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 * Walks the immediate children of the specified parent
124 * cgroup_subsys_state. Unlike BPF_CGROUP_ITER_DESCENDANTS_PRE,
125 * BPF_CGROUP_ITER_DESCENDANTS_POST, and BPF_CGROUP_ITER_ANCESTORS_UP
126 * the iterator does not include the specified parent as one of the
127 * returned iterator elements.
128 */
129 BPF_CGROUP_ITER_CHILDREN,
130};
131
132union bpf_iter_link_info {
133 struct {
134 __u32 map_fd;
135 } map;
136 struct {
137 enum bpf_cgroup_iter_order order;
138
139 /* At most one of cgroup_fd and cgroup_id can be non-zero. If
140 * both are zero, the walk starts from the default cgroup v2
141 * root. For walking v1 hierarchy, one should always explicitly
142 * specify cgroup_fd.
143 */
144 __u32 cgroup_fd;
145 __u64 cgroup_id;
146 } cgroup;
147 /* Parameters of task iterators. */
148 struct {
149 __u32 tid;
150 __u32 pid;
151 __u32 pid_fd;
152 } task;
153};
154
155/* BPF syscall commands, see bpf(2) man-page for more details. */
156/**
157 * DOC: eBPF Syscall Preamble
158 *
159 * The operation to be performed by the **bpf**\ () system call is determined
160 * by the *cmd* argument. Each operation takes an accompanying argument,
161 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
162 * below). The size argument is the size of the union pointed to by *attr*.
163 */
164/**
165 * DOC: eBPF Syscall Commands
166 *
167 * BPF_MAP_CREATE
168 * Description
169 * Create a map and return a file descriptor that refers to the
170 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
171 * is automatically enabled for the new file descriptor.
172 *
173 * Applying **close**\ (2) to the file descriptor returned by
174 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
175 *
176 * Return
177 * A new file descriptor (a nonnegative integer), or -1 if an
178 * error occurred (in which case, *errno* is set appropriately).
179 *
180 * BPF_MAP_LOOKUP_ELEM
181 * Description
182 * Look up an element with a given *key* in the map referred to
183 * by the file descriptor *map_fd*.
184 *
185 * The *flags* argument may be specified as one of the
186 * following:
187 *
188 * **BPF_F_LOCK**
189 * Look up the value of a spin-locked map without
190 * returning the lock. This must be specified if the
191 * elements contain a spinlock.
192 *
193 * Return
194 * Returns zero on success. On error, -1 is returned and *errno*
195 * is set appropriately.
196 *
197 * BPF_MAP_UPDATE_ELEM
198 * Description
199 * Create or update an element (key/value pair) in a specified map.
200 *
201 * The *flags* argument should be specified as one of the
202 * following:
203 *
204 * **BPF_ANY**
205 * Create a new element or update an existing element.
206 * **BPF_NOEXIST**
207 * Create a new element only if it did not exist.
208 * **BPF_EXIST**
209 * Update an existing element.
210 * **BPF_F_LOCK**
211 * Update a spin_lock-ed map element.
212 *
213 * Return
214 * Returns zero on success. On error, -1 is returned and *errno*
215 * is set appropriately.
216 *
217 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
218 * **E2BIG**, **EEXIST**, or **ENOENT**.
219 *
220 * **E2BIG**
221 * The number of elements in the map reached the
222 * *max_entries* limit specified at map creation time.
223 * **EEXIST**
224 * If *flags* specifies **BPF_NOEXIST** and the element
225 * with *key* already exists in the map.
226 * **ENOENT**
227 * If *flags* specifies **BPF_EXIST** and the element with
228 * *key* does not exist in the map.
229 *
230 * BPF_MAP_DELETE_ELEM
231 * Description
232 * Look up and delete an element by key in a specified map.
233 *
234 * Return
235 * Returns zero on success. On error, -1 is returned and *errno*
236 * is set appropriately.
237 *
238 * BPF_MAP_GET_NEXT_KEY
239 * Description
240 * Look up an element by key in a specified map and return the key
241 * of the next element. Can be used to iterate over all elements
242 * in the map.
243 *
244 * Return
245 * Returns zero on success. On error, -1 is returned and *errno*
246 * is set appropriately.
247 *
248 * The following cases can be used to iterate over all elements of
249 * the map:
250 *
251 * * If *key* is not found, the operation returns zero and sets
252 * the *next_key* pointer to the key of the first element.
253 * * If *key* is found, the operation returns zero and sets the
254 * *next_key* pointer to the key of the next element.
255 * * If *key* is the last element, returns -1 and *errno* is set
256 * to **ENOENT**.
257 *
258 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
259 * **EINVAL** on error.
260 *
261 * BPF_PROG_LOAD
262 * Description
263 * Verify and load an eBPF program, returning a new file
264 * descriptor associated with the program.
265 *
266 * Applying **close**\ (2) to the file descriptor returned by
267 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
268 *
269 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
270 * automatically enabled for the new file descriptor.
271 *
272 * Return
273 * A new file descriptor (a nonnegative integer), or -1 if an
274 * error occurred (in which case, *errno* is set appropriately).
275 *
276 * BPF_OBJ_PIN
277 * Description
278 * Pin an eBPF program or map referred by the specified *bpf_fd*
279 * to the provided *pathname* on the filesystem.
280 *
281 * The *pathname* argument must not contain a dot (".").
282 *
283 * On success, *pathname* retains a reference to the eBPF object,
284 * preventing deallocation of the object when the original
285 * *bpf_fd* is closed. This allow the eBPF object to live beyond
286 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
287 * process.
288 *
289 * Applying **unlink**\ (2) or similar calls to the *pathname*
290 * unpins the object from the filesystem, removing the reference.
291 * If no other file descriptors or filesystem nodes refer to the
292 * same object, it will be deallocated (see NOTES).
293 *
294 * The filesystem type for the parent directory of *pathname* must
295 * be **BPF_FS_MAGIC**.
296 *
297 * Return
298 * Returns zero on success. On error, -1 is returned and *errno*
299 * is set appropriately.
300 *
301 * BPF_OBJ_GET
302 * Description
303 * Open a file descriptor for the eBPF object pinned to the
304 * specified *pathname*.
305 *
306 * Return
307 * A new file descriptor (a nonnegative integer), or -1 if an
308 * error occurred (in which case, *errno* is set appropriately).
309 *
310 * BPF_PROG_ATTACH
311 * Description
312 * Attach an eBPF program to a *target_fd* at the specified
313 * *attach_type* hook.
314 *
315 * The *attach_type* specifies the eBPF attachment point to
316 * attach the program to, and must be one of *bpf_attach_type*
317 * (see below).
318 *
319 * The *attach_bpf_fd* must be a valid file descriptor for a
320 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
321 * or sock_ops type corresponding to the specified *attach_type*.
322 *
323 * The *target_fd* must be a valid file descriptor for a kernel
324 * object which depends on the attach type of *attach_bpf_fd*:
325 *
326 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
327 * **BPF_PROG_TYPE_CGROUP_SKB**,
328 * **BPF_PROG_TYPE_CGROUP_SOCK**,
329 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
330 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
331 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
332 * **BPF_PROG_TYPE_SOCK_OPS**
333 *
334 * Control Group v2 hierarchy with the eBPF controller
335 * enabled. Requires the kernel to be compiled with
336 * **CONFIG_CGROUP_BPF**.
337 *
338 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
339 *
340 * Network namespace (eg /proc/self/ns/net).
341 *
342 * **BPF_PROG_TYPE_LIRC_MODE2**
343 *
344 * LIRC device path (eg /dev/lircN). Requires the kernel
345 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
346 *
347 * **BPF_PROG_TYPE_SK_SKB**,
348 * **BPF_PROG_TYPE_SK_MSG**
349 *
350 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
351 *
352 * Return
353 * Returns zero on success. On error, -1 is returned and *errno*
354 * is set appropriately.
355 *
356 * BPF_PROG_DETACH
357 * Description
358 * Detach the eBPF program associated with the *target_fd* at the
359 * hook specified by *attach_type*. The program must have been
360 * previously attached using **BPF_PROG_ATTACH**.
361 *
362 * Return
363 * Returns zero on success. On error, -1 is returned and *errno*
364 * is set appropriately.
365 *
366 * BPF_PROG_TEST_RUN
367 * Description
368 * Run the eBPF program associated with the *prog_fd* a *repeat*
369 * number of times against a provided program context *ctx_in* and
370 * data *data_in*, and return the modified program context
371 * *ctx_out*, *data_out* (for example, packet data), result of the
372 * execution *retval*, and *duration* of the test run.
373 *
374 * The sizes of the buffers provided as input and output
375 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
376 * be provided in the corresponding variables *ctx_size_in*,
377 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
378 * of these parameters are not provided (ie set to NULL), the
379 * corresponding size field must be zero.
380 *
381 * Some program types have particular requirements:
382 *
383 * **BPF_PROG_TYPE_SK_LOOKUP**
384 * *data_in* and *data_out* must be NULL.
385 *
386 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
387 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
388 *
389 * *ctx_out*, *data_in* and *data_out* must be NULL.
390 * *repeat* must be zero.
391 *
392 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
393 *
394 * Return
395 * Returns zero on success. On error, -1 is returned and *errno*
396 * is set appropriately.
397 *
398 * **ENOSPC**
399 * Either *data_size_out* or *ctx_size_out* is too small.
400 * **ENOTSUPP**
401 * This command is not supported by the program type of
402 * the program referred to by *prog_fd*.
403 *
404 * BPF_PROG_GET_NEXT_ID
405 * Description
406 * Fetch the next eBPF program currently loaded into the kernel.
407 *
408 * Looks for the eBPF program with an id greater than *start_id*
409 * and updates *next_id* on success. If no other eBPF programs
410 * remain with ids higher than *start_id*, returns -1 and sets
411 * *errno* to **ENOENT**.
412 *
413 * Return
414 * Returns zero on success. On error, or when no id remains, -1
415 * is returned and *errno* is set appropriately.
416 *
417 * BPF_MAP_GET_NEXT_ID
418 * Description
419 * Fetch the next eBPF map currently loaded into the kernel.
420 *
421 * Looks for the eBPF map with an id greater than *start_id*
422 * and updates *next_id* on success. If no other eBPF maps
423 * remain with ids higher than *start_id*, returns -1 and sets
424 * *errno* to **ENOENT**.
425 *
426 * Return
427 * Returns zero on success. On error, or when no id remains, -1
428 * is returned and *errno* is set appropriately.
429 *
430 * BPF_PROG_GET_FD_BY_ID
431 * Description
432 * Open a file descriptor for the eBPF program corresponding to
433 * *prog_id*.
434 *
435 * Return
436 * A new file descriptor (a nonnegative integer), or -1 if an
437 * error occurred (in which case, *errno* is set appropriately).
438 *
439 * BPF_MAP_GET_FD_BY_ID
440 * Description
441 * Open a file descriptor for the eBPF map corresponding to
442 * *map_id*.
443 *
444 * Return
445 * A new file descriptor (a nonnegative integer), or -1 if an
446 * error occurred (in which case, *errno* is set appropriately).
447 *
448 * BPF_OBJ_GET_INFO_BY_FD
449 * Description
450 * Obtain information about the eBPF object corresponding to
451 * *bpf_fd*.
452 *
453 * Populates up to *info_len* bytes of *info*, which will be in
454 * one of the following formats depending on the eBPF object type
455 * of *bpf_fd*:
456 *
457 * * **struct bpf_prog_info**
458 * * **struct bpf_map_info**
459 * * **struct bpf_btf_info**
460 * * **struct bpf_link_info**
461 * * **struct bpf_token_info**
462 *
463 * Return
464 * Returns zero on success. On error, -1 is returned and *errno*
465 * is set appropriately.
466 *
467 * BPF_PROG_QUERY
468 * Description
469 * Obtain information about eBPF programs associated with the
470 * specified *attach_type* hook.
471 *
472 * The *target_fd* must be a valid file descriptor for a kernel
473 * object which depends on the attach type of *attach_bpf_fd*:
474 *
475 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
476 * **BPF_PROG_TYPE_CGROUP_SKB**,
477 * **BPF_PROG_TYPE_CGROUP_SOCK**,
478 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
479 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
480 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
481 * **BPF_PROG_TYPE_SOCK_OPS**
482 *
483 * Control Group v2 hierarchy with the eBPF controller
484 * enabled. Requires the kernel to be compiled with
485 * **CONFIG_CGROUP_BPF**.
486 *
487 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
488 *
489 * Network namespace (eg /proc/self/ns/net).
490 *
491 * **BPF_PROG_TYPE_LIRC_MODE2**
492 *
493 * LIRC device path (eg /dev/lircN). Requires the kernel
494 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
495 *
496 * **BPF_PROG_QUERY** always fetches the number of programs
497 * attached and the *attach_flags* which were used to attach those
498 * programs. Additionally, if *prog_ids* is nonzero and the number
499 * of attached programs is less than *prog_cnt*, populates
500 * *prog_ids* with the eBPF program ids of the programs attached
501 * at *target_fd*.
502 *
503 * The following flags may alter the result:
504 *
505 * **BPF_F_QUERY_EFFECTIVE**
506 * Only return information regarding programs which are
507 * currently effective at the specified *target_fd*.
508 *
509 * Return
510 * Returns zero on success. On error, -1 is returned and *errno*
511 * is set appropriately.
512 *
513 * BPF_RAW_TRACEPOINT_OPEN
514 * Description
515 * Attach an eBPF program to a tracepoint *name* to access kernel
516 * internal arguments of the tracepoint in their raw form.
517 *
518 * The *prog_fd* must be a valid file descriptor associated with
519 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
520 *
521 * No ABI guarantees are made about the content of tracepoint
522 * arguments exposed to the corresponding eBPF program.
523 *
524 * Applying **close**\ (2) to the file descriptor returned by
525 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
526 *
527 * Return
528 * A new file descriptor (a nonnegative integer), or -1 if an
529 * error occurred (in which case, *errno* is set appropriately).
530 *
531 * BPF_BTF_LOAD
532 * Description
533 * Verify and load BPF Type Format (BTF) metadata into the kernel,
534 * returning a new file descriptor associated with the metadata.
535 * BTF is described in more detail at
536 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
537 *
538 * The *btf* parameter must point to valid memory providing
539 * *btf_size* bytes of BTF binary metadata.
540 *
541 * The returned file descriptor can be passed to other **bpf**\ ()
542 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
543 * associate the BTF with those objects.
544 *
545 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
546 * parameters to specify a *btf_log_buf*, *btf_log_size* and
547 * *btf_log_level* which allow the kernel to return freeform log
548 * output regarding the BTF verification process.
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_BTF_GET_FD_BY_ID
555 * Description
556 * Open a file descriptor for the BPF Type Format (BTF)
557 * corresponding to *btf_id*.
558 *
559 * Return
560 * A new file descriptor (a nonnegative integer), or -1 if an
561 * error occurred (in which case, *errno* is set appropriately).
562 *
563 * BPF_TASK_FD_QUERY
564 * Description
565 * Obtain information about eBPF programs associated with the
566 * target process identified by *pid* and *fd*.
567 *
568 * If the *pid* and *fd* are associated with a tracepoint, kprobe
569 * or uprobe perf event, then the *prog_id* and *fd_type* will
570 * be populated with the eBPF program id and file descriptor type
571 * of type **bpf_task_fd_type**. If associated with a kprobe or
572 * uprobe, the *probe_offset* and *probe_addr* will also be
573 * populated. Optionally, if *buf* is provided, then up to
574 * *buf_len* bytes of *buf* will be populated with the name of
575 * the tracepoint, kprobe or uprobe.
576 *
577 * The resulting *prog_id* may be introspected in deeper detail
578 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
579 *
580 * Return
581 * Returns zero on success. On error, -1 is returned and *errno*
582 * is set appropriately.
583 *
584 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
585 * Description
586 * Look up an element with the given *key* in the map referred to
587 * by the file descriptor *fd*, and if found, delete the element.
588 *
589 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
590 * types, the *flags* argument needs to be set to 0, but for other
591 * map types, it may be specified as:
592 *
593 * **BPF_F_LOCK**
594 * Look up and delete the value of a spin-locked map
595 * without returning the lock. This must be specified if
596 * the elements contain a spinlock.
597 *
598 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
599 * implement this command as a "pop" operation, deleting the top
600 * element rather than one corresponding to *key*.
601 * The *key* and *key_len* parameters should be zeroed when
602 * issuing this operation for these map types.
603 *
604 * This command is only valid for the following map types:
605 * * **BPF_MAP_TYPE_QUEUE**
606 * * **BPF_MAP_TYPE_STACK**
607 * * **BPF_MAP_TYPE_HASH**
608 * * **BPF_MAP_TYPE_PERCPU_HASH**
609 * * **BPF_MAP_TYPE_LRU_HASH**
610 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
611 *
612 * Return
613 * Returns zero on success. On error, -1 is returned and *errno*
614 * is set appropriately.
615 *
616 * BPF_MAP_FREEZE
617 * Description
618 * Freeze the permissions of the specified map.
619 *
620 * Write permissions may be frozen by passing zero *flags*.
621 * Upon success, no future syscall invocations may alter the
622 * map state of *map_fd*. Write operations from eBPF programs
623 * are still possible for a frozen map.
624 *
625 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
626 *
627 * Return
628 * Returns zero on success. On error, -1 is returned and *errno*
629 * is set appropriately.
630 *
631 * BPF_BTF_GET_NEXT_ID
632 * Description
633 * Fetch the next BPF Type Format (BTF) object currently loaded
634 * into the kernel.
635 *
636 * Looks for the BTF object with an id greater than *start_id*
637 * and updates *next_id* on success. If no other BTF objects
638 * remain with ids higher than *start_id*, returns -1 and sets
639 * *errno* to **ENOENT**.
640 *
641 * Return
642 * Returns zero on success. On error, or when no id remains, -1
643 * is returned and *errno* is set appropriately.
644 *
645 * BPF_MAP_LOOKUP_BATCH
646 * Description
647 * Iterate and fetch multiple elements in a map.
648 *
649 * Two opaque values are used to manage batch operations,
650 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
651 * to NULL to begin the batched operation. After each subsequent
652 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
653 * *out_batch* as the *in_batch* for the next operation to
654 * continue iteration from the current point. Both *in_batch* and
655 * *out_batch* must point to memory large enough to hold a key,
656 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,
657 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters
658 * must be at least 4 bytes wide regardless of key size.
659 *
660 * The *keys* and *values* are output parameters which must point
661 * to memory large enough to hold *count* items based on the key
662 * and value size of the map *map_fd*. The *keys* buffer must be
663 * of *key_size* * *count*. The *values* buffer must be of
664 * *value_size* * *count*.
665 *
666 * The *elem_flags* argument may be specified as one of the
667 * following:
668 *
669 * **BPF_F_LOCK**
670 * Look up the value of a spin-locked map without
671 * returning the lock. This must be specified if the
672 * elements contain a spinlock.
673 *
674 * On success, *count* elements from the map are copied into the
675 * user buffer, with the keys copied into *keys* and the values
676 * copied into the corresponding indices in *values*.
677 *
678 * If an error is returned and *errno* is not **EFAULT**, *count*
679 * is set to the number of successfully processed elements.
680 *
681 * Return
682 * Returns zero on success. On error, -1 is returned and *errno*
683 * is set appropriately.
684 *
685 * May set *errno* to **ENOSPC** to indicate that *keys* or
686 * *values* is too small to dump an entire bucket during
687 * iteration of a hash-based map type.
688 *
689 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
690 * Description
691 * Iterate and delete all elements in a map.
692 *
693 * This operation has the same behavior as
694 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
695 *
696 * * Every element that is successfully returned is also deleted
697 * from the map. This is at least *count* elements. Note that
698 * *count* is both an input and an output parameter.
699 * * Upon returning with *errno* set to **EFAULT**, up to
700 * *count* elements may be deleted without returning the keys
701 * and values of the deleted elements.
702 *
703 * Return
704 * Returns zero on success. On error, -1 is returned and *errno*
705 * is set appropriately.
706 *
707 * BPF_MAP_UPDATE_BATCH
708 * Description
709 * Update multiple elements in a map by *key*.
710 *
711 * The *keys* and *values* are input parameters which must point
712 * to memory large enough to hold *count* items based on the key
713 * and value size of the map *map_fd*. The *keys* buffer must be
714 * of *key_size* * *count*. The *values* buffer must be of
715 * *value_size* * *count*.
716 *
717 * Each element specified in *keys* is sequentially updated to the
718 * value in the corresponding index in *values*. The *in_batch*
719 * and *out_batch* parameters are ignored and should be zeroed.
720 *
721 * The *elem_flags* argument should be specified as one of the
722 * following:
723 *
724 * **BPF_ANY**
725 * Create new elements or update a existing elements.
726 * **BPF_NOEXIST**
727 * Create new elements only if they do not exist.
728 * **BPF_EXIST**
729 * Update existing elements.
730 * **BPF_F_LOCK**
731 * Update spin_lock-ed map elements. This must be
732 * specified if the map value contains a spinlock.
733 *
734 * On success, *count* elements from the map are updated.
735 *
736 * If an error is returned and *errno* is not **EFAULT**, *count*
737 * is set to the number of successfully processed elements.
738 *
739 * Return
740 * Returns zero on success. On error, -1 is returned and *errno*
741 * is set appropriately.
742 *
743 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
744 * **E2BIG**. **E2BIG** indicates that the number of elements in
745 * the map reached the *max_entries* limit specified at map
746 * creation time.
747 *
748 * May set *errno* to one of the following error codes under
749 * specific circumstances:
750 *
751 * **EEXIST**
752 * If *flags* specifies **BPF_NOEXIST** and the element
753 * with *key* already exists in the map.
754 * **ENOENT**
755 * If *flags* specifies **BPF_EXIST** and the element with
756 * *key* does not exist in the map.
757 *
758 * BPF_MAP_DELETE_BATCH
759 * Description
760 * Delete multiple elements in a map by *key*.
761 *
762 * The *keys* parameter is an input parameter which must point
763 * to memory large enough to hold *count* items based on the key
764 * size of the map *map_fd*, that is, *key_size* * *count*.
765 *
766 * Each element specified in *keys* is sequentially deleted. The
767 * *in_batch*, *out_batch*, and *values* parameters are ignored
768 * and should be zeroed.
769 *
770 * The *elem_flags* argument may be specified as one of the
771 * following:
772 *
773 * **BPF_F_LOCK**
774 * Look up the value of a spin-locked map without
775 * returning the lock. This must be specified if the
776 * elements contain a spinlock.
777 *
778 * On success, *count* elements from the map are updated.
779 *
780 * If an error is returned and *errno* is not **EFAULT**, *count*
781 * is set to the number of successfully processed elements. If
782 * *errno* is **EFAULT**, up to *count* elements may be been
783 * deleted.
784 *
785 * Return
786 * Returns zero on success. On error, -1 is returned and *errno*
787 * is set appropriately.
788 *
789 * BPF_LINK_CREATE
790 * Description
791 * Attach an eBPF program to a *target_fd* at the specified
792 * *attach_type* hook and return a file descriptor handle for
793 * managing the link.
794 *
795 * Return
796 * A new file descriptor (a nonnegative integer), or -1 if an
797 * error occurred (in which case, *errno* is set appropriately).
798 *
799 * BPF_LINK_UPDATE
800 * Description
801 * Update the eBPF program in the specified *link_fd* to
802 * *new_prog_fd*.
803 *
804 * Return
805 * Returns zero on success. On error, -1 is returned and *errno*
806 * is set appropriately.
807 *
808 * BPF_LINK_GET_FD_BY_ID
809 * Description
810 * Open a file descriptor for the eBPF Link corresponding to
811 * *link_id*.
812 *
813 * Return
814 * A new file descriptor (a nonnegative integer), or -1 if an
815 * error occurred (in which case, *errno* is set appropriately).
816 *
817 * BPF_LINK_GET_NEXT_ID
818 * Description
819 * Fetch the next eBPF link currently loaded into the kernel.
820 *
821 * Looks for the eBPF link with an id greater than *start_id*
822 * and updates *next_id* on success. If no other eBPF links
823 * remain with ids higher than *start_id*, returns -1 and sets
824 * *errno* to **ENOENT**.
825 *
826 * Return
827 * Returns zero on success. On error, or when no id remains, -1
828 * is returned and *errno* is set appropriately.
829 *
830 * BPF_ENABLE_STATS
831 * Description
832 * Enable eBPF runtime statistics gathering.
833 *
834 * Runtime statistics gathering for the eBPF runtime is disabled
835 * by default to minimize the corresponding performance overhead.
836 * This command enables statistics globally.
837 *
838 * Multiple programs may independently enable statistics.
839 * After gathering the desired statistics, eBPF runtime statistics
840 * may be disabled again by calling **close**\ (2) for the file
841 * descriptor returned by this function. Statistics will only be
842 * disabled system-wide when all outstanding file descriptors
843 * returned by prior calls for this subcommand are closed.
844 *
845 * Return
846 * A new file descriptor (a nonnegative integer), or -1 if an
847 * error occurred (in which case, *errno* is set appropriately).
848 *
849 * BPF_ITER_CREATE
850 * Description
851 * Create an iterator on top of the specified *link_fd* (as
852 * previously created using **BPF_LINK_CREATE**) and return a
853 * file descriptor that can be used to trigger the iteration.
854 *
855 * If the resulting file descriptor is pinned to the filesystem
856 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
857 * for that path will trigger the iterator to read kernel state
858 * using the eBPF program attached to *link_fd*.
859 *
860 * Return
861 * A new file descriptor (a nonnegative integer), or -1 if an
862 * error occurred (in which case, *errno* is set appropriately).
863 *
864 * BPF_LINK_DETACH
865 * Description
866 * Forcefully detach the specified *link_fd* from its
867 * corresponding attachment point.
868 *
869 * Return
870 * Returns zero on success. On error, -1 is returned and *errno*
871 * is set appropriately.
872 *
873 * BPF_PROG_BIND_MAP
874 * Description
875 * Bind a map to the lifetime of an eBPF program.
876 *
877 * The map identified by *map_fd* is bound to the program
878 * identified by *prog_fd* and only released when *prog_fd* is
879 * released. This may be used in cases where metadata should be
880 * associated with a program which otherwise does not contain any
881 * references to the map (for example, embedded in the eBPF
882 * program instructions).
883 *
884 * Return
885 * Returns zero on success. On error, -1 is returned and *errno*
886 * is set appropriately.
887 *
888 * BPF_TOKEN_CREATE
889 * Description
890 * Create BPF token with embedded information about what
891 * BPF-related functionality it allows:
892 * - a set of allowed bpf() syscall commands;
893 * - a set of allowed BPF map types to be created with
894 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;
895 * - a set of allowed BPF program types and BPF program attach
896 * types to be loaded with BPF_PROG_LOAD command, if
897 * BPF_PROG_LOAD itself is allowed.
898 *
899 * BPF token is created (derived) from an instance of BPF FS,
900 * assuming it has necessary delegation mount options specified.
901 * This BPF token can be passed as an extra parameter to various
902 * bpf() syscall commands to grant BPF subsystem functionality to
903 * unprivileged processes.
904 *
905 * When created, BPF token is "associated" with the owning
906 * user namespace of BPF FS instance (super block) that it was
907 * derived from, and subsequent BPF operations performed with
908 * BPF token would be performing capabilities checks (i.e.,
909 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within
910 * that user namespace. Without BPF token, such capabilities
911 * have to be granted in init user namespace, making bpf()
912 * syscall incompatible with user namespace, for the most part.
913 *
914 * Return
915 * A new file descriptor (a nonnegative integer), or -1 if an
916 * error occurred (in which case, *errno* is set appropriately).
917 *
918 * BPF_PROG_STREAM_READ_BY_FD
919 * Description
920 * Read data of a program's BPF stream. The program is identified
921 * by *prog_fd*, and the stream is identified by the *stream_id*.
922 * The data is copied to a buffer pointed to by *stream_buf*, and
923 * filled less than or equal to *stream_buf_len* bytes.
924 *
925 * Return
926 * Number of bytes read from the stream on success, or -1 if an
927 * error occurred (in which case, *errno* is set appropriately).
928 *
929 * BPF_PROG_ASSOC_STRUCT_OPS
930 * Description
931 * Associate a BPF program with a struct_ops map. The struct_ops
932 * map is identified by *map_fd* and the BPF program is
933 * identified by *prog_fd*.
934 *
935 * Return
936 * 0 on success or -1 if an error occurred (in which case,
937 * *errno* is set appropriately).
938 *
939 * NOTES
940 * eBPF objects (maps and programs) can be shared between processes.
941 *
942 * * After **fork**\ (2), the child inherits file descriptors
943 * referring to the same eBPF objects.
944 * * File descriptors referring to eBPF objects can be transferred over
945 * **unix**\ (7) domain sockets.
946 * * File descriptors referring to eBPF objects can be duplicated in the
947 * usual way, using **dup**\ (2) and similar calls.
948 * * File descriptors referring to eBPF objects can be pinned to the
949 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
950 *
951 * An eBPF object is deallocated only after all file descriptors referring
952 * to the object have been closed and no references remain pinned to the
953 * filesystem or attached (for example, bound to a program or device).
954 */
955enum bpf_cmd {
956 BPF_MAP_CREATE,
957 BPF_MAP_LOOKUP_ELEM,
958 BPF_MAP_UPDATE_ELEM,
959 BPF_MAP_DELETE_ELEM,
960 BPF_MAP_GET_NEXT_KEY,
961 BPF_PROG_LOAD,
962 BPF_OBJ_PIN,
963 BPF_OBJ_GET,
964 BPF_PROG_ATTACH,
965 BPF_PROG_DETACH,
966 BPF_PROG_TEST_RUN,
967 BPF_PROG_RUN = BPF_PROG_TEST_RUN,
968 BPF_PROG_GET_NEXT_ID,
969 BPF_MAP_GET_NEXT_ID,
970 BPF_PROG_GET_FD_BY_ID,
971 BPF_MAP_GET_FD_BY_ID,
972 BPF_OBJ_GET_INFO_BY_FD,
973 BPF_PROG_QUERY,
974 BPF_RAW_TRACEPOINT_OPEN,
975 BPF_BTF_LOAD,
976 BPF_BTF_GET_FD_BY_ID,
977 BPF_TASK_FD_QUERY,
978 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
979 BPF_MAP_FREEZE,
980 BPF_BTF_GET_NEXT_ID,
981 BPF_MAP_LOOKUP_BATCH,
982 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
983 BPF_MAP_UPDATE_BATCH,
984 BPF_MAP_DELETE_BATCH,
985 BPF_LINK_CREATE,
986 BPF_LINK_UPDATE,
987 BPF_LINK_GET_FD_BY_ID,
988 BPF_LINK_GET_NEXT_ID,
989 BPF_ENABLE_STATS,
990 BPF_ITER_CREATE,
991 BPF_LINK_DETACH,
992 BPF_PROG_BIND_MAP,
993 BPF_TOKEN_CREATE,
994 BPF_PROG_STREAM_READ_BY_FD,
995 BPF_PROG_ASSOC_STRUCT_OPS,
996 __MAX_BPF_CMD,
997};
998
999enum bpf_map_type {
1000 BPF_MAP_TYPE_UNSPEC,
1001 BPF_MAP_TYPE_HASH,
1002 BPF_MAP_TYPE_ARRAY,
1003 BPF_MAP_TYPE_PROG_ARRAY,
1004 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
1005 BPF_MAP_TYPE_PERCPU_HASH,
1006 BPF_MAP_TYPE_PERCPU_ARRAY,
1007 BPF_MAP_TYPE_STACK_TRACE,
1008 BPF_MAP_TYPE_CGROUP_ARRAY,
1009 BPF_MAP_TYPE_LRU_HASH,
1010 BPF_MAP_TYPE_LRU_PERCPU_HASH,
1011 BPF_MAP_TYPE_LPM_TRIE,
1012 BPF_MAP_TYPE_ARRAY_OF_MAPS,
1013 BPF_MAP_TYPE_HASH_OF_MAPS,
1014 BPF_MAP_TYPE_DEVMAP,
1015 BPF_MAP_TYPE_SOCKMAP,
1016 BPF_MAP_TYPE_CPUMAP,
1017 BPF_MAP_TYPE_XSKMAP,
1018 BPF_MAP_TYPE_SOCKHASH,
1019 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1020 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
1021 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
1022 * both cgroup-attached and other progs and supports all functionality
1023 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
1024 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
1025 */
1026 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
1027 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
1028 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1029 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
1030 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
1031 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1032 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1033 * deprecated.
1034 */
1035 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
1036 BPF_MAP_TYPE_QUEUE,
1037 BPF_MAP_TYPE_STACK,
1038 BPF_MAP_TYPE_SK_STORAGE,
1039 BPF_MAP_TYPE_DEVMAP_HASH,
1040 BPF_MAP_TYPE_STRUCT_OPS,
1041 BPF_MAP_TYPE_RINGBUF,
1042 BPF_MAP_TYPE_INODE_STORAGE,
1043 BPF_MAP_TYPE_TASK_STORAGE,
1044 BPF_MAP_TYPE_BLOOM_FILTER,
1045 BPF_MAP_TYPE_USER_RINGBUF,
1046 BPF_MAP_TYPE_CGRP_STORAGE,
1047 BPF_MAP_TYPE_ARENA,
1048 BPF_MAP_TYPE_INSN_ARRAY,
1049 __MAX_BPF_MAP_TYPE
1050};
1051
1052/* Note that tracing related programs such as
1053 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
1054 * are not subject to a stable API since kernel internal data
1055 * structures can change from release to release and may
1056 * therefore break existing tracing BPF programs. Tracing BPF
1057 * programs correspond to /a/ specific kernel which is to be
1058 * analyzed, and not /a/ specific kernel /and/ all future ones.
