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 __LINUX_BPF_H__
9#define __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_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
23#define BPF_XADD 0xc0 /* exclusive add - legacy name */
24
25/* alu/jmp fields */
26#define BPF_MOV 0xb0 /* mov reg to reg */
27#define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
28
29/* change endianness of a register */
30#define BPF_END 0xd0 /* flags for endianness conversion: */
31#define BPF_TO_LE 0x00 /* convert to little-endian */
32#define BPF_TO_BE 0x08 /* convert to big-endian */
33#define BPF_FROM_LE BPF_TO_LE
34#define BPF_FROM_BE BPF_TO_BE
35
36/* jmp encodings */
37#define BPF_JNE 0x50 /* jump != */
38#define BPF_JLT 0xa0 /* LT is unsigned, '<' */
39#define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
40#define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
41#define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
42#define BPF_JSLT 0xc0 /* SLT is signed, '<' */
43#define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
44#define BPF_CALL 0x80 /* function call */
45#define BPF_EXIT 0x90 /* function return */
46
47/* atomic op type fields (stored in immediate) */
48#define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
49#define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
50#define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
51
52/* Register numbers */
53enum {
54 BPF_REG_0 = 0,
55 BPF_REG_1,
56 BPF_REG_2,
57 BPF_REG_3,
58 BPF_REG_4,
59 BPF_REG_5,
60 BPF_REG_6,
61 BPF_REG_7,
62 BPF_REG_8,
63 BPF_REG_9,
64 BPF_REG_10,
65 __MAX_BPF_REG,
66};
67
68/* BPF has 10 general purpose 64-bit registers and stack frame. */
69#define MAX_BPF_REG __MAX_BPF_REG
70
71struct bpf_insn {
72 __u8 code; /* opcode */
73 __u8 dst_reg:4; /* dest register */
74 __u8 src_reg:4; /* source register */
75 __s16 off; /* signed offset */
76 __s32 imm; /* signed immediate constant */
77};
78
79/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80struct bpf_lpm_trie_key {
81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
82 __u8 data[0]; /* Arbitrary size */
83};
84
85struct bpf_cgroup_storage_key {
86 __u64 cgroup_inode_id; /* cgroup inode id */
87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */
88};
89
90union bpf_iter_link_info {
91 struct {
92 __u32 map_fd;
93 } map;
94};
95
96/* BPF syscall commands, see bpf(2) man-page for more details. */
97/**
98 * DOC: eBPF Syscall Preamble
99 *
100 * The operation to be performed by the **bpf**\ () system call is determined
101 * by the *cmd* argument. Each operation takes an accompanying argument,
102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
103 * below). The size argument is the size of the union pointed to by *attr*.
104 */
105/**
106 * DOC: eBPF Syscall Commands
107 *
108 * BPF_MAP_CREATE
109 * Description
110 * Create a map and return a file descriptor that refers to the
111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
112 * is automatically enabled for the new file descriptor.
113 *
114 * Applying **close**\ (2) to the file descriptor returned by
115 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
116 *
117 * Return
118 * A new file descriptor (a nonnegative integer), or -1 if an
119 * error occurred (in which case, *errno* is set appropriately).
120 *
121 * BPF_MAP_LOOKUP_ELEM
122 * Description
123 * Look up an element with a given *key* in the map referred to
124 * by the file descriptor *map_fd*.
125 *
126 * The *flags* argument may be specified as one of the
127 * following:
128 *
129 * **BPF_F_LOCK**
130 * Look up the value of a spin-locked map without
131 * returning the lock. This must be specified if the
132 * elements contain a spinlock.
133 *
134 * Return
135 * Returns zero on success. On error, -1 is returned and *errno*
136 * is set appropriately.
137 *
138 * BPF_MAP_UPDATE_ELEM
139 * Description
140 * Create or update an element (key/value pair) in a specified map.
141 *
142 * The *flags* argument should be specified as one of the
143 * following:
144 *
145 * **BPF_ANY**
146 * Create a new element or update an existing element.
147 * **BPF_NOEXIST**
148 * Create a new element only if it did not exist.
149 * **BPF_EXIST**
150 * Update an existing element.
151 * **BPF_F_LOCK**
152 * Update a spin_lock-ed map element.
153 *
154 * Return
155 * Returns zero on success. On error, -1 is returned and *errno*
156 * is set appropriately.
157 *
158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
159 * **E2BIG**, **EEXIST**, or **ENOENT**.
160 *
161 * **E2BIG**
162 * The number of elements in the map reached the
163 * *max_entries* limit specified at map creation time.
164 * **EEXIST**
165 * If *flags* specifies **BPF_NOEXIST** and the element
166 * with *key* already exists in the map.
167 * **ENOENT**
168 * If *flags* specifies **BPF_EXIST** and the element with
169 * *key* does not exist in the map.
170 *
171 * BPF_MAP_DELETE_ELEM
172 * Description
173 * Look up and delete an element by key in a specified map.
174 *
175 * Return
176 * Returns zero on success. On error, -1 is returned and *errno*
177 * is set appropriately.
178 *
179 * BPF_MAP_GET_NEXT_KEY
180 * Description
181 * Look up an element by key in a specified map and return the key
182 * of the next element. Can be used to iterate over all elements
183 * in the map.
184 *
185 * Return
186 * Returns zero on success. On error, -1 is returned and *errno*
187 * is set appropriately.
188 *
189 * The following cases can be used to iterate over all elements of
190 * the map:
191 *
192 * * If *key* is not found, the operation returns zero and sets
193 * the *next_key* pointer to the key of the first element.
194 * * If *key* is found, the operation returns zero and sets the
195 * *next_key* pointer to the key of the next element.
196 * * If *key* is the last element, returns -1 and *errno* is set
197 * to **ENOENT**.
198 *
199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
200 * **EINVAL** on error.
201 *
202 * BPF_PROG_LOAD
203 * Description
204 * Verify and load an eBPF program, returning a new file
205 * descriptor associated with the program.
206 *
207 * Applying **close**\ (2) to the file descriptor returned by
208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
209 *
210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
211 * automatically enabled for the new file descriptor.
212 *
213 * Return
214 * A new file descriptor (a nonnegative integer), or -1 if an
215 * error occurred (in which case, *errno* is set appropriately).
216 *
217 * BPF_OBJ_PIN
218 * Description
219 * Pin an eBPF program or map referred by the specified *bpf_fd*
220 * to the provided *pathname* on the filesystem.
221 *
222 * The *pathname* argument must not contain a dot (".").
223 *
224 * On success, *pathname* retains a reference to the eBPF object,
225 * preventing deallocation of the object when the original
226 * *bpf_fd* is closed. This allow the eBPF object to live beyond
227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
228 * process.
229 *
230 * Applying **unlink**\ (2) or similar calls to the *pathname*
231 * unpins the object from the filesystem, removing the reference.
232 * If no other file descriptors or filesystem nodes refer to the
233 * same object, it will be deallocated (see NOTES).
234 *
235 * The filesystem type for the parent directory of *pathname* must
236 * be **BPF_FS_MAGIC**.
237 *
238 * Return
239 * Returns zero on success. On error, -1 is returned and *errno*
240 * is set appropriately.
241 *
242 * BPF_OBJ_GET
243 * Description
244 * Open a file descriptor for the eBPF object pinned to the
245 * specified *pathname*.
246 *
247 * Return
248 * A new file descriptor (a nonnegative integer), or -1 if an
249 * error occurred (in which case, *errno* is set appropriately).
250 *
251 * BPF_PROG_ATTACH
252 * Description
253 * Attach an eBPF program to a *target_fd* at the specified
254 * *attach_type* hook.
255 *
256 * The *attach_type* specifies the eBPF attachment point to
257 * attach the program to, and must be one of *bpf_attach_type*
258 * (see below).
259 *
260 * The *attach_bpf_fd* must be a valid file descriptor for a
261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
262 * or sock_ops type corresponding to the specified *attach_type*.
263 *
264 * The *target_fd* must be a valid file descriptor for a kernel
265 * object which depends on the attach type of *attach_bpf_fd*:
266 *
267 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
268 * **BPF_PROG_TYPE_CGROUP_SKB**,
269 * **BPF_PROG_TYPE_CGROUP_SOCK**,
270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
273 * **BPF_PROG_TYPE_SOCK_OPS**
274 *
275 * Control Group v2 hierarchy with the eBPF controller
276 * enabled. Requires the kernel to be compiled with
277 * **CONFIG_CGROUP_BPF**.
278 *
279 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
280 *
281 * Network namespace (eg /proc/self/ns/net).
282 *
283 * **BPF_PROG_TYPE_LIRC_MODE2**
284 *
285 * LIRC device path (eg /dev/lircN). Requires the kernel
286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
287 *
288 * **BPF_PROG_TYPE_SK_SKB**,
289 * **BPF_PROG_TYPE_SK_MSG**
290 *
291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
292 *
293 * Return
294 * Returns zero on success. On error, -1 is returned and *errno*
295 * is set appropriately.
296 *
297 * BPF_PROG_DETACH
298 * Description
299 * Detach the eBPF program associated with the *target_fd* at the
300 * hook specified by *attach_type*. The program must have been
301 * previously attached using **BPF_PROG_ATTACH**.
302 *
303 * Return
304 * Returns zero on success. On error, -1 is returned and *errno*
305 * is set appropriately.
306 *
307 * BPF_PROG_TEST_RUN
308 * Description
309 * Run the eBPF program associated with the *prog_fd* a *repeat*
310 * number of times against a provided program context *ctx_in* and
311 * data *data_in*, and return the modified program context
312 * *ctx_out*, *data_out* (for example, packet data), result of the
313 * execution *retval*, and *duration* of the test run.
314 *
315 * The sizes of the buffers provided as input and output
316 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
317 * be provided in the corresponding variables *ctx_size_in*,
318 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
319 * of these parameters are not provided (ie set to NULL), the
320 * corresponding size field must be zero.
321 *
322 * Some program types have particular requirements:
323 *
324 * **BPF_PROG_TYPE_SK_LOOKUP**
325 * *data_in* and *data_out* must be NULL.
326 *
327 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
328 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
329 *
330 * *ctx_out*, *data_in* and *data_out* must be NULL.
331 * *repeat* must be zero.
332 *
333 * Return
334 * Returns zero on success. On error, -1 is returned and *errno*
335 * is set appropriately.
336 *
337 * **ENOSPC**
338 * Either *data_size_out* or *ctx_size_out* is too small.
339 * **ENOTSUPP**
340 * This command is not supported by the program type of
341 * the program referred to by *prog_fd*.
342 *
343 * BPF_PROG_GET_NEXT_ID
344 * Description
345 * Fetch the next eBPF program currently loaded into the kernel.
346 *
347 * Looks for the eBPF program with an id greater than *start_id*
348 * and updates *next_id* on success. If no other eBPF programs
349 * remain with ids higher than *start_id*, returns -1 and sets
350 * *errno* to **ENOENT**.
351 *
352 * Return
353 * Returns zero on success. On error, or when no id remains, -1
354 * is returned and *errno* is set appropriately.
355 *
356 * BPF_MAP_GET_NEXT_ID
357 * Description
358 * Fetch the next eBPF map currently loaded into the kernel.
359 *
360 * Looks for the eBPF map with an id greater than *start_id*
361 * and updates *next_id* on success. If no other eBPF maps
362 * remain with ids higher than *start_id*, returns -1 and sets
363 * *errno* to **ENOENT**.
364 *
365 * Return
366 * Returns zero on success. On error, or when no id remains, -1
367 * is returned and *errno* is set appropriately.
368 *
369 * BPF_PROG_GET_FD_BY_ID
370 * Description
371 * Open a file descriptor for the eBPF program corresponding to
372 * *prog_id*.
373 *
374 * Return
375 * A new file descriptor (a nonnegative integer), or -1 if an
376 * error occurred (in which case, *errno* is set appropriately).
377 *
378 * BPF_MAP_GET_FD_BY_ID
379 * Description
380 * Open a file descriptor for the eBPF map corresponding to
381 * *map_id*.
382 *
383 * Return
384 * A new file descriptor (a nonnegative integer), or -1 if an
385 * error occurred (in which case, *errno* is set appropriately).
386 *
387 * BPF_OBJ_GET_INFO_BY_FD
388 * Description
389 * Obtain information about the eBPF object corresponding to
390 * *bpf_fd*.
391 *
392 * Populates up to *info_len* bytes of *info*, which will be in
393 * one of the following formats depending on the eBPF object type
394 * of *bpf_fd*:
395 *
396 * * **struct bpf_prog_info**
397 * * **struct bpf_map_info**
398 * * **struct bpf_btf_info**
399 * * **struct bpf_link_info**
400 *
401 * Return
402 * Returns zero on success. On error, -1 is returned and *errno*
403 * is set appropriately.
404 *
405 * BPF_PROG_QUERY
406 * Description
407 * Obtain information about eBPF programs associated with the
408 * specified *attach_type* hook.
409 *
410 * The *target_fd* must be a valid file descriptor for a kernel
411 * object which depends on the attach type of *attach_bpf_fd*:
412 *
413 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
414 * **BPF_PROG_TYPE_CGROUP_SKB**,
415 * **BPF_PROG_TYPE_CGROUP_SOCK**,
416 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
417 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
418 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
419 * **BPF_PROG_TYPE_SOCK_OPS**
420 *
421 * Control Group v2 hierarchy with the eBPF controller
422 * enabled. Requires the kernel to be compiled with
423 * **CONFIG_CGROUP_BPF**.
424 *
425 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
426 *
427 * Network namespace (eg /proc/self/ns/net).
428 *
429 * **BPF_PROG_TYPE_LIRC_MODE2**
430 *
431 * LIRC device path (eg /dev/lircN). Requires the kernel
432 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
433 *
434 * **BPF_PROG_QUERY** always fetches the number of programs
435 * attached and the *attach_flags* which were used to attach those
436 * programs. Additionally, if *prog_ids* is nonzero and the number
437 * of attached programs is less than *prog_cnt*, populates
438 * *prog_ids* with the eBPF program ids of the programs attached
439 * at *target_fd*.
440 *
441 * The following flags may alter the result:
442 *
443 * **BPF_F_QUERY_EFFECTIVE**
444 * Only return information regarding programs which are
445 * currently effective at the specified *target_fd*.
446 *
447 * Return
448 * Returns zero on success. On error, -1 is returned and *errno*
449 * is set appropriately.
450 *
451 * BPF_RAW_TRACEPOINT_OPEN
452 * Description
453 * Attach an eBPF program to a tracepoint *name* to access kernel
454 * internal arguments of the tracepoint in their raw form.
455 *
456 * The *prog_fd* must be a valid file descriptor associated with
457 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
458 *
459 * No ABI guarantees are made about the content of tracepoint
460 * arguments exposed to the corresponding eBPF program.
461 *
462 * Applying **close**\ (2) to the file descriptor returned by
463 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
464 *
465 * Return
466 * A new file descriptor (a nonnegative integer), or -1 if an
467 * error occurred (in which case, *errno* is set appropriately).
468 *
469 * BPF_BTF_LOAD
470 * Description
471 * Verify and load BPF Type Format (BTF) metadata into the kernel,
472 * returning a new file descriptor associated with the metadata.
473 * BTF is described in more detail at
474 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
475 *
476 * The *btf* parameter must point to valid memory providing
477 * *btf_size* bytes of BTF binary metadata.
478 *
479 * The returned file descriptor can be passed to other **bpf**\ ()
480 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
481 * associate the BTF with those objects.
482 *
483 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
484 * parameters to specify a *btf_log_buf*, *btf_log_size* and
485 * *btf_log_level* which allow the kernel to return freeform log
486 * output regarding the BTF verification process.
487 *
488 * Return
489 * A new file descriptor (a nonnegative integer), or -1 if an
490 * error occurred (in which case, *errno* is set appropriately).
491 *
492 * BPF_BTF_GET_FD_BY_ID
493 * Description
494 * Open a file descriptor for the BPF Type Format (BTF)
495 * corresponding to *btf_id*.
496 *
497 * Return
498 * A new file descriptor (a nonnegative integer), or -1 if an
499 * error occurred (in which case, *errno* is set appropriately).
