG_TYPE_EXT,
1071 BPF_PROG_TYPE_LSM,
1072 BPF_PROG_TYPE_SK_LOOKUP,
1073 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
1074 BPF_PROG_TYPE_NETFILTER,
1075 __MAX_BPF_PROG_TYPE
1076};
1077
1078enum bpf_attach_type {
1079 BPF_CGROUP_INET_INGRESS,
1080 BPF_CGROUP_INET_EGRESS,
1081 BPF_CGROUP_INET_SOCK_CREATE,
1082 BPF_CGROUP_SOCK_OPS,
1083 BPF_SK_SKB_STREAM_PARSER,
1084 BPF_SK_SKB_STREAM_VERDICT,
1085 BPF_CGROUP_DEVICE,
1086 BPF_SK_MSG_VERDICT,
1087 BPF_CGROUP_INET4_BIND,
1088 BPF_CGROUP_INET6_BIND,
1089 BPF_CGROUP_INET4_CONNECT,
1090 BPF_CGROUP_INET6_CONNECT,
1091 BPF_CGROUP_INET4_POST_BIND,
1092 BPF_CGROUP_INET6_POST_BIND,
1093 BPF_CGROUP_UDP4_SENDMSG,
1094 BPF_CGROUP_UDP6_SENDMSG,
1095 BPF_LIRC_MODE2,
1096 BPF_FLOW_DISSECTOR,
1097 BPF_CGROUP_SYSCTL,
1098 BPF_CGROUP_UDP4_RECVMSG,
1099 BPF_CGROUP_UDP6_RECVMSG,
1100 BPF_CGROUP_GETSOCKOPT,
1101 BPF_CGROUP_SETSOCKOPT,
1102 BPF_TRACE_RAW_TP,
1103 BPF_TRACE_FENTRY,
1104 BPF_TRACE_FEXIT,
1105 BPF_MODIFY_RETURN,
1106 BPF_LSM_MAC,
1107 BPF_TRACE_ITER,
1108 BPF_CGROUP_INET4_GETPEERNAME,
1109 BPF_CGROUP_INET6_GETPEERNAME,
1110 BPF_CGROUP_INET4_GETSOCKNAME,
1111 BPF_CGROUP_INET6_GETSOCKNAME,
1112 BPF_XDP_DEVMAP,
1113 BPF_CGROUP_INET_SOCK_RELEASE,
1114 BPF_XDP_CPUMAP,
1115 BPF_SK_LOOKUP,
1116 BPF_XDP,
1117 BPF_SK_SKB_VERDICT,
1118 BPF_SK_REUSEPORT_SELECT,
1119 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1120 BPF_PERF_EVENT,
1121 BPF_TRACE_KPROBE_MULTI,
1122 BPF_LSM_CGROUP,
1123 BPF_STRUCT_OPS,
1124 BPF_NETFILTER,
1125 BPF_TCX_INGRESS,
1126 BPF_TCX_EGRESS,
1127 BPF_TRACE_UPROBE_MULTI,
1128 BPF_CGROUP_UNIX_CONNECT,
1129 BPF_CGROUP_UNIX_SENDMSG,
1130 BPF_CGROUP_UNIX_RECVMSG,
1131 BPF_CGROUP_UNIX_GETPEERNAME,
1132 BPF_CGROUP_UNIX_GETSOCKNAME,
1133 BPF_NETKIT_PRIMARY,
1134 BPF_NETKIT_PEER,
1135 BPF_TRACE_KPROBE_SESSION,
1136 BPF_TRACE_UPROBE_SESSION,
1137 __MAX_BPF_ATTACH_TYPE
1138};
1139
1140#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1141
1142/* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]
1143 * in sync with the definitions below.
1144 */
1145enum bpf_link_type {
1146 BPF_LINK_TYPE_UNSPEC = 0,
1147 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1148 BPF_LINK_TYPE_TRACING = 2,
1149 BPF_LINK_TYPE_CGROUP = 3,
1150 BPF_LINK_TYPE_ITER = 4,
1151 BPF_LINK_TYPE_NETNS = 5,
1152 BPF_LINK_TYPE_XDP = 6,
1153 BPF_LINK_TYPE_PERF_EVENT = 7,
1154 BPF_LINK_TYPE_KPROBE_MULTI = 8,
1155 BPF_LINK_TYPE_STRUCT_OPS = 9,
1156 BPF_LINK_TYPE_NETFILTER = 10,
1157 BPF_LINK_TYPE_TCX = 11,
1158 BPF_LINK_TYPE_UPROBE_MULTI = 12,
1159 BPF_LINK_TYPE_NETKIT = 13,
1160 BPF_LINK_TYPE_SOCKMAP = 14,
1161 __MAX_BPF_LINK_TYPE,
1162};
1163
1164#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
1165
1166enum bpf_perf_event_type {
1167 BPF_PERF_EVENT_UNSPEC = 0,
1168 BPF_PERF_EVENT_UPROBE = 1,
1169 BPF_PERF_EVENT_URETPROBE = 2,
1170 BPF_PERF_EVENT_KPROBE = 3,
1171 BPF_PERF_EVENT_KRETPROBE = 4,
1172 BPF_PERF_EVENT_TRACEPOINT = 5,
1173 BPF_PERF_EVENT_EVENT = 6,
1174};
1175
1176/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1177 *
1178 * NONE(default): No further bpf programs allowed in the subtree.
1179 *
1180 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1181 * the program in this cgroup yields to sub-cgroup program.
1182 *
1183 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1184 * that cgroup program gets run in addition to the program in this cgroup.
1185 *
1186 * Only one program is allowed to be attached to a cgroup with
1187 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1188 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1189 * release old program and attach the new one. Attach flags has to match.
1190 *
1191 * Multiple programs are allowed to be attached to a cgroup with
1192 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1193 * (those that were attached first, run first)
1194 * The programs of sub-cgroup are executed first, then programs of
1195 * this cgroup and then programs of parent cgroup.
1196 * When children program makes decision (like picking TCP CA or sock bind)
1197 * parent program has a chance to override it.
1198 *
1199 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1200 * programs for a cgroup. Though it's possible to replace an old program at
1201 * any position by also specifying BPF_F_REPLACE flag and position itself in
1202 * replace_bpf_fd attribute. Old program at this position will be released.
1203 *
1204 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1205 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1206 * Ex1:
1207 * cgrp1 (MULTI progs A, B) ->
1208 * cgrp2 (OVERRIDE prog C) ->
1209 * cgrp3 (MULTI prog D) ->
1210 * cgrp4 (OVERRIDE prog E) ->
1211 * cgrp5 (NONE prog F)
1212 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1213 * if prog F is detached, the execution is E,D,A,B
1214 * if prog F and D are detached, the execution is E,A,B
1215 * if prog F, E and D are detached, the execution is C,A,B
1216 *
1217 * All eligible programs are executed regardless of return code from
1218 * earlier programs.
1219 */
1220#define BPF_F_ALLOW_OVERRIDE (1U << 0)
1221#define BPF_F_ALLOW_MULTI (1U << 1)
1222/* Generic attachment flags. */
1223#define BPF_F_REPLACE (1U << 2)
1224#define BPF_F_BEFORE (1U << 3)
1225#define BPF_F_AFTER (1U << 4)
1226#define BPF_F_ID (1U << 5)
1227#define BPF_F_PREORDER (1U << 6)
1228#define BPF_F_LINK BPF_F_LINK /* 1 << 13 */
1229
1230/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1231 * verifier will perform strict alignment checking as if the kernel
1232 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1233 * and NET_IP_ALIGN defined to 2.
1234 */
1235#define BPF_F_STRICT_ALIGNMENT (1U << 0)
1236
1237/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1238 * verifier will allow any alignment whatsoever. On platforms
1239 * with strict alignment requirements for loads ands stores (such
1240 * as sparc and mips) the verifier validates that all loads and
1241 * stores provably follow this requirement. This flag turns that
1242 * checking and enforcement off.
1243 *
1244 * It is mostly used for testing when we want to validate the
1245 * context and memory access aspects of the verifier, but because
1246 * of an unaligned access the alignment check would trigger before
1247 * the one we are interested in.
1248 */
1249#define BPF_F_ANY_ALIGNMENT (1U << 1)
1250
1251/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1252 * Verifier does sub-register def/use analysis and identifies instructions whose
1253 * def only matters for low 32-bit, high 32-bit is never referenced later
1254 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1255 * that it is safe to ignore clearing high 32-bit for these instructions. This
1256 * saves some back-ends a lot of code-gen. However such optimization is not
1257 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1258 * hence hasn't used verifier's analysis result. But, we really want to have a
1259 * way to be able to verify the correctness of the described optimization on
1260 * x86_64 on which testsuites are frequently exercised.
1261 *
1262 * So, this flag is introduced. Once it is set, verifier will randomize high
1263 * 32-bit for those instructions who has been identified as safe to ignore them.
1264 * Then, if verifier is not doing correct analysis, such randomization will
1265 * regress tests to expose bugs.
1266 */
1267#define BPF_F_TEST_RND_HI32 (1U << 2)
1268
1269/* The verifier internal test flag. Behavior is undefined */
1270#define BPF_F_TEST_STATE_FREQ (1U << 3)
1271
1272/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1273 * restrict map and helper usage for such programs. Sleepable BPF programs can
1274 * only be attached to hooks where kernel execution context allows sleeping.
1275 * Such programs are allowed to use helpers that may sleep like
1276 * bpf_copy_from_user().
1277 */
1278#define BPF_F_SLEEPABLE (1U << 4)
1279
1280/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1281 * fully support xdp frags.
1282 */
1283#define BPF_F_XDP_HAS_FRAGS (1U << 5)
1284
1285/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1286 * program becomes device-bound but can access XDP metadata.
1287 */
1288#define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6)
1289
1290/* The verifier internal test flag. Behavior is undefined */
1291#define BPF_F_TEST_REG_INVARIANTS (1U << 7)
1292
1293/* link_create.kprobe_multi.flags used in LINK_CREATE command for
1294 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1295 */
1296enum {
1297 BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
1298};
1299
1300/* link_create.uprobe_multi.flags used in LINK_CREATE command for
1301 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1302 */
1303enum {
1304 BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
1305};
1306
1307/* link_create.netfilter.flags used in LINK_CREATE command for
1308 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1309 */
1310#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1311
1312/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1313 * the following extensions:
1314 *
1315 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1316 * insn[0].imm: map fd or fd_idx
1317 * insn[1].imm: 0
1318 * insn[0].off: 0
1319 * insn[1].off: 0
1320 * ldimm64 rewrite: address of map
1321 * verifier type: CONST_PTR_TO_MAP
1322 */
1323#define BPF_PSEUDO_MAP_FD 1
1324#define BPF_PSEUDO_MAP_IDX 5
1325
1326/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1327 * insn[0].imm: map fd or fd_idx
1328 * insn[1].imm: offset into value
1329 * insn[0].off: 0
1330 * insn[1].off: 0
1331 * ldimm64 rewrite: address of map[0]+offset
1332 * verifier type: PTR_TO_MAP_VALUE
1333 */
1334#define BPF_PSEUDO_MAP_VALUE 2
1335#define BPF_PSEUDO_MAP_IDX_VALUE 6
1336
1337/* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1338 * insn[0].imm: kernel btd id of VAR
1339 * insn[1].imm: 0
1340 * insn[0].off: 0
1341 * insn[1].off: 0
1342 * ldimm64 rewrite: address of the kernel variable
1343 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1344 * is struct/union.
1345 */
1346#define BPF_PSEUDO_BTF_ID 3
1347/* insn[0].src_reg: BPF_PSEUDO_FUNC
1348 * insn[0].imm: insn offset to the func
1349 * insn[1].imm: 0
1350 * insn[0].off: 0
1351 * insn[1].off: 0
1352 * ldimm64 rewrite: address of the function
1353 * verifier type: PTR_TO_FUNC.
1354 */
1355#define BPF_PSEUDO_FUNC 4
1356
1357/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1358 * offset to another bpf function
1359 */
1360#define BPF_PSEUDO_CALL 1
1361/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1362 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1363 */
1364#define BPF_PSEUDO_KFUNC_CALL 2
1365
1366enum bpf_addr_space_cast {
1367 BPF_ADDR_SPACE_CAST = 1,
1368};
1369
1370/* flags for BPF_MAP_UPDATE_ELEM command */
1371enum {
1372 BPF_ANY = 0, /* create new element or update existing */
1373 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1374 BPF_EXIST = 2, /* update existing element */
1375 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1376};
1377
1378/* flags for BPF_MAP_CREATE command */
1379enum {
1380 BPF_F_NO_PREALLOC = (1U << 0),
1381/* Instead of having one common LRU list in the
1382 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1383 * which can scale and perform better.
1384 * Note, the LRU nodes (including free nodes) cannot be moved
1385 * across different LRU lists.
1386 */
1387 BPF_F_NO_COMMON_LRU = (1U << 1),
1388/* Specify numa node during map creation */
1389 BPF_F_NUMA_NODE = (1U << 2),
1390
1391/* Flags for accessing BPF object from syscall side. */
1392 BPF_F_RDONLY = (1U << 3),
1393 BPF_F_WRONLY = (1U << 4),
1394
1395/* Flag for stack_map, store build_id+offset instead of pointer */
1396 BPF_F_STACK_BUILD_ID = (1U << 5),
1397
1398/* Zero-initialize hash function seed. This should only be used for testing. */
1399 BPF_F_ZERO_SEED = (1U << 6),
1400
1401/* Flags for accessing BPF object from program side. */
1402 BPF_F_RDONLY_PROG = (1U << 7),
1403 BPF_F_WRONLY_PROG = (1U << 8),
1404
1405/* Clone map from listener for newly accepted socket */
1406 BPF_F_CLONE = (1U << 9),
1407
1408/* Enable memory-mapping BPF map */
1409 BPF_F_MMAPABLE = (1U << 10),
1410
1411/* Share perf_event among processes */
1412 BPF_F_PRESERVE_ELEMS = (1U << 11),
1413
1414/* Create a map that is suitable to be an inner map with dynamic max entries */
1415 BPF_F_INNER_MAP = (1U << 12),
1416
1417/* Create a map that will be registered/unregesitered by the backed bpf_link */
1418 BPF_F_LINK = (1U << 13),
1419
1420/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1421 BPF_F_PATH_FD = (1U << 14),
1422
1423/* Flag for value_type_btf_obj_fd, the fd is available */
1424 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15),
1425
1426/* BPF token FD is passed in a corresponding command's token_fd field */
1427 BPF_F_TOKEN_FD = (1U << 16),
1428
1429/* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */
1430 BPF_F_SEGV_ON_FAULT = (1U << 17),
1431
1432/* Do not translate kernel bpf_arena pointers to user pointers */
1433 BPF_F_NO_USER_CONV = (1U << 18),
1434
1435/* Enable BPF ringbuf overwrite mode */
1436 BPF_F_RB_OVERWRITE = (1U << 19),
1437};
1438
1439/* Flags for BPF_PROG_QUERY. */
1440
1441/* Query effective (directly attached + inherited from ancestor cgroups)
1442 * programs that will be executed for events within a cgroup.
1443 * attach_flags with this flag are always returned 0.
1444 */
1445#define BPF_F_QUERY_EFFECTIVE (1U << 0)
1446
1447/* Flags for BPF_PROG_TEST_RUN */
1448
1449/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1450#define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1451/* If set, XDP frames will be transmitted after processing */
1452#define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
1453/* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */
1454#define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2)
1455
1456/* type for BPF_ENABLE_STATS */
1457enum bpf_stats_type {
1458 /* enabled run_time_ns and run_cnt */
1459 BPF_STATS_RUN_TIME = 0,
1460};
1461
1462enum bpf_stack_build_id_status {
1463 /* user space need an empty entry to identify end of a trace */
1464 BPF_STACK_BUILD_ID_EMPTY = 0,
1465 /* with valid build_id and offset */
1466 BPF_STACK_BUILD_ID_VALID = 1,
1467 /* couldn't get build_id, fallback to ip */
1468 BPF_STACK_BUILD_ID_IP = 2,
1469};
1470
1471#define BPF_BUILD_ID_SIZE 20
1472struct bpf_stack_build_id {
1473 __s32 status;
1474 unsigned char build_id[BPF_BUILD_ID_SIZE];
1475 union {
1476 __u64 offset;
1477 __u64 ip;
1478 };
1479};
1480
1481#define BPF_OBJ_NAME_LEN 16U
1482
1483enum {
1484 BPF_STREAM_STDOUT = 1,
1485 BPF_STREAM_STDERR = 2,
1486};
1487
1488union bpf_attr {
1489 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1490 __u32 map_type; /* one of enum bpf_map_type */
1491 __u32 key_size; /* size of key in bytes */
1492 __u32 value_size; /* size of value in bytes */
1493 __u32 max_entries; /* max number of entries in a map */
1494 __u32 map_flags; /* BPF_MAP_CREATE related
1495 * flags defined above.
1496 */
1497 __u32 inner_map_fd; /* fd pointing to the inner map */
1498 __u32 numa_node; /* numa node (effective only if
1499 * BPF_F_NUMA_NODE is set).
1500 */
1501 char map_name[BPF_OBJ_NAME_LEN];
1502 __u32 map_ifindex; /* ifindex of netdev to create on */
1503 __u32 btf_fd; /* fd pointing to a BTF type data */
1504 __u32 btf_key_type_id; /* BTF type_id of the key */
1505 __u32 btf_value_type_id; /* BTF type_id of the value */
1506 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1507 * struct stored as the
1508 * map value
1509 */
1510 /* Any per-map-type extra fields
1511 *
1512 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1513 * number of hash functions (if 0, the bloom filter will default
1514 * to using 5 hash functions).
1515 *
1516 * BPF_MAP_TYPE_ARENA - contains the address where user space
1517 * is going to mmap() the arena. It has to be page aligned.
1518 */
1519 __u64 map_extra;
1520
1521 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF
1522 * type data for
1523 * btf_vmlinux_value_type_id.
1524 */
1525 /* BPF token FD to use with BPF_MAP_CREATE operation.
1526 * If provided, map_flags should have BPF_F_TOKEN_FD flag set.
1527 */
1528 __s32 map_token_fd;
1529
1530 /* Hash of the program that has exclusive access to the map.
1531 */
1532 __aligned_u64 excl_prog_hash;
1533 /* Size of the passed excl_prog_hash. */
1534 __u32 excl_prog_hash_size;
1535 };
1536
1537 struct { /* anonymous struct used by BPF_MAP_*_ELEM and BPF_MAP_FREEZE commands */
1538 __u32 map_fd;
1539 __aligned_u64 key;
1540 union {
1541 __aligned_u64 value;
1542 __aligned_u64 next_key;
1543 };
1544 __u64 flags;
1545 };
1546
1547 struct { /* struct used by BPF_MAP_*_BATCH commands */
1548 __aligned_u64 in_batch; /* start batch,
1549 * NULL to start from beginning
1550 */
1551 __aligned_u64 out_batch; /* output: next start batch */
1552 __aligned_u64 keys;
1553 __aligned_u64 values;
1554 __u32 count; /* input/output:
1555 * input: # of key/value
1556 * elements
1557 * output: # of filled elements
1558 */
1559 __u32 map_fd;
1560 __u64 elem_flags;
1561 __u64 flags;
1562 } batch;
1563
1564 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1565 __u32 prog_type; /* one of enum bpf_prog_type */
1566 __u32 insn_cnt;
1567 __aligned_u64 insns;
1568 __aligned_u64 license;
1569 __u32 log_level; /* verbosity level of verifier */
1570 __u32 log_size; /* size of user buffer */
1571 __aligned_u64 log_buf; /* user supplied buffer */
1572 __u32 kern_version; /* not used */
1573 __u32 prog_flags;
1574 char prog_name[BPF_OBJ_NAME_LEN];
1575 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1576 /* For some prog types expected attach type must be known at
1577 * load time to verify attach type specific parts of prog
1578 * (context accesses, allowed helpers, etc).
1579 */
1580 __u32 expected_attach_type;
1581 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1582 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1583 __aligned_u64 func_info; /* func info */
1584 __u32 func_info_cnt; /* number of bpf_func_info records */
1585 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1586 __aligned_u64 line_info; /* line info */
1587 __u32 line_info_cnt; /* number of bpf_line_info records */
1588 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1589 union {
1590 /* valid prog_fd to attach to bpf prog */
1591 __u32 attach_prog_fd;
1592 /* or valid module BTF object fd or 0 to attach to vmlinux */
1593 __u32 attach_btf_obj_fd;
1594 };
1595 __u32 core_relo_cnt; /* number of bpf_core_relo */
1596 __aligned_u64 fd_array; /* array of FDs */
1597 __aligned_u64 core_relos;
1598 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1599 /* output: actual total log contents size (including termintaing zero).
1600 * It could be both larger than original log_size (if log was
1601 * truncated), or smaller (if log buffer wasn't filled completely).
1602 */
1603 __u32 log_true_size;
1604 /* BPF token FD to use with BPF_PROG_LOAD operation.
1605 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.
1606 */
1607 __s32 prog_token_fd;
1608 /* The fd_array_cnt can be used to pass the length of the
1609 * fd_array array. In this case all the [map] file descriptors
1610 * passed in this array will be bound to the program, even if
1611 * the maps are not referenced directly. The functionality is
1612 * similar to the BPF_PROG_BIND_MAP syscall, but maps can be
1613 * used by the verifier during the program load. If provided,
1614 * then the fd_array[0,...,fd_array_cnt-1] is expected to be
1615 * continuous.
1616 */
1617 __u32 fd_array_cnt;
1618 /* Pointer to a buffer containing the signature of the BPF
1619 * program.
1620 */
1621 __aligned_u64 signature;
1622 /* Size of the signature buffer in bytes. */
1623 __u32 signature_size;
1624 /* ID of the kernel keyring to be used for signature
1625 * verification.
1626 */
1627 __s32 keyring_id;
1628 };
1629
1630 struct { /* anonymous struct used by BPF_OBJ_* commands */
1631 __aligned_u64 pathname;
1632 __u32 bpf_fd;
1633 __u32 file_flags;
1634 /* Same as dirfd in openat() syscall; see openat(2)
1635 * manpage for details of path FD and pathname semantics;
1636 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1637 * file_flags field, otherwise it should be set to zero;
1638 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1639 */
1640 __s32 path_fd;
1641 };
1642
1643 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1644 union {
1645 __u32 target_fd; /* target object to attach to or ... */
1646 __u32 target_ifindex; /* target ifindex */
1647 };
1648 __u32 attach_bpf_fd;
1649 __u32 attach_type;
1650 __u32 attach_flags;
1651 __u32 replace_bpf_fd;
1652 union {
1653 __u32 relative_fd;
1654 __u32 relative_id;
1655 };
1656 __u64 expected_revision;
1657 };
1658
1659 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1660 __u32 prog_fd;
1661 __u32 retval;
1662 __u32 data_size_in; /* input: len of data_in */
1663 __u32 data_size_out; /* input/output: len of data_out
1664 * returns ENOSPC if data_out
1665 * is too small.
1666 */
1667 __aligned_u64 data_in;
1668 __aligned_u64 data_out;
1669 __u32 repeat;
1670 __u32 duration;
1671 __u32 ctx_size_in; /* input: len of ctx_in */
1672 __u32 ctx_size_out; /* input/output: len of ctx_out
1673 * returns ENOSPC if ctx_out
1674 * is too small.
1675 */
1676 __aligned_u64 ctx_in;
1677 __aligned_u64 ctx_out;
1678 __u32 flags;
1679 __u32 cpu;
1680 __u32 batch_size;
1681 } test;
1682
1683 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1684 union {
1685 __u32 start_id;
1686 __u32 prog_id;
1687 __u32 map_id;
1688 __u32 btf_id;
1689 __u32 link_id;
1690 };
1691 __u32 next_id;
1692 __u32 open_flags;
1693 __s32 fd_by_id_token_fd;
1694 };
1695
1696 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1697 __u32 bpf_fd;
1698 __u32 info_len;
1699 __aligned_u64 info;
1700 } info;
1701
1702 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1703 union {
1704 __u32 target_fd; /* target object to query or ... */
1705 __u32 target_ifindex; /* target ifindex */
1706 };
1707 __u32 attach_type;
1708 __u32 query_flags;
1709 __u32 attach_flags;
1710 __aligned_u64 prog_ids;
1711 union {
1712 __u32 prog_cnt;
1713 __u32 count;
1714 };
1715 __u32 :32;
1716 /* output: per-program attach_flags.
1717 * not allowed to be set during effective query.
1718 */
1719 __aligned_u64 prog_attach_flags;
1720 __aligned_u64 link_ids;
1721 __aligned_u64 link_attach_flags;
1722 __u64 revision;
1723 } query;
1724
1725 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1726 __u64 name;
1727 __u32 prog_fd;
1728 __u32 :32;
1729 __aligned_u64 cookie;
1730 } raw_tracepoint;
1731
1732 struct { /* anonymous struct for BPF_BTF_LOAD */
1733 __aligned_u64 btf;
1734 __aligned_u64 btf_log_buf;
1735 __u32 btf_size;
1736 __u32 btf_log_size;
1737 __u32 btf_log_level;
1738 /* output: actual total log contents size (including termintaing zero).
1739 * It could be both larger than original log_size (if log was
1740 * truncated), or smaller (if log buffer wasn't filled completely).
1741 */
1742 __u32 btf_log_true_size;
1743 __u32 btf_flags;
1744 /* BPF token FD to use with BPF_BTF_LOAD operation.
1745 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.
1746 */
1747 __s32 btf_token_fd;
1748 };
1749
1750 struct {
1751 __u32 pid; /* input: pid */
1752 __u32 fd; /* input: fd */
1753 __u32 flags; /* input: flags */
1754 __u32 buf_len; /* input/output: buf len */
1755 __aligned_u64 buf; /* input/output:
1756 * tp_name for tracepoint
1757 * symbol for kprobe
1758 * filename for uprobe
1759 */
1760 __u32 prog_id; /* output: prod_id */
1761 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1762 __u64 probe_offset; /* output: probe_offset */
1763 __u64 probe_addr; /* output: probe_addr */
1764 } task_fd_query;
1765
1766 struct { /* struct used by BPF_LINK_CREATE command */
1767 union {
1768 __u32 prog_fd; /* eBPF program to attach */
1769 __u32 map_fd; /* struct_ops to attach */
1770 };
1771 union {
1772 __u32 target_fd; /* target object to attach to or ... */
1773 __u32 target_ifindex; /* target ifindex */
1774 };
1775 __u32 attach_type; /* attach type */
1776 __u32 flags; /* extra flags */
1777 union {
1778 __u32 target_btf_id; /* btf_id of target to attach to */
1779 struct {
1780 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1781 __u32 iter_info_len; /* iter_info length */
1782 };
1783 struct {
1784 /* black box user-provided value passed through
1785 * to BPF program at the execution time and
1786 * accessible through bpf_get_attach_cookie() BPF helper
1787 */
1788 __u64 bpf_cookie;
1789 } perf_event;
1790 struct {
1791 __u32 flags;
1792 __u32 cnt;
1793 __aligned_u64 syms;
1794 __aligned_u64 addrs;
1795 __aligned_u64 cookies;
1796 } kprobe_multi;
1797 struct {
1798 /* this is overlaid with the target_btf_id above. */
1799 __u32 target_btf_id;
1800 /* black box user-provided value passed through
1801 * to BPF program at the execution time and
1802 * accessible through bpf_get_attach_cookie() BPF helper
1803 */
1804 __u64 cookie;
1805 } tracing;
1806 struct {
1807 __u32 pf;
1808 __u32 hooknum;
1809 __s32 priority;
1810 __u32 flags;
1811 } netfilter;
1812 struct {
1813 union {
1814 __u32 relative_fd;
1815 __u32 relative_id;
1816 };
1817 __u64 expected_revision;
1818 } tcx;
1819 struct {
1820 __aligned_u64 path;
1821 __aligned_u64 offsets;
1822 __aligned_u64 ref_ctr_offsets;
1823 __aligned_u64 cookies;
1824 __u32 cnt;
1825 __u32 flags;
1826 __u32 pid;
1827 } uprobe_multi;
1828 struct {
1829 union {
1830 __u32 relative_fd;
1831 __u32 relative_id;
1832 };
1833 __u64 expected_revision;
1834 } netkit;
1835 struct {
1836 union {
1837 __u32 relative_fd;
1838 __u32 relative_id;
1839 };
1840 __u64 expected_revision;
1841 } cgroup;
1842 };
1843 } link_create;
1844
1845 struct { /* struct used by BPF_LINK_UPDATE command */
1846 __u32 link_fd; /* link fd */
1847 union {
1848 /* new program fd to update link with */
1849 __u32 new_prog_fd;
1850 /* new struct_ops map fd to update link with */
1851 __u32 new_map_fd;
1852 };
1853 __u32 flags; /* extra flags */
1854 union {
1855 /* expected link's program fd; is specified only if
1856 * BPF_F_REPLACE flag is set in flags.
1857 */
1858 __u32 old_prog_fd;
1859 /* expected link's map fd; is specified only
1860 * if BPF_F_REPLACE flag is set.