1059 */
1060enum bpf_prog_type {
1061 BPF_PROG_TYPE_UNSPEC,
1062 BPF_PROG_TYPE_SOCKET_FILTER,
1063 BPF_PROG_TYPE_KPROBE,
1064 BPF_PROG_TYPE_SCHED_CLS,
1065 BPF_PROG_TYPE_SCHED_ACT,
1066 BPF_PROG_TYPE_TRACEPOINT,
1067 BPF_PROG_TYPE_XDP,
1068 BPF_PROG_TYPE_PERF_EVENT,
1069 BPF_PROG_TYPE_CGROUP_SKB,
1070 BPF_PROG_TYPE_CGROUP_SOCK,
1071 BPF_PROG_TYPE_LWT_IN,
1072 BPF_PROG_TYPE_LWT_OUT,
1073 BPF_PROG_TYPE_LWT_XMIT,
1074 BPF_PROG_TYPE_SOCK_OPS,
1075 BPF_PROG_TYPE_SK_SKB,
1076 BPF_PROG_TYPE_CGROUP_DEVICE,
1077 BPF_PROG_TYPE_SK_MSG,
1078 BPF_PROG_TYPE_RAW_TRACEPOINT,
1079 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
1080 BPF_PROG_TYPE_LWT_SEG6LOCAL,
1081 BPF_PROG_TYPE_LIRC_MODE2,
1082 BPF_PROG_TYPE_SK_REUSEPORT,
1083 BPF_PROG_TYPE_FLOW_DISSECTOR,
1084 BPF_PROG_TYPE_CGROUP_SYSCTL,
1085 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
1086 BPF_PROG_TYPE_CGROUP_SOCKOPT,
1087 BPF_PROG_TYPE_TRACING,
1088 BPF_PROG_TYPE_STRUCT_OPS,
1089 BPF_PROG_TYPE_EXT,
1090 BPF_PROG_TYPE_LSM,
1091 BPF_PROG_TYPE_SK_LOOKUP,
1092 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
1093 BPF_PROG_TYPE_NETFILTER,
1094 __MAX_BPF_PROG_TYPE
1095};
1096
1097enum bpf_attach_type {
1098 BPF_CGROUP_INET_INGRESS,
1099 BPF_CGROUP_INET_EGRESS,
1100 BPF_CGROUP_INET_SOCK_CREATE,
1101 BPF_CGROUP_SOCK_OPS,
1102 BPF_SK_SKB_STREAM_PARSER,
1103 BPF_SK_SKB_STREAM_VERDICT,
1104 BPF_CGROUP_DEVICE,
1105 BPF_SK_MSG_VERDICT,
1106 BPF_CGROUP_INET4_BIND,
1107 BPF_CGROUP_INET6_BIND,
1108 BPF_CGROUP_INET4_CONNECT,
1109 BPF_CGROUP_INET6_CONNECT,
1110 BPF_CGROUP_INET4_POST_BIND,
1111 BPF_CGROUP_INET6_POST_BIND,
1112 BPF_CGROUP_UDP4_SENDMSG,
1113 BPF_CGROUP_UDP6_SENDMSG,
1114 BPF_LIRC_MODE2,
1115 BPF_FLOW_DISSECTOR,
1116 BPF_CGROUP_SYSCTL,
1117 BPF_CGROUP_UDP4_RECVMSG,
1118 BPF_CGROUP_UDP6_RECVMSG,
1119 BPF_CGROUP_GETSOCKOPT,
1120 BPF_CGROUP_SETSOCKOPT,
1121 BPF_TRACE_RAW_TP,
1122 BPF_TRACE_FENTRY,
1123 BPF_TRACE_FEXIT,
1124 BPF_MODIFY_RETURN,
1125 BPF_LSM_MAC,
1126 BPF_TRACE_ITER,
1127 BPF_CGROUP_INET4_GETPEERNAME,
1128 BPF_CGROUP_INET6_GETPEERNAME,
1129 BPF_CGROUP_INET4_GETSOCKNAME,
1130 BPF_CGROUP_INET6_GETSOCKNAME,
1131 BPF_XDP_DEVMAP,
1132 BPF_CGROUP_INET_SOCK_RELEASE,
1133 BPF_XDP_CPUMAP,
1134 BPF_SK_LOOKUP,
1135 BPF_XDP,
1136 BPF_SK_SKB_VERDICT,
1137 BPF_SK_REUSEPORT_SELECT,
1138 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1139 BPF_PERF_EVENT,
1140 BPF_TRACE_KPROBE_MULTI,
1141 BPF_LSM_CGROUP,
1142 BPF_STRUCT_OPS,
1143 BPF_NETFILTER,
1144 BPF_TCX_INGRESS,
1145 BPF_TCX_EGRESS,
1146 BPF_TRACE_UPROBE_MULTI,
1147 BPF_CGROUP_UNIX_CONNECT,
1148 BPF_CGROUP_UNIX_SENDMSG,
1149 BPF_CGROUP_UNIX_RECVMSG,
1150 BPF_CGROUP_UNIX_GETPEERNAME,
1151 BPF_CGROUP_UNIX_GETSOCKNAME,
1152 BPF_NETKIT_PRIMARY,
1153 BPF_NETKIT_PEER,
1154 BPF_TRACE_KPROBE_SESSION,
1155 BPF_TRACE_UPROBE_SESSION,
1156 BPF_TRACE_FSESSION,
1157 __MAX_BPF_ATTACH_TYPE
1158};
1159
1160#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1161
1162/* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]
1163 * in sync with the definitions below.
1164 */
1165enum bpf_link_type {
1166 BPF_LINK_TYPE_UNSPEC = 0,
1167 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1168 BPF_LINK_TYPE_TRACING = 2,
1169 BPF_LINK_TYPE_CGROUP = 3,
1170 BPF_LINK_TYPE_ITER = 4,
1171 BPF_LINK_TYPE_NETNS = 5,
1172 BPF_LINK_TYPE_XDP = 6,
1173 BPF_LINK_TYPE_PERF_EVENT = 7,
1174 BPF_LINK_TYPE_KPROBE_MULTI = 8,
1175 BPF_LINK_TYPE_STRUCT_OPS = 9,
1176 BPF_LINK_TYPE_NETFILTER = 10,
1177 BPF_LINK_TYPE_TCX = 11,
1178 BPF_LINK_TYPE_UPROBE_MULTI = 12,
1179 BPF_LINK_TYPE_NETKIT = 13,
1180 BPF_LINK_TYPE_SOCKMAP = 14,
1181 __MAX_BPF_LINK_TYPE,
1182};
1183
1184#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
1185
1186enum bpf_perf_event_type {
1187 BPF_PERF_EVENT_UNSPEC = 0,
1188 BPF_PERF_EVENT_UPROBE = 1,
1189 BPF_PERF_EVENT_URETPROBE = 2,
1190 BPF_PERF_EVENT_KPROBE = 3,
1191 BPF_PERF_EVENT_KRETPROBE = 4,
1192 BPF_PERF_EVENT_TRACEPOINT = 5,
1193 BPF_PERF_EVENT_EVENT = 6,
1194};
1195
1196/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1197 *
1198 * NONE(default): No further bpf programs allowed in the subtree.
1199 *
1200 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1201 * the program in this cgroup yields to sub-cgroup program.
1202 *
1203 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1204 * that cgroup program gets run in addition to the program in this cgroup.
1205 *
1206 * Only one program is allowed to be attached to a cgroup with
1207 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1208 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1209 * release old program and attach the new one. Attach flags has to match.
1210 *
1211 * Multiple programs are allowed to be attached to a cgroup with
1212 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1213 * (those that were attached first, run first)
1214 * The programs of sub-cgroup are executed first, then programs of
1215 * this cgroup and then programs of parent cgroup.
1216 * When children program makes decision (like picking TCP CA or sock bind)
1217 * parent program has a chance to override it.
1218 *
1219 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1220 * programs for a cgroup. Though it's possible to replace an old program at
1221 * any position by also specifying BPF_F_REPLACE flag and position itself in
1222 * replace_bpf_fd attribute. Old program at this position will be released.
1223 *
1224 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1225 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1226 * Ex1:
1227 * cgrp1 (MULTI progs A, B) ->
1228 * cgrp2 (OVERRIDE prog C) ->
1229 * cgrp3 (MULTI prog D) ->
1230 * cgrp4 (OVERRIDE prog E) ->
1231 * cgrp5 (NONE prog F)
1232 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1233 * if prog F is detached, the execution is E,D,A,B
1234 * if prog F and D are detached, the execution is E,A,B
1235 * if prog F, E and D are detached, the execution is C,A,B
1236 *
1237 * All eligible programs are executed regardless of return code from
1238 * earlier programs.
1239 */
1240#define BPF_F_ALLOW_OVERRIDE (1U << 0)
1241#define BPF_F_ALLOW_MULTI (1U << 1)
1242/* Generic attachment flags. */
1243#define BPF_F_REPLACE (1U << 2)
1244#define BPF_F_BEFORE (1U << 3)
1245#define BPF_F_AFTER (1U << 4)
1246#define BPF_F_ID (1U << 5)
1247#define BPF_F_PREORDER (1U << 6)
1248#define BPF_F_LINK BPF_F_LINK /* 1 << 13 */
1249
1250/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1251 * verifier will perform strict alignment checking as if the kernel
1252 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1253 * and NET_IP_ALIGN defined to 2.
1254 */
1255#define BPF_F_STRICT_ALIGNMENT (1U << 0)
1256
1257/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1258 * verifier will allow any alignment whatsoever. On platforms
1259 * with strict alignment requirements for loads ands stores (such
1260 * as sparc and mips) the verifier validates that all loads and
1261 * stores provably follow this requirement. This flag turns that
1262 * checking and enforcement off.
1263 *
1264 * It is mostly used for testing when we want to validate the
1265 * context and memory access aspects of the verifier, but because
1266 * of an unaligned access the alignment check would trigger before
1267 * the one we are interested in.
1268 */
1269#define BPF_F_ANY_ALIGNMENT (1U << 1)
1270
1271/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1272 * Verifier does sub-register def/use analysis and identifies instructions whose
1273 * def only matters for low 32-bit, high 32-bit is never referenced later
1274 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1275 * that it is safe to ignore clearing high 32-bit for these instructions. This
1276 * saves some back-ends a lot of code-gen. However such optimization is not
1277 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1278 * hence hasn't used verifier's analysis result. But, we really want to have a
1279 * way to be able to verify the correctness of the described optimization on
1280 * x86_64 on which testsuites are frequently exercised.
1281 *
1282 * So, this flag is introduced. Once it is set, verifier will randomize high
1283 * 32-bit for those instructions who has been identified as safe to ignore them.
1284 * Then, if verifier is not doing correct analysis, such randomization will
1285 * regress tests to expose bugs.
1286 */
1287#define BPF_F_TEST_RND_HI32 (1U << 2)
1288
1289/* The verifier internal test flag. Behavior is undefined */
1290#define BPF_F_TEST_STATE_FREQ (1U << 3)
1291
1292/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1293 * restrict map and helper usage for such programs. Sleepable BPF programs can
1294 * only be attached to hooks where kernel execution context allows sleeping.
1295 * Such programs are allowed to use helpers that may sleep like
1296 * bpf_copy_from_user().
1297 */
1298#define BPF_F_SLEEPABLE (1U << 4)
1299
1300/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1301 * fully support xdp frags.
1302 */
1303#define BPF_F_XDP_HAS_FRAGS (1U << 5)
1304
1305/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1306 * program becomes device-bound but can access XDP metadata.
1307 */
1308#define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6)
1309
1310/* The verifier internal test flag. Behavior is undefined */
1311#define BPF_F_TEST_REG_INVARIANTS (1U << 7)
1312
1313/* link_create.kprobe_multi.flags used in LINK_CREATE command for
1314 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1315 */
1316enum {
1317 BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
1318};
1319
1320/* link_create.uprobe_multi.flags used in LINK_CREATE command for
1321 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1322 */
1323enum {
1324 BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
1325};
1326
1327/* link_create.netfilter.flags used in LINK_CREATE command for
1328 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1329 */
1330#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1331
1332/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1333 * the following extensions:
1334 *
1335 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1336 * insn[0].imm: map fd or fd_idx
1337 * insn[1].imm: 0
1338 * insn[0].off: 0
1339 * insn[1].off: 0
1340 * ldimm64 rewrite: address of map
1341 * verifier type: CONST_PTR_TO_MAP
1342 */
1343#define BPF_PSEUDO_MAP_FD 1
1344#define BPF_PSEUDO_MAP_IDX 5
1345
1346/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1347 * insn[0].imm: map fd or fd_idx
1348 * insn[1].imm: offset into value
1349 * insn[0].off: 0
1350 * insn[1].off: 0
1351 * ldimm64 rewrite: address of map[0]+offset
1352 * verifier type: PTR_TO_MAP_VALUE
1353 */
1354#define BPF_PSEUDO_MAP_VALUE 2
1355#define BPF_PSEUDO_MAP_IDX_VALUE 6
1356
1357/* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1358 * insn[0].imm: kernel btd id of VAR
1359 * insn[1].imm: 0
1360 * insn[0].off: 0
1361 * insn[1].off: 0
1362 * ldimm64 rewrite: address of the kernel variable
1363 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1364 * is struct/union.
1365 */
1366#define BPF_PSEUDO_BTF_ID 3
1367/* insn[0].src_reg: BPF_PSEUDO_FUNC
1368 * insn[0].imm: insn offset to the func
1369 * insn[1].imm: 0
1370 * insn[0].off: 0
1371 * insn[1].off: 0
1372 * ldimm64 rewrite: address of the function
1373 * verifier type: PTR_TO_FUNC.
1374 */
1375#define BPF_PSEUDO_FUNC 4
1376
1377/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1378 * offset to another bpf function
1379 */
1380#define BPF_PSEUDO_CALL 1
1381/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1382 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1383 */
1384#define BPF_PSEUDO_KFUNC_CALL 2
1385
1386enum bpf_addr_space_cast {
1387 BPF_ADDR_SPACE_CAST = 1,
1388};
1389
1390/* flags for BPF_MAP_UPDATE_ELEM command */
1391enum {
1392 BPF_ANY = 0, /* create new element or update existing */
1393 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1394 BPF_EXIST = 2, /* update existing element */
1395 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1396 BPF_F_CPU = 8, /* cpu flag for percpu maps, upper 32-bit of flags is a cpu number */
1397 BPF_F_ALL_CPUS = 16, /* update value across all CPUs for percpu maps */
1398};
1399
1400/* flags for BPF_MAP_CREATE command */
1401enum {
1402 BPF_F_NO_PREALLOC = (1U << 0),
1403/* Instead of having one common LRU list in the
1404 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1405 * which can scale and perform better.
1406 * Note, the LRU nodes (including free nodes) cannot be moved
1407 * across different LRU lists.
1408 */
1409 BPF_F_NO_COMMON_LRU = (1U << 1),
1410/* Specify numa node during map creation */
1411 BPF_F_NUMA_NODE = (1U << 2),
1412
1413/* Flags for accessing BPF object from syscall side. */
1414 BPF_F_RDONLY = (1U << 3),
1415 BPF_F_WRONLY = (1U << 4),
1416
1417/* Flag for stack_map, store build_id+offset instead of pointer */
1418 BPF_F_STACK_BUILD_ID = (1U << 5),
1419
1420/* Zero-initialize hash function seed. This should only be used for testing. */
1421 BPF_F_ZERO_SEED = (1U << 6),
1422
1423/* Flags for accessing BPF object from program side. */
1424 BPF_F_RDONLY_PROG = (1U << 7),
1425 BPF_F_WRONLY_PROG = (1U << 8),
1426
1427/* Clone map from listener for newly accepted socket */
1428 BPF_F_CLONE = (1U << 9),
1429
1430/* Enable memory-mapping BPF map */
1431 BPF_F_MMAPABLE = (1U << 10),
1432
1433/* Share perf_event among processes */
1434 BPF_F_PRESERVE_ELEMS = (1U << 11),
1435
1436/* Create a map that is suitable to be an inner map with dynamic max entries */
1437 BPF_F_INNER_MAP = (1U << 12),
1438
1439/* Create a map that will be registered/unregesitered by the backed bpf_link */
1440 BPF_F_LINK = (1U << 13),
1441
1442/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1443 BPF_F_PATH_FD = (1U << 14),
1444
1445/* Flag for value_type_btf_obj_fd, the fd is available */
1446 BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15),
1447
1448/* BPF token FD is passed in a corresponding command's token_fd field */
1449 BPF_F_TOKEN_FD = (1U << 16),
1450
1451/* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */
1452 BPF_F_SEGV_ON_FAULT = (1U << 17),
1453
1454/* Do not translate kernel bpf_arena pointers to user pointers */
1455 BPF_F_NO_USER_CONV = (1U << 18),
1456
1457/* Enable BPF ringbuf overwrite mode */
1458 BPF_F_RB_OVERWRITE = (1U << 19),
1459};
1460
1461/* Flags for BPF_PROG_QUERY. */
1462
1463/* Query effective (directly attached + inherited from ancestor cgroups)
1464 * programs that will be executed for events within a cgroup.
1465 * attach_flags with this flag are always returned 0.
1466 */
1467#define BPF_F_QUERY_EFFECTIVE (1U << 0)
1468
1469/* Flags for BPF_PROG_TEST_RUN */
1470
1471/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1472#define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1473/* If set, XDP frames will be transmitted after processing */
1474#define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
1475/* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */
1476#define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2)
1477
1478/* type for BPF_ENABLE_STATS */
1479enum bpf_stats_type {
1480 /* enabled run_time_ns and run_cnt */
1481 BPF_STATS_RUN_TIME = 0,
1482};
1483
1484enum bpf_stack_build_id_status {
1485 /* user space need an empty entry to identify end of a trace */
1486 BPF_STACK_BUILD_ID_EMPTY = 0,
1487 /* with valid build_id and offset */
1488 BPF_STACK_BUILD_ID_VALID = 1,
1489 /* couldn't get build_id, fallback to ip */
1490 BPF_STACK_BUILD_ID_IP = 2,
1491};
1492
1493#define BPF_BUILD_ID_SIZE 20
1494struct bpf_stack_build_id {
1495 __s32 status;
1496 unsigned char build_id[BPF_BUILD_ID_SIZE];
1497 union {
1498 __u64 offset;
1499 __u64 ip;
1500 };
1501};
1502
1503#define BPF_OBJ_NAME_LEN 16U
1504
1505enum {
1506 BPF_STREAM_STDOUT = 1,
1507 BPF_STREAM_STDERR = 2,
1508};
1509
1510union bpf_attr {
1511 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1512 __u32 map_type; /* one of enum bpf_map_type */
1513 __u32 key_size; /* size of key in bytes */
1514 __u32 value_size; /* size of value in bytes */
1515 __u32 max_entries; /* max number of entries in a map */
1516 __u32 map_flags; /* BPF_MAP_CREATE related
1517 * flags defined above.
1518 */
1519 __u32 inner_map_fd; /* fd pointing to the inner map */
1520 __u32 numa_node; /* numa node (effective only if
1521 * BPF_F_NUMA_NODE is set).
1522 */
1523 char map_name[BPF_OBJ_NAME_LEN];
1524 __u32 map_ifindex; /* ifindex of netdev to create on */
1525 __u32 btf_fd; /* fd pointing to a BTF type data */
1526 __u32 btf_key_type_id; /* BTF type_id of the key */
1527 __u32 btf_value_type_id; /* BTF type_id of the value */
1528 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1529 * struct stored as the
1530 * map value
1531 */
1532 /* Any per-map-type extra fields
1533 *
1534 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1535 * number of hash functions (if 0, the bloom filter will default
1536 * to using 5 hash functions).
1537 *
1538 * BPF_MAP_TYPE_ARENA - contains the address where user space
1539 * is going to mmap() the arena. It has to be page aligned.
1540 */
1541 __u64 map_extra;
1542
1543 __s32 value_type_btf_obj_fd; /* fd pointing to a BTF
1544 * type data for
1545 * btf_vmlinux_value_type_id.
1546 */
1547 /* BPF token FD to use with BPF_MAP_CREATE operation.
1548 * If provided, map_flags should have BPF_F_TOKEN_FD flag set.
1549 */
1550 __s32 map_token_fd;
1551
1552 /* Hash of the program that has exclusive access to the map.
1553 */
1554 __aligned_u64 excl_prog_hash;
1555 /* Size of the passed excl_prog_hash. */
1556 __u32 excl_prog_hash_size;
1557 };
1558
1559 struct { /* anonymous struct used by BPF_MAP_*_ELEM and BPF_MAP_FREEZE commands */
1560 __u32 map_fd;
1561 __aligned_u64 key;
1562 union {
1563 __aligned_u64 value;
1564 __aligned_u64 next_key;
1565 };
1566 __u64 flags;
1567 };
1568
1569 struct { /* struct used by BPF_MAP_*_BATCH commands */
1570 __aligned_u64 in_batch; /* start batch,
1571 * NULL to start from beginning
1572 */
1573 __aligned_u64 out_batch; /* output: next start batch */
1574 __aligned_u64 keys;
1575 __aligned_u64 values;
1576 __u32 count; /* input/output:
1577 * input: # of key/value
1578 * elements
1579 * output: # of filled elements
1580 */
1581 __u32 map_fd;
1582 __u64 elem_flags;
1583 __u64 flags;
1584 } batch;
1585
1586 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1587 __u32 prog_type; /* one of enum bpf_prog_type */
1588 __u32 insn_cnt;
1589 __aligned_u64 insns;
1590 __aligned_u64 license;
1591 __u32 log_level; /* verbosity level of verifier */
1592 __u32 log_size; /* size of user buffer */
1593 __aligned_u64 log_buf; /* user supplied buffer */
1594 __u32 kern_version; /* not used */
1595 __u32 prog_flags;
1596 char prog_name[BPF_OBJ_NAME_LEN];
1597 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1598 /* For some prog types expected attach type must be known at
1599 * load time to verify attach type specific parts of prog
1600 * (context accesses, allowed helpers, etc).
1601 */
1602 __u32 expected_attach_type;
1603 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1604 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1605 __aligned_u64 func_info; /* func info */
1606 __u32 func_info_cnt; /* number of bpf_func_info records */
1607 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1608 __aligned_u64 line_info; /* line info */
1609 __u32 line_info_cnt; /* number of bpf_line_info records */
1610 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1611 union {
1612 /* valid prog_fd to attach to bpf prog */
1613 __u32 attach_prog_fd;
1614 /* or valid module BTF object fd or 0 to attach to vmlinux */
1615 __u32 attach_btf_obj_fd;
1616 };
1617 __u32 core_relo_cnt; /* number of bpf_core_relo */
1618 __aligned_u64 fd_array; /* array of FDs */
1619 __aligned_u64 core_relos;
1620 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1621 /* output: actual total log contents size (including termintaing zero).
1622 * It could be both larger than original log_size (if log was
1623 * truncated), or smaller (if log buffer wasn't filled completely).
1624 */
1625 __u32 log_true_size;
1626 /* BPF token FD to use with BPF_PROG_LOAD operation.
1627 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.
1628 */
1629 __s32 prog_token_fd;
1630 /* The fd_array_cnt can be used to pass the length of the
1631 * fd_array array. In this case all the [map] file descriptors
1632 * passed in this array will be bound to the program, even if
1633 * the maps are not referenced directly. The functionality is
1634 * similar to the BPF_PROG_BIND_MAP syscall, but maps can be
1635 * used by the verifier during the program load. If provided,
1636 * then the fd_array[0,...,fd_array_cnt-1] is expected to be
1637 * continuous.
1638 */
1639 __u32 fd_array_cnt;
1640 /* Pointer to a buffer containing the signature of the BPF
1641 * program.
1642 */
1643 __aligned_u64 signature;
1644 /* Size of the signature buffer in bytes. */
1645 __u32 signature_size;
1646 /* ID of the kernel keyring to be used for signature
1647 * verification.
1648 */
1649 __s32 keyring_id;
1650 };
1651
1652 struct { /* anonymous struct used by BPF_OBJ_* commands */
1653 __aligned_u64 pathname;
1654 __u32 bpf_fd;
1655 __u32 file_flags;
1656 /* Same as dirfd in openat() syscall; see openat(2)
1657 * manpage for details of path FD and pathname semantics;
1658 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1659 * file_flags field, otherwise it should be set to zero;
1660 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1661 */
1662 __s32 path_fd;
1663 };
1664
1665 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1666 union {
1667 __u32 target_fd; /* target object to attach to or ... */
1668 __u32 target_ifindex; /* target ifindex */
1669 };
1670 __u32 attach_bpf_fd;
1671 __u32 attach_type;
1672 __u32 attach_flags;
1673 __u32 replace_bpf_fd;
1674 union {
1675 __u32 relative_fd;
1676 __u32 relative_id;
1677 };
1678 __u64 expected_revision;
1679 };
1680
1681 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1682 __u32 prog_fd;
1683 __u32 retval;
1684 __u32 data_size_in; /* input: len of data_in */
1685 __u32 data_size_out; /* input/output: len of data_out
1686 * returns ENOSPC if data_out
1687 * is too small.
1688 */
1689 __aligned_u64 data_in;
1690 __aligned_u64 data_out;
1691 __u32 repeat;
1692 __u32 duration;
1693 __u32 ctx_size_in; /* input: len of ctx_in */
1694 __u32 ctx_size_out; /* input/output: len of ctx_out
1695 * returns ENOSPC if ctx_out
1696 * is too small.
1697 */
1698 __aligned_u64 ctx_in;
1699 __aligned_u64 ctx_out;
1700 __u32 flags;
1701 __u32 cpu;
1702 __u32 batch_size;
1703 } test;
1704
1705 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1706 union {
1707 __u32 start_id;
1708 __u32 prog_id;
1709 __u32 map_id;
1710 __u32 btf_id;
1711 __u32 link_id;
1712 };
1713 __u32 next_id;
1714 __u32 open_flags;
1715 __s32 fd_by_id_token_fd;
1716 };
1717
1718 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1719 __u32 bpf_fd;
1720 __u32 info_len;
1721 __aligned_u64 info;
1722 } info;
1723
1724 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1725 union {
1726 __u32 target_fd; /* target object to query or ... */
1727 __u32 target_ifindex; /* target ifindex */
1728 };
1729 __u32 attach_type;
1730 __u32 query_flags;
1731 __u32 attach_flags;
1732 __aligned_u64 prog_ids;
1733 union {
1734 __u32 prog_cnt;
1735 __u32 count;
1736 };
1737 __u32 :32;
1738 /* output: per-program attach_flags.
1739 * not allowed to be set during effective query.
1740 */
1741 __aligned_u64 prog_attach_flags;
1742 __aligned_u64 link_ids;
1743 __aligned_u64 link_attach_flags;
1744 __u64 revision;
1745 } query;
1746
1747 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1748 __u64 name;
1749 __u32 prog_fd;
1750 __u32 :32;
1751 __aligned_u64 cookie;
1752 } raw_tracepoint;
1753
1754 struct { /* anonymous struct for BPF_BTF_LOAD */
1755 __aligned_u64 btf;
1756 __aligned_u64 btf_log_buf;
1757 __u32 btf_size;
1758 __u32 btf_log_size;
1759 __u32 btf_log_level;
1760 /* output: actual total log contents size (including termintaing zero).
1761 * It could be both larger than original log_size (if log was
1762 * truncated), or smaller (if log buffer wasn't filled completely).
1763 */
1764 __u32 btf_log_true_size;
1765 __u32 btf_flags;
1766 /* BPF token FD to use with BPF_BTF_LOAD operation.
1767 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.
1768 */
1769 __s32 btf_token_fd;
1770 };
1771
1772 struct {
1773 __u32 pid; /* input: pid */
1774 __u32 fd; /* input: fd */
1775 __u32 flags; /* input: flags */
1776 __u32 buf_len; /* input/output: buf len */
1777 __aligned_u64 buf; /* input/output:
1778 * tp_name for tracepoint
1779 * symbol for kprobe
1780 * filename for uprobe
1781 */
1782 __u32 prog_id; /* output: prod_id */
1783 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1784 __u64 probe_offset; /* output: probe_offset */
1785 __u64 probe_addr; /* output: probe_addr */
1786 } task_fd_query;
1787
1788 struct { /* struct used by BPF_LINK_CREATE command */
1789 union {
1790 __u32 prog_fd; /* eBPF program to attach */
1791 __u32 map_fd; /* struct_ops to attach */
1792 };
1793 union {
1794 __u32 target_fd; /* target object to attach to or ... */
1795 __u32 target_ifindex; /* target ifindex */
1796 };
1797 __u32 attach_type; /* attach type */
1798 __u32 flags; /* extra flags */
1799 union {
1800 __u32 target_btf_id; /* btf_id of target to attach to */
1801 struct {
1802 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1803 __u32 iter_info_len; /* iter_info length */
1804 };
1805 struct {
1806 /* black box user-provided value passed through
1807 * to BPF program at the execution time and
1808 * accessible through bpf_get_attach_cookie() BPF helper
1809 */
1810 __u64 bpf_cookie;
1811 } perf_event;
1812 struct {
1813 __u32 flags;
1814 __u32 cnt;
1815 __aligned_u64 syms;
1816 __aligned_u64 addrs;
1817 __aligned_u64 cookies;
1818 } kprobe_multi;
1819 struct {
1820 /* this is overlaid with the target_btf_id above. */
1821 __u32 target_btf_id;
1822 /* black box user-provided value passed through
1823 * to BPF program at the execution time and
1824 * accessible through bpf_get_attach_cookie() BPF helper
1825 */
1826 __u64 cookie;
1827 } tracing;
1828 struct {
1829 __u32 pf;
1830 __u32 hooknum;
1831 __s32 priority;
1832 __u32 flags;
1833 } netfilter;
1834 struct {
1835 union {
1836 __u32 relative_fd;
1837 __u32 relative_id;
1838 };
1839 __u64 expected_revision;
1840 } tcx;
1841 struct {
1842 __aligned_u64 path;
1843 __aligned_u64 offsets;
1844 __aligned_u64 ref_ctr_offsets;
1845 __aligned_u64 cookies;
1846 __u32 cnt;
1847 __u32 flags;
1848 __u32 pid;
1849 } uprobe_multi;
1850 struct {
1851 union {
1852 __u32 relative_fd;
1853 __u32 relative_id;
1854 };
1855 __u64 expected_revision;
1856 } netkit;
1857 struct {
1858 union {
1859 __u32 relative_fd;
1860 __u32 relative_id;
1861 };
1862 __u64 expected_revision;
1863 } cgroup;
1864 };
1865 } link_create;
1866
1867 struct { /* struct used by BPF_LINK_UPDATE command */
1868 __u32 link_fd; /* link fd */
1869 union {
1870 /* new program fd to update link with */
1871 __u32 new_prog_fd;
1872 /* new struct_ops map fd to update link with */
1873 __u32 new_map_fd;
1874 };
1875 __u32 flags; /* extra flags */
1876 union {
1877 /* expected link's program fd; is specified only if
1878 * BPF_F_REPLACE flag is set in flags.
1879 */
1880 __u32 old_prog_fd;
1881 /* expected link's map fd; is specified only
1882 * if BPF_F_REPLACE flag is set.
1883 */
1884 __u32 old_map_fd;
1885 };
1886 } link_update;
1887
1888 struct {
1889 __u32 link_fd;
1890 } link_detach;
1891
1892 struct { /* struct used by BPF_ENABLE_STATS command */
1893 __u32 type;
1894 } enable_stats;
1895
1896 struct { /* struct used by BPF_ITER_CREATE command */
1897 __u32 link_fd;
1898 __u32 flags;
1899 } iter_create;
1900
1901 struct { /* struct used by BPF_PROG_BIND_MAP command */
1902 __u32 prog_fd;
1903 __u32 map_fd;
1904 __u32 flags; /* extra flags */
1905 } prog_bind_map;
1906
1907 struct { /* struct used by BPF_TOKEN_CREATE command */
1908 __u32 flags;
1909 __u32 bpffs_fd;
1910 } token_create;
1911
1912 struct {
1913 __aligned_u64 stream_buf;
1914 __u32 stream_buf_len;
1915 __u32 stream_id;
1916 __u32 prog_fd;
1917 } prog_stream_read;
1918
1919 struct {
1920 __u32 map_fd;
1921 __u32 prog_fd;
1922 __u32 flags;
1923 } prog_assoc_struct_ops;
1924
1925} __attribute__((aligned(8)));
1926
1927/* The description below is an attempt at providing documentation to eBPF
1928 * developers about the multiple available eBPF helper functions. It can be
1929 * parsed and used to produce a manual page. The workflow is the following,
1930 * and requires the rst2man utility:
1931 *
1932 * $ ./scripts/bpf_doc.py \
1933 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1934 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1935 * $ man /tmp/bpf-helpers.7
1936 *
1937 * Note that in order to produce this external documentation, some RST
1938 * formatting is used in the descriptions to get "bold" and "italics" in
1939 * manual pages. Also note that the few trailing white spaces are
1940 * intentional, removing them would break paragraphs for rst2man.
1941 *
1942 * Start of BPF helper function descriptions:
1943 *
1944 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1945 * Description
1946 * Perform a lookup in *map* for an entry associated to *key*.
1947 * Return
1948 * Map value associated to *key*, or **NULL** if no entry was
1949 * found.
1950 *
1951 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1952 * Description
1953 * Add or update the value of the entry associated to *key* in
1954 * *map* with *value*. *flags* is one of:
1955 *
1956 * **BPF_NOEXIST**
1957 * The entry for *key* must not exist in the map.
1958 * **BPF_EXIST**
1959 * The entry for *key* must already exist in the map.
1960 * **BPF_ANY**
1961 * No condition on the existence of the entry for *key*.
1962 *
1963 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1964 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1965 * elements always exist), the helper would return an error.
1966 * Return
1967 * 0 on success, or a negative error in case of failure.
1968 *
1969 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1970 * Description
1971 * Delete entry with *key* from *map*.
1972 * Return
1973 * 0 on success, or a negative error in case of failure.
1974 *
1975 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1976 * Description
1977 * For tracing programs, safely attempt to read *size* bytes from
1978 * kernel space address *unsafe_ptr* and store the data in *dst*.
1979 *
1980 * Generally, use **bpf_probe_read_user**\ () or
1981 * **bpf_probe_read_kernel**\ () instead.