500 *
501 * BPF_TASK_FD_QUERY
502 * Description
503 * Obtain information about eBPF programs associated with the
504 * target process identified by *pid* and *fd*.
505 *
506 * If the *pid* and *fd* are associated with a tracepoint, kprobe
507 * or uprobe perf event, then the *prog_id* and *fd_type* will
508 * be populated with the eBPF program id and file descriptor type
509 * of type **bpf_task_fd_type**. If associated with a kprobe or
510 * uprobe, the *probe_offset* and *probe_addr* will also be
511 * populated. Optionally, if *buf* is provided, then up to
512 * *buf_len* bytes of *buf* will be populated with the name of
513 * the tracepoint, kprobe or uprobe.
514 *
515 * The resulting *prog_id* may be introspected in deeper detail
516 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
517 *
518 * Return
519 * Returns zero on success. On error, -1 is returned and *errno*
520 * is set appropriately.
521 *
522 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
523 * Description
524 * Look up an element with the given *key* in the map referred to
525 * by the file descriptor *fd*, and if found, delete the element.
526 *
527 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
528 * types, the *flags* argument needs to be set to 0, but for other
529 * map types, it may be specified as:
530 *
531 * **BPF_F_LOCK**
532 * Look up and delete the value of a spin-locked map
533 * without returning the lock. This must be specified if
534 * the elements contain a spinlock.
535 *
536 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
537 * implement this command as a "pop" operation, deleting the top
538 * element rather than one corresponding to *key*.
539 * The *key* and *key_len* parameters should be zeroed when
540 * issuing this operation for these map types.
541 *
542 * This command is only valid for the following map types:
543 * * **BPF_MAP_TYPE_QUEUE**
544 * * **BPF_MAP_TYPE_STACK**
545 * * **BPF_MAP_TYPE_HASH**
546 * * **BPF_MAP_TYPE_PERCPU_HASH**
547 * * **BPF_MAP_TYPE_LRU_HASH**
548 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
549 *
550 * Return
551 * Returns zero on success. On error, -1 is returned and *errno*
552 * is set appropriately.
553 *
554 * BPF_MAP_FREEZE
555 * Description
556 * Freeze the permissions of the specified map.
557 *
558 * Write permissions may be frozen by passing zero *flags*.
559 * Upon success, no future syscall invocations may alter the
560 * map state of *map_fd*. Write operations from eBPF programs
561 * are still possible for a frozen map.
562 *
563 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
564 *
565 * Return
566 * Returns zero on success. On error, -1 is returned and *errno*
567 * is set appropriately.
568 *
569 * BPF_BTF_GET_NEXT_ID
570 * Description
571 * Fetch the next BPF Type Format (BTF) object currently loaded
572 * into the kernel.
573 *
574 * Looks for the BTF object with an id greater than *start_id*
575 * and updates *next_id* on success. If no other BTF objects
576 * remain with ids higher than *start_id*, returns -1 and sets
577 * *errno* to **ENOENT**.
578 *
579 * Return
580 * Returns zero on success. On error, or when no id remains, -1
581 * is returned and *errno* is set appropriately.
582 *
583 * BPF_MAP_LOOKUP_BATCH
584 * Description
585 * Iterate and fetch multiple elements in a map.
586 *
587 * Two opaque values are used to manage batch operations,
588 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
589 * to NULL to begin the batched operation. After each subsequent
590 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
591 * *out_batch* as the *in_batch* for the next operation to
592 * continue iteration from the current point.
593 *
594 * The *keys* and *values* are output parameters which must point
595 * to memory large enough to hold *count* items based on the key
596 * and value size of the map *map_fd*. The *keys* buffer must be
597 * of *key_size* * *count*. The *values* buffer must be of
598 * *value_size* * *count*.
599 *
600 * The *elem_flags* argument may be specified as one of the
601 * following:
602 *
603 * **BPF_F_LOCK**
604 * Look up the value of a spin-locked map without
605 * returning the lock. This must be specified if the
606 * elements contain a spinlock.
607 *
608 * On success, *count* elements from the map are copied into the
609 * user buffer, with the keys copied into *keys* and the values
610 * copied into the corresponding indices in *values*.
611 *
612 * If an error is returned and *errno* is not **EFAULT**, *count*
613 * is set to the number of successfully processed elements.
614 *
615 * Return
616 * Returns zero on success. On error, -1 is returned and *errno*
617 * is set appropriately.
618 *
619 * May set *errno* to **ENOSPC** to indicate that *keys* or
620 * *values* is too small to dump an entire bucket during
621 * iteration of a hash-based map type.
622 *
623 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
624 * Description
625 * Iterate and delete all elements in a map.
626 *
627 * This operation has the same behavior as
628 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
629 *
630 * * Every element that is successfully returned is also deleted
631 * from the map. This is at least *count* elements. Note that
632 * *count* is both an input and an output parameter.
633 * * Upon returning with *errno* set to **EFAULT**, up to
634 * *count* elements may be deleted without returning the keys
635 * and values of the deleted elements.
636 *
637 * Return
638 * Returns zero on success. On error, -1 is returned and *errno*
639 * is set appropriately.
640 *
641 * BPF_MAP_UPDATE_BATCH
642 * Description
643 * Update multiple elements in a map by *key*.
644 *
645 * The *keys* and *values* are input parameters which must point
646 * to memory large enough to hold *count* items based on the key
647 * and value size of the map *map_fd*. The *keys* buffer must be
648 * of *key_size* * *count*. The *values* buffer must be of
649 * *value_size* * *count*.
650 *
651 * Each element specified in *keys* is sequentially updated to the
652 * value in the corresponding index in *values*. The *in_batch*
653 * and *out_batch* parameters are ignored and should be zeroed.
654 *
655 * The *elem_flags* argument should be specified as one of the
656 * following:
657 *
658 * **BPF_ANY**
659 * Create new elements or update a existing elements.
660 * **BPF_NOEXIST**
661 * Create new elements only if they do not exist.
662 * **BPF_EXIST**
663 * Update existing elements.
664 * **BPF_F_LOCK**
665 * Update spin_lock-ed map elements. This must be
666 * specified if the map value contains a spinlock.
667 *
668 * On success, *count* elements from the map are updated.
669 *
670 * If an error is returned and *errno* is not **EFAULT**, *count*
671 * is set to the number of successfully processed elements.
672 *
673 * Return
674 * Returns zero on success. On error, -1 is returned and *errno*
675 * is set appropriately.
676 *
677 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
678 * **E2BIG**. **E2BIG** indicates that the number of elements in
679 * the map reached the *max_entries* limit specified at map
680 * creation time.
681 *
682 * May set *errno* to one of the following error codes under
683 * specific circumstances:
684 *
685 * **EEXIST**
686 * If *flags* specifies **BPF_NOEXIST** and the element
687 * with *key* already exists in the map.
688 * **ENOENT**
689 * If *flags* specifies **BPF_EXIST** and the element with
690 * *key* does not exist in the map.
691 *
692 * BPF_MAP_DELETE_BATCH
693 * Description
694 * Delete multiple elements in a map by *key*.
695 *
696 * The *keys* parameter is an input parameter which must point
697 * to memory large enough to hold *count* items based on the key
698 * size of the map *map_fd*, that is, *key_size* * *count*.
699 *
700 * Each element specified in *keys* is sequentially deleted. The
701 * *in_batch*, *out_batch*, and *values* parameters are ignored
702 * and should be zeroed.
703 *
704 * The *elem_flags* argument may be specified as one of the
705 * following:
706 *
707 * **BPF_F_LOCK**
708 * Look up the value of a spin-locked map without
709 * returning the lock. This must be specified if the
710 * elements contain a spinlock.
711 *
712 * On success, *count* elements from the map are updated.
713 *
714 * If an error is returned and *errno* is not **EFAULT**, *count*
715 * is set to the number of successfully processed elements. If
716 * *errno* is **EFAULT**, up to *count* elements may be been
717 * deleted.
718 *
719 * Return
720 * Returns zero on success. On error, -1 is returned and *errno*
721 * is set appropriately.
722 *
723 * BPF_LINK_CREATE
724 * Description
725 * Attach an eBPF program to a *target_fd* at the specified
726 * *attach_type* hook and return a file descriptor handle for
727 * managing the link.
728 *
729 * Return
730 * A new file descriptor (a nonnegative integer), or -1 if an
731 * error occurred (in which case, *errno* is set appropriately).
732 *
733 * BPF_LINK_UPDATE
734 * Description
735 * Update the eBPF program in the specified *link_fd* to
736 * *new_prog_fd*.
737 *
738 * Return
739 * Returns zero on success. On error, -1 is returned and *errno*
740 * is set appropriately.
741 *
742 * BPF_LINK_GET_FD_BY_ID
743 * Description
744 * Open a file descriptor for the eBPF Link corresponding to
745 * *link_id*.
746 *
747 * Return
748 * A new file descriptor (a nonnegative integer), or -1 if an
749 * error occurred (in which case, *errno* is set appropriately).
750 *
751 * BPF_LINK_GET_NEXT_ID
752 * Description
753 * Fetch the next eBPF link currently loaded into the kernel.
754 *
755 * Looks for the eBPF link with an id greater than *start_id*
756 * and updates *next_id* on success. If no other eBPF links
757 * remain with ids higher than *start_id*, returns -1 and sets
758 * *errno* to **ENOENT**.
759 *
760 * Return
761 * Returns zero on success. On error, or when no id remains, -1
762 * is returned and *errno* is set appropriately.
763 *
764 * BPF_ENABLE_STATS
765 * Description
766 * Enable eBPF runtime statistics gathering.
767 *
768 * Runtime statistics gathering for the eBPF runtime is disabled
769 * by default to minimize the corresponding performance overhead.
770 * This command enables statistics globally.
771 *
772 * Multiple programs may independently enable statistics.
773 * After gathering the desired statistics, eBPF runtime statistics
774 * may be disabled again by calling **close**\ (2) for the file
775 * descriptor returned by this function. Statistics will only be
776 * disabled system-wide when all outstanding file descriptors
777 * returned by prior calls for this subcommand are closed.
778 *
779 * Return
780 * A new file descriptor (a nonnegative integer), or -1 if an
781 * error occurred (in which case, *errno* is set appropriately).
782 *
783 * BPF_ITER_CREATE
784 * Description
785 * Create an iterator on top of the specified *link_fd* (as
786 * previously created using **BPF_LINK_CREATE**) and return a
787 * file descriptor that can be used to trigger the iteration.
788 *
789 * If the resulting file descriptor is pinned to the filesystem
790 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
791 * for that path will trigger the iterator to read kernel state
792 * using the eBPF program attached to *link_fd*.
793 *
794 * Return
795 * A new file descriptor (a nonnegative integer), or -1 if an
796 * error occurred (in which case, *errno* is set appropriately).
797 *
798 * BPF_LINK_DETACH
799 * Description
800 * Forcefully detach the specified *link_fd* from its
801 * corresponding attachment point.
802 *
803 * Return
804 * Returns zero on success. On error, -1 is returned and *errno*
805 * is set appropriately.
806 *
807 * BPF_PROG_BIND_MAP
808 * Description
809 * Bind a map to the lifetime of an eBPF program.
810 *
811 * The map identified by *map_fd* is bound to the program
812 * identified by *prog_fd* and only released when *prog_fd* is
813 * released. This may be used in cases where metadata should be
814 * associated with a program which otherwise does not contain any
815 * references to the map (for example, embedded in the eBPF
816 * program instructions).
817 *
818 * Return
819 * Returns zero on success. On error, -1 is returned and *errno*
820 * is set appropriately.
821 *
822 * NOTES
823 * eBPF objects (maps and programs) can be shared between processes.
824 *
825 * * After **fork**\ (2), the child inherits file descriptors
826 * referring to the same eBPF objects.
827 * * File descriptors referring to eBPF objects can be transferred over
828 * **unix**\ (7) domain sockets.
829 * * File descriptors referring to eBPF objects can be duplicated in the
830 * usual way, using **dup**\ (2) and similar calls.
831 * * File descriptors referring to eBPF objects can be pinned to the
832 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
833 *
834 * An eBPF object is deallocated only after all file descriptors referring
835 * to the object have been closed and no references remain pinned to the
836 * filesystem or attached (for example, bound to a program or device).
837 */
838enum bpf_cmd {
839 BPF_MAP_CREATE,
840 BPF_MAP_LOOKUP_ELEM,
841 BPF_MAP_UPDATE_ELEM,
842 BPF_MAP_DELETE_ELEM,
843 BPF_MAP_GET_NEXT_KEY,
844 BPF_PROG_LOAD,
845 BPF_OBJ_PIN,
846 BPF_OBJ_GET,
847 BPF_PROG_ATTACH,
848 BPF_PROG_DETACH,
849 BPF_PROG_TEST_RUN,
850 BPF_PROG_RUN = BPF_PROG_TEST_RUN,
851 BPF_PROG_GET_NEXT_ID,
852 BPF_MAP_GET_NEXT_ID,
853 BPF_PROG_GET_FD_BY_ID,
854 BPF_MAP_GET_FD_BY_ID,
855 BPF_OBJ_GET_INFO_BY_FD,
856 BPF_PROG_QUERY,
857 BPF_RAW_TRACEPOINT_OPEN,
858 BPF_BTF_LOAD,
859 BPF_BTF_GET_FD_BY_ID,
860 BPF_TASK_FD_QUERY,
861 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
862 BPF_MAP_FREEZE,
863 BPF_BTF_GET_NEXT_ID,
864 BPF_MAP_LOOKUP_BATCH,
865 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
866 BPF_MAP_UPDATE_BATCH,
867 BPF_MAP_DELETE_BATCH,
868 BPF_LINK_CREATE,
869 BPF_LINK_UPDATE,
870 BPF_LINK_GET_FD_BY_ID,
871 BPF_LINK_GET_NEXT_ID,
872 BPF_ENABLE_STATS,
873 BPF_ITER_CREATE,
874 BPF_LINK_DETACH,
875 BPF_PROG_BIND_MAP,
876};
877
878enum bpf_map_type {
879 BPF_MAP_TYPE_UNSPEC,
880 BPF_MAP_TYPE_HASH,
881 BPF_MAP_TYPE_ARRAY,
882 BPF_MAP_TYPE_PROG_ARRAY,
883 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
884 BPF_MAP_TYPE_PERCPU_HASH,
885 BPF_MAP_TYPE_PERCPU_ARRAY,
886 BPF_MAP_TYPE_STACK_TRACE,
887 BPF_MAP_TYPE_CGROUP_ARRAY,
888 BPF_MAP_TYPE_LRU_HASH,
889 BPF_MAP_TYPE_LRU_PERCPU_HASH,
890 BPF_MAP_TYPE_LPM_TRIE,
891 BPF_MAP_TYPE_ARRAY_OF_MAPS,
892 BPF_MAP_TYPE_HASH_OF_MAPS,
893 BPF_MAP_TYPE_DEVMAP,
894 BPF_MAP_TYPE_SOCKMAP,
895 BPF_MAP_TYPE_CPUMAP,
896 BPF_MAP_TYPE_XSKMAP,
897 BPF_MAP_TYPE_SOCKHASH,
898 BPF_MAP_TYPE_CGROUP_STORAGE,
899 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
900 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
901 BPF_MAP_TYPE_QUEUE,
902 BPF_MAP_TYPE_STACK,
903 BPF_MAP_TYPE_SK_STORAGE,
904 BPF_MAP_TYPE_DEVMAP_HASH,
905 BPF_MAP_TYPE_STRUCT_OPS,
906 BPF_MAP_TYPE_RINGBUF,
907 BPF_MAP_TYPE_INODE_STORAGE,
908 BPF_MAP_TYPE_TASK_STORAGE,
909};
910
911/* Note that tracing related programs such as
912 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
913 * are not subject to a stable API since kernel internal data
914 * structures can change from release to release and may
915 * therefore break existing tracing BPF programs. Tracing BPF
916 * programs correspond to /a/ specific kernel which is to be
917 * analyzed, and not /a/ specific kernel /and/ all future ones.