1861 */
1862 __u32 old_map_fd;
1863 };
1864 } link_update;
1865
1866 struct {
1867 __u32 link_fd;
1868 } link_detach;
1869
1870 struct { /* struct used by BPF_ENABLE_STATS command */
1871 __u32 type;
1872 } enable_stats;
1873
1874 struct { /* struct used by BPF_ITER_CREATE command */
1875 __u32 link_fd;
1876 __u32 flags;
1877 } iter_create;
1878
1879 struct { /* struct used by BPF_PROG_BIND_MAP command */
1880 __u32 prog_fd;
1881 __u32 map_fd;
1882 __u32 flags; /* extra flags */
1883 } prog_bind_map;
1884
1885 struct { /* struct used by BPF_TOKEN_CREATE command */
1886 __u32 flags;
1887 __u32 bpffs_fd;
1888 } token_create;
1889
1890 struct {
1891 __aligned_u64 stream_buf;
1892 __u32 stream_buf_len;
1893 __u32 stream_id;
1894 __u32 prog_fd;
1895 } prog_stream_read;
1896
1897} __attribute__((aligned(8)));
1898
1899/* The description below is an attempt at providing documentation to eBPF
1900 * developers about the multiple available eBPF helper functions. It can be
1901 * parsed and used to produce a manual page. The workflow is the following,
1902 * and requires the rst2man utility:
1903 *
1904 * $ ./scripts/bpf_doc.py \
1905 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1906 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1907 * $ man /tmp/bpf-helpers.7
1908 *
1909 * Note that in order to produce this external documentation, some RST
1910 * formatting is used in the descriptions to get "bold" and "italics" in
1911 * manual pages. Also note that the few trailing white spaces are
1912 * intentional, removing them would break paragraphs for rst2man.
1913 *
1914 * Start of BPF helper function descriptions:
1915 *
1916 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1917 * Description
1918 * Perform a lookup in *map* for an entry associated to *key*.
1919 * Return
1920 * Map value associated to *key*, or **NULL** if no entry was
1921 * found.
1922 *
1923 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1924 * Description
1925 * Add or update the value of the entry associated to *key* in
1926 * *map* with *value*. *flags* is one of:
1927 *
1928 * **BPF_NOEXIST**
1929 * The entry for *key* must not exist in the map.
1930 * **BPF_EXIST**
1931 * The entry for *key* must already exist in the map.
1932 * **BPF_ANY**
1933 * No condition on the existence of the entry for *key*.
1934 *
1935 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1936 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1937 * elements always exist), the helper would return an error.
1938 * Return
1939 * 0 on success, or a negative error in case of failure.
1940 *
1941 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1942 * Description
1943 * Delete entry with *key* from *map*.
1944 * Return
1945 * 0 on success, or a negative error in case of failure.
1946 *
1947 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1948 * Description
1949 * For tracing programs, safely attempt to read *size* bytes from
1950 * kernel space address *unsafe_ptr* and store the data in *dst*.
1951 *
1952 * Generally, use **bpf_probe_read_user**\ () or
1953 * **bpf_probe_read_kernel**\ () instead.
1954 * Return
1955 * 0 on success, or a negative error in case of failure.
1956 *
1957 * u64 bpf_ktime_get_ns(void)
1958 * Description
1959 * Return the time elapsed since system boot, in nanoseconds.
1960 * Does not include time the system was suspended.
1961 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1962 * Return
1963 * Current *ktime*.
1964 *
1965 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1966 * Description
1967 * This helper is a "printk()-like" facility for debugging. It
1968 * prints a message defined by format *fmt* (of size *fmt_size*)
1969 * to file *\/sys/kernel/tracing/trace* from TraceFS, if
1970 * available. It can take up to three additional **u64**
1971 * arguments (as an eBPF helpers, the total number of arguments is
1972 * limited to five).
1973 *
1974 * Each time the helper is called, it appends a line to the trace.
1975 * Lines are discarded while *\/sys/kernel/tracing/trace* is
1976 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
1977 * The format of the trace is customizable, and the exact output
1978 * one will get depends on the options set in
1979 * *\/sys/kernel/tracing/trace_options* (see also the
1980 * *README* file under the same directory). However, it usually
1981 * defaults to something like:
1982 *
1983 * ::
1984 *
1985 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1986 *
1987 * In the above:
1988 *
1989 * * ``telnet`` is the name of the current task.
1990 * * ``470`` is the PID of the current task.
1991 * * ``001`` is the CPU number on which the task is
1992 * running.
1993 * * In ``.N..``, each character refers to a set of
1994 * options (whether irqs are enabled, scheduling
1995 * options, whether hard/softirqs are running, level of
1996 * preempt_disabled respectively). **N** means that
1997 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1998 * are set.
1999 * * ``419421.045894`` is a timestamp.
2000 * * ``0x00000001`` is a fake value used by BPF for the
2001 * instruction pointer register.
2002 * * ``<formatted msg>`` is the message formatted with
2003 * *fmt*.
2004 *
2005 * The conversion specifiers supported by *fmt* are similar, but
2006 * more limited than for printk(). They are **%d**, **%i**,
2007 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
2008 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
2009 * of field, padding with zeroes, etc.) is available, and the
2010 * helper will return **-EINVAL** (but print nothing) if it
2011 * encounters an unknown specifier.
2012 *
2013 * Also, note that **bpf_trace_printk**\ () is slow, and should
2014 * only be used for debugging purposes. For this reason, a notice
2015 * block (spanning several lines) is printed to kernel logs and
2016 * states that the helper should not be used "for production use"
2017 * the first time this helper is used (or more precisely, when
2018 * **trace_printk**\ () buffers are allocated). For passing values
2019 * to user space, perf events should be preferred.
2020 * Return
2021 * The number of bytes written to the buffer, or a negative error
2022 * in case of failure.
2023 *
2024 * u32 bpf_get_prandom_u32(void)
2025 * Description
2026 * Get a pseudo-random number.
2027 *
2028 * From a security point of view, this helper uses its own
2029 * pseudo-random internal state, and cannot be used to infer the
2030 * seed of other random functions in the kernel. However, it is
2031 * essential to note that the generator used by the helper is not
2032 * cryptographically secure.
2033 * Return
2034 * A random 32-bit unsigned value.
2035 *
2036 * u32 bpf_get_smp_processor_id(void)
2037 * Description
2038 * Get the SMP (symmetric multiprocessing) processor id. Note that
2039 * all programs run with migration disabled, which means that the
2040 * SMP processor id is stable during all the execution of the
2041 * program.
2042 * Return
2043 * The SMP id of the processor running the program.
2044 * Attributes
2045 * __bpf_fastcall
2046 *
2047 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
2048 * Description
2049 * Store *len* bytes from address *from* into the packet
2050 * associated to *skb*, at *offset*. The *flags* are a combination
2051 * of the following values:
2052 *
2053 * **BPF_F_RECOMPUTE_CSUM**
2054 * Automatically update *skb*\ **->csum** after storing the
2055 * bytes.
2056 * **BPF_F_INVALIDATE_HASH**
2057 * Set *skb*\ **->hash**, *skb*\ **->swhash** and *skb*\
2058 * **->l4hash** to 0.
2059 *
2060 * A call to this helper is susceptible to change the underlying
2061 * packet buffer. Therefore, at load time, all checks on pointers
2062 * previously done by the verifier are invalidated and must be
2063 * performed again, if the helper is used in combination with
2064 * direct packet access.
2065 * Return
2066 * 0 on success, or a negative error in case of failure.
2067 *
2068 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
2069 * Description
2070 * Recompute the layer 3 (e.g. IP) checksum for the packet
2071 * associated to *skb*. Computation is incremental, so the helper
2072 * must know the former value of the header field that was
2073 * modified (*from*), the new value of this field (*to*), and the
2074 * number of bytes (2 or 4) for this field, stored in *size*.
2075 * Alternatively, it is possible to store the difference between
2076 * the previous and the new values of the header field in *to*, by
2077 * setting *from* and *size* to 0. For both methods, *offset*
2078 * indicates the location of the IP checksum within the packet.
2079 *
2080 * This helper works in combination with **bpf_csum_diff**\ (),
2081 * which does not update the checksum in-place, but offers more
2082 * flexibility and can handle sizes larger than 2 or 4 for the
2083 * checksum to update.
2084 *
2085 * A call to this helper is susceptible to change the underlying
2086 * packet buffer. Therefore, at load time, all checks on pointers
2087 * previously done by the verifier are invalidated and must be
2088 * performed again, if the helper is used in combination with
2089 * direct packet access.
2090 * Return
2091 * 0 on success, or a negative error in case of failure.
2092 *
2093 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
2094 * Description
2095 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
2096 * packet associated to *skb*. Computation is incremental, so the
2097 * helper must know the former value of the header field that was
2098 * modified (*from*), the new value of this field (*to*), and the
2099 * number of bytes (2 or 4) for this field, stored on the lowest
2100 * four bits of *flags*. Alternatively, it is possible to store
2101 * the difference between the previous and the new values of the
2102 * header field in *to*, by setting *from* and the four lowest
2103 * bits of *flags* to 0. For both methods, *offset* indicates the
2104 * location of the IP checksum within the packet. In addition to
2105 * the size of the field, *flags* can be added (bitwise OR) actual
2106 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
2107 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
2108 * for updates resulting in a null checksum the value is set to
2109 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
2110 * that the modified header field is part of the pseudo-header.
2111 * Flag **BPF_F_IPV6** should be set for IPv6 packets.
2112 *
2113 * This helper works in combination with **bpf_csum_diff**\ (),
2114 * which does not update the checksum in-place, but offers more
2115 * flexibility and can handle sizes larger than 2 or 4 for the
2116 * checksum to update.
2117 *
2118 * A call to this helper is susceptible to change the underlying
2119 * packet buffer. Therefore, at load time, all checks on pointers
2120 * previously done by the verifier are invalidated and must be
2121 * performed again, if the helper is used in combination with
2122 * direct packet access.
2123 * Return
2124 * 0 on success, or a negative error in case of failure.
2125 *
2126 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
2127 * Description
2128 * This special helper is used to trigger a "tail call", or in
2129 * other words, to jump into another eBPF program. The same stack
2130 * frame is used (but values on stack and in registers for the
2131 * caller are not accessible to the callee). This mechanism allows
2132 * for program chaining, either for raising the maximum number of
2133 * available eBPF instructions, or to execute given programs in
2134 * conditional blocks. For security reasons, there is an upper
2135 * limit to the number of successive tail calls that can be
2136 * performed.
2137 *
2138 * Upon call of this helper, the program attempts to jump into a
2139 * program referenced at index *index* in *prog_array_map*, a
2140 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
2141 * *ctx*, a pointer to the context.
2142 *
2143 * If the call succeeds, the kernel immediately runs the first
2144 * instruction of the new program. This is not a function call,
2145 * and it never returns to the previous program. If the call
2146 * fails, then the helper has no effect, and the caller continues
2147 * to run its subsequent instructions. A call can fail if the
2148 * destination program for the jump does not exist (i.e. *index*
2149 * is superior to the number of entries in *prog_array_map*), or
2150 * if the maximum number of tail calls has been reached for this
2151 * chain of programs. This limit is defined in the kernel by the
2152 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
2153 * which is currently set to 33.
2154 * Return
2155 * 0 on success, or a negative error in case of failure.
2156 *
2157 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
2158 * Description
2159 * Clone and redirect the packet associated to *skb* to another
2160 * net device of index *ifindex*. Both ingress and egress
2161 * interfaces can be used for redirection. The **BPF_F_INGRESS**
2162 * value in *flags* is used to make the distinction (ingress path
2163 * is selected if the flag is present, egress path otherwise).
2164 * This is the only flag supported for now.
2165 *
2166 * In comparison with **bpf_redirect**\ () helper,
2167 * **bpf_clone_redirect**\ () has the associated cost of
2168 * duplicating the packet buffer, but this can be executed out of
2169 * the eBPF program. Conversely, **bpf_redirect**\ () is more
2170 * efficient, but it is handled through an action code where the
2171 * redirection happens only after the eBPF program has returned.
2172 *
2173 * A call to this helper is susceptible to change the underlying
2174 * packet buffer. Therefore, at load time, all checks on pointers
2175 * previously done by the verifier are invalidated and must be
2176 * performed again, if the helper is used in combination with
2177 * direct packet access.
2178 * Return
2179 * 0 on success, or a negative error in case of failure. Positive
2180 * error indicates a potential drop or congestion in the target
2181 * device. The particular positive error codes are not defined.
2182 *
2183 * u64 bpf_get_current_pid_tgid(void)
2184 * Description
2185 * Get the current pid and tgid.
2186 * Return
2187 * A 64-bit integer containing the current tgid and pid, and
2188 * created as such:
2189 * *current_task*\ **->tgid << 32 \|**
2190 * *current_task*\ **->pid**.
2191 *
2192 * u64 bpf_get_current_uid_gid(void)
2193 * Description
2194 * Get the current uid and gid.
2195 * Return
2196 * A 64-bit integer containing the current GID and UID, and
2197 * created as such: *current_gid* **<< 32 \|** *current_uid*.
2198 *
2199 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
2200 * Description
2201 * Copy the **comm** attribute of the current task into *buf* of
2202 * *size_of_buf*. The **comm** attribute contains the name of
2203 * the executable (excluding the path) for the current task. The
2204 * *size_of_buf* must be strictly positive. On success, the
2205 * helper makes sure that the *buf* is NUL-terminated. On failure,
2206 * it is filled with zeroes.
2207 * Return
2208 * 0 on success, or a negative error in case of failure.
2209 *
2210 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
2211 * Description
2212 * Retrieve the classid for the current task, i.e. for the net_cls
2213 * cgroup to which *skb* belongs.
2214 *
2215 * This helper can be used on TC egress path, but not on ingress.
2216 *
2217 * The net_cls cgroup provides an interface to tag network packets
2218 * based on a user-provided identifier for all traffic coming from
2219 * the tasks belonging to the related cgroup. See also the related
2220 * kernel documentation, available from the Linux sources in file
2221 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2222 *
2223 * The Linux kernel has two versions for cgroups: there are
2224 * cgroups v1 and cgroups v2. Both are available to users, who can
2225 * use a mixture of them, but note that the net_cls cgroup is for
2226 * cgroup v1 only. This makes it incompatible with BPF programs
2227 * run on cgroups, which is a cgroup-v2-only feature (a socket can
2228 * only hold data for one version of cgroups at a time).
2229 *
2230 * This helper is only available is the kernel was compiled with
2231 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2232 * "**y**" or to "**m**".
2233 * Return
2234 * The classid, or 0 for the default unconfigured classid.
2235 *
2236 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2237 * Description
2238 * Push a *vlan_tci* (VLAN tag control information) of protocol
2239 * *vlan_proto* to the packet associated to *skb*, then update
2240 * the checksum. Note that if *vlan_proto* is different from
2241 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2242 * be **ETH_P_8021Q**.
2243 *
2244 * A call to this helper is susceptible to change the underlying
2245 * packet buffer. Therefore, at load time, all checks on pointers
2246 * previously done by the verifier are invalidated and must be
2247 * performed again, if the helper is used in combination with
2248 * direct packet access.
2249 * Return
2250 * 0 on success, or a negative error in case of failure.
2251 *
2252 * long bpf_skb_vlan_pop(struct sk_buff *skb)
2253 * Description
2254 * Pop a VLAN header from the packet associated to *skb*.
2255 *
2256 * A call to this helper is susceptible to change the underlying
2257 * packet buffer. Therefore, at load time, all checks on pointers
2258 * previously done by the verifier are invalidated and must be
2259 * performed again, if the helper is used in combination with
2260 * direct packet access.
2261 * Return
2262 * 0 on success, or a negative error in case of failure.
2263 *
2264 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2265 * Description
2266 * Get tunnel metadata. This helper takes a pointer *key* to an
2267 * empty **struct bpf_tunnel_key** of **size**, that will be
2268 * filled with tunnel metadata for the packet associated to *skb*.
2269 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2270 * indicates that the tunnel is based on IPv6 protocol instead of
2271 * IPv4.
2272 *
2273 * The **struct bpf_tunnel_key** is an object that generalizes the
2274 * principal parameters used by various tunneling protocols into a
2275 * single struct. This way, it can be used to easily make a
2276 * decision based on the contents of the encapsulation header,
2277 * "summarized" in this struct. In particular, it holds the IP
2278 * address of the remote end (IPv4 or IPv6, depending on the case)
2279 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2280 * this struct exposes the *key*\ **->tunnel_id**, which is
2281 * generally mapped to a VNI (Virtual Network Identifier), making
2282 * it programmable together with the **bpf_skb_set_tunnel_key**\
2283 * () helper.
2284 *
2285 * Let's imagine that the following code is part of a program
2286 * attached to the TC ingress interface, on one end of a GRE
2287 * tunnel, and is supposed to filter out all messages coming from
2288 * remote ends with IPv4 address other than 10.0.0.1:
2289 *
2290 * ::
2291 *
2292 * int ret;
2293 * struct bpf_tunnel_key key = {};
2294 *
2295 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2296 * if (ret < 0)
2297 * return TC_ACT_SHOT; // drop packet
2298 *
2299 * if (key.remote_ipv4 != 0x0a000001)
2300 * return TC_ACT_SHOT; // drop packet
2301 *
2302 * return TC_ACT_OK; // accept packet
2303 *
2304 * This interface can also be used with all encapsulation devices
2305 * that can operate in "collect metadata" mode: instead of having
2306 * one network device per specific configuration, the "collect
2307 * metadata" mode only requires a single device where the
2308 * configuration can be extracted from this helper.
2309 *
2310 * This can be used together with various tunnels such as VXLan,
2311 * Geneve, GRE or IP in IP (IPIP).
2312 * Return
2313 * 0 on success, or a negative error in case of failure.
2314 *
2315 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2316 * Description
2317 * Populate tunnel metadata for packet associated to *skb.* The
2318 * tunnel metadata is set to the contents of *key*, of *size*. The
2319 * *flags* can be set to a combination of the following values:
2320 *
2321 * **BPF_F_TUNINFO_IPV6**
2322 * Indicate that the tunnel is based on IPv6 protocol
2323 * instead of IPv4.
2324 * **BPF_F_ZERO_CSUM_TX**
2325 * For IPv4 packets, add a flag to tunnel metadata
2326 * indicating that checksum computation should be skipped
2327 * and checksum set to zeroes.
2328 * **BPF_F_DONT_FRAGMENT**
2329 * Add a flag to tunnel metadata indicating that the
2330 * packet should not be fragmented.
2331 * **BPF_F_SEQ_NUMBER**
2332 * Add a flag to tunnel metadata indicating that a
2333 * sequence number should be added to tunnel header before
2334 * sending the packet. This flag was added for GRE
2335 * encapsulation, but might be used with other protocols
2336 * as well in the future.
2337 * **BPF_F_NO_TUNNEL_KEY**
2338 * Add a flag to tunnel metadata indicating that no tunnel
2339 * key should be set in the resulting tunnel header.
2340 *
2341 * Here is a typical usage on the transmit path:
2342 *
2343 * ::
2344 *
2345 * struct bpf_tunnel_key key;
2346 * populate key ...
2347 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2348 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2349 *
2350 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
2351 * helper for additional information.
2352 * Return
2353 * 0 on success, or a negative error in case of failure.
2354 *
2355 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2356 * Description
2357 * Read the value of a perf event counter. This helper relies on a
2358 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2359 * the perf event counter is selected when *map* is updated with
2360 * perf event file descriptors. The *map* is an array whose size
2361 * is the number of available CPUs, and each cell contains a value
2362 * relative to one CPU. The value to retrieve is indicated by
2363 * *flags*, that contains the index of the CPU to look up, masked
2364 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2365 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2366 * current CPU should be retrieved.
2367 *
2368 * Note that before Linux 4.13, only hardware perf event can be
2369 * retrieved.
2370 *
2371 * Also, be aware that the newer helper
2372 * **bpf_perf_event_read_value**\ () is recommended over
2373 * **bpf_perf_event_read**\ () in general. The latter has some ABI
2374 * quirks where error and counter value are used as a return code
2375 * (which is wrong to do since ranges may overlap). This issue is
2376 * fixed with **bpf_perf_event_read_value**\ (), which at the same
2377 * time provides more features over the **bpf_perf_event_read**\
2378 * () interface. Please refer to the description of
2379 * **bpf_perf_event_read_value**\ () for details.
2380 * Return
2381 * The value of the perf event counter read from the map, or a
2382 * negative error code in case of failure.
2383 *
2384 * long bpf_redirect(u32 ifindex, u64 flags)
2385 * Description
2386 * Redirect the packet to another net device of index *ifindex*.
2387 * This helper is somewhat similar to **bpf_clone_redirect**\
2388 * (), except that the packet is not cloned, which provides
2389 * increased performance.
2390 *
2391 * Except for XDP, both ingress and egress interfaces can be used
2392 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
2393 * to make the distinction (ingress path is selected if the flag
2394 * is present, egress path otherwise). Currently, XDP only
2395 * supports redirection to the egress interface, and accepts no
2396 * flag at all.
2397 *
2398 * The same effect can also be attained with the more generic
2399 * **bpf_redirect_map**\ (), which uses a BPF map to store the
2400 * redirect target instead of providing it directly to the helper.
2401 * Return
2402 * For XDP, the helper returns **XDP_REDIRECT** on success or
2403 * **XDP_ABORTED** on error. For other program types, the values
2404 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2405 * error.
2406 *
2407 * u32 bpf_get_route_realm(struct sk_buff *skb)
2408 * Description
2409 * Retrieve the realm or the route, that is to say the
2410 * **tclassid** field of the destination for the *skb*. The
2411 * identifier retrieved is a user-provided tag, similar to the
2412 * one used with the net_cls cgroup (see description for
2413 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
2414 * held by a route (a destination entry), not by a task.
2415 *
2416 * Retrieving this identifier works with the clsact TC egress hook
2417 * (see also **tc-bpf(8)**), or alternatively on conventional
2418 * classful egress qdiscs, but not on TC ingress path. In case of
2419 * clsact TC egress hook, this has the advantage that, internally,
2420 * the destination entry has not been dropped yet in the transmit
2421 * path. Therefore, the destination entry does not need to be
2422 * artificially held via **netif_keep_dst**\ () for a classful
2423 * qdisc until the *skb* is freed.
2424 *
2425 * This helper is available only if the kernel was compiled with
2426 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2427 * Return
2428 * The realm of the route for the packet associated to *skb*, or 0
2429 * if none was found.
2430 *
2431 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2432 * Description
2433 * Write raw *data* blob into a special BPF perf event held by
2434 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2435 * event must have the following attributes: **PERF_SAMPLE_RAW**
2436 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2437 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2438 *
2439 * The *flags* are used to indicate the index in *map* for which
2440 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2441 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2442 * to indicate that the index of the current CPU core should be
2443 * used.
2444 *
2445 * The value to write, of *size*, is passed through eBPF stack and
2446 * pointed by *data*.
2447 *
2448 * The context of the program *ctx* needs also be passed to the
2449 * helper.
2450 *
2451 * On user space, a program willing to read the values needs to
2452 * call **perf_event_open**\ () on the perf event (either for
2453 * one or for all CPUs) and to store the file descriptor into the
2454 * *map*. This must be done before the eBPF program can send data
2455 * into it. An example is available in file
2456 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2457 * tree (the eBPF program counterpart is in
2458 * *samples/bpf/trace_output.bpf.c*).
2459 *
2460 * **bpf_perf_event_output**\ () achieves better performance
2461 * than **bpf_trace_printk**\ () for sharing data with user
2462 * space, and is much better suitable for streaming data from eBPF
2463 * programs.
2464 *
2465 * Note that this helper is not restricted to tracing use cases
2466 * and can be used with programs attached to TC or XDP as well,
2467 * where it allows for passing data to user space listeners. Data
2468 * can be:
2469 *
2470 * * Only custom structs,
2471 * * Only the packet payload, or
2472 * * A combination of both.
2473 * Return
2474 * 0 on success, or a negative error in case of failure.
2475 *
2476 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2477 * Description
2478 * This helper was provided as an easy way to load data from a
2479 * packet. It can be used to load *len* bytes from *offset* from
2480 * the packet associated to *skb*, into the buffer pointed by
2481 * *to*.
2482 *
2483 * Since Linux 4.7, usage of this helper has mostly been replaced
2484 * by "direct packet access", enabling packet data to be
2485 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2486 * pointing respectively to the first byte of packet data and to
2487 * the byte after the last byte of packet data. However, it
2488 * remains useful if one wishes to read large quantities of data
2489 * at once from a packet into the eBPF stack.
2490 * Return
2491 * 0 on success, or a negative error in case of failure.
2492 *
2493 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2494 * Description
2495 * Walk a user or a kernel stack and return its id. To achieve
2496 * this, the helper needs *ctx*, which is a pointer to the context
2497 * on which the tracing program is executed, and a pointer to a
2498 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2499 *
2500 * The last argument, *flags*, holds the number of stack frames to
2501 * skip (from 0 to 255), masked with
2502 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2503 * a combination of the following flags:
2504 *
2505 * **BPF_F_USER_STACK**
2506 * Collect a user space stack instead of a kernel stack.
2507 * **BPF_F_FAST_STACK_CMP**
2508 * Compare stacks by hash only.
2509 * **BPF_F_REUSE_STACKID**
2510 * If two different stacks hash into the same *stackid*,
2511 * discard the old one.
2512 *
2513 * The stack id retrieved is a 32 bit long integer handle which
2514 * can be further combined with other data (including other stack
2515 * ids) and used as a key into maps. This can be useful for
2516 * generating a variety of graphs (such as flame graphs or off-cpu
2517 * graphs).
2518 *
2519 * For walking a stack, this helper is an improvement over
2520 * **bpf_probe_read**\ (), which can be used with unrolled loops
2521 * but is not efficient and consumes a lot of eBPF instructions.
2522 * Instead, **bpf_get_stackid**\ () can collect up to
2523 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2524 * this limit can be controlled with the **sysctl** program, and
2525 * that it should be manually increased in order to profile long
2526 * user stacks (such as stacks for Java programs). To do so, use:
2527 *
2528 * ::
2529 *
2530 * # sysctl kernel.perf_event_max_stack=<new value>
2531 * Return
2532 * The positive or null stack id on success, or a negative error
2533 * in case of failure.
2534 *
2535 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2536 * Description
2537 * Compute a checksum difference, from the raw buffer pointed by
2538 * *from*, of length *from_size* (that must be a multiple of 4),
2539 * towards the raw buffer pointed by *to*, of size *to_size*
2540 * (same remark). An optional *seed* can be added to the value
2541 * (this can be cascaded, the seed may come from a previous call
2542 * to the helper).
2543 *
2544 * This is flexible enough to be used in several ways:
2545 *
2546 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2547 * checksum, it can be used when pushing new data.
2548 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2549 * checksum, it can be used when removing data from a packet.
2550 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2551 * can be used to compute a diff. Note that *from_size* and
2552 * *to_size* do not need to be equal.
2553 *
2554 * This helper can be used in combination with
2555 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2556 * which one can feed in the difference computed with
2557 * **bpf_csum_diff**\ ().
2558 * Return
2559 * The checksum result, or a negative error code in case of
2560 * failure.
2561 *
2562 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2563 * Description
2564 * Retrieve tunnel options metadata for the packet associated to
2565 * *skb*, and store the raw tunnel option data to the buffer *opt*
2566 * of *size*.
2567 *
2568 * This helper can be used with encapsulation devices that can
2569 * operate in "collect metadata" mode (please refer to the related
2570 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2571 * more details). A particular example where this can be used is
2572 * in combination with the Geneve encapsulation protocol, where it
2573 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2574 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2575 * the eBPF program. This allows for full customization of these
2576 * headers.
2577 * Return
2578 * The size of the option data retrieved.
2579 *
2580 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2581 * Description
2582 * Set tunnel options metadata for the packet associated to *skb*
2583 * to the option data contained in the raw buffer *opt* of *size*.
2584 *
2585 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2586 * helper for additional information.
2587 * Return
2588 * 0 on success, or a negative error in case of failure.
2589 *
2590 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2591 * Description
2592 * Change the protocol of the *skb* to *proto*. Currently
2593 * supported are transition from IPv4 to IPv6, and from IPv6 to
2594 * IPv4. The helper takes care of the groundwork for the
2595 * transition, including resizing the socket buffer. The eBPF
2596 * program is expected to fill the new headers, if any, via
2597 * **skb_store_bytes**\ () and to recompute the checksums with
2598 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2599 * (). The main case for this helper is to perform NAT64
2600 * operations out of an eBPF program.
2601 *
2602 * Internally, the GSO type is marked as dodgy so that headers are
2603 * checked and segments are recalculated by the GSO/GRO engine.
2604 * The size for GSO target is adapted as well.
2605 *
2606 * All values for *flags* are reserved for future usage, and must
2607 * be left at zero.
2608 *
2609 * A call to this helper is susceptible to change the underlying
2610 * packet buffer. Therefore, at load time, all checks on pointers
2611 * previously done by the verifier are invalidated and must be
2612 * performed again, if the helper is used in combination with
2613 * direct packet access.
2614 * Return
2615 * 0 on success, or a negative error in case of failure.
2616 *
2617 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2618 * Description
2619 * Change the packet type for the packet associated to *skb*. This
2620 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2621 * the eBPF program does not have a write access to *skb*\
2622 * **->pkt_type** beside this helper. Using a helper here allows
2623 * for graceful handling of errors.
2624 *
2625 * The major use case is to change incoming *skb*s to
2626 * **PACKET_HOST** in a programmatic way instead of having to
2627 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2628 * example.
2629 *
2630 * Note that *type* only allows certain values. At this time, they
2631 * are:
2632 *
2633 * **PACKET_HOST**
2634 * Packet is for us.
2635 * **PACKET_BROADCAST**
2636 * Send packet to all.