1982 * Return
1983 * 0 on success, or a negative error in case of failure.
1984 *
1985 * u64 bpf_ktime_get_ns(void)
1986 * Description
1987 * Return the time elapsed since system boot, in nanoseconds.
1988 * Does not include time the system was suspended.
1989 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1990 * Return
1991 * Current *ktime*.
1992 *
1993 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1994 * Description
1995 * This helper is a "printk()-like" facility for debugging. It
1996 * prints a message defined by format *fmt* (of size *fmt_size*)
1997 * to file *\/sys/kernel/tracing/trace* from TraceFS, if
1998 * available. It can take up to three additional **u64**
1999 * arguments (as an eBPF helpers, the total number of arguments is
2000 * limited to five).
2001 *
2002 * Each time the helper is called, it appends a line to the trace.
2003 * Lines are discarded while *\/sys/kernel/tracing/trace* is
2004 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
2005 * The format of the trace is customizable, and the exact output
2006 * one will get depends on the options set in
2007 * *\/sys/kernel/tracing/trace_options* (see also the
2008 * *README* file under the same directory). However, it usually
2009 * defaults to something like:
2010 *
2011 * ::
2012 *
2013 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
2014 *
2015 * In the above:
2016 *
2017 * * ``telnet`` is the name of the current task.
2018 * * ``470`` is the PID of the current task.
2019 * * ``001`` is the CPU number on which the task is
2020 * running.
2021 * * In ``.N..``, each character refers to a set of
2022 * options (whether irqs are enabled, scheduling
2023 * options, whether hard/softirqs are running, level of
2024 * preempt_disabled respectively). **N** means that
2025 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
2026 * are set.
2027 * * ``419421.045894`` is a timestamp.
2028 * * ``0x00000001`` is a fake value used by BPF for the
2029 * instruction pointer register.
2030 * * ``<formatted msg>`` is the message formatted with
2031 * *fmt*.
2032 *
2033 * The conversion specifiers supported by *fmt* are similar, but
2034 * more limited than for printk(). They are **%d**, **%i**,
2035 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
2036 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
2037 * of field, padding with zeroes, etc.) is available, and the
2038 * helper will return **-EINVAL** (but print nothing) if it
2039 * encounters an unknown specifier.
2040 *
2041 * Also, note that **bpf_trace_printk**\ () is slow, and should
2042 * only be used for debugging purposes. For this reason, a notice
2043 * block (spanning several lines) is printed to kernel logs and
2044 * states that the helper should not be used "for production use"
2045 * the first time this helper is used (or more precisely, when
2046 * **trace_printk**\ () buffers are allocated). For passing values
2047 * to user space, perf events should be preferred.
2048 * Return
2049 * The number of bytes written to the buffer, or a negative error
2050 * in case of failure.
2051 *
2052 * u32 bpf_get_prandom_u32(void)
2053 * Description
2054 * Get a pseudo-random number.
2055 *
2056 * From a security point of view, this helper uses its own
2057 * pseudo-random internal state, and cannot be used to infer the
2058 * seed of other random functions in the kernel. However, it is
2059 * essential to note that the generator used by the helper is not
2060 * cryptographically secure.
2061 * Return
2062 * A random 32-bit unsigned value.
2063 *
2064 * u32 bpf_get_smp_processor_id(void)
2065 * Description
2066 * Get the SMP (symmetric multiprocessing) processor id. Note that
2067 * all programs run with migration disabled, which means that the
2068 * SMP processor id is stable during all the execution of the
2069 * program.
2070 * Return
2071 * The SMP id of the processor running the program.
2072 * Attributes
2073 * __bpf_fastcall
2074 *
2075 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
2076 * Description
2077 * Store *len* bytes from address *from* into the packet
2078 * associated to *skb*, at *offset*. The *flags* are a combination
2079 * of the following values:
2080 *
2081 * **BPF_F_RECOMPUTE_CSUM**
2082 * Automatically update *skb*\ **->csum** after storing the
2083 * bytes.
2084 * **BPF_F_INVALIDATE_HASH**
2085 * Set *skb*\ **->hash**, *skb*\ **->swhash** and *skb*\
2086 * **->l4hash** to 0.
2087 *
2088 * A call to this helper is susceptible to change the underlying
2089 * packet buffer. Therefore, at load time, all checks on pointers
2090 * previously done by the verifier are invalidated and must be
2091 * performed again, if the helper is used in combination with
2092 * direct packet access.
2093 * Return
2094 * 0 on success, or a negative error in case of failure.
2095 *
2096 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
2097 * Description
2098 * Recompute the layer 3 (e.g. IP) checksum for the packet
2099 * associated to *skb*. Computation is incremental, so the helper
2100 * must know the former value of the header field that was
2101 * modified (*from*), the new value of this field (*to*), and the
2102 * number of bytes (2 or 4) for this field, stored in *size*.
2103 * Alternatively, it is possible to store the difference between
2104 * the previous and the new values of the header field in *to*, by
2105 * setting *from* and *size* to 0. For both methods, *offset*
2106 * indicates the location of the IP checksum within the packet.
2107 *
2108 * This helper works in combination with **bpf_csum_diff**\ (),
2109 * which does not update the checksum in-place, but offers more
2110 * flexibility and can handle sizes larger than 2 or 4 for the
2111 * checksum to update.
2112 *
2113 * A call to this helper is susceptible to change the underlying
2114 * packet buffer. Therefore, at load time, all checks on pointers
2115 * previously done by the verifier are invalidated and must be
2116 * performed again, if the helper is used in combination with
2117 * direct packet access.
2118 * Return
2119 * 0 on success, or a negative error in case of failure.
2120 *
2121 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
2122 * Description
2123 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
2124 * packet associated to *skb*. Computation is incremental, so the
2125 * helper must know the former value of the header field that was
2126 * modified (*from*), the new value of this field (*to*), and the
2127 * number of bytes (2 or 4) for this field, stored on the lowest
2128 * four bits of *flags*. Alternatively, it is possible to store
2129 * the difference between the previous and the new values of the
2130 * header field in *to*, by setting *from* and the four lowest
2131 * bits of *flags* to 0. For both methods, *offset* indicates the
2132 * location of the IP checksum within the packet. In addition to
2133 * the size of the field, *flags* can be added (bitwise OR) actual
2134 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
2135 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
2136 * for updates resulting in a null checksum the value is set to
2137 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
2138 * that the modified header field is part of the pseudo-header.
2139 * Flag **BPF_F_IPV6** should be set for IPv6 packets.
2140 *
2141 * This helper works in combination with **bpf_csum_diff**\ (),
2142 * which does not update the checksum in-place, but offers more
2143 * flexibility and can handle sizes larger than 2 or 4 for the
2144 * checksum to update.
2145 *
2146 * A call to this helper is susceptible to change the underlying
2147 * packet buffer. Therefore, at load time, all checks on pointers
2148 * previously done by the verifier are invalidated and must be
2149 * performed again, if the helper is used in combination with
2150 * direct packet access.
2151 * Return
2152 * 0 on success, or a negative error in case of failure.
2153 *
2154 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
2155 * Description
2156 * This special helper is used to trigger a "tail call", or in
2157 * other words, to jump into another eBPF program. The same stack
2158 * frame is used (but values on stack and in registers for the
2159 * caller are not accessible to the callee). This mechanism allows
2160 * for program chaining, either for raising the maximum number of
2161 * available eBPF instructions, or to execute given programs in
2162 * conditional blocks. For security reasons, there is an upper
2163 * limit to the number of successive tail calls that can be
2164 * performed.
2165 *
2166 * Upon call of this helper, the program attempts to jump into a
2167 * program referenced at index *index* in *prog_array_map*, a
2168 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
2169 * *ctx*, a pointer to the context.
2170 *
2171 * If the call succeeds, the kernel immediately runs the first
2172 * instruction of the new program. This is not a function call,
2173 * and it never returns to the previous program. If the call
2174 * fails, then the helper has no effect, and the caller continues
2175 * to run its subsequent instructions. A call can fail if the
2176 * destination program for the jump does not exist (i.e. *index*
2177 * is superior to the number of entries in *prog_array_map*), or
2178 * if the maximum number of tail calls has been reached for this
2179 * chain of programs. This limit is defined in the kernel by the
2180 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
2181 * which is currently set to 33.
2182 * Return
2183 * 0 on success, or a negative error in case of failure.
2184 *
2185 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
2186 * Description
2187 * Clone and redirect the packet associated to *skb* to another
2188 * net device of index *ifindex*. Both ingress and egress
2189 * interfaces can be used for redirection. The **BPF_F_INGRESS**
2190 * value in *flags* is used to make the distinction (ingress path
2191 * is selected if the flag is present, egress path otherwise).
2192 * This is the only flag supported for now.
2193 *
2194 * In comparison with **bpf_redirect**\ () helper,
2195 * **bpf_clone_redirect**\ () has the associated cost of
2196 * duplicating the packet buffer, but this can be executed out of
2197 * the eBPF program. Conversely, **bpf_redirect**\ () is more
2198 * efficient, but it is handled through an action code where the
2199 * redirection happens only after the eBPF program has returned.
2200 *
2201 * A call to this helper is susceptible to change the underlying
2202 * packet buffer. Therefore, at load time, all checks on pointers
2203 * previously done by the verifier are invalidated and must be
2204 * performed again, if the helper is used in combination with
2205 * direct packet access.
2206 * Return
2207 * 0 on success, or a negative error in case of failure. Positive
2208 * error indicates a potential drop or congestion in the target
2209 * device. The particular positive error codes are not defined.
2210 *
2211 * u64 bpf_get_current_pid_tgid(void)
2212 * Description
2213 * Get the current pid and tgid.
2214 * Return
2215 * A 64-bit integer containing the current tgid and pid, and
2216 * created as such:
2217 * *current_task*\ **->tgid << 32 \|**
2218 * *current_task*\ **->pid**.
2219 *
2220 * u64 bpf_get_current_uid_gid(void)
2221 * Description
2222 * Get the current uid and gid.
2223 * Return
2224 * A 64-bit integer containing the current GID and UID, and
2225 * created as such: *current_gid* **<< 32 \|** *current_uid*.
2226 *
2227 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
2228 * Description
2229 * Copy the **comm** attribute of the current task into *buf* of
2230 * *size_of_buf*. The **comm** attribute contains the name of
2231 * the executable (excluding the path) for the current task. The
2232 * *size_of_buf* must be strictly positive. On success, the
2233 * helper makes sure that the *buf* is NUL-terminated. On failure,
2234 * it is filled with zeroes.
2235 * Return
2236 * 0 on success, or a negative error in case of failure.
2237 *
2238 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
2239 * Description
2240 * Retrieve the classid for the current task, i.e. for the net_cls
2241 * cgroup to which *skb* belongs.
2242 *
2243 * This helper can be used on TC egress path, but not on ingress.
2244 *
2245 * The net_cls cgroup provides an interface to tag network packets
2246 * based on a user-provided identifier for all traffic coming from
2247 * the tasks belonging to the related cgroup. See also the related
2248 * kernel documentation, available from the Linux sources in file
2249 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2250 *
2251 * The Linux kernel has two versions for cgroups: there are
2252 * cgroups v1 and cgroups v2. Both are available to users, who can
2253 * use a mixture of them, but note that the net_cls cgroup is for
2254 * cgroup v1 only. This makes it incompatible with BPF programs
2255 * run on cgroups, which is a cgroup-v2-only feature (a socket can
2256 * only hold data for one version of cgroups at a time).
2257 *
2258 * This helper is only available is the kernel was compiled with
2259 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2260 * "**y**" or to "**m**".
2261 * Return
2262 * The classid, or 0 for the default unconfigured classid.
2263 *
2264 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2265 * Description
2266 * Push a *vlan_tci* (VLAN tag control information) of protocol
2267 * *vlan_proto* to the packet associated to *skb*, then update
2268 * the checksum. Note that if *vlan_proto* is different from
2269 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2270 * be **ETH_P_8021Q**.
2271 *
2272 * A call to this helper is susceptible to change the underlying
2273 * packet buffer. Therefore, at load time, all checks on pointers
2274 * previously done by the verifier are invalidated and must be
2275 * performed again, if the helper is used in combination with
2276 * direct packet access.
2277 * Return
2278 * 0 on success, or a negative error in case of failure.
2279 *
2280 * long bpf_skb_vlan_pop(struct sk_buff *skb)
2281 * Description
2282 * Pop a VLAN header from the packet associated to *skb*.
2283 *
2284 * A call to this helper is susceptible to change the underlying
2285 * packet buffer. Therefore, at load time, all checks on pointers
2286 * previously done by the verifier are invalidated and must be
2287 * performed again, if the helper is used in combination with
2288 * direct packet access.
2289 * Return
2290 * 0 on success, or a negative error in case of failure.
2291 *
2292 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2293 * Description
2294 * Get tunnel metadata. This helper takes a pointer *key* to an
2295 * empty **struct bpf_tunnel_key** of **size**, that will be
2296 * filled with tunnel metadata for the packet associated to *skb*.
2297 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2298 * indicates that the tunnel is based on IPv6 protocol instead of
2299 * IPv4.
2300 *
2301 * The **struct bpf_tunnel_key** is an object that generalizes the
2302 * principal parameters used by various tunneling protocols into a
2303 * single struct. This way, it can be used to easily make a
2304 * decision based on the contents of the encapsulation header,
2305 * "summarized" in this struct. In particular, it holds the IP
2306 * address of the remote end (IPv4 or IPv6, depending on the case)
2307 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2308 * this struct exposes the *key*\ **->tunnel_id**, which is
2309 * generally mapped to a VNI (Virtual Network Identifier), making
2310 * it programmable together with the **bpf_skb_set_tunnel_key**\
2311 * () helper.
2312 *
2313 * Let's imagine that the following code is part of a program
2314 * attached to the TC ingress interface, on one end of a GRE
2315 * tunnel, and is supposed to filter out all messages coming from
2316 * remote ends with IPv4 address other than 10.0.0.1:
2317 *
2318 * ::
2319 *
2320 * int ret;
2321 * struct bpf_tunnel_key key = {};
2322 *
2323 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2324 * if (ret < 0)
2325 * return TC_ACT_SHOT; // drop packet
2326 *
2327 * if (key.remote_ipv4 != 0x0a000001)
2328 * return TC_ACT_SHOT; // drop packet
2329 *
2330 * return TC_ACT_OK; // accept packet
2331 *
2332 * This interface can also be used with all encapsulation devices
2333 * that can operate in "collect metadata" mode: instead of having
2334 * one network device per specific configuration, the "collect
2335 * metadata" mode only requires a single device where the
2336 * configuration can be extracted from this helper.
2337 *
2338 * This can be used together with various tunnels such as VXLan,
2339 * Geneve, GRE or IP in IP (IPIP).
2340 * Return
2341 * 0 on success, or a negative error in case of failure.
2342 *
2343 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2344 * Description
2345 * Populate tunnel metadata for packet associated to *skb.* The
2346 * tunnel metadata is set to the contents of *key*, of *size*. The
2347 * *flags* can be set to a combination of the following values:
2348 *
2349 * **BPF_F_TUNINFO_IPV6**
2350 * Indicate that the tunnel is based on IPv6 protocol
2351 * instead of IPv4.
2352 * **BPF_F_ZERO_CSUM_TX**
2353 * For IPv4 packets, add a flag to tunnel metadata
2354 * indicating that checksum computation should be skipped
2355 * and checksum set to zeroes.
2356 * **BPF_F_DONT_FRAGMENT**
2357 * Add a flag to tunnel metadata indicating that the
2358 * packet should not be fragmented.
2359 * **BPF_F_SEQ_NUMBER**
2360 * Add a flag to tunnel metadata indicating that a
2361 * sequence number should be added to tunnel header before
2362 * sending the packet. This flag was added for GRE
2363 * encapsulation, but might be used with other protocols
2364 * as well in the future.
2365 * **BPF_F_NO_TUNNEL_KEY**
2366 * Add a flag to tunnel metadata indicating that no tunnel
2367 * key should be set in the resulting tunnel header.
2368 *
2369 * Here is a typical usage on the transmit path:
2370 *
2371 * ::
2372 *
2373 * struct bpf_tunnel_key key;
2374 * populate key ...
2375 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2376 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2377 *
2378 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
2379 * helper for additional information.
2380 * Return
2381 * 0 on success, or a negative error in case of failure.
2382 *
2383 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2384 * Description
2385 * Read the value of a perf event counter. This helper relies on a
2386 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2387 * the perf event counter is selected when *map* is updated with
2388 * perf event file descriptors. The *map* is an array whose size
2389 * is the number of available CPUs, and each cell contains a value
2390 * relative to one CPU. The value to retrieve is indicated by
2391 * *flags*, that contains the index of the CPU to look up, masked
2392 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2393 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2394 * current CPU should be retrieved.
2395 *
2396 * Note that before Linux 4.13, only hardware perf event can be
2397 * retrieved.
2398 *
2399 * Also, be aware that the newer helper
2400 * **bpf_perf_event_read_value**\ () is recommended over
2401 * **bpf_perf_event_read**\ () in general. The latter has some ABI
2402 * quirks where error and counter value are used as a return code
2403 * (which is wrong to do since ranges may overlap). This issue is
2404 * fixed with **bpf_perf_event_read_value**\ (), which at the same
2405 * time provides more features over the **bpf_perf_event_read**\
2406 * () interface. Please refer to the description of
2407 * **bpf_perf_event_read_value**\ () for details.
2408 * Return
2409 * The value of the perf event counter read from the map, or a
2410 * negative error code in case of failure.
2411 *
2412 * long bpf_redirect(u32 ifindex, u64 flags)
2413 * Description
2414 * Redirect the packet to another net device of index *ifindex*.
2415 * This helper is somewhat similar to **bpf_clone_redirect**\
2416 * (), except that the packet is not cloned, which provides
2417 * increased performance.
2418 *
2419 * Except for XDP, both ingress and egress interfaces can be used
2420 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
2421 * to make the distinction (ingress path is selected if the flag
2422 * is present, egress path otherwise). Currently, XDP only
2423 * supports redirection to the egress interface, and accepts no
2424 * flag at all.
2425 *
2426 * The same effect can also be attained with the more generic
2427 * **bpf_redirect_map**\ (), which uses a BPF map to store the
2428 * redirect target instead of providing it directly to the helper.
2429 * Return
2430 * For XDP, the helper returns **XDP_REDIRECT** on success or
2431 * **XDP_ABORTED** on error. For other program types, the values
2432 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2433 * error.
2434 *
2435 * u32 bpf_get_route_realm(struct sk_buff *skb)
2436 * Description
2437 * Retrieve the realm or the route, that is to say the
2438 * **tclassid** field of the destination for the *skb*. The
2439 * identifier retrieved is a user-provided tag, similar to the
2440 * one used with the net_cls cgroup (see description for
2441 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
2442 * held by a route (a destination entry), not by a task.
2443 *
2444 * Retrieving this identifier works with the clsact TC egress hook
2445 * (see also **tc-bpf(8)**), or alternatively on conventional
2446 * classful egress qdiscs, but not on TC ingress path. In case of
2447 * clsact TC egress hook, this has the advantage that, internally,
2448 * the destination entry has not been dropped yet in the transmit
2449 * path. Therefore, the destination entry does not need to be
2450 * artificially held via **netif_keep_dst**\ () for a classful
2451 * qdisc until the *skb* is freed.
2452 *
2453 * This helper is available only if the kernel was compiled with
2454 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2455 * Return
2456 * The realm of the route for the packet associated to *skb*, or 0
2457 * if none was found.
2458 *
2459 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2460 * Description
2461 * Write raw *data* blob into a special BPF perf event held by
2462 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2463 * event must have the following attributes: **PERF_SAMPLE_RAW**
2464 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2465 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2466 *
2467 * The *flags* are used to indicate the index in *map* for which
2468 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2469 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2470 * to indicate that the index of the current CPU core should be
2471 * used.
2472 *
2473 * The value to write, of *size*, is passed through eBPF stack and
2474 * pointed by *data*.
2475 *
2476 * The context of the program *ctx* needs also be passed to the
2477 * helper.
2478 *
2479 * On user space, a program willing to read the values needs to
2480 * call **perf_event_open**\ () on the perf event (either for
2481 * one or for all CPUs) and to store the file descriptor into the
2482 * *map*. This must be done before the eBPF program can send data
2483 * into it. An example is available in file
2484 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2485 * tree (the eBPF program counterpart is in
2486 * *samples/bpf/trace_output.bpf.c*).
2487 *
2488 * **bpf_perf_event_output**\ () achieves better performance
2489 * than **bpf_trace_printk**\ () for sharing data with user
2490 * space, and is much better suitable for streaming data from eBPF
2491 * programs.
2492 *
2493 * Note that this helper is not restricted to tracing use cases
2494 * and can be used with programs attached to TC or XDP as well,
2495 * where it allows for passing data to user space listeners. Data
2496 * can be:
2497 *
2498 * * Only custom structs,
2499 * * Only the packet payload, or
2500 * * A combination of both.
2501 * Return
2502 * 0 on success, or a negative error in case of failure.
2503 *
2504 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2505 * Description
2506 * This helper was provided as an easy way to load data from a
2507 * packet. It can be used to load *len* bytes from *offset* from
2508 * the packet associated to *skb*, into the buffer pointed by
2509 * *to*.
2510 *
2511 * Since Linux 4.7, usage of this helper has mostly been replaced
2512 * by "direct packet access", enabling packet data to be
2513 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2514 * pointing respectively to the first byte of packet data and to
2515 * the byte after the last byte of packet data. However, it
2516 * remains useful if one wishes to read large quantities of data
2517 * at once from a packet into the eBPF stack.
2518 * Return
2519 * 0 on success, or a negative error in case of failure.
2520 *
2521 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2522 * Description
2523 * Walk a user or a kernel stack and return its id. To achieve
2524 * this, the helper needs *ctx*, which is a pointer to the context
2525 * on which the tracing program is executed, and a pointer to a
2526 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2527 *
2528 * The last argument, *flags*, holds the number of stack frames to
2529 * skip (from 0 to 255), masked with
2530 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2531 * a combination of the following flags:
2532 *
2533 * **BPF_F_USER_STACK**
2534 * Collect a user space stack instead of a kernel stack.
2535 * **BPF_F_FAST_STACK_CMP**
2536 * Compare stacks by hash only.
2537 * **BPF_F_REUSE_STACKID**
2538 * If two different stacks hash into the same *stackid*,
2539 * discard the old one.
2540 *
2541 * The stack id retrieved is a 32 bit long integer handle which
2542 * can be further combined with other data (including other stack
2543 * ids) and used as a key into maps. This can be useful for
2544 * generating a variety of graphs (such as flame graphs or off-cpu
2545 * graphs).
2546 *
2547 * For walking a stack, this helper is an improvement over
2548 * **bpf_probe_read**\ (), which can be used with unrolled loops
2549 * but is not efficient and consumes a lot of eBPF instructions.
2550 * Instead, **bpf_get_stackid**\ () can collect up to
2551 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2552 * this limit can be controlled with the **sysctl** program, and
2553 * that it should be manually increased in order to profile long
2554 * user stacks (such as stacks for Java programs). To do so, use:
2555 *
2556 * ::
2557 *
2558 * # sysctl kernel.perf_event_max_stack=<new value>
2559 * Return
2560 * The positive or null stack id on success, or a negative error
2561 * in case of failure.
2562 *
2563 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2564 * Description
2565 * Compute a checksum difference, from the raw buffer pointed by
2566 * *from*, of length *from_size* (that must be a multiple of 4),
2567 * towards the raw buffer pointed by *to*, of size *to_size*
2568 * (same remark). An optional *seed* can be added to the value
2569 * (this can be cascaded, the seed may come from a previous call
2570 * to the helper).
2571 *
2572 * This is flexible enough to be used in several ways:
2573 *
2574 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2575 * checksum, it can be used when pushing new data.
2576 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2577 * checksum, it can be used when removing data from a packet.
2578 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2579 * can be used to compute a diff. Note that *from_size* and
2580 * *to_size* do not need to be equal.
2581 *
2582 * This helper can be used in combination with
2583 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2584 * which one can feed in the difference computed with
2585 * **bpf_csum_diff**\ ().
2586 * Return
2587 * The checksum result, or a negative error code in case of
2588 * failure.
2589 *
2590 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2591 * Description
2592 * Retrieve tunnel options metadata for the packet associated to
2593 * *skb*, and store the raw tunnel option data to the buffer *opt*
2594 * of *size*.
2595 *
2596 * This helper can be used with encapsulation devices that can
2597 * operate in "collect metadata" mode (please refer to the related
2598 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2599 * more details). A particular example where this can be used is
2600 * in combination with the Geneve encapsulation protocol, where it
2601 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2602 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2603 * the eBPF program. This allows for full customization of these
2604 * headers.
2605 * Return
2606 * The size of the option data retrieved.
2607 *
2608 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2609 * Description
2610 * Set tunnel options metadata for the packet associated to *skb*
2611 * to the option data contained in the raw buffer *opt* of *size*.
2612 *
2613 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2614 * helper for additional information.
2615 * Return
2616 * 0 on success, or a negative error in case of failure.
2617 *
2618 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2619 * Description
2620 * Change the protocol of the *skb* to *proto*. Currently
2621 * supported are transition from IPv4 to IPv6, and from IPv6 to
2622 * IPv4. The helper takes care of the groundwork for the
2623 * transition, including resizing the socket buffer. The eBPF
2624 * program is expected to fill the new headers, if any, via
2625 * **skb_store_bytes**\ () and to recompute the checksums with
2626 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2627 * (). The main case for this helper is to perform NAT64
2628 * operations out of an eBPF program.
2629 *
2630 * Internally, the GSO type is marked as dodgy so that headers are
2631 * checked and segments are recalculated by the GSO/GRO engine.
2632 * The size for GSO target is adapted as well.
2633 *
2634 * All values for *flags* are reserved for future usage, and must
2635 * be left at zero.
2636 *
2637 * A call to this helper is susceptible to change the underlying
2638 * packet buffer. Therefore, at load time, all checks on pointers
2639 * previously done by the verifier are invalidated and must be
2640 * performed again, if the helper is used in combination with
2641 * direct packet access.
2642 * Return
2643 * 0 on success, or a negative error in case of failure.
2644 *
2645 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2646 * Description
2647 * Change the packet type for the packet associated to *skb*. This
2648 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2649 * the eBPF program does not have a write access to *skb*\
2650 * **->pkt_type** beside this helper. Using a helper here allows
2651 * for graceful handling of errors.
2652 *
2653 * The major use case is to change incoming *skb*s to
2654 * **PACKET_HOST** in a programmatic way instead of having to
2655 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2656 * example.
2657 *
2658 * Note that *type* only allows certain values. At this time, they
2659 * are:
2660 *
2661 * **PACKET_HOST**
2662 * Packet is for us.
2663 * **PACKET_BROADCAST**
2664 * Send packet to all.
2665 * **PACKET_MULTICAST**
2666 * Send packet to group.
2667 * **PACKET_OTHERHOST**
2668 * Send packet to someone else.
2669 * Return
2670 * 0 on success, or a negative error in case of failure.
2671 *
2672 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2673 * Description
2674 * Check whether *skb* is a descendant of the cgroup2 held by
2675 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2676 * Return
2677 * The return value depends on the result of the test, and can be:
2678 *
2679 * * 0, if the *skb* failed the cgroup2 descendant test.
2680 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2681 * * A negative error code, if an error occurred.
2682 *
2683 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2684 * Description
2685 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2686 * not set, in particular if the hash was cleared due to mangling,
2687 * recompute this hash. Later accesses to the hash can be done
2688 * directly with *skb*\ **->hash**.
2689 *
2690 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2691 * prototype with **bpf_skb_change_proto**\ (), or calling
2692 * **bpf_skb_store_bytes**\ () with the
2693 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2694 * the hash and to trigger a new computation for the next call to
2695 * **bpf_get_hash_recalc**\ ().
2696 * Return
2697 * The 32-bit hash.
2698 *
2699 * u64 bpf_get_current_task(void)
2700 * Description
2701 * Get the current task.
2702 * Return
2703 * A pointer to the current task struct.
2704 *
2705 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2706 * Description
2707 * Attempt in a safe way to write *len* bytes from the buffer
2708 * *src* to *dst* in memory. It only works for threads that are in
2709 * user context, and *dst* must be a valid user space address.
2710 *
2711 * This helper should not be used to implement any kind of
2712 * security mechanism because of TOC-TOU attacks, but rather to
2713 * debug, divert, and manipulate execution of semi-cooperative
2714 * processes.
2715 *
2716 * Keep in mind that this feature is meant for experiments, and it
2717 * has a risk of crashing the system and running programs.
2718 * Therefore, when an eBPF program using this helper is attached,
2719 * a warning including PID and process name is printed to kernel
2720 * logs.
2721 * Return
2722 * 0 on success, or a negative error in case of failure.
2723 *
2724 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2725 * Description
2726 * Check whether the probe is being run is the context of a given
2727 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2728 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2729 * Return
2730 * The return value depends on the result of the test, and can be:
2731 *
2732 * * 1, if current task belongs to the cgroup2.
2733 * * 0, if current task does not belong to the cgroup2.
2734 * * A negative error code, if an error occurred.
2735 *
2736 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2737 * Description
2738 * Resize (trim or grow) the packet associated to *skb* to the
2739 * new *len*. The *flags* are reserved for future usage, and must
2740 * be left at zero.
2741 *
2742 * The basic idea is that the helper performs the needed work to
2743 * change the size of the packet, then the eBPF program rewrites
2744 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2745 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2746 * and others. This helper is a slow path utility intended for
2747 * replies with control messages. And because it is targeted for
2748 * slow path, the helper itself can afford to be slow: it
2749 * implicitly linearizes, unclones and drops offloads from the
2750 * *skb*.
2751 *
2752 * A call to this helper is susceptible to change the underlying
2753 * packet buffer. Therefore, at load time, all checks on pointers
2754 * previously done by the verifier are invalidated and must be
2755 * performed again, if the helper is used in combination with
2756 * direct packet access.
2757 * Return
2758 * 0 on success, or a negative error in case of failure.
2759 *
2760 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2761 * Description
2762 * Pull in non-linear data in case the *skb* is non-linear and not
2763 * all of *len* are part of the linear section. Make *len* bytes
2764 * from *skb* readable and writable. If a zero value is passed for
2765 * *len*, then all bytes in the linear part of *skb* will be made
2766 * readable and writable.
2767 *
2768 * This helper is only needed for reading and writing with direct
2769 * packet access.
2770 *
2771 * For direct packet access, testing that offsets to access
2772 * are within packet boundaries (test on *skb*\ **->data_end**) is
2773 * susceptible to fail if offsets are invalid, or if the requested
2774 * data is in non-linear parts of the *skb*. On failure the
2775 * program can just bail out, or in the case of a non-linear
2776 * buffer, use a helper to make the data available. The
2777 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2778 * the data. Another one consists in using **bpf_skb_pull_data**
2779 * to pull in once the non-linear parts, then retesting and
2780 * eventually access the data.
2781 *
2782 * At the same time, this also makes sure the *skb* is uncloned,
2783 * which is a necessary condition for direct write. As this needs
2784 * to be an invariant for the write part only, the verifier
2785 * detects writes and adds a prologue that is calling
2786 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2787 * the very beginning in case it is indeed cloned.
2788 *
2789 * A call to this helper is susceptible to change the underlying
2790 * packet buffer. Therefore, at load time, all checks on pointers
2791 * previously done by the verifier are invalidated and must be
2792 * performed again, if the helper is used in combination with
2793 * direct packet access.
2794 * Return
2795 * 0 on success, or a negative error in case of failure.
2796 *
2797 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2798 * Description
2799 * Add the checksum *csum* into *skb*\ **->csum** in case the
2800 * driver has supplied a checksum for the entire packet into that
2801 * field. Return an error otherwise. This helper is intended to be
2802 * used in combination with **bpf_csum_diff**\ (), in particular
2803 * when the checksum needs to be updated after data has been
2804 * written into the packet through direct packet access.
2805 * Return
2806 * The checksum on success, or a negative error code in case of
2807 * failure.
2808 *
2809 * void bpf_set_hash_invalid(struct sk_buff *skb)
2810 * Description
2811 * Invalidate the current *skb*\ **->hash**. It can be used after
2812 * mangling on headers through direct packet access, in order to
2813 * indicate that the hash is outdated and to trigger a
2814 * recalculation the next time the kernel tries to access this
2815 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2816 * Return
2817 * void.
2818 *
2819 * long bpf_get_numa_node_id(void)
2820 * Description
2821 * Return the id of the current NUMA node. The primary use case
2822 * for this helper is the selection of sockets for the local NUMA
2823 * node, when the program is attached to sockets using the
2824 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2825 * but the helper is also available to other eBPF program types,
2826 * similarly to **bpf_get_smp_processor_id**\ ().
2827 * Return
2828 * The id of current NUMA node.
2829 *
2830 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2831 * Description
2832 * Grows headroom of packet associated to *skb* and adjusts the
2833 * offset of the MAC header accordingly, adding *len* bytes of
2834 * space. It automatically extends and reallocates memory as
2835 * required.
2836 *
2837 * This helper can be used on a layer 3 *skb* to push a MAC header
2838 * for redirection into a layer 2 device.
2839 *
2840 * All values for *flags* are reserved for future usage, and must
2841 * be left at zero.
2842 *
2843 * A call to this helper is susceptible to change the underlying
2844 * packet buffer. Therefore, at load time, all checks on pointers
2845 * previously done by the verifier are invalidated and must be
2846 * performed again, if the helper is used in combination with
2847 * direct packet access.
2848 * Return
2849 * 0 on success, or a negative error in case of failure.
2850 *
2851 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2852 * Description
2853 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2854 * it is possible to use a negative value for *delta*. This helper
2855 * can be used to prepare the packet for pushing or popping
2856 * headers.
2857 *
2858 * A call to this helper is susceptible to change the underlying
2859 * packet buffer. Therefore, at load time, all checks on pointers
2860 * previously done by the verifier are invalidated and must be
2861 * performed again, if the helper is used in combination with
2862 * direct packet access.