918 */
919enum bpf_prog_type {
920 BPF_PROG_TYPE_UNSPEC,
921 BPF_PROG_TYPE_SOCKET_FILTER,
922 BPF_PROG_TYPE_KPROBE,
923 BPF_PROG_TYPE_SCHED_CLS,
924 BPF_PROG_TYPE_SCHED_ACT,
925 BPF_PROG_TYPE_TRACEPOINT,
926 BPF_PROG_TYPE_XDP,
927 BPF_PROG_TYPE_PERF_EVENT,
928 BPF_PROG_TYPE_CGROUP_SKB,
929 BPF_PROG_TYPE_CGROUP_SOCK,
930 BPF_PROG_TYPE_LWT_IN,
931 BPF_PROG_TYPE_LWT_OUT,
932 BPF_PROG_TYPE_LWT_XMIT,
933 BPF_PROG_TYPE_SOCK_OPS,
934 BPF_PROG_TYPE_SK_SKB,
935 BPF_PROG_TYPE_CGROUP_DEVICE,
936 BPF_PROG_TYPE_SK_MSG,
937 BPF_PROG_TYPE_RAW_TRACEPOINT,
938 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
939 BPF_PROG_TYPE_LWT_SEG6LOCAL,
940 BPF_PROG_TYPE_LIRC_MODE2,
941 BPF_PROG_TYPE_SK_REUSEPORT,
942 BPF_PROG_TYPE_FLOW_DISSECTOR,
943 BPF_PROG_TYPE_CGROUP_SYSCTL,
944 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
945 BPF_PROG_TYPE_CGROUP_SOCKOPT,
946 BPF_PROG_TYPE_TRACING,
947 BPF_PROG_TYPE_STRUCT_OPS,
948 BPF_PROG_TYPE_EXT,
949 BPF_PROG_TYPE_LSM,
950 BPF_PROG_TYPE_SK_LOOKUP,
951 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
952};
953
954enum bpf_attach_type {
955 BPF_CGROUP_INET_INGRESS,
956 BPF_CGROUP_INET_EGRESS,
957 BPF_CGROUP_INET_SOCK_CREATE,
958 BPF_CGROUP_SOCK_OPS,
959 BPF_SK_SKB_STREAM_PARSER,
960 BPF_SK_SKB_STREAM_VERDICT,
961 BPF_CGROUP_DEVICE,
962 BPF_SK_MSG_VERDICT,
963 BPF_CGROUP_INET4_BIND,
964 BPF_CGROUP_INET6_BIND,
965 BPF_CGROUP_INET4_CONNECT,
966 BPF_CGROUP_INET6_CONNECT,
967 BPF_CGROUP_INET4_POST_BIND,
968 BPF_CGROUP_INET6_POST_BIND,
969 BPF_CGROUP_UDP4_SENDMSG,
970 BPF_CGROUP_UDP6_SENDMSG,
971 BPF_LIRC_MODE2,
972 BPF_FLOW_DISSECTOR,
973 BPF_CGROUP_SYSCTL,
974 BPF_CGROUP_UDP4_RECVMSG,
975 BPF_CGROUP_UDP6_RECVMSG,
976 BPF_CGROUP_GETSOCKOPT,
977 BPF_CGROUP_SETSOCKOPT,
978 BPF_TRACE_RAW_TP,
979 BPF_TRACE_FENTRY,
980 BPF_TRACE_FEXIT,
981 BPF_MODIFY_RETURN,
982 BPF_LSM_MAC,
983 BPF_TRACE_ITER,
984 BPF_CGROUP_INET4_GETPEERNAME,
985 BPF_CGROUP_INET6_GETPEERNAME,
986 BPF_CGROUP_INET4_GETSOCKNAME,
987 BPF_CGROUP_INET6_GETSOCKNAME,
988 BPF_XDP_DEVMAP,
989 BPF_CGROUP_INET_SOCK_RELEASE,
990 BPF_XDP_CPUMAP,
991 BPF_SK_LOOKUP,
992 BPF_XDP,
993 BPF_SK_SKB_VERDICT,
994 BPF_SK_REUSEPORT_SELECT,
995 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
996 BPF_PERF_EVENT,
997 __MAX_BPF_ATTACH_TYPE
998};
999
1000#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1001
1002enum bpf_link_type {
1003 BPF_LINK_TYPE_UNSPEC = 0,
1004 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1005 BPF_LINK_TYPE_TRACING = 2,
1006 BPF_LINK_TYPE_CGROUP = 3,
1007 BPF_LINK_TYPE_ITER = 4,
1008 BPF_LINK_TYPE_NETNS = 5,
1009 BPF_LINK_TYPE_XDP = 6,
1010 BPF_LINK_TYPE_PERF_EVENT = 7,
1011
1012 MAX_BPF_LINK_TYPE,
1013};
1014
1015/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1016 *
1017 * NONE(default): No further bpf programs allowed in the subtree.
1018 *
1019 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1020 * the program in this cgroup yields to sub-cgroup program.
1021 *
1022 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1023 * that cgroup program gets run in addition to the program in this cgroup.
1024 *
1025 * Only one program is allowed to be attached to a cgroup with
1026 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1027 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1028 * release old program and attach the new one. Attach flags has to match.
1029 *
1030 * Multiple programs are allowed to be attached to a cgroup with
1031 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1032 * (those that were attached first, run first)
1033 * The programs of sub-cgroup are executed first, then programs of
1034 * this cgroup and then programs of parent cgroup.
1035 * When children program makes decision (like picking TCP CA or sock bind)
1036 * parent program has a chance to override it.
1037 *
1038 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1039 * programs for a cgroup. Though it's possible to replace an old program at
1040 * any position by also specifying BPF_F_REPLACE flag and position itself in
1041 * replace_bpf_fd attribute. Old program at this position will be released.
1042 *
1043 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1044 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1045 * Ex1:
1046 * cgrp1 (MULTI progs A, B) ->
1047 * cgrp2 (OVERRIDE prog C) ->
1048 * cgrp3 (MULTI prog D) ->
1049 * cgrp4 (OVERRIDE prog E) ->
1050 * cgrp5 (NONE prog F)
1051 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1052 * if prog F is detached, the execution is E,D,A,B
1053 * if prog F and D are detached, the execution is E,A,B
1054 * if prog F, E and D are detached, the execution is C,A,B
1055 *
1056 * All eligible programs are executed regardless of return code from
1057 * earlier programs.
1058 */
1059#define BPF_F_ALLOW_OVERRIDE (1U << 0)
1060#define BPF_F_ALLOW_MULTI (1U << 1)
1061#define BPF_F_REPLACE (1U << 2)
1062
1063/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1064 * verifier will perform strict alignment checking as if the kernel
1065 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1066 * and NET_IP_ALIGN defined to 2.
1067 */
1068#define BPF_F_STRICT_ALIGNMENT (1U << 0)
1069
1070/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1071 * verifier will allow any alignment whatsoever. On platforms
1072 * with strict alignment requirements for loads ands stores (such
1073 * as sparc and mips) the verifier validates that all loads and
1074 * stores provably follow this requirement. This flag turns that
1075 * checking and enforcement off.
1076 *
1077 * It is mostly used for testing when we want to validate the
1078 * context and memory access aspects of the verifier, but because
1079 * of an unaligned access the alignment check would trigger before
1080 * the one we are interested in.
1081 */
1082#define BPF_F_ANY_ALIGNMENT (1U << 1)
1083
1084/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1085 * Verifier does sub-register def/use analysis and identifies instructions whose
1086 * def only matters for low 32-bit, high 32-bit is never referenced later
1087 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1088 * that it is safe to ignore clearing high 32-bit for these instructions. This
1089 * saves some back-ends a lot of code-gen. However such optimization is not
1090 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1091 * hence hasn't used verifier's analysis result. But, we really want to have a
1092 * way to be able to verify the correctness of the described optimization on
1093 * x86_64 on which testsuites are frequently exercised.
1094 *
1095 * So, this flag is introduced. Once it is set, verifier will randomize high
1096 * 32-bit for those instructions who has been identified as safe to ignore them.
1097 * Then, if verifier is not doing correct analysis, such randomization will
1098 * regress tests to expose bugs.
1099 */
1100#define BPF_F_TEST_RND_HI32 (1U << 2)
1101
1102/* The verifier internal test flag. Behavior is undefined */
1103#define BPF_F_TEST_STATE_FREQ (1U << 3)
1104
1105/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1106 * restrict map and helper usage for such programs. Sleepable BPF programs can
1107 * only be attached to hooks where kernel execution context allows sleeping.
1108 * Such programs are allowed to use helpers that may sleep like
1109 * bpf_copy_from_user().
1110 */
1111#define BPF_F_SLEEPABLE (1U << 4)
1112
1113/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1114 * the following extensions:
1115 *
1116 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1117 * insn[0].imm: map fd or fd_idx
1118 * insn[1].imm: 0
1119 * insn[0].off: 0
1120 * insn[1].off: 0
1121 * ldimm64 rewrite: address of map
1122 * verifier type: CONST_PTR_TO_MAP
1123 */
1124#define BPF_PSEUDO_MAP_FD 1
1125#define BPF_PSEUDO_MAP_IDX 5
1126
1127/* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1128 * insn[0].imm: map fd or fd_idx
1129 * insn[1].imm: offset into value
1130 * insn[0].off: 0
1131 * insn[1].off: 0
1132 * ldimm64 rewrite: address of map[0]+offset
1133 * verifier type: PTR_TO_MAP_VALUE
1134 */
1135#define BPF_PSEUDO_MAP_VALUE 2
1136#define BPF_PSEUDO_MAP_IDX_VALUE 6
1137
1138/* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1139 * insn[0].imm: kernel btd id of VAR
1140 * insn[1].imm: 0
1141 * insn[0].off: 0
1142 * insn[1].off: 0
1143 * ldimm64 rewrite: address of the kernel variable
1144 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1145 * is struct/union.
1146 */
1147#define BPF_PSEUDO_BTF_ID 3
1148/* insn[0].src_reg: BPF_PSEUDO_FUNC
1149 * insn[0].imm: insn offset to the func
1150 * insn[1].imm: 0
1151 * insn[0].off: 0
1152 * insn[1].off: 0
1153 * ldimm64 rewrite: address of the function
1154 * verifier type: PTR_TO_FUNC.
1155 */
1156#define BPF_PSEUDO_FUNC 4
1157
1158/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1159 * offset to another bpf function
1160 */
1161#define BPF_PSEUDO_CALL 1
1162/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1163 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1164 */
1165#define BPF_PSEUDO_KFUNC_CALL 2
1166
1167/* flags for BPF_MAP_UPDATE_ELEM command */
1168enum {
1169 BPF_ANY = 0, /* create new element or update existing */
1170 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1171 BPF_EXIST = 2, /* update existing element */
1172 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1173};
1174
1175/* flags for BPF_MAP_CREATE command */
1176enum {
1177 BPF_F_NO_PREALLOC = (1U << 0),
1178/* Instead of having one common LRU list in the
1179 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1180 * which can scale and perform better.
1181 * Note, the LRU nodes (including free nodes) cannot be moved
1182 * across different LRU lists.
1183 */
1184 BPF_F_NO_COMMON_LRU = (1U << 1),
1185/* Specify numa node during map creation */
1186 BPF_F_NUMA_NODE = (1U << 2),
1187
1188/* Flags for accessing BPF object from syscall side. */
1189 BPF_F_RDONLY = (1U << 3),
1190 BPF_F_WRONLY = (1U << 4),
1191
1192/* Flag for stack_map, store build_id+offset instead of pointer */
1193 BPF_F_STACK_BUILD_ID = (1U << 5),
1194
1195/* Zero-initialize hash function seed. This should only be used for testing. */
1196 BPF_F_ZERO_SEED = (1U << 6),
1197
1198/* Flags for accessing BPF object from program side. */
1199 BPF_F_RDONLY_PROG = (1U << 7),
1200 BPF_F_WRONLY_PROG = (1U << 8),
1201
1202/* Clone map from listener for newly accepted socket */
1203 BPF_F_CLONE = (1U << 9),
1204
1205/* Enable memory-mapping BPF map */
1206 BPF_F_MMAPABLE = (1U << 10),
1207
1208/* Share perf_event among processes */
1209 BPF_F_PRESERVE_ELEMS = (1U << 11),
1210
1211/* Create a map that is suitable to be an inner map with dynamic max entries */
1212 BPF_F_INNER_MAP = (1U << 12),
1213};
1214
1215/* Flags for BPF_PROG_QUERY. */
1216
1217/* Query effective (directly attached + inherited from ancestor cgroups)
1218 * programs that will be executed for events within a cgroup.
1219 * attach_flags with this flag are returned only for directly attached programs.
1220 */
1221#define BPF_F_QUERY_EFFECTIVE (1U << 0)
1222
1223/* Flags for BPF_PROG_TEST_RUN */
1224
1225/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1226#define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1227
1228/* type for BPF_ENABLE_STATS */
1229enum bpf_stats_type {
1230 /* enabled run_time_ns and run_cnt */
1231 BPF_STATS_RUN_TIME = 0,
1232};
1233
1234enum bpf_stack_build_id_status {
1235 /* user space need an empty entry to identify end of a trace */
1236 BPF_STACK_BUILD_ID_EMPTY = 0,
1237 /* with valid build_id and offset */
1238 BPF_STACK_BUILD_ID_VALID = 1,
1239 /* couldn't get build_id, fallback to ip */
1240 BPF_STACK_BUILD_ID_IP = 2,
1241};
1242
1243#define BPF_BUILD_ID_SIZE 20
1244struct bpf_stack_build_id {
1245 __s32 status;
1246 unsigned char build_id[BPF_BUILD_ID_SIZE];
1247 union {
1248 __u64 offset;
1249 __u64 ip;
1250 };
1251};
1252
1253#define BPF_OBJ_NAME_LEN 16U
1254
1255union bpf_attr {
1256 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1257 __u32 map_type; /* one of enum bpf_map_type */
1258 __u32 key_size; /* size of key in bytes */
1259 __u32 value_size; /* size of value in bytes */
1260 __u32 max_entries; /* max number of entries in a map */
1261 __u32 map_flags; /* BPF_MAP_CREATE related
1262 * flags defined above.
1263 */
1264 __u32 inner_map_fd; /* fd pointing to the inner map */
1265 __u32 numa_node; /* numa node (effective only if
1266 * BPF_F_NUMA_NODE is set).
1267 */
1268 char map_name[BPF_OBJ_NAME_LEN];
1269 __u32 map_ifindex; /* ifindex of netdev to create on */
1270 __u32 btf_fd; /* fd pointing to a BTF type data */
1271 __u32 btf_key_type_id; /* BTF type_id of the key */
1272 __u32 btf_value_type_id; /* BTF type_id of the value */
1273 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1274 * struct stored as the
1275 * map value
1276 */
1277 };
1278
1279 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1280 __u32 map_fd;
1281 __aligned_u64 key;
1282 union {
1283 __aligned_u64 value;
1284 __aligned_u64 next_key;
1285 };
1286 __u64 flags;
1287 };
1288
1289 struct { /* struct used by BPF_MAP_*_BATCH commands */
1290 __aligned_u64 in_batch; /* start batch,
1291 * NULL to start from beginning
1292 */
1293 __aligned_u64 out_batch; /* output: next start batch */
1294 __aligned_u64 keys;
1295 __aligned_u64 values;
1296 __u32 count; /* input/output:
1297 * input: # of key/value
1298 * elements
1299 * output: # of filled elements
1300 */
1301 __u32 map_fd;
1302 __u64 elem_flags;
1303 __u64 flags;
1304 } batch;
1305
1306 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1307 __u32 prog_type; /* one of enum bpf_prog_type */
1308 __u32 insn_cnt;
1309 __aligned_u64 insns;
1310 __aligned_u64 license;
1311 __u32 log_level; /* verbosity level of verifier */
1312 __u32 log_size; /* size of user buffer */
1313 __aligned_u64 log_buf; /* user supplied buffer */
1314 __u32 kern_version; /* not used */
1315 __u32 prog_flags;
1316 char prog_name[BPF_OBJ_NAME_LEN];
1317 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1318 /* For some prog types expected attach type must be known at
1319 * load time to verify attach type specific parts of prog
1320 * (context accesses, allowed helpers, etc).