2637 * **PACKET_MULTICAST**
2638 * Send packet to group.
2639 * **PACKET_OTHERHOST**
2640 * Send packet to someone else.
2641 * Return
2642 * 0 on success, or a negative error in case of failure.
2643 *
2644 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2645 * Description
2646 * Check whether *skb* is a descendant of the cgroup2 held by
2647 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2648 * Return
2649 * The return value depends on the result of the test, and can be:
2650 *
2651 * * 0, if the *skb* failed the cgroup2 descendant test.
2652 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2653 * * A negative error code, if an error occurred.
2654 *
2655 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2656 * Description
2657 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2658 * not set, in particular if the hash was cleared due to mangling,
2659 * recompute this hash. Later accesses to the hash can be done
2660 * directly with *skb*\ **->hash**.
2661 *
2662 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2663 * prototype with **bpf_skb_change_proto**\ (), or calling
2664 * **bpf_skb_store_bytes**\ () with the
2665 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2666 * the hash and to trigger a new computation for the next call to
2667 * **bpf_get_hash_recalc**\ ().
2668 * Return
2669 * The 32-bit hash.
2670 *
2671 * u64 bpf_get_current_task(void)
2672 * Description
2673 * Get the current task.
2674 * Return
2675 * A pointer to the current task struct.
2676 *
2677 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2678 * Description
2679 * Attempt in a safe way to write *len* bytes from the buffer
2680 * *src* to *dst* in memory. It only works for threads that are in
2681 * user context, and *dst* must be a valid user space address.
2682 *
2683 * This helper should not be used to implement any kind of
2684 * security mechanism because of TOC-TOU attacks, but rather to
2685 * debug, divert, and manipulate execution of semi-cooperative
2686 * processes.
2687 *
2688 * Keep in mind that this feature is meant for experiments, and it
2689 * has a risk of crashing the system and running programs.
2690 * Therefore, when an eBPF program using this helper is attached,
2691 * a warning including PID and process name is printed to kernel
2692 * logs.
2693 * Return
2694 * 0 on success, or a negative error in case of failure.
2695 *
2696 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2697 * Description
2698 * Check whether the probe is being run is the context of a given
2699 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2700 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2701 * Return
2702 * The return value depends on the result of the test, and can be:
2703 *
2704 * * 1, if current task belongs to the cgroup2.
2705 * * 0, if current task does not belong to the cgroup2.
2706 * * A negative error code, if an error occurred.
2707 *
2708 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2709 * Description
2710 * Resize (trim or grow) the packet associated to *skb* to the
2711 * new *len*. The *flags* are reserved for future usage, and must
2712 * be left at zero.
2713 *
2714 * The basic idea is that the helper performs the needed work to
2715 * change the size of the packet, then the eBPF program rewrites
2716 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2717 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2718 * and others. This helper is a slow path utility intended for
2719 * replies with control messages. And because it is targeted for
2720 * slow path, the helper itself can afford to be slow: it
2721 * implicitly linearizes, unclones and drops offloads from the
2722 * *skb*.
2723 *
2724 * A call to this helper is susceptible to change the underlying
2725 * packet buffer. Therefore, at load time, all checks on pointers
2726 * previously done by the verifier are invalidated and must be
2727 * performed again, if the helper is used in combination with
2728 * direct packet access.
2729 * Return
2730 * 0 on success, or a negative error in case of failure.
2731 *
2732 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2733 * Description
2734 * Pull in non-linear data in case the *skb* is non-linear and not
2735 * all of *len* are part of the linear section. Make *len* bytes
2736 * from *skb* readable and writable. If a zero value is passed for
2737 * *len*, then all bytes in the linear part of *skb* will be made
2738 * readable and writable.
2739 *
2740 * This helper is only needed for reading and writing with direct
2741 * packet access.
2742 *
2743 * For direct packet access, testing that offsets to access
2744 * are within packet boundaries (test on *skb*\ **->data_end**) is
2745 * susceptible to fail if offsets are invalid, or if the requested
2746 * data is in non-linear parts of the *skb*. On failure the
2747 * program can just bail out, or in the case of a non-linear
2748 * buffer, use a helper to make the data available. The
2749 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2750 * the data. Another one consists in using **bpf_skb_pull_data**
2751 * to pull in once the non-linear parts, then retesting and
2752 * eventually access the data.
2753 *
2754 * At the same time, this also makes sure the *skb* is uncloned,
2755 * which is a necessary condition for direct write. As this needs
2756 * to be an invariant for the write part only, the verifier
2757 * detects writes and adds a prologue that is calling
2758 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2759 * the very beginning in case it is indeed cloned.
2760 *
2761 * A call to this helper is susceptible to change the underlying
2762 * packet buffer. Therefore, at load time, all checks on pointers
2763 * previously done by the verifier are invalidated and must be
2764 * performed again, if the helper is used in combination with
2765 * direct packet access.
2766 * Return
2767 * 0 on success, or a negative error in case of failure.
2768 *
2769 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2770 * Description
2771 * Add the checksum *csum* into *skb*\ **->csum** in case the
2772 * driver has supplied a checksum for the entire packet into that
2773 * field. Return an error otherwise. This helper is intended to be
2774 * used in combination with **bpf_csum_diff**\ (), in particular
2775 * when the checksum needs to be updated after data has been
2776 * written into the packet through direct packet access.
2777 * Return
2778 * The checksum on success, or a negative error code in case of
2779 * failure.
2780 *
2781 * void bpf_set_hash_invalid(struct sk_buff *skb)
2782 * Description
2783 * Invalidate the current *skb*\ **->hash**. It can be used after
2784 * mangling on headers through direct packet access, in order to
2785 * indicate that the hash is outdated and to trigger a
2786 * recalculation the next time the kernel tries to access this
2787 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2788 * Return
2789 * void.
2790 *
2791 * long bpf_get_numa_node_id(void)
2792 * Description
2793 * Return the id of the current NUMA node. The primary use case
2794 * for this helper is the selection of sockets for the local NUMA
2795 * node, when the program is attached to sockets using the
2796 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2797 * but the helper is also available to other eBPF program types,
2798 * similarly to **bpf_get_smp_processor_id**\ ().
2799 * Return
2800 * The id of current NUMA node.
2801 *
2802 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2803 * Description
2804 * Grows headroom of packet associated to *skb* and adjusts the
2805 * offset of the MAC header accordingly, adding *len* bytes of
2806 * space. It automatically extends and reallocates memory as
2807 * required.
2808 *
2809 * This helper can be used on a layer 3 *skb* to push a MAC header
2810 * for redirection into a layer 2 device.
2811 *
2812 * All values for *flags* are reserved for future usage, and must
2813 * be left at zero.
2814 *
2815 * A call to this helper is susceptible to change the underlying
2816 * packet buffer. Therefore, at load time, all checks on pointers
2817 * previously done by the verifier are invalidated and must be
2818 * performed again, if the helper is used in combination with
2819 * direct packet access.
2820 * Return
2821 * 0 on success, or a negative error in case of failure.
2822 *
2823 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2824 * Description
2825 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2826 * it is possible to use a negative value for *delta*. This helper
2827 * can be used to prepare the packet for pushing or popping
2828 * headers.
2829 *
2830 * A call to this helper is susceptible to change the underlying
2831 * packet buffer. Therefore, at load time, all checks on pointers
2832 * previously done by the verifier are invalidated and must be
2833 * performed again, if the helper is used in combination with
2834 * direct packet access.
2835 * Return
2836 * 0 on success, or a negative error in case of failure.
2837 *
2838 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2839 * Description
2840 * Copy a NUL terminated string from an unsafe kernel address
2841 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2842 * more details.
2843 *
2844 * Generally, use **bpf_probe_read_user_str**\ () or
2845 * **bpf_probe_read_kernel_str**\ () instead.
2846 * Return
2847 * On success, the strictly positive length of the string,
2848 * including the trailing NUL character. On error, a negative
2849 * value.
2850 *
2851 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2852 * Description
2853 * If the **struct sk_buff** pointed by *skb* has a known socket,
2854 * retrieve the cookie (generated by the kernel) of this socket.
2855 * If no cookie has been set yet, generate a new cookie. Once
2856 * generated, the socket cookie remains stable for the life of the
2857 * socket. This helper can be useful for monitoring per socket
2858 * networking traffic statistics as it provides a global socket
2859 * identifier that can be assumed unique.
2860 * Return
2861 * A 8-byte long unique number on success, or 0 if the socket
2862 * field is missing inside *skb*.
2863 *
2864 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2865 * Description
2866 * Equivalent to bpf_get_socket_cookie() helper that accepts
2867 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2868 * Return
2869 * A 8-byte long unique number.
2870 *
2871 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2872 * Description
2873 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2874 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2875 * Return
2876 * A 8-byte long unique number.
2877 *
2878 * u64 bpf_get_socket_cookie(struct sock *sk)
2879 * Description
2880 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2881 * *sk*, but gets socket from a BTF **struct sock**. This helper
2882 * also works for sleepable programs.
2883 * Return
2884 * A 8-byte long unique number or 0 if *sk* is NULL.
2885 *
2886 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2887 * Description
2888 * Get the owner UID of the socked associated to *skb*.
2889 * Return
2890 * The owner UID of the socket associated to *skb*. If the socket
2891 * is **NULL**, or if it is not a full socket (i.e. if it is a
2892 * time-wait or a request socket instead), **overflowuid** value
2893 * is returned (note that **overflowuid** might also be the actual
2894 * UID value for the socket).
2895 *
2896 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2897 * Description
2898 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2899 * to value *hash*.
2900 * Return
2901 * 0
2902 *
2903 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2904 * Description
2905 * Emulate a call to **setsockopt()** on the socket associated to
2906 * *bpf_socket*, which must be a full socket. The *level* at
2907 * which the option resides and the name *optname* of the option
2908 * must be specified, see **setsockopt(2)** for more information.
2909 * The option value of length *optlen* is pointed by *optval*.
2910 *
2911 * *bpf_socket* should be one of the following:
2912 *
2913 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2914 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
2915 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
2916 *
2917 * This helper actually implements a subset of **setsockopt()**.
2918 * It supports the following *level*\ s:
2919 *
2920 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2921 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2922 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2923 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2924 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2925 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2926 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2927 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2928 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2929 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2930 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2931 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2932 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.
2933 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2934 * * **IPPROTO_IPV6**, which supports the following *optname*\ s:
2935 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2936 * Return
2937 * 0 on success, or a negative error in case of failure.
2938 *
2939 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2940 * Description
2941 * Grow or shrink the room for data in the packet associated to
2942 * *skb* by *len_diff*, and according to the selected *mode*.
2943 *
2944 * By default, the helper will reset any offloaded checksum
2945 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2946 * by the following flag:
2947 *
2948 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2949 * checksum data of the skb to CHECKSUM_NONE.
2950 *
2951 * There are two supported modes at this time:
2952 *
2953 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2954 * (room space is added or removed between the layer 2 and
2955 * layer 3 headers).
2956 *
2957 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2958 * (room space is added or removed between the layer 3 and
2959 * layer 4 headers).
2960 *
2961 * The following flags are supported at this time:
2962 *
2963 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2964 * Adjusting mss in this way is not allowed for datagrams.
2965 *
2966 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2967 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2968 * Any new space is reserved to hold a tunnel header.
2969 * Configure skb offsets and other fields accordingly.
2970 *
2971 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2972 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2973 * Use with ENCAP_L3 flags to further specify the tunnel type.
2974 *
2975 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2976 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2977 * type; *len* is the length of the inner MAC header.
2978 *
2979 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2980 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2981 * L2 type as Ethernet.
2982 *
2983 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
2984 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
2985 * Indicate the new IP header version after decapsulating the outer
2986 * IP header. Used when the inner and outer IP versions are different.
2987 *
2988 * A call to this helper is susceptible to change the underlying
2989 * packet buffer. Therefore, at load time, all checks on pointers
2990 * previously done by the verifier are invalidated and must be
2991 * performed again, if the helper is used in combination with
2992 * direct packet access.
2993 * Return
2994 * 0 on success, or a negative error in case of failure.
2995 *
2996 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2997 * Description
2998 * Redirect the packet to the endpoint referenced by *map* at
2999 * index *key*. Depending on its type, this *map* can contain
3000 * references to net devices (for forwarding packets through other
3001 * ports), or to CPUs (for redirecting XDP frames to another CPU;
3002 * but this is only implemented for native XDP (with driver
3003 * support) as of this writing).
3004 *
3005 * The lower two bits of *flags* are used as the return code if
3006 * the map lookup fails. This is so that the return value can be
3007 * one of the XDP program return codes up to **XDP_TX**, as chosen
3008 * by the caller. The higher bits of *flags* can be set to
3009 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
3010 *
3011 * With BPF_F_BROADCAST the packet will be broadcasted to all the
3012 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
3013 * interface will be excluded when do broadcasting.
3014 *
3015 * See also **bpf_redirect**\ (), which only supports redirecting
3016 * to an ifindex, but doesn't require a map to do so.
3017 * Return
3018 * **XDP_REDIRECT** on success, or the value of the two lower bits
3019 * of the *flags* argument on error.
3020 *
3021 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
3022 * Description
3023 * Redirect the packet to the socket referenced by *map* (of type
3024 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3025 * egress interfaces can be used for redirection. The
3026 * **BPF_F_INGRESS** value in *flags* is used to make the
3027 * distinction (ingress path is selected if the flag is present,
3028 * egress path otherwise). This is the only flag supported for now.
3029 * Return
3030 * **SK_PASS** on success, or **SK_DROP** on error.
3031 *
3032 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3033 * Description
3034 * Add an entry to, or update a *map* referencing sockets. The
3035 * *skops* is used as a new value for the entry associated to
3036 * *key*. *flags* is one of:
3037 *
3038 * **BPF_NOEXIST**
3039 * The entry for *key* must not exist in the map.
3040 * **BPF_EXIST**
3041 * The entry for *key* must already exist in the map.
3042 * **BPF_ANY**
3043 * No condition on the existence of the entry for *key*.
3044 *
3045 * If the *map* has eBPF programs (parser and verdict), those will
3046 * be inherited by the socket being added. If the socket is
3047 * already attached to eBPF programs, this results in an error.
3048 * Return
3049 * 0 on success, or a negative error in case of failure.
3050 *
3051 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
3052 * Description
3053 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
3054 * *delta* (which can be positive or negative). Note that this
3055 * operation modifies the address stored in *xdp_md*\ **->data**,
3056 * so the latter must be loaded only after the helper has been
3057 * called.
3058 *
3059 * The use of *xdp_md*\ **->data_meta** is optional and programs
3060 * are not required to use it. The rationale is that when the
3061 * packet is processed with XDP (e.g. as DoS filter), it is
3062 * possible to push further meta data along with it before passing
3063 * to the stack, and to give the guarantee that an ingress eBPF
3064 * program attached as a TC classifier on the same device can pick
3065 * this up for further post-processing. Since TC works with socket
3066 * buffers, it remains possible to set from XDP the **mark** or
3067 * **priority** pointers, or other pointers for the socket buffer.
3068 * Having this scratch space generic and programmable allows for
3069 * more flexibility as the user is free to store whatever meta
3070 * data they need.
3071 *
3072 * A call to this helper is susceptible to change the underlying
3073 * packet buffer. Therefore, at load time, all checks on pointers
3074 * previously done by the verifier are invalidated and must be
3075 * performed again, if the helper is used in combination with
3076 * direct packet access.
3077 * Return
3078 * 0 on success, or a negative error in case of failure.
3079 *
3080 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
3081 * Description
3082 * Read the value of a perf event counter, and store it into *buf*
3083 * of size *buf_size*. This helper relies on a *map* of type
3084 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
3085 * counter is selected when *map* is updated with perf event file
3086 * descriptors. The *map* is an array whose size is the number of
3087 * available CPUs, and each cell contains a value relative to one
3088 * CPU. The value to retrieve is indicated by *flags*, that
3089 * contains the index of the CPU to look up, masked with
3090 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
3091 * **BPF_F_CURRENT_CPU** to indicate that the value for the
3092 * current CPU should be retrieved.
3093 *
3094 * This helper behaves in a way close to
3095 * **bpf_perf_event_read**\ () helper, save that instead of
3096 * just returning the value observed, it fills the *buf*
3097 * structure. This allows for additional data to be retrieved: in
3098 * particular, the enabled and running times (in *buf*\
3099 * **->enabled** and *buf*\ **->running**, respectively) are
3100 * copied. In general, **bpf_perf_event_read_value**\ () is
3101 * recommended over **bpf_perf_event_read**\ (), which has some
3102 * ABI issues and provides fewer functionalities.
3103 *
3104 * These values are interesting, because hardware PMU (Performance
3105 * Monitoring Unit) counters are limited resources. When there are
3106 * more PMU based perf events opened than available counters,
3107 * kernel will multiplex these events so each event gets certain
3108 * percentage (but not all) of the PMU time. In case that
3109 * multiplexing happens, the number of samples or counter value
3110 * will not reflect the case compared to when no multiplexing
3111 * occurs. This makes comparison between different runs difficult.
3112 * Typically, the counter value should be normalized before
3113 * comparing to other experiments. The usual normalization is done
3114 * as follows.
3115 *
3116 * ::
3117 *
3118 * normalized_counter = counter * t_enabled / t_running
3119 *
3120 * Where t_enabled is the time enabled for event and t_running is
3121 * the time running for event since last normalization. The
3122 * enabled and running times are accumulated since the perf event
3123 * open. To achieve scaling factor between two invocations of an
3124 * eBPF program, users can use CPU id as the key (which is
3125 * typical for perf array usage model) to remember the previous
3126 * value and do the calculation inside the eBPF program.
3127 * Return
3128 * 0 on success, or a negative error in case of failure.
3129 *
3130 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
3131 * Description
3132 * For an eBPF program attached to a perf event, retrieve the
3133 * value of the event counter associated to *ctx* and store it in
3134 * the structure pointed by *buf* and of size *buf_size*. Enabled
3135 * and running times are also stored in the structure (see
3136 * description of helper **bpf_perf_event_read_value**\ () for
3137 * more details).
3138 * Return
3139 * 0 on success, or a negative error in case of failure.
3140 *
3141 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
3142 * Description
3143 * Emulate a call to **getsockopt()** on the socket associated to
3144 * *bpf_socket*, which must be a full socket. The *level* at
3145 * which the option resides and the name *optname* of the option
3146 * must be specified, see **getsockopt(2)** for more information.
3147 * The retrieved value is stored in the structure pointed by
3148 * *opval* and of length *optlen*.
3149 *
3150 * *bpf_socket* should be one of the following:
3151 *
3152 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
3153 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
3154 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
3155 *
3156 * This helper actually implements a subset of **getsockopt()**.
3157 * It supports the same set of *optname*\ s that is supported by
3158 * the **bpf_setsockopt**\ () helper. The exceptions are
3159 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and
3160 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
3161 * Return
3162 * 0 on success, or a negative error in case of failure.
3163 *
3164 * long bpf_override_return(struct pt_regs *regs, u64 rc)
3165 * Description
3166 * Used for error injection, this helper uses kprobes to override
3167 * the return value of the probed function, and to set it to *rc*.
3168 * The first argument is the context *regs* on which the kprobe
3169 * works.
3170 *
3171 * This helper works by setting the PC (program counter)
3172 * to an override function which is run in place of the original
3173 * probed function. This means the probed function is not run at
3174 * all. The replacement function just returns with the required
3175 * value.
3176 *
3177 * This helper has security implications, and thus is subject to
3178 * restrictions. It is only available if the kernel was compiled
3179 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
3180 * option, and in this case it only works on functions tagged with
3181 * **ALLOW_ERROR_INJECTION** in the kernel code.
3182 * Return
3183 * 0
3184 *
3185 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
3186 * Description
3187 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
3188 * for the full TCP socket associated to *bpf_sock_ops* to
3189 * *argval*.
3190 *
3191 * The primary use of this field is to determine if there should
3192 * be calls to eBPF programs of type
3193 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
3194 * code. A program of the same type can change its value, per
3195 * connection and as necessary, when the connection is
3196 * established. This field is directly accessible for reading, but
3197 * this helper must be used for updates in order to return an
3198 * error if an eBPF program tries to set a callback that is not
3199 * supported in the current kernel.
3200 *
3201 * *argval* is a flag array which can combine these flags:
3202 *
3203 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
3204 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
3205 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
3206 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
3207 *
3208 * Therefore, this function can be used to clear a callback flag by
3209 * setting the appropriate bit to zero. e.g. to disable the RTO
3210 * callback:
3211 *
3212 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
3213 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3214 *
3215 * Here are some examples of where one could call such eBPF
3216 * program:
3217 *
3218 * * When RTO fires.
3219 * * When a packet is retransmitted.
3220 * * When the connection terminates.
3221 * * When a packet is sent.
3222 * * When a packet is received.
3223 * Return
3224 * Code **-EINVAL** if the socket is not a full TCP socket;
3225 * otherwise, a positive number containing the bits that could not
3226 * be set is returned (which comes down to 0 if all bits were set
3227 * as required).
3228 *
3229 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3230 * Description
3231 * This helper is used in programs implementing policies at the
3232 * socket level. If the message *msg* is allowed to pass (i.e. if
3233 * the verdict eBPF program returns **SK_PASS**), redirect it to
3234 * the socket referenced by *map* (of type
3235 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3236 * egress interfaces can be used for redirection. The
3237 * **BPF_F_INGRESS** value in *flags* is used to make the
3238 * distinction (ingress path is selected if the flag is present,
3239 * egress path otherwise). This is the only flag supported for now.
3240 * Return
3241 * **SK_PASS** on success, or **SK_DROP** on error.
3242 *
3243 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3244 * Description
3245 * For socket policies, apply the verdict of the eBPF program to
3246 * the next *bytes* (number of bytes) of message *msg*.
3247 *
3248 * For example, this helper can be used in the following cases:
3249 *
3250 * * A single **sendmsg**\ () or **sendfile**\ () system call
3251 * contains multiple logical messages that the eBPF program is
3252 * supposed to read and for which it should apply a verdict.
3253 * * An eBPF program only cares to read the first *bytes* of a
3254 * *msg*. If the message has a large payload, then setting up
3255 * and calling the eBPF program repeatedly for all bytes, even
3256 * though the verdict is already known, would create unnecessary
3257 * overhead.
3258 *
3259 * When called from within an eBPF program, the helper sets a
3260 * counter internal to the BPF infrastructure, that is used to
3261 * apply the last verdict to the next *bytes*. If *bytes* is
3262 * smaller than the current data being processed from a
3263 * **sendmsg**\ () or **sendfile**\ () system call, the first
3264 * *bytes* will be sent and the eBPF program will be re-run with
3265 * the pointer for start of data pointing to byte number *bytes*
3266 * **+ 1**. If *bytes* is larger than the current data being
3267 * processed, then the eBPF verdict will be applied to multiple
3268 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3269 * consumed.
3270 *
3271 * Note that if a socket closes with the internal counter holding
3272 * a non-zero value, this is not a problem because data is not
3273 * being buffered for *bytes* and is sent as it is received.
3274 * Return
3275 * 0
3276 *
3277 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3278 * Description
3279 * For socket policies, prevent the execution of the verdict eBPF
3280 * program for message *msg* until *bytes* (byte number) have been
3281 * accumulated.
3282 *
3283 * This can be used when one needs a specific number of bytes
3284 * before a verdict can be assigned, even if the data spans
3285 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3286 * case would be a user calling **sendmsg**\ () repeatedly with
3287 * 1-byte long message segments. Obviously, this is bad for
3288 * performance, but it is still valid. If the eBPF program needs
3289 * *bytes* bytes to validate a header, this helper can be used to
3290 * prevent the eBPF program to be called again until *bytes* have
3291 * been accumulated.
3292 * Return
3293 * 0
3294 *
3295 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3296 * Description
3297 * For socket policies, pull in non-linear data from user space
3298 * for *msg* and set pointers *msg*\ **->data** and *msg*\
3299 * **->data_end** to *start* and *end* bytes offsets into *msg*,
3300 * respectively.
3301 *
3302 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3303 * *msg* it can only parse data that the (**data**, **data_end**)
3304 * pointers have already consumed. For **sendmsg**\ () hooks this
3305 * is likely the first scatterlist element. But for calls relying
3306 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3307 * be the range (**0**, **0**) because the data is shared with
3308 * user space and by default the objective is to avoid allowing
3309 * user space to modify data while (or after) eBPF verdict is
3310 * being decided. This helper can be used to pull in data and to
3311 * set the start and end pointer to given values. Data will be
3312 * copied if necessary (i.e. if data was not linear and if start
3313 * and end pointers do not point to the same chunk).
3314 *
3315 * A call to this helper is susceptible to change the underlying
3316 * packet buffer. Therefore, at load time, all checks on pointers
3317 * previously done by the verifier are invalidated and must be
3318 * performed again, if the helper is used in combination with
3319 * direct packet access.
3320 *
3321 * All values for *flags* are reserved for future usage, and must
3322 * be left at zero.
3323 * Return
3324 * 0 on success, or a negative error in case of failure.
3325 *
3326 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3327 * Description
3328 * Bind the socket associated to *ctx* to the address pointed by
3329 * *addr*, of length *addr_len*. This allows for making outgoing
3330 * connection from the desired IP address, which can be useful for
3331 * example when all processes inside a cgroup should use one
3332 * single IP address on a host that has multiple IP configured.
3333 *
3334 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3335 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3336 * **AF_INET6**). It's advised to pass zero port (**sin_port**
3337 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3338 * behavior and lets the kernel efficiently pick up an unused
3339 * port as long as 4-tuple is unique. Passing non-zero port might
3340 * lead to degraded performance.
3341 * Return
3342 * 0 on success, or a negative error in case of failure.
3343 *
3344 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3345 * Description
3346 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3347 * possible to both shrink and grow the packet tail.
3348 * Shrink done via *delta* being a negative integer.
3349 *
3350 * A call to this helper is susceptible to change the underlying
3351 * packet buffer. Therefore, at load time, all checks on pointers
3352 * previously done by the verifier are invalidated and must be
3353 * performed again, if the helper is used in combination with
3354 * direct packet access.
3355 * Return
3356 * 0 on success, or a negative error in case of failure.
3357 *
3358 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3359 * Description
3360 * Retrieve the XFRM state (IP transform framework, see also
3361 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3362 *
3363 * The retrieved value is stored in the **struct bpf_xfrm_state**
3364 * pointed by *xfrm_state* and of length *size*.
3365 *
3366 * All values for *flags* are reserved for future usage, and must
3367 * be left at zero.
3368 *
3369 * This helper is available only if the kernel was compiled with
3370 * **CONFIG_XFRM** configuration option.
3371 * Return
3372 * 0 on success, or a negative error in case of failure.
3373 *
3374 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3375 * Description
3376 * Return a user or a kernel stack in bpf program provided buffer.
3377 * To achieve this, the helper needs *ctx*, which is a pointer
3378 * to the context on which the tracing program is executed.
3379 * To store the stacktrace, the bpf program provides *buf* with
3380 * a nonnegative *size*.
3381 *
3382 * The last argument, *flags*, holds the number of stack frames to
3383 * skip (from 0 to 255), masked with
3384 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3385 * the following flags:
3386 *
3387 * **BPF_F_USER_STACK**
3388 * Collect a user space stack instead of a kernel stack.
3389 * **BPF_F_USER_BUILD_ID**
3390 * Collect (build_id, file_offset) instead of ips for user
3391 * stack, only valid if **BPF_F_USER_STACK** is also
3392 * specified.
3393 *
3394 * *file_offset* is an offset relative to the beginning
3395 * of the executable or shared object file backing the vma
3396 * which the *ip* falls in. It is *not* an offset relative
3397 * to that object's base address. Accordingly, it must be
3398 * adjusted by adding (sh_addr - sh_offset), where
3399 * sh_{addr,offset} correspond to the executable section
3400 * containing *file_offset* in the object, for comparisons
3401 * to symbols' st_value to be valid.
3402 *
3403 * **bpf_get_stack**\ () can collect up to
3404 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3405 * to sufficient large buffer size. Note that
3406 * this limit can be controlled with the **sysctl** program, and
3407 * that it should be manually increased in order to profile long
3408 * user stacks (such as stacks for Java programs). To do so, use:
3409 *
3410 * ::
3411 *
3412 * # sysctl kernel.perf_event_max_stack=<new value>
3413 * Return
3414 * The non-negative copied *buf* length equal to or less than
3415 * *size* on success, or a negative error in case of failure.
3416 *
3417 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3418 * Description
3419 * This helper is similar to **bpf_skb_load_bytes**\ () in that
3420 * it provides an easy way to load *len* bytes from *offset*
3421 * from the packet associated to *skb*, into the buffer pointed
3422 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3423 * a fifth argument *start_header* exists in order to select a
3424 * base offset to start from. *start_header* can be one of:
3425 *
3426 * **BPF_HDR_START_MAC**
3427 * Base offset to load data from is *skb*'s mac header.
3428 * **BPF_HDR_START_NET**
3429 * Base offset to load data from is *skb*'s network header.
3430 *
3431 * In general, "direct packet access" is the preferred method to
3432 * access packet data, however, this helper is in particular useful
3433 * in socket filters where *skb*\ **->data** does not always point
3434 * to the start of the mac header and where "direct packet access"
3435 * is not available.
3436 * Return
3437 * 0 on success, or a negative error in case of failure.
3438 *
3439 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3440 * Description
3441 * Do FIB lookup in kernel tables using parameters in *params*.