2863 * Return
2864 * 0 on success, or a negative error in case of failure.
2865 *
2866 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2867 * Description
2868 * Copy a NUL terminated string from an unsafe kernel address
2869 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2870 * more details.
2871 *
2872 * Generally, use **bpf_probe_read_user_str**\ () or
2873 * **bpf_probe_read_kernel_str**\ () instead.
2874 * Return
2875 * On success, the strictly positive length of the string,
2876 * including the trailing NUL character. On error, a negative
2877 * value.
2878 *
2879 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2880 * Description
2881 * If the **struct sk_buff** pointed by *skb* has a known socket,
2882 * retrieve the cookie (generated by the kernel) of this socket.
2883 * If no cookie has been set yet, generate a new cookie. Once
2884 * generated, the socket cookie remains stable for the life of the
2885 * socket. This helper can be useful for monitoring per socket
2886 * networking traffic statistics as it provides a global socket
2887 * identifier that can be assumed unique.
2888 * Return
2889 * A 8-byte long unique number on success, or 0 if the socket
2890 * field is missing inside *skb*.
2891 *
2892 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2893 * Description
2894 * Equivalent to bpf_get_socket_cookie() helper that accepts
2895 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2896 * Return
2897 * A 8-byte long unique number.
2898 *
2899 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2900 * Description
2901 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2902 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2903 * Return
2904 * A 8-byte long unique number.
2905 *
2906 * u64 bpf_get_socket_cookie(struct sock *sk)
2907 * Description
2908 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2909 * *sk*, but gets socket from a BTF **struct sock**. This helper
2910 * also works for sleepable programs.
2911 * Return
2912 * A 8-byte long unique number or 0 if *sk* is NULL.
2913 *
2914 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2915 * Description
2916 * Get the owner UID of the socked associated to *skb*.
2917 * Return
2918 * The owner UID of the socket associated to *skb*. If the socket
2919 * is **NULL**, or if it is not a full socket (i.e. if it is a
2920 * time-wait or a request socket instead), **overflowuid** value
2921 * is returned (note that **overflowuid** might also be the actual
2922 * UID value for the socket).
2923 *
2924 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2925 * Description
2926 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2927 * to value *hash*.
2928 * Return
2929 * 0
2930 *
2931 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2932 * Description
2933 * Emulate a call to **setsockopt()** on the socket associated to
2934 * *bpf_socket*, which must be a full socket. The *level* at
2935 * which the option resides and the name *optname* of the option
2936 * must be specified, see **setsockopt(2)** for more information.
2937 * The option value of length *optlen* is pointed by *optval*.
2938 *
2939 * *bpf_socket* should be one of the following:
2940 *
2941 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2942 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
2943 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
2944 *
2945 * This helper actually implements a subset of **setsockopt()**.
2946 * It supports the following *level*\ s:
2947 *
2948 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2949 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2950 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2951 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2952 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2953 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2954 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2955 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2956 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2957 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2958 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2959 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2960 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.
2961 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2962 * * **IPPROTO_IPV6**, which supports the following *optname*\ s:
2963 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2964 * Return
2965 * 0 on success, or a negative error in case of failure.
2966 *
2967 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2968 * Description
2969 * Grow or shrink the room for data in the packet associated to
2970 * *skb* by *len_diff*, and according to the selected *mode*.
2971 *
2972 * By default, the helper will reset any offloaded checksum
2973 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2974 * by the following flag:
2975 *
2976 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2977 * checksum data of the skb to CHECKSUM_NONE.
2978 *
2979 * There are two supported modes at this time:
2980 *
2981 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2982 * (room space is added or removed between the layer 2 and
2983 * layer 3 headers).
2984 *
2985 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2986 * (room space is added or removed between the layer 3 and
2987 * layer 4 headers).
2988 *
2989 * The following flags are supported at this time:
2990 *
2991 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2992 * Adjusting mss in this way is not allowed for datagrams.
2993 *
2994 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2995 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2996 * Any new space is reserved to hold a tunnel header.
2997 * Configure skb offsets and other fields accordingly.
2998 *
2999 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
3000 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
3001 * Use with ENCAP_L3 flags to further specify the tunnel type.
3002 *
3003 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
3004 * Use with ENCAP_L3/L4 flags to further specify the tunnel
3005 * type; *len* is the length of the inner MAC header.
3006 *
3007 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
3008 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
3009 * L2 type as Ethernet.
3010 *
3011 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
3012 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
3013 * Indicate the new IP header version after decapsulating the outer
3014 * IP header. Used when the inner and outer IP versions are different.
3015 *
3016 * A call to this helper is susceptible to change the underlying
3017 * packet buffer. Therefore, at load time, all checks on pointers
3018 * previously done by the verifier are invalidated and must be
3019 * performed again, if the helper is used in combination with
3020 * direct packet access.
3021 * Return
3022 * 0 on success, or a negative error in case of failure.
3023 *
3024 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
3025 * Description
3026 * Redirect the packet to the endpoint referenced by *map* at
3027 * index *key*. Depending on its type, this *map* can contain
3028 * references to net devices (for forwarding packets through other
3029 * ports), or to CPUs (for redirecting XDP frames to another CPU;
3030 * but this is only implemented for native XDP (with driver
3031 * support) as of this writing).
3032 *
3033 * The lower two bits of *flags* are used as the return code if
3034 * the map lookup fails. This is so that the return value can be
3035 * one of the XDP program return codes up to **XDP_TX**, as chosen
3036 * by the caller. The higher bits of *flags* can be set to
3037 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
3038 *
3039 * With BPF_F_BROADCAST the packet will be broadcasted to all the
3040 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
3041 * interface will be excluded when do broadcasting.
3042 *
3043 * See also **bpf_redirect**\ (), which only supports redirecting
3044 * to an ifindex, but doesn't require a map to do so.
3045 * Return
3046 * **XDP_REDIRECT** on success, or the value of the two lower bits
3047 * of the *flags* argument on error.
3048 *
3049 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
3050 * Description
3051 * Redirect the packet to the socket referenced by *map* (of type
3052 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3053 * egress interfaces can be used for redirection. The
3054 * **BPF_F_INGRESS** value in *flags* is used to make the
3055 * distinction (ingress path is selected if the flag is present,
3056 * egress path otherwise). This is the only flag supported for now.
3057 * Return
3058 * **SK_PASS** on success, or **SK_DROP** on error.
3059 *
3060 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3061 * Description
3062 * Add an entry to, or update a *map* referencing sockets. The
3063 * *skops* is used as a new value for the entry associated to
3064 * *key*. *flags* is one of:
3065 *
3066 * **BPF_NOEXIST**
3067 * The entry for *key* must not exist in the map.
3068 * **BPF_EXIST**
3069 * The entry for *key* must already exist in the map.
3070 * **BPF_ANY**
3071 * No condition on the existence of the entry for *key*.
3072 *
3073 * If the *map* has eBPF programs (parser and verdict), those will
3074 * be inherited by the socket being added. If the socket is
3075 * already attached to eBPF programs, this results in an error.
3076 * Return
3077 * 0 on success, or a negative error in case of failure.
3078 *
3079 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
3080 * Description
3081 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
3082 * *delta* (which can be positive or negative). Note that this
3083 * operation modifies the address stored in *xdp_md*\ **->data**,
3084 * so the latter must be loaded only after the helper has been
3085 * called.
3086 *
3087 * The use of *xdp_md*\ **->data_meta** is optional and programs
3088 * are not required to use it. The rationale is that when the
3089 * packet is processed with XDP (e.g. as DoS filter), it is
3090 * possible to push further meta data along with it before passing
3091 * to the stack, and to give the guarantee that an ingress eBPF
3092 * program attached as a TC classifier on the same device can pick
3093 * this up for further post-processing. Since TC works with socket
3094 * buffers, it remains possible to set from XDP the **mark** or
3095 * **priority** pointers, or other pointers for the socket buffer.
3096 * Having this scratch space generic and programmable allows for
3097 * more flexibility as the user is free to store whatever meta
3098 * data they need.
3099 *
3100 * A call to this helper is susceptible to change the underlying
3101 * packet buffer. Therefore, at load time, all checks on pointers
3102 * previously done by the verifier are invalidated and must be
3103 * performed again, if the helper is used in combination with
3104 * direct packet access.
3105 * Return
3106 * 0 on success, or a negative error in case of failure.
3107 *
3108 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
3109 * Description
3110 * Read the value of a perf event counter, and store it into *buf*
3111 * of size *buf_size*. This helper relies on a *map* of type
3112 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
3113 * counter is selected when *map* is updated with perf event file
3114 * descriptors. The *map* is an array whose size is the number of
3115 * available CPUs, and each cell contains a value relative to one
3116 * CPU. The value to retrieve is indicated by *flags*, that
3117 * contains the index of the CPU to look up, masked with
3118 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
3119 * **BPF_F_CURRENT_CPU** to indicate that the value for the
3120 * current CPU should be retrieved.
3121 *
3122 * This helper behaves in a way close to
3123 * **bpf_perf_event_read**\ () helper, save that instead of
3124 * just returning the value observed, it fills the *buf*
3125 * structure. This allows for additional data to be retrieved: in
3126 * particular, the enabled and running times (in *buf*\
3127 * **->enabled** and *buf*\ **->running**, respectively) are
3128 * copied. In general, **bpf_perf_event_read_value**\ () is
3129 * recommended over **bpf_perf_event_read**\ (), which has some
3130 * ABI issues and provides fewer functionalities.
3131 *
3132 * These values are interesting, because hardware PMU (Performance
3133 * Monitoring Unit) counters are limited resources. When there are
3134 * more PMU based perf events opened than available counters,
3135 * kernel will multiplex these events so each event gets certain
3136 * percentage (but not all) of the PMU time. In case that
3137 * multiplexing happens, the number of samples or counter value
3138 * will not reflect the case compared to when no multiplexing
3139 * occurs. This makes comparison between different runs difficult.
3140 * Typically, the counter value should be normalized before
3141 * comparing to other experiments. The usual normalization is done
3142 * as follows.
3143 *
3144 * ::
3145 *
3146 * normalized_counter = counter * t_enabled / t_running
3147 *
3148 * Where t_enabled is the time enabled for event and t_running is
3149 * the time running for event since last normalization. The
3150 * enabled and running times are accumulated since the perf event
3151 * open. To achieve scaling factor between two invocations of an
3152 * eBPF program, users can use CPU id as the key (which is
3153 * typical for perf array usage model) to remember the previous
3154 * value and do the calculation inside the eBPF program.
3155 * Return
3156 * 0 on success, or a negative error in case of failure.
3157 *
3158 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
3159 * Description
3160 * For an eBPF program attached to a perf event, retrieve the
3161 * value of the event counter associated to *ctx* and store it in
3162 * the structure pointed by *buf* and of size *buf_size*. Enabled
3163 * and running times are also stored in the structure (see
3164 * description of helper **bpf_perf_event_read_value**\ () for
3165 * more details).
3166 * Return
3167 * 0 on success, or a negative error in case of failure.
3168 *
3169 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
3170 * Description
3171 * Emulate a call to **getsockopt()** on the socket associated to
3172 * *bpf_socket*, which must be a full socket. The *level* at
3173 * which the option resides and the name *optname* of the option
3174 * must be specified, see **getsockopt(2)** for more information.
3175 * The retrieved value is stored in the structure pointed by
3176 * *opval* and of length *optlen*.
3177 *
3178 * *bpf_socket* should be one of the following:
3179 *
3180 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
3181 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
3182 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
3183 *
3184 * This helper actually implements a subset of **getsockopt()**.
3185 * It supports the same set of *optname*\ s that is supported by
3186 * the **bpf_setsockopt**\ () helper. The exceptions are
3187 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and
3188 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
3189 * Return
3190 * 0 on success, or a negative error in case of failure.
3191 *
3192 * long bpf_override_return(struct pt_regs *regs, u64 rc)
3193 * Description
3194 * Used for error injection, this helper uses kprobes to override
3195 * the return value of the probed function, and to set it to *rc*.
3196 * The first argument is the context *regs* on which the kprobe
3197 * works.
3198 *
3199 * This helper works by setting the PC (program counter)
3200 * to an override function which is run in place of the original
3201 * probed function. This means the probed function is not run at
3202 * all. The replacement function just returns with the required
3203 * value.
3204 *
3205 * This helper has security implications, and thus is subject to
3206 * restrictions. It is only available if the kernel was compiled
3207 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
3208 * option, and in this case it only works on functions tagged with
3209 * **ALLOW_ERROR_INJECTION** in the kernel code.
3210 * Return
3211 * 0
3212 *
3213 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
3214 * Description
3215 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
3216 * for the full TCP socket associated to *bpf_sock_ops* to
3217 * *argval*.
3218 *
3219 * The primary use of this field is to determine if there should
3220 * be calls to eBPF programs of type
3221 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
3222 * code. A program of the same type can change its value, per
3223 * connection and as necessary, when the connection is
3224 * established. This field is directly accessible for reading, but
3225 * this helper must be used for updates in order to return an
3226 * error if an eBPF program tries to set a callback that is not
3227 * supported in the current kernel.
3228 *
3229 * *argval* is a flag array which can combine these flags:
3230 *
3231 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
3232 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
3233 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
3234 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
3235 *
3236 * Therefore, this function can be used to clear a callback flag by
3237 * setting the appropriate bit to zero. e.g. to disable the RTO
3238 * callback:
3239 *
3240 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
3241 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3242 *
3243 * Here are some examples of where one could call such eBPF
3244 * program:
3245 *
3246 * * When RTO fires.
3247 * * When a packet is retransmitted.
3248 * * When the connection terminates.
3249 * * When a packet is sent.
3250 * * When a packet is received.
3251 * Return
3252 * Code **-EINVAL** if the socket is not a full TCP socket;
3253 * otherwise, a positive number containing the bits that could not
3254 * be set is returned (which comes down to 0 if all bits were set
3255 * as required).
3256 *
3257 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3258 * Description
3259 * This helper is used in programs implementing policies at the
3260 * socket level. If the message *msg* is allowed to pass (i.e. if
3261 * the verdict eBPF program returns **SK_PASS**), redirect it to
3262 * the socket referenced by *map* (of type
3263 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3264 * egress interfaces can be used for redirection. The
3265 * **BPF_F_INGRESS** value in *flags* is used to make the
3266 * distinction (ingress path is selected if the flag is present,
3267 * egress path otherwise). This is the only flag supported for now.
3268 * Return
3269 * **SK_PASS** on success, or **SK_DROP** on error.
3270 *
3271 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3272 * Description
3273 * For socket policies, apply the verdict of the eBPF program to
3274 * the next *bytes* (number of bytes) of message *msg*.
3275 *
3276 * For example, this helper can be used in the following cases:
3277 *
3278 * * A single **sendmsg**\ () or **sendfile**\ () system call
3279 * contains multiple logical messages that the eBPF program is
3280 * supposed to read and for which it should apply a verdict.
3281 * * An eBPF program only cares to read the first *bytes* of a
3282 * *msg*. If the message has a large payload, then setting up
3283 * and calling the eBPF program repeatedly for all bytes, even
3284 * though the verdict is already known, would create unnecessary
3285 * overhead.
3286 *
3287 * When called from within an eBPF program, the helper sets a
3288 * counter internal to the BPF infrastructure, that is used to
3289 * apply the last verdict to the next *bytes*. If *bytes* is
3290 * smaller than the current data being processed from a
3291 * **sendmsg**\ () or **sendfile**\ () system call, the first
3292 * *bytes* will be sent and the eBPF program will be re-run with
3293 * the pointer for start of data pointing to byte number *bytes*
3294 * **+ 1**. If *bytes* is larger than the current data being
3295 * processed, then the eBPF verdict will be applied to multiple
3296 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3297 * consumed.
3298 *
3299 * Note that if a socket closes with the internal counter holding
3300 * a non-zero value, this is not a problem because data is not
3301 * being buffered for *bytes* and is sent as it is received.
3302 * Return
3303 * 0
3304 *
3305 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3306 * Description
3307 * For socket policies, prevent the execution of the verdict eBPF
3308 * program for message *msg* until *bytes* (byte number) have been
3309 * accumulated.
3310 *
3311 * This can be used when one needs a specific number of bytes
3312 * before a verdict can be assigned, even if the data spans
3313 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3314 * case would be a user calling **sendmsg**\ () repeatedly with
3315 * 1-byte long message segments. Obviously, this is bad for
3316 * performance, but it is still valid. If the eBPF program needs
3317 * *bytes* bytes to validate a header, this helper can be used to
3318 * prevent the eBPF program to be called again until *bytes* have
3319 * been accumulated.
3320 * Return
3321 * 0
3322 *
3323 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3324 * Description
3325 * For socket policies, pull in non-linear data from user space
3326 * for *msg* and set pointers *msg*\ **->data** and *msg*\
3327 * **->data_end** to *start* and *end* bytes offsets into *msg*,
3328 * respectively.
3329 *
3330 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3331 * *msg* it can only parse data that the (**data**, **data_end**)
3332 * pointers have already consumed. For **sendmsg**\ () hooks this
3333 * is likely the first scatterlist element. But for calls relying
3334 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3335 * be the range (**0**, **0**) because the data is shared with
3336 * user space and by default the objective is to avoid allowing
3337 * user space to modify data while (or after) eBPF verdict is
3338 * being decided. This helper can be used to pull in data and to
3339 * set the start and end pointer to given values. Data will be
3340 * copied if necessary (i.e. if data was not linear and if start
3341 * and end pointers do not point to the same chunk).
3342 *
3343 * A call to this helper is susceptible to change the underlying
3344 * packet buffer. Therefore, at load time, all checks on pointers
3345 * previously done by the verifier are invalidated and must be
3346 * performed again, if the helper is used in combination with
3347 * direct packet access.
3348 *
3349 * All values for *flags* are reserved for future usage, and must
3350 * be left at zero.
3351 * Return
3352 * 0 on success, or a negative error in case of failure.
3353 *
3354 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3355 * Description
3356 * Bind the socket associated to *ctx* to the address pointed by
3357 * *addr*, of length *addr_len*. This allows for making outgoing
3358 * connection from the desired IP address, which can be useful for
3359 * example when all processes inside a cgroup should use one
3360 * single IP address on a host that has multiple IP configured.
3361 *
3362 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3363 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3364 * **AF_INET6**). It's advised to pass zero port (**sin_port**
3365 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3366 * behavior and lets the kernel efficiently pick up an unused
3367 * port as long as 4-tuple is unique. Passing non-zero port might
3368 * lead to degraded performance.
3369 * Return
3370 * 0 on success, or a negative error in case of failure.
3371 *
3372 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3373 * Description
3374 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3375 * possible to both shrink and grow the packet tail.
3376 * Shrink done via *delta* being a negative integer.
3377 *
3378 * A call to this helper is susceptible to change the underlying
3379 * packet buffer. Therefore, at load time, all checks on pointers
3380 * previously done by the verifier are invalidated and must be
3381 * performed again, if the helper is used in combination with
3382 * direct packet access.
3383 * Return
3384 * 0 on success, or a negative error in case of failure.
3385 *
3386 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3387 * Description
3388 * Retrieve the XFRM state (IP transform framework, see also
3389 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3390 *
3391 * The retrieved value is stored in the **struct bpf_xfrm_state**
3392 * pointed by *xfrm_state* and of length *size*.
3393 *
3394 * All values for *flags* are reserved for future usage, and must
3395 * be left at zero.
3396 *
3397 * This helper is available only if the kernel was compiled with
3398 * **CONFIG_XFRM** configuration option.
3399 * Return
3400 * 0 on success, or a negative error in case of failure.
3401 *
3402 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3403 * Description
3404 * Return a user or a kernel stack in bpf program provided buffer.
3405 * To achieve this, the helper needs *ctx*, which is a pointer
3406 * to the context on which the tracing program is executed.
3407 * To store the stacktrace, the bpf program provides *buf* with
3408 * a nonnegative *size*.
3409 *
3410 * The last argument, *flags*, holds the number of stack frames to
3411 * skip (from 0 to 255), masked with
3412 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3413 * the following flags:
3414 *
3415 * **BPF_F_USER_STACK**
3416 * Collect a user space stack instead of a kernel stack.
3417 * **BPF_F_USER_BUILD_ID**
3418 * Collect (build_id, file_offset) instead of ips for user
3419 * stack, only valid if **BPF_F_USER_STACK** is also
3420 * specified.
3421 *
3422 * *file_offset* is an offset relative to the beginning
3423 * of the executable or shared object file backing the vma
3424 * which the *ip* falls in. It is *not* an offset relative
3425 * to that object's base address. Accordingly, it must be
3426 * adjusted by adding (sh_addr - sh_offset), where
3427 * sh_{addr,offset} correspond to the executable section
3428 * containing *file_offset* in the object, for comparisons
3429 * to symbols' st_value to be valid.
3430 *
3431 * **bpf_get_stack**\ () can collect up to
3432 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3433 * to sufficient large buffer size. Note that
3434 * this limit can be controlled with the **sysctl** program, and
3435 * that it should be manually increased in order to profile long
3436 * user stacks (such as stacks for Java programs). To do so, use:
3437 *
3438 * ::
3439 *
3440 * # sysctl kernel.perf_event_max_stack=<new value>
3441 * Return
3442 * The non-negative copied *buf* length equal to or less than
3443 * *size* on success, or a negative error in case of failure.
3444 *
3445 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3446 * Description
3447 * This helper is similar to **bpf_skb_load_bytes**\ () in that
3448 * it provides an easy way to load *len* bytes from *offset*
3449 * from the packet associated to *skb*, into the buffer pointed
3450 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3451 * a fifth argument *start_header* exists in order to select a
3452 * base offset to start from. *start_header* can be one of:
3453 *
3454 * **BPF_HDR_START_MAC**
3455 * Base offset to load data from is *skb*'s mac header.
3456 * **BPF_HDR_START_NET**
3457 * Base offset to load data from is *skb*'s network header.
3458 *
3459 * In general, "direct packet access" is the preferred method to
3460 * access packet data, however, this helper is in particular useful
3461 * in socket filters where *skb*\ **->data** does not always point
3462 * to the start of the mac header and where "direct packet access"
3463 * is not available.
3464 * Return
3465 * 0 on success, or a negative error in case of failure.
3466 *
3467 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3468 * Description
3469 * Do FIB lookup in kernel tables using parameters in *params*.
3470 * If lookup is successful and result shows packet is to be
3471 * forwarded, the neighbor tables are searched for the nexthop.
3472 * If successful (ie., FIB lookup shows forwarding and nexthop
3473 * is resolved), the nexthop address is returned in ipv4_dst
3474 * or ipv6_dst based on family, smac is set to mac address of
3475 * egress device, dmac is set to nexthop mac address, rt_metric
3476 * is set to metric from route (IPv4/IPv6 only), and ifindex
3477 * is set to the device index of the nexthop from the FIB lookup.
3478 *
3479 * *plen* argument is the size of the passed in struct.
3480 * *flags* argument can be a combination of one or more of the
3481 * following values:
3482 *
3483 * **BPF_FIB_LOOKUP_DIRECT**
3484 * Do a direct table lookup vs full lookup using FIB
3485 * rules.
3486 * **BPF_FIB_LOOKUP_TBID**
3487 * Used with BPF_FIB_LOOKUP_DIRECT.
3488 * Use the routing table ID present in *params*->tbid
3489 * for the fib lookup.
3490 * **BPF_FIB_LOOKUP_OUTPUT**
3491 * Perform lookup from an egress perspective (default is
3492 * ingress).
3493 * **BPF_FIB_LOOKUP_SKIP_NEIGH**
3494 * Skip the neighbour table lookup. *params*->dmac
3495 * and *params*->smac will not be set as output. A common
3496 * use case is to call **bpf_redirect_neigh**\ () after
3497 * doing **bpf_fib_lookup**\ ().
3498 * **BPF_FIB_LOOKUP_SRC**
3499 * Derive and set source IP addr in *params*->ipv{4,6}_src
3500 * for the nexthop. If the src addr cannot be derived,
3501 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3502 * case, *params*->dmac and *params*->smac are not set either.
3503 * **BPF_FIB_LOOKUP_MARK**
3504 * Use the mark present in *params*->mark for the fib lookup.
3505 * This option should not be used with BPF_FIB_LOOKUP_DIRECT,
3506 * as it only has meaning for full lookups.
3507 *
3508 * *ctx* is either **struct xdp_md** for XDP programs or
3509 * **struct sk_buff** tc cls_act programs.
3510 * Return
3511 * * < 0 if any input argument is invalid
3512 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3513 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3514 * packet is not forwarded or needs assist from full stack
3515 *
3516 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3517 * was exceeded and output params->mtu_result contains the MTU.
3518 *
3519 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3520 * Description
3521 * Add an entry to, or update a sockhash *map* referencing sockets.
3522 * The *skops* is used as a new value for the entry associated to
3523 * *key*. *flags* is one of:
3524 *
3525 * **BPF_NOEXIST**
3526 * The entry for *key* must not exist in the map.
3527 * **BPF_EXIST**
3528 * The entry for *key* must already exist in the map.
3529 * **BPF_ANY**
3530 * No condition on the existence of the entry for *key*.
3531 *
3532 * If the *map* has eBPF programs (parser and verdict), those will
3533 * be inherited by the socket being added. If the socket is
3534 * already attached to eBPF programs, this results in an error.
3535 * Return
3536 * 0 on success, or a negative error in case of failure.
3537 *
3538 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3539 * Description
3540 * This helper is used in programs implementing policies at the
3541 * socket level. If the message *msg* is allowed to pass (i.e. if
3542 * the verdict eBPF program returns **SK_PASS**), redirect it to
3543 * the socket referenced by *map* (of type
3544 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3545 * egress interfaces can be used for redirection. The
3546 * **BPF_F_INGRESS** value in *flags* is used to make the
3547 * distinction (ingress path is selected if the flag is present,
3548 * egress path otherwise). This is the only flag supported for now.
3549 * Return
3550 * **SK_PASS** on success, or **SK_DROP** on error.
3551 *
3552 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3553 * Description
3554 * This helper is used in programs implementing policies at the
3555 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3556 * if the verdict eBPF program returns **SK_PASS**), redirect it
3557 * to the socket referenced by *map* (of type
3558 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3559 * egress interfaces can be used for redirection. The
3560 * **BPF_F_INGRESS** value in *flags* is used to make the
3561 * distinction (ingress path is selected if the flag is present,
3562 * egress otherwise). This is the only flag supported for now.
3563 * Return
3564 * **SK_PASS** on success, or **SK_DROP** on error.
3565 *
3566 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3567 * Description
3568 * Encapsulate the packet associated to *skb* within a Layer 3
3569 * protocol header. This header is provided in the buffer at
3570 * address *hdr*, with *len* its size in bytes. *type* indicates
3571 * the protocol of the header and can be one of:
3572 *
3573 * **BPF_LWT_ENCAP_SEG6**
3574 * IPv6 encapsulation with Segment Routing Header
3575 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3576 * the IPv6 header is computed by the kernel.
3577 * **BPF_LWT_ENCAP_SEG6_INLINE**
3578 * Only works if *skb* contains an IPv6 packet. Insert a
3579 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3580 * the IPv6 header.
3581 * **BPF_LWT_ENCAP_IP**
3582 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3583 * must be IPv4 or IPv6, followed by zero or more
3584 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3585 * total bytes in all prepended headers. Please note that
3586 * if **skb_is_gso**\ (*skb*) is true, no more than two
3587 * headers can be prepended, and the inner header, if
3588 * present, should be either GRE or UDP/GUE.
3589 *
3590 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3591 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3592 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3593 * **BPF_PROG_TYPE_LWT_XMIT**.
3594 *
3595 * A call to this helper is susceptible to change the underlying
3596 * packet buffer. Therefore, at load time, all checks on pointers
3597 * previously done by the verifier are invalidated and must be
3598 * performed again, if the helper is used in combination with
3599 * direct packet access.
3600 * Return
3601 * 0 on success, or a negative error in case of failure.
3602 *
3603 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3604 * Description
3605 * Store *len* bytes from address *from* into the packet
3606 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3607 * inside the outermost IPv6 Segment Routing Header can be
3608 * modified through this helper.
3609 *
3610 * A call to this helper is susceptible to change the underlying
3611 * packet buffer. Therefore, at load time, all checks on pointers
3612 * previously done by the verifier are invalidated and must be
3613 * performed again, if the helper is used in combination with
3614 * direct packet access.
3615 * Return
3616 * 0 on success, or a negative error in case of failure.
3617 *
3618 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3619 * Description
3620 * Adjust the size allocated to TLVs in the outermost IPv6
3621 * Segment Routing Header contained in the packet associated to
3622 * *skb*, at position *offset* by *delta* bytes. Only offsets
3623 * after the segments are accepted. *delta* can be as well
3624 * positive (growing) as negative (shrinking).
3625 *
3626 * A call to this helper is susceptible to change the underlying
3627 * packet buffer. Therefore, at load time, all checks on pointers
3628 * previously done by the verifier are invalidated and must be
3629 * performed again, if the helper is used in combination with
3630 * direct packet access.
3631 * Return
3632 * 0 on success, or a negative error in case of failure.
3633 *
3634 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3635 * Description
3636 * Apply an IPv6 Segment Routing action of type *action* to the
3637 * packet associated to *skb*. Each action takes a parameter
3638 * contained at address *param*, and of length *param_len* bytes.
3639 * *action* can be one of:
3640 *
3641 * **SEG6_LOCAL_ACTION_END_X**
3642 * End.X action: Endpoint with Layer-3 cross-connect.
3643 * Type of *param*: **struct in6_addr**.
3644 * **SEG6_LOCAL_ACTION_END_T**
3645 * End.T action: Endpoint with specific IPv6 table lookup.
3646 * Type of *param*: **int**.
3647 * **SEG6_LOCAL_ACTION_END_B6**
3648 * End.B6 action: Endpoint bound to an SRv6 policy.
3649 * Type of *param*: **struct ipv6_sr_hdr**.
3650 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3651 * End.B6.Encap action: Endpoint bound to an SRv6
3652 * encapsulation policy.
3653 * Type of *param*: **struct ipv6_sr_hdr**.
3654 *
3655 * A call to this helper is susceptible to change the underlying
3656 * packet buffer. Therefore, at load time, all checks on pointers
3657 * previously done by the verifier are invalidated and must be
3658 * performed again, if the helper is used in combination with
3659 * direct packet access.
3660 * Return
3661 * 0 on success, or a negative error in case of failure.
3662 *
3663 * long bpf_rc_repeat(void *ctx)
3664 * Description
3665 * This helper is used in programs implementing IR decoding, to
3666 * report a successfully decoded repeat key message. This delays
3667 * the generation of a key up event for previously generated
3668 * key down event.
3669 *
3670 * Some IR protocols like NEC have a special IR message for
3671 * repeating last button, for when a button is held down.
3672 *
3673 * The *ctx* should point to the lirc sample as passed into
3674 * the program.
3675 *
3676 * This helper is only available is the kernel was compiled with
3677 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3678 * "**y**".
3679 * Return
3680 * 0
3681 *
3682 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3683 * Description
3684 * This helper is used in programs implementing IR decoding, to
3685 * report a successfully decoded key press with *scancode*,
3686 * *toggle* value in the given *protocol*. The scancode will be
3687 * translated to a keycode using the rc keymap, and reported as
3688 * an input key down event. After a period a key up event is
3689 * generated. This period can be extended by calling either
3690 * **bpf_rc_keydown**\ () again with the same values, or calling
3691 * **bpf_rc_repeat**\ ().
3692 *
3693 * Some protocols include a toggle bit, in case the button was
3694 * released and pressed again between consecutive scancodes.
3695 *
3696 * The *ctx* should point to the lirc sample as passed into
3697 * the program.
3698 *
3699 * The *protocol* is the decoded protocol number (see
3700 * **enum rc_proto** for some predefined values).
3701 *
3702 * This helper is only available is the kernel was compiled with
3703 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3704 * "**y**".
3705 * Return
3706 * 0
3707 *
3708 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3709 * Description
3710 * Return the cgroup v2 id of the socket associated with the *skb*.
3711 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3712 * helper for cgroup v1 by providing a tag resp. identifier that
3713 * can be matched on or used for map lookups e.g. to implement
3714 * policy. The cgroup v2 id of a given path in the hierarchy is
3715 * exposed in user space through the f_handle API in order to get
3716 * to the same 64-bit id.
3717 *
3718 * This helper can be used on TC egress path, but not on ingress,
3719 * and is available only if the kernel was compiled with the
3720 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3721 * Return
3722 * The id is returned or 0 in case the id could not be retrieved.
3723 *
3724 * u64 bpf_get_current_cgroup_id(void)
3725 * Description
3726 * Get the current cgroup id based on the cgroup within which
3727 * the current task is running.
3728 * Return
3729 * A 64-bit integer containing the current cgroup id based
3730 * on the cgroup within which the current task is running.
3731 *
3732 * void *bpf_get_local_storage(void *map, u64 flags)
3733 * Description
3734 * Get the pointer to the local storage area.
3735 * The type and the size of the local storage is defined
3736 * by the *map* argument.
3737 * The *flags* meaning is specific for each map type,
3738 * and has to be 0 for cgroup local storage.
3739 *
3740 * Depending on the BPF program type, a local storage area
3741 * can be shared between multiple instances of the BPF program,
3742 * running simultaneously.
3743 *
3744 * A user should care about the synchronization by himself.
3745 * For example, by using the **BPF_ATOMIC** instructions to alter
3746 * the shared data.
3747 * Return
3748 * A pointer to the local storage area.
3749 *
3750 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3751 * Description
3752 * Select a **SO_REUSEPORT** socket from a
3753 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3754 * It checks the selected socket is matching the incoming
3755 * request in the socket buffer.
3756 * Return
3757 * 0 on success, or a negative error in case of failure.