1321 */
1322 __u32 expected_attach_type;
1323 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1324 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1325 __aligned_u64 func_info; /* func info */
1326 __u32 func_info_cnt; /* number of bpf_func_info records */
1327 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1328 __aligned_u64 line_info; /* line info */
1329 __u32 line_info_cnt; /* number of bpf_line_info records */
1330 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1331 union {
1332 /* valid prog_fd to attach to bpf prog */
1333 __u32 attach_prog_fd;
1334 /* or valid module BTF object fd or 0 to attach to vmlinux */
1335 __u32 attach_btf_obj_fd;
1336 };
1337 __u32 :32; /* pad */
1338 __aligned_u64 fd_array; /* array of FDs */
1339 };
1340
1341 struct { /* anonymous struct used by BPF_OBJ_* commands */
1342 __aligned_u64 pathname;
1343 __u32 bpf_fd;
1344 __u32 file_flags;
1345 };
1346
1347 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1348 __u32 target_fd; /* container object to attach to */
1349 __u32 attach_bpf_fd; /* eBPF program to attach */
1350 __u32 attach_type;
1351 __u32 attach_flags;
1352 __u32 replace_bpf_fd; /* previously attached eBPF
1353 * program to replace if
1354 * BPF_F_REPLACE is used
1355 */
1356 };
1357
1358 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1359 __u32 prog_fd;
1360 __u32 retval;
1361 __u32 data_size_in; /* input: len of data_in */
1362 __u32 data_size_out; /* input/output: len of data_out
1363 * returns ENOSPC if data_out
1364 * is too small.
1365 */
1366 __aligned_u64 data_in;
1367 __aligned_u64 data_out;
1368 __u32 repeat;
1369 __u32 duration;
1370 __u32 ctx_size_in; /* input: len of ctx_in */
1371 __u32 ctx_size_out; /* input/output: len of ctx_out
1372 * returns ENOSPC if ctx_out
1373 * is too small.
1374 */
1375 __aligned_u64 ctx_in;
1376 __aligned_u64 ctx_out;
1377 __u32 flags;
1378 __u32 cpu;
1379 } test;
1380
1381 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1382 union {
1383 __u32 start_id;
1384 __u32 prog_id;
1385 __u32 map_id;
1386 __u32 btf_id;
1387 __u32 link_id;
1388 };
1389 __u32 next_id;
1390 __u32 open_flags;
1391 };
1392
1393 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1394 __u32 bpf_fd;
1395 __u32 info_len;
1396 __aligned_u64 info;
1397 } info;
1398
1399 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1400 __u32 target_fd; /* container object to query */
1401 __u32 attach_type;
1402 __u32 query_flags;
1403 __u32 attach_flags;
1404 __aligned_u64 prog_ids;
1405 __u32 prog_cnt;
1406 } query;
1407
1408 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1409 __u64 name;
1410 __u32 prog_fd;
1411 } raw_tracepoint;
1412
1413 struct { /* anonymous struct for BPF_BTF_LOAD */
1414 __aligned_u64 btf;
1415 __aligned_u64 btf_log_buf;
1416 __u32 btf_size;
1417 __u32 btf_log_size;
1418 __u32 btf_log_level;
1419 };
1420
1421 struct {
1422 __u32 pid; /* input: pid */
1423 __u32 fd; /* input: fd */
1424 __u32 flags; /* input: flags */
1425 __u32 buf_len; /* input/output: buf len */
1426 __aligned_u64 buf; /* input/output:
1427 * tp_name for tracepoint
1428 * symbol for kprobe
1429 * filename for uprobe
1430 */
1431 __u32 prog_id; /* output: prod_id */
1432 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1433 __u64 probe_offset; /* output: probe_offset */
1434 __u64 probe_addr; /* output: probe_addr */
1435 } task_fd_query;
1436
1437 struct { /* struct used by BPF_LINK_CREATE command */
1438 __u32 prog_fd; /* eBPF program to attach */
1439 union {
1440 __u32 target_fd; /* object to attach to */
1441 __u32 target_ifindex; /* target ifindex */
1442 };
1443 __u32 attach_type; /* attach type */
1444 __u32 flags; /* extra flags */
1445 union {
1446 __u32 target_btf_id; /* btf_id of target to attach to */
1447 struct {
1448 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1449 __u32 iter_info_len; /* iter_info length */
1450 };
1451 struct {
1452 /* black box user-provided value passed through
1453 * to BPF program at the execution time and
1454 * accessible through bpf_get_attach_cookie() BPF helper
1455 */
1456 __u64 bpf_cookie;
1457 } perf_event;
1458 };
1459 } link_create;
1460
1461 struct { /* struct used by BPF_LINK_UPDATE command */
1462 __u32 link_fd; /* link fd */
1463 /* new program fd to update link with */
1464 __u32 new_prog_fd;
1465 __u32 flags; /* extra flags */
1466 /* expected link's program fd; is specified only if
1467 * BPF_F_REPLACE flag is set in flags */
1468 __u32 old_prog_fd;
1469 } link_update;
1470
1471 struct {
1472 __u32 link_fd;
1473 } link_detach;
1474
1475 struct { /* struct used by BPF_ENABLE_STATS command */
1476 __u32 type;
1477 } enable_stats;
1478
1479 struct { /* struct used by BPF_ITER_CREATE command */
1480 __u32 link_fd;
1481 __u32 flags;
1482 } iter_create;
1483
1484 struct { /* struct used by BPF_PROG_BIND_MAP command */
1485 __u32 prog_fd;
1486 __u32 map_fd;
1487 __u32 flags; /* extra flags */
1488 } prog_bind_map;
1489
1490} __attribute__((aligned(8)));
1491
1492/* The description below is an attempt at providing documentation to eBPF
1493 * developers about the multiple available eBPF helper functions. It can be
1494 * parsed and used to produce a manual page. The workflow is the following,
1495 * and requires the rst2man utility:
1496 *
1497 * $ ./scripts/bpf_doc.py \
1498 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1499 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1500 * $ man /tmp/bpf-helpers.7
1501 *
1502 * Note that in order to produce this external documentation, some RST
1503 * formatting is used in the descriptions to get "bold" and "italics" in
1504 * manual pages. Also note that the few trailing white spaces are
1505 * intentional, removing them would break paragraphs for rst2man.
1506 *
1507 * Start of BPF helper function descriptions:
1508 *
1509 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1510 * Description
1511 * Perform a lookup in *map* for an entry associated to *key*.
1512 * Return
1513 * Map value associated to *key*, or **NULL** if no entry was
1514 * found.
1515 *
1516 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1517 * Description
1518 * Add or update the value of the entry associated to *key* in
1519 * *map* with *value*. *flags* is one of:
1520 *
1521 * **BPF_NOEXIST**
1522 * The entry for *key* must not exist in the map.
1523 * **BPF_EXIST**
1524 * The entry for *key* must already exist in the map.
1525 * **BPF_ANY**
1526 * No condition on the existence of the entry for *key*.
1527 *
1528 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1529 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1530 * elements always exist), the helper would return an error.
1531 * Return
1532 * 0 on success, or a negative error in case of failure.
1533 *
1534 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1535 * Description
1536 * Delete entry with *key* from *map*.
1537 * Return
1538 * 0 on success, or a negative error in case of failure.
1539 *
1540 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1541 * Description
1542 * For tracing programs, safely attempt to read *size* bytes from
1543 * kernel space address *unsafe_ptr* and store the data in *dst*.
1544 *
1545 * Generally, use **bpf_probe_read_user**\ () or
1546 * **bpf_probe_read_kernel**\ () instead.
1547 * Return
1548 * 0 on success, or a negative error in case of failure.
1549 *
1550 * u64 bpf_ktime_get_ns(void)
1551 * Description
1552 * Return the time elapsed since system boot, in nanoseconds.
1553 * Does not include time the system was suspended.
1554 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1555 * Return
1556 * Current *ktime*.
1557 *
1558 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1559 * Description
1560 * This helper is a "printk()-like" facility for debugging. It
1561 * prints a message defined by format *fmt* (of size *fmt_size*)
1562 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1563 * available. It can take up to three additional **u64**
1564 * arguments (as an eBPF helpers, the total number of arguments is
1565 * limited to five).
1566 *
1567 * Each time the helper is called, it appends a line to the trace.
1568 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1569 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1570 * The format of the trace is customizable, and the exact output
1571 * one will get depends on the options set in
1572 * *\/sys/kernel/debug/tracing/trace_options* (see also the
1573 * *README* file under the same directory). However, it usually
1574 * defaults to something like:
1575 *
1576 * ::
1577 *
1578 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1579 *
1580 * In the above:
1581 *
1582 * * ``telnet`` is the name of the current task.
1583 * * ``470`` is the PID of the current task.
1584 * * ``001`` is the CPU number on which the task is
1585 * running.
1586 * * In ``.N..``, each character refers to a set of
1587 * options (whether irqs are enabled, scheduling
1588 * options, whether hard/softirqs are running, level of
1589 * preempt_disabled respectively). **N** means that
1590 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1591 * are set.
1592 * * ``419421.045894`` is a timestamp.
1593 * * ``0x00000001`` is a fake value used by BPF for the
1594 * instruction pointer register.
1595 * * ``<formatted msg>`` is the message formatted with
1596 * *fmt*.
1597 *
1598 * The conversion specifiers supported by *fmt* are similar, but
1599 * more limited than for printk(). They are **%d**, **%i**,
1600 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1601 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1602 * of field, padding with zeroes, etc.) is available, and the
1603 * helper will return **-EINVAL** (but print nothing) if it
1604 * encounters an unknown specifier.
1605 *
1606 * Also, note that **bpf_trace_printk**\ () is slow, and should
1607 * only be used for debugging purposes. For this reason, a notice
1608 * block (spanning several lines) is printed to kernel logs and
1609 * states that the helper should not be used "for production use"
1610 * the first time this helper is used (or more precisely, when
1611 * **trace_printk**\ () buffers are allocated). For passing values
1612 * to user space, perf events should be preferred.
1613 * Return
1614 * The number of bytes written to the buffer, or a negative error
1615 * in case of failure.
1616 *
1617 * u32 bpf_get_prandom_u32(void)
1618 * Description
1619 * Get a pseudo-random number.
1620 *
1621 * From a security point of view, this helper uses its own
1622 * pseudo-random internal state, and cannot be used to infer the
1623 * seed of other random functions in the kernel. However, it is
1624 * essential to note that the generator used by the helper is not
1625 * cryptographically secure.
1626 * Return
1627 * A random 32-bit unsigned value.
1628 *
1629 * u32 bpf_get_smp_processor_id(void)
1630 * Description
1631 * Get the SMP (symmetric multiprocessing) processor id. Note that
1632 * all programs run with preemption disabled, which means that the
1633 * SMP processor id is stable during all the execution of the
1634 * program.
1635 * Return
1636 * The SMP id of the processor running the program.
1637 *
1638 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1639 * Description
1640 * Store *len* bytes from address *from* into the packet
1641 * associated to *skb*, at *offset*. *flags* are a combination of
1642 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1643 * checksum for the packet after storing the bytes) and
1644 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1645 * **->swhash** and *skb*\ **->l4hash** to 0).
1646 *
1647 * A call to this helper is susceptible to change the underlying
1648 * packet buffer. Therefore, at load time, all checks on pointers
1649 * previously done by the verifier are invalidated and must be
1650 * performed again, if the helper is used in combination with
1651 * direct packet access.
1652 * Return
1653 * 0 on success, or a negative error in case of failure.
1654 *
1655 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1656 * Description
1657 * Recompute the layer 3 (e.g. IP) checksum for the packet
1658 * associated to *skb*. Computation is incremental, so the helper
1659 * must know the former value of the header field that was
1660 * modified (*from*), the new value of this field (*to*), and the
1661 * number of bytes (2 or 4) for this field, stored in *size*.
1662 * Alternatively, it is possible to store the difference between
1663 * the previous and the new values of the header field in *to*, by
1664 * setting *from* and *size* to 0. For both methods, *offset*
1665 * indicates the location of the IP checksum within the packet.
1666 *
1667 * This helper works in combination with **bpf_csum_diff**\ (),
1668 * which does not update the checksum in-place, but offers more
1669 * flexibility and can handle sizes larger than 2 or 4 for the
1670 * checksum to update.
1671 *
1672 * A call to this helper is susceptible to change the underlying
1673 * packet buffer. Therefore, at load time, all checks on pointers
1674 * previously done by the verifier are invalidated and must be
1675 * performed again, if the helper is used in combination with
1676 * direct packet access.
1677 * Return
1678 * 0 on success, or a negative error in case of failure.
1679 *
1680 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1681 * Description
1682 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1683 * packet associated to *skb*. Computation is incremental, so the
1684 * helper must know the former value of the header field that was
1685 * modified (*from*), the new value of this field (*to*), and the
1686 * number of bytes (2 or 4) for this field, stored on the lowest
1687 * four bits of *flags*. Alternatively, it is possible to store
1688 * the difference between the previous and the new values of the
1689 * header field in *to*, by setting *from* and the four lowest
1690 * bits of *flags* to 0. For both methods, *offset* indicates the
1691 * location of the IP checksum within the packet. In addition to
1692 * the size of the field, *flags* can be added (bitwise OR) actual
1693 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1694 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1695 * for updates resulting in a null checksum the value is set to
1696 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1697 * the checksum is to be computed against a pseudo-header.
1698 *
1699 * This helper works in combination with **bpf_csum_diff**\ (),
1700 * which does not update the checksum in-place, but offers more
1701 * flexibility and can handle sizes larger than 2 or 4 for the
1702 * checksum to update.
1703 *
1704 * A call to this helper is susceptible to change the underlying
1705 * packet buffer. Therefore, at load time, all checks on pointers
1706 * previously done by the verifier are invalidated and must be
1707 * performed again, if the helper is used in combination with
1708 * direct packet access.
1709 * Return
1710 * 0 on success, or a negative error in case of failure.
1711 *
1712 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1713 * Description
1714 * This special helper is used to trigger a "tail call", or in
1715 * other words, to jump into another eBPF program. The same stack
1716 * frame is used (but values on stack and in registers for the
1717 * caller are not accessible to the callee). This mechanism allows
1718 * for program chaining, either for raising the maximum number of
1719 * available eBPF instructions, or to execute given programs in
1720 * conditional blocks. For security reasons, there is an upper
1721 * limit to the number of successive tail calls that can be
1722 * performed.
1723 *
1724 * Upon call of this helper, the program attempts to jump into a
1725 * program referenced at index *index* in *prog_array_map*, a
1726 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1727 * *ctx*, a pointer to the context.
1728 *
1729 * If the call succeeds, the kernel immediately runs the first
1730 * instruction of the new program. This is not a function call,
1731 * and it never returns to the previous program. If the call
1732 * fails, then the helper has no effect, and the caller continues
1733 * to run its subsequent instructions. A call can fail if the
1734 * destination program for the jump does not exist (i.e. *index*
1735 * is superior to the number of entries in *prog_array_map*), or
1736 * if the maximum number of tail calls has been reached for this
1737 * chain of programs. This limit is defined in the kernel by the
1738 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1739 * which is currently set to 32.
1740 * Return
1741 * 0 on success, or a negative error in case of failure.
1742 *
1743 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1744 * Description
1745 * Clone and redirect the packet associated to *skb* to another
1746 * net device of index *ifindex*. Both ingress and egress
1747 * interfaces can be used for redirection. The **BPF_F_INGRESS**
1748 * value in *flags* is used to make the distinction (ingress path
1749 * is selected if the flag is present, egress path otherwise).
1750 * This is the only flag supported for now.
1751 *
1752 * In comparison with **bpf_redirect**\ () helper,
1753 * **bpf_clone_redirect**\ () has the associated cost of
1754 * duplicating the packet buffer, but this can be executed out of
1755 * the eBPF program. Conversely, **bpf_redirect**\ () is more
1756 * efficient, but it is handled through an action code where the
1757 * redirection happens only after the eBPF program has returned.
