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