3442 * If lookup is successful and result shows packet is to be
3443 * forwarded, the neighbor tables are searched for the nexthop.
3444 * If successful (ie., FIB lookup shows forwarding and nexthop
3445 * is resolved), the nexthop address is returned in ipv4_dst
3446 * or ipv6_dst based on family, smac is set to mac address of
3447 * egress device, dmac is set to nexthop mac address, rt_metric
3448 * is set to metric from route (IPv4/IPv6 only), and ifindex
3449 * is set to the device index of the nexthop from the FIB lookup.
3450 *
3451 * *plen* argument is the size of the passed in struct.
3452 * *flags* argument can be a combination of one or more of the
3453 * following values:
3454 *
3455 * **BPF_FIB_LOOKUP_DIRECT**
3456 * Do a direct table lookup vs full lookup using FIB
3457 * rules.
3458 * **BPF_FIB_LOOKUP_TBID**
3459 * Used with BPF_FIB_LOOKUP_DIRECT.
3460 * Use the routing table ID present in *params*->tbid
3461 * for the fib lookup.
3462 * **BPF_FIB_LOOKUP_OUTPUT**
3463 * Perform lookup from an egress perspective (default is
3464 * ingress).
3465 * **BPF_FIB_LOOKUP_SKIP_NEIGH**
3466 * Skip the neighbour table lookup. *params*->dmac
3467 * and *params*->smac will not be set as output. A common
3468 * use case is to call **bpf_redirect_neigh**\ () after
3469 * doing **bpf_fib_lookup**\ ().
3470 * **BPF_FIB_LOOKUP_SRC**
3471 * Derive and set source IP addr in *params*->ipv{4,6}_src
3472 * for the nexthop. If the src addr cannot be derived,
3473 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3474 * case, *params*->dmac and *params*->smac are not set either.
3475 * **BPF_FIB_LOOKUP_MARK**
3476 * Use the mark present in *params*->mark for the fib lookup.
3477 * This option should not be used with BPF_FIB_LOOKUP_DIRECT,
3478 * as it only has meaning for full lookups.
3479 *
3480 * *ctx* is either **struct xdp_md** for XDP programs or
3481 * **struct sk_buff** tc cls_act programs.
3482 * Return
3483 * * < 0 if any input argument is invalid
3484 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3485 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3486 * packet is not forwarded or needs assist from full stack
3487 *
3488 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3489 * was exceeded and output params->mtu_result contains the MTU.
3490 *
3491 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3492 * Description
3493 * Add an entry to, or update a sockhash *map* referencing sockets.
3494 * The *skops* is used as a new value for the entry associated to
3495 * *key*. *flags* is one of:
3496 *
3497 * **BPF_NOEXIST**
3498 * The entry for *key* must not exist in the map.
3499 * **BPF_EXIST**
3500 * The entry for *key* must already exist in the map.
3501 * **BPF_ANY**
3502 * No condition on the existence of the entry for *key*.
3503 *
3504 * If the *map* has eBPF programs (parser and verdict), those will
3505 * be inherited by the socket being added. If the socket is
3506 * already attached to eBPF programs, this results in an error.
3507 * Return
3508 * 0 on success, or a negative error in case of failure.
3509 *
3510 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3511 * Description
3512 * This helper is used in programs implementing policies at the
3513 * socket level. If the message *msg* is allowed to pass (i.e. if
3514 * the verdict eBPF program returns **SK_PASS**), redirect it to
3515 * the socket referenced by *map* (of type
3516 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3517 * egress interfaces can be used for redirection. The
3518 * **BPF_F_INGRESS** value in *flags* is used to make the
3519 * distinction (ingress path is selected if the flag is present,
3520 * egress path otherwise). This is the only flag supported for now.
3521 * Return
3522 * **SK_PASS** on success, or **SK_DROP** on error.
3523 *
3524 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3525 * Description
3526 * This helper is used in programs implementing policies at the
3527 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3528 * if the verdict eBPF program returns **SK_PASS**), redirect it
3529 * to the socket referenced by *map* (of type
3530 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3531 * egress interfaces can be used for redirection. The
3532 * **BPF_F_INGRESS** value in *flags* is used to make the
3533 * distinction (ingress path is selected if the flag is present,
3534 * egress otherwise). This is the only flag supported for now.
3535 * Return
3536 * **SK_PASS** on success, or **SK_DROP** on error.
3537 *
3538 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3539 * Description
3540 * Encapsulate the packet associated to *skb* within a Layer 3
3541 * protocol header. This header is provided in the buffer at
3542 * address *hdr*, with *len* its size in bytes. *type* indicates
3543 * the protocol of the header and can be one of:
3544 *
3545 * **BPF_LWT_ENCAP_SEG6**
3546 * IPv6 encapsulation with Segment Routing Header
3547 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3548 * the IPv6 header is computed by the kernel.
3549 * **BPF_LWT_ENCAP_SEG6_INLINE**
3550 * Only works if *skb* contains an IPv6 packet. Insert a
3551 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3552 * the IPv6 header.
3553 * **BPF_LWT_ENCAP_IP**
3554 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3555 * must be IPv4 or IPv6, followed by zero or more
3556 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3557 * total bytes in all prepended headers. Please note that
3558 * if **skb_is_gso**\ (*skb*) is true, no more than two
3559 * headers can be prepended, and the inner header, if
3560 * present, should be either GRE or UDP/GUE.
3561 *
3562 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3563 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3564 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3565 * **BPF_PROG_TYPE_LWT_XMIT**.
3566 *
3567 * A call to this helper is susceptible to change the underlying
3568 * packet buffer. Therefore, at load time, all checks on pointers
3569 * previously done by the verifier are invalidated and must be
3570 * performed again, if the helper is used in combination with
3571 * direct packet access.
3572 * Return
3573 * 0 on success, or a negative error in case of failure.
3574 *
3575 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3576 * Description
3577 * Store *len* bytes from address *from* into the packet
3578 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3579 * inside the outermost IPv6 Segment Routing Header can be
3580 * modified through this helper.
3581 *
3582 * A call to this helper is susceptible to change the underlying
3583 * packet buffer. Therefore, at load time, all checks on pointers
3584 * previously done by the verifier are invalidated and must be
3585 * performed again, if the helper is used in combination with
3586 * direct packet access.
3587 * Return
3588 * 0 on success, or a negative error in case of failure.
3589 *
3590 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3591 * Description
3592 * Adjust the size allocated to TLVs in the outermost IPv6
3593 * Segment Routing Header contained in the packet associated to
3594 * *skb*, at position *offset* by *delta* bytes. Only offsets
3595 * after the segments are accepted. *delta* can be as well
3596 * positive (growing) as negative (shrinking).
3597 *
3598 * A call to this helper is susceptible to change the underlying
3599 * packet buffer. Therefore, at load time, all checks on pointers
3600 * previously done by the verifier are invalidated and must be
3601 * performed again, if the helper is used in combination with
3602 * direct packet access.
3603 * Return
3604 * 0 on success, or a negative error in case of failure.
3605 *
3606 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3607 * Description
3608 * Apply an IPv6 Segment Routing action of type *action* to the
3609 * packet associated to *skb*. Each action takes a parameter
3610 * contained at address *param*, and of length *param_len* bytes.
3611 * *action* can be one of:
3612 *
3613 * **SEG6_LOCAL_ACTION_END_X**
3614 * End.X action: Endpoint with Layer-3 cross-connect.
3615 * Type of *param*: **struct in6_addr**.
3616 * **SEG6_LOCAL_ACTION_END_T**
3617 * End.T action: Endpoint with specific IPv6 table lookup.
3618 * Type of *param*: **int**.
3619 * **SEG6_LOCAL_ACTION_END_B6**
3620 * End.B6 action: Endpoint bound to an SRv6 policy.
3621 * Type of *param*: **struct ipv6_sr_hdr**.
3622 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3623 * End.B6.Encap action: Endpoint bound to an SRv6
3624 * encapsulation policy.
3625 * Type of *param*: **struct ipv6_sr_hdr**.
3626 *
3627 * A call to this helper is susceptible to change the underlying
3628 * packet buffer. Therefore, at load time, all checks on pointers
3629 * previously done by the verifier are invalidated and must be
3630 * performed again, if the helper is used in combination with
3631 * direct packet access.
3632 * Return
3633 * 0 on success, or a negative error in case of failure.
3634 *
3635 * long bpf_rc_repeat(void *ctx)
3636 * Description
3637 * This helper is used in programs implementing IR decoding, to
3638 * report a successfully decoded repeat key message. This delays
3639 * the generation of a key up event for previously generated
3640 * key down event.
3641 *
3642 * Some IR protocols like NEC have a special IR message for
3643 * repeating last button, for when a button is held down.
3644 *
3645 * The *ctx* should point to the lirc sample as passed into
3646 * the program.
3647 *
3648 * This helper is only available is the kernel was compiled with
3649 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3650 * "**y**".
3651 * Return
3652 * 0
3653 *
3654 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3655 * Description
3656 * This helper is used in programs implementing IR decoding, to
3657 * report a successfully decoded key press with *scancode*,
3658 * *toggle* value in the given *protocol*. The scancode will be
3659 * translated to a keycode using the rc keymap, and reported as
3660 * an input key down event. After a period a key up event is
3661 * generated. This period can be extended by calling either
3662 * **bpf_rc_keydown**\ () again with the same values, or calling
3663 * **bpf_rc_repeat**\ ().
3664 *
3665 * Some protocols include a toggle bit, in case the button was
3666 * released and pressed again between consecutive scancodes.
3667 *
3668 * The *ctx* should point to the lirc sample as passed into
3669 * the program.
3670 *
3671 * The *protocol* is the decoded protocol number (see
3672 * **enum rc_proto** for some predefined values).
3673 *
3674 * This helper is only available is the kernel was compiled with
3675 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3676 * "**y**".
3677 * Return
3678 * 0
3679 *
3680 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3681 * Description
3682 * Return the cgroup v2 id of the socket associated with the *skb*.
3683 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3684 * helper for cgroup v1 by providing a tag resp. identifier that
3685 * can be matched on or used for map lookups e.g. to implement
3686 * policy. The cgroup v2 id of a given path in the hierarchy is
3687 * exposed in user space through the f_handle API in order to get
3688 * to the same 64-bit id.
3689 *
3690 * This helper can be used on TC egress path, but not on ingress,
3691 * and is available only if the kernel was compiled with the
3692 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3693 * Return
3694 * The id is returned or 0 in case the id could not be retrieved.
3695 *
3696 * u64 bpf_get_current_cgroup_id(void)
3697 * Description
3698 * Get the current cgroup id based on the cgroup within which
3699 * the current task is running.
3700 * Return
3701 * A 64-bit integer containing the current cgroup id based
3702 * on the cgroup within which the current task is running.
3703 *
3704 * void *bpf_get_local_storage(void *map, u64 flags)
3705 * Description
3706 * Get the pointer to the local storage area.
3707 * The type and the size of the local storage is defined
3708 * by the *map* argument.
3709 * The *flags* meaning is specific for each map type,
3710 * and has to be 0 for cgroup local storage.
3711 *
3712 * Depending on the BPF program type, a local storage area
3713 * can be shared between multiple instances of the BPF program,
3714 * running simultaneously.
3715 *
3716 * A user should care about the synchronization by himself.
3717 * For example, by using the **BPF_ATOMIC** instructions to alter
3718 * the shared data.
3719 * Return
3720 * A pointer to the local storage area.
3721 *
3722 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3723 * Description
3724 * Select a **SO_REUSEPORT** socket from a
3725 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3726 * It checks the selected socket is matching the incoming
3727 * request in the socket buffer.
3728 * Return
3729 * 0 on success, or a negative error in case of failure.
3730 *
3731 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3732 * Description
3733 * Return id of cgroup v2 that is ancestor of cgroup associated
3734 * with the *skb* at the *ancestor_level*. The root cgroup is at
3735 * *ancestor_level* zero and each step down the hierarchy
3736 * increments the level. If *ancestor_level* == level of cgroup
3737 * associated with *skb*, then return value will be same as that
3738 * of **bpf_skb_cgroup_id**\ ().
3739 *
3740 * The helper is useful to implement policies based on cgroups
3741 * that are upper in hierarchy than immediate cgroup associated
3742 * with *skb*.
3743 *
3744 * The format of returned id and helper limitations are same as in
3745 * **bpf_skb_cgroup_id**\ ().
3746 * Return
3747 * The id is returned or 0 in case the id could not be retrieved.
3748 *
3749 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3750 * Description
3751 * Look for TCP socket matching *tuple*, optionally in a child
3752 * network namespace *netns*. The return value must be checked,
3753 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3754 *
3755 * The *ctx* should point to the context of the program, such as
3756 * the skb or socket (depending on the hook in use). This is used
3757 * to determine the base network namespace for the lookup.
3758 *
3759 * *tuple_size* must be one of:
3760 *
3761 * **sizeof**\ (*tuple*\ **->ipv4**)
3762 * Look for an IPv4 socket.
3763 * **sizeof**\ (*tuple*\ **->ipv6**)
3764 * Look for an IPv6 socket.
3765 *
3766 * If the *netns* is a negative signed 32-bit integer, then the
3767 * socket lookup table in the netns associated with the *ctx*
3768 * will be used. For the TC hooks, this is the netns of the device
3769 * in the skb. For socket hooks, this is the netns of the socket.
3770 * If *netns* is any other signed 32-bit value greater than or
3771 * equal to zero then it specifies the ID of the netns relative to
3772 * the netns associated with the *ctx*. *netns* values beyond the
3773 * range of 32-bit integers are reserved for future use.
3774 *
3775 * All values for *flags* are reserved for future usage, and must
3776 * be left at zero.
3777 *
3778 * This helper is available only if the kernel was compiled with
3779 * **CONFIG_NET** configuration option.
3780 * Return
3781 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3782 * For sockets with reuseport option, the **struct bpf_sock**
3783 * result is from *reuse*\ **->socks**\ [] using the hash of the
3784 * tuple.
3785 *
3786 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3787 * Description
3788 * Look for UDP socket matching *tuple*, optionally in a child
3789 * network namespace *netns*. The return value must be checked,
3790 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3791 *
3792 * The *ctx* should point to the context of the program, such as
3793 * the skb or socket (depending on the hook in use). This is used
3794 * to determine the base network namespace for the lookup.
3795 *
3796 * *tuple_size* must be one of:
3797 *
3798 * **sizeof**\ (*tuple*\ **->ipv4**)
3799 * Look for an IPv4 socket.
3800 * **sizeof**\ (*tuple*\ **->ipv6**)
3801 * Look for an IPv6 socket.
3802 *
3803 * If the *netns* is a negative signed 32-bit integer, then the
3804 * socket lookup table in the netns associated with the *ctx*
3805 * will be used. For the TC hooks, this is the netns of the device
3806 * in the skb. For socket hooks, this is the netns of the socket.
3807 * If *netns* is any other signed 32-bit value greater than or
3808 * equal to zero then it specifies the ID of the netns relative to
3809 * the netns associated with the *ctx*. *netns* values beyond the
3810 * range of 32-bit integers are reserved for future use.
3811 *
3812 * All values for *flags* are reserved for future usage, and must
3813 * be left at zero.
3814 *
3815 * This helper is available only if the kernel was compiled with
3816 * **CONFIG_NET** configuration option.
3817 * Return
3818 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3819 * For sockets with reuseport option, the **struct bpf_sock**
3820 * result is from *reuse*\ **->socks**\ [] using the hash of the
3821 * tuple.
3822 *
3823 * long bpf_sk_release(void *sock)
3824 * Description
3825 * Release the reference held by *sock*. *sock* must be a
3826 * non-**NULL** pointer that was returned from
3827 * **bpf_sk_lookup_xxx**\ ().
3828 * Return
3829 * 0 on success, or a negative error in case of failure.
3830 *
3831 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3832 * Description
3833 * Push an element *value* in *map*. *flags* is one of:
3834 *
3835 * **BPF_EXIST**
3836 * If the queue/stack is full, the oldest element is
3837 * removed to make room for this.
3838 * Return
3839 * 0 on success, or a negative error in case of failure.
3840 *
3841 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3842 * Description
3843 * Pop an element from *map*.
3844 * Return
3845 * 0 on success, or a negative error in case of failure.
3846 *
3847 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3848 * Description
3849 * Get an element from *map* without removing it.
3850 * Return
3851 * 0 on success, or a negative error in case of failure.
3852 *
3853 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3854 * Description
3855 * For socket policies, insert *len* bytes into *msg* at offset
3856 * *start*.
3857 *
3858 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3859 * *msg* it may want to insert metadata or options into the *msg*.
3860 * This can later be read and used by any of the lower layer BPF
3861 * hooks.
3862 *
3863 * This helper may fail if under memory pressure (a malloc
3864 * fails) in these cases BPF programs will get an appropriate
3865 * error and BPF programs will need to handle them.
3866 * Return
3867 * 0 on success, or a negative error in case of failure.
3868 *
3869 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3870 * Description
3871 * Will remove *len* bytes from a *msg* starting at byte *start*.
3872 * This may result in **ENOMEM** errors under certain situations if
3873 * an allocation and copy are required due to a full ring buffer.
3874 * However, the helper will try to avoid doing the allocation
3875 * if possible. Other errors can occur if input parameters are
3876 * invalid either due to *start* byte not being valid part of *msg*
3877 * payload and/or *pop* value being to large.
3878 * Return
3879 * 0 on success, or a negative error in case of failure.
3880 *
3881 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3882 * Description
3883 * This helper is used in programs implementing IR decoding, to
3884 * report a successfully decoded pointer movement.
3885 *
3886 * The *ctx* should point to the lirc sample as passed into
3887 * the program.
3888 *
3889 * This helper is only available is the kernel was compiled with
3890 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3891 * "**y**".
3892 * Return
3893 * 0
3894 *
3895 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3896 * Description
3897 * Acquire a spinlock represented by the pointer *lock*, which is
3898 * stored as part of a value of a map. Taking the lock allows to
3899 * safely update the rest of the fields in that value. The
3900 * spinlock can (and must) later be released with a call to
3901 * **bpf_spin_unlock**\ (\ *lock*\ ).
3902 *
3903 * Spinlocks in BPF programs come with a number of restrictions
3904 * and constraints:
3905 *
3906 * * **bpf_spin_lock** objects are only allowed inside maps of
3907 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3908 * list could be extended in the future).
3909 * * BTF description of the map is mandatory.
3910 * * The BPF program can take ONE lock at a time, since taking two
3911 * or more could cause dead locks.
3912 * * Only one **struct bpf_spin_lock** is allowed per map element.
3913 * * When the lock is taken, calls (either BPF to BPF or helpers)
3914 * are not allowed.
3915 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3916 * allowed inside a spinlock-ed region.
3917 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3918 * the lock, on all execution paths, before it returns.
3919 * * The BPF program can access **struct bpf_spin_lock** only via
3920 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3921 * helpers. Loading or storing data into the **struct
3922 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3923 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3924 * of the map value must be a struct and have **struct
3925 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3926 * Nested lock inside another struct is not allowed.
3927 * * The **struct bpf_spin_lock** *lock* field in a map value must
3928 * be aligned on a multiple of 4 bytes in that value.
3929 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3930 * the **bpf_spin_lock** field to user space.
3931 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3932 * a BPF program, do not update the **bpf_spin_lock** field.
3933 * * **bpf_spin_lock** cannot be on the stack or inside a
3934 * networking packet (it can only be inside of a map values).
3935 * * **bpf_spin_lock** is available to root only.
3936 * * Tracing programs and socket filter programs cannot use
3937 * **bpf_spin_lock**\ () due to insufficient preemption checks
3938 * (but this may change in the future).
3939 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3940 * Return
3941 * 0
3942 *
3943 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3944 * Description
3945 * Release the *lock* previously locked by a call to
3946 * **bpf_spin_lock**\ (\ *lock*\ ).
3947 * Return
3948 * 0
3949 *
3950 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3951 * Description
3952 * This helper gets a **struct bpf_sock** pointer such
3953 * that all the fields in this **bpf_sock** can be accessed.
3954 * Return
3955 * A **struct bpf_sock** pointer on success, or **NULL** in
3956 * case of failure.
3957 *
3958 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3959 * Description
3960 * This helper gets a **struct bpf_tcp_sock** pointer from a
3961 * **struct bpf_sock** pointer.
3962 * Return
3963 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3964 * case of failure.
3965 *
3966 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3967 * Description
3968 * Set ECN (Explicit Congestion Notification) field of IP header
3969 * to **CE** (Congestion Encountered) if current value is **ECT**
3970 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3971 * and IPv4.
3972 * Return
3973 * 1 if the **CE** flag is set (either by the current helper call
3974 * or because it was already present), 0 if it is not set.
3975 *
3976 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3977 * Description
3978 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3979 * **bpf_sk_release**\ () is unnecessary and not allowed.
3980 * Return
3981 * A **struct bpf_sock** pointer on success, or **NULL** in
3982 * case of failure.
3983 *
3984 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3985 * Description
3986 * Look for TCP socket matching *tuple*, optionally in a child
3987 * network namespace *netns*. The return value must be checked,
3988 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3989 *
3990 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3991 * that it also returns timewait or request sockets. Use
3992 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3993 * full structure.
3994 *
3995 * This helper is available only if the kernel was compiled with
3996 * **CONFIG_NET** configuration option.
3997 * Return
3998 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3999 * For sockets with reuseport option, the **struct bpf_sock**
4000 * result is from *reuse*\ **->socks**\ [] using the hash of the
4001 * tuple.
4002 *
4003 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4004 * Description
4005 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
4006 * the listening socket in *sk*.
4007 *
4008 * *iph* points to the start of the IPv4 or IPv6 header, while
4009 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
4010 * **sizeof**\ (**struct ipv6hdr**).
4011 *
4012 * *th* points to the start of the TCP header, while *th_len*
4013 * contains the length of the TCP header (at least
4014 * **sizeof**\ (**struct tcphdr**)).
4015 * Return
4016 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
4017 * error otherwise.
4018 *
4019 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
4020 * Description
4021 * Get name of sysctl in /proc/sys/ and copy it into provided by
4022 * program buffer *buf* of size *buf_len*.
4023 *
4024 * The buffer is always NUL terminated, unless it's zero-sized.
4025 *
4026 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
4027 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
4028 * only (e.g. "tcp_mem").
4029 * Return
4030 * Number of character copied (not including the trailing NUL).
4031 *
4032 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4033 * truncated name in this case).
4034 *
4035 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4036 * Description
4037 * Get current value of sysctl as it is presented in /proc/sys
4038 * (incl. newline, etc), and copy it as a string into provided
4039 * by program buffer *buf* of size *buf_len*.
4040 *
4041 * The whole value is copied, no matter what file position user
4042 * space issued e.g. sys_read at.
4043 *
4044 * The buffer is always NUL terminated, unless it's zero-sized.
4045 * Return
4046 * Number of character copied (not including the trailing NUL).
4047 *
4048 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4049 * truncated name in this case).
4050 *
4051 * **-EINVAL** if current value was unavailable, e.g. because
4052 * sysctl is uninitialized and read returns -EIO for it.
4053 *
4054 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4055 * Description
4056 * Get new value being written by user space to sysctl (before
4057 * the actual write happens) and copy it as a string into
4058 * provided by program buffer *buf* of size *buf_len*.
4059 *
4060 * User space may write new value at file position > 0.
4061 *
4062 * The buffer is always NUL terminated, unless it's zero-sized.
4063 * Return
4064 * Number of character copied (not including the trailing NUL).
4065 *
4066 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4067 * truncated name in this case).
4068 *
4069 * **-EINVAL** if sysctl is being read.
4070 *
4071 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
4072 * Description
4073 * Override new value being written by user space to sysctl with
4074 * value provided by program in buffer *buf* of size *buf_len*.
4075 *
4076 * *buf* should contain a string in same form as provided by user
4077 * space on sysctl write.
4078 *
4079 * User space may write new value at file position > 0. To override
4080 * the whole sysctl value file position should be set to zero.
4081 * Return
4082 * 0 on success.
4083 *
4084 * **-E2BIG** if the *buf_len* is too big.
4085 *
4086 * **-EINVAL** if sysctl is being read.
4087 *
4088 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
4089 * Description
4090 * Convert the initial part of the string from buffer *buf* of
4091 * size *buf_len* to a long integer according to the given base
4092 * and save the result in *res*.
4093 *
4094 * The string may begin with an arbitrary amount of white space
4095 * (as determined by **isspace**\ (3)) followed by a single
4096 * optional '**-**' sign.
4097 *
4098 * Five least significant bits of *flags* encode base, other bits
4099 * are currently unused.
4100 *
4101 * Base must be either 8, 10, 16 or 0 to detect it automatically
4102 * similar to user space **strtol**\ (3).
4103 * Return
4104 * Number of characters consumed on success. Must be positive but
4105 * no more than *buf_len*.
4106 *
4107 * **-EINVAL** if no valid digits were found or unsupported base
4108 * was provided.
4109 *
4110 * **-ERANGE** if resulting value was out of range.
4111 *
4112 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
4113 * Description
4114 * Convert the initial part of the string from buffer *buf* of
4115 * size *buf_len* to an unsigned long integer according to the
4116 * given base and save the result in *res*.
4117 *
4118 * The string may begin with an arbitrary amount of white space
4119 * (as determined by **isspace**\ (3)).
4120 *
4121 * Five least significant bits of *flags* encode base, other bits
4122 * are currently unused.
4123 *
4124 * Base must be either 8, 10, 16 or 0 to detect it automatically
4125 * similar to user space **strtoul**\ (3).
4126 * Return
4127 * Number of characters consumed on success. Must be positive but
4128 * no more than *buf_len*.
4129 *
4130 * **-EINVAL** if no valid digits were found or unsupported base
4131 * was provided.
4132 *
4133 * **-ERANGE** if resulting value was out of range.
4134 *
4135 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
4136 * Description
4137 * Get a bpf-local-storage from a *sk*.
4138 *
4139 * Logically, it could be thought of getting the value from
4140 * a *map* with *sk* as the **key**. From this
4141 * perspective, the usage is not much different from
4142 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
4143 * helper enforces the key must be a full socket and the map must
4144 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
4145 *
4146 * Underneath, the value is stored locally at *sk* instead of
4147 * the *map*. The *map* is used as the bpf-local-storage
4148 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4149 * searched against all bpf-local-storages residing at *sk*.
4150 *
4151 * *sk* is a kernel **struct sock** pointer for LSM program.
4152 * *sk* is a **struct bpf_sock** pointer for other program types.
4153 *
4154 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
4155 * used such that a new bpf-local-storage will be
4156 * created if one does not exist. *value* can be used
4157 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
4158 * the initial value of a bpf-local-storage. If *value* is
4159 * **NULL**, the new bpf-local-storage will be zero initialized.
4160 * Return
4161 * A bpf-local-storage pointer is returned on success.
4162 *
4163 * **NULL** if not found or there was an error in adding
4164 * a new bpf-local-storage.
4165 *
4166 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
4167 * Description
4168 * Delete a bpf-local-storage from a *sk*.
4169 * Return
4170 * 0 on success.
4171 *
4172 * **-ENOENT** if the bpf-local-storage cannot be found.
4173 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
4174 *
4175 * long bpf_send_signal(u32 sig)
4176 * Description
4177 * Send signal *sig* to the process of the current task.
4178 * The signal may be delivered to any of this process's threads.
4179 * Return
4180 * 0 on success or successfully queued.
4181 *
4182 * **-EBUSY** if work queue under nmi is full.
4183 *
4184 * **-EINVAL** if *sig* is invalid.
4185 *
4186 * **-EPERM** if no permission to send the *sig*.
4187 *
4188 * **-EAGAIN** if bpf program can try again.
4189 *
4190 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4191 * Description
4192 * Try to issue a SYN cookie for the packet with corresponding
4193 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
4194 *
4195 * *iph* points to the start of the IPv4 or IPv6 header, while
4196 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
4197 * **sizeof**\ (**struct ipv6hdr**).
4198 *
4199 * *th* points to the start of the TCP header, while *th_len*
4200 * contains the length of the TCP header with options (at least
4201 * **sizeof**\ (**struct tcphdr**)).
4202 * Return
4203 * On success, lower 32 bits hold the generated SYN cookie in
4204 * followed by 16 bits which hold the MSS value for that cookie,
4205 * and the top 16 bits are unused.
4206 *
4207 * On failure, the returned value is one of the following:
4208 *
4209 * **-EINVAL** SYN cookie cannot be issued due to error
4210 *
4211 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
4212 *
4213 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
4214 *
4215 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
4216 *
4217 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4218 * Description
4219 * Write raw *data* blob into a special BPF perf event held by
4220 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4221 * event must have the following attributes: **PERF_SAMPLE_RAW**
4222 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4223 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4224 *
4225 * The *flags* are used to indicate the index in *map* for which
4226 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4227 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4228 * to indicate that the index of the current CPU core should be
4229 * used.
4230 *
4231 * The value to write, of *size*, is passed through eBPF stack and
4232 * pointed by *data*.
4233 *
4234 * *ctx* is a pointer to in-kernel struct sk_buff.
4235 *
4236 * This helper is similar to **bpf_perf_event_output**\ () but
4237 * restricted to raw_tracepoint bpf programs.
4238 * Return
4239 * 0 on success, or a negative error in case of failure.
4240 *
4241 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4242 * Description
4243 * Safely attempt to read *size* bytes from user space address
4244 * *unsafe_ptr* and store the data in *dst*.
4245 * Return
4246 * 0 on success, or a negative error in case of failure.