1758 *
1759 * A call to this helper is susceptible to change the underlying
1760 * packet buffer. Therefore, at load time, all checks on pointers
1761 * previously done by the verifier are invalidated and must be
1762 * performed again, if the helper is used in combination with
1763 * direct packet access.
1764 * Return
1765 * 0 on success, or a negative error in case of failure. Positive
1766 * error indicates a potential drop or congestion in the target
1767 * device. The particular positive error codes are not defined.
1768 *
1769 * u64 bpf_get_current_pid_tgid(void)
1770 * Return
1771 * A 64-bit integer containing the current tgid and pid, and
1772 * created as such:
1773 * *current_task*\ **->tgid << 32 \|**
1774 * *current_task*\ **->pid**.
1775 *
1776 * u64 bpf_get_current_uid_gid(void)
1777 * Return
1778 * A 64-bit integer containing the current GID and UID, and
1779 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1780 *
1781 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1782 * Description
1783 * Copy the **comm** attribute of the current task into *buf* of
1784 * *size_of_buf*. The **comm** attribute contains the name of
1785 * the executable (excluding the path) for the current task. The
1786 * *size_of_buf* must be strictly positive. On success, the
1787 * helper makes sure that the *buf* is NUL-terminated. On failure,
1788 * it is filled with zeroes.
1789 * Return
1790 * 0 on success, or a negative error in case of failure.
1791 *
1792 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1793 * Description
1794 * Retrieve the classid for the current task, i.e. for the net_cls
1795 * cgroup to which *skb* belongs.
1796 *
1797 * This helper can be used on TC egress path, but not on ingress.
1798 *
1799 * The net_cls cgroup provides an interface to tag network packets
1800 * based on a user-provided identifier for all traffic coming from
1801 * the tasks belonging to the related cgroup. See also the related
1802 * kernel documentation, available from the Linux sources in file
1803 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1804 *
1805 * The Linux kernel has two versions for cgroups: there are
1806 * cgroups v1 and cgroups v2. Both are available to users, who can
1807 * use a mixture of them, but note that the net_cls cgroup is for
1808 * cgroup v1 only. This makes it incompatible with BPF programs
1809 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1810 * only hold data for one version of cgroups at a time).
1811 *
1812 * This helper is only available is the kernel was compiled with
1813 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1814 * "**y**" or to "**m**".
1815 * Return
1816 * The classid, or 0 for the default unconfigured classid.
1817 *
1818 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1819 * Description
1820 * Push a *vlan_tci* (VLAN tag control information) of protocol
1821 * *vlan_proto* to the packet associated to *skb*, then update
1822 * the checksum. Note that if *vlan_proto* is different from
1823 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1824 * be **ETH_P_8021Q**.
1825 *
1826 * A call to this helper is susceptible to change the underlying
1827 * packet buffer. Therefore, at load time, all checks on pointers
1828 * previously done by the verifier are invalidated and must be
1829 * performed again, if the helper is used in combination with
1830 * direct packet access.
1831 * Return
1832 * 0 on success, or a negative error in case of failure.
1833 *
1834 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1835 * Description
1836 * Pop a VLAN header from the packet associated to *skb*.
1837 *
1838 * A call to this helper is susceptible to change the underlying
1839 * packet buffer. Therefore, at load time, all checks on pointers
1840 * previously done by the verifier are invalidated and must be
1841 * performed again, if the helper is used in combination with
1842 * direct packet access.
1843 * Return
1844 * 0 on success, or a negative error in case of failure.
1845 *
1846 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1847 * Description
1848 * Get tunnel metadata. This helper takes a pointer *key* to an
1849 * empty **struct bpf_tunnel_key** of **size**, that will be
1850 * filled with tunnel metadata for the packet associated to *skb*.
1851 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1852 * indicates that the tunnel is based on IPv6 protocol instead of
1853 * IPv4.
1854 *
1855 * The **struct bpf_tunnel_key** is an object that generalizes the
1856 * principal parameters used by various tunneling protocols into a
1857 * single struct. This way, it can be used to easily make a
1858 * decision based on the contents of the encapsulation header,
1859 * "summarized" in this struct. In particular, it holds the IP
1860 * address of the remote end (IPv4 or IPv6, depending on the case)
1861 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1862 * this struct exposes the *key*\ **->tunnel_id**, which is
1863 * generally mapped to a VNI (Virtual Network Identifier), making
1864 * it programmable together with the **bpf_skb_set_tunnel_key**\
1865 * () helper.
1866 *
1867 * Let's imagine that the following code is part of a program
1868 * attached to the TC ingress interface, on one end of a GRE
1869 * tunnel, and is supposed to filter out all messages coming from
1870 * remote ends with IPv4 address other than 10.0.0.1:
1871 *
1872 * ::
1873 *
1874 * int ret;
1875 * struct bpf_tunnel_key key = {};
1876 *
1877 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1878 * if (ret < 0)
1879 * return TC_ACT_SHOT; // drop packet
1880 *
1881 * if (key.remote_ipv4 != 0x0a000001)
1882 * return TC_ACT_SHOT; // drop packet
1883 *
1884 * return TC_ACT_OK; // accept packet
1885 *
1886 * This interface can also be used with all encapsulation devices
1887 * that can operate in "collect metadata" mode: instead of having
1888 * one network device per specific configuration, the "collect
1889 * metadata" mode only requires a single device where the
1890 * configuration can be extracted from this helper.
1891 *
1892 * This can be used together with various tunnels such as VXLan,
1893 * Geneve, GRE or IP in IP (IPIP).
1894 * Return
1895 * 0 on success, or a negative error in case of failure.
1896 *
1897 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1898 * Description
1899 * Populate tunnel metadata for packet associated to *skb.* The
1900 * tunnel metadata is set to the contents of *key*, of *size*. The
1901 * *flags* can be set to a combination of the following values:
1902 *
1903 * **BPF_F_TUNINFO_IPV6**
1904 * Indicate that the tunnel is based on IPv6 protocol
1905 * instead of IPv4.
1906 * **BPF_F_ZERO_CSUM_TX**
1907 * For IPv4 packets, add a flag to tunnel metadata
1908 * indicating that checksum computation should be skipped
1909 * and checksum set to zeroes.
1910 * **BPF_F_DONT_FRAGMENT**
1911 * Add a flag to tunnel metadata indicating that the
1912 * packet should not be fragmented.
1913 * **BPF_F_SEQ_NUMBER**
1914 * Add a flag to tunnel metadata indicating that a
1915 * sequence number should be added to tunnel header before
1916 * sending the packet. This flag was added for GRE
1917 * encapsulation, but might be used with other protocols
1918 * as well in the future.
1919 *
1920 * Here is a typical usage on the transmit path:
1921 *
1922 * ::
1923 *
1924 * struct bpf_tunnel_key key;
1925 * populate key ...
1926 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1927 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1928 *
1929 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1930 * helper for additional information.
1931 * Return
1932 * 0 on success, or a negative error in case of failure.
1933 *
1934 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1935 * Description
1936 * Read the value of a perf event counter. This helper relies on a
1937 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1938 * the perf event counter is selected when *map* is updated with
1939 * perf event file descriptors. The *map* is an array whose size
1940 * is the number of available CPUs, and each cell contains a value
1941 * relative to one CPU. The value to retrieve is indicated by
1942 * *flags*, that contains the index of the CPU to look up, masked
1943 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1944 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1945 * current CPU should be retrieved.
1946 *
1947 * Note that before Linux 4.13, only hardware perf event can be
1948 * retrieved.
1949 *
1950 * Also, be aware that the newer helper
1951 * **bpf_perf_event_read_value**\ () is recommended over
1952 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1953 * quirks where error and counter value are used as a return code
1954 * (which is wrong to do since ranges may overlap). This issue is
1955 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1956 * time provides more features over the **bpf_perf_event_read**\
1957 * () interface. Please refer to the description of
1958 * **bpf_perf_event_read_value**\ () for details.
1959 * Return
1960 * The value of the perf event counter read from the map, or a
1961 * negative error code in case of failure.
1962 *
1963 * long bpf_redirect(u32 ifindex, u64 flags)
1964 * Description
1965 * Redirect the packet to another net device of index *ifindex*.
1966 * This helper is somewhat similar to **bpf_clone_redirect**\
1967 * (), except that the packet is not cloned, which provides
1968 * increased performance.
1969 *
1970 * Except for XDP, both ingress and egress interfaces can be used
1971 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1972 * to make the distinction (ingress path is selected if the flag
1973 * is present, egress path otherwise). Currently, XDP only
1974 * supports redirection to the egress interface, and accepts no
1975 * flag at all.
1976 *
1977 * The same effect can also be attained with the more generic
1978 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1979 * redirect target instead of providing it directly to the helper.
1980 * Return
1981 * For XDP, the helper returns **XDP_REDIRECT** on success or
1982 * **XDP_ABORTED** on error. For other program types, the values
1983 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1984 * error.
1985 *
1986 * u32 bpf_get_route_realm(struct sk_buff *skb)
1987 * Description
1988 * Retrieve the realm or the route, that is to say the
1989 * **tclassid** field of the destination for the *skb*. The
1990 * identifier retrieved is a user-provided tag, similar to the
1991 * one used with the net_cls cgroup (see description for
1992 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1993 * held by a route (a destination entry), not by a task.
1994 *
1995 * Retrieving this identifier works with the clsact TC egress hook
1996 * (see also **tc-bpf(8)**), or alternatively on conventional
1997 * classful egress qdiscs, but not on TC ingress path. In case of
1998 * clsact TC egress hook, this has the advantage that, internally,
1999 * the destination entry has not been dropped yet in the transmit
2000 * path. Therefore, the destination entry does not need to be
2001 * artificially held via **netif_keep_dst**\ () for a classful
2002 * qdisc until the *skb* is freed.
2003 *
2004 * This helper is available only if the kernel was compiled with
2005 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2006 * Return
2007 * The realm of the route for the packet associated to *skb*, or 0
2008 * if none was found.
2009 *
2010 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2011 * Description
2012 * Write raw *data* blob into a special BPF perf event held by
2013 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2014 * event must have the following attributes: **PERF_SAMPLE_RAW**
2015 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2016 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2017 *
2018 * The *flags* are used to indicate the index in *map* for which
2019 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2020 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2021 * to indicate that the index of the current CPU core should be
2022 * used.
2023 *
2024 * The value to write, of *size*, is passed through eBPF stack and
2025 * pointed by *data*.
2026 *
2027 * The context of the program *ctx* needs also be passed to the
2028 * helper.
2029 *
2030 * On user space, a program willing to read the values needs to
2031 * call **perf_event_open**\ () on the perf event (either for
2032 * one or for all CPUs) and to store the file descriptor into the
2033 * *map*. This must be done before the eBPF program can send data
2034 * into it. An example is available in file
2035 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2036 * tree (the eBPF program counterpart is in
2037 * *samples/bpf/trace_output_kern.c*).
2038 *
2039 * **bpf_perf_event_output**\ () achieves better performance
2040 * than **bpf_trace_printk**\ () for sharing data with user
2041 * space, and is much better suitable for streaming data from eBPF
2042 * programs.
2043 *
2044 * Note that this helper is not restricted to tracing use cases
2045 * and can be used with programs attached to TC or XDP as well,
2046 * where it allows for passing data to user space listeners. Data
2047 * can be:
2048 *
2049 * * Only custom structs,
2050 * * Only the packet payload, or
2051 * * A combination of both.
2052 * Return
2053 * 0 on success, or a negative error in case of failure.
2054 *
2055 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2056 * Description
2057 * This helper was provided as an easy way to load data from a
2058 * packet. It can be used to load *len* bytes from *offset* from
2059 * the packet associated to *skb*, into the buffer pointed by
2060 * *to*.
2061 *
2062 * Since Linux 4.7, usage of this helper has mostly been replaced
2063 * by "direct packet access", enabling packet data to be
2064 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2065 * pointing respectively to the first byte of packet data and to
2066 * the byte after the last byte of packet data. However, it
2067 * remains useful if one wishes to read large quantities of data
2068 * at once from a packet into the eBPF stack.
2069 * Return
2070 * 0 on success, or a negative error in case of failure.
2071 *
2072 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2073 * Description
2074 * Walk a user or a kernel stack and return its id. To achieve
2075 * this, the helper needs *ctx*, which is a pointer to the context
2076 * on which the tracing program is executed, and a pointer to a
2077 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2078 *
2079 * The last argument, *flags*, holds the number of stack frames to
2080 * skip (from 0 to 255), masked with
2081 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2082 * a combination of the following flags:
2083 *
2084 * **BPF_F_USER_STACK**
2085 * Collect a user space stack instead of a kernel stack.
2086 * **BPF_F_FAST_STACK_CMP**
2087 * Compare stacks by hash only.
2088 * **BPF_F_REUSE_STACKID**
2089 * If two different stacks hash into the same *stackid*,
2090 * discard the old one.
2091 *
2092 * The stack id retrieved is a 32 bit long integer handle which
2093 * can be further combined with other data (including other stack
2094 * ids) and used as a key into maps. This can be useful for
2095 * generating a variety of graphs (such as flame graphs or off-cpu
2096 * graphs).
2097 *
2098 * For walking a stack, this helper is an improvement over
2099 * **bpf_probe_read**\ (), which can be used with unrolled loops
2100 * but is not efficient and consumes a lot of eBPF instructions.
2101 * Instead, **bpf_get_stackid**\ () can collect up to
2102 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2103 * this limit can be controlled with the **sysctl** program, and
2104 * that it should be manually increased in order to profile long
2105 * user stacks (such as stacks for Java programs). To do so, use:
2106 *
2107 * ::
2108 *
2109 * # sysctl kernel.perf_event_max_stack=<new value>
2110 * Return
2111 * The positive or null stack id on success, or a negative error
2112 * in case of failure.
2113 *
2114 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2115 * Description
2116 * Compute a checksum difference, from the raw buffer pointed by
2117 * *from*, of length *from_size* (that must be a multiple of 4),
2118 * towards the raw buffer pointed by *to*, of size *to_size*
2119 * (same remark). An optional *seed* can be added to the value
2120 * (this can be cascaded, the seed may come from a previous call
2121 * to the helper).
2122 *
2123 * This is flexible enough to be used in several ways:
2124 *
2125 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2126 * checksum, it can be used when pushing new data.
2127 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2128 * checksum, it can be used when removing data from a packet.
2129 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2130 * can be used to compute a diff. Note that *from_size* and
2131 * *to_size* do not need to be equal.
2132 *
2133 * This helper can be used in combination with
2134 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2135 * which one can feed in the difference computed with
2136 * **bpf_csum_diff**\ ().
2137 * Return
2138 * The checksum result, or a negative error code in case of
2139 * failure.
2140 *
2141 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2142 * Description
2143 * Retrieve tunnel options metadata for the packet associated to
2144 * *skb*, and store the raw tunnel option data to the buffer *opt*
2145 * of *size*.
2146 *
2147 * This helper can be used with encapsulation devices that can
2148 * operate in "collect metadata" mode (please refer to the related
2149 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2150 * more details). A particular example where this can be used is
2151 * in combination with the Geneve encapsulation protocol, where it
2152 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2153 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2154 * the eBPF program. This allows for full customization of these
2155 * headers.
2156 * Return
2157 * The size of the option data retrieved.
2158 *
2159 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2160 * Description
2161 * Set tunnel options metadata for the packet associated to *skb*
2162 * to the option data contained in the raw buffer *opt* of *size*.
2163 *
2164 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2165 * helper for additional information.
2166 * Return
2167 * 0 on success, or a negative error in case of failure.
2168 *
2169 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2170 * Description
2171 * Change the protocol of the *skb* to *proto*. Currently
2172 * supported are transition from IPv4 to IPv6, and from IPv6 to
2173 * IPv4. The helper takes care of the groundwork for the
2174 * transition, including resizing the socket buffer. The eBPF
2175 * program is expected to fill the new headers, if any, via
2176 * **skb_store_bytes**\ () and to recompute the checksums with
2177 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2178 * (). The main case for this helper is to perform NAT64
2179 * operations out of an eBPF program.
2180 *
2181 * Internally, the GSO type is marked as dodgy so that headers are
2182 * checked and segments are recalculated by the GSO/GRO engine.