3758 *
3759 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3760 * Description
3761 * Return id of cgroup v2 that is ancestor of cgroup associated
3762 * with the *skb* at the *ancestor_level*. The root cgroup is at
3763 * *ancestor_level* zero and each step down the hierarchy
3764 * increments the level. If *ancestor_level* == level of cgroup
3765 * associated with *skb*, then return value will be same as that
3766 * of **bpf_skb_cgroup_id**\ ().
3767 *
3768 * The helper is useful to implement policies based on cgroups
3769 * that are upper in hierarchy than immediate cgroup associated
3770 * with *skb*.
3771 *
3772 * The format of returned id and helper limitations are same as in
3773 * **bpf_skb_cgroup_id**\ ().
3774 * Return
3775 * The id is returned or 0 in case the id could not be retrieved.
3776 *
3777 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3778 * Description
3779 * Look for TCP socket matching *tuple*, optionally in a child
3780 * network namespace *netns*. The return value must be checked,
3781 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3782 *
3783 * The *ctx* should point to the context of the program, such as
3784 * the skb or socket (depending on the hook in use). This is used
3785 * to determine the base network namespace for the lookup.
3786 *
3787 * *tuple_size* must be one of:
3788 *
3789 * **sizeof**\ (*tuple*\ **->ipv4**)
3790 * Look for an IPv4 socket.
3791 * **sizeof**\ (*tuple*\ **->ipv6**)
3792 * Look for an IPv6 socket.
3793 *
3794 * If the *netns* is a negative signed 32-bit integer, then the
3795 * socket lookup table in the netns associated with the *ctx*
3796 * will be used. For the TC hooks, this is the netns of the device
3797 * in the skb. For socket hooks, this is the netns of the socket.
3798 * If *netns* is any other signed 32-bit value greater than or
3799 * equal to zero then it specifies the ID of the netns relative to
3800 * the netns associated with the *ctx*. *netns* values beyond the
3801 * range of 32-bit integers are reserved for future use.
3802 *
3803 * All values for *flags* are reserved for future usage, and must
3804 * be left at zero.
3805 *
3806 * This helper is available only if the kernel was compiled with
3807 * **CONFIG_NET** configuration option.
3808 * Return
3809 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3810 * For sockets with reuseport option, the **struct bpf_sock**
3811 * result is from *reuse*\ **->socks**\ [] using the hash of the
3812 * tuple.
3813 *
3814 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3815 * Description
3816 * Look for UDP socket matching *tuple*, optionally in a child
3817 * network namespace *netns*. The return value must be checked,
3818 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3819 *
3820 * The *ctx* should point to the context of the program, such as
3821 * the skb or socket (depending on the hook in use). This is used
3822 * to determine the base network namespace for the lookup.
3823 *
3824 * *tuple_size* must be one of:
3825 *
3826 * **sizeof**\ (*tuple*\ **->ipv4**)
3827 * Look for an IPv4 socket.
3828 * **sizeof**\ (*tuple*\ **->ipv6**)
3829 * Look for an IPv6 socket.
3830 *
3831 * If the *netns* is a negative signed 32-bit integer, then the
3832 * socket lookup table in the netns associated with the *ctx*
3833 * will be used. For the TC hooks, this is the netns of the device
3834 * in the skb. For socket hooks, this is the netns of the socket.
3835 * If *netns* is any other signed 32-bit value greater than or
3836 * equal to zero then it specifies the ID of the netns relative to
3837 * the netns associated with the *ctx*. *netns* values beyond the
3838 * range of 32-bit integers are reserved for future use.
3839 *
3840 * All values for *flags* are reserved for future usage, and must
3841 * be left at zero.
3842 *
3843 * This helper is available only if the kernel was compiled with
3844 * **CONFIG_NET** configuration option.
3845 * Return
3846 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3847 * For sockets with reuseport option, the **struct bpf_sock**
3848 * result is from *reuse*\ **->socks**\ [] using the hash of the
3849 * tuple.
3850 *
3851 * long bpf_sk_release(void *sock)
3852 * Description
3853 * Release the reference held by *sock*. *sock* must be a
3854 * non-**NULL** pointer that was returned from
3855 * **bpf_sk_lookup_xxx**\ ().
3856 * Return
3857 * 0 on success, or a negative error in case of failure.
3858 *
3859 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3860 * Description
3861 * Push an element *value* in *map*. *flags* is one of:
3862 *
3863 * **BPF_EXIST**
3864 * If the queue/stack is full, the oldest element is
3865 * removed to make room for this.
3866 * Return
3867 * 0 on success, or a negative error in case of failure.
3868 *
3869 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3870 * Description
3871 * Pop an element from *map*.
3872 * Return
3873 * 0 on success, or a negative error in case of failure.
3874 *
3875 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3876 * Description
3877 * Get an element from *map* without removing it.
3878 * Return
3879 * 0 on success, or a negative error in case of failure.
3880 *
3881 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3882 * Description
3883 * For socket policies, insert *len* bytes into *msg* at offset
3884 * *start*.
3885 *
3886 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3887 * *msg* it may want to insert metadata or options into the *msg*.
3888 * This can later be read and used by any of the lower layer BPF
3889 * hooks.
3890 *
3891 * This helper may fail if under memory pressure (a malloc
3892 * fails) in these cases BPF programs will get an appropriate
3893 * error and BPF programs will need to handle them.
3894 * Return
3895 * 0 on success, or a negative error in case of failure.
3896 *
3897 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3898 * Description
3899 * Will remove *len* bytes from a *msg* starting at byte *start*.
3900 * This may result in **ENOMEM** errors under certain situations if
3901 * an allocation and copy are required due to a full ring buffer.
3902 * However, the helper will try to avoid doing the allocation
3903 * if possible. Other errors can occur if input parameters are
3904 * invalid either due to *start* byte not being valid part of *msg*
3905 * payload and/or *pop* value being to large.
3906 * Return
3907 * 0 on success, or a negative error in case of failure.
3908 *
3909 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3910 * Description
3911 * This helper is used in programs implementing IR decoding, to
3912 * report a successfully decoded pointer movement.
3913 *
3914 * The *ctx* should point to the lirc sample as passed into
3915 * the program.
3916 *
3917 * This helper is only available is the kernel was compiled with
3918 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3919 * "**y**".
3920 * Return
3921 * 0
3922 *
3923 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3924 * Description
3925 * Acquire a spinlock represented by the pointer *lock*, which is
3926 * stored as part of a value of a map. Taking the lock allows to
3927 * safely update the rest of the fields in that value. The
3928 * spinlock can (and must) later be released with a call to
3929 * **bpf_spin_unlock**\ (\ *lock*\ ).
3930 *
3931 * Spinlocks in BPF programs come with a number of restrictions
3932 * and constraints:
3933 *
3934 * * **bpf_spin_lock** objects are only allowed inside maps of
3935 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3936 * list could be extended in the future).
3937 * * BTF description of the map is mandatory.
3938 * * The BPF program can take ONE lock at a time, since taking two
3939 * or more could cause dead locks.
3940 * * Only one **struct bpf_spin_lock** is allowed per map element.
3941 * * When the lock is taken, calls (either BPF to BPF or helpers)
3942 * are not allowed.
3943 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3944 * allowed inside a spinlock-ed region.
3945 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3946 * the lock, on all execution paths, before it returns.
3947 * * The BPF program can access **struct bpf_spin_lock** only via
3948 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3949 * helpers. Loading or storing data into the **struct
3950 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3951 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3952 * of the map value must be a struct and have **struct
3953 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3954 * Nested lock inside another struct is not allowed.
3955 * * The **struct bpf_spin_lock** *lock* field in a map value must
3956 * be aligned on a multiple of 4 bytes in that value.
3957 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3958 * the **bpf_spin_lock** field to user space.
3959 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3960 * a BPF program, do not update the **bpf_spin_lock** field.
3961 * * **bpf_spin_lock** cannot be on the stack or inside a
3962 * networking packet (it can only be inside of a map values).
3963 * * **bpf_spin_lock** is available to root only.
3964 * * Tracing programs and socket filter programs cannot use
3965 * **bpf_spin_lock**\ () due to insufficient preemption checks
3966 * (but this may change in the future).
3967 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3968 * Return
3969 * 0
3970 *
3971 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3972 * Description
3973 * Release the *lock* previously locked by a call to
3974 * **bpf_spin_lock**\ (\ *lock*\ ).
3975 * Return
3976 * 0
3977 *
3978 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3979 * Description
3980 * This helper gets a **struct bpf_sock** pointer such
3981 * that all the fields in this **bpf_sock** can be accessed.
3982 * Return
3983 * A **struct bpf_sock** pointer on success, or **NULL** in
3984 * case of failure.
3985 *
3986 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3987 * Description
3988 * This helper gets a **struct bpf_tcp_sock** pointer from a
3989 * **struct bpf_sock** pointer.
3990 * Return
3991 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3992 * case of failure.
3993 *
3994 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3995 * Description
3996 * Set ECN (Explicit Congestion Notification) field of IP header
3997 * to **CE** (Congestion Encountered) if current value is **ECT**
3998 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3999 * and IPv4.
4000 * Return
4001 * 1 if the **CE** flag is set (either by the current helper call
4002 * or because it was already present), 0 if it is not set.
4003 *
4004 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
4005 * Description
4006 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
4007 * **bpf_sk_release**\ () is unnecessary and not allowed.
4008 * Return
4009 * A **struct bpf_sock** pointer on success, or **NULL** in
4010 * case of failure.
4011 *
4012 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
4013 * Description
4014 * Look for TCP socket matching *tuple*, optionally in a child
4015 * network namespace *netns*. The return value must be checked,
4016 * and if non-**NULL**, released via **bpf_sk_release**\ ().
4017 *
4018 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
4019 * that it also returns timewait or request sockets. Use
4020 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
4021 * full structure.
4022 *
4023 * This helper is available only if the kernel was compiled with
4024 * **CONFIG_NET** configuration option.
4025 * Return
4026 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
4027 * For sockets with reuseport option, the **struct bpf_sock**
4028 * result is from *reuse*\ **->socks**\ [] using the hash of the
4029 * tuple.
4030 *
4031 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4032 * Description
4033 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
4034 * the listening socket in *sk*.
4035 *
4036 * *iph* points to the start of the IPv4 or IPv6 header, while
4037 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
4038 * **sizeof**\ (**struct ipv6hdr**).
4039 *
4040 * *th* points to the start of the TCP header, while *th_len*
4041 * contains the length of the TCP header (at least
4042 * **sizeof**\ (**struct tcphdr**)).
4043 * Return
4044 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
4045 * error otherwise.
4046 *
4047 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
4048 * Description
4049 * Get name of sysctl in /proc/sys/ and copy it into provided by
4050 * program buffer *buf* of size *buf_len*.
4051 *
4052 * The buffer is always NUL terminated, unless it's zero-sized.
4053 *
4054 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
4055 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
4056 * only (e.g. "tcp_mem").
4057 * Return
4058 * Number of character copied (not including the trailing NUL).
4059 *
4060 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4061 * truncated name in this case).
4062 *
4063 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4064 * Description
4065 * Get current value of sysctl as it is presented in /proc/sys
4066 * (incl. newline, etc), and copy it as a string into provided
4067 * by program buffer *buf* of size *buf_len*.
4068 *
4069 * The whole value is copied, no matter what file position user
4070 * space issued e.g. sys_read at.
4071 *
4072 * The buffer is always NUL terminated, unless it's zero-sized.
4073 * Return
4074 * Number of character copied (not including the trailing NUL).
4075 *
4076 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4077 * truncated name in this case).
4078 *
4079 * **-EINVAL** if current value was unavailable, e.g. because
4080 * sysctl is uninitialized and read returns -EIO for it.
4081 *
4082 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
4083 * Description
4084 * Get new value being written by user space to sysctl (before
4085 * the actual write happens) and copy it as a string into
4086 * provided by program buffer *buf* of size *buf_len*.
4087 *
4088 * User space may write new value at file position > 0.
4089 *
4090 * The buffer is always NUL terminated, unless it's zero-sized.
4091 * Return
4092 * Number of character copied (not including the trailing NUL).
4093 *
4094 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
4095 * truncated name in this case).
4096 *
4097 * **-EINVAL** if sysctl is being read.
4098 *
4099 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
4100 * Description
4101 * Override new value being written by user space to sysctl with
4102 * value provided by program in buffer *buf* of size *buf_len*.
4103 *
4104 * *buf* should contain a string in same form as provided by user
4105 * space on sysctl write.
4106 *
4107 * User space may write new value at file position > 0. To override
4108 * the whole sysctl value file position should be set to zero.
4109 * Return
4110 * 0 on success.
4111 *
4112 * **-E2BIG** if the *buf_len* is too big.
4113 *
4114 * **-EINVAL** if sysctl is being read.
4115 *
4116 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
4117 * Description
4118 * Convert the initial part of the string from buffer *buf* of
4119 * size *buf_len* to a long integer according to the given base
4120 * and save the result in *res*.
4121 *
4122 * The string may begin with an arbitrary amount of white space
4123 * (as determined by **isspace**\ (3)) followed by a single
4124 * optional '**-**' sign.
4125 *
4126 * Five least significant bits of *flags* encode base, other bits
4127 * are currently unused.
4128 *
4129 * Base must be either 8, 10, 16 or 0 to detect it automatically
4130 * similar to user space **strtol**\ (3).
4131 * Return
4132 * Number of characters consumed on success. Must be positive but
4133 * no more than *buf_len*.
4134 *
4135 * **-EINVAL** if no valid digits were found or unsupported base
4136 * was provided.
4137 *
4138 * **-ERANGE** if resulting value was out of range.
4139 *
4140 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
4141 * Description
4142 * Convert the initial part of the string from buffer *buf* of
4143 * size *buf_len* to an unsigned long integer according to the
4144 * given base and save the result in *res*.
4145 *
4146 * The string may begin with an arbitrary amount of white space
4147 * (as determined by **isspace**\ (3)).
4148 *
4149 * Five least significant bits of *flags* encode base, other bits
4150 * are currently unused.
4151 *
4152 * Base must be either 8, 10, 16 or 0 to detect it automatically
4153 * similar to user space **strtoul**\ (3).
4154 * Return
4155 * Number of characters consumed on success. Must be positive but
4156 * no more than *buf_len*.
4157 *
4158 * **-EINVAL** if no valid digits were found or unsupported base
4159 * was provided.
4160 *
4161 * **-ERANGE** if resulting value was out of range.
4162 *
4163 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
4164 * Description
4165 * Get a bpf-local-storage from a *sk*.
4166 *
4167 * Logically, it could be thought of getting the value from
4168 * a *map* with *sk* as the **key**. From this
4169 * perspective, the usage is not much different from
4170 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
4171 * helper enforces the key must be a full socket and the map must
4172 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
4173 *
4174 * Underneath, the value is stored locally at *sk* instead of
4175 * the *map*. The *map* is used as the bpf-local-storage
4176 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4177 * searched against all bpf-local-storages residing at *sk*.
4178 *
4179 * *sk* is a kernel **struct sock** pointer for LSM program.
4180 * *sk* is a **struct bpf_sock** pointer for other program types.
4181 *
4182 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
4183 * used such that a new bpf-local-storage will be
4184 * created if one does not exist. *value* can be used
4185 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
4186 * the initial value of a bpf-local-storage. If *value* is
4187 * **NULL**, the new bpf-local-storage will be zero initialized.
4188 * Return
4189 * A bpf-local-storage pointer is returned on success.
4190 *
4191 * **NULL** if not found or there was an error in adding
4192 * a new bpf-local-storage.
4193 *
4194 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
4195 * Description
4196 * Delete a bpf-local-storage from a *sk*.
4197 * Return
4198 * 0 on success.
4199 *
4200 * **-ENOENT** if the bpf-local-storage cannot be found.
4201 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
4202 *
4203 * long bpf_send_signal(u32 sig)
4204 * Description
4205 * Send signal *sig* to the process of the current task.
4206 * The signal may be delivered to any of this process's threads.
4207 * Return
4208 * 0 on success or successfully queued.
4209 *
4210 * **-EBUSY** if work queue under nmi is full.
4211 *
4212 * **-EINVAL** if *sig* is invalid.
4213 *
4214 * **-EPERM** if no permission to send the *sig*.
4215 *
4216 * **-EAGAIN** if bpf program can try again.
4217 *
4218 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4219 * Description
4220 * Try to issue a SYN cookie for the packet with corresponding
4221 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
4222 *
4223 * *iph* points to the start of the IPv4 or IPv6 header, while
4224 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
4225 * **sizeof**\ (**struct ipv6hdr**).
4226 *
4227 * *th* points to the start of the TCP header, while *th_len*
4228 * contains the length of the TCP header with options (at least
4229 * **sizeof**\ (**struct tcphdr**)).
4230 * Return
4231 * On success, lower 32 bits hold the generated SYN cookie in
4232 * followed by 16 bits which hold the MSS value for that cookie,
4233 * and the top 16 bits are unused.
4234 *
4235 * On failure, the returned value is one of the following:
4236 *
4237 * **-EINVAL** SYN cookie cannot be issued due to error
4238 *
4239 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
4240 *
4241 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
4242 *
4243 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
4244 *
4245 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4246 * Description
4247 * Write raw *data* blob into a special BPF perf event held by
4248 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4249 * event must have the following attributes: **PERF_SAMPLE_RAW**
4250 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4251 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4252 *
4253 * The *flags* are used to indicate the index in *map* for which
4254 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4255 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4256 * to indicate that the index of the current CPU core should be
4257 * used.
4258 *
4259 * The value to write, of *size*, is passed through eBPF stack and
4260 * pointed by *data*.
4261 *
4262 * *ctx* is a pointer to in-kernel struct sk_buff.
4263 *
4264 * This helper is similar to **bpf_perf_event_output**\ () but
4265 * restricted to raw_tracepoint bpf programs.
4266 * Return
4267 * 0 on success, or a negative error in case of failure.
4268 *
4269 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4270 * Description
4271 * Safely attempt to read *size* bytes from user space address
4272 * *unsafe_ptr* and store the data in *dst*.
4273 * Return
4274 * 0 on success, or a negative error in case of failure.
4275 *
4276 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4277 * Description
4278 * Safely attempt to read *size* bytes from kernel space address
4279 * *unsafe_ptr* and store the data in *dst*.
4280 * Return
4281 * 0 on success, or a negative error in case of failure.
4282 *
4283 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4284 * Description
4285 * Copy a NUL terminated string from an unsafe user address
4286 * *unsafe_ptr* to *dst*. The *size* should include the
4287 * terminating NUL byte. In case the string length is smaller than
4288 * *size*, the target is not padded with further NUL bytes. If the
4289 * string length is larger than *size*, just *size*-1 bytes are
4290 * copied and the last byte is set to NUL.
4291 *
4292 * On success, returns the number of bytes that were written,
4293 * including the terminal NUL. This makes this helper useful in
4294 * tracing programs for reading strings, and more importantly to
4295 * get its length at runtime. See the following snippet:
4296 *
4297 * ::
4298 *
4299 * SEC("kprobe/sys_open")
4300 * void bpf_sys_open(struct pt_regs *ctx)
4301 * {
4302 * char buf[PATHLEN]; // PATHLEN is defined to 256
4303 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
4304 * ctx->di);
4305 *
4306 * // Consume buf, for example push it to
4307 * // userspace via bpf_perf_event_output(); we
4308 * // can use res (the string length) as event
4309 * // size, after checking its boundaries.
4310 * }
4311 *
4312 * In comparison, using **bpf_probe_read_user**\ () helper here
4313 * instead to read the string would require to estimate the length
4314 * at compile time, and would often result in copying more memory
4315 * than necessary.
4316 *
4317 * Another useful use case is when parsing individual process
4318 * arguments or individual environment variables navigating
4319 * *current*\ **->mm->arg_start** and *current*\
4320 * **->mm->env_start**: using this helper and the return value,
4321 * one can quickly iterate at the right offset of the memory area.
4322 * Return
4323 * On success, the strictly positive length of the output string,
4324 * including the trailing NUL character. On error, a negative
4325 * value.
4326 *
4327 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4328 * Description
4329 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4330 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4331 * Return
4332 * On success, the strictly positive length of the string, including
4333 * the trailing NUL character. On error, a negative value.
4334 *
4335 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4336 * Description
4337 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4338 * *rcv_nxt* is the ack_seq to be sent out.
4339 * Return
4340 * 0 on success, or a negative error in case of failure.
4341 *
4342 * long bpf_send_signal_thread(u32 sig)
4343 * Description
4344 * Send signal *sig* to the thread corresponding to the current task.
4345 * Return
4346 * 0 on success or successfully queued.
4347 *
4348 * **-EBUSY** if work queue under nmi is full.
4349 *
4350 * **-EINVAL** if *sig* is invalid.
4351 *
4352 * **-EPERM** if no permission to send the *sig*.
4353 *
4354 * **-EAGAIN** if bpf program can try again.
4355 *
4356 * u64 bpf_jiffies64(void)
4357 * Description
4358 * Obtain the 64bit jiffies
4359 * Return
4360 * The 64 bit jiffies
4361 *
4362 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4363 * Description
4364 * For an eBPF program attached to a perf event, retrieve the
4365 * branch records (**struct perf_branch_entry**) associated to *ctx*
4366 * and store it in the buffer pointed by *buf* up to size
4367 * *size* bytes.
4368 * Return
4369 * On success, number of bytes written to *buf*. On error, a
4370 * negative value.
4371 *
4372 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4373 * instead return the number of bytes required to store all the
4374 * branch entries. If this flag is set, *buf* may be NULL.
4375 *
4376 * **-EINVAL** if arguments invalid or **size** not a multiple
4377 * of **sizeof**\ (**struct perf_branch_entry**\ ).
4378 *
4379 * **-ENOENT** if architecture does not support branch records.
4380 *
4381 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4382 * Description
4383 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
4384 * *namespace* will be returned in *nsdata*.
4385 * Return
4386 * 0 on success, or one of the following in case of failure:
4387 *
4388 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
4389 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
4390 *
4391 * **-ENOENT** if pidns does not exists for the current task.
4392 *
4393 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4394 * Description
4395 * Write raw *data* blob into a special BPF perf event held by
4396 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4397 * event must have the following attributes: **PERF_SAMPLE_RAW**
4398 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4399 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4400 *
4401 * The *flags* are used to indicate the index in *map* for which
4402 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4403 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4404 * to indicate that the index of the current CPU core should be
4405 * used.
4406 *
4407 * The value to write, of *size*, is passed through eBPF stack and
4408 * pointed by *data*.
4409 *
4410 * *ctx* is a pointer to in-kernel struct xdp_buff.
4411 *
4412 * This helper is similar to **bpf_perf_eventoutput**\ () but
4413 * restricted to raw_tracepoint bpf programs.
4414 * Return
4415 * 0 on success, or a negative error in case of failure.
4416 *
4417 * u64 bpf_get_netns_cookie(void *ctx)
4418 * Description
4419 * Retrieve the cookie (generated by the kernel) of the network
4420 * namespace the input *ctx* is associated with. The network
4421 * namespace cookie remains stable for its lifetime and provides
4422 * a global identifier that can be assumed unique. If *ctx* is
4423 * NULL, then the helper returns the cookie for the initial
4424 * network namespace. The cookie itself is very similar to that
4425 * of **bpf_get_socket_cookie**\ () helper, but for network
4426 * namespaces instead of sockets.
4427 * Return
4428 * A 8-byte long opaque number.
4429 *
4430 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4431 * Description
4432 * Return id of cgroup v2 that is ancestor of the cgroup associated
4433 * with the current task at the *ancestor_level*. The root cgroup
4434 * is at *ancestor_level* zero and each step down the hierarchy
4435 * increments the level. If *ancestor_level* == level of cgroup
4436 * associated with the current task, then return value will be the
4437 * same as that of **bpf_get_current_cgroup_id**\ ().
4438 *
4439 * The helper is useful to implement policies based on cgroups
4440 * that are upper in hierarchy than immediate cgroup associated
4441 * with the current task.
4442 *
4443 * The format of returned id and helper limitations are same as in
4444 * **bpf_get_current_cgroup_id**\ ().
4445 * Return
4446 * The id is returned or 0 in case the id could not be retrieved.
4447 *
4448 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4449 * Description
4450 * Helper is overloaded depending on BPF program type. This
4451 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4452 * **BPF_PROG_TYPE_SCHED_ACT** programs.
4453 *
4454 * Assign the *sk* to the *skb*. When combined with appropriate
4455 * routing configuration to receive the packet towards the socket,
4456 * will cause *skb* to be delivered to the specified socket.
4457 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
4458 * **bpf_clone_redirect**\ () or other methods outside of BPF may
4459 * interfere with successful delivery to the socket.
4460 *
4461 * This operation is only valid from TC ingress path.
4462 *
4463 * The *flags* argument must be zero.
4464 * Return
4465 * 0 on success, or a negative error in case of failure:
4466 *
4467 * **-EINVAL** if specified *flags* are not supported.
4468 *
4469 * **-ENOENT** if the socket is unavailable for assignment.
4470 *
4471 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
4472 *
4473 * **-EOPNOTSUPP** if the operation is not supported, for example
4474 * a call from outside of TC ingress.
4475 *
4476 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4477 * Description
4478 * Helper is overloaded depending on BPF program type. This
4479 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4480 *
4481 * Select the *sk* as a result of a socket lookup.
4482 *
4483 * For the operation to succeed passed socket must be compatible
4484 * with the packet description provided by the *ctx* object.
4485 *
4486 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4487 * be an exact match. While IP family (**AF_INET** or
4488 * **AF_INET6**) must be compatible, that is IPv6 sockets
4489 * that are not v6-only can be selected for IPv4 packets.
4490 *
4491 * Only TCP listeners and UDP unconnected sockets can be
4492 * selected. *sk* can also be NULL to reset any previous
4493 * selection.
4494 *
4495 * *flags* argument can combination of following values:
4496 *
4497 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4498 * socket selection, potentially done by a BPF program
4499 * that ran before us.
4500 *
4501 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4502 * load-balancing within reuseport group for the socket
4503 * being selected.
4504 *
4505 * On success *ctx->sk* will point to the selected socket.
4506 *
4507 * Return
4508 * 0 on success, or a negative errno in case of failure.
4509 *
4510 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4511 * not compatible with packet family (*ctx->family*).
4512 *
4513 * * **-EEXIST** if socket has been already selected,
4514 * potentially by another program, and
4515 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4516 *
4517 * * **-EINVAL** if unsupported flags were specified.
4518 *
4519 * * **-EPROTOTYPE** if socket L4 protocol
4520 * (*sk->protocol*) doesn't match packet protocol
4521 * (*ctx->protocol*).
4522 *
4523 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4524 * state (TCP listening or UDP unconnected).
4525 *
4526 * u64 bpf_ktime_get_boot_ns(void)
4527 * Description
4528 * Return the time elapsed since system boot, in nanoseconds.
4529 * Does include the time the system was suspended.
4530 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4531 * Return
4532 * Current *ktime*.
4533 *
4534 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4535 * Description
4536 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4537 * out the format string.
4538 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4539 * the format string itself. The *data* and *data_len* are format string
4540 * arguments. The *data* are a **u64** array and corresponding format string
4541 * values are stored in the array. For strings and pointers where pointees
4542 * are accessed, only the pointer values are stored in the *data* array.
4543 * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4544 *
4545 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4546 * Reading kernel memory may fail due to either invalid address or
4547 * valid address but requiring a major memory fault. If reading kernel memory
4548 * fails, the string for **%s** will be an empty string, and the ip
4549 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4550 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4551 * Return
4552 * 0 on success, or a negative error in case of failure:
4553 *
4554 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4555 * by returning 1 from bpf program.
4556 *
4557 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4558 *
4559 * **-E2BIG** if *fmt* contains too many format specifiers.
4560 *
4561 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4562 *
4563 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4564 * Description
4565 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4566 * The *m* represents the seq_file. The *data* and *len* represent the
4567 * data to write in bytes.
4568 * Return
4569 * 0 on success, or a negative error in case of failure:
4570 *
4571 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4572 *
4573 * u64 bpf_sk_cgroup_id(void *sk)
4574 * Description
4575 * Return the cgroup v2 id of the socket *sk*.
4576 *
4577 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4578 * returned from **bpf_sk_lookup_xxx**\ (),
4579 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4580 * same as in **bpf_skb_cgroup_id**\ ().
4581 *
4582 * This helper is available only if the kernel was compiled with
4583 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4584 * Return
4585 * The id is returned or 0 in case the id could not be retrieved.
4586 *
4587 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4588 * Description
4589 * Return id of cgroup v2 that is ancestor of cgroup associated
4590 * with the *sk* at the *ancestor_level*. The root cgroup is at
4591 * *ancestor_level* zero and each step down the hierarchy
4592 * increments the level. If *ancestor_level* == level of cgroup
4593 * associated with *sk*, then return value will be same as that
4594 * of **bpf_sk_cgroup_id**\ ().
4595 *
4596 * The helper is useful to implement policies based on cgroups
4597 * that are upper in hierarchy than immediate cgroup associated
4598 * with *sk*.
4599 *
4600 * The format of returned id and helper limitations are same as in
4601 * **bpf_sk_cgroup_id**\ ().
4602 * Return
4603 * The id is returned or 0 in case the id could not be retrieved.
4604 *
4605 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4606 * Description
4607 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4608 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4609 * of new data availability is sent.
4610 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4611 * of new data availability is sent unconditionally.
4612 * If **0** is specified in *flags*, an adaptive notification
4613 * of new data availability is sent.
4614 *
4615 * An adaptive notification is a notification sent whenever the user-space
4616 * process has caught up and consumed all available payloads. In case the user-space
4617 * process is still processing a previous payload, then no notification is needed
4618 * as it will process the newly added payload automatically.
4619 * Return
4620 * 0 on success, or a negative error in case of failure.
4621 *
4622 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4623 * Description
4624 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4625 * *flags* must be 0.
4626 * Return
4627 * Valid pointer with *size* bytes of memory available; NULL,
4628 * otherwise.
4629 *
4630 * void bpf_ringbuf_submit(void *data, u64 flags)
4631 * Description
4632 * Submit reserved ring buffer sample, pointed to by *data*.
4633 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4634 * of new data availability is sent.
4635 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4636 * of new data availability is sent unconditionally.
4637 * If **0** is specified in *flags*, an adaptive notification
4638 * of new data availability is sent.
4639 *
4640 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4641 * Return
4642 * Nothing. Always succeeds.
4643 *
4644 * void bpf_ringbuf_discard(void *data, u64 flags)
4645 * Description
4646 * Discard reserved ring buffer sample, pointed to by *data*.
4647 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4648 * of new data availability is sent. Discarded records remain in
4649 * the ring buffer until consumed by user space, so a later submit
4650 * using adaptive wakeup might not wake up the consumer.
4651 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4652 * of new data availability is sent unconditionally.
4653 * If **0** is specified in *flags*, an adaptive notification
4654 * of new data availability is sent.
4655 *
4656 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4657 * Return
4658 * Nothing. Always succeeds.
4659 *
4660 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4661 * Description
4662 * Query various characteristics of provided ring buffer. What
4663 * exactly is queries is determined by *flags*:
4664 *
4665 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4666 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4667 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4668 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4669 *
4670 * Data returned is just a momentary snapshot of actual values
4671 * and could be inaccurate, so this facility should be used to
4672 * power heuristics and for reporting, not to make 100% correct
4673 * calculation.
4674 * Return
4675 * Requested value, or 0, if *flags* are not recognized.
4676 *
4677 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4678 * Description
4679 * Change the skbs checksum level by one layer up or down, or
4680 * reset it entirely to none in order to have the stack perform
4681 * checksum validation. The level is applicable to the following
4682 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4683 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4684 * through **bpf_skb_adjust_room**\ () helper with passing in
4685 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4686 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4687 * the UDP header is removed. Similarly, an encap of the latter
4688 * into the former could be accompanied by a helper call to
4689 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4690 * skb is still intended to be processed in higher layers of the
4691 * stack instead of just egressing at tc.
4692 *
4693 * There are three supported level settings at this time:
4694 *
4695 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4696 * with CHECKSUM_UNNECESSARY.
4697 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4698 * with CHECKSUM_UNNECESSARY.
4699 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4700 * sets CHECKSUM_NONE to force checksum validation by the stack.
4701 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4702 * skb->csum_level.
4703 * Return
4704 * 0 on success, or a negative error in case of failure. In the
4705 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4706 * is returned or the error code -EACCES in case the skb is not
4707 * subject to CHECKSUM_UNNECESSARY.
4708 *
4709 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4710 * Description
4711 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4712 * Return
4713 * *sk* if casting is valid, or **NULL** otherwise.
4714 *
4715 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4716 * Description
4717 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4718 * Return
4719 * *sk* if casting is valid, or **NULL** otherwise.
4720 *
4721 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4722 * Description
4723 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4724 * Return
4725 * *sk* if casting is valid, or **NULL** otherwise.
4726 *
4727 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4728 * Description
4729 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4730 * Return
4731 * *sk* if casting is valid, or **NULL** otherwise.
4732 *
4733 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4734 * Description
4735 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4736 * Return
4737 * *sk* if casting is valid, or **NULL** otherwise.
4738 *
4739 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4740 * Description
4741 * Return a user or a kernel stack in bpf program provided buffer.
4742 * Note: the user stack will only be populated if the *task* is
4743 * the current task; all other tasks will return -EOPNOTSUPP.
4744 * To achieve this, the helper needs *task*, which is a valid
4745 * pointer to **struct task_struct**. To store the stacktrace, the
4746 * bpf program provides *buf* with a nonnegative *size*.
4747 *
4748 * The last argument, *flags*, holds the number of stack frames to
4749 * skip (from 0 to 255), masked with
4750 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4751 * the following flags:
4752 *
4753 * **BPF_F_USER_STACK**
4754 * Collect a user space stack instead of a kernel stack.
4755 * The *task* must be the current task.
4756 * **BPF_F_USER_BUILD_ID**
4757 * Collect buildid+offset instead of ips for user stack,
4758 * only valid if **BPF_F_USER_STACK** is also specified.
