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