2183 * The size for GSO target is adapted as well.
2184 *
2185 * All values for *flags* are reserved for future usage, and must
2186 * be left at zero.
2187 *
2188 * A call to this helper is susceptible to change the underlying
2189 * packet buffer. Therefore, at load time, all checks on pointers
2190 * previously done by the verifier are invalidated and must be
2191 * performed again, if the helper is used in combination with
2192 * direct packet access.
2193 * Return
2194 * 0 on success, or a negative error in case of failure.
2195 *
2196 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2197 * Description
2198 * Change the packet type for the packet associated to *skb*. This
2199 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2200 * the eBPF program does not have a write access to *skb*\
2201 * **->pkt_type** beside this helper. Using a helper here allows
2202 * for graceful handling of errors.
2203 *
2204 * The major use case is to change incoming *skb*s to
2205 * **PACKET_HOST** in a programmatic way instead of having to
2206 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2207 * example.
2208 *
2209 * Note that *type* only allows certain values. At this time, they
2210 * are:
2211 *
2212 * **PACKET_HOST**
2213 * Packet is for us.
2214 * **PACKET_BROADCAST**
2215 * Send packet to all.
2216 * **PACKET_MULTICAST**
2217 * Send packet to group.
2218 * **PACKET_OTHERHOST**
2219 * Send packet to someone else.
2220 * Return
2221 * 0 on success, or a negative error in case of failure.
2222 *
2223 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2224 * Description
2225 * Check whether *skb* is a descendant of the cgroup2 held by
2226 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2227 * Return
2228 * The return value depends on the result of the test, and can be:
2229 *
2230 * * 0, if the *skb* failed the cgroup2 descendant test.
2231 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2232 * * A negative error code, if an error occurred.
2233 *
2234 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2235 * Description
2236 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2237 * not set, in particular if the hash was cleared due to mangling,
2238 * recompute this hash. Later accesses to the hash can be done
2239 * directly with *skb*\ **->hash**.
2240 *
2241 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2242 * prototype with **bpf_skb_change_proto**\ (), or calling
2243 * **bpf_skb_store_bytes**\ () with the
2244 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2245 * the hash and to trigger a new computation for the next call to
2246 * **bpf_get_hash_recalc**\ ().
2247 * Return
2248 * The 32-bit hash.
2249 *
2250 * u64 bpf_get_current_task(void)
2251 * Return
2252 * A pointer to the current task struct.
2253 *
2254 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2255 * Description
2256 * Attempt in a safe way to write *len* bytes from the buffer
2257 * *src* to *dst* in memory. It only works for threads that are in
2258 * user context, and *dst* must be a valid user space address.
2259 *
2260 * This helper should not be used to implement any kind of
2261 * security mechanism because of TOC-TOU attacks, but rather to
2262 * debug, divert, and manipulate execution of semi-cooperative
2263 * processes.
2264 *
2265 * Keep in mind that this feature is meant for experiments, and it
2266 * has a risk of crashing the system and running programs.
2267 * Therefore, when an eBPF program using this helper is attached,
2268 * a warning including PID and process name is printed to kernel
2269 * logs.
2270 * Return
2271 * 0 on success, or a negative error in case of failure.
2272 *
2273 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2274 * Description
2275 * Check whether the probe is being run is the context of a given
2276 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2277 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2278 * Return
2279 * The return value depends on the result of the test, and can be:
2280 *
2281 * * 1, if current task belongs to the cgroup2.
2282 * * 0, if current task does not belong to the cgroup2.
2283 * * A negative error code, if an error occurred.
2284 *
2285 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2286 * Description
2287 * Resize (trim or grow) the packet associated to *skb* to the
2288 * new *len*. The *flags* are reserved for future usage, and must
2289 * be left at zero.
2290 *
2291 * The basic idea is that the helper performs the needed work to
2292 * change the size of the packet, then the eBPF program rewrites
2293 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2294 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2295 * and others. This helper is a slow path utility intended for
2296 * replies with control messages. And because it is targeted for
2297 * slow path, the helper itself can afford to be slow: it
2298 * implicitly linearizes, unclones and drops offloads from the
2299 * *skb*.
2300 *
2301 * A call to this helper is susceptible to change the underlying
2302 * packet buffer. Therefore, at load time, all checks on pointers
2303 * previously done by the verifier are invalidated and must be
2304 * performed again, if the helper is used in combination with
2305 * direct packet access.
2306 * Return
2307 * 0 on success, or a negative error in case of failure.
2308 *
2309 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2310 * Description
2311 * Pull in non-linear data in case the *skb* is non-linear and not
2312 * all of *len* are part of the linear section. Make *len* bytes
2313 * from *skb* readable and writable. If a zero value is passed for
2314 * *len*, then the whole length of the *skb* is pulled.
2315 *
2316 * This helper is only needed for reading and writing with direct
2317 * packet access.
2318 *
2319 * For direct packet access, testing that offsets to access
2320 * are within packet boundaries (test on *skb*\ **->data_end**) is
2321 * susceptible to fail if offsets are invalid, or if the requested
2322 * data is in non-linear parts of the *skb*. On failure the
2323 * program can just bail out, or in the case of a non-linear
2324 * buffer, use a helper to make the data available. The
2325 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2326 * the data. Another one consists in using **bpf_skb_pull_data**
2327 * to pull in once the non-linear parts, then retesting and
2328 * eventually access the data.
2329 *
2330 * At the same time, this also makes sure the *skb* is uncloned,
2331 * which is a necessary condition for direct write. As this needs
2332 * to be an invariant for the write part only, the verifier
2333 * detects writes and adds a prologue that is calling
2334 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2335 * the very beginning in case it is indeed cloned.
2336 *
2337 * A call to this helper is susceptible to change the underlying
2338 * packet buffer. Therefore, at load time, all checks on pointers
2339 * previously done by the verifier are invalidated and must be
2340 * performed again, if the helper is used in combination with
2341 * direct packet access.
2342 * Return
2343 * 0 on success, or a negative error in case of failure.
2344 *
2345 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2346 * Description
2347 * Add the checksum *csum* into *skb*\ **->csum** in case the
2348 * driver has supplied a checksum for the entire packet into that
2349 * field. Return an error otherwise. This helper is intended to be
2350 * used in combination with **bpf_csum_diff**\ (), in particular
2351 * when the checksum needs to be updated after data has been
2352 * written into the packet through direct packet access.
2353 * Return
2354 * The checksum on success, or a negative error code in case of
2355 * failure.
2356 *
2357 * void bpf_set_hash_invalid(struct sk_buff *skb)
2358 * Description
2359 * Invalidate the current *skb*\ **->hash**. It can be used after
2360 * mangling on headers through direct packet access, in order to
2361 * indicate that the hash is outdated and to trigger a
2362 * recalculation the next time the kernel tries to access this
2363 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2364 *
2365 * long bpf_get_numa_node_id(void)
2366 * Description
2367 * Return the id of the current NUMA node. The primary use case
2368 * for this helper is the selection of sockets for the local NUMA
2369 * node, when the program is attached to sockets using the
2370 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2371 * but the helper is also available to other eBPF program types,
2372 * similarly to **bpf_get_smp_processor_id**\ ().
2373 * Return
2374 * The id of current NUMA node.
2375 *
2376 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2377 * Description
2378 * Grows headroom of packet associated to *skb* and adjusts the
2379 * offset of the MAC header accordingly, adding *len* bytes of
2380 * space. It automatically extends and reallocates memory as
2381 * required.
2382 *
2383 * This helper can be used on a layer 3 *skb* to push a MAC header
2384 * for redirection into a layer 2 device.
2385 *
2386 * All values for *flags* are reserved for future usage, and must
2387 * be left at zero.
2388 *
2389 * A call to this helper is susceptible to change the underlying
2390 * packet buffer. Therefore, at load time, all checks on pointers
2391 * previously done by the verifier are invalidated and must be
2392 * performed again, if the helper is used in combination with
2393 * direct packet access.
2394 * Return
2395 * 0 on success, or a negative error in case of failure.
2396 *
2397 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2398 * Description
2399 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2400 * it is possible to use a negative value for *delta*. This helper
2401 * can be used to prepare the packet for pushing or popping
2402 * headers.
2403 *
2404 * A call to this helper is susceptible to change the underlying
2405 * packet buffer. Therefore, at load time, all checks on pointers
2406 * previously done by the verifier are invalidated and must be
2407 * performed again, if the helper is used in combination with
2408 * direct packet access.
2409 * Return
2410 * 0 on success, or a negative error in case of failure.
2411 *
2412 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2413 * Description
2414 * Copy a NUL terminated string from an unsafe kernel address
2415 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2416 * more details.
2417 *
2418 * Generally, use **bpf_probe_read_user_str**\ () or
2419 * **bpf_probe_read_kernel_str**\ () instead.
2420 * Return
2421 * On success, the strictly positive length of the string,
2422 * including the trailing NUL character. On error, a negative
2423 * value.
2424 *
2425 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2426 * Description
2427 * If the **struct sk_buff** pointed by *skb* has a known socket,
2428 * retrieve the cookie (generated by the kernel) of this socket.
2429 * If no cookie has been set yet, generate a new cookie. Once
2430 * generated, the socket cookie remains stable for the life of the
2431 * socket. This helper can be useful for monitoring per socket
2432 * networking traffic statistics as it provides a global socket
2433 * identifier that can be assumed unique.
2434 * Return
2435 * A 8-byte long unique number on success, or 0 if the socket
2436 * field is missing inside *skb*.
2437 *
2438 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2439 * Description
2440 * Equivalent to bpf_get_socket_cookie() helper that accepts
2441 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2442 * Return
2443 * A 8-byte long unique number.
2444 *
2445 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2446 * Description
2447 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2448 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2449 * Return
2450 * A 8-byte long unique number.
2451 *
2452 * u64 bpf_get_socket_cookie(struct sock *sk)
2453 * Description
2454 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2455 * *sk*, but gets socket from a BTF **struct sock**. This helper
2456 * also works for sleepable programs.
2457 * Return
2458 * A 8-byte long unique number or 0 if *sk* is NULL.
2459 *
2460 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2461 * Return
2462 * The owner UID of the socket associated to *skb*. If the socket
2463 * is **NULL**, or if it is not a full socket (i.e. if it is a
2464 * time-wait or a request socket instead), **overflowuid** value
2465 * is returned (note that **overflowuid** might also be the actual
2466 * UID value for the socket).
2467 *
2468 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2469 * Description
2470 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2471 * to value *hash*.
2472 * Return
2473 * 0
2474 *
2475 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2476 * Description
2477 * Emulate a call to **setsockopt()** on the socket associated to
2478 * *bpf_socket*, which must be a full socket. The *level* at
2479 * which the option resides and the name *optname* of the option
2480 * must be specified, see **setsockopt(2)** for more information.
2481 * The option value of length *optlen* is pointed by *optval*.
2482 *
2483 * *bpf_socket* should be one of the following:
2484 *
2485 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2486 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2487 * and **BPF_CGROUP_INET6_CONNECT**.
2488 *
2489 * This helper actually implements a subset of **setsockopt()**.
2490 * It supports the following *level*\ s:
2491 *
2492 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2493 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2494 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2495 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2496 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2497 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2498 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2499 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2500 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2501 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2502 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2503 * Return
2504 * 0 on success, or a negative error in case of failure.
2505 *
2506 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2507 * Description
2508 * Grow or shrink the room for data in the packet associated to
2509 * *skb* by *len_diff*, and according to the selected *mode*.
2510 *
2511 * By default, the helper will reset any offloaded checksum
2512 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2513 * by the following flag:
2514 *
2515 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2516 * checksum data of the skb to CHECKSUM_NONE.
2517 *
2518 * There are two supported modes at this time:
2519 *
2520 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2521 * (room space is added or removed below the layer 2 header).
2522 *
2523 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2524 * (room space is added or removed below the layer 3 header).
2525 *
2526 * The following flags are supported at this time:
2527 *
2528 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2529 * Adjusting mss in this way is not allowed for datagrams.
2530 *
2531 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2532 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2533 * Any new space is reserved to hold a tunnel header.
2534 * Configure skb offsets and other fields accordingly.
2535 *
2536 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2537 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2538 * Use with ENCAP_L3 flags to further specify the tunnel type.
2539 *
2540 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2541 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2542 * type; *len* is the length of the inner MAC header.
2543 *
2544 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2545 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2546 * L2 type as Ethernet.
2547 *
2548 * A call to this helper is susceptible to change the underlying
2549 * packet buffer. Therefore, at load time, all checks on pointers
2550 * previously done by the verifier are invalidated and must be
2551 * performed again, if the helper is used in combination with
2552 * direct packet access.
2553 * Return
2554 * 0 on success, or a negative error in case of failure.
2555 *
2556 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2557 * Description
2558 * Redirect the packet to the endpoint referenced by *map* at
2559 * index *key*. Depending on its type, this *map* can contain
2560 * references to net devices (for forwarding packets through other
2561 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2562 * but this is only implemented for native XDP (with driver
2563 * support) as of this writing).
2564 *
2565 * The lower two bits of *flags* are used as the return code if
2566 * the map lookup fails. This is so that the return value can be
2567 * one of the XDP program return codes up to **XDP_TX**, as chosen
2568 * by the caller. The higher bits of *flags* can be set to
2569 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2570 *
2571 * With BPF_F_BROADCAST the packet will be broadcasted to all the
2572 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2573 * interface will be excluded when do broadcasting.
2574 *
2575 * See also **bpf_redirect**\ (), which only supports redirecting
2576 * to an ifindex, but doesn't require a map to do so.
2577 * Return
2578 * **XDP_REDIRECT** on success, or the value of the two lower bits
2579 * of the *flags* argument on error.
2580 *
2581 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2582 * Description
2583 * Redirect the packet to the socket referenced by *map* (of type
2584 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2585 * egress interfaces can be used for redirection. The
2586 * **BPF_F_INGRESS** value in *flags* is used to make the
2587 * distinction (ingress path is selected if the flag is present,
2588 * egress path otherwise). This is the only flag supported for now.
2589 * Return
2590 * **SK_PASS** on success, or **SK_DROP** on error.
2591 *
2592 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2593 * Description
2594 * Add an entry to, or update a *map* referencing sockets. The
2595 * *skops* is used as a new value for the entry associated to
2596 * *key*. *flags* is one of:
2597 *
2598 * **BPF_NOEXIST**
2599 * The entry for *key* must not exist in the map.
2600 * **BPF_EXIST**
2601 * The entry for *key* must already exist in the map.
2602 * **BPF_ANY**
2603 * No condition on the existence of the entry for *key*.
2604 *
2605 * If the *map* has eBPF programs (parser and verdict), those will
2606 * be inherited by the socket being added. If the socket is
2607 * already attached to eBPF programs, this results in an error.
2608 * Return
2609 * 0 on success, or a negative error in case of failure.
2610 *
2611 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2612 * Description
2613 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2614 * *delta* (which can be positive or negative). Note that this
2615 * operation modifies the address stored in *xdp_md*\ **->data**,
2616 * so the latter must be loaded only after the helper has been
2617 * called.
2618 *
2619 * The use of *xdp_md*\ **->data_meta** is optional and programs
2620 * are not required to use it. The rationale is that when the
2621 * packet is processed with XDP (e.g. as DoS filter), it is
2622 * possible to push further meta data along with it before passing
2623 * to the stack, and to give the guarantee that an ingress eBPF
2624 * program attached as a TC classifier on the same device can pick
2625 * this up for further post-processing. Since TC works with socket
2626 * buffers, it remains possible to set from XDP the **mark** or
2627 * **priority** pointers, or other pointers for the socket buffer.
2628 * Having this scratch space generic and programmable allows for
2629 * more flexibility as the user is free to store whatever meta
2630 * data they need.
2631 *
2632 * A call to this helper is susceptible to change the underlying
2633 * packet buffer. Therefore, at load time, all checks on pointers
2634 * previously done by the verifier are invalidated and must be
2635 * performed again, if the helper is used in combination with
2636 * direct packet access.
2637 * Return
2638 * 0 on success, or a negative error in case of failure.