4759 *
4760 * **bpf_get_task_stack**\ () can collect up to
4761 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4762 * to sufficient large buffer size. Note that
4763 * this limit can be controlled with the **sysctl** program, and
4764 * that it should be manually increased in order to profile long
4765 * user stacks (such as stacks for Java programs). To do so, use:
4766 *
4767 * ::
4768 *
4769 * # sysctl kernel.perf_event_max_stack=<new value>
4770 * Return
4771 * The non-negative copied *buf* length equal to or less than
4772 * *size* on success, or a negative error in case of failure.
4773 *
4774 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4775 * Description
4776 * Load header option. Support reading a particular TCP header
4777 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4778 *
4779 * If *flags* is 0, it will search the option from the
4780 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4781 * has details on what skb_data contains under different
4782 * *skops*\ **->op**.
4783 *
4784 * The first byte of the *searchby_res* specifies the
4785 * kind that it wants to search.
4786 *
4787 * If the searching kind is an experimental kind
4788 * (i.e. 253 or 254 according to RFC6994). It also
4789 * needs to specify the "magic" which is either
4790 * 2 bytes or 4 bytes. It then also needs to
4791 * specify the size of the magic by using
4792 * the 2nd byte which is "kind-length" of a TCP
4793 * header option and the "kind-length" also
4794 * includes the first 2 bytes "kind" and "kind-length"
4795 * itself as a normal TCP header option also does.
4796 *
4797 * For example, to search experimental kind 254 with
4798 * 2 byte magic 0xeB9F, the searchby_res should be
4799 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4800 *
4801 * To search for the standard window scale option (3),
4802 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4803 * Note, kind-length must be 0 for regular option.
4804 *
4805 * Searching for No-Op (0) and End-of-Option-List (1) are
4806 * not supported.
4807 *
4808 * *len* must be at least 2 bytes which is the minimal size
4809 * of a header option.
4810 *
4811 * Supported flags:
4812 *
4813 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4814 * saved_syn packet or the just-received syn packet.
4815 *
4816 * Return
4817 * > 0 when found, the header option is copied to *searchby_res*.
4818 * The return value is the total length copied. On failure, a
4819 * negative error code is returned:
4820 *
4821 * **-EINVAL** if a parameter is invalid.
4822 *
4823 * **-ENOMSG** if the option is not found.
4824 *
4825 * **-ENOENT** if no syn packet is available when
4826 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4827 *
4828 * **-ENOSPC** if there is not enough space. Only *len* number of
4829 * bytes are copied.
4830 *
4831 * **-EFAULT** on failure to parse the header options in the
4832 * packet.
4833 *
4834 * **-EPERM** if the helper cannot be used under the current
4835 * *skops*\ **->op**.
4836 *
4837 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4838 * Description
4839 * Store header option. The data will be copied
4840 * from buffer *from* with length *len* to the TCP header.
4841 *
4842 * The buffer *from* should have the whole option that
4843 * includes the kind, kind-length, and the actual
4844 * option data. The *len* must be at least kind-length
4845 * long. The kind-length does not have to be 4 byte
4846 * aligned. The kernel will take care of the padding
4847 * and setting the 4 bytes aligned value to th->doff.
4848 *
4849 * This helper will check for duplicated option
4850 * by searching the same option in the outgoing skb.
4851 *
4852 * This helper can only be called during
4853 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4854 *
4855 * Return
4856 * 0 on success, or negative error in case of failure:
4857 *
4858 * **-EINVAL** If param is invalid.
4859 *
4860 * **-ENOSPC** if there is not enough space in the header.
4861 * Nothing has been written
4862 *
4863 * **-EEXIST** if the option already exists.
4864 *
4865 * **-EFAULT** on failure to parse the existing header options.
4866 *
4867 * **-EPERM** if the helper cannot be used under the current
4868 * *skops*\ **->op**.
4869 *
4870 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4871 * Description
4872 * Reserve *len* bytes for the bpf header option. The
4873 * space will be used by **bpf_store_hdr_opt**\ () later in
4874 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4875 *
4876 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4877 * the total number of bytes will be reserved.
4878 *
4879 * This helper can only be called during
4880 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4881 *
4882 * Return
4883 * 0 on success, or negative error in case of failure:
4884 *
4885 * **-EINVAL** if a parameter is invalid.
4886 *
4887 * **-ENOSPC** if there is not enough space in the header.
4888 *
4889 * **-EPERM** if the helper cannot be used under the current
4890 * *skops*\ **->op**.
4891 *
4892 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4893 * Description
4894 * Get a bpf_local_storage from an *inode*.
4895 *
4896 * Logically, it could be thought of as getting the value from
4897 * a *map* with *inode* as the **key**. From this
4898 * perspective, the usage is not much different from
4899 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4900 * helper enforces the key must be an inode and the map must also
4901 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4902 *
4903 * Underneath, the value is stored locally at *inode* instead of
4904 * the *map*. The *map* is used as the bpf-local-storage
4905 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4906 * searched against all bpf_local_storage residing at *inode*.
4907 *
4908 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4909 * used such that a new bpf_local_storage will be
4910 * created if one does not exist. *value* can be used
4911 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4912 * the initial value of a bpf_local_storage. If *value* is
4913 * **NULL**, the new bpf_local_storage will be zero initialized.
4914 * Return
4915 * A bpf_local_storage pointer is returned on success.
4916 *
4917 * **NULL** if not found or there was an error in adding
4918 * a new bpf_local_storage.
4919 *
4920 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4921 * Description
4922 * Delete a bpf_local_storage from an *inode*.
4923 * Return
4924 * 0 on success.
4925 *
4926 * **-ENOENT** if the bpf_local_storage cannot be found.
4927 *
4928 * long bpf_d_path(const struct path *path, char *buf, u32 sz)
4929 * Description
4930 * Return full path for given **struct path** object, which
4931 * needs to be the kernel BTF *path* object. The path is
4932 * returned in the provided buffer *buf* of size *sz* and
4933 * is zero terminated.
4934 *
4935 * Return
4936 * On success, the strictly positive length of the string,
4937 * including the trailing NUL character. On error, a negative
4938 * value.
4939 *
4940 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4941 * Description
4942 * Read *size* bytes from user space address *user_ptr* and store
4943 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4944 * Return
4945 * 0 on success, or a negative error in case of failure.
4946 *
4947 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4948 * Description
4949 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4950 * using *ptr*->type_id. This value should specify the type
4951 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4952 * can be used to look up vmlinux BTF type ids. Traversing the
4953 * data structure using BTF, the type information and values are
4954 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4955 * the pointer data is carried out to avoid kernel crashes during
4956 * operation. Smaller types can use string space on the stack;
4957 * larger programs can use map data to store the string
4958 * representation.
4959 *
4960 * The string can be subsequently shared with userspace via
4961 * bpf_perf_event_output() or ring buffer interfaces.
4962 * bpf_trace_printk() is to be avoided as it places too small
4963 * a limit on string size to be useful.
4964 *
4965 * *flags* is a combination of
4966 *
4967 * **BTF_F_COMPACT**
4968 * no formatting around type information
4969 * **BTF_F_NONAME**
4970 * no struct/union member names/types
4971 * **BTF_F_PTR_RAW**
4972 * show raw (unobfuscated) pointer values;
4973 * equivalent to printk specifier %px.
4974 * **BTF_F_ZERO**
4975 * show zero-valued struct/union members; they
4976 * are not displayed by default
4977 *
4978 * Return
4979 * The number of bytes that were written (or would have been
4980 * written if output had to be truncated due to string size),
4981 * or a negative error in cases of failure.
4982 *
4983 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4984 * Description
4985 * Use BTF to write to seq_write a string representation of
4986 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4987 * *flags* are identical to those used for bpf_snprintf_btf.
4988 * Return
4989 * 0 on success or a negative error in case of failure.
4990 *
4991 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4992 * Description
4993 * See **bpf_get_cgroup_classid**\ () for the main description.
4994 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4995 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4996 * associated socket instead of the current process.
4997 * Return
4998 * The id is returned or 0 in case the id could not be retrieved.
4999 *
5000 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
5001 * Description
5002 * Redirect the packet to another net device of index *ifindex*
5003 * and fill in L2 addresses from neighboring subsystem. This helper
5004 * is somewhat similar to **bpf_redirect**\ (), except that it
5005 * populates L2 addresses as well, meaning, internally, the helper
5006 * relies on the neighbor lookup for the L2 address of the nexthop.
5007 *
5008 * The helper will perform a FIB lookup based on the skb's
5009 * networking header to get the address of the next hop, unless
5010 * this is supplied by the caller in the *params* argument. The
5011 * *plen* argument indicates the len of *params* and should be set
5012 * to 0 if *params* is NULL.
5013 *
5014 * The *flags* argument is reserved and must be 0. The helper is
5015 * currently only supported for tc BPF program types, and enabled
5016 * for IPv4 and IPv6 protocols.
5017 * Return
5018 * The helper returns **TC_ACT_REDIRECT** on success or
5019 * **TC_ACT_SHOT** on error.
5020 *
5021 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
5022 * Description
5023 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
5024 * pointer to the percpu kernel variable on *cpu*. A ksym is an
5025 * extern variable decorated with '__ksym'. For ksym, there is a
5026 * global var (either static or global) defined of the same name
5027 * in the kernel. The ksym is percpu if the global var is percpu.
5028 * The returned pointer points to the global percpu var on *cpu*.
5029 *
5030 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
5031 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
5032 * happens if *cpu* is larger than nr_cpu_ids. The caller of
5033 * bpf_per_cpu_ptr() must check the returned value.
5034 * Return
5035 * A pointer pointing to the kernel percpu variable on *cpu*, or
5036 * NULL, if *cpu* is invalid.
5037 *
5038 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
5039 * Description
5040 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
5041 * pointer to the percpu kernel variable on this cpu. See the
5042 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
5043 *
5044 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
5045 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
5046 * never return NULL.
5047 * Return
5048 * A pointer pointing to the kernel percpu variable on this cpu.
5049 *
5050 * long bpf_redirect_peer(u32 ifindex, u64 flags)
5051 * Description
5052 * Redirect the packet to another net device of index *ifindex*.
5053 * This helper is somewhat similar to **bpf_redirect**\ (), except
5054 * that the redirection happens to the *ifindex*' peer device and
5055 * the netns switch takes place from ingress to ingress without
5056 * going through the CPU's backlog queue.
5057 *
5058 * *skb*\ **->mark** and *skb*\ **->tstamp** are not cleared during
5059 * the netns switch.
5060 *
5061 * The *flags* argument is reserved and must be 0. The helper is
5062 * currently only supported for tc BPF program types at the
5063 * ingress hook and for veth and netkit target device types. The
5064 * peer device must reside in a different network namespace.
5065 * Return
5066 * The helper returns **TC_ACT_REDIRECT** on success or
5067 * **TC_ACT_SHOT** on error.
5068 *
5069 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
5070 * Description
5071 * Get a bpf_local_storage from the *task*.
5072 *
5073 * Logically, it could be thought of as getting the value from
5074 * a *map* with *task* as the **key**. From this
5075 * perspective, the usage is not much different from
5076 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
5077 * helper enforces the key must be a task_struct and the map must also
5078 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
5079 *
5080 * Underneath, the value is stored locally at *task* instead of
5081 * the *map*. The *map* is used as the bpf-local-storage
5082 * "type". The bpf-local-storage "type" (i.e. the *map*) is
5083 * searched against all bpf_local_storage residing at *task*.
5084 *
5085 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5086 * used such that a new bpf_local_storage will be
5087 * created if one does not exist. *value* can be used
5088 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5089 * the initial value of a bpf_local_storage. If *value* is
5090 * **NULL**, the new bpf_local_storage will be zero initialized.
5091 * Return
5092 * A bpf_local_storage pointer is returned on success.
5093 *
5094 * **NULL** if not found or there was an error in adding
5095 * a new bpf_local_storage.
5096 *
5097 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
5098 * Description
5099 * Delete a bpf_local_storage from a *task*.
5100 * Return
5101 * 0 on success.
5102 *
5103 * **-ENOENT** if the bpf_local_storage cannot be found.
5104 *
5105 * struct task_struct *bpf_get_current_task_btf(void)
5106 * Description
5107 * Return a BTF pointer to the "current" task.
5108 * This pointer can also be used in helpers that accept an
5109 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
5110 * Return
5111 * Pointer to the current task.
5112 *
5113 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
5114 * Description
5115 * Set or clear certain options on *bprm*:
5116 *
5117 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
5118 * which sets the **AT_SECURE** auxv for glibc. The bit
5119 * is cleared if the flag is not specified.
5120 * Return
5121 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
5122 *
5123 * u64 bpf_ktime_get_coarse_ns(void)
5124 * Description
5125 * Return a coarse-grained version of the time elapsed since
5126 * system boot, in nanoseconds. Does not include time the system
5127 * was suspended.
5128 *
5129 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
5130 * Return
5131 * Current *ktime*.
5132 *
5133 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
5134 * Description
5135 * Returns the stored IMA hash of the *inode* (if it's available).
5136 * If the hash is larger than *size*, then only *size*
5137 * bytes will be copied to *dst*
5138 * Return
5139 * The **hash_algo** is returned on success,
5140 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if
5141 * invalid arguments are passed.
5142 *
5143 * struct socket *bpf_sock_from_file(struct file *file)
5144 * Description
5145 * If the given file represents a socket, returns the associated
5146 * socket.
5147 * Return
5148 * A pointer to a struct socket on success or NULL if the file is
5149 * not a socket.
5150 *
5151 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
5152 * Description
5153 * Check packet size against exceeding MTU of net device (based
5154 * on *ifindex*). This helper will likely be used in combination
5155 * with helpers that adjust/change the packet size.
5156 *
5157 * The argument *len_diff* can be used for querying with a planned
5158 * size change. This allows to check MTU prior to changing packet
5159 * ctx. Providing a *len_diff* adjustment that is larger than the
5160 * actual packet size (resulting in negative packet size) will in
5161 * principle not exceed the MTU, which is why it is not considered
5162 * a failure. Other BPF helpers are needed for performing the
5163 * planned size change; therefore the responsibility for catching
5164 * a negative packet size belongs in those helpers.
5165 *
5166 * Specifying *ifindex* zero means the MTU check is performed
5167 * against the current net device. This is practical if this isn't
5168 * used prior to redirect.
5169 *
5170 * On input *mtu_len* must be a valid pointer, else verifier will
5171 * reject BPF program. If the value *mtu_len* is initialized to
5172 * zero then the ctx packet size is use. When value *mtu_len* is
5173 * provided as input this specify the L3 length that the MTU check
5174 * is done against. Remember XDP and TC length operate at L2, but
5175 * this value is L3 as this correlate to MTU and IP-header tot_len
5176 * values which are L3 (similar behavior as bpf_fib_lookup).
5177 *
5178 * The Linux kernel route table can configure MTUs on a more
5179 * specific per route level, which is not provided by this helper.
5180 * For route level MTU checks use the **bpf_fib_lookup**\ ()
5181 * helper.
5182 *
5183 * *ctx* is either **struct xdp_md** for XDP programs or
5184 * **struct sk_buff** for tc cls_act programs.
5185 *
5186 * The *flags* argument can be a combination of one or more of the
5187 * following values:
5188 *
5189 * **BPF_MTU_CHK_SEGS**
5190 * This flag will only works for *ctx* **struct sk_buff**.
5191 * If packet context contains extra packet segment buffers
5192 * (often knows as GSO skb), then MTU check is harder to
5193 * check at this point, because in transmit path it is
5194 * possible for the skb packet to get re-segmented
5195 * (depending on net device features). This could still be
5196 * a MTU violation, so this flag enables performing MTU
5197 * check against segments, with a different violation
5198 * return code to tell it apart. Check cannot use len_diff.
5199 *
5200 * On return *mtu_len* pointer contains the MTU value of the net
5201 * device. Remember the net device configured MTU is the L3 size,
5202 * which is returned here and XDP and TC length operate at L2.
5203 * Helper take this into account for you, but remember when using
5204 * MTU value in your BPF-code.
5205 *
5206 * Return
5207 * * 0 on success, and populate MTU value in *mtu_len* pointer.
5208 *
5209 * * < 0 if any input argument is invalid (*mtu_len* not updated)
5210 *
5211 * MTU violations return positive values, but also populate MTU
5212 * value in *mtu_len* pointer, as this can be needed for
5213 * implementing PMTU handing:
5214 *
5215 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
5216 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
5217 *
5218 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
5219 * Description
5220 * For each element in **map**, call **callback_fn** function with
5221 * **map**, **callback_ctx** and other map-specific parameters.
5222 * The **callback_fn** should be a static function and
5223 * the **callback_ctx** should be a pointer to the stack.
5224 * The **flags** is used to control certain aspects of the helper.
5225 * Currently, the **flags** must be 0.
5226 *
5227 * The following are a list of supported map types and their
5228 * respective expected callback signatures:
5229 *
5230 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
5231 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
5232 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
5233 *
5234 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
5235 *
5236 * For per_cpu maps, the map_value is the value on the cpu where the
5237 * bpf_prog is running.
5238 *
5239 * If **callback_fn** return 0, the helper will continue to the next
5240 * element. If return value is 1, the helper will skip the rest of
5241 * elements and return. Other return values are not used now.
5242 *
5243 * Return
5244 * The number of traversed map elements for success, **-EINVAL** for
5245 * invalid **flags**.
5246 *
5247 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
5248 * Description
5249 * Outputs a string into the **str** buffer of size **str_size**
5250 * based on a format string stored in a read-only map pointed by
5251 * **fmt**.
5252 *
5253 * Each format specifier in **fmt** corresponds to one u64 element
5254 * in the **data** array. For strings and pointers where pointees
5255 * are accessed, only the pointer values are stored in the *data*
5256 * array. The *data_len* is the size of *data* in bytes - must be
5257 * a multiple of 8.
5258 *
5259 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5260 * memory. Reading kernel memory may fail due to either invalid
5261 * address or valid address but requiring a major memory fault. If
5262 * reading kernel memory fails, the string for **%s** will be an
5263 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5264 * Not returning error to bpf program is consistent with what
5265 * **bpf_trace_printk**\ () does for now.
5266 *
5267 * Return
5268 * The strictly positive length of the formatted string, including
5269 * the trailing zero character. If the return value is greater than
5270 * **str_size**, **str** contains a truncated string, guaranteed to
5271 * be zero-terminated except when **str_size** is 0.
5272 *
5273 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5274 *
5275 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5276 * Description
5277 * Execute bpf syscall with given arguments.
5278 * Return
5279 * A syscall result.
5280 *
5281 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5282 * Description
5283 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5284 * Return
5285 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5286 *
5287 * long bpf_sys_close(u32 fd)
5288 * Description
5289 * Execute close syscall for given FD.
5290 * Return
5291 * A syscall result.
5292 *
5293 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5294 * Description
5295 * Initialize the timer.
5296 * First 4 bits of *flags* specify clockid.
5297 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5298 * All other bits of *flags* are reserved.
5299 * The verifier will reject the program if *timer* is not from
5300 * the same *map*.
5301 * Return
5302 * 0 on success.
5303 * **-EBUSY** if *timer* is already initialized.
5304 * **-EINVAL** if invalid *flags* are passed.
5305 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5306 * The user space should either hold a file descriptor to a map with timers
5307 * or pin such map in bpffs. When map is unpinned or file descriptor is
5308 * closed all timers in the map will be cancelled and freed.
5309 *
5310 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5311 * Description
5312 * Configure the timer to call *callback_fn* static function.
5313 * Return
5314 * 0 on success.
5315 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5316 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5317 * The user space should either hold a file descriptor to a map with timers
5318 * or pin such map in bpffs. When map is unpinned or file descriptor is
5319 * closed all timers in the map will be cancelled and freed.
5320 *
5321 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5322 * Description
5323 * Set timer expiration N nanoseconds from the current time. The
5324 * configured callback will be invoked in soft irq context on some cpu
5325 * and will not repeat unless another bpf_timer_start() is made.
5326 * In such case the next invocation can migrate to a different cpu.
5327 * Since struct bpf_timer is a field inside map element the map
5328 * owns the timer. The bpf_timer_set_callback() will increment refcnt
5329 * of BPF program to make sure that callback_fn code stays valid.
5330 * When user space reference to a map reaches zero all timers
5331 * in a map are cancelled and corresponding program's refcnts are
5332 * decremented. This is done to make sure that Ctrl-C of a user
5333 * process doesn't leave any timers running. If map is pinned in
5334 * bpffs the callback_fn can re-arm itself indefinitely.
5335 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5336 * cancel and free the timer in the given map element.
5337 * The map can contain timers that invoke callback_fn-s from different
5338 * programs. The same callback_fn can serve different timers from
5339 * different maps if key/value layout matches across maps.
5340 * Every bpf_timer_set_callback() can have different callback_fn.
5341 *
5342 * *flags* can be one of:
5343 *
5344 * **BPF_F_TIMER_ABS**
5345 * Start the timer in absolute expire value instead of the
5346 * default relative one.
5347 * **BPF_F_TIMER_CPU_PIN**
5348 * Timer will be pinned to the CPU of the caller.
5349 *
5350 * Return
5351 * 0 on success.
5352 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5353 * or invalid *flags* are passed.
5354 *
5355 * long bpf_timer_cancel(struct bpf_timer *timer)
5356 * Description
5357 * Cancel the timer and wait for callback_fn to finish if it was running.
5358 * Return
5359 * 0 if the timer was not active.
5360 * 1 if the timer was active.
5361 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5362 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5363 * own timer which would have led to a deadlock otherwise.
5364 *
5365 * u64 bpf_get_func_ip(void *ctx)
5366 * Description
5367 * Get address of the traced function (for tracing and kprobe programs).
5368 *
5369 * When called for kprobe program attached as uprobe it returns
5370 * probe address for both entry and return uprobe.
5371 *
5372 * Return
5373 * Address of the traced function for kprobe.
5374 * 0 for kprobes placed within the function (not at the entry).
5375 * Address of the probe for uprobe and return uprobe.
5376 *
5377 * u64 bpf_get_attach_cookie(void *ctx)
5378 * Description
5379 * Get bpf_cookie value provided (optionally) during the program
5380 * attachment. It might be different for each individual
5381 * attachment, even if BPF program itself is the same.
5382 * Expects BPF program context *ctx* as a first argument.
5383 *
5384 * Supported for the following program types:
5385 * - kprobe/uprobe;
5386 * - tracepoint;
5387 * - perf_event.
5388 * Return
5389 * Value specified by user at BPF link creation/attachment time
5390 * or 0, if it was not specified.
5391 *
5392 * long bpf_task_pt_regs(struct task_struct *task)
5393 * Description
5394 * Get the struct pt_regs associated with **task**.
5395 * Return
5396 * A pointer to struct pt_regs.
5397 *
5398 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5399 * Description
5400 * Get branch trace from hardware engines like Intel LBR. The
5401 * hardware engine is stopped shortly after the helper is
5402 * called. Therefore, the user need to filter branch entries
5403 * based on the actual use case. To capture branch trace
5404 * before the trigger point of the BPF program, the helper
5405 * should be called at the beginning of the BPF program.
5406 *
5407 * The data is stored as struct perf_branch_entry into output
5408 * buffer *entries*. *size* is the size of *entries* in bytes.
5409 * *flags* is reserved for now and must be zero.
5410 *
5411 * Return
5412 * On success, number of bytes written to *buf*. On error, a
5413 * negative value.
5414 *
5415 * **-EINVAL** if *flags* is not zero.
5416 *
5417 * **-ENOENT** if architecture does not support branch records.
5418 *
5419 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5420 * Description
5421 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5422 * to format and can handle more format args as a result.
5423 *
5424 * Arguments are to be used as in **bpf_seq_printf**\ () helper.
5425 * Return
5426 * The number of bytes written to the buffer, or a negative error
5427 * in case of failure.
5428 *
5429 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5430 * Description
5431 * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5432 * Return
5433 * *sk* if casting is valid, or **NULL** otherwise.
5434 *
5435 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5436 * Description
5437 * Get the address of a kernel symbol, returned in *res*. *res* is
5438 * set to 0 if the symbol is not found.
5439 * Return
5440 * On success, zero. On error, a negative value.
5441 *
5442 * **-EINVAL** if *flags* is not zero.
5443 *
5444 * **-EINVAL** if string *name* is not the same size as *name_sz*.
5445 *
5446 * **-ENOENT** if symbol is not found.
5447 *
5448 * **-EPERM** if caller does not have permission to obtain kernel address.
5449 *
5450 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5451 * Description
5452 * Find vma of *task* that contains *addr*, call *callback_fn*
5453 * function with *task*, *vma*, and *callback_ctx*.
5454 * The *callback_fn* should be a static function and
5455 * the *callback_ctx* should be a pointer to the stack.
5456 * The *flags* is used to control certain aspects of the helper.
5457 * Currently, the *flags* must be 0.
5458 *
5459 * The expected callback signature is
5460 *
5461 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5462 *
5463 * Return
5464 * 0 on success.
5465 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5466 * **-EBUSY** if failed to try lock mmap_lock.
5467 * **-EINVAL** for invalid **flags**.
5468 *
5469 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5470 * Description
5471 * For **nr_loops**, call **callback_fn** function
5472 * with **callback_ctx** as the context parameter.
5473 * The **callback_fn** should be a static function and
5474 * the **callback_ctx** should be a pointer to the stack.
5475 * The **flags** is used to control certain aspects of the helper.
5476 * Currently, the **flags** must be 0. Currently, nr_loops is
5477 * limited to 1 << 23 (~8 million) loops.
5478 *
5479 * long (\*callback_fn)(u64 index, void \*ctx);
5480 *
5481 * where **index** is the current index in the loop. The index
5482 * is zero-indexed.
5483 *
5484 * If **callback_fn** returns 0, the helper will continue to the next
5485 * loop. If return value is 1, the helper will skip the rest of
5486 * the loops and return. Other return values are not used now,
5487 * and will be rejected by the verifier.
5488 *
5489 * Return
5490 * The number of loops performed, **-EINVAL** for invalid **flags**,
5491 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5492 *
5493 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5494 * Description
5495 * Do strncmp() between **s1** and **s2**. **s1** doesn't need
5496 * to be null-terminated and **s1_sz** is the maximum storage
5497 * size of **s1**. **s2** must be a read-only string.
5498 * Return
5499 * An integer less than, equal to, or greater than zero
5500 * if the first **s1_sz** bytes of **s1** is found to be
5501 * less than, to match, or be greater than **s2**.
5502 *
5503 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5504 * Description
5505 * Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5506 * returned in **value**.
5507 *
5508 * Return
5509 * 0 on success.
5510 * **-EINVAL** if n >= argument register count of traced function.
5511 *
5512 * long bpf_get_func_ret(void *ctx, u64 *value)
5513 * Description
5514 * Get return value of the traced function (for tracing programs)
5515 * in **value**.
5516 *
5517 * Return
5518 * 0 on success.
5519 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5520 *
5521 * long bpf_get_func_arg_cnt(void *ctx)
5522 * Description
5523 * Get number of registers of the traced function (for tracing programs) where
5524 * function arguments are stored in these registers.
5525 *
5526 * Return
5527 * The number of argument registers of the traced function.
5528 *
5529 * int bpf_get_retval(void)
5530 * Description
5531 * Get the BPF program's return value that will be returned to the upper layers.
5532 *
5533 * This helper is currently supported by cgroup programs and only by the hooks
5534 * where BPF program's return value is returned to the userspace via errno.
5535 * Return
5536 * The BPF program's return value.
5537 *
5538 * int bpf_set_retval(int retval)
5539 * Description
5540 * Set the BPF program's return value that will be returned to the upper layers.
5541 *
5542 * This helper is currently supported by cgroup programs and only by the hooks
5543 * where BPF program's return value is returned to the userspace via errno.
5544 *
5545 * Note that there is the following corner case where the program exports an error
5546 * via bpf_set_retval but signals success via 'return 1':
5547 *
5548 * bpf_set_retval(-EPERM);
5549 * return 1;
5550 *
5551 * In this case, the BPF program's return value will use helper's -EPERM. This
5552 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5553 *
5554 * Return
5555 * 0 on success, or a negative error in case of failure.
5556 *
5557 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5558 * Description
5559 * Get the total size of a given xdp buff (linear and paged area)
5560 * Return
5561 * The total size of a given xdp buffer.
5562 *
5563 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5564 * Description
5565 * This helper is provided as an easy way to load data from a
5566 * xdp buffer. It can be used to load *len* bytes from *offset* from
5567 * the frame associated to *xdp_md*, into the buffer pointed by
5568 * *buf*.
5569 * Return
5570 * 0 on success, or a negative error in case of failure.
5571 *
5572 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5573 * Description
5574 * Store *len* bytes from buffer *buf* into the frame
5575 * associated to *xdp_md*, at *offset*.
5576 * Return
5577 * 0 on success, or a negative error in case of failure.
5578 *
5579 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5580 * Description
5581 * Read *size* bytes from user space address *user_ptr* in *tsk*'s
5582 * address space, and stores the data in *dst*. *flags* is not
5583 * used yet and is provided for future extensibility. This helper
5584 * can only be used by sleepable programs.
5585 * Return
5586 * 0 on success, or a negative error in case of failure. On error
5587 * *dst* buffer is zeroed out.
5588 *
5589 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5590 * Description
5591 * Change the __sk_buff->tstamp_type to *tstamp_type*
5592 * and set *tstamp* to the __sk_buff->tstamp together.
5593 *
5594 * If there is no need to change the __sk_buff->tstamp_type,
5595 * the tstamp value can be directly written to __sk_buff->tstamp
5596 * instead.
5597 *
5598 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5599 * will be kept during bpf_redirect_*(). A non zero
5600 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5601 * *tstamp_type*.
5602 *
5603 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5604 * with a zero *tstamp*.
5605 *
5606 * Only IPv4 and IPv6 skb->protocol are supported.
5607 *
5608 * This function is most useful when it needs to set a
5609 * mono delivery time to __sk_buff->tstamp and then
5610 * bpf_redirect_*() to the egress of an iface. For example,
5611 * changing the (rcv) timestamp in __sk_buff->tstamp at
5612 * ingress to a mono delivery time and then bpf_redirect_*()
5613 * to sch_fq@phy-dev.
5614 * Return
5615 * 0 on success.
5616 * **-EINVAL** for invalid input
5617 * **-EOPNOTSUPP** for unsupported protocol
5618 *
5619 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5620 * Description
5621 * Returns a calculated IMA hash of the *file*.
5622 * If the hash is larger than *size*, then only *size*
5623 * bytes will be copied to *dst*
5624 * Return
5625 * The **hash_algo** is returned on success,
5626 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if
5627 * invalid arguments are passed.
5628 *
5629 * void *bpf_kptr_xchg(void *dst, void *ptr)
5630 * Description
5631 * Exchange kptr at pointer *dst* with *ptr*, and return the old value.
5632 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise
5633 * it must be a referenced pointer which will be released when this helper
5634 * is called.
5635 * Return
5636 * The old value of kptr (which can be NULL). The returned pointer
5637 * if not NULL, is a reference which must be released using its
5638 * corresponding release function, or moved into a BPF map before
5639 * program exit.
5640 *
5641 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5642 * Description
5643 * Perform a lookup in *percpu map* for an entry associated to
5644 * *key* on *cpu*.
5645 * Return
5646 * Map value associated to *key* on *cpu*, or **NULL** if no entry
5647 * was found or *cpu* is invalid.
5648 *
5649 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5650 * Description
5651 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5652 * Return
5653 * *sk* if casting is valid, or **NULL** otherwise.
5654 *
5655 * long bpf_dynptr_from_mem(void *data, u64 size, u64 flags, struct bpf_dynptr *ptr)
5656 * Description
5657 * Get a dynptr to local memory *data*.
5658 *
5659 * *data* must be a ptr to a map value.
5660 * The maximum *size* supported is DYNPTR_MAX_SIZE.
5661 * *flags* is currently unused.
5662 * Return
5663 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5664 * -EINVAL if flags is not 0.
5665 *
5666 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5667 * Description
5668 * Reserve *size* bytes of payload in a ring buffer *ringbuf*
5669 * through the dynptr interface. *flags* must be 0.
5670 *
5671 * Please note that a corresponding bpf_ringbuf_submit_dynptr or
5672 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5673 * reservation fails. This is enforced by the verifier.
5674 * Return
5675 * 0 on success, or a negative error in case of failure.
5676 *
5677 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5678 * Description
5679 * Submit reserved ring buffer sample, pointed to by *data*,
5680 * through the dynptr interface. This is a no-op if the dynptr is
5681 * invalid/null.
5682 *
5683 * For more information on *flags*, please see
5684 * 'bpf_ringbuf_submit'.
5685 * Return
5686 * Nothing. Always succeeds.
5687 *
5688 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5689 * Description
5690 * Discard reserved ring buffer sample through the dynptr
5691 * interface. This is a no-op if the dynptr is invalid/null.
5692 *
5693 * For more information on *flags*, please see
5694 * 'bpf_ringbuf_discard'.
5695 * Return
5696 * Nothing. Always succeeds.
5697 *
5698 * long bpf_dynptr_read(void *dst, u64 len, const struct bpf_dynptr *src, u64 offset, u64 flags)
5699 * Description
5700 * Read *len* bytes from *src* into *dst*, starting from *offset*
5701 * into *src*.
5702 * *flags* is currently unused.
5703 * Return
5704 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5705 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5706 * *flags* is not 0.
5707 *
5708 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u64 offset, void *src, u64 len, u64 flags)
5709 * Description
5710 * Write *len* bytes from *src* into *dst*, starting from *offset*
5711 * into *dst*.
5712 *
5713 * *flags* must be 0 except for skb-type dynptrs.
5714 *
5715 * For skb-type dynptrs:
5716 * * All data slices of the dynptr are automatically
5717 * invalidated after **bpf_dynptr_write**\ (). This is
5718 * because writing may pull the skb and change the
5719 * underlying packet buffer.
5720 *
5721 * * For *flags*, please see the flags accepted by
5722 * **bpf_skb_store_bytes**\ ().
5723 * Return
5724 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5725 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5726 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5727 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5728 *
5729 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u64 offset, u64 len)
5730 * Description
5731 * Get a pointer to the underlying dynptr data.