4247 *
4248 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4249 * Description
4250 * Safely attempt to read *size* bytes from kernel space address
4251 * *unsafe_ptr* and store the data in *dst*.
4252 * Return
4253 * 0 on success, or a negative error in case of failure.
4254 *
4255 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4256 * Description
4257 * Copy a NUL terminated string from an unsafe user address
4258 * *unsafe_ptr* to *dst*. The *size* should include the
4259 * terminating NUL byte. In case the string length is smaller than
4260 * *size*, the target is not padded with further NUL bytes. If the
4261 * string length is larger than *size*, just *size*-1 bytes are
4262 * copied and the last byte is set to NUL.
4263 *
4264 * On success, returns the number of bytes that were written,
4265 * including the terminal NUL. This makes this helper useful in
4266 * tracing programs for reading strings, and more importantly to
4267 * get its length at runtime. See the following snippet:
4268 *
4269 * ::
4270 *
4271 * SEC("kprobe/sys_open")
4272 * void bpf_sys_open(struct pt_regs *ctx)
4273 * {
4274 * char buf[PATHLEN]; // PATHLEN is defined to 256
4275 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
4276 * ctx->di);
4277 *
4278 * // Consume buf, for example push it to
4279 * // userspace via bpf_perf_event_output(); we
4280 * // can use res (the string length) as event
4281 * // size, after checking its boundaries.
4282 * }
4283 *
4284 * In comparison, using **bpf_probe_read_user**\ () helper here
4285 * instead to read the string would require to estimate the length
4286 * at compile time, and would often result in copying more memory
4287 * than necessary.
4288 *
4289 * Another useful use case is when parsing individual process
4290 * arguments or individual environment variables navigating
4291 * *current*\ **->mm->arg_start** and *current*\
4292 * **->mm->env_start**: using this helper and the return value,
4293 * one can quickly iterate at the right offset of the memory area.
4294 * Return
4295 * On success, the strictly positive length of the output string,
4296 * including the trailing NUL character. On error, a negative
4297 * value.
4298 *
4299 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4300 * Description
4301 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4302 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4303 * Return
4304 * On success, the strictly positive length of the string, including
4305 * the trailing NUL character. On error, a negative value.
4306 *
4307 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4308 * Description
4309 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4310 * *rcv_nxt* is the ack_seq to be sent out.
4311 * Return
4312 * 0 on success, or a negative error in case of failure.
4313 *
4314 * long bpf_send_signal_thread(u32 sig)
4315 * Description
4316 * Send signal *sig* to the thread corresponding to the current task.
4317 * Return
4318 * 0 on success or successfully queued.
4319 *
4320 * **-EBUSY** if work queue under nmi is full.
4321 *
4322 * **-EINVAL** if *sig* is invalid.
4323 *
4324 * **-EPERM** if no permission to send the *sig*.
4325 *
4326 * **-EAGAIN** if bpf program can try again.
4327 *
4328 * u64 bpf_jiffies64(void)
4329 * Description
4330 * Obtain the 64bit jiffies
4331 * Return
4332 * The 64 bit jiffies
4333 *
4334 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4335 * Description
4336 * For an eBPF program attached to a perf event, retrieve the
4337 * branch records (**struct perf_branch_entry**) associated to *ctx*
4338 * and store it in the buffer pointed by *buf* up to size
4339 * *size* bytes.
4340 * Return
4341 * On success, number of bytes written to *buf*. On error, a
4342 * negative value.
4343 *
4344 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4345 * instead return the number of bytes required to store all the
4346 * branch entries. If this flag is set, *buf* may be NULL.
4347 *
4348 * **-EINVAL** if arguments invalid or **size** not a multiple
4349 * of **sizeof**\ (**struct perf_branch_entry**\ ).
4350 *
4351 * **-ENOENT** if architecture does not support branch records.
4352 *
4353 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4354 * Description
4355 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
4356 * *namespace* will be returned in *nsdata*.
4357 * Return
4358 * 0 on success, or one of the following in case of failure:
4359 *
4360 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
4361 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
4362 *
4363 * **-ENOENT** if pidns does not exists for the current task.
4364 *
4365 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4366 * Description
4367 * Write raw *data* blob into a special BPF perf event held by
4368 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4369 * event must have the following attributes: **PERF_SAMPLE_RAW**
4370 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4371 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4372 *
4373 * The *flags* are used to indicate the index in *map* for which
4374 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4375 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4376 * to indicate that the index of the current CPU core should be
4377 * used.
4378 *
4379 * The value to write, of *size*, is passed through eBPF stack and
4380 * pointed by *data*.
4381 *
4382 * *ctx* is a pointer to in-kernel struct xdp_buff.
4383 *
4384 * This helper is similar to **bpf_perf_eventoutput**\ () but
4385 * restricted to raw_tracepoint bpf programs.
4386 * Return
4387 * 0 on success, or a negative error in case of failure.
4388 *
4389 * u64 bpf_get_netns_cookie(void *ctx)
4390 * Description
4391 * Retrieve the cookie (generated by the kernel) of the network
4392 * namespace the input *ctx* is associated with. The network
4393 * namespace cookie remains stable for its lifetime and provides
4394 * a global identifier that can be assumed unique. If *ctx* is
4395 * NULL, then the helper returns the cookie for the initial
4396 * network namespace. The cookie itself is very similar to that
4397 * of **bpf_get_socket_cookie**\ () helper, but for network
4398 * namespaces instead of sockets.
4399 * Return
4400 * A 8-byte long opaque number.
4401 *
4402 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4403 * Description
4404 * Return id of cgroup v2 that is ancestor of the cgroup associated
4405 * with the current task at the *ancestor_level*. The root cgroup
4406 * is at *ancestor_level* zero and each step down the hierarchy
4407 * increments the level. If *ancestor_level* == level of cgroup
4408 * associated with the current task, then return value will be the
4409 * same as that of **bpf_get_current_cgroup_id**\ ().
4410 *
4411 * The helper is useful to implement policies based on cgroups
4412 * that are upper in hierarchy than immediate cgroup associated
4413 * with the current task.
4414 *
4415 * The format of returned id and helper limitations are same as in
4416 * **bpf_get_current_cgroup_id**\ ().
4417 * Return
4418 * The id is returned or 0 in case the id could not be retrieved.
4419 *
4420 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4421 * Description
4422 * Helper is overloaded depending on BPF program type. This
4423 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4424 * **BPF_PROG_TYPE_SCHED_ACT** programs.
4425 *
4426 * Assign the *sk* to the *skb*. When combined with appropriate
4427 * routing configuration to receive the packet towards the socket,
4428 * will cause *skb* to be delivered to the specified socket.
4429 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
4430 * **bpf_clone_redirect**\ () or other methods outside of BPF may
4431 * interfere with successful delivery to the socket.
4432 *
4433 * This operation is only valid from TC ingress path.
4434 *
4435 * The *flags* argument must be zero.
4436 * Return
4437 * 0 on success, or a negative error in case of failure:
4438 *
4439 * **-EINVAL** if specified *flags* are not supported.
4440 *
4441 * **-ENOENT** if the socket is unavailable for assignment.
4442 *
4443 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
4444 *
4445 * **-EOPNOTSUPP** if the operation is not supported, for example
4446 * a call from outside of TC ingress.
4447 *
4448 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4449 * Description
4450 * Helper is overloaded depending on BPF program type. This
4451 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4452 *
4453 * Select the *sk* as a result of a socket lookup.
4454 *
4455 * For the operation to succeed passed socket must be compatible
4456 * with the packet description provided by the *ctx* object.
4457 *
4458 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4459 * be an exact match. While IP family (**AF_INET** or
4460 * **AF_INET6**) must be compatible, that is IPv6 sockets
4461 * that are not v6-only can be selected for IPv4 packets.
4462 *
4463 * Only TCP listeners and UDP unconnected sockets can be
4464 * selected. *sk* can also be NULL to reset any previous
4465 * selection.
4466 *
4467 * *flags* argument can combination of following values:
4468 *
4469 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4470 * socket selection, potentially done by a BPF program
4471 * that ran before us.
4472 *
4473 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4474 * load-balancing within reuseport group for the socket
4475 * being selected.
4476 *
4477 * On success *ctx->sk* will point to the selected socket.
4478 *
4479 * Return
4480 * 0 on success, or a negative errno in case of failure.
4481 *
4482 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4483 * not compatible with packet family (*ctx->family*).
4484 *
4485 * * **-EEXIST** if socket has been already selected,
4486 * potentially by another program, and
4487 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4488 *
4489 * * **-EINVAL** if unsupported flags were specified.
4490 *
4491 * * **-EPROTOTYPE** if socket L4 protocol
4492 * (*sk->protocol*) doesn't match packet protocol
4493 * (*ctx->protocol*).
4494 *
4495 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4496 * state (TCP listening or UDP unconnected).
4497 *
4498 * u64 bpf_ktime_get_boot_ns(void)
4499 * Description
4500 * Return the time elapsed since system boot, in nanoseconds.
4501 * Does include the time the system was suspended.
4502 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4503 * Return
4504 * Current *ktime*.
4505 *
4506 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4507 * Description
4508 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4509 * out the format string.
4510 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4511 * the format string itself. The *data* and *data_len* are format string
4512 * arguments. The *data* are a **u64** array and corresponding format string
4513 * values are stored in the array. For strings and pointers where pointees
4514 * are accessed, only the pointer values are stored in the *data* array.
4515 * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4516 *
4517 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4518 * Reading kernel memory may fail due to either invalid address or
4519 * valid address but requiring a major memory fault. If reading kernel memory
4520 * fails, the string for **%s** will be an empty string, and the ip
4521 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4522 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4523 * Return
4524 * 0 on success, or a negative error in case of failure:
4525 *
4526 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4527 * by returning 1 from bpf program.
4528 *
4529 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4530 *
4531 * **-E2BIG** if *fmt* contains too many format specifiers.
4532 *
4533 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4534 *
4535 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4536 * Description
4537 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4538 * The *m* represents the seq_file. The *data* and *len* represent the
4539 * data to write in bytes.
4540 * Return
4541 * 0 on success, or a negative error in case of failure:
4542 *
4543 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4544 *
4545 * u64 bpf_sk_cgroup_id(void *sk)
4546 * Description
4547 * Return the cgroup v2 id of the socket *sk*.
4548 *
4549 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4550 * returned from **bpf_sk_lookup_xxx**\ (),
4551 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4552 * same as in **bpf_skb_cgroup_id**\ ().
4553 *
4554 * This helper is available only if the kernel was compiled with
4555 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4556 * Return
4557 * The id is returned or 0 in case the id could not be retrieved.
4558 *
4559 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4560 * Description
4561 * Return id of cgroup v2 that is ancestor of cgroup associated
4562 * with the *sk* at the *ancestor_level*. The root cgroup is at
4563 * *ancestor_level* zero and each step down the hierarchy
4564 * increments the level. If *ancestor_level* == level of cgroup
4565 * associated with *sk*, then return value will be same as that
4566 * of **bpf_sk_cgroup_id**\ ().
4567 *
4568 * The helper is useful to implement policies based on cgroups
4569 * that are upper in hierarchy than immediate cgroup associated
4570 * with *sk*.
4571 *
4572 * The format of returned id and helper limitations are same as in
4573 * **bpf_sk_cgroup_id**\ ().
4574 * Return
4575 * The id is returned or 0 in case the id could not be retrieved.
4576 *
4577 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4578 * Description
4579 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4580 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4581 * of new data availability is sent.
4582 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4583 * of new data availability is sent unconditionally.
4584 * If **0** is specified in *flags*, an adaptive notification
4585 * of new data availability is sent.
4586 *
4587 * An adaptive notification is a notification sent whenever the user-space
4588 * process has caught up and consumed all available payloads. In case the user-space
4589 * process is still processing a previous payload, then no notification is needed
4590 * as it will process the newly added payload automatically.
4591 * Return
4592 * 0 on success, or a negative error in case of failure.
4593 *
4594 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4595 * Description
4596 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4597 * *flags* must be 0.
4598 * Return
4599 * Valid pointer with *size* bytes of memory available; NULL,
4600 * otherwise.
4601 *
4602 * void bpf_ringbuf_submit(void *data, u64 flags)
4603 * Description
4604 * Submit reserved ring buffer sample, pointed to by *data*.
4605 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4606 * of new data availability is sent.
4607 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4608 * of new data availability is sent unconditionally.
4609 * If **0** is specified in *flags*, an adaptive notification
4610 * of new data availability is sent.
4611 *
4612 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4613 * Return
4614 * Nothing. Always succeeds.
4615 *
4616 * void bpf_ringbuf_discard(void *data, u64 flags)
4617 * Description
4618 * Discard reserved ring buffer sample, pointed to by *data*.
4619 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4620 * of new data availability is sent.
4621 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4622 * of new data availability is sent unconditionally.
4623 * If **0** is specified in *flags*, an adaptive notification
4624 * of new data availability is sent.
4625 *
4626 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4627 * Return
4628 * Nothing. Always succeeds.
4629 *
4630 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4631 * Description
4632 * Query various characteristics of provided ring buffer. What
4633 * exactly is queries is determined by *flags*:
4634 *
4635 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4636 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4637 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4638 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4639 *
4640 * Data returned is just a momentary snapshot of actual values
4641 * and could be inaccurate, so this facility should be used to
4642 * power heuristics and for reporting, not to make 100% correct
4643 * calculation.
4644 * Return
4645 * Requested value, or 0, if *flags* are not recognized.
4646 *
4647 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4648 * Description
4649 * Change the skbs checksum level by one layer up or down, or
4650 * reset it entirely to none in order to have the stack perform
4651 * checksum validation. The level is applicable to the following
4652 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4653 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4654 * through **bpf_skb_adjust_room**\ () helper with passing in
4655 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4656 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4657 * the UDP header is removed. Similarly, an encap of the latter
4658 * into the former could be accompanied by a helper call to
4659 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4660 * skb is still intended to be processed in higher layers of the
4661 * stack instead of just egressing at tc.
4662 *
4663 * There are three supported level settings at this time:
4664 *
4665 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4666 * with CHECKSUM_UNNECESSARY.
4667 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4668 * with CHECKSUM_UNNECESSARY.
4669 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4670 * sets CHECKSUM_NONE to force checksum validation by the stack.
4671 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4672 * skb->csum_level.
4673 * Return
4674 * 0 on success, or a negative error in case of failure. In the
4675 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4676 * is returned or the error code -EACCES in case the skb is not
4677 * subject to CHECKSUM_UNNECESSARY.
4678 *
4679 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4680 * Description
4681 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4682 * Return
4683 * *sk* if casting is valid, or **NULL** otherwise.
4684 *
4685 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4686 * Description
4687 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4688 * Return
4689 * *sk* if casting is valid, or **NULL** otherwise.
4690 *
4691 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4692 * Description
4693 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4694 * Return
4695 * *sk* if casting is valid, or **NULL** otherwise.
4696 *
4697 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4698 * Description
4699 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4700 * Return
4701 * *sk* if casting is valid, or **NULL** otherwise.
4702 *
4703 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4704 * Description
4705 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4706 * Return
4707 * *sk* if casting is valid, or **NULL** otherwise.
4708 *
4709 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4710 * Description
4711 * Return a user or a kernel stack in bpf program provided buffer.
4712 * Note: the user stack will only be populated if the *task* is
4713 * the current task; all other tasks will return -EOPNOTSUPP.
4714 * To achieve this, the helper needs *task*, which is a valid
4715 * pointer to **struct task_struct**. To store the stacktrace, the
4716 * bpf program provides *buf* with a nonnegative *size*.
4717 *
4718 * The last argument, *flags*, holds the number of stack frames to
4719 * skip (from 0 to 255), masked with
4720 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4721 * the following flags:
4722 *
4723 * **BPF_F_USER_STACK**
4724 * Collect a user space stack instead of a kernel stack.
4725 * The *task* must be the current task.
4726 * **BPF_F_USER_BUILD_ID**
4727 * Collect buildid+offset instead of ips for user stack,
4728 * only valid if **BPF_F_USER_STACK** is also specified.
4729 *
4730 * **bpf_get_task_stack**\ () can collect up to
4731 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4732 * to sufficient large buffer size. Note that
4733 * this limit can be controlled with the **sysctl** program, and
4734 * that it should be manually increased in order to profile long
4735 * user stacks (such as stacks for Java programs). To do so, use:
4736 *
4737 * ::
4738 *
4739 * # sysctl kernel.perf_event_max_stack=<new value>
4740 * Return
4741 * The non-negative copied *buf* length equal to or less than
4742 * *size* on success, or a negative error in case of failure.
4743 *
4744 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4745 * Description
4746 * Load header option. Support reading a particular TCP header
4747 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4748 *
4749 * If *flags* is 0, it will search the option from the
4750 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4751 * has details on what skb_data contains under different
4752 * *skops*\ **->op**.
4753 *
4754 * The first byte of the *searchby_res* specifies the
4755 * kind that it wants to search.
4756 *
4757 * If the searching kind is an experimental kind
4758 * (i.e. 253 or 254 according to RFC6994). It also
4759 * needs to specify the "magic" which is either
4760 * 2 bytes or 4 bytes. It then also needs to
4761 * specify the size of the magic by using
4762 * the 2nd byte which is "kind-length" of a TCP
4763 * header option and the "kind-length" also
4764 * includes the first 2 bytes "kind" and "kind-length"
4765 * itself as a normal TCP header option also does.
4766 *
4767 * For example, to search experimental kind 254 with
4768 * 2 byte magic 0xeB9F, the searchby_res should be
4769 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4770 *
4771 * To search for the standard window scale option (3),
4772 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4773 * Note, kind-length must be 0 for regular option.
4774 *
4775 * Searching for No-Op (0) and End-of-Option-List (1) are
4776 * not supported.
4777 *
4778 * *len* must be at least 2 bytes which is the minimal size
4779 * of a header option.
4780 *
4781 * Supported flags:
4782 *
4783 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4784 * saved_syn packet or the just-received syn packet.
4785 *
4786 * Return
4787 * > 0 when found, the header option is copied to *searchby_res*.
4788 * The return value is the total length copied. On failure, a
4789 * negative error code is returned:
4790 *
4791 * **-EINVAL** if a parameter is invalid.
4792 *
4793 * **-ENOMSG** if the option is not found.
4794 *
4795 * **-ENOENT** if no syn packet is available when
4796 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4797 *
4798 * **-ENOSPC** if there is not enough space. Only *len* number of
4799 * bytes are copied.
4800 *
4801 * **-EFAULT** on failure to parse the header options in the
4802 * packet.
4803 *
4804 * **-EPERM** if the helper cannot be used under the current
4805 * *skops*\ **->op**.
4806 *
4807 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4808 * Description
4809 * Store header option. The data will be copied
4810 * from buffer *from* with length *len* to the TCP header.
4811 *
4812 * The buffer *from* should have the whole option that
4813 * includes the kind, kind-length, and the actual
4814 * option data. The *len* must be at least kind-length
4815 * long. The kind-length does not have to be 4 byte
4816 * aligned. The kernel will take care of the padding
4817 * and setting the 4 bytes aligned value to th->doff.
4818 *
4819 * This helper will check for duplicated option
4820 * by searching the same option in the outgoing skb.
4821 *
4822 * This helper can only be called during
4823 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4824 *
4825 * Return
4826 * 0 on success, or negative error in case of failure:
4827 *
4828 * **-EINVAL** If param is invalid.
4829 *
4830 * **-ENOSPC** if there is not enough space in the header.
4831 * Nothing has been written
4832 *
4833 * **-EEXIST** if the option already exists.
4834 *
4835 * **-EFAULT** on failure to parse the existing header options.
4836 *
4837 * **-EPERM** if the helper cannot be used under the current
4838 * *skops*\ **->op**.
4839 *
4840 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4841 * Description
4842 * Reserve *len* bytes for the bpf header option. The
4843 * space will be used by **bpf_store_hdr_opt**\ () later in
4844 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4845 *
4846 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4847 * the total number of bytes will be reserved.
4848 *
4849 * This helper can only be called during
4850 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4851 *
4852 * Return
4853 * 0 on success, or negative error in case of failure:
4854 *
4855 * **-EINVAL** if a parameter is invalid.
4856 *
4857 * **-ENOSPC** if there is not enough space in the header.
4858 *
4859 * **-EPERM** if the helper cannot be used under the current
4860 * *skops*\ **->op**.
4861 *
4862 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4863 * Description
4864 * Get a bpf_local_storage from an *inode*.
4865 *
4866 * Logically, it could be thought of as getting the value from
4867 * a *map* with *inode* as the **key**. From this
4868 * perspective, the usage is not much different from
4869 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4870 * helper enforces the key must be an inode and the map must also
4871 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4872 *
4873 * Underneath, the value is stored locally at *inode* instead of
4874 * the *map*. The *map* is used as the bpf-local-storage
4875 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4876 * searched against all bpf_local_storage residing at *inode*.
4877 *
4878 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4879 * used such that a new bpf_local_storage will be
4880 * created if one does not exist. *value* can be used
4881 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4882 * the initial value of a bpf_local_storage. If *value* is
4883 * **NULL**, the new bpf_local_storage will be zero initialized.
4884 * Return
4885 * A bpf_local_storage pointer is returned on success.
4886 *
4887 * **NULL** if not found or there was an error in adding
4888 * a new bpf_local_storage.
4889 *
4890 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4891 * Description
4892 * Delete a bpf_local_storage from an *inode*.
4893 * Return
4894 * 0 on success.
4895 *
4896 * **-ENOENT** if the bpf_local_storage cannot be found.
4897 *
4898 * long bpf_d_path(const struct path *path, char *buf, u32 sz)
4899 * Description
4900 * Return full path for given **struct path** object, which
4901 * needs to be the kernel BTF *path* object. The path is
4902 * returned in the provided buffer *buf* of size *sz* and
4903 * is zero terminated.
4904 *
4905 * Return
4906 * On success, the strictly positive length of the string,
4907 * including the trailing NUL character. On error, a negative
4908 * value.
4909 *
4910 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4911 * Description
4912 * Read *size* bytes from user space address *user_ptr* and store
4913 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4914 * Return
4915 * 0 on success, or a negative error in case of failure.
4916 *
4917 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4918 * Description
4919 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4920 * using *ptr*->type_id. This value should specify the type
4921 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4922 * can be used to look up vmlinux BTF type ids. Traversing the
4923 * data structure using BTF, the type information and values are
4924 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4925 * the pointer data is carried out to avoid kernel crashes during
4926 * operation. Smaller types can use string space on the stack;
4927 * larger programs can use map data to store the string
4928 * representation.
4929 *
4930 * The string can be subsequently shared with userspace via
4931 * bpf_perf_event_output() or ring buffer interfaces.
4932 * bpf_trace_printk() is to be avoided as it places too small
4933 * a limit on string size to be useful.
4934 *
4935 * *flags* is a combination of
4936 *
4937 * **BTF_F_COMPACT**
4938 * no formatting around type information
4939 * **BTF_F_NONAME**
4940 * no struct/union member names/types
4941 * **BTF_F_PTR_RAW**
4942 * show raw (unobfuscated) pointer values;
4943 * equivalent to printk specifier %px.
4944 * **BTF_F_ZERO**
4945 * show zero-valued struct/union members; they
4946 * are not displayed by default
4947 *
4948 * Return
4949 * The number of bytes that were written (or would have been
4950 * written if output had to be truncated due to string size),
4951 * or a negative error in cases of failure.
4952 *
4953 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4954 * Description
4955 * Use BTF to write to seq_write a string representation of
4956 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4957 * *flags* are identical to those used for bpf_snprintf_btf.
4958 * Return
4959 * 0 on success or a negative error in case of failure.
4960 *
4961 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4962 * Description
4963 * See **bpf_get_cgroup_classid**\ () for the main description.
4964 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4965 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4966 * associated socket instead of the current process.
4967 * Return
4968 * The id is returned or 0 in case the id could not be retrieved.
4969 *
4970 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4971 * Description
4972 * Redirect the packet to another net device of index *ifindex*
4973 * and fill in L2 addresses from neighboring subsystem. This helper
4974 * is somewhat similar to **bpf_redirect**\ (), except that it
4975 * populates L2 addresses as well, meaning, internally, the helper
4976 * relies on the neighbor lookup for the L2 address of the nexthop.
4977 *
4978 * The helper will perform a FIB lookup based on the skb's
4979 * networking header to get the address of the next hop, unless
4980 * this is supplied by the caller in the *params* argument. The
4981 * *plen* argument indicates the len of *params* and should be set
4982 * to 0 if *params* is NULL.
4983 *
4984 * The *flags* argument is reserved and must be 0. The helper is
4985 * currently only supported for tc BPF program types, and enabled
4986 * for IPv4 and IPv6 protocols.
4987 * Return
4988 * The helper returns **TC_ACT_REDIRECT** on success or
4989 * **TC_ACT_SHOT** on error.
4990 *
4991 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4992 * Description
4993 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4994 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4995 * extern variable decorated with '__ksym'. For ksym, there is a
4996 * global var (either static or global) defined of the same name
4997 * in the kernel. The ksym is percpu if the global var is percpu.
4998 * The returned pointer points to the global percpu var on *cpu*.
4999 *
5000 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
5001 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
5002 * happens if *cpu* is larger than nr_cpu_ids. The caller of
5003 * bpf_per_cpu_ptr() must check the returned value.
5004 * Return
5005 * A pointer pointing to the kernel percpu variable on *cpu*, or
5006 * NULL, if *cpu* is invalid.
5007 *
5008 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
5009 * Description
5010 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
5011 * pointer to the percpu kernel variable on this cpu. See the
5012 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
5013 *
5014 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
5015 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
5016 * never return NULL.
5017 * Return
5018 * A pointer pointing to the kernel percpu variable on this cpu.
5019 *
5020 * long bpf_redirect_peer(u32 ifindex, u64 flags)
5021 * Description
5022 * Redirect the packet to another net device of index *ifindex*.
5023 * This helper is somewhat similar to **bpf_redirect**\ (), except
5024 * that the redirection happens to the *ifindex*' peer device and
5025 * the netns switch takes place from ingress to ingress without
5026 * going through the CPU's backlog queue.
5027 *
5028 * *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during
5029 * the netns switch.
5030 *
5031 * The *flags* argument is reserved and must be 0. The helper is
5032 * currently only supported for tc BPF program types at the
5033 * ingress hook and for veth and netkit target device types. The
5034 * peer device must reside in a different network namespace.
5035 * Return
5036 * The helper returns **TC_ACT_REDIRECT** on success or
5037 * **TC_ACT_SHOT** on error.
5038 *
5039 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
5040 * Description
5041 * Get a bpf_local_storage from the *task*.
5042 *
5043 * Logically, it could be thought of as getting the value from
5044 * a *map* with *task* as the **key**. From this
5045 * perspective, the usage is not much different from
5046 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
5047 * helper enforces the key must be a task_struct and the map must also
5048 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
5049 *
5050 * Underneath, the value is stored locally at *task* instead of
5051 * the *map*. The *map* is used as the bpf-local-storage
5052 * "type". The bpf-local-storage "type" (i.e. the *map*) is
5053 * searched against all bpf_local_storage residing at *task*.
5054 *
5055 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5056 * used such that a new bpf_local_storage will be
5057 * created if one does not exist. *value* can be used
5058 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5059 * the initial value of a bpf_local_storage. If *value* is
5060 * **NULL**, the new bpf_local_storage will be zero initialized.
5061 * Return
5062 * A bpf_local_storage pointer is returned on success.
5063 *
5064 * **NULL** if not found or there was an error in adding
5065 * a new bpf_local_storage.
5066 *
5067 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
5068 * Description
5069 * Delete a bpf_local_storage from a *task*.
5070 * Return
5071 * 0 on success.
5072 *
5073 * **-ENOENT** if the bpf_local_storage cannot be found.
5074 *
5075 * struct task_struct *bpf_get_current_task_btf(void)
5076 * Description
5077 * Return a BTF pointer to the "current" task.
5078 * This pointer can also be used in helpers that accept an
5079 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
5080 * Return
5081 * Pointer to the current task.
5082 *
5083 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
5084 * Description
5085 * Set or clear certain options on *bprm*:
5086 *
5087 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
5088 * which sets the **AT_SECURE** auxv for glibc. The bit
5089 * is cleared if the flag is not specified.
5090 * Return
5091 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
5092 *
5093 * u64 bpf_ktime_get_coarse_ns(void)
5094 * Description
5095 * Return a coarse-grained version of the time elapsed since
5096 * system boot, in nanoseconds. Does not include time the system
5097 * was suspended.
5098 *
5099 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
5100 * Return
5101 * Current *ktime*.
5102 *
5103 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
5104 * Description
5105 * Returns the stored IMA hash of the *inode* (if it's available).
5106 * If the hash is larger than *size*, then only *size*
5107 * bytes will be copied to *dst*
5108 * Return
5109 * The **hash_algo** is returned on success,
5110 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if
5111 * invalid arguments are passed.
5112 *
5113 * struct socket *bpf_sock_from_file(struct file *file)
5114 * Description
5115 * If the given file represents a socket, returns the associated
5116 * socket.
5117 * Return
5118 * A pointer to a struct socket on success or NULL if the file is
5119 * not a socket.
5120 *
5121 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
5122 * Description
5123 * Check packet size against exceeding MTU of net device (based
5124 * on *ifindex*). This helper will likely be used in combination
5125 * with helpers that adjust/change the packet size.