2639 *
2640 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2641 * Description
2642 * Read the value of a perf event counter, and store it into *buf*
2643 * of size *buf_size*. This helper relies on a *map* of type
2644 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2645 * counter is selected when *map* is updated with perf event file
2646 * descriptors. The *map* is an array whose size is the number of
2647 * available CPUs, and each cell contains a value relative to one
2648 * CPU. The value to retrieve is indicated by *flags*, that
2649 * contains the index of the CPU to look up, masked with
2650 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2651 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2652 * current CPU should be retrieved.
2653 *
2654 * This helper behaves in a way close to
2655 * **bpf_perf_event_read**\ () helper, save that instead of
2656 * just returning the value observed, it fills the *buf*
2657 * structure. This allows for additional data to be retrieved: in
2658 * particular, the enabled and running times (in *buf*\
2659 * **->enabled** and *buf*\ **->running**, respectively) are
2660 * copied. In general, **bpf_perf_event_read_value**\ () is
2661 * recommended over **bpf_perf_event_read**\ (), which has some
2662 * ABI issues and provides fewer functionalities.
2663 *
2664 * These values are interesting, because hardware PMU (Performance
2665 * Monitoring Unit) counters are limited resources. When there are
2666 * more PMU based perf events opened than available counters,
2667 * kernel will multiplex these events so each event gets certain
2668 * percentage (but not all) of the PMU time. In case that
2669 * multiplexing happens, the number of samples or counter value
2670 * will not reflect the case compared to when no multiplexing
2671 * occurs. This makes comparison between different runs difficult.
2672 * Typically, the counter value should be normalized before
2673 * comparing to other experiments. The usual normalization is done
2674 * as follows.
2675 *
2676 * ::
2677 *
2678 * normalized_counter = counter * t_enabled / t_running
2679 *
2680 * Where t_enabled is the time enabled for event and t_running is
2681 * the time running for event since last normalization. The
2682 * enabled and running times are accumulated since the perf event
2683 * open. To achieve scaling factor between two invocations of an
2684 * eBPF program, users can use CPU id as the key (which is
2685 * typical for perf array usage model) to remember the previous
2686 * value and do the calculation inside the eBPF program.
2687 * Return
2688 * 0 on success, or a negative error in case of failure.
2689 *
2690 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2691 * Description
2692 * For en eBPF program attached to a perf event, retrieve the
2693 * value of the event counter associated to *ctx* and store it in
2694 * the structure pointed by *buf* and of size *buf_size*. Enabled
2695 * and running times are also stored in the structure (see
2696 * description of helper **bpf_perf_event_read_value**\ () for
2697 * more details).
2698 * Return
2699 * 0 on success, or a negative error in case of failure.
2700 *
2701 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2702 * Description
2703 * Emulate a call to **getsockopt()** on the socket associated to
2704 * *bpf_socket*, which must be a full socket. The *level* at
2705 * which the option resides and the name *optname* of the option
2706 * must be specified, see **getsockopt(2)** for more information.
2707 * The retrieved value is stored in the structure pointed by
2708 * *opval* and of length *optlen*.
2709 *
2710 * *bpf_socket* should be one of the following:
2711 *
2712 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2713 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2714 * and **BPF_CGROUP_INET6_CONNECT**.
2715 *
2716 * This helper actually implements a subset of **getsockopt()**.
2717 * It supports the following *level*\ s:
2718 *
2719 * * **IPPROTO_TCP**, which supports *optname*
2720 * **TCP_CONGESTION**.
2721 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2722 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2723 * Return
2724 * 0 on success, or a negative error in case of failure.
2725 *
2726 * long bpf_override_return(struct pt_regs *regs, u64 rc)
2727 * Description
2728 * Used for error injection, this helper uses kprobes to override
2729 * the return value of the probed function, and to set it to *rc*.
2730 * The first argument is the context *regs* on which the kprobe
2731 * works.
2732 *
2733 * This helper works by setting the PC (program counter)
2734 * to an override function which is run in place of the original
2735 * probed function. This means the probed function is not run at
2736 * all. The replacement function just returns with the required
2737 * value.
2738 *
2739 * This helper has security implications, and thus is subject to
2740 * restrictions. It is only available if the kernel was compiled
2741 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2742 * option, and in this case it only works on functions tagged with
2743 * **ALLOW_ERROR_INJECTION** in the kernel code.
2744 *
2745 * Also, the helper is only available for the architectures having
2746 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2747 * x86 architecture is the only one to support this feature.
2748 * Return
2749 * 0
2750 *
2751 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2752 * Description
2753 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2754 * for the full TCP socket associated to *bpf_sock_ops* to
2755 * *argval*.
2756 *
2757 * The primary use of this field is to determine if there should
2758 * be calls to eBPF programs of type
2759 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2760 * code. A program of the same type can change its value, per
2761 * connection and as necessary, when the connection is
2762 * established. This field is directly accessible for reading, but
2763 * this helper must be used for updates in order to return an
2764 * error if an eBPF program tries to set a callback that is not
2765 * supported in the current kernel.
2766 *
2767 * *argval* is a flag array which can combine these flags:
2768 *
2769 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2770 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2771 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2772 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2773 *
2774 * Therefore, this function can be used to clear a callback flag by
2775 * setting the appropriate bit to zero. e.g. to disable the RTO
2776 * callback:
2777 *
2778 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
2779 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2780 *
2781 * Here are some examples of where one could call such eBPF
2782 * program:
2783 *
2784 * * When RTO fires.
2785 * * When a packet is retransmitted.
2786 * * When the connection terminates.
2787 * * When a packet is sent.
2788 * * When a packet is received.
2789 * Return
2790 * Code **-EINVAL** if the socket is not a full TCP socket;
2791 * otherwise, a positive number containing the bits that could not
2792 * be set is returned (which comes down to 0 if all bits were set
2793 * as required).
2794 *
2795 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2796 * Description
2797 * This helper is used in programs implementing policies at the
2798 * socket level. If the message *msg* is allowed to pass (i.e. if
2799 * the verdict eBPF program returns **SK_PASS**), redirect it to
2800 * the socket referenced by *map* (of type
2801 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2802 * egress interfaces can be used for redirection. The
2803 * **BPF_F_INGRESS** value in *flags* is used to make the
2804 * distinction (ingress path is selected if the flag is present,
2805 * egress path otherwise). This is the only flag supported for now.
2806 * Return
2807 * **SK_PASS** on success, or **SK_DROP** on error.
2808 *
2809 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2810 * Description
2811 * For socket policies, apply the verdict of the eBPF program to
2812 * the next *bytes* (number of bytes) of message *msg*.
2813 *
2814 * For example, this helper can be used in the following cases:
2815 *
2816 * * A single **sendmsg**\ () or **sendfile**\ () system call
2817 * contains multiple logical messages that the eBPF program is
2818 * supposed to read and for which it should apply a verdict.
2819 * * An eBPF program only cares to read the first *bytes* of a
2820 * *msg*. If the message has a large payload, then setting up
2821 * and calling the eBPF program repeatedly for all bytes, even
2822 * though the verdict is already known, would create unnecessary
2823 * overhead.
2824 *
2825 * When called from within an eBPF program, the helper sets a
2826 * counter internal to the BPF infrastructure, that is used to
2827 * apply the last verdict to the next *bytes*. If *bytes* is
2828 * smaller than the current data being processed from a
2829 * **sendmsg**\ () or **sendfile**\ () system call, the first
2830 * *bytes* will be sent and the eBPF program will be re-run with
2831 * the pointer for start of data pointing to byte number *bytes*
2832 * **+ 1**. If *bytes* is larger than the current data being
2833 * processed, then the eBPF verdict will be applied to multiple
2834 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2835 * consumed.
2836 *
2837 * Note that if a socket closes with the internal counter holding
2838 * a non-zero value, this is not a problem because data is not
2839 * being buffered for *bytes* and is sent as it is received.
2840 * Return
2841 * 0
2842 *
2843 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2844 * Description
2845 * For socket policies, prevent the execution of the verdict eBPF
2846 * program for message *msg* until *bytes* (byte number) have been
2847 * accumulated.
2848 *
2849 * This can be used when one needs a specific number of bytes
2850 * before a verdict can be assigned, even if the data spans
2851 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2852 * case would be a user calling **sendmsg**\ () repeatedly with
2853 * 1-byte long message segments. Obviously, this is bad for
2854 * performance, but it is still valid. If the eBPF program needs
2855 * *bytes* bytes to validate a header, this helper can be used to
2856 * prevent the eBPF program to be called again until *bytes* have
2857 * been accumulated.
2858 * Return
2859 * 0
2860 *
2861 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2862 * Description
2863 * For socket policies, pull in non-linear data from user space
2864 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2865 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2866 * respectively.
2867 *
2868 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2869 * *msg* it can only parse data that the (**data**, **data_end**)
2870 * pointers have already consumed. For **sendmsg**\ () hooks this
2871 * is likely the first scatterlist element. But for calls relying
2872 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2873 * be the range (**0**, **0**) because the data is shared with
2874 * user space and by default the objective is to avoid allowing
2875 * user space to modify data while (or after) eBPF verdict is
2876 * being decided. This helper can be used to pull in data and to
2877 * set the start and end pointer to given values. Data will be
2878 * copied if necessary (i.e. if data was not linear and if start
2879 * and end pointers do not point to the same chunk).
2880 *
2881 * A call to this helper is susceptible to change the underlying
2882 * packet buffer. Therefore, at load time, all checks on pointers
2883 * previously done by the verifier are invalidated and must be
2884 * performed again, if the helper is used in combination with
2885 * direct packet access.
2886 *
2887 * All values for *flags* are reserved for future usage, and must
2888 * be left at zero.
2889 * Return
2890 * 0 on success, or a negative error in case of failure.
2891 *
2892 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2893 * Description
2894 * Bind the socket associated to *ctx* to the address pointed by
2895 * *addr*, of length *addr_len*. This allows for making outgoing
2896 * connection from the desired IP address, which can be useful for
2897 * example when all processes inside a cgroup should use one
2898 * single IP address on a host that has multiple IP configured.
2899 *
2900 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2901 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2902 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2903 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2904 * behavior and lets the kernel efficiently pick up an unused
2905 * port as long as 4-tuple is unique. Passing non-zero port might
2906 * lead to degraded performance.
2907 * Return
2908 * 0 on success, or a negative error in case of failure.
2909 *
2910 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2911 * Description
2912 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2913 * possible to both shrink and grow the packet tail.
2914 * Shrink done via *delta* being a negative integer.
2915 *
2916 * A call to this helper is susceptible to change the underlying
2917 * packet buffer. Therefore, at load time, all checks on pointers
2918 * previously done by the verifier are invalidated and must be
2919 * performed again, if the helper is used in combination with
2920 * direct packet access.
2921 * Return
2922 * 0 on success, or a negative error in case of failure.
2923 *
2924 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2925 * Description
2926 * Retrieve the XFRM state (IP transform framework, see also
2927 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2928 *
2929 * The retrieved value is stored in the **struct bpf_xfrm_state**
2930 * pointed by *xfrm_state* and of length *size*.
2931 *
2932 * All values for *flags* are reserved for future usage, and must
2933 * be left at zero.
2934 *
2935 * This helper is available only if the kernel was compiled with
2936 * **CONFIG_XFRM** configuration option.
2937 * Return
2938 * 0 on success, or a negative error in case of failure.
2939 *
2940 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2941 * Description
2942 * Return a user or a kernel stack in bpf program provided buffer.
2943 * To achieve this, the helper needs *ctx*, which is a pointer
2944 * to the context on which the tracing program is executed.
2945 * To store the stacktrace, the bpf program provides *buf* with
2946 * a nonnegative *size*.
2947 *
2948 * The last argument, *flags*, holds the number of stack frames to
2949 * skip (from 0 to 255), masked with
2950 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2951 * the following flags:
2952 *
2953 * **BPF_F_USER_STACK**
2954 * Collect a user space stack instead of a kernel stack.
2955 * **BPF_F_USER_BUILD_ID**
2956 * Collect buildid+offset instead of ips for user stack,
2957 * only valid if **BPF_F_USER_STACK** is also specified.
2958 *
2959 * **bpf_get_stack**\ () can collect up to
2960 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2961 * to sufficient large buffer size. Note that
2962 * this limit can be controlled with the **sysctl** program, and
2963 * that it should be manually increased in order to profile long
2964 * user stacks (such as stacks for Java programs). To do so, use:
2965 *
2966 * ::
2967 *
2968 * # sysctl kernel.perf_event_max_stack=<new value>
2969 * Return
2970 * The non-negative copied *buf* length equal to or less than
2971 * *size* on success, or a negative error in case of failure.
2972 *
2973 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2974 * Description
2975 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2976 * it provides an easy way to load *len* bytes from *offset*
2977 * from the packet associated to *skb*, into the buffer pointed
2978 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2979 * a fifth argument *start_header* exists in order to select a
2980 * base offset to start from. *start_header* can be one of:
2981 *
2982 * **BPF_HDR_START_MAC**
2983 * Base offset to load data from is *skb*'s mac header.
2984 * **BPF_HDR_START_NET**
2985 * Base offset to load data from is *skb*'s network header.
2986 *
2987 * In general, "direct packet access" is the preferred method to
2988 * access packet data, however, this helper is in particular useful
2989 * in socket filters where *skb*\ **->data** does not always point
2990 * to the start of the mac header and where "direct packet access"
2991 * is not available.
2992 * Return
2993 * 0 on success, or a negative error in case of failure.
2994 *
2995 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2996 * Description
2997 * Do FIB lookup in kernel tables using parameters in *params*.
2998 * If lookup is successful and result shows packet is to be
2999 * forwarded, the neighbor tables are searched for the nexthop.
3000 * If successful (ie., FIB lookup shows forwarding and nexthop
3001 * is resolved), the nexthop address is returned in ipv4_dst
3002 * or ipv6_dst based on family, smac is set to mac address of
3003 * egress device, dmac is set to nexthop mac address, rt_metric
3004 * is set to metric from route (IPv4/IPv6 only), and ifindex
3005 * is set to the device index of the nexthop from the FIB lookup.
3006 *
3007 * *plen* argument is the size of the passed in struct.
3008 * *flags* argument can be a combination of one or more of the
3009 * following values:
3010 *
3011 * **BPF_FIB_LOOKUP_DIRECT**
3012 * Do a direct table lookup vs full lookup using FIB
3013 * rules.
3014 * **BPF_FIB_LOOKUP_TBID**
3015 * Used with BPF_FIB_LOOKUP_DIRECT.
3016 * Use the routing table ID present in *params*->tbid
3017 * for the fib lookup.
3018 * **BPF_FIB_LOOKUP_OUTPUT**
3019 * Perform lookup from an egress perspective (default is
3020 * ingress).
3021 * **BPF_FIB_LOOKUP_SKIP_NEIGH**
3022 * Skip the neighbour table lookup. *params*->dmac
3023 * and *params*->smac will not be set as output. A common
3024 * use case is to call **bpf_redirect_neigh**\ () after
3025 * doing **bpf_fib_lookup**\ ().
3026 * **BPF_FIB_LOOKUP_SRC**
3027 * Derive and set source IP addr in *params*->ipv{4,6}_src
3028 * for the nexthop. If the src addr cannot be derived,
3029 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3030 * case, *params*->dmac and *params*->smac are not set either.
3031 *
3032 * *ctx* is either **struct xdp_md** for XDP programs or
3033 * **struct sk_buff** tc cls_act programs.
3034 * Return
3035 * * < 0 if any input argument is invalid
3036 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3037 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3038 * packet is not forwarded or needs assist from full stack
3039 *
3040 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3041 * was exceeded and output params->mtu_result contains the MTU.
3042 *
3043 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3044 * Description
3045 * Add an entry to, or update a sockhash *map* referencing sockets.
3046 * The *skops* is used as a new value for the entry associated to
3047 * *key*. *flags* is one of:
3048 *
3049 * **BPF_NOEXIST**
3050 * The entry for *key* must not exist in the map.
3051 * **BPF_EXIST**
3052 * The entry for *key* must already exist in the map.