5732 *
5733 * *len* must be a statically known value. The returned data slice
5734 * is invalidated whenever the dynptr is invalidated.
5735 *
5736 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5737 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5738 * Return
5739 * Pointer to the underlying dynptr data, NULL if the dynptr is
5740 * read-only, if the dynptr is invalid, or if the offset and length
5741 * is out of bounds.
5742 *
5743 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5744 * Description
5745 * Try to issue a SYN cookie for the packet with corresponding
5746 * IPv4/TCP headers, *iph* and *th*, without depending on a
5747 * listening socket.
5748 *
5749 * *iph* points to the IPv4 header.
5750 *
5751 * *th* points to the start of the TCP header, while *th_len*
5752 * contains the length of the TCP header (at least
5753 * **sizeof**\ (**struct tcphdr**)).
5754 * Return
5755 * On success, lower 32 bits hold the generated SYN cookie in
5756 * followed by 16 bits which hold the MSS value for that cookie,
5757 * and the top 16 bits are unused.
5758 *
5759 * On failure, the returned value is one of the following:
5760 *
5761 * **-EINVAL** if *th_len* is invalid.
5762 *
5763 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5764 * Description
5765 * Try to issue a SYN cookie for the packet with corresponding
5766 * IPv6/TCP headers, *iph* and *th*, without depending on a
5767 * listening socket.
5768 *
5769 * *iph* points to the IPv6 header.
5770 *
5771 * *th* points to the start of the TCP header, while *th_len*
5772 * contains the length of the TCP header (at least
5773 * **sizeof**\ (**struct tcphdr**)).
5774 * Return
5775 * On success, lower 32 bits hold the generated SYN cookie in
5776 * followed by 16 bits which hold the MSS value for that cookie,
5777 * and the top 16 bits are unused.
5778 *
5779 * On failure, the returned value is one of the following:
5780 *
5781 * **-EINVAL** if *th_len* is invalid.
5782 *
5783 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5784 *
5785 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5786 * Description
5787 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5788 * without depending on a listening socket.
5789 *
5790 * *iph* points to the IPv4 header.
5791 *
5792 * *th* points to the TCP header.
5793 * Return
5794 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5795 *
5796 * On failure, the returned value is one of the following:
5797 *
5798 * **-EACCES** if the SYN cookie is not valid.
5799 *
5800 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5801 * Description
5802 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5803 * without depending on a listening socket.
5804 *
5805 * *iph* points to the IPv6 header.
5806 *
5807 * *th* points to the TCP header.
5808 * Return
5809 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5810 *
5811 * On failure, the returned value is one of the following:
5812 *
5813 * **-EACCES** if the SYN cookie is not valid.
5814 *
5815 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5816 *
5817 * u64 bpf_ktime_get_tai_ns(void)
5818 * Description
5819 * A nonsettable system-wide clock derived from wall-clock time but
5820 * ignoring leap seconds. This clock does not experience
5821 * discontinuities and backwards jumps caused by NTP inserting leap
5822 * seconds as CLOCK_REALTIME does.
5823 *
5824 * See: **clock_gettime**\ (**CLOCK_TAI**)
5825 * Return
5826 * Current *ktime*.
5827 *
5828 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5829 * Description
5830 * Drain samples from the specified user ring buffer, and invoke
5831 * the provided callback for each such sample:
5832 *
5833 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5834 *
5835 * If **callback_fn** returns 0, the helper will continue to try
5836 * and drain the next sample, up to a maximum of
5837 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5838 * the helper will skip the rest of the samples and return. Other
5839 * return values are not used now, and will be rejected by the
5840 * verifier.
5841 * Return
5842 * The number of drained samples if no error was encountered while
5843 * draining samples, or 0 if no samples were present in the ring
5844 * buffer. If a user-space producer was epoll-waiting on this map,
5845 * and at least one sample was drained, they will receive an event
5846 * notification notifying them of available space in the ring
5847 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5848 * function, no wakeup notification will be sent. If the
5849 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5850 * be sent even if no sample was drained.
5851 *
5852 * On failure, the returned value is one of the following:
5853 *
5854 * **-EBUSY** if the ring buffer is contended, and another calling
5855 * context was concurrently draining the ring buffer.
5856 *
5857 * **-EINVAL** if user-space is not properly tracking the ring
5858 * buffer due to the producer position not being aligned to 8
5859 * bytes, a sample not being aligned to 8 bytes, or the producer
5860 * position not matching the advertised length of a sample.
5861 *
5862 * **-E2BIG** if user-space has tried to publish a sample which is
5863 * larger than the size of the ring buffer, or which cannot fit
5864 * within a struct bpf_dynptr.
5865 *
5866 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5867 * Description
5868 * Get a bpf_local_storage from the *cgroup*.
5869 *
5870 * Logically, it could be thought of as getting the value from
5871 * a *map* with *cgroup* as the **key**. From this
5872 * perspective, the usage is not much different from
5873 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5874 * helper enforces the key must be a cgroup struct and the map must also
5875 * be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5876 *
5877 * In reality, the local-storage value is embedded directly inside of the
5878 * *cgroup* object itself, rather than being located in the
5879 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5880 * queried for some *map* on a *cgroup* object, the kernel will perform an
5881 * O(n) iteration over all of the live local-storage values for that
5882 * *cgroup* object until the local-storage value for the *map* is found.
5883 *
5884 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5885 * used such that a new bpf_local_storage will be
5886 * created if one does not exist. *value* can be used
5887 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5888 * the initial value of a bpf_local_storage. If *value* is
5889 * **NULL**, the new bpf_local_storage will be zero initialized.
5890 * Return
5891 * A bpf_local_storage pointer is returned on success.
5892 *
5893 * **NULL** if not found or there was an error in adding
5894 * a new bpf_local_storage.
5895 *
5896 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5897 * Description
5898 * Delete a bpf_local_storage from a *cgroup*.
5899 * Return
5900 * 0 on success.
5901 *
5902 * **-ENOENT** if the bpf_local_storage cannot be found.
5903 */
5904#define ___BPF_FUNC_MAPPER(FN, ctx...) \
5905 FN(unspec, 0, ##ctx) \
5906 FN(map_lookup_elem, 1, ##ctx) \
5907 FN(map_update_elem, 2, ##ctx) \
5908 FN(map_delete_elem, 3, ##ctx) \
5909 FN(probe_read, 4, ##ctx) \
5910 FN(ktime_get_ns, 5, ##ctx) \
5911 FN(trace_printk, 6, ##ctx) \
5912 FN(get_prandom_u32, 7, ##ctx) \
5913 FN(get_smp_processor_id, 8, ##ctx) \
5914 FN(skb_store_bytes, 9, ##ctx) \
5915 FN(l3_csum_replace, 10, ##ctx) \
5916 FN(l4_csum_replace, 11, ##ctx) \
5917 FN(tail_call, 12, ##ctx) \
5918 FN(clone_redirect, 13, ##ctx) \
5919 FN(get_current_pid_tgid, 14, ##ctx) \
5920 FN(get_current_uid_gid, 15, ##ctx) \
5921 FN(get_current_comm, 16, ##ctx) \
5922 FN(get_cgroup_classid, 17, ##ctx) \
5923 FN(skb_vlan_push, 18, ##ctx) \
5924 FN(skb_vlan_pop, 19, ##ctx) \
5925 FN(skb_get_tunnel_key, 20, ##ctx) \
5926 FN(skb_set_tunnel_key, 21, ##ctx) \
5927 FN(perf_event_read, 22, ##ctx) \
5928 FN(redirect, 23, ##ctx) \
5929 FN(get_route_realm, 24, ##ctx) \
5930 FN(perf_event_output, 25, ##ctx) \
5931 FN(skb_load_bytes, 26, ##ctx) \
5932 FN(get_stackid, 27, ##ctx) \
5933 FN(csum_diff, 28, ##ctx) \
5934 FN(skb_get_tunnel_opt, 29, ##ctx) \
5935 FN(skb_set_tunnel_opt, 30, ##ctx) \
5936 FN(skb_change_proto, 31, ##ctx) \
5937 FN(skb_change_type, 32, ##ctx) \
5938 FN(skb_under_cgroup, 33, ##ctx) \
5939 FN(get_hash_recalc, 34, ##ctx) \
5940 FN(get_current_task, 35, ##ctx) \
5941 FN(probe_write_user, 36, ##ctx) \
5942 FN(current_task_under_cgroup, 37, ##ctx) \
5943 FN(skb_change_tail, 38, ##ctx) \
5944 FN(skb_pull_data, 39, ##ctx) \
5945 FN(csum_update, 40, ##ctx) \
5946 FN(set_hash_invalid, 41, ##ctx) \
5947 FN(get_numa_node_id, 42, ##ctx) \
5948 FN(skb_change_head, 43, ##ctx) \
5949 FN(xdp_adjust_head, 44, ##ctx) \
5950 FN(probe_read_str, 45, ##ctx) \
5951 FN(get_socket_cookie, 46, ##ctx) \
5952 FN(get_socket_uid, 47, ##ctx) \
5953 FN(set_hash, 48, ##ctx) \
5954 FN(setsockopt, 49, ##ctx) \
5955 FN(skb_adjust_room, 50, ##ctx) \
5956 FN(redirect_map, 51, ##ctx) \
5957 FN(sk_redirect_map, 52, ##ctx) \
5958 FN(sock_map_update, 53, ##ctx) \
5959 FN(xdp_adjust_meta, 54, ##ctx) \
5960 FN(perf_event_read_value, 55, ##ctx) \
5961 FN(perf_prog_read_value, 56, ##ctx) \
5962 FN(getsockopt, 57, ##ctx) \
5963 FN(override_return, 58, ##ctx) \
5964 FN(sock_ops_cb_flags_set, 59, ##ctx) \
5965 FN(msg_redirect_map, 60, ##ctx) \
5966 FN(msg_apply_bytes, 61, ##ctx) \
5967 FN(msg_cork_bytes, 62, ##ctx) \
5968 FN(msg_pull_data, 63, ##ctx) \
5969 FN(bind, 64, ##ctx) \
5970 FN(xdp_adjust_tail, 65, ##ctx) \
5971 FN(skb_get_xfrm_state, 66, ##ctx) \
5972 FN(get_stack, 67, ##ctx) \
5973 FN(skb_load_bytes_relative, 68, ##ctx) \
5974 FN(fib_lookup, 69, ##ctx) \
5975 FN(sock_hash_update, 70, ##ctx) \
5976 FN(msg_redirect_hash, 71, ##ctx) \
5977 FN(sk_redirect_hash, 72, ##ctx) \
5978 FN(lwt_push_encap, 73, ##ctx) \
5979 FN(lwt_seg6_store_bytes, 74, ##ctx) \
5980 FN(lwt_seg6_adjust_srh, 75, ##ctx) \
5981 FN(lwt_seg6_action, 76, ##ctx) \
5982 FN(rc_repeat, 77, ##ctx) \
5983 FN(rc_keydown, 78, ##ctx) \
5984 FN(skb_cgroup_id, 79, ##ctx) \
5985 FN(get_current_cgroup_id, 80, ##ctx) \
5986 FN(get_local_storage, 81, ##ctx) \
5987 FN(sk_select_reuseport, 82, ##ctx) \
5988 FN(skb_ancestor_cgroup_id, 83, ##ctx) \
5989 FN(sk_lookup_tcp, 84, ##ctx) \
5990 FN(sk_lookup_udp, 85, ##ctx) \
5991 FN(sk_release, 86, ##ctx) \
5992 FN(map_push_elem, 87, ##ctx) \
5993 FN(map_pop_elem, 88, ##ctx) \
5994 FN(map_peek_elem, 89, ##ctx) \
5995 FN(msg_push_data, 90, ##ctx) \
5996 FN(msg_pop_data, 91, ##ctx) \
5997 FN(rc_pointer_rel, 92, ##ctx) \
5998 FN(spin_lock, 93, ##ctx) \
5999 FN(spin_unlock, 94, ##ctx) \
6000 FN(sk_fullsock, 95, ##ctx) \
6001 FN(tcp_sock, 96, ##ctx) \
6002 FN(skb_ecn_set_ce, 97, ##ctx) \
6003 FN(get_listener_sock, 98, ##ctx) \
6004 FN(skc_lookup_tcp, 99, ##ctx) \
6005 FN(tcp_check_syncookie, 100, ##ctx) \
6006 FN(sysctl_get_name, 101, ##ctx) \
6007 FN(sysctl_get_current_value, 102, ##ctx) \
6008 FN(sysctl_get_new_value, 103, ##ctx) \
6009 FN(sysctl_set_new_value, 104, ##ctx) \
6010 FN(strtol, 105, ##ctx) \
6011 FN(strtoul, 106, ##ctx) \
6012 FN(sk_storage_get, 107, ##ctx) \
6013 FN(sk_storage_delete, 108, ##ctx) \
6014 FN(send_signal, 109, ##ctx) \
6015 FN(tcp_gen_syncookie, 110, ##ctx) \
6016 FN(skb_output, 111, ##ctx) \
6017 FN(probe_read_user, 112, ##ctx) \
6018 FN(probe_read_kernel, 113, ##ctx) \
6019 FN(probe_read_user_str, 114, ##ctx) \
6020 FN(probe_read_kernel_str, 115, ##ctx) \
6021 FN(tcp_send_ack, 116, ##ctx) \
6022 FN(send_signal_thread, 117, ##ctx) \
6023 FN(jiffies64, 118, ##ctx) \
6024 FN(read_branch_records, 119, ##ctx) \
6025 FN(get_ns_current_pid_tgid, 120, ##ctx) \
6026 FN(xdp_output, 121, ##ctx) \
6027 FN(get_netns_cookie, 122, ##ctx) \
6028 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \
6029 FN(sk_assign, 124, ##ctx) \
6030 FN(ktime_get_boot_ns, 125, ##ctx) \
6031 FN(seq_printf, 126, ##ctx) \
6032 FN(seq_write, 127, ##ctx) \
6033 FN(sk_cgroup_id, 128, ##ctx) \
6034 FN(sk_ancestor_cgroup_id, 129, ##ctx) \
6035 FN(ringbuf_output, 130, ##ctx) \
6036 FN(ringbuf_reserve, 131, ##ctx) \
6037 FN(ringbuf_submit, 132, ##ctx) \
6038 FN(ringbuf_discard, 133, ##ctx) \
6039 FN(ringbuf_query, 134, ##ctx) \
6040 FN(csum_level, 135, ##ctx) \
6041 FN(skc_to_tcp6_sock, 136, ##ctx) \
6042 FN(skc_to_tcp_sock, 137, ##ctx) \
6043 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \
6044 FN(skc_to_tcp_request_sock, 139, ##ctx) \
6045 FN(skc_to_udp6_sock, 140, ##ctx) \
6046 FN(get_task_stack, 141, ##ctx) \
6047 FN(load_hdr_opt, 142, ##ctx) \
6048 FN(store_hdr_opt, 143, ##ctx) \
6049 FN(reserve_hdr_opt, 144, ##ctx) \
6050 FN(inode_storage_get, 145, ##ctx) \
6051 FN(inode_storage_delete, 146, ##ctx) \
6052 FN(d_path, 147, ##ctx) \
6053 FN(copy_from_user, 148, ##ctx) \
6054 FN(snprintf_btf, 149, ##ctx) \
6055 FN(seq_printf_btf, 150, ##ctx) \
6056 FN(skb_cgroup_classid, 151, ##ctx) \
6057 FN(redirect_neigh, 152, ##ctx) \
6058 FN(per_cpu_ptr, 153, ##ctx) \
6059 FN(this_cpu_ptr, 154, ##ctx) \
6060 FN(redirect_peer, 155, ##ctx) \
6061 FN(task_storage_get, 156, ##ctx) \
6062 FN(task_storage_delete, 157, ##ctx) \
6063 FN(get_current_task_btf, 158, ##ctx) \
6064 FN(bprm_opts_set, 159, ##ctx) \
6065 FN(ktime_get_coarse_ns, 160, ##ctx) \
6066 FN(ima_inode_hash, 161, ##ctx) \
6067 FN(sock_from_file, 162, ##ctx) \
6068 FN(check_mtu, 163, ##ctx) \
6069 FN(for_each_map_elem, 164, ##ctx) \
6070 FN(snprintf, 165, ##ctx) \
6071 FN(sys_bpf, 166, ##ctx) \
6072 FN(btf_find_by_name_kind, 167, ##ctx) \
6073 FN(sys_close, 168, ##ctx) \
6074 FN(timer_init, 169, ##ctx) \
6075 FN(timer_set_callback, 170, ##ctx) \
6076 FN(timer_start, 171, ##ctx) \
6077 FN(timer_cancel, 172, ##ctx) \
6078 FN(get_func_ip, 173, ##ctx) \
6079 FN(get_attach_cookie, 174, ##ctx) \
6080 FN(task_pt_regs, 175, ##ctx) \
6081 FN(get_branch_snapshot, 176, ##ctx) \
6082 FN(trace_vprintk, 177, ##ctx) \
6083 FN(skc_to_unix_sock, 178, ##ctx) \
6084 FN(kallsyms_lookup_name, 179, ##ctx) \
6085 FN(find_vma, 180, ##ctx) \
6086 FN(loop, 181, ##ctx) \
6087 FN(strncmp, 182, ##ctx) \
6088 FN(get_func_arg, 183, ##ctx) \
6089 FN(get_func_ret, 184, ##ctx) \
6090 FN(get_func_arg_cnt, 185, ##ctx) \
6091 FN(get_retval, 186, ##ctx) \
6092 FN(set_retval, 187, ##ctx) \
6093 FN(xdp_get_buff_len, 188, ##ctx) \
6094 FN(xdp_load_bytes, 189, ##ctx) \
6095 FN(xdp_store_bytes, 190, ##ctx) \
6096 FN(copy_from_user_task, 191, ##ctx) \
6097 FN(skb_set_tstamp, 192, ##ctx) \
6098 FN(ima_file_hash, 193, ##ctx) \
6099 FN(kptr_xchg, 194, ##ctx) \
6100 FN(map_lookup_percpu_elem, 195, ##ctx) \
6101 FN(skc_to_mptcp_sock, 196, ##ctx) \
6102 FN(dynptr_from_mem, 197, ##ctx) \
6103 FN(ringbuf_reserve_dynptr, 198, ##ctx) \
6104 FN(ringbuf_submit_dynptr, 199, ##ctx) \
6105 FN(ringbuf_discard_dynptr, 200, ##ctx) \
6106 FN(dynptr_read, 201, ##ctx) \
6107 FN(dynptr_write, 202, ##ctx) \
6108 FN(dynptr_data, 203, ##ctx) \
6109 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \
6110 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \
6111 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \
6112 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \
6113 FN(ktime_get_tai_ns, 208, ##ctx) \
6114 FN(user_ringbuf_drain, 209, ##ctx) \
6115 FN(cgrp_storage_get, 210, ##ctx) \
6116 FN(cgrp_storage_delete, 211, ##ctx) \
6117 /* This helper list is effectively frozen. If you are trying to \
6118 * add a new helper, you should add a kfunc instead which has \
6119 * less stability guarantees. See Documentation/bpf/kfuncs.rst \
6120 */
6121
6122/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
6123 * know or care about integer value that is now passed as second argument
6124 */
6125#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
6126#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
6127
6128/* integer value in 'imm' field of BPF_CALL instruction selects which helper
6129 * function eBPF program intends to call
6130 */
6131#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
6132enum bpf_func_id {
6133 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
6134 __BPF_FUNC_MAX_ID,
6135};
6136#undef __BPF_ENUM_FN
6137
6138/* All flags used by eBPF helper functions, placed here. */
6139
6140/* BPF_FUNC_skb_store_bytes flags. */
6141enum {
6142 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
6143 BPF_F_INVALIDATE_HASH = (1ULL << 1),
6144};
6145
6146/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
6147 * First 4 bits are for passing the header field size.
6148 */
6149enum {
6150 BPF_F_HDR_FIELD_MASK = 0xfULL,
6151};
6152
6153/* BPF_FUNC_l4_csum_replace flags. */
6154enum {
6155 BPF_F_PSEUDO_HDR = (1ULL << 4),
6156 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
6157 BPF_F_MARK_ENFORCE = (1ULL << 6),
6158 BPF_F_IPV6 = (1ULL << 7),
6159};
6160
6161/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
6162enum {
6163 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
6164};
6165
6166/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
6167enum {
6168 BPF_F_SKIP_FIELD_MASK = 0xffULL,
6169 BPF_F_USER_STACK = (1ULL << 8),
6170/* flags used by BPF_FUNC_get_stackid only. */
6171 BPF_F_FAST_STACK_CMP = (1ULL << 9),
6172 BPF_F_REUSE_STACKID = (1ULL << 10),
6173/* flags used by BPF_FUNC_get_stack only. */
6174 BPF_F_USER_BUILD_ID = (1ULL << 11),
6175};
6176
6177/* BPF_FUNC_skb_set_tunnel_key flags. */
6178enum {
6179 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
6180 BPF_F_DONT_FRAGMENT = (1ULL << 2),
6181 BPF_F_SEQ_NUMBER = (1ULL << 3),
6182 BPF_F_NO_TUNNEL_KEY = (1ULL << 4),
6183};
6184
6185/* BPF_FUNC_skb_get_tunnel_key flags. */
6186enum {
6187 BPF_F_TUNINFO_FLAGS = (1ULL << 4),
6188};
6189
6190/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
6191 * BPF_FUNC_perf_event_read_value flags.
6192 */
6193enum {
6194 BPF_F_INDEX_MASK = 0xffffffffULL,
6195 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
6196/* BPF_FUNC_perf_event_output for sk_buff input context. */
6197 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
6198};
6199
6200/* Current network namespace */
6201enum {
6202 BPF_F_CURRENT_NETNS = (-1L),
6203};
6204
6205/* BPF_FUNC_csum_level level values. */
6206enum {
6207 BPF_CSUM_LEVEL_QUERY,
6208 BPF_CSUM_LEVEL_INC,
6209 BPF_CSUM_LEVEL_DEC,
6210 BPF_CSUM_LEVEL_RESET,
6211};
6212
6213/* BPF_FUNC_skb_adjust_room flags. */
6214enum {
6215 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
6216 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
6217 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
6218 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
6219 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
6220 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
6221 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
6222 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7),
6223 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8),
6224};
6225
6226enum {
6227 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
6228 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
6229};
6230
6231#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
6232 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
6233 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6234
6235/* BPF_FUNC_sysctl_get_name flags. */
6236enum {
6237 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
6238};
6239
6240/* BPF_FUNC_<kernel_obj>_storage_get flags */
6241enum {
6242 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
6243 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
6244 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
6245 */
6246 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
6247};
6248
6249/* BPF_FUNC_read_branch_records flags. */
6250enum {
6251 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
6252};
6253
6254/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
6255 * BPF_FUNC_bpf_ringbuf_output flags.
6256 */
6257enum {
6258 BPF_RB_NO_WAKEUP = (1ULL << 0),
6259 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
6260};
6261
6262/* BPF_FUNC_bpf_ringbuf_query flags */
6263enum {
6264 BPF_RB_AVAIL_DATA = 0,
6265 BPF_RB_RING_SIZE = 1,
6266 BPF_RB_CONS_POS = 2,
6267 BPF_RB_PROD_POS = 3,
6268 BPF_RB_OVERWRITE_POS = 4,
6269};
6270
6271/* BPF ring buffer constants */
6272enum {
6273 BPF_RINGBUF_BUSY_BIT = (1U << 31),
6274 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
6275 BPF_RINGBUF_HDR_SZ = 8,
6276};
6277
6278/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6279enum {
6280 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
6281 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
6282};
6283
6284/* Mode for BPF_FUNC_skb_adjust_room helper. */
6285enum bpf_adj_room_mode {
6286 BPF_ADJ_ROOM_NET,
6287 BPF_ADJ_ROOM_MAC,
6288};
6289
6290/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6291enum bpf_hdr_start_off {
6292 BPF_HDR_START_MAC,
6293 BPF_HDR_START_NET,
6294};
6295
6296/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6297enum bpf_lwt_encap_mode {
6298 BPF_LWT_ENCAP_SEG6,
6299 BPF_LWT_ENCAP_SEG6_INLINE,
6300 BPF_LWT_ENCAP_IP,
6301};
6302
6303/* Flags for bpf_bprm_opts_set helper */
6304enum {
6305 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
6306};
6307
6308/* Flags for bpf_redirect and bpf_redirect_map helpers */
6309enum {
6310 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */
6311 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */
6312 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */
6313#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
6314};
6315
6316#define __bpf_md_ptr(type, name) \
6317union { \
6318 type name; \
6319 __u64 :64; \
6320} __attribute__((aligned(8)))
6321
6322/* The enum used in skb->tstamp_type. It specifies the clock type
6323 * of the time stored in the skb->tstamp.
6324 */
6325enum {
6326 BPF_SKB_TSTAMP_UNSPEC = 0, /* DEPRECATED */
6327 BPF_SKB_TSTAMP_DELIVERY_MONO = 1, /* DEPRECATED */
6328 BPF_SKB_CLOCK_REALTIME = 0,
6329 BPF_SKB_CLOCK_MONOTONIC = 1,
6330 BPF_SKB_CLOCK_TAI = 2,
6331 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,
6332 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.
6333 */
6334};
6335
6336/* user accessible mirror of in-kernel sk_buff.
6337 * new fields can only be added to the end of this structure
6338 */
6339struct __sk_buff {
6340 __u32 len;
6341 __u32 pkt_type;
6342 __u32 mark;
6343 __u32 queue_mapping;
6344 __u32 protocol;
6345 __u32 vlan_present;
6346 __u32 vlan_tci;
6347 __u32 vlan_proto;
6348 __u32 priority;
6349 __u32 ingress_ifindex;
6350 __u32 ifindex;
6351 __u32 tc_index;
6352 __u32 cb[5];
6353 __u32 hash;
6354 __u32 tc_classid;
6355 __u32 data;
6356 __u32 data_end;
6357 __u32 napi_id;
6358
6359 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6360 __u32 family;
6361 __u32 remote_ip4; /* Stored in network byte order */
6362 __u32 local_ip4; /* Stored in network byte order */
6363 __u32 remote_ip6[4]; /* Stored in network byte order */
6364 __u32 local_ip6[4]; /* Stored in network byte order */
6365 __u32 remote_port; /* Stored in network byte order */
6366 __u32 local_port; /* stored in host byte order */
6367 /* ... here. */
6368
6369 __u32 data_meta;
6370 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
6371 __u64 tstamp;
6372 __u32 wire_len;
6373 __u32 gso_segs;
6374 __bpf_md_ptr(struct bpf_sock *, sk);
6375 __u32 gso_size;
6376 __u8 tstamp_type;
6377 __u32 :24; /* Padding, future use. */
6378 __u64 hwtstamp;
6379};
6380
6381struct bpf_tunnel_key {
6382 __u32 tunnel_id;
6383 union {
6384 __u32 remote_ipv4;
6385 __u32 remote_ipv6[4];
6386 };
6387 __u8 tunnel_tos;
6388 __u8 tunnel_ttl;
6389 union {
6390 __u16 tunnel_ext; /* compat */
6391 __be16 tunnel_flags;
6392 };
6393 __u32 tunnel_label;
6394 union {
6395 __u32 local_ipv4;
6396 __u32 local_ipv6[4];
6397 };
6398};
6399
6400/* user accessible mirror of in-kernel xfrm_state.
6401 * new fields can only be added to the end of this structure
6402 */
6403struct bpf_xfrm_state {
6404 __u32 reqid;
6405 __u32 spi; /* Stored in network byte order */
6406 __u16 family;
6407 __u16 ext; /* Padding, future use. */
6408 union {
6409 __u32 remote_ipv4; /* Stored in network byte order */
6410 __u32 remote_ipv6[4]; /* Stored in network byte order */
6411 };
6412};
6413
6414/* Generic BPF return codes which all BPF program types may support.
6415 * The values are binary compatible with their TC_ACT_* counter-part to
6416 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6417 * programs.
6418 *
6419 * XDP is handled seprately, see XDP_*.
6420 */
6421enum bpf_ret_code {
6422 BPF_OK = 0,
6423 /* 1 reserved */
6424 BPF_DROP = 2,
6425 /* 3-6 reserved */
6426 BPF_REDIRECT = 7,
6427 /* >127 are reserved for prog type specific return codes.
6428 *
6429 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6430 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6431 * changed and should be routed based on its new L3 header.
6432 * (This is an L3 redirect, as opposed to L2 redirect
6433 * represented by BPF_REDIRECT above).
6434 */
6435 BPF_LWT_REROUTE = 128,
6436 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6437 * to indicate that no custom dissection was performed, and
6438 * fallback to standard dissector is requested.
6439 */
6440 BPF_FLOW_DISSECTOR_CONTINUE = 129,
6441};
6442
6443struct bpf_sock {
6444 __u32 bound_dev_if;
6445 __u32 family;
6446 __u32 type;
6447 __u32 protocol;
6448 __u32 mark;
6449 __u32 priority;
6450 /* IP address also allows 1 and 2 bytes access */
6451 __u32 src_ip4;
6452 __u32 src_ip6[4];
6453 __u32 src_port; /* host byte order */
6454 __be16 dst_port; /* network byte order */
6455 __u16 :16; /* zero padding */
6456 __u32 dst_ip4;
6457 __u32 dst_ip6[4];
6458 __u32 state;
6459 __s32 rx_queue_mapping;
6460};
6461
6462struct bpf_tcp_sock {
6463 __u32 snd_cwnd; /* Sending congestion window */
6464 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
6465 __u32 rtt_min;
6466 __u32 snd_ssthresh; /* Slow start size threshold */
6467 __u32 rcv_nxt; /* What we want to receive next */
6468 __u32 snd_nxt; /* Next sequence we send */
6469 __u32 snd_una; /* First byte we want an ack for */
6470 __u32 mss_cache; /* Cached effective mss, not including SACKS */
6471 __u32 ecn_flags; /* ECN status bits. */
6472 __u32 rate_delivered; /* saved rate sample: packets delivered */
6473 __u32 rate_interval_us; /* saved rate sample: time elapsed */
6474 __u32 packets_out; /* Packets which are "in flight" */
6475 __u32 retrans_out; /* Retransmitted packets out */
6476 __u32 total_retrans; /* Total retransmits for entire connection */
6477 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
6478 * total number of segments in.
6479 */
6480 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
6481 * total number of data segments in.
6482 */
6483 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
6484 * The total number of segments sent.
6485 */
6486 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
6487 * total number of data segments sent.
6488 */
6489 __u32 lost_out; /* Lost packets */
6490 __u32 sacked_out; /* SACK'd packets */
6491 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
6492 * sum(delta(rcv_nxt)), or how many bytes
6493 * were acked.
6494 */
6495 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
6496 * sum(delta(snd_una)), or how many bytes
6497 * were acked.
6498 */
6499 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
6500 * total number of DSACK blocks received
6501 */
6502 __u32 delivered; /* Total data packets delivered incl. rexmits */
6503 __u32 delivered_ce; /* Like the above but only ECE marked packets */
6504 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
6505};
6506
6507struct bpf_sock_tuple {
6508 union {
6509 struct {
6510 __be32 saddr;
6511 __be32 daddr;
6512 __be16 sport;
6513 __be16 dport;
6514 } ipv4;
6515 struct {
6516 __be32 saddr[4];
6517 __be32 daddr[4];
6518 __be16 sport;
6519 __be16 dport;
6520 } ipv6;
6521 };
6522};
6523
6524/* (Simplified) user return codes for tcx prog type.
6525 * A valid tcx program must return one of these defined values. All other
6526 * return codes are reserved for future use. Must remain compatible with
6527 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6528 * return codes are mapped to TCX_NEXT.
6529 */
6530enum tcx_action_base {
6531 TCX_NEXT = -1,
6532 TCX_PASS = 0,
6533 TCX_DROP = 2,
6534 TCX_REDIRECT = 7,
6535};
6536
6537struct bpf_xdp_sock {
6538 __u32 queue_id;
6539};
6540
6541#define XDP_PACKET_HEADROOM 256
6542
6543/* User return codes for XDP prog type.
6544 * A valid XDP program must return one of these defined values. All other
6545 * return codes are reserved for future use. Unknown return codes will
6546 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6547 */
6548enum xdp_action {
6549 XDP_ABORTED = 0,
6550 XDP_DROP,
6551 XDP_PASS,
6552 XDP_TX,
6553 XDP_REDIRECT,
6554};
6555
6556/* user accessible metadata for XDP packet hook
6557 * new fields must be added to the end of this structure
6558 */
6559struct xdp_md {
6560 __u32 data;
6561 __u32 data_end;
6562 __u32 data_meta;
6563 /* Below access go through struct xdp_rxq_info */
6564 __u32 ingress_ifindex; /* rxq->dev->ifindex */
6565 __u32 rx_queue_index; /* rxq->queue_index */
6566
6567 __u32 egress_ifindex; /* txq->dev->ifindex */
6568};
6569
6570/* DEVMAP map-value layout
6571 *
6572 * The struct data-layout of map-value is a configuration interface.
6573 * New members can only be added to the end of this structure.
6574 */
6575struct bpf_devmap_val {
6576 __u32 ifindex; /* device index */
6577 union {
6578 int fd; /* prog fd on map write */
6579 __u32 id; /* prog id on map read */
6580 } bpf_prog;
6581};
6582
6583/* CPUMAP map-value layout
6584 *
6585 * The struct data-layout of map-value is a configuration interface.
6586 * New members can only be added to the end of this structure.