5126 *
5127 * The argument *len_diff* can be used for querying with a planned
5128 * size change. This allows to check MTU prior to changing packet
5129 * ctx. Providing a *len_diff* adjustment that is larger than the
5130 * actual packet size (resulting in negative packet size) will in
5131 * principle not exceed the MTU, which is why it is not considered
5132 * a failure. Other BPF helpers are needed for performing the
5133 * planned size change; therefore the responsibility for catching
5134 * a negative packet size belongs in those helpers.
5135 *
5136 * Specifying *ifindex* zero means the MTU check is performed
5137 * against the current net device. This is practical if this isn't
5138 * used prior to redirect.
5139 *
5140 * On input *mtu_len* must be a valid pointer, else verifier will
5141 * reject BPF program. If the value *mtu_len* is initialized to
5142 * zero then the ctx packet size is use. When value *mtu_len* is
5143 * provided as input this specify the L3 length that the MTU check
5144 * is done against. Remember XDP and TC length operate at L2, but
5145 * this value is L3 as this correlate to MTU and IP-header tot_len
5146 * values which are L3 (similar behavior as bpf_fib_lookup).
5147 *
5148 * The Linux kernel route table can configure MTUs on a more
5149 * specific per route level, which is not provided by this helper.
5150 * For route level MTU checks use the **bpf_fib_lookup**\ ()
5151 * helper.
5152 *
5153 * *ctx* is either **struct xdp_md** for XDP programs or
5154 * **struct sk_buff** for tc cls_act programs.
5155 *
5156 * The *flags* argument can be a combination of one or more of the
5157 * following values:
5158 *
5159 * **BPF_MTU_CHK_SEGS**
5160 * This flag will only works for *ctx* **struct sk_buff**.
5161 * If packet context contains extra packet segment buffers
5162 * (often knows as GSO skb), then MTU check is harder to
5163 * check at this point, because in transmit path it is
5164 * possible for the skb packet to get re-segmented
5165 * (depending on net device features). This could still be
5166 * a MTU violation, so this flag enables performing MTU
5167 * check against segments, with a different violation
5168 * return code to tell it apart. Check cannot use len_diff.
5169 *
5170 * On return *mtu_len* pointer contains the MTU value of the net
5171 * device. Remember the net device configured MTU is the L3 size,
5172 * which is returned here and XDP and TC length operate at L2.
5173 * Helper take this into account for you, but remember when using
5174 * MTU value in your BPF-code.
5175 *
5176 * Return
5177 * * 0 on success, and populate MTU value in *mtu_len* pointer.
5178 *
5179 * * < 0 if any input argument is invalid (*mtu_len* not updated)
5180 *
5181 * MTU violations return positive values, but also populate MTU
5182 * value in *mtu_len* pointer, as this can be needed for
5183 * implementing PMTU handing:
5184 *
5185 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
5186 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
5187 *
5188 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
5189 * Description
5190 * For each element in **map**, call **callback_fn** function with
5191 * **map**, **callback_ctx** and other map-specific parameters.
5192 * The **callback_fn** should be a static function and
5193 * the **callback_ctx** should be a pointer to the stack.
5194 * The **flags** is used to control certain aspects of the helper.
5195 * Currently, the **flags** must be 0.
5196 *
5197 * The following are a list of supported map types and their
5198 * respective expected callback signatures:
5199 *
5200 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
5201 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
5202 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
5203 *
5204 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
5205 *
5206 * For per_cpu maps, the map_value is the value on the cpu where the
5207 * bpf_prog is running.
5208 *
5209 * If **callback_fn** return 0, the helper will continue to the next
5210 * element. If return value is 1, the helper will skip the rest of
5211 * elements and return. Other return values are not used now.
5212 *
5213 * Return
5214 * The number of traversed map elements for success, **-EINVAL** for
5215 * invalid **flags**.
5216 *
5217 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
5218 * Description
5219 * Outputs a string into the **str** buffer of size **str_size**
5220 * based on a format string stored in a read-only map pointed by
5221 * **fmt**.
5222 *
5223 * Each format specifier in **fmt** corresponds to one u64 element
5224 * in the **data** array. For strings and pointers where pointees
5225 * are accessed, only the pointer values are stored in the *data*
5226 * array. The *data_len* is the size of *data* in bytes - must be
5227 * a multiple of 8.
5228 *
5229 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5230 * memory. Reading kernel memory may fail due to either invalid
5231 * address or valid address but requiring a major memory fault. If
5232 * reading kernel memory fails, the string for **%s** will be an
5233 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5234 * Not returning error to bpf program is consistent with what
5235 * **bpf_trace_printk**\ () does for now.
5236 *
5237 * Return
5238 * The strictly positive length of the formatted string, including
5239 * the trailing zero character. If the return value is greater than
5240 * **str_size**, **str** contains a truncated string, guaranteed to
5241 * be zero-terminated except when **str_size** is 0.
5242 *
5243 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5244 *
5245 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5246 * Description
5247 * Execute bpf syscall with given arguments.
5248 * Return
5249 * A syscall result.
5250 *
5251 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5252 * Description
5253 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5254 * Return
5255 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5256 *
5257 * long bpf_sys_close(u32 fd)
5258 * Description
5259 * Execute close syscall for given FD.
5260 * Return
5261 * A syscall result.
5262 *
5263 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5264 * Description
5265 * Initialize the timer.
5266 * First 4 bits of *flags* specify clockid.
5267 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5268 * All other bits of *flags* are reserved.
5269 * The verifier will reject the program if *timer* is not from
5270 * the same *map*.
5271 * Return
5272 * 0 on success.
5273 * **-EBUSY** if *timer* is already initialized.
5274 * **-EINVAL** if invalid *flags* are passed.
5275 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5276 * The user space should either hold a file descriptor to a map with timers
5277 * or pin such map in bpffs. When map is unpinned or file descriptor is
5278 * closed all timers in the map will be cancelled and freed.
5279 *
5280 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5281 * Description
5282 * Configure the timer to call *callback_fn* static function.
5283 * Return
5284 * 0 on success.
5285 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5286 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5287 * The user space should either hold a file descriptor to a map with timers
5288 * or pin such map in bpffs. When map is unpinned or file descriptor is
5289 * closed all timers in the map will be cancelled and freed.
5290 *
5291 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5292 * Description
5293 * Set timer expiration N nanoseconds from the current time. The
5294 * configured callback will be invoked in soft irq context on some cpu
5295 * and will not repeat unless another bpf_timer_start() is made.
5296 * In such case the next invocation can migrate to a different cpu.
5297 * Since struct bpf_timer is a field inside map element the map
5298 * owns the timer. The bpf_timer_set_callback() will increment refcnt
5299 * of BPF program to make sure that callback_fn code stays valid.
5300 * When user space reference to a map reaches zero all timers
5301 * in a map are cancelled and corresponding program's refcnts are
5302 * decremented. This is done to make sure that Ctrl-C of a user
5303 * process doesn't leave any timers running. If map is pinned in
5304 * bpffs the callback_fn can re-arm itself indefinitely.
5305 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5306 * cancel and free the timer in the given map element.
5307 * The map can contain timers that invoke callback_fn-s from different
5308 * programs. The same callback_fn can serve different timers from
5309 * different maps if key/value layout matches across maps.
5310 * Every bpf_timer_set_callback() can have different callback_fn.
5311 *
5312 * *flags* can be one of:
5313 *
5314 * **BPF_F_TIMER_ABS**
5315 * Start the timer in absolute expire value instead of the
5316 * default relative one.
5317 * **BPF_F_TIMER_CPU_PIN**
5318 * Timer will be pinned to the CPU of the caller.
5319 *
5320 * Return
5321 * 0 on success.
5322 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5323 * or invalid *flags* are passed.
5324 *
5325 * long bpf_timer_cancel(struct bpf_timer *timer)
5326 * Description
5327 * Cancel the timer and wait for callback_fn to finish if it was running.
5328 * Return
5329 * 0 if the timer was not active.
5330 * 1 if the timer was active.
5331 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5332 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5333 * own timer which would have led to a deadlock otherwise.
5334 *
5335 * u64 bpf_get_func_ip(void *ctx)
5336 * Description
5337 * Get address of the traced function (for tracing and kprobe programs).
5338 *
5339 * When called for kprobe program attached as uprobe it returns
5340 * probe address for both entry and return uprobe.
5341 *
5342 * Return
5343 * Address of the traced function for kprobe.
5344 * 0 for kprobes placed within the function (not at the entry).
5345 * Address of the probe for uprobe and return uprobe.
5346 *
5347 * u64 bpf_get_attach_cookie(void *ctx)
5348 * Description
5349 * Get bpf_cookie value provided (optionally) during the program
5350 * attachment. It might be different for each individual
5351 * attachment, even if BPF program itself is the same.
5352 * Expects BPF program context *ctx* as a first argument.
5353 *
5354 * Supported for the following program types:
5355 * - kprobe/uprobe;
5356 * - tracepoint;
5357 * - perf_event.
5358 * Return
5359 * Value specified by user at BPF link creation/attachment time
5360 * or 0, if it was not specified.
5361 *
5362 * long bpf_task_pt_regs(struct task_struct *task)
5363 * Description
5364 * Get the struct pt_regs associated with **task**.
5365 * Return
5366 * A pointer to struct pt_regs.
5367 *
5368 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5369 * Description
5370 * Get branch trace from hardware engines like Intel LBR. The
5371 * hardware engine is stopped shortly after the helper is
5372 * called. Therefore, the user need to filter branch entries
5373 * based on the actual use case. To capture branch trace
5374 * before the trigger point of the BPF program, the helper
5375 * should be called at the beginning of the BPF program.
5376 *
5377 * The data is stored as struct perf_branch_entry into output
5378 * buffer *entries*. *size* is the size of *entries* in bytes.
5379 * *flags* is reserved for now and must be zero.
5380 *
5381 * Return
5382 * On success, number of bytes written to *buf*. On error, a
5383 * negative value.
5384 *
5385 * **-EINVAL** if *flags* is not zero.
5386 *
5387 * **-ENOENT** if architecture does not support branch records.
5388 *
5389 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5390 * Description
5391 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5392 * to format and can handle more format args as a result.
5393 *
5394 * Arguments are to be used as in **bpf_seq_printf**\ () helper.
5395 * Return
5396 * The number of bytes written to the buffer, or a negative error
5397 * in case of failure.
5398 *
5399 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5400 * Description
5401 * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5402 * Return
5403 * *sk* if casting is valid, or **NULL** otherwise.
5404 *
5405 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5406 * Description
5407 * Get the address of a kernel symbol, returned in *res*. *res* is
5408 * set to 0 if the symbol is not found.
5409 * Return
5410 * On success, zero. On error, a negative value.
5411 *
5412 * **-EINVAL** if *flags* is not zero.
5413 *
5414 * **-EINVAL** if string *name* is not the same size as *name_sz*.
5415 *
5416 * **-ENOENT** if symbol is not found.
5417 *
5418 * **-EPERM** if caller does not have permission to obtain kernel address.
5419 *
5420 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5421 * Description
5422 * Find vma of *task* that contains *addr*, call *callback_fn*
5423 * function with *task*, *vma*, and *callback_ctx*.
5424 * The *callback_fn* should be a static function and
5425 * the *callback_ctx* should be a pointer to the stack.
5426 * The *flags* is used to control certain aspects of the helper.
5427 * Currently, the *flags* must be 0.
5428 *
5429 * The expected callback signature is
5430 *
5431 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5432 *
5433 * Return
5434 * 0 on success.
5435 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5436 * **-EBUSY** if failed to try lock mmap_lock.
5437 * **-EINVAL** for invalid **flags**.
5438 *
5439 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5440 * Description
5441 * For **nr_loops**, call **callback_fn** function
5442 * with **callback_ctx** as the context parameter.
5443 * The **callback_fn** should be a static function and
5444 * the **callback_ctx** should be a pointer to the stack.
5445 * The **flags** is used to control certain aspects of the helper.
5446 * Currently, the **flags** must be 0. Currently, nr_loops is
5447 * limited to 1 << 23 (~8 million) loops.
5448 *
5449 * long (\*callback_fn)(u64 index, void \*ctx);
5450 *
5451 * where **index** is the current index in the loop. The index
5452 * is zero-indexed.
5453 *
5454 * If **callback_fn** returns 0, the helper will continue to the next
5455 * loop. If return value is 1, the helper will skip the rest of
5456 * the loops and return. Other return values are not used now,
5457 * and will be rejected by the verifier.
5458 *
5459 * Return
5460 * The number of loops performed, **-EINVAL** for invalid **flags**,
5461 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5462 *
5463 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5464 * Description
5465 * Do strncmp() between **s1** and **s2**. **s1** doesn't need
5466 * to be null-terminated and **s1_sz** is the maximum storage
5467 * size of **s1**. **s2** must be a read-only string.
5468 * Return
5469 * An integer less than, equal to, or greater than zero
5470 * if the first **s1_sz** bytes of **s1** is found to be
5471 * less than, to match, or be greater than **s2**.
5472 *
5473 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5474 * Description
5475 * Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5476 * returned in **value**.
5477 *
5478 * Return
5479 * 0 on success.
5480 * **-EINVAL** if n >= argument register count of traced function.
5481 *
5482 * long bpf_get_func_ret(void *ctx, u64 *value)
5483 * Description
5484 * Get return value of the traced function (for tracing programs)
5485 * in **value**.
5486 *
5487 * Return
5488 * 0 on success.
5489 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5490 *
5491 * long bpf_get_func_arg_cnt(void *ctx)
5492 * Description
5493 * Get number of registers of the traced function (for tracing programs) where
5494 * function arguments are stored in these registers.
5495 *
5496 * Return
5497 * The number of argument registers of the traced function.
5498 *
5499 * int bpf_get_retval(void)
5500 * Description
5501 * Get the BPF program's return value that will be returned to the upper layers.
5502 *
5503 * This helper is currently supported by cgroup programs and only by the hooks
5504 * where BPF program's return value is returned to the userspace via errno.
5505 * Return
5506 * The BPF program's return value.
5507 *
5508 * int bpf_set_retval(int retval)
5509 * Description
5510 * Set the BPF program's return value that will be returned to the upper layers.
5511 *
5512 * This helper is currently supported by cgroup programs and only by the hooks
5513 * where BPF program's return value is returned to the userspace via errno.
5514 *
5515 * Note that there is the following corner case where the program exports an error
5516 * via bpf_set_retval but signals success via 'return 1':
5517 *
5518 * bpf_set_retval(-EPERM);
5519 * return 1;
5520 *
5521 * In this case, the BPF program's return value will use helper's -EPERM. This
5522 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5523 *
5524 * Return
5525 * 0 on success, or a negative error in case of failure.
5526 *
5527 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5528 * Description
5529 * Get the total size of a given xdp buff (linear and paged area)
5530 * Return
5531 * The total size of a given xdp buffer.
5532 *
5533 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5534 * Description
5535 * This helper is provided as an easy way to load data from a
5536 * xdp buffer. It can be used to load *len* bytes from *offset* from
5537 * the frame associated to *xdp_md*, into the buffer pointed by
5538 * *buf*.
5539 * Return
5540 * 0 on success, or a negative error in case of failure.
5541 *
5542 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5543 * Description
5544 * Store *len* bytes from buffer *buf* into the frame
5545 * associated to *xdp_md*, at *offset*.
5546 * Return
5547 * 0 on success, or a negative error in case of failure.
5548 *
5549 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5550 * Description
5551 * Read *size* bytes from user space address *user_ptr* in *tsk*'s
5552 * address space, and stores the data in *dst*. *flags* is not
5553 * used yet and is provided for future extensibility. This helper
5554 * can only be used by sleepable programs.
5555 * Return
5556 * 0 on success, or a negative error in case of failure. On error
5557 * *dst* buffer is zeroed out.
5558 *
5559 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5560 * Description
5561 * Change the __sk_buff->tstamp_type to *tstamp_type*
5562 * and set *tstamp* to the __sk_buff->tstamp together.
5563 *
5564 * If there is no need to change the __sk_buff->tstamp_type,
5565 * the tstamp value can be directly written to __sk_buff->tstamp
5566 * instead.
5567 *
5568 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5569 * will be kept during bpf_redirect_*(). A non zero
5570 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5571 * *tstamp_type*.
5572 *
5573 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5574 * with a zero *tstamp*.
5575 *
5576 * Only IPv4 and IPv6 skb->protocol are supported.
5577 *
5578 * This function is most useful when it needs to set a
5579 * mono delivery time to __sk_buff->tstamp and then
5580 * bpf_redirect_*() to the egress of an iface. For example,
5581 * changing the (rcv) timestamp in __sk_buff->tstamp at
5582 * ingress to a mono delivery time and then bpf_redirect_*()
5583 * to sch_fq@phy-dev.
5584 * Return
5585 * 0 on success.
5586 * **-EINVAL** for invalid input
5587 * **-EOPNOTSUPP** for unsupported protocol
5588 *
5589 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5590 * Description
5591 * Returns a calculated IMA hash of the *file*.
5592 * If the hash is larger than *size*, then only *size*
5593 * bytes will be copied to *dst*
5594 * Return
5595 * The **hash_algo** is returned on success,
5596 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if
5597 * invalid arguments are passed.
5598 *
5599 * void *bpf_kptr_xchg(void *dst, void *ptr)
5600 * Description
5601 * Exchange kptr at pointer *dst* with *ptr*, and return the old value.
5602 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise
5603 * it must be a referenced pointer which will be released when this helper
5604 * is called.
5605 * Return
5606 * The old value of kptr (which can be NULL). The returned pointer
5607 * if not NULL, is a reference which must be released using its
5608 * corresponding release function, or moved into a BPF map before
5609 * program exit.
5610 *
5611 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5612 * Description
5613 * Perform a lookup in *percpu map* for an entry associated to
5614 * *key* on *cpu*.
5615 * Return
5616 * Map value associated to *key* on *cpu*, or **NULL** if no entry
5617 * was found or *cpu* is invalid.
5618 *
5619 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5620 * Description
5621 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5622 * Return
5623 * *sk* if casting is valid, or **NULL** otherwise.
5624 *
5625 * long bpf_dynptr_from_mem(void *data, u64 size, u64 flags, struct bpf_dynptr *ptr)
5626 * Description
5627 * Get a dynptr to local memory *data*.
5628 *
5629 * *data* must be a ptr to a map value.
5630 * The maximum *size* supported is DYNPTR_MAX_SIZE.
5631 * *flags* is currently unused.
5632 * Return
5633 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5634 * -EINVAL if flags is not 0.
5635 *
5636 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5637 * Description
5638 * Reserve *size* bytes of payload in a ring buffer *ringbuf*
5639 * through the dynptr interface. *flags* must be 0.
5640 *
5641 * Please note that a corresponding bpf_ringbuf_submit_dynptr or
5642 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5643 * reservation fails. This is enforced by the verifier.
5644 * Return
5645 * 0 on success, or a negative error in case of failure.
5646 *
5647 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5648 * Description
5649 * Submit reserved ring buffer sample, pointed to by *data*,
5650 * through the dynptr interface. This is a no-op if the dynptr is
5651 * invalid/null.
5652 *
5653 * For more information on *flags*, please see
5654 * 'bpf_ringbuf_submit'.
5655 * Return
5656 * Nothing. Always succeeds.
5657 *
5658 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5659 * Description
5660 * Discard reserved ring buffer sample through the dynptr
5661 * interface. This is a no-op if the dynptr is invalid/null.
5662 *
5663 * For more information on *flags*, please see
5664 * 'bpf_ringbuf_discard'.
5665 * Return
5666 * Nothing. Always succeeds.
5667 *
5668 * long bpf_dynptr_read(void *dst, u64 len, const struct bpf_dynptr *src, u64 offset, u64 flags)
5669 * Description
5670 * Read *len* bytes from *src* into *dst*, starting from *offset*
5671 * into *src*.
5672 * *flags* is currently unused.
5673 * Return
5674 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5675 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5676 * *flags* is not 0.
5677 *
5678 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u64 offset, void *src, u64 len, u64 flags)
5679 * Description
5680 * Write *len* bytes from *src* into *dst*, starting from *offset*
5681 * into *dst*.
5682 *
5683 * *flags* must be 0 except for skb-type dynptrs.
5684 *
5685 * For skb-type dynptrs:
5686 * * All data slices of the dynptr are automatically
5687 * invalidated after **bpf_dynptr_write**\ (). This is
5688 * because writing may pull the skb and change the
5689 * underlying packet buffer.
5690 *
5691 * * For *flags*, please see the flags accepted by
5692 * **bpf_skb_store_bytes**\ ().
5693 * Return
5694 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5695 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5696 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5697 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5698 *
5699 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u64 offset, u64 len)
5700 * Description
5701 * Get a pointer to the underlying dynptr data.
5702 *
5703 * *len* must be a statically known value. The returned data slice
5704 * is invalidated whenever the dynptr is invalidated.
5705 *
5706 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5707 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5708 * Return
5709 * Pointer to the underlying dynptr data, NULL if the dynptr is
5710 * read-only, if the dynptr is invalid, or if the offset and length
5711 * is out of bounds.
5712 *
5713 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5714 * Description
5715 * Try to issue a SYN cookie for the packet with corresponding
5716 * IPv4/TCP headers, *iph* and *th*, without depending on a
5717 * listening socket.
5718 *
5719 * *iph* points to the IPv4 header.
5720 *
5721 * *th* points to the start of the TCP header, while *th_len*
5722 * contains the length of the TCP header (at least
5723 * **sizeof**\ (**struct tcphdr**)).
5724 * Return
5725 * On success, lower 32 bits hold the generated SYN cookie in
5726 * followed by 16 bits which hold the MSS value for that cookie,
5727 * and the top 16 bits are unused.
5728 *
5729 * On failure, the returned value is one of the following:
5730 *
5731 * **-EINVAL** if *th_len* is invalid.
5732 *
5733 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5734 * Description
5735 * Try to issue a SYN cookie for the packet with corresponding
5736 * IPv6/TCP headers, *iph* and *th*, without depending on a
5737 * listening socket.
5738 *
5739 * *iph* points to the IPv6 header.
5740 *
5741 * *th* points to the start of the TCP header, while *th_len*
5742 * contains the length of the TCP header (at least
5743 * **sizeof**\ (**struct tcphdr**)).
5744 * Return
5745 * On success, lower 32 bits hold the generated SYN cookie in
5746 * followed by 16 bits which hold the MSS value for that cookie,
5747 * and the top 16 bits are unused.
5748 *
5749 * On failure, the returned value is one of the following:
5750 *
5751 * **-EINVAL** if *th_len* is invalid.
5752 *
5753 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5754 *
5755 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5756 * Description
5757 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5758 * without depending on a listening socket.
5759 *
5760 * *iph* points to the IPv4 header.
5761 *
5762 * *th* points to the TCP header.
5763 * Return
5764 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5765 *
5766 * On failure, the returned value is one of the following:
5767 *
5768 * **-EACCES** if the SYN cookie is not valid.
5769 *
5770 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5771 * Description
5772 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5773 * without depending on a listening socket.
5774 *
5775 * *iph* points to the IPv6 header.
5776 *
5777 * *th* points to the TCP header.
5778 * Return
5779 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5780 *
5781 * On failure, the returned value is one of the following:
5782 *
5783 * **-EACCES** if the SYN cookie is not valid.
5784 *
5785 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5786 *
5787 * u64 bpf_ktime_get_tai_ns(void)
5788 * Description
5789 * A nonsettable system-wide clock derived from wall-clock time but
5790 * ignoring leap seconds. This clock does not experience
5791 * discontinuities and backwards jumps caused by NTP inserting leap
5792 * seconds as CLOCK_REALTIME does.
5793 *
5794 * See: **clock_gettime**\ (**CLOCK_TAI**)
5795 * Return
5796 * Current *ktime*.
5797 *
5798 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5799 * Description
5800 * Drain samples from the specified user ring buffer, and invoke
5801 * the provided callback for each such sample:
5802 *
5803 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5804 *
5805 * If **callback_fn** returns 0, the helper will continue to try
5806 * and drain the next sample, up to a maximum of
5807 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5808 * the helper will skip the rest of the samples and return. Other
5809 * return values are not used now, and will be rejected by the
5810 * verifier.
5811 * Return
5812 * The number of drained samples if no error was encountered while
5813 * draining samples, or 0 if no samples were present in the ring
5814 * buffer. If a user-space producer was epoll-waiting on this map,
5815 * and at least one sample was drained, they will receive an event
5816 * notification notifying them of available space in the ring
5817 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5818 * function, no wakeup notification will be sent. If the
5819 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5820 * be sent even if no sample was drained.
5821 *
5822 * On failure, the returned value is one of the following:
5823 *
5824 * **-EBUSY** if the ring buffer is contended, and another calling
5825 * context was concurrently draining the ring buffer.
5826 *
5827 * **-EINVAL** if user-space is not properly tracking the ring
5828 * buffer due to the producer position not being aligned to 8
5829 * bytes, a sample not being aligned to 8 bytes, or the producer
5830 * position not matching the advertised length of a sample.
5831 *
5832 * **-E2BIG** if user-space has tried to publish a sample which is
5833 * larger than the size of the ring buffer, or which cannot fit
5834 * within a struct bpf_dynptr.
5835 *
5836 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5837 * Description
5838 * Get a bpf_local_storage from the *cgroup*.
5839 *
5840 * Logically, it could be thought of as getting the value from
5841 * a *map* with *cgroup* as the **key**. From this
5842 * perspective, the usage is not much different from
5843 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5844 * helper enforces the key must be a cgroup struct and the map must also
5845 * be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5846 *
5847 * In reality, the local-storage value is embedded directly inside of the
5848 * *cgroup* object itself, rather than being located in the
5849 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5850 * queried for some *map* on a *cgroup* object, the kernel will perform an
5851 * O(n) iteration over all of the live local-storage values for that
5852 * *cgroup* object until the local-storage value for the *map* is found.
5853 *
5854 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5855 * used such that a new bpf_local_storage will be
5856 * created if one does not exist. *value* can be used
5857 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5858 * the initial value of a bpf_local_storage. If *value* is
5859 * **NULL**, the new bpf_local_storage will be zero initialized.
5860 * Return
5861 * A bpf_local_storage pointer is returned on success.
5862 *
5863 * **NULL** if not found or there was an error in adding
5864 * a new bpf_local_storage.
5865 *
5866 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5867 * Description
5868 * Delete a bpf_local_storage from a *cgroup*.
5869 * Return
5870 * 0 on success.
5871 *
5872 * **-ENOENT** if the bpf_local_storage cannot be found.
5873 */
5874#define ___BPF_FUNC_MAPPER(FN, ctx...) \
5875 FN(unspec, 0, ##ctx) \
5876 FN(map_lookup_elem, 1, ##ctx) \
5877 FN(map_update_elem, 2, ##ctx) \
5878 FN(map_delete_elem, 3, ##ctx) \
5879 FN(probe_read, 4, ##ctx) \
5880 FN(ktime_get_ns, 5, ##ctx) \
5881 FN(trace_printk, 6, ##ctx) \
5882 FN(get_prandom_u32, 7, ##ctx) \
5883 FN(get_smp_processor_id, 8, ##ctx) \
5884 FN(skb_store_bytes, 9, ##ctx) \
5885 FN(l3_csum_replace, 10, ##ctx) \
5886 FN(l4_csum_replace, 11, ##ctx) \
5887 FN(tail_call, 12, ##ctx) \
5888 FN(clone_redirect, 13, ##ctx) \
5889 FN(get_current_pid_tgid, 14, ##ctx) \
5890 FN(get_current_uid_gid, 15, ##ctx) \
5891 FN(get_current_comm, 16, ##ctx) \
5892 FN(get_cgroup_classid, 17, ##ctx) \
5893 FN(skb_vlan_push, 18, ##ctx) \
5894 FN(skb_vlan_pop, 19, ##ctx) \
5895 FN(skb_get_tunnel_key, 20, ##ctx) \
5896 FN(skb_set_tunnel_key, 21, ##ctx) \
5897 FN(perf_event_read, 22, ##ctx) \
5898 FN(redirect, 23, ##ctx) \
5899 FN(get_route_realm, 24, ##ctx) \
5900 FN(perf_event_output, 25, ##ctx) \
5901 FN(skb_load_bytes, 26, ##ctx) \
5902 FN(get_stackid, 27, ##ctx) \
5903 FN(csum_diff, 28, ##ctx) \
5904 FN(skb_get_tunnel_opt, 29, ##ctx) \
5905 FN(skb_set_tunnel_opt, 30, ##ctx) \
5906 FN(skb_change_proto, 31, ##ctx) \
5907 FN(skb_change_type, 32, ##ctx) \
5908 FN(skb_under_cgroup, 33, ##ctx) \
5909 FN(get_hash_recalc, 34, ##ctx) \
5910 FN(get_current_task, 35, ##ctx) \
5911 FN(probe_write_user, 36, ##ctx) \
5912 FN(current_task_under_cgroup, 37, ##ctx) \
5913 FN(skb_change_tail, 38, ##ctx) \
5914 FN(skb_pull_data, 39, ##ctx) \
5915 FN(csum_update, 40, ##ctx) \
5916 FN(set_hash_invalid, 41, ##ctx) \
5917 FN(get_numa_node_id, 42, ##ctx) \
5918 FN(skb_change_head, 43, ##ctx) \
5919 FN(xdp_adjust_head, 44, ##ctx) \
5920 FN(probe_read_str, 45, ##ctx) \
5921 FN(get_socket_cookie, 46, ##ctx) \
5922 FN(get_socket_uid, 47, ##ctx) \
5923 FN(set_hash, 48, ##ctx) \
5924 FN(setsockopt, 49, ##ctx) \
5925 FN(skb_adjust_room, 50, ##ctx) \
5926 FN(redirect_map, 51, ##ctx) \
5927 FN(sk_redirect_map, 52, ##ctx) \
5928 FN(sock_map_update, 53, ##ctx) \
5929 FN(xdp_adjust_meta, 54, ##ctx) \
5930 FN(perf_event_read_value, 55, ##ctx) \
5931 FN(perf_prog_read_value, 56, ##ctx) \
5932 FN(getsockopt, 57, ##ctx) \
5933 FN(override_return, 58, ##ctx) \
5934 FN(sock_ops_cb_flags_set, 59, ##ctx) \
5935 FN(msg_redirect_map, 60, ##ctx) \
5936 FN(msg_apply_bytes, 61, ##ctx) \
5937 FN(msg_cork_bytes, 62, ##ctx) \
5938 FN(msg_pull_data, 63, ##ctx) \
5939 FN(bind, 64, ##ctx) \
5940 FN(xdp_adjust_tail, 65, ##ctx) \
5941 FN(skb_get_xfrm_state, 66, ##ctx) \
5942 FN(get_stack, 67, ##ctx) \
5943 FN(skb_load_bytes_relative, 68, ##ctx) \
5944 FN(fib_lookup, 69, ##ctx) \
5945 FN(sock_hash_update, 70, ##ctx) \
5946 FN(msg_redirect_hash, 71, ##ctx) \
5947 FN(sk_redirect_hash, 72, ##ctx) \
5948 FN(lwt_push_encap, 73, ##ctx) \
5949 FN(lwt_seg6_store_bytes, 74, ##ctx) \
5950 FN(lwt_seg6_adjust_srh, 75, ##ctx) \
5951 FN(lwt_seg6_action, 76, ##ctx) \
5952 FN(rc_repeat, 77, ##ctx) \
5953 FN(rc_keydown, 78, ##ctx) \
5954 FN(skb_cgroup_id, 79, ##ctx) \
5955 FN(get_current_cgroup_id, 80, ##ctx) \
5956 FN(get_local_storage, 81, ##ctx) \
5957 FN(sk_select_reuseport, 82, ##ctx) \
5958 FN(skb_ancestor_cgroup_id, 83, ##ctx) \
5959 FN(sk_lookup_tcp, 84, ##ctx) \
5960 FN(sk_lookup_udp, 85, ##ctx) \
5961 FN(sk_release, 86, ##ctx) \
5962 FN(map_push_elem, 87, ##ctx) \
5963 FN(map_pop_elem, 88, ##ctx) \
5964 FN(map_peek_elem, 89, ##ctx) \
5965 FN(msg_push_data, 90, ##ctx) \
5966 FN(msg_pop_data, 91, ##ctx) \
5967 FN(rc_pointer_rel, 92, ##ctx) \
5968 FN(spin_lock, 93, ##ctx) \
5969 FN(spin_unlock, 94, ##ctx) \
5970 FN(sk_fullsock, 95, ##ctx) \
5971 FN(tcp_sock, 96, ##ctx) \
5972 FN(skb_ecn_set_ce, 97, ##ctx) \
5973 FN(get_listener_sock, 98, ##ctx) \
5974 FN(skc_lookup_tcp, 99, ##ctx) \
5975 FN(tcp_check_syncookie, 100, ##ctx) \
5976 FN(sysctl_get_name, 101, ##ctx) \
5977 FN(sysctl_get_current_value, 102, ##ctx) \
5978 FN(sysctl_get_new_value, 103, ##ctx) \
5979 FN(sysctl_set_new_value, 104, ##ctx) \
5980 FN(strtol, 105, ##ctx) \
5981 FN(strtoul, 106, ##ctx) \
5982 FN(sk_storage_get, 107, ##ctx) \
5983 FN(sk_storage_delete, 108, ##ctx) \
5984 FN(send_signal, 109, ##ctx) \
5985 FN(tcp_gen_syncookie, 110, ##ctx) \
5986 FN(skb_output, 111, ##ctx) \
5987 FN(probe_read_user, 112, ##ctx) \
5988 FN(probe_read_kernel, 113, ##ctx) \
5989 FN(probe_read_user_str, 114, ##ctx) \
5990 FN(probe_read_kernel_str, 115, ##ctx) \
5991 FN(tcp_send_ack, 116, ##ctx) \
5992 FN(send_signal_thread, 117, ##ctx) \
5993 FN(jiffies64, 118, ##ctx) \
5994 FN(read_branch_records, 119, ##ctx) \
5995 FN(get_ns_current_pid_tgid, 120, ##ctx) \
5996 FN(xdp_output, 121, ##ctx) \
5997 FN(get_netns_cookie, 122, ##ctx) \
5998 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \
5999 FN(sk_assign, 124, ##ctx) \
6000 FN(ktime_get_boot_ns, 125, ##ctx) \
6001 FN(seq_printf, 126, ##ctx) \
6002 FN(seq_write, 127, ##ctx) \
6003 FN(sk_cgroup_id, 128, ##ctx) \
6004 FN(sk_ancestor_cgroup_id, 129, ##ctx) \
6005 FN(ringbuf_output, 130, ##ctx) \
6006 FN(ringbuf_reserve, 131, ##ctx) \
6007 FN(ringbuf_submit, 132, ##ctx) \
6008 FN(ringbuf_discard, 133, ##ctx) \
6009 FN(ringbuf_query, 134, ##ctx) \
6010 FN(csum_level, 135, ##ctx) \
6011 FN(skc_to_tcp6_sock, 136, ##ctx) \
6012 FN(skc_to_tcp_sock, 137, ##ctx) \
6013 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \
6014 FN(skc_to_tcp_request_sock, 139, ##ctx) \
6015 FN(skc_to_udp6_sock, 140, ##ctx) \
6016 FN(get_task_stack, 141, ##ctx) \
6017 FN(load_hdr_opt, 142, ##ctx) \
6018 FN(store_hdr_opt, 143, ##ctx) \
6019 FN(reserve_hdr_opt, 144, ##ctx) \
6020 FN(inode_storage_get, 145, ##ctx) \
6021 FN(inode_storage_delete, 146, ##ctx) \
6022 FN(d_path, 147, ##ctx) \
6023 FN(copy_from_user, 148, ##ctx) \
6024 FN(snprintf_btf, 149, ##ctx) \
6025 FN(seq_printf_btf, 150, ##ctx) \
6026 FN(skb_cgroup_classid, 151, ##ctx) \
6027 FN(redirect_neigh, 152, ##ctx) \
6028 FN(per_cpu_ptr, 153, ##ctx) \
6029 FN(this_cpu_ptr, 154, ##ctx) \
6030 FN(redirect_peer, 155, ##ctx) \
6031 FN(task_storage_get, 156, ##ctx) \
6032 FN(task_storage_delete, 157, ##ctx) \
6033 FN(get_current_task_btf, 158, ##ctx) \
6034 FN(bprm_opts_set, 159, ##ctx) \
6035 FN(ktime_get_coarse_ns, 160, ##ctx) \
6036 FN(ima_inode_hash, 161, ##ctx) \
6037 FN(sock_from_file, 162, ##ctx) \
6038 FN(check_mtu, 163, ##ctx) \
6039 FN(for_each_map_elem, 164, ##ctx) \
6040 FN(snprintf, 165, ##ctx) \
6041 FN(sys_bpf, 166, ##ctx) \
6042 FN(btf_find_by_name_kind, 167, ##ctx) \
6043 FN(sys_close, 168, ##ctx) \
6044 FN(timer_init, 169, ##ctx) \
6045 FN(timer_set_callback, 170, ##ctx) \
6046 FN(timer_start, 171, ##ctx) \
6047 FN(timer_cancel, 172, ##ctx) \
6048 FN(get_func_ip, 173, ##ctx) \
6049 FN(get_attach_cookie, 174, ##ctx) \
6050 FN(task_pt_regs, 175, ##ctx) \
6051 FN(get_branch_snapshot, 176, ##ctx) \
6052 FN(trace_vprintk, 177, ##ctx) \
6053 FN(skc_to_unix_sock, 178, ##ctx) \
6054 FN(kallsyms_lookup_name, 179, ##ctx) \
6055 FN(find_vma, 180, ##ctx) \
6056 FN(loop, 181, ##ctx) \
6057 FN(strncmp, 182, ##ctx) \
6058 FN(get_func_arg, 183, ##ctx) \
6059 FN(get_func_ret, 184, ##ctx) \
6060 FN(get_func_arg_cnt, 185, ##ctx) \
6061 FN(get_retval, 186, ##ctx) \
6062 FN(set_retval, 187, ##ctx) \
6063 FN(xdp_get_buff_len, 188, ##ctx) \
6064 FN(xdp_load_bytes, 189, ##ctx) \
6065 FN(xdp_store_bytes, 190, ##ctx) \
6066 FN(copy_from_user_task, 191, ##ctx) \
6067 FN(skb_set_tstamp, 192, ##ctx) \
6068 FN(ima_file_hash, 193, ##ctx) \
6069 FN(kptr_xchg, 194, ##ctx) \
6070 FN(map_lookup_percpu_elem, 195, ##ctx) \
6071 FN(skc_to_mptcp_sock, 196, ##ctx) \
6072 FN(dynptr_from_mem, 197, ##ctx) \
6073 FN(ringbuf_reserve_dynptr, 198, ##ctx) \
6074 FN(ringbuf_submit_dynptr, 199, ##ctx) \
6075 FN(ringbuf_discard_dynptr, 200, ##ctx) \
6076 FN(dynptr_read, 201, ##ctx) \
6077 FN(dynptr_write, 202, ##ctx) \
6078 FN(dynptr_data, 203, ##ctx) \
6079 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \
6080 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \
6081 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \
6082 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \
6083 FN(ktime_get_tai_ns, 208, ##ctx) \
6084 FN(user_ringbuf_drain, 209, ##ctx) \
6085 FN(cgrp_storage_get, 210, ##ctx) \
6086 FN(cgrp_storage_delete, 211, ##ctx) \
6087 /* This helper list is effectively frozen. If you are trying to \
6088 * add a new helper, you should add a kfunc instead which has \
6089 * less stability guarantees. See Documentation/bpf/kfuncs.rst \
6090 */
6091
6092/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
6093 * know or care about integer value that is now passed as second argument
6094 */
6095#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
6096#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
6097
6098/* integer value in 'imm' field of BPF_CALL instruction selects which helper
6099 * function eBPF program intends to call
6100 */
6101#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
6102enum bpf_func_id {
6103 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
6104 __BPF_FUNC_MAX_ID,
6105};
6106#undef __BPF_ENUM_FN
6107
6108/* All flags used by eBPF helper functions, placed here. */
6109
6110/* BPF_FUNC_skb_store_bytes flags. */
6111enum {
6112 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
6113 BPF_F_INVALIDATE_HASH = (1ULL << 1),
6114};
6115
6116/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
6117 * First 4 bits are for passing the header field size.
6118 */
6119enum {
6120 BPF_F_HDR_FIELD_MASK = 0xfULL,
6121};
6122
6123/* BPF_FUNC_l4_csum_replace flags. */
6124enum {
6125 BPF_F_PSEUDO_HDR = (1ULL << 4),
6126 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
6127 BPF_F_MARK_ENFORCE = (1ULL << 6),
6128 BPF_F_IPV6 = (1ULL << 7),
6129};
6130
6131/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
6132enum {
6133 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
6134};
6135
6136/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
6137enum {
6138 BPF_F_SKIP_FIELD_MASK = 0xffULL,
6139 BPF_F_USER_STACK = (1ULL << 8),
6140/* flags used by BPF_FUNC_get_stackid only. */
6141 BPF_F_FAST_STACK_CMP = (1ULL << 9),
6142 BPF_F_REUSE_STACKID = (1ULL << 10),
6143/* flags used by BPF_FUNC_get_stack only. */
6144 BPF_F_USER_BUILD_ID = (1ULL << 11),
6145};
6146
6147/* BPF_FUNC_skb_set_tunnel_key flags. */
6148enum {
6149 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
6150 BPF_F_DONT_FRAGMENT = (1ULL << 2),
6151 BPF_F_SEQ_NUMBER = (1ULL << 3),
6152 BPF_F_NO_TUNNEL_KEY = (1ULL << 4),
6153};
6154
6155/* BPF_FUNC_skb_get_tunnel_key flags. */
6156enum {
6157 BPF_F_TUNINFO_FLAGS = (1ULL << 4),
6158};
6159
6160/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
6161 * BPF_FUNC_perf_event_read_value flags.
6162 */
6163enum {
6164 BPF_F_INDEX_MASK = 0xffffffffULL,
6165 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
6166/* BPF_FUNC_perf_event_output for sk_buff input context. */
6167 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
6168};
6169
6170/* Current network namespace */
6171enum {
6172 BPF_F_CURRENT_NETNS = (-1L),
6173};
6174
6175/* BPF_FUNC_csum_level level values. */
6176enum {
6177 BPF_CSUM_LEVEL_QUERY,
6178 BPF_CSUM_LEVEL_INC,
6179 BPF_CSUM_LEVEL_DEC,
6180 BPF_CSUM_LEVEL_RESET,
6181};
6182
6183/* BPF_FUNC_skb_adjust_room flags. */
6184enum {
6185 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
6186 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
6187 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
6188 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
6189 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
6190 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
6191 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
6192 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7),
6193 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8),
6194};
6195
6196enum {
6197 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
6198 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
6199};
6200
6201#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
6202 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
6203 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6204
6205/* BPF_FUNC_sysctl_get_name flags. */
6206enum {
6207 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
6208};
6209
6210/* BPF_FUNC_<kernel_obj>_storage_get flags */
6211enum {
6212 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
6213 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
6214 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
6215 */
6216 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
6217};
6218
6219/* BPF_FUNC_read_branch_records flags. */
6220enum {
6221 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
6222};
6223
6224/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
6225 * BPF_FUNC_bpf_ringbuf_output flags.
6226 */
6227enum {
6228 BPF_RB_NO_WAKEUP = (1ULL << 0),
6229 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
6230};
6231
6232/* BPF_FUNC_bpf_ringbuf_query flags */
6233enum {
6234 BPF_RB_AVAIL_DATA = 0,
6235 BPF_RB_RING_SIZE = 1,
6236 BPF_RB_CONS_POS = 2,
6237 BPF_RB_PROD_POS = 3,
6238 BPF_RB_OVERWRITE_POS = 4,
6239};
6240
6241/* BPF ring buffer constants */
6242enum {
6243 BPF_RINGBUF_BUSY_BIT = (1U << 31),
6244 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
6245 BPF_RINGBUF_HDR_SZ = 8,
6246};
6247
6248/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6249enum {
6250 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
6251 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
6252};
6253
6254/* Mode for BPF_FUNC_skb_adjust_room helper. */
6255enum bpf_adj_room_mode {
6256 BPF_ADJ_ROOM_NET,
6257 BPF_ADJ_ROOM_MAC,
6258};
6259
6260/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6261enum bpf_hdr_start_off {
6262 BPF_HDR_START_MAC,
6263 BPF_HDR_START_NET,
6264};
6265
6266/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6267enum bpf_lwt_encap_mode {
6268 BPF_LWT_ENCAP_SEG6,
6269 BPF_LWT_ENCAP_SEG6_INLINE,
6270 BPF_LWT_ENCAP_IP,
6271};
6272
6273/* Flags for bpf_bprm_opts_set helper */
6274enum {
6275 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
6276};
6277
6278/* Flags for bpf_redirect and bpf_redirect_map helpers */
6279enum {
6280 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */
6281 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */
6282 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */
6283#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
6284};
6285
6286#define __bpf_md_ptr(type, name) \
6287union { \
6288 type name; \
6289 __u64 :64; \
6290} __attribute__((aligned(8)))
6291
6292/* The enum used in skb->tstamp_type. It specifies the clock type
6293 * of the time stored in the skb->tstamp.
6294 */
6295enum {
6296 BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */
6297 BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */
6298 BPF_SKB_CLOCK_REALTIME = 0,
6299 BPF_SKB_CLOCK_MONOTONIC = 1,
6300 BPF_SKB_CLOCK_TAI = 2,
6301 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,
6302 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.
6303 */
6304};
6305
6306/* user accessible mirror of in-kernel sk_buff.
6307 * new fields can only be added to the end of this structure
6308 */
6309struct __sk_buff {
6310 __u32 len;
6311 __u32 pkt_type;
6312 __u32 mark;
6313 __u32 queue_mapping;
6314 __u32 protocol;
6315 __u32 vlan_present;
6316 __u32 vlan_tci;
6317 __u32 vlan_proto;
6318 __u32 priority;
6319 __u32 ingress_ifindex;
6320 __u32 ifindex;
6321 __u32 tc_index;
6322 __u32 cb[5];
6323 __u32 hash;
6324 __u32 tc_classid;
6325 __u32 data;
6326 __u32 data_end;
6327 __u32 napi_id;
6328
6329 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6330 __u32 family;
6331 __u32 remote_ip4; /* Stored in network byte order */
6332 __u32 local_ip4; /* Stored in network byte order */
6333 __u32 remote_ip6[4]; /* Stored in network byte order */
6334 __u32 local_ip6[4]; /* Stored in network byte order */
6335 __u32 remote_port; /* Stored in network byte order */
6336 __u32 local_port; /* stored in host byte order */
6337 /* ... here. */
6338
6339 __u32 data_meta;
6340 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
6341 __u64 tstamp;
6342 __u32 wire_len;
6343 __u32 gso_segs;
6344 __bpf_md_ptr(struct bpf_sock *, sk);
6345 __u32 gso_size;
6346 __u8 tstamp_type;
6347 __u32 :24; /* Padding, future use. */
6348 __u64 hwtstamp;
6349};
6350
6351struct bpf_tunnel_key {
6352 __u32 tunnel_id;
6353 union {
6354 __u32 remote_ipv4;
6355 __u32 remote_ipv6[4];
6356 };
6357 __u8 tunnel_tos;
6358 __u8 tunnel_ttl;
6359 union {
6360 __u16 tunnel_ext; /* compat */
6361 __be16 tunnel_flags;
6362 };
6363 __u32 tunnel_label;
6364 union {
6365 __u32 local_ipv4;
6366 __u32 local_ipv6[4];
6367 };
6368};
6369
6370/* user accessible mirror of in-kernel xfrm_state.
6371 * new fields can only be added to the end of this structure
6372 */
6373struct bpf_xfrm_state {
6374 __u32 reqid;
6375 __u32 spi; /* Stored in network byte order */
6376 __u16 family;
6377 __u16 ext; /* Padding, future use. */
6378 union {
6379 __u32 remote_ipv4; /* Stored in network byte order */
6380 __u32 remote_ipv6[4]; /* Stored in network byte order */
6381 };
6382};
6383
6384/* Generic BPF return codes which all BPF program types may support.
6385 * The values are binary compatible with their TC_ACT_* counter-part to
6386 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6387 * programs.
6388 *
6389 * XDP is handled seprately, see XDP_*.
6390 */
6391enum bpf_ret_code {
6392 BPF_OK = 0,
6393 /* 1 reserved */
6394 BPF_DROP = 2,
6395 /* 3-6 reserved */
6396 BPF_REDIRECT = 7,
6397 /* >127 are reserved for prog type specific return codes.
6398 *
6399 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6400 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6401 * changed and should be routed based on its new L3 header.
6402 * (This is an L3 redirect, as opposed to L2 redirect
6403 * represented by BPF_REDIRECT above).
6404 */
6405 BPF_LWT_REROUTE = 128,
6406 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6407 * to indicate that no custom dissection was performed, and
6408 * fallback to standard dissector is requested.
6409 */
6410 BPF_FLOW_DISSECTOR_CONTINUE = 129,
6411};
6412
6413struct bpf_sock {
6414 __u32 bound_dev_if;
6415 __u32 family;
6416 __u32 type;
6417 __u32 protocol;
6418 __u32 mark;
6419 __u32 priority;
6420 /* IP address also allows 1 and 2 bytes access */
6421 __u32 src_ip4;
6422 __u32 src_ip6[4];
6423 __u32 src_port; /* host byte order */
6424 __be16 dst_port; /* network byte order */
6425 __u16 :16; /* zero padding */
6426 __u32 dst_ip4;
6427 __u32 dst_ip6[4];
6428 __u32 state;
6429 __s32 rx_queue_mapping;
6430};
6431
6432struct bpf_tcp_sock {
6433 __u32 snd_cwnd; /* Sending congestion window */
6434 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
6435 __u32 rtt_min;
6436 __u32 snd_ssthresh; /* Slow start size threshold */
6437 __u32 rcv_nxt; /* What we want to receive next */
6438 __u32 snd_nxt; /* Next sequence we send */
6439 __u32 snd_una; /* First byte we want an ack for */
6440 __u32 mss_cache; /* Cached effective mss, not including SACKS */
6441 __u32 ecn_flags; /* ECN status bits. */
6442 __u32 rate_delivered; /* saved rate sample: packets delivered */
6443 __u32 rate_interval_us; /* saved rate sample: time elapsed */
6444 __u32 packets_out; /* Packets which are "in flight" */
6445 __u32 retrans_out; /* Retransmitted packets out */
6446 __u32 total_retrans; /* Total retransmits for entire connection */
6447 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
6448 * total number of segments in.
6449 */
6450 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
6451 * total number of data segments in.
6452 */
6453 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
6454 * The total number of segments sent.
6455 */
6456 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
6457 * total number of data segments sent.
6458 */
6459 __u32 lost_out; /* Lost packets */
6460 __u32 sacked_out; /* SACK'd packets */
6461 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
6462 * sum(delta(rcv_nxt)), or how many bytes
6463 * were acked.
6464 */
6465 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
6466 * sum(delta(snd_una)), or how many bytes
6467 * were acked.
6468 */
6469 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
6470 * total number of DSACK blocks received
6471 */
6472 __u32 delivered; /* Total data packets delivered incl. rexmits */
6473 __u32 delivered_ce; /* Like the above but only ECE marked packets */
6474 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
6475};
6476
6477struct bpf_sock_tuple {
6478 union {
6479 struct {
6480 __be32 saddr;
6481 __be32 daddr;
6482 __be16 sport;
6483 __be16 dport;
6484 } ipv4;
6485 struct {
6486 __be32 saddr[4];
6487 __be32 daddr[4];
6488 __be16 sport;
6489 __be16 dport;
6490 } ipv6;
6491 };
6492};
6493
6494/* (Simplified) user return codes for tcx prog type.
6495 * A valid tcx program must return one of these defined values. All other
6496 * return codes are reserved for future use. Must remain compatible with
6497 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6498 * return codes are mapped to TCX_NEXT.
6499 */
6500enum tcx_action_base {
6501 TCX_NEXT = -1,
6502 TCX_PASS = 0,
6503 TCX_DROP = 2,
6504 TCX_REDIRECT = 7,
6505};
6506
6507struct bpf_xdp_sock {
6508 __u32 queue_id;
6509};
6510
6511#define XDP_PACKET_HEADROOM 256
6512
6513/* User return codes for XDP prog type.
6514 * A valid XDP program must return one of these defined values. All other
6515 * return codes are reserved for future use. Unknown return codes will
6516 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6517 */
6518enum xdp_action {
6519 XDP_ABORTED = 0,
6520 XDP_DROP,
6521 XDP_PASS,
6522 XDP_TX,
6523 XDP_REDIRECT,
6524};
6525
6526/* user accessible metadata for XDP packet hook
6527 * new fields must be added to the end of this structure
6528 */
6529struct xdp_md {
6530 __u32 data;
6531 __u32 data_end;
6532 __u32 data_meta;
6533 /* Below access go through struct xdp_rxq_info */
6534 __u32 ingress_ifindex; /* rxq->dev->ifindex */
6535 __u32 rx_queue_index; /* rxq->queue_index */
6536
6537 __u32 egress_ifindex; /* txq->dev->ifindex */
6538};
6539
6540/* DEVMAP map-value layout
6541 *
6542 * The struct data-layout of map-value is a configuration interface.
6543 * New members can only be added to the end of this structure.
6544 */
6545struct bpf_devmap_val {
6546 __u32 ifindex; /* device index */
6547 union {
6548 int fd; /* prog fd on map write */
6549 __u32 id; /* prog id on map read */
6550 } bpf_prog;
6551};
6552
6553/* CPUMAP map-value layout
6554 *
6555 * The struct data-layout of map-value is a configuration interface.
6556 * New members can only be added to the end of this structure.
6557 */
6558struct bpf_cpumap_val {
6559 __u32 qsize; /* queue size to remote target CPU */
6560 union {
6561 int fd; /* prog fd on map write */
6562 __u32 id; /* prog id on map read */
6563 } bpf_prog;
6564};
6565
6566enum sk_action {
6567 SK_DROP = 0,
6568 SK_PASS,
6569};
6570
6571/* user accessible metadata for SK_MSG packet hook, new fields must
6572 * be added to the end of this structure
6573 */
6574struct sk_msg_md {
6575 __bpf_md_ptr(void *, data);
6576 __bpf_md_ptr(void *, data_end);
6577
6578 __u32 family;
6579 __u32 remote_ip4; /* Stored in network byte order */
6580 __u32 local_ip4; /* Stored in network byte order */
6581 __u32 remote_ip6[4]; /* Stored in network byte order */
6582 __u32 local_ip6[4]; /* Stored in network byte order */
6583 __u32 remote_port; /* Stored in network byte order */
6584 __u32 local_port; /* stored in host byte order */
6585 __u32 size; /* Total size of sk_msg */
6586
6587 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6588};
6589
6590struct sk_reuseport_md {
6591 /*
6592 * Start of directly accessible data. It begins from
6593 * the tcp/udp header.
6594 */
6595 __bpf_md_ptr(void *, data);
6596 /* End of directly accessible data */
6597 __bpf_md_ptr(void *, data_end);
6598 /*
6599 * Total length of packet (starting from the tcp/udp header).
6600 * Note that the directly accessible bytes (data_end - data)
6601 * could be less than this "len". Those bytes could be
6602 * indirectly read by a helper "bpf_skb_load_bytes()".
6603 */
6604 __u32 len;
6605 /*
6606 * Eth protocol in the mac header (network byte order). e.g.
6607 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6608 */
6609 __u32 eth_protocol;
6610 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6611 __u32 bind_inany; /* Is sock bound to an INANY address? */
6612 __u32 hash; /* A hash of the packet 4 tuples */
6613 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6614 * new incoming connection request (e.g. selecting a listen sk for
6615 * the received SYN in the TCP case). reuse->sk is one of the sk
6616 * in the reuseport group. The bpf prog can use reuse->sk to learn
6617 * the local listening ip/port without looking into the skb.
6618 *
6619 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6620 * reuse->migrating_sk is the socket that needs to be migrated
6621 * to another listening socket. migrating_sk could be a fullsock
6622 * sk that is fully established or a reqsk that is in-the-middle
6623 * of 3-way handshake.
6624 */
6625 __bpf_md_ptr(struct bpf_sock *, sk);
6626 __bpf_md_ptr(struct bpf_sock *, migrating_sk);
6627};
6628
6629#define BPF_TAG_SIZE 8
6630
6631struct bpf_prog_info {
6632 __u32 type;
6633 __u32 id;
6634 __u8 tag[BPF_TAG_SIZE];
6635 __u32 jited_prog_len;
6636 __u32 xlated_prog_len;
6637 __aligned_u64 jited_prog_insns;
6638 __aligned_u64 xlated_prog_insns;
6639 __u64 load_time; /* ns since boottime */
6640 __u32 created_by_uid;
6641 __u32 nr_map_ids;
6642 __aligned_u64 map_ids;
6643 char name[BPF_OBJ_NAME_LEN];
6644 __u32 ifindex;
6645 __u32 gpl_compatible:1;
6646 __u32 :31; /* alignment pad */
6647 __u64 netns_dev;
6648 __u64 netns_ino;
6649 __u32 nr_jited_ksyms;
6650 __u32 nr_jited_func_lens;
6651 __aligned_u64 jited_ksyms;
6652 __aligned_u64 jited_func_lens;
6653 __u32 btf_id;
6654 __u32 func_info_rec_size;
6655 __aligned_u64 func_info;
6656 __u32 nr_func_info;
6657 __u32 nr_line_info;
6658 __aligned_u64 line_info;
6659 __aligned_u64 jited_line_info;
6660 __u32 nr_jited_line_info;
6661 __u32 line_info_rec_size;
6662 __u32 jited_line_info_rec_size;
6663 __u32 nr_prog_tags;
6664 __aligned_u64 prog_tags;
6665 __u64 run_time_ns;
6666 __u64 run_cnt;
6667 __u64 recursion_misses;
6668 __u32 verified_insns;
6669 __u32 attach_btf_obj_id;
6670 __u32 attach_btf_id;
6671} __attribute__((aligned(8)));
6672
6673struct bpf_map_info {
6674 __u32 type;
6675 __u32 id;
6676 __u32 key_size;
6677 __u32 value_size;
6678 __u32 max_entries;
6679 __u32 map_flags;
6680 char name[BPF_OBJ_NAME_LEN];
6681 __u32 ifindex;
6682 __u32 btf_vmlinux_value_type_id;
6683 __u64 netns_dev;
6684 __u64 netns_ino;
6685 __u32 btf_id;
6686 __u32 btf_key_type_id;
6687 __u32 btf_value_type_id;
6688 __u32 btf_vmlinux_id;
6689 __u64 map_extra;
6690 __aligned_u64 hash;
6691 __u32 hash_size;
6692} __attribute__((aligned(8)));
6693
6694struct bpf_btf_info {
6695 __aligned_u64 btf;
6696 __u32 btf_size;
6697 __u32 id;
6698 __aligned_u64 name;
6699 __u32 name_len;
6700 __u32 kernel_btf;
6701} __attribute__((aligned(8)));
6702
6703struct bpf_link_info {
6704 __u32 type;
6705 __u32 id;
6706 __u32 prog_id;
6707 union {
6708 struct {
6709 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6710 __u32 tp_name_len; /* in/out: tp_name buffer len */
6711 __u32 :32;
6712 __u64 cookie;
6713 } raw_tracepoint;
6714 struct {
6715 __u32 attach_type;
6716 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6717 __u32 target_btf_id; /* BTF type id inside the object */
6718 __u32 :32;
6719 __u64 cookie;
6720 } tracing;
6721 struct {
6722 __u64 cgroup_id;
6723 __u32 attach_type;
6724 } cgroup;
6725 struct {
6726 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
6727 __u32 target_name_len; /* in/out: target_name buffer len */
6728
6729 /* If the iter specific field is 32 bits, it can be put
6730 * in the first or second union. Otherwise it should be
6731 * put in the second union.