6587 */
6588struct bpf_cpumap_val {
6589 __u32 qsize; /* queue size to remote target CPU */
6590 union {
6591 int fd; /* prog fd on map write */
6592 __u32 id; /* prog id on map read */
6593 } bpf_prog;
6594};
6595
6596enum sk_action {
6597 SK_DROP = 0,
6598 SK_PASS,
6599};
6600
6601/* user accessible metadata for SK_MSG packet hook, new fields must
6602 * be added to the end of this structure
6603 */
6604struct sk_msg_md {
6605 __bpf_md_ptr(void *, data);
6606 __bpf_md_ptr(void *, data_end);
6607
6608 __u32 family;
6609 __u32 remote_ip4; /* Stored in network byte order */
6610 __u32 local_ip4; /* Stored in network byte order */
6611 __u32 remote_ip6[4]; /* Stored in network byte order */
6612 __u32 local_ip6[4]; /* Stored in network byte order */
6613 __u32 remote_port; /* Stored in network byte order */
6614 __u32 local_port; /* stored in host byte order */
6615 __u32 size; /* Total size of sk_msg */
6616
6617 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6618};
6619
6620struct sk_reuseport_md {
6621 /*
6622 * Start of directly accessible data. It begins from
6623 * the tcp/udp header.
6624 */
6625 __bpf_md_ptr(void *, data);
6626 /* End of directly accessible data */
6627 __bpf_md_ptr(void *, data_end);
6628 /*
6629 * Total length of packet (starting from the tcp/udp header).
6630 * Note that the directly accessible bytes (data_end - data)
6631 * could be less than this "len". Those bytes could be
6632 * indirectly read by a helper "bpf_skb_load_bytes()".
6633 */
6634 __u32 len;
6635 /*
6636 * Eth protocol in the mac header (network byte order). e.g.
6637 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6638 */
6639 __u32 eth_protocol;
6640 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6641 __u32 bind_inany; /* Is sock bound to an INANY address? */
6642 __u32 hash; /* A hash of the packet 4 tuples */
6643 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6644 * new incoming connection request (e.g. selecting a listen sk for
6645 * the received SYN in the TCP case). reuse->sk is one of the sk
6646 * in the reuseport group. The bpf prog can use reuse->sk to learn
6647 * the local listening ip/port without looking into the skb.
6648 *
6649 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6650 * reuse->migrating_sk is the socket that needs to be migrated
6651 * to another listening socket. migrating_sk could be a fullsock
6652 * sk that is fully established or a reqsk that is in-the-middle
6653 * of 3-way handshake.
6654 */
6655 __bpf_md_ptr(struct bpf_sock *, sk);
6656 __bpf_md_ptr(struct bpf_sock *, migrating_sk);
6657};
6658
6659#define BPF_TAG_SIZE 8
6660
6661struct bpf_prog_info {
6662 __u32 type;
6663 __u32 id;
6664 __u8 tag[BPF_TAG_SIZE];
6665 __u32 jited_prog_len;
6666 __u32 xlated_prog_len;
6667 __aligned_u64 jited_prog_insns;
6668 __aligned_u64 xlated_prog_insns;
6669 __u64 load_time; /* ns since boottime */
6670 __u32 created_by_uid;
6671 __u32 nr_map_ids;
6672 __aligned_u64 map_ids;
6673 char name[BPF_OBJ_NAME_LEN];
6674 __u32 ifindex;
6675 __u32 gpl_compatible:1;
6676 __u32 :31; /* alignment pad */
6677 __u64 netns_dev;
6678 __u64 netns_ino;
6679 __u32 nr_jited_ksyms;
6680 __u32 nr_jited_func_lens;
6681 __aligned_u64 jited_ksyms;
6682 __aligned_u64 jited_func_lens;
6683 __u32 btf_id;
6684 __u32 func_info_rec_size;
6685 __aligned_u64 func_info;
6686 __u32 nr_func_info;
6687 __u32 nr_line_info;
6688 __aligned_u64 line_info;
6689 __aligned_u64 jited_line_info;
6690 __u32 nr_jited_line_info;
6691 __u32 line_info_rec_size;
6692 __u32 jited_line_info_rec_size;
6693 __u32 nr_prog_tags;
6694 __aligned_u64 prog_tags;
6695 __u64 run_time_ns;
6696 __u64 run_cnt;
6697 __u64 recursion_misses;
6698 __u32 verified_insns;
6699 __u32 attach_btf_obj_id;
6700 __u32 attach_btf_id;
6701} __attribute__((aligned(8)));
6702
6703struct bpf_map_info {
6704 __u32 type;
6705 __u32 id;
6706 __u32 key_size;
6707 __u32 value_size;
6708 __u32 max_entries;
6709 __u32 map_flags;
6710 char name[BPF_OBJ_NAME_LEN];
6711 __u32 ifindex;
6712 __u32 btf_vmlinux_value_type_id;
6713 __u64 netns_dev;
6714 __u64 netns_ino;
6715 __u32 btf_id;
6716 __u32 btf_key_type_id;
6717 __u32 btf_value_type_id;
6718 __u32 btf_vmlinux_id;
6719 __u64 map_extra;
6720 __aligned_u64 hash;
6721 __u32 hash_size;
6722} __attribute__((aligned(8)));
6723
6724struct bpf_btf_info {
6725 __aligned_u64 btf;
6726 __u32 btf_size;
6727 __u32 id;
6728 __aligned_u64 name;
6729 __u32 name_len;
6730 __u32 kernel_btf;
6731} __attribute__((aligned(8)));
6732
6733struct bpf_link_info {
6734 __u32 type;
6735 __u32 id;
6736 __u32 prog_id;
6737 union {
6738 struct {
6739 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6740 __u32 tp_name_len; /* in/out: tp_name buffer len */
6741 __u32 :32;
6742 __u64 cookie;
6743 } raw_tracepoint;
6744 struct {
6745 __u32 attach_type;
6746 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6747 __u32 target_btf_id; /* BTF type id inside the object */
6748 __u32 :32;
6749 __u64 cookie;
6750 } tracing;
6751 struct {
6752 __u64 cgroup_id;
6753 __u32 attach_type;
6754 } cgroup;
6755 struct {
6756 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
6757 __u32 target_name_len; /* in/out: target_name buffer len */
6758
6759 /* If the iter specific field is 32 bits, it can be put
6760 * in the first or second union. Otherwise it should be
6761 * put in the second union.
6762 */
6763 union {
6764 struct {
6765 __u32 map_id;
6766 } map;
6767 };
6768 union {
6769 struct {
6770 __u64 cgroup_id;
6771 __u32 order;
6772 } cgroup;
6773 struct {
6774 __u32 tid;
6775 __u32 pid;
6776 } task;
6777 };
6778 } iter;
6779 struct {
6780 __u32 netns_ino;
6781 __u32 attach_type;
6782 } netns;
6783 struct {
6784 __u32 ifindex;
6785 } xdp;
6786 struct {
6787 __u32 map_id;
6788 } struct_ops;
6789 struct {
6790 __u32 pf;
6791 __u32 hooknum;
6792 __s32 priority;
6793 __u32 flags;
6794 } netfilter;
6795 struct {
6796 __aligned_u64 addrs;
6797 __u32 count; /* in/out: kprobe_multi function count */
6798 __u32 flags;
6799 __u64 missed;
6800 __aligned_u64 cookies;
6801 } kprobe_multi;
6802 struct {
6803 __aligned_u64 path;
6804 __aligned_u64 offsets;
6805 __aligned_u64 ref_ctr_offsets;
6806 __aligned_u64 cookies;
6807 __u32 path_size; /* in/out: real path size on success, including zero byte */
6808 __u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
6809 __u32 flags;
6810 __u32 pid;
6811 } uprobe_multi;
6812 struct {
6813 __u32 type; /* enum bpf_perf_event_type */
6814 __u32 :32;
6815 union {
6816 struct {
6817 __aligned_u64 file_name; /* in/out */
6818 __u32 name_len;
6819 __u32 offset; /* offset from file_name */
6820 __u64 cookie;
6821 __u64 ref_ctr_offset;
6822 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6823 struct {
6824 __aligned_u64 func_name; /* in/out */
6825 __u32 name_len;
6826 __u32 offset; /* offset from func_name */
6827 __u64 addr;
6828 __u64 missed;
6829 __u64 cookie;
6830 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6831 struct {
6832 __aligned_u64 tp_name; /* in/out */
6833 __u32 name_len;
6834 __u32 :32;
6835 __u64 cookie;
6836 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
6837 struct {
6838 __u64 config;
6839 __u32 type;
6840 __u32 :32;
6841 __u64 cookie;
6842 } event; /* BPF_PERF_EVENT_EVENT */
6843 };
6844 } perf_event;
6845 struct {
6846 __u32 ifindex;
6847 __u32 attach_type;
6848 } tcx;
6849 struct {
6850 __u32 ifindex;
6851 __u32 attach_type;
6852 } netkit;
6853 struct {
6854 __u32 map_id;
6855 __u32 attach_type;
6856 } sockmap;
6857 };
6858} __attribute__((aligned(8)));
6859
6860struct bpf_token_info {
6861 __u64 allowed_cmds;
6862 __u64 allowed_maps;
6863 __u64 allowed_progs;
6864 __u64 allowed_attachs;
6865} __attribute__((aligned(8)));
6866
6867/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6868 * by user and intended to be used by socket (e.g. to bind to, depends on
6869 * attach type).
6870 */
6871struct bpf_sock_addr {
6872 __u32 user_family; /* Allows 4-byte read, but no write. */
6873 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6874 * Stored in network byte order.
6875 */
6876 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6877 * Stored in network byte order.
6878 */
6879 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
6880 * Stored in network byte order
6881 */
6882 __u32 family; /* Allows 4-byte read, but no write */
6883 __u32 type; /* Allows 4-byte read, but no write */
6884 __u32 protocol; /* Allows 4-byte read, but no write */
6885 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6886 * Stored in network byte order.
6887 */
6888 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6889 * Stored in network byte order.
6890 */
6891 __bpf_md_ptr(struct bpf_sock *, sk);
6892};
6893
6894/* User bpf_sock_ops struct to access socket values and specify request ops
6895 * and their replies.
6896 * Some of this fields are in network (bigendian) byte order and may need
6897 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6898 * New fields can only be added at the end of this structure
6899 */
6900struct bpf_sock_ops {
6901 __u32 op;
6902 union {
6903 __u32 args[4]; /* Optionally passed to bpf program */
6904 __u32 reply; /* Returned by bpf program */
6905 __u32 replylong[4]; /* Optionally returned by bpf prog */
6906 };
6907 __u32 family;
6908 __u32 remote_ip4; /* Stored in network byte order */
6909 __u32 local_ip4; /* Stored in network byte order */
6910 __u32 remote_ip6[4]; /* Stored in network byte order */
6911 __u32 local_ip6[4]; /* Stored in network byte order */
6912 __u32 remote_port; /* Stored in network byte order */
6913 __u32 local_port; /* stored in host byte order */
6914 __u32 is_fullsock; /* Some TCP fields are only valid if
6915 * there is a full socket. If not, the
6916 * fields read as zero.
6917 */
6918 __u32 snd_cwnd;
6919 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
6920 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6921 __u32 state;
6922 __u32 rtt_min;
6923 __u32 snd_ssthresh;
6924 __u32 rcv_nxt;
6925 __u32 snd_nxt;
6926 __u32 snd_una;
6927 __u32 mss_cache;
6928 __u32 ecn_flags;
6929 __u32 rate_delivered;
6930 __u32 rate_interval_us;
6931 __u32 packets_out;
6932 __u32 retrans_out;
6933 __u32 total_retrans;
6934 __u32 segs_in;
6935 __u32 data_segs_in;
6936 __u32 segs_out;
6937 __u32 data_segs_out;
6938 __u32 lost_out;
6939 __u32 sacked_out;
6940 __u32 sk_txhash;
6941 __u64 bytes_received;
6942 __u64 bytes_acked;
6943 __bpf_md_ptr(struct bpf_sock *, sk);
6944 /* [skb_data, skb_data_end) covers the whole TCP header.
6945 *
6946 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6947 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
6948 * header has not been written.
6949 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6950 * been written so far.
6951 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
6952 * the 3WHS.
6953 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6954 * the 3WHS.
6955 *
6956 * bpf_load_hdr_opt() can also be used to read a particular option.
6957 */
6958 __bpf_md_ptr(void *, skb_data);
6959 __bpf_md_ptr(void *, skb_data_end);
6960 __u32 skb_len; /* The total length of a packet.
6961 * It includes the header, options,
6962 * and payload.
6963 */
6964 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
6965 * an easy way to check for tcp_flags
6966 * without parsing skb_data.
6967 *
6968 * In particular, the skb_tcp_flags
6969 * will still be available in
6970 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6971 * the outgoing header has not
6972 * been written yet.
6973 */
6974 __u64 skb_hwtstamp;
6975};
6976
6977/* Definitions for bpf_sock_ops_cb_flags */
6978enum {
6979 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
6980 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
6981 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
6982 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
6983 /* Call bpf for all received TCP headers. The bpf prog will be
6984 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6985 *
6986 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6987 * for the header option related helpers that will be useful
6988 * to the bpf programs.
6989 *
6990 * It could be used at the client/active side (i.e. connect() side)
6991 * when the server told it that the server was in syncookie
6992 * mode and required the active side to resend the bpf-written
6993 * options. The active side can keep writing the bpf-options until
6994 * it received a valid packet from the server side to confirm
6995 * the earlier packet (and options) has been received. The later
6996 * example patch is using it like this at the active side when the
6997 * server is in syncookie mode.
6998 *
6999 * The bpf prog will usually turn this off in the common cases.
7000 */
7001 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
7002 /* Call bpf when kernel has received a header option that
7003 * the kernel cannot handle. The bpf prog will be called under
7004 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
7005 *
7006 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
7007 * for the header option related helpers that will be useful
7008 * to the bpf programs.
7009 */
7010 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
7011 /* Call bpf when the kernel is writing header options for the
7012 * outgoing packet. The bpf prog will first be called
7013 * to reserve space in a skb under
7014 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
7015 * the bpf prog will be called to write the header option(s)
7016 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7017 *
7018 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
7019 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
7020 * related helpers that will be useful to the bpf programs.
7021 *
7022 * The kernel gets its chance to reserve space and write
7023 * options first before the BPF program does.
7024 */
7025 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
7026/* Mask of all currently supported cb flags */
7027 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
7028};
7029
7030enum {
7031 SK_BPF_CB_TX_TIMESTAMPING = 1<<0,
7032 SK_BPF_CB_MASK = (SK_BPF_CB_TX_TIMESTAMPING - 1) |
7033 SK_BPF_CB_TX_TIMESTAMPING
7034};
7035
7036/* List of known BPF sock_ops operators.
7037 * New entries can only be added at the end
7038 */
7039enum {
7040 BPF_SOCK_OPS_VOID,
7041 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
7042 * -1 if default value should be used
7043 */
7044 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
7045 * window (in packets) or -1 if default
7046 * value should be used
7047 */
7048 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
7049 * active connection is initialized
7050 */
7051 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
7052 * active connection is
7053 * established
7054 */
7055 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
7056 * passive connection is
7057 * established
7058 */
7059 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
7060 * needs ECN
7061 */
7062 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
7063 * based on the path and may be
7064 * dependent on the congestion control
7065 * algorithm. In general it indicates
7066 * a congestion threshold. RTTs above
7067 * this indicate congestion
7068 */
7069 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
7070 * Arg1: value of icsk_retransmits
7071 * Arg2: value of icsk_rto
7072 * Arg3: whether RTO has expired
7073 */
7074 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
7075 * Arg1: sequence number of 1st byte
7076 * Arg2: # segments
7077 * Arg3: return value of
7078 * tcp_transmit_skb (0 => success)
7079 */
7080 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
7081 * Arg1: old_state
7082 * Arg2: new_state
7083 */
7084 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
7085 * socket transition to LISTEN state.
7086 */
7087 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
7088 * Arg1: measured RTT input (mrtt)
7089 * Arg2: updated srtt
7090 */
7091 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
7092 * It will be called to handle
7093 * the packets received at
7094 * an already established
7095 * connection.
7096 *
7097 * sock_ops->skb_data:
7098 * Referring to the received skb.
7099 * It covers the TCP header only.
7100 *
7101 * bpf_load_hdr_opt() can also
7102 * be used to search for a
7103 * particular option.
7104 */
7105 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
7106 * header option later in
7107 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7108 * Arg1: bool want_cookie. (in
7109 * writing SYNACK only)
7110 *
7111 * sock_ops->skb_data:
7112 * Not available because no header has
7113 * been written yet.
7114 *
7115 * sock_ops->skb_tcp_flags:
7116 * The tcp_flags of the
7117 * outgoing skb. (e.g. SYN, ACK, FIN).
7118 *
7119 * bpf_reserve_hdr_opt() should
7120 * be used to reserve space.
7121 */
7122 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
7123 * Arg1: bool want_cookie. (in
7124 * writing SYNACK only)
7125 *
7126 * sock_ops->skb_data:
7127 * Referring to the outgoing skb.
7128 * It covers the TCP header
7129 * that has already been written
7130 * by the kernel and the
7131 * earlier bpf-progs.
7132 *
7133 * sock_ops->skb_tcp_flags:
7134 * The tcp_flags of the outgoing
7135 * skb. (e.g. SYN, ACK, FIN).
7136 *
7137 * bpf_store_hdr_opt() should
7138 * be used to write the
7139 * option.
7140 *
7141 * bpf_load_hdr_opt() can also
7142 * be used to search for a
7143 * particular option that
7144 * has already been written
7145 * by the kernel or the
7146 * earlier bpf-progs.
7147 */
7148 BPF_SOCK_OPS_TSTAMP_SCHED_CB, /* Called when skb is passing
7149 * through dev layer when
7150 * SK_BPF_CB_TX_TIMESTAMPING
7151 * feature is on.
7152 */
7153 BPF_SOCK_OPS_TSTAMP_SND_SW_CB, /* Called when skb is about to send
7154 * to the nic when SK_BPF_CB_TX_TIMESTAMPING
7155 * feature is on.
7156 */
7157 BPF_SOCK_OPS_TSTAMP_SND_HW_CB, /* Called in hardware phase when
7158 * SK_BPF_CB_TX_TIMESTAMPING feature
7159 * is on.
7160 */
7161 BPF_SOCK_OPS_TSTAMP_ACK_CB, /* Called when all the skbs in the
7162 * same sendmsg call are acked
7163 * when SK_BPF_CB_TX_TIMESTAMPING
7164 * feature is on.
7165 */
7166 BPF_SOCK_OPS_TSTAMP_SENDMSG_CB, /* Called when every sendmsg syscall
7167 * is triggered. It's used to correlate
7168 * sendmsg timestamp with corresponding
7169 * tskey.
7170 */
7171};
7172
7173/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
7174 * changes between the TCP and BPF versions. Ideally this should never happen.
7175 * If it does, we need to add code to convert them before calling
7176 * the BPF sock_ops function.
7177 */
7178enum {
7179 BPF_TCP_ESTABLISHED = 1,
7180 BPF_TCP_SYN_SENT,
7181 BPF_TCP_SYN_RECV,
7182 BPF_TCP_FIN_WAIT1,
7183 BPF_TCP_FIN_WAIT2,
7184 BPF_TCP_TIME_WAIT,
7185 BPF_TCP_CLOSE,
7186 BPF_TCP_CLOSE_WAIT,
7187 BPF_TCP_LAST_ACK,
7188 BPF_TCP_LISTEN,
7189 BPF_TCP_CLOSING, /* Now a valid state */
7190 BPF_TCP_NEW_SYN_RECV,
7191 BPF_TCP_BOUND_INACTIVE,
7192
7193 BPF_TCP_MAX_STATES /* Leave at the end! */
7194};
7195
7196enum {
7197 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
7198 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
7199 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
7200 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
7201 /* Copy the SYN pkt to optval
7202 *
7203 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
7204 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
7205 * to only getting from the saved_syn. It can either get the
7206 * syn packet from:
7207 *
7208 * 1. the just-received SYN packet (only available when writing the
7209 * SYNACK). It will be useful when it is not necessary to
7210 * save the SYN packet for latter use. It is also the only way
7211 * to get the SYN during syncookie mode because the syn
7212 * packet cannot be saved during syncookie.
7213 *
7214 * OR
7215 *
7216 * 2. the earlier saved syn which was done by
7217 * bpf_setsockopt(TCP_SAVE_SYN).
7218 *
7219 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
7220 * SYN packet is obtained.
7221 *
7222 * If the bpf-prog does not need the IP[46] header, the
7223 * bpf-prog can avoid parsing the IP header by using
7224 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
7225 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
7226 *
7227 * >0: Total number of bytes copied
7228 * -ENOSPC: Not enough space in optval. Only optlen number of
7229 * bytes is copied.
7230 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
7231 * is not saved by setsockopt(TCP_SAVE_SYN).
7232 */
7233 TCP_BPF_SYN = 1005, /* Copy the TCP header */
7234 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
7235 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
7236 TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */
7237 SK_BPF_CB_FLAGS = 1009, /* Get or set sock ops flags in socket */
7238 SK_BPF_BYPASS_PROT_MEM = 1010, /* Get or Set sk->sk_bypass_prot_mem */
7239
7240};
7241
7242enum {
7243 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
7244};
7245
7246/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
7247 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7248 */
7249enum {
7250 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
7251 * total option spaces
7252 * required for an established
7253 * sk in order to calculate the
7254 * MSS. No skb is actually
7255 * sent.
7256 */
7257 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
7258 * when sending a SYN.
7259 */
7260};
7261
7262struct bpf_perf_event_value {
7263 __u64 counter;
7264 __u64 enabled;
7265 __u64 running;
7266};
7267
7268enum {
7269 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
7270 BPF_DEVCG_ACC_READ = (1ULL << 1),
7271 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
7272};
7273
7274enum {
7275 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
7276 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
7277};
7278
7279struct bpf_cgroup_dev_ctx {
7280 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
7281 __u32 access_type;
7282 __u32 major;
7283 __u32 minor;
7284};
7285
7286struct bpf_raw_tracepoint_args {
7287 __u64 args[0];
7288};
7289
7290/* DIRECT: Skip the FIB rules and go to FIB table associated with device
7291 * OUTPUT: Do lookup from egress perspective; default is ingress
7292 */
7293enum {
7294 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
7295 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
7296 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
7297 BPF_FIB_LOOKUP_TBID = (1U << 3),
7298 BPF_FIB_LOOKUP_SRC = (1U << 4),
7299 BPF_FIB_LOOKUP_MARK = (1U << 5),
7300};
7301
7302enum {
7303 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
7304 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
7305 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
7306 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
7307 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
7308 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
7309 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
7310 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
7311 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
7312 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */
7313};
7314
7315struct bpf_fib_lookup {
7316 /* input: network family for lookup (AF_INET, AF_INET6)
7317 * output: network family of egress nexthop
7318 */
7319 __u8 family;
7320
7321 /* set if lookup is to consider L4 data - e.g., FIB rules */
7322 __u8 l4_protocol;
7323 __be16 sport;
7324 __be16 dport;
7325
7326 union { /* used for MTU check */
7327 /* input to lookup */
7328 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
7329
7330 /* output: MTU value */
7331 __u16 mtu_result;
7332 } __attribute__((packed, aligned(2)));
7333 /* input: L3 device index for lookup
7334 * output: device index from FIB lookup
7335 */
7336 __u32 ifindex;
7337
7338 union {
7339 /* inputs to lookup */
7340 __u8 tos; /* AF_INET */
7341 __be32 flowinfo; /* AF_INET6, flow_label + priority */
7342
7343 /* output: metric of fib result (IPv4/IPv6 only) */
7344 __u32 rt_metric;
7345 };
7346
7347 /* input: source address to consider for lookup
7348 * output: source address result from lookup
7349 */
7350 union {
7351 __be32 ipv4_src;
7352 __u32 ipv6_src[4]; /* in6_addr; network order */
7353 };
7354
7355 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7356 * network header. output: bpf_fib_lookup sets to gateway address
7357 * if FIB lookup returns gateway route
7358 */
7359 union {
7360 __be32 ipv4_dst;
7361 __u32 ipv6_dst[4]; /* in6_addr; network order */
7362 };
7363
7364 union {
7365 struct {
7366 /* output */
7367 __be16 h_vlan_proto;
7368 __be16 h_vlan_TCI;
7369 };
7370 /* input: when accompanied with the
7371 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7372 * specific routing table to use for the fib lookup.
7373 */
7374 __u32 tbid;
7375 };
7376
7377 union {
7378 /* input */
7379 struct {
7380 __u32 mark; /* policy routing */
7381 /* 2 4-byte holes for input */
7382 };
7383
7384 /* output: source and dest mac */
7385 struct {
7386 __u8 smac[6]; /* ETH_ALEN */
7387 __u8 dmac[6]; /* ETH_ALEN */
7388 };
7389 };
7390};
7391
7392struct bpf_redir_neigh {
7393 /* network family for lookup (AF_INET, AF_INET6) */
7394 __u32 nh_family;
7395 /* network address of nexthop; skips fib lookup to find gateway */
7396 union {
7397 __be32 ipv4_nh;
7398 __u32 ipv6_nh[4]; /* in6_addr; network order */
7399 };
7400};
7401
7402/* bpf_check_mtu flags*/
7403enum bpf_check_mtu_flags {
7404 BPF_MTU_CHK_SEGS = (1U << 0),
7405};
7406
7407enum bpf_check_mtu_ret {
7408 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
7409 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
7410 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
7411};
7412
7413enum bpf_task_fd_type {
7414 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
7415 BPF_FD_TYPE_TRACEPOINT, /* tp name */
7416 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
7417 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
7418 BPF_FD_TYPE_UPROBE, /* filename + offset */
7419 BPF_FD_TYPE_URETPROBE, /* filename + offset */
7420};
7421
7422enum {
7423 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
7424 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
7425 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
7426};
7427
7428struct bpf_flow_keys {
7429 __u16 nhoff;
7430 __u16 thoff;
7431 __u16 addr_proto; /* ETH_P_* of valid addrs */
7432 __u8 is_frag;
7433 __u8 is_first_frag;
7434 __u8 is_encap;
7435 __u8 ip_proto;
7436 __be16 n_proto;
7437 __be16 sport;
7438 __be16 dport;
7439 union {
7440 struct {
7441 __be32 ipv4_src;
7442 __be32 ipv4_dst;
7443 };
7444 struct {
7445 __u32 ipv6_src[4]; /* in6_addr; network order */
7446 __u32 ipv6_dst[4]; /* in6_addr; network order */
7447 };
7448 };
7449 __u32 flags;
7450 __be32 flow_label;
7451};
7452
7453struct bpf_func_info {
7454 __u32 insn_off;
7455 __u32 type_id;
7456};
7457
7458#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
7459#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
7460
7461struct bpf_line_info {
7462 __u32 insn_off;
7463 __u32 file_name_off;
7464 __u32 line_off;
7465 __u32 line_col;
7466};
7467
7468struct bpf_spin_lock {
7469 __u32 val;
7470};
7471
7472struct bpf_timer {
7473 __u64 __opaque[2];
7474} __attribute__((aligned(8)));
7475
7476struct bpf_task_work {
7477 __u64 __opaque;
7478} __attribute__((aligned(8)));
7479
7480struct bpf_wq {
7481 __u64 __opaque[2];
7482} __attribute__((aligned(8)));
7483
7484struct bpf_dynptr {
7485 __u64 __opaque[2];
7486} __attribute__((aligned(8)));
7487
7488struct bpf_list_head {
7489 __u64 __opaque[2];
7490} __attribute__((aligned(8)));
7491
7492struct bpf_list_node {
7493 __u64 __opaque[3];
7494} __attribute__((aligned(8)));
7495
7496struct bpf_rb_root {
7497 __u64 __opaque[2];
7498} __attribute__((aligned(8)));
7499
7500struct bpf_rb_node {
7501 __u64 __opaque[4];
7502} __attribute__((aligned(8)));
7503
7504struct bpf_refcount {
7505 __u32 __opaque[1];
7506} __attribute__((aligned(4)));
7507
7508struct bpf_sysctl {
7509 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
7510 * Allows 1,2,4-byte read, but no write.
7511 */
7512 __u32 file_pos; /* Sysctl file position to read from, write to.
7513 * Allows 1,2,4-byte read an 4-byte write.
7514 */
7515};
7516
7517struct bpf_sockopt {
7518 __bpf_md_ptr(struct bpf_sock *, sk);
7519 __bpf_md_ptr(void *, optval);
7520 __bpf_md_ptr(void *, optval_end);
7521
7522 __s32 level;
7523 __s32 optname;
7524 __s32 optlen;
7525 __s32 retval;
7526};
7527
7528struct bpf_pidns_info {
7529 __u32 pid;
7530 __u32 tgid;
7531};
7532
7533/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7534struct bpf_sk_lookup {
7535 union {
7536 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
7537 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
7538 };
7539
7540 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
7541 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7542 __u32 remote_ip4; /* Network byte order */
7543 __u32 remote_ip6[4]; /* Network byte order */
7544 __be16 remote_port; /* Network byte order */
7545 __u16 :16; /* Zero padding */
7546 __u32 local_ip4; /* Network byte order */
7547 __u32 local_ip6[4]; /* Network byte order */
7548 __u32 local_port; /* Host byte order */
7549 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */
7550};
7551
7552/*
7553 * struct btf_ptr is used for typed pointer representation; the
7554 * type id is used to render the pointer data as the appropriate type
7555 * via the bpf_snprintf_btf() helper described above. A flags field -
7556 * potentially to specify additional details about the BTF pointer
7557 * (rather than its mode of display) - is included for future use.
7558 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7559 */
7560struct btf_ptr {
7561 void *ptr;
7562 __u32 type_id;
7563 __u32 flags; /* BTF ptr flags; unused at present. */
7564};
7565
7566/*
7567 * Flags to control bpf_snprintf_btf() behaviour.
7568 * - BTF_F_COMPACT: no formatting around type information
7569 * - BTF_F_NONAME: no struct/union member names/types
7570 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7571 * equivalent to %px.
7572 * - BTF_F_ZERO: show zero-valued struct/union members; they
7573 * are not displayed by default
7574 */
7575enum {
7576 BTF_F_COMPACT = (1ULL << 0),
7577 BTF_F_NONAME = (1ULL << 1),
7578 BTF_F_PTR_RAW = (1ULL << 2),
7579 BTF_F_ZERO = (1ULL << 3),
7580};
7581
7582/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7583 * has to be adjusted by relocations. It is emitted by llvm and passed to
7584 * libbpf and later to the kernel.
7585 */
7586enum bpf_core_relo_kind {
7587 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */
7588 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */
7589 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */
7590 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */
7591 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */
7592 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */
7593 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */
7594 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */
7595 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */
7596 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */
7597 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */
7598 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */
7599 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */
7600};
7601
7602/*
7603 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7604 * and from libbpf to the kernel.
7605 *
7606 * CO-RE relocation captures the following data:
7607 * - insn_off - instruction offset (in bytes) within a BPF program that needs
7608 * its insn->imm field to be relocated with actual field info;
7609 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7610 * type or field;
7611 * - access_str_off - offset into corresponding .BTF string section. String
7612 * interpretation depends on specific relocation kind:
7613 * - for field-based relocations, string encodes an accessed field using
7614 * a sequence of field and array indices, separated by colon (:). It's
7615 * conceptually very close to LLVM's getelementptr ([0]) instruction's
7616 * arguments for identifying offset to a field.
7617 * - for type-based relocations, strings is expected to be just "0";
7618 * - for enum value-based relocations, string contains an index of enum
7619 * value within its enum type;
7620 * - kind - one of enum bpf_core_relo_kind;
7621 *
7622 * Example:
7623 * struct sample {
7624 * int a;
7625 * struct {
7626 * int b[10];
7627 * };
7628 * };
7629 *
7630 * struct sample *s = ...;
7631 * int *x = &s->a; // encoded as "0:0" (a is field #0)
7632 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
7633 * // b is field #0 inside anon struct, accessing elem #5)
7634 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7635 *
7636 * type_id for all relocs in this example will capture BTF type id of
7637 * `struct sample`.
7638 *
7639 * Such relocation is emitted when using __builtin_preserve_access_index()
7640 * Clang built-in, passing expression that captures field address, e.g.:
7641 *
7642 * bpf_probe_read(&dst, sizeof(dst),
7643 * __builtin_preserve_access_index(&src->a.b.c));
7644 *
7645 * In this case Clang will emit field relocation recording necessary data to
7646 * be able to find offset of embedded `a.b.c` field within `src` struct.
7647 *
7648 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7649 */
7650struct bpf_core_relo {
7651 __u32 insn_off;
7652 __u32 type_id;
7653 __u32 access_str_off;
7654 enum bpf_core_relo_kind kind;
7655};
7656
7657/*
7658 * Flags to control bpf_timer_start() behaviour.
7659 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7660 * relative to current time.
7661 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
7662 */
7663enum {
7664 BPF_F_TIMER_ABS = (1ULL << 0),
7665 BPF_F_TIMER_CPU_PIN = (1ULL << 1),
7666};
7667
7668/* BPF numbers iterator state */
7669struct bpf_iter_num {
7670 /* opaque iterator state; having __u64 here allows to preserve correct
7671 * alignment requirements in vmlinux.h, generated from BTF
7672 */
7673 __u64 __opaque[1];
7674} __attribute__((aligned(8)));
7675
7676/*
7677 * Flags to control BPF kfunc behaviour.
7678 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective
7679 * helper documentation for details.)
7680 */
7681enum bpf_kfunc_flags {
7682 BPF_F_PAD_ZEROS = (1ULL << 0),
7683};
7684
7685/*
7686 * Values of a BPF_MAP_TYPE_INSN_ARRAY entry must be of this type.
7687 *
7688 * Before the map is used the orig_off field should point to an
7689 * instruction inside the program being loaded. The other fields
7690 * must be set to 0.
7691 *
7692 * After the program is loaded, the xlated_off will be adjusted
7693 * by the verifier to point to the index of the original instruction
7694 * in the xlated program. If the instruction is deleted, it will
7695 * be set to (u32)-1. The jitted_off will be set to the corresponding
7696 * offset in the jitted image of the program.
7697 */
7698struct bpf_insn_array_value {
7699 __u32 orig_off;
7700 __u32 xlated_off;
7701 __u32 jitted_off;
7702 __u32 :32;
7703};
7704
7705#endif /* _UAPI__LINUX_BPF_H__ */