6732 */
6733 union {
6734 struct {
6735 __u32 map_id;
6736 } map;
6737 };
6738 union {
6739 struct {
6740 __u64 cgroup_id;
6741 __u32 order;
6742 } cgroup;
6743 struct {
6744 __u32 tid;
6745 __u32 pid;
6746 } task;
6747 };
6748 } iter;
6749 struct {
6750 __u32 netns_ino;
6751 __u32 attach_type;
6752 } netns;
6753 struct {
6754 __u32 ifindex;
6755 } xdp;
6756 struct {
6757 __u32 map_id;
6758 } struct_ops;
6759 struct {
6760 __u32 pf;
6761 __u32 hooknum;
6762 __s32 priority;
6763 __u32 flags;
6764 } netfilter;
6765 struct {
6766 __aligned_u64 addrs;
6767 __u32 count; /* in/out: kprobe_multi function count */
6768 __u32 flags;
6769 __u64 missed;
6770 __aligned_u64 cookies;
6771 } kprobe_multi;
6772 struct {
6773 __aligned_u64 path;
6774 __aligned_u64 offsets;
6775 __aligned_u64 ref_ctr_offsets;
6776 __aligned_u64 cookies;
6777 __u32 path_size; /* in/out: real path size on success, including zero byte */
6778 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
6779 __u32 flags;
6780 __u32 pid;
6781 } uprobe_multi;
6782 struct {
6783 __u32 type; /* enum bpf_perf_event_type */
6784 __u32 :32;
6785 union {
6786 struct {
6787 __aligned_u64 file_name; /* in/out */
6788 __u32 name_len;
6789 __u32 offset; /* offset from file_name */
6790 __u64 cookie;
6791 __u64 ref_ctr_offset;
6792 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6793 struct {
6794 __aligned_u64 func_name; /* in/out */
6795 __u32 name_len;
6796 __u32 offset; /* offset from func_name */
6797 __u64 addr;
6798 __u64 missed;
6799 __u64 cookie;
6800 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6801 struct {
6802 __aligned_u64 tp_name; /* in/out */
6803 __u32 name_len;
6804 __u32 :32;
6805 __u64 cookie;
6806 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
6807 struct {
6808 __u64 config;
6809 __u32 type;
6810 __u32 :32;
6811 __u64 cookie;
6812 } event; /* BPF_PERF_EVENT_EVENT */
6813 };
6814 } perf_event;
6815 struct {
6816 __u32 ifindex;
6817 __u32 attach_type;
6818 } tcx;
6819 struct {
6820 __u32 ifindex;
6821 __u32 attach_type;
6822 } netkit;
6823 struct {
6824 __u32 map_id;
6825 __u32 attach_type;
6826 } sockmap;
6827 };
6828} __attribute__((aligned(8)));
6829
6830struct bpf_token_info {
6831 __u64 allowed_cmds;
6832 __u64 allowed_maps;
6833 __u64 allowed_progs;
6834 __u64 allowed_attachs;
6835} __attribute__((aligned(8)));
6836
6837/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6838 * by user and intended to be used by socket (e.g. to bind to, depends on
6839 * attach type).
6840 */
6841struct bpf_sock_addr {
6842 __u32 user_family; /* Allows 4-byte read, but no write. */
6843 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6844 * Stored in network byte order.
6845 */
6846 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6847 * Stored in network byte order.
6848 */
6849 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
6850 * Stored in network byte order
6851 */
6852 __u32 family; /* Allows 4-byte read, but no write */
6853 __u32 type; /* Allows 4-byte read, but no write */
6854 __u32 protocol; /* Allows 4-byte read, but no write */
6855 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6856 * Stored in network byte order.
6857 */
6858 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6859 * Stored in network byte order.
6860 */
6861 __bpf_md_ptr(struct bpf_sock *, sk);
6862};
6863
6864/* User bpf_sock_ops struct to access socket values and specify request ops
6865 * and their replies.
6866 * Some of this fields are in network (bigendian) byte order and may need
6867 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6868 * New fields can only be added at the end of this structure
6869 */
6870struct bpf_sock_ops {
6871 __u32 op;
6872 union {
6873 __u32 args[4]; /* Optionally passed to bpf program */
6874 __u32 reply; /* Returned by bpf program */
6875 __u32 replylong[4]; /* Optionally returned by bpf prog */
6876 };
6877 __u32 family;
6878 __u32 remote_ip4; /* Stored in network byte order */
6879 __u32 local_ip4; /* Stored in network byte order */
6880 __u32 remote_ip6[4]; /* Stored in network byte order */
6881 __u32 local_ip6[4]; /* Stored in network byte order */
6882 __u32 remote_port; /* Stored in network byte order */
6883 __u32 local_port; /* stored in host byte order */
6884 __u32 is_fullsock; /* Some TCP fields are only valid if
6885 * there is a full socket. If not, the
6886 * fields read as zero.
6887 */
6888 __u32 snd_cwnd;
6889 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
6890 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6891 __u32 state;
6892 __u32 rtt_min;
6893 __u32 snd_ssthresh;
6894 __u32 rcv_nxt;
6895 __u32 snd_nxt;
6896 __u32 snd_una;
6897 __u32 mss_cache;
6898 __u32 ecn_flags;
6899 __u32 rate_delivered;
6900 __u32 rate_interval_us;
6901 __u32 packets_out;
6902 __u32 retrans_out;
6903 __u32 total_retrans;
6904 __u32 segs_in;
6905 __u32 data_segs_in;
6906 __u32 segs_out;
6907 __u32 data_segs_out;
6908 __u32 lost_out;
6909 __u32 sacked_out;
6910 __u32 sk_txhash;
6911 __u64 bytes_received;
6912 __u64 bytes_acked;
6913 __bpf_md_ptr(struct bpf_sock *, sk);
6914 /* [skb_data, skb_data_end) covers the whole TCP header.
6915 *
6916 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6917 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
6918 * header has not been written.
6919 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6920 * been written so far.
6921 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
6922 * the 3WHS.
6923 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6924 * the 3WHS.
6925 *
6926 * bpf_load_hdr_opt() can also be used to read a particular option.
6927 */
6928 __bpf_md_ptr(void *, skb_data);
6929 __bpf_md_ptr(void *, skb_data_end);
6930 __u32 skb_len; /* The total length of a packet.
6931 * It includes the header, options,
6932 * and payload.
6933 */
6934 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
6935 * an easy way to check for tcp_flags
6936 * without parsing skb_data.
6937 *
6938 * In particular, the skb_tcp_flags
6939 * will still be available in
6940 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6941 * the outgoing header has not
6942 * been written yet.
6943 */
6944 __u64 skb_hwtstamp;
6945};
6946
6947/* Definitions for bpf_sock_ops_cb_flags */
6948enum {
6949 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
6950 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
6951 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
6952 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
6953 /* Call bpf for all received TCP headers. The bpf prog will be
6954 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6955 *
6956 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6957 * for the header option related helpers that will be useful
6958 * to the bpf programs.
6959 *
6960 * It could be used at the client/active side (i.e. connect() side)
6961 * when the server told it that the server was in syncookie
6962 * mode and required the active side to resend the bpf-written
6963 * options. The active side can keep writing the bpf-options until
6964 * it received a valid packet from the server side to confirm
6965 * the earlier packet (and options) has been received. The later
6966 * example patch is using it like this at the active side when the
6967 * server is in syncookie mode.
6968 *
6969 * The bpf prog will usually turn this off in the common cases.
6970 */
6971 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
6972 /* Call bpf when kernel has received a header option that
6973 * the kernel cannot handle. The bpf prog will be called under
6974 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6975 *
6976 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6977 * for the header option related helpers that will be useful
6978 * to the bpf programs.
6979 */
6980 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6981 /* Call bpf when the kernel is writing header options for the
6982 * outgoing packet. The bpf prog will first be called
6983 * to reserve space in a skb under
6984 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
6985 * the bpf prog will be called to write the header option(s)
6986 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6987 *
6988 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6989 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6990 * related helpers that will be useful to the bpf programs.
6991 *
6992 * The kernel gets its chance to reserve space and write
6993 * options first before the BPF program does.
6994 */
6995 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6996/* Mask of all currently supported cb flags */
6997 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
6998};
6999
7000enum {
7001 SK_BPF_CB_TX_TIMESTAMPING = 1<<0,
7002 SK_BPF_CB_MASK = (SK_BPF_CB_TX_TIMESTAMPING - 1) |
7003 SK_BPF_CB_TX_TIMESTAMPING
7004};
7005
7006/* List of known BPF sock_ops operators.
7007 * New entries can only be added at the end
7008 */
7009enum {
7010 BPF_SOCK_OPS_VOID,
7011 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
7012 * -1 if default value should be used
7013 */
7014 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
7015 * window (in packets) or -1 if default
7016 * value should be used
7017 */
7018 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
7019 * active connection is initialized
7020 */
7021 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
7022 * active connection is
7023 * established
7024 */
7025 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
7026 * passive connection is
7027 * established
7028 */
7029 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
7030 * needs ECN
7031 */
7032 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
7033 * based on the path and may be
7034 * dependent on the congestion control
7035 * algorithm. In general it indicates
7036 * a congestion threshold. RTTs above
7037 * this indicate congestion
7038 */
7039 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
7040 * Arg1: value of icsk_retransmits
7041 * Arg2: value of icsk_rto
7042 * Arg3: whether RTO has expired
7043 */
7044 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
7045 * Arg1: sequence number of 1st byte
7046 * Arg2: # segments
7047 * Arg3: return value of
7048 * tcp_transmit_skb (0 => success)
7049 */
7050 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
7051 * Arg1: old_state
7052 * Arg2: new_state
7053 */
7054 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
7055 * socket transition to LISTEN state.
7056 */
7057 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
7058 * Arg1: measured RTT input (mrtt)
7059 * Arg2: updated srtt
7060 */
7061 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
7062 * It will be called to handle
7063 * the packets received at
7064 * an already established
7065 * connection.
7066 *
7067 * sock_ops->skb_data:
7068 * Referring to the received skb.
7069 * It covers the TCP header only.
7070 *
7071 * bpf_load_hdr_opt() can also
7072 * be used to search for a
7073 * particular option.
7074 */
7075 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
7076 * header option later in
7077 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7078 * Arg1: bool want_cookie. (in
7079 * writing SYNACK only)
7080 *
7081 * sock_ops->skb_data:
7082 * Not available because no header has
7083 * been written yet.
7084 *
7085 * sock_ops->skb_tcp_flags:
7086 * The tcp_flags of the
7087 * outgoing skb. (e.g. SYN, ACK, FIN).
7088 *
7089 * bpf_reserve_hdr_opt() should
7090 * be used to reserve space.
7091 */
7092 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
7093 * Arg1: bool want_cookie. (in
7094 * writing SYNACK only)
7095 *
7096 * sock_ops->skb_data:
7097 * Referring to the outgoing skb.
7098 * It covers the TCP header
7099 * that has already been written
7100 * by the kernel and the
7101 * earlier bpf-progs.
7102 *
7103 * sock_ops->skb_tcp_flags:
7104 * The tcp_flags of the outgoing
7105 * skb. (e.g. SYN, ACK, FIN).
7106 *
7107 * bpf_store_hdr_opt() should
7108 * be used to write the
7109 * option.
7110 *
7111 * bpf_load_hdr_opt() can also
7112 * be used to search for a
7113 * particular option that
7114 * has already been written
7115 * by the kernel or the
7116 * earlier bpf-progs.
7117 */
7118 BPF_SOCK_OPS_TSTAMP_SCHED_CB, /* Called when skb is passing
7119 * through dev layer when
7120 * SK_BPF_CB_TX_TIMESTAMPING
7121 * feature is on.
7122 */
7123 BPF_SOCK_OPS_TSTAMP_SND_SW_CB, /* Called when skb is about to send
7124 * to the nic when SK_BPF_CB_TX_TIMESTAMPING
7125 * feature is on.
7126 */
7127 BPF_SOCK_OPS_TSTAMP_SND_HW_CB, /* Called in hardware phase when
7128 * SK_BPF_CB_TX_TIMESTAMPING feature
7129 * is on.
7130 */
7131 BPF_SOCK_OPS_TSTAMP_ACK_CB, /* Called when all the skbs in the
7132 * same sendmsg call are acked
7133 * when SK_BPF_CB_TX_TIMESTAMPING
7134 * feature is on.
7135 */
7136 BPF_SOCK_OPS_TSTAMP_SENDMSG_CB, /* Called when every sendmsg syscall
7137 * is triggered. It's used to correlate
7138 * sendmsg timestamp with corresponding
7139 * tskey.
7140 */
7141};
7142
7143/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
7144 * changes between the TCP and BPF versions. Ideally this should never happen.
7145 * If it does, we need to add code to convert them before calling
7146 * the BPF sock_ops function.
7147 */
7148enum {
7149 BPF_TCP_ESTABLISHED = 1,
7150 BPF_TCP_SYN_SENT,
7151 BPF_TCP_SYN_RECV,
7152 BPF_TCP_FIN_WAIT1,
7153 BPF_TCP_FIN_WAIT2,
7154 BPF_TCP_TIME_WAIT,
7155 BPF_TCP_CLOSE,
7156 BPF_TCP_CLOSE_WAIT,
7157 BPF_TCP_LAST_ACK,
7158 BPF_TCP_LISTEN,
7159 BPF_TCP_CLOSING, /* Now a valid state */
7160 BPF_TCP_NEW_SYN_RECV,
7161 BPF_TCP_BOUND_INACTIVE,
7162
7163 BPF_TCP_MAX_STATES /* Leave at the end! */
7164};
7165
7166enum {
7167 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
7168 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
7169 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
7170 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
7171 /* Copy the SYN pkt to optval
7172 *
7173 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
7174 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
7175 * to only getting from the saved_syn. It can either get the
7176 * syn packet from:
7177 *
7178 * 1. the just-received SYN packet (only available when writing the
7179 * SYNACK). It will be useful when it is not necessary to
7180 * save the SYN packet for latter use. It is also the only way
7181 * to get the SYN during syncookie mode because the syn
7182 * packet cannot be saved during syncookie.
7183 *
7184 * OR
7185 *
7186 * 2. the earlier saved syn which was done by
7187 * bpf_setsockopt(TCP_SAVE_SYN).
7188 *
7189 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
7190 * SYN packet is obtained.
7191 *
7192 * If the bpf-prog does not need the IP[46] header, the
7193 * bpf-prog can avoid parsing the IP header by using
7194 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
7195 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
7196 *
7197 * >0: Total number of bytes copied
7198 * -ENOSPC: Not enough space in optval. Only optlen number of
7199 * bytes is copied.
7200 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
7201 * is not saved by setsockopt(TCP_SAVE_SYN).
7202 */
7203 TCP_BPF_SYN = 1005, /* Copy the TCP header */
7204 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
7205 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
7206 TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */
7207 SK_BPF_CB_FLAGS = 1009, /* Get or set sock ops flags in socket */
7208 SK_BPF_BYPASS_PROT_MEM = 1010, /* Get or Set sk->sk_bypass_prot_mem */
7209
7210};
7211
7212enum {
7213 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
7214};
7215
7216/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
7217 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7218 */
7219enum {
7220 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
7221 * total option spaces
7222 * required for an established
7223 * sk in order to calculate the
7224 * MSS. No skb is actually
7225 * sent.
7226 */
7227 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
7228 * when sending a SYN.
7229 */
7230};
7231
7232struct bpf_perf_event_value {
7233 __u64 counter;
7234 __u64 enabled;
7235 __u64 running;
7236};
7237
7238enum {
7239 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
7240 BPF_DEVCG_ACC_READ = (1ULL << 1),
7241 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
7242};
7243
7244enum {
7245 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
7246 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
7247};
7248
7249struct bpf_cgroup_dev_ctx {
7250 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
7251 __u32 access_type;
7252 __u32 major;
7253 __u32 minor;
7254};
7255
7256struct bpf_raw_tracepoint_args {
7257 __u64 args[0];
7258};
7259
7260/* DIRECT: Skip the FIB rules and go to FIB table associated with device
7261 * OUTPUT: Do lookup from egress perspective; default is ingress
7262 */
7263enum {
7264 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
7265 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
7266 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
7267 BPF_FIB_LOOKUP_TBID = (1U << 3),
7268 BPF_FIB_LOOKUP_SRC = (1U << 4),
7269 BPF_FIB_LOOKUP_MARK = (1U << 5),
7270};
7271
7272enum {
7273 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
7274 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
7275 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
7276 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
7277 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
7278 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
7279 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
7280 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
7281 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
7282 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */
7283};
7284
7285struct bpf_fib_lookup {
7286 /* input: network family for lookup (AF_INET, AF_INET6)
7287 * output: network family of egress nexthop
7288 */
7289 __u8 family;
7290
7291 /* set if lookup is to consider L4 data - e.g., FIB rules */
7292 __u8 l4_protocol;
7293 __be16 sport;
7294 __be16 dport;
7295
7296 union { /* used for MTU check */
7297 /* input to lookup */
7298 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
7299
7300 /* output: MTU value */
7301 __u16 mtu_result;
7302 } __attribute__((packed, aligned(2)));
7303 /* input: L3 device index for lookup
7304 * output: device index from FIB lookup
7305 */
7306 __u32 ifindex;
7307
7308 union {
7309 /* inputs to lookup */
7310 __u8 tos; /* AF_INET */
7311 __be32 flowinfo; /* AF_INET6, flow_label + priority */
7312
7313 /* output: metric of fib result (IPv4/IPv6 only) */
7314 __u32 rt_metric;
7315 };
7316
7317 /* input: source address to consider for lookup
7318 * output: source address result from lookup
7319 */
7320 union {
7321 __be32 ipv4_src;
7322 __u32 ipv6_src[4]; /* in6_addr; network order */
7323 };
7324
7325 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7326 * network header. output: bpf_fib_lookup sets to gateway address
7327 * if FIB lookup returns gateway route
7328 */
7329 union {
7330 __be32 ipv4_dst;
7331 __u32 ipv6_dst[4]; /* in6_addr; network order */
7332 };
7333
7334 union {
7335 struct {
7336 /* output */
7337 __be16 h_vlan_proto;
7338 __be16 h_vlan_TCI;
7339 };
7340 /* input: when accompanied with the
7341 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7342 * specific routing table to use for the fib lookup.
7343 */
7344 __u32 tbid;
7345 };
7346
7347 union {
7348 /* input */
7349 struct {
7350 __u32 mark; /* policy routing */
7351 /* 2 4-byte holes for input */
7352 };
7353
7354 /* output: source and dest mac */
7355 struct {
7356 __u8 smac[6]; /* ETH_ALEN */
7357 __u8 dmac[6]; /* ETH_ALEN */
7358 };
7359 };
7360};
7361
7362struct bpf_redir_neigh {
7363 /* network family for lookup (AF_INET, AF_INET6) */
7364 __u32 nh_family;
7365 /* network address of nexthop; skips fib lookup to find gateway */
7366 union {
7367 __be32 ipv4_nh;
7368 __u32 ipv6_nh[4]; /* in6_addr; network order */
7369 };
7370};
7371
7372/* bpf_check_mtu flags*/
7373enum bpf_check_mtu_flags {
7374 BPF_MTU_CHK_SEGS = (1U << 0),
7375};
7376
7377enum bpf_check_mtu_ret {
7378 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
7379 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
7380 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
7381};
7382
7383enum bpf_task_fd_type {
7384 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
7385 BPF_FD_TYPE_TRACEPOINT, /* tp name */
7386 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
7387 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
7388 BPF_FD_TYPE_UPROBE, /* filename + offset */
7389 BPF_FD_TYPE_URETPROBE, /* filename + offset */
7390};
7391
7392enum {
7393 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
7394 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
7395 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
7396};
7397
7398struct bpf_flow_keys {
7399 __u16 nhoff;
7400 __u16 thoff;
7401 __u16 addr_proto; /* ETH_P_* of valid addrs */
7402 __u8 is_frag;
7403 __u8 is_first_frag;
7404 __u8 is_encap;
7405 __u8 ip_proto;
7406 __be16 n_proto;
7407 __be16 sport;
7408 __be16 dport;
7409 union {
7410 struct {
7411 __be32 ipv4_src;
7412 __be32 ipv4_dst;
7413 };
7414 struct {
7415 __u32 ipv6_src[4]; /* in6_addr; network order */
7416 __u32 ipv6_dst[4]; /* in6_addr; network order */
7417 };
7418 };
7419 __u32 flags;
7420 __be32 flow_label;
7421};
7422
7423struct bpf_func_info {
7424 __u32 insn_off;
7425 __u32 type_id;
7426};
7427
7428#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
7429#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
7430
7431struct bpf_line_info {
7432 __u32 insn_off;
7433 __u32 file_name_off;
7434 __u32 line_off;
7435 __u32 line_col;
7436};
7437
7438struct bpf_spin_lock {
7439 __u32 val;
7440};
7441
7442struct bpf_timer {
7443 __u64 __opaque[2];
7444} __attribute__((aligned(8)));
7445
7446struct bpf_task_work {
7447 __u64 __opaque;
7448} __attribute__((aligned(8)));
7449
7450struct bpf_wq {
7451 __u64 __opaque[2];
7452} __attribute__((aligned(8)));
7453
7454struct bpf_dynptr {
7455 __u64 __opaque[2];
7456} __attribute__((aligned(8)));
7457
7458struct bpf_list_head {
7459 __u64 __opaque[2];
7460} __attribute__((aligned(8)));
7461
7462struct bpf_list_node {
7463 __u64 __opaque[3];
7464} __attribute__((aligned(8)));
7465
7466struct bpf_rb_root {
7467 __u64 __opaque[2];
7468} __attribute__((aligned(8)));
7469
7470struct bpf_rb_node {
7471 __u64 __opaque[4];
7472} __attribute__((aligned(8)));
7473
7474struct bpf_refcount {
7475 __u32 __opaque[1];
7476} __attribute__((aligned(4)));
7477
7478struct bpf_sysctl {
7479 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
7480 * Allows 1,2,4-byte read, but no write.
7481 */
7482 __u32 file_pos; /* Sysctl file position to read from, write to.
7483 * Allows 1,2,4-byte read an 4-byte write.
7484 */
7485};
7486
7487struct bpf_sockopt {
7488 __bpf_md_ptr(struct bpf_sock *, sk);
7489 __bpf_md_ptr(void *, optval);
7490 __bpf_md_ptr(void *, optval_end);
7491
7492 __s32 level;
7493 __s32 optname;
7494 __s32 optlen;
7495 __s32 retval;
7496};
7497
7498struct bpf_pidns_info {
7499 __u32 pid;
7500 __u32 tgid;
7501};
7502
7503/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7504struct bpf_sk_lookup {
7505 union {
7506 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
7507 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
7508 };
7509
7510 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
7511 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7512 __u32 remote_ip4; /* Network byte order */
7513 __u32 remote_ip6[4]; /* Network byte order */
7514 __be16 remote_port; /* Network byte order */
7515 __u16 :16; /* Zero padding */
7516 __u32 local_ip4; /* Network byte order */
7517 __u32 local_ip6[4]; /* Network byte order */
7518 __u32 local_port; /* Host byte order */
7519 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */
7520};
7521
7522/*
7523 * struct btf_ptr is used for typed pointer representation; the
7524 * type id is used to render the pointer data as the appropriate type
7525 * via the bpf_snprintf_btf() helper described above. A flags field -
7526 * potentially to specify additional details about the BTF pointer
7527 * (rather than its mode of display) - is included for future use.
7528 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7529 */
7530struct btf_ptr {
7531 void *ptr;
7532 __u32 type_id;
7533 __u32 flags; /* BTF ptr flags; unused at present. */
7534};
7535
7536/*
7537 * Flags to control bpf_snprintf_btf() behaviour.
7538 * - BTF_F_COMPACT: no formatting around type information
7539 * - BTF_F_NONAME: no struct/union member names/types
7540 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7541 * equivalent to %px.
7542 * - BTF_F_ZERO: show zero-valued struct/union members; they
7543 * are not displayed by default
7544 */
7545enum {
7546 BTF_F_COMPACT = (1ULL << 0),
7547 BTF_F_NONAME = (1ULL << 1),
7548 BTF_F_PTR_RAW = (1ULL << 2),
7549 BTF_F_ZERO = (1ULL << 3),
7550};
7551
7552/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7553 * has to be adjusted by relocations. It is emitted by llvm and passed to
7554 * libbpf and later to the kernel.
7555 */
7556enum bpf_core_relo_kind {
7557 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */
7558 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */
7559 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */
7560 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */
7561 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */
7562 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */
7563 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */
7564 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */
7565 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */
7566 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */
7567 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */
7568 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */
7569 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */
7570};
7571
7572/*
7573 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7574 * and from libbpf to the kernel.
7575 *
7576 * CO-RE relocation captures the following data:
7577 * - insn_off - instruction offset (in bytes) within a BPF program that needs
7578 * its insn->imm field to be relocated with actual field info;
7579 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7580 * type or field;
7581 * - access_str_off - offset into corresponding .BTF string section. String
7582 * interpretation depends on specific relocation kind:
7583 * - for field-based relocations, string encodes an accessed field using
7584 * a sequence of field and array indices, separated by colon (:). It's
7585 * conceptually very close to LLVM's getelementptr ([0]) instruction's
7586 * arguments for identifying offset to a field.
7587 * - for type-based relocations, strings is expected to be just "0";
7588 * - for enum value-based relocations, string contains an index of enum
7589 * value within its enum type;
7590 * - kind - one of enum bpf_core_relo_kind;
7591 *
7592 * Example:
7593 * struct sample {
7594 * int a;
7595 * struct {
7596 * int b[10];
7597 * };
7598 * };
7599 *
7600 * struct sample *s = ...;
7601 * int *x = &s->a; // encoded as "0:0" (a is field #0)
7602 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
7603 * // b is field #0 inside anon struct, accessing elem #5)
7604 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7605 *
7606 * type_id for all relocs in this example will capture BTF type id of
7607 * `struct sample`.
7608 *
7609 * Such relocation is emitted when using __builtin_preserve_access_index()
7610 * Clang built-in, passing expression that captures field address, e.g.:
7611 *
7612 * bpf_probe_read(&dst, sizeof(dst),
7613 * __builtin_preserve_access_index(&src->a.b.c));
7614 *
7615 * In this case Clang will emit field relocation recording necessary data to
7616 * be able to find offset of embedded `a.b.c` field within `src` struct.
7617 *
7618 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7619 */
7620struct bpf_core_relo {
7621 __u32 insn_off;
7622 __u32 type_id;
7623 __u32 access_str_off;
7624 enum bpf_core_relo_kind kind;
7625};
7626
7627/*
7628 * Flags to control bpf_timer_start() behaviour.
7629 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7630 * relative to current time.
7631 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
7632 */
7633enum {
7634 BPF_F_TIMER_ABS = (1ULL << 0),
7635 BPF_F_TIMER_CPU_PIN = (1ULL << 1),
7636};
7637
7638/* BPF numbers iterator state */
7639struct bpf_iter_num {
7640 /* opaque iterator state; having __u64 here allows to preserve correct
7641 * alignment requirements in vmlinux.h, generated from BTF
7642 */
7643 __u64 __opaque[1];
7644} __attribute__((aligned(8)));
7645
7646/*
7647 * Flags to control BPF kfunc behaviour.
7648 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective
7649 * helper documentation for details.)
7650 */
7651enum bpf_kfunc_flags {
7652 BPF_F_PAD_ZEROS = (1ULL << 0),
7653};
7654
7655/*
7656 * Values of a BPF_MAP_TYPE_INSN_ARRAY entry must be of this type.
7657 *
7658 * Before the map is used the orig_off field should point to an
7659 * instruction inside the program being loaded. The other fields
7660 * must be set to 0.
7661 *
7662 * After the program is loaded, the xlated_off will be adjusted
7663 * by the verifier to point to the index of the original instruction
7664 * in the xlated program. If the instruction is deleted, it will
7665 * be set to (u32)-1. The jitted_off will be set to the corresponding
7666 * offset in the jitted image of the program.
7667 */
7668struct bpf_insn_array_value {
7669 __u32 orig_off;
7670 __u32 xlated_off;
7671 __u32 jitted_off;
7672 __u32 :32;
7673};
7674
7675#endif /* _UAPI__LINUX_BPF_H__ */