3053 * **BPF_ANY**
3054 * No condition on the existence of the entry for *key*.
3055 *
3056 * If the *map* has eBPF programs (parser and verdict), those will
3057 * be inherited by the socket being added. If the socket is
3058 * already attached to eBPF programs, this results in an error.
3059 * Return
3060 * 0 on success, or a negative error in case of failure.
3061 *
3062 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3063 * Description
3064 * This helper is used in programs implementing policies at the
3065 * socket level. If the message *msg* is allowed to pass (i.e. if
3066 * the verdict eBPF program returns **SK_PASS**), redirect it to
3067 * the socket referenced by *map* (of type
3068 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3069 * egress interfaces can be used for redirection. The
3070 * **BPF_F_INGRESS** value in *flags* is used to make the
3071 * distinction (ingress path is selected if the flag is present,
3072 * egress path otherwise). This is the only flag supported for now.
3073 * Return
3074 * **SK_PASS** on success, or **SK_DROP** on error.
3075 *
3076 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3077 * Description
3078 * This helper is used in programs implementing policies at the
3079 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3080 * if the verdict eBPF program returns **SK_PASS**), redirect it
3081 * to the socket referenced by *map* (of type
3082 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3083 * egress interfaces can be used for redirection. The
3084 * **BPF_F_INGRESS** value in *flags* is used to make the
3085 * distinction (ingress path is selected if the flag is present,
3086 * egress otherwise). This is the only flag supported for now.
3087 * Return
3088 * **SK_PASS** on success, or **SK_DROP** on error.
3089 *
3090 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3091 * Description
3092 * Encapsulate the packet associated to *skb* within a Layer 3
3093 * protocol header. This header is provided in the buffer at
3094 * address *hdr*, with *len* its size in bytes. *type* indicates
3095 * the protocol of the header and can be one of:
3096 *
3097 * **BPF_LWT_ENCAP_SEG6**
3098 * IPv6 encapsulation with Segment Routing Header
3099 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3100 * the IPv6 header is computed by the kernel.
3101 * **BPF_LWT_ENCAP_SEG6_INLINE**
3102 * Only works if *skb* contains an IPv6 packet. Insert a
3103 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3104 * the IPv6 header.
3105 * **BPF_LWT_ENCAP_IP**
3106 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3107 * must be IPv4 or IPv6, followed by zero or more
3108 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3109 * total bytes in all prepended headers. Please note that
3110 * if **skb_is_gso**\ (*skb*) is true, no more than two
3111 * headers can be prepended, and the inner header, if
3112 * present, should be either GRE or UDP/GUE.
3113 *
3114 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3115 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3116 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3117 * **BPF_PROG_TYPE_LWT_XMIT**.
3118 *
3119 * A call to this helper is susceptible to change the underlying
3120 * packet buffer. Therefore, at load time, all checks on pointers
3121 * previously done by the verifier are invalidated and must be
3122 * performed again, if the helper is used in combination with
3123 * direct packet access.
3124 * Return
3125 * 0 on success, or a negative error in case of failure.
3126 *
3127 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3128 * Description
3129 * Store *len* bytes from address *from* into the packet
3130 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3131 * inside the outermost IPv6 Segment Routing Header can be
3132 * modified through this helper.
3133 *
3134 * A call to this helper is susceptible to change the underlying
3135 * packet buffer. Therefore, at load time, all checks on pointers
3136 * previously done by the verifier are invalidated and must be
3137 * performed again, if the helper is used in combination with
3138 * direct packet access.
3139 * Return
3140 * 0 on success, or a negative error in case of failure.
3141 *
3142 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3143 * Description
3144 * Adjust the size allocated to TLVs in the outermost IPv6
3145 * Segment Routing Header contained in the packet associated to
3146 * *skb*, at position *offset* by *delta* bytes. Only offsets
3147 * after the segments are accepted. *delta* can be as well
3148 * positive (growing) as negative (shrinking).
3149 *
3150 * A call to this helper is susceptible to change the underlying
3151 * packet buffer. Therefore, at load time, all checks on pointers
3152 * previously done by the verifier are invalidated and must be
3153 * performed again, if the helper is used in combination with
3154 * direct packet access.
3155 * Return
3156 * 0 on success, or a negative error in case of failure.
3157 *
3158 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3159 * Description
3160 * Apply an IPv6 Segment Routing action of type *action* to the
3161 * packet associated to *skb*. Each action takes a parameter
3162 * contained at address *param*, and of length *param_len* bytes.
3163 * *action* can be one of:
3164 *
3165 * **SEG6_LOCAL_ACTION_END_X**
3166 * End.X action: Endpoint with Layer-3 cross-connect.
3167 * Type of *param*: **struct in6_addr**.
3168 * **SEG6_LOCAL_ACTION_END_T**
3169 * End.T action: Endpoint with specific IPv6 table lookup.
3170 * Type of *param*: **int**.
3171 * **SEG6_LOCAL_ACTION_END_B6**
3172 * End.B6 action: Endpoint bound to an SRv6 policy.
3173 * Type of *param*: **struct ipv6_sr_hdr**.
3174 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3175 * End.B6.Encap action: Endpoint bound to an SRv6
3176 * encapsulation policy.
3177 * Type of *param*: **struct ipv6_sr_hdr**.
3178 *
3179 * A call to this helper is susceptible to change the underlying
3180 * packet buffer. Therefore, at load time, all checks on pointers
3181 * previously done by the verifier are invalidated and must be
3182 * performed again, if the helper is used in combination with
3183 * direct packet access.
3184 * Return
3185 * 0 on success, or a negative error in case of failure.
3186 *
3187 * long bpf_rc_repeat(void *ctx)
3188 * Description
3189 * This helper is used in programs implementing IR decoding, to
3190 * report a successfully decoded repeat key message. This delays
3191 * the generation of a key up event for previously generated
3192 * key down event.
3193 *
3194 * Some IR protocols like NEC have a special IR message for
3195 * repeating last button, for when a button is held down.
3196 *
3197 * The *ctx* should point to the lirc sample as passed into
3198 * the program.
3199 *
3200 * This helper is only available is the kernel was compiled with
3201 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3202 * "**y**".
3203 * Return
3204 * 0
3205 *
3206 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3207 * Description
3208 * This helper is used in programs implementing IR decoding, to
3209 * report a successfully decoded key press with *scancode*,
3210 * *toggle* value in the given *protocol*. The scancode will be
3211 * translated to a keycode using the rc keymap, and reported as
3212 * an input key down event. After a period a key up event is
3213 * generated. This period can be extended by calling either
3214 * **bpf_rc_keydown**\ () again with the same values, or calling
3215 * **bpf_rc_repeat**\ ().
3216 *
3217 * Some protocols include a toggle bit, in case the button was
3218 * released and pressed again between consecutive scancodes.
3219 *
3220 * The *ctx* should point to the lirc sample as passed into
3221 * the program.
3222 *
3223 * The *protocol* is the decoded protocol number (see
3224 * **enum rc_proto** for some predefined values).
3225 *
3226 * This helper is only available is the kernel was compiled with
3227 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3228 * "**y**".
3229 * Return
3230 * 0
3231 *
3232 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3233 * Description
3234 * Return the cgroup v2 id of the socket associated with the *skb*.
3235 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3236 * helper for cgroup v1 by providing a tag resp. identifier that
3237 * can be matched on or used for map lookups e.g. to implement
3238 * policy. The cgroup v2 id of a given path in the hierarchy is
3239 * exposed in user space through the f_handle API in order to get
3240 * to the same 64-bit id.
3241 *
3242 * This helper can be used on TC egress path, but not on ingress,
3243 * and is available only if the kernel was compiled with the
3244 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3245 * Return
3246 * The id is returned or 0 in case the id could not be retrieved.
3247 *
3248 * u64 bpf_get_current_cgroup_id(void)
3249 * Return
3250 * A 64-bit integer containing the current cgroup id based
3251 * on the cgroup within which the current task is running.
3252 *
3253 * void *bpf_get_local_storage(void *map, u64 flags)
3254 * Description
3255 * Get the pointer to the local storage area.
3256 * The type and the size of the local storage is defined
3257 * by the *map* argument.
3258 * The *flags* meaning is specific for each map type,
3259 * and has to be 0 for cgroup local storage.
3260 *
3261 * Depending on the BPF program type, a local storage area
3262 * can be shared between multiple instances of the BPF program,
3263 * running simultaneously.
3264 *
3265 * A user should care about the synchronization by himself.
3266 * For example, by using the **BPF_ATOMIC** instructions to alter
3267 * the shared data.
3268 * Return
3269 * A pointer to the local storage area.
3270 *
3271 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3272 * Description
3273 * Select a **SO_REUSEPORT** socket from a
3274 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3275 * It checks the selected socket is matching the incoming
3276 * request in the socket buffer.
3277 * Return
3278 * 0 on success, or a negative error in case of failure.
3279 *
3280 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3281 * Description
3282 * Return id of cgroup v2 that is ancestor of cgroup associated
3283 * with the *skb* at the *ancestor_level*. The root cgroup is at
3284 * *ancestor_level* zero and each step down the hierarchy
3285 * increments the level. If *ancestor_level* == level of cgroup
3286 * associated with *skb*, then return value will be same as that
3287 * of **bpf_skb_cgroup_id**\ ().
3288 *
3289 * The helper is useful to implement policies based on cgroups
3290 * that are upper in hierarchy than immediate cgroup associated
3291 * with *skb*.
3292 *
3293 * The format of returned id and helper limitations are same as in
3294 * **bpf_skb_cgroup_id**\ ().
3295 * Return
3296 * The id is returned or 0 in case the id could not be retrieved.
3297 *
3298 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3299 * Description
3300 * Look for TCP socket matching *tuple*, optionally in a child
3301 * network namespace *netns*. The return value must be checked,
3302 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3303 *
3304 * The *ctx* should point to the context of the program, such as
3305 * the skb or socket (depending on the hook in use). This is used
3306 * to determine the base network namespace for the lookup.
3307 *
3308 * *tuple_size* must be one of:
3309 *
3310 * **sizeof**\ (*tuple*\ **->ipv4**)
3311 * Look for an IPv4 socket.
3312 * **sizeof**\ (*tuple*\ **->ipv6**)
3313 * Look for an IPv6 socket.
3314 *
3315 * If the *netns* is a negative signed 32-bit integer, then the
3316 * socket lookup table in the netns associated with the *ctx*
3317 * will be used. For the TC hooks, this is the netns of the device
3318 * in the skb. For socket hooks, this is the netns of the socket.
3319 * If *netns* is any other signed 32-bit value greater than or
3320 * equal to zero then it specifies the ID of the netns relative to
3321 * the netns associated with the *ctx*. *netns* values beyond the
3322 * range of 32-bit integers are reserved for future use.
3323 *
3324 * All values for *flags* are reserved for future usage, and must
3325 * be left at zero.
3326 *
3327 * This helper is available only if the kernel was compiled with
3328 * **CONFIG_NET** configuration option.
3329 * Return
3330 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3331 * For sockets with reuseport option, the **struct bpf_sock**
3332 * result is from *reuse*\ **->socks**\ [] using the hash of the
3333 * tuple.
3334 *
3335 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3336 * Description
3337 * Look for UDP socket matching *tuple*, optionally in a child
3338 * network namespace *netns*. The return value must be checked,
3339 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3340 *
3341 * The *ctx* should point to the context of the program, such as
3342 * the skb or socket (depending on the hook in use). This is used
3343 * to determine the base network namespace for the lookup.
3344 *
3345 * *tuple_size* must be one of:
3346 *
3347 * **sizeof**\ (*tuple*\ **->ipv4**)
3348 * Look for an IPv4 socket.
3349 * **sizeof**\ (*tuple*\ **->ipv6**)
3350 * Look for an IPv6 socket.
3351 *
3352 * If the *netns* is a negative signed 32-bit integer, then the
3353 * socket lookup table in the netns associated with the *ctx*
3354 * will be used. For the TC hooks, this is the netns of the device
3355 * in the skb. For socket hooks, this is the netns of the socket.
3356 * If *netns* is any other signed 32-bit value greater than or
3357 * equal to zero then it specifies the ID of the netns relative to
3358 * the netns associated with the *ctx*. *netns* values beyond the
3359 * range of 32-bit integers are reserved for future use.
3360 *
3361 * All values for *flags* are reserved for future usage, and must
3362 * be left at zero.
3363 *
3364 * This helper is available only if the kernel was compiled with
3365 * **CONFIG_NET** configuration option.
3366 * Return
3367 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3368 * For sockets with reuseport option, the **struct bpf_sock**
3369 * result is from *reuse*\ **->socks**\ [] using the hash of the
3370 * tuple.
3371 *
3372 * long bpf_sk_release(void *sock)
3373 * Description
3374 * Release the reference held by *sock*. *sock* must be a
3375 * non-**NULL** pointer that was returned from
3376 * **bpf_sk_lookup_xxx**\ ().
3377 * Return
3378 * 0 on success, or a negative error in case of failure.
3379 *
3380 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3381 * Description
3382 * Push an element *value* in *map*. *flags* is one of:
3383 *
3384 * **BPF_EXIST**
3385 * If the queue/stack is full, the oldest element is
3386 * removed to make room for this.
3387 * Return
3388 * 0 on success, or a negative error in case of failure.
3389 *
3390 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3391 * Description
3392 * Pop an element from *map*.
3393 * Return
3394 * 0 on success, or a negative error in case of failure.
3395 *
3396 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3397 * Description
3398 * Get an element from *map* without removing it.
3399 * Return
3400 * 0 on success, or a negative error in case of failure.
3401 *
3402 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3403 * Description
3404 * For socket policies, insert *len* bytes into *msg* at offset
3405 * *start*.
3406 *
3407 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3408 * *msg* it may want to insert metadata or options into the *msg*.
3409 * This can later be read and used by any of the lower layer BPF
3410 * hooks.
3411 *
3412 * This helper may fail if under memory pressure (a malloc
3413 * fails) in these cases BPF programs will get an appropriate
3414 * error and BPF programs will need to handle them.
3415 * Return
3416 * 0 on success, or a negative error in case of failure.
3417 *
3418 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3419 * Description
3420 * Will remove *len* bytes from a *msg* starting at byte *start*.
3421 * This may result in **ENOMEM** errors under certain situations if
3422 * an allocation and copy are required due to a full ring buffer.
3423 * However, the helper will try to avoid doing the allocation
3424 * if possible. Other errors can occur if input parameters are
3425 * invalid either due to *start* byte not being valid part of *msg*
3426 * payload and/or *pop* value being to large.
3427 * Return
3428 * 0 on success, or a negative error in case of failure.
3429 *
3430 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3431 * Description
3432 * This helper is used in programs implementing IR decoding, to
3433 * report a successfully decoded pointer movement.
3434 *
3435 * The *ctx* should point to the lirc sample as passed into
3436 * the program.
3437 *
3438 * This helper is only available is the kernel was compiled with
3439 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3440 * "**y**".
3441 * Return
3442 * 0
3443 *
3444 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3445 * Description
3446 * Acquire a spinlock represented by the pointer *lock*, which is
3447 * stored as part of a value of a map. Taking the lock allows to
3448 * safely update the rest of the fields in that value. The
3449 * spinlock can (and must) later be released with a call to
3450 * **bpf_spin_unlock**\ (\ *lock*\ ).
3451 *
3452 * Spinlocks in BPF programs come with a number of restrictions
3453 * and constraints:
3454 *
3455 * * **bpf_spin_lock** objects are only allowed inside maps of
3456 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3457 * list could be extended in the future).
3458 * * BTF description of the map is mandatory.
3459 * * The BPF program can take ONE lock at a time, since taking two
3460 * or more could cause dead locks.
3461 * * Only one **struct bpf_spin_lock** is allowed per map element.
3462 * * When the lock is taken, calls (either BPF to BPF or helpers)
3463 * are not allowed.
3464 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3465 * allowed inside a spinlock-ed region.
3466 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3467 * the lock, on all execution paths, before it returns.
3468 * * The BPF program can access **struct bpf_spin_lock** only via
3469 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3470 * helpers. Loading or storing data into the **struct
3471 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3472 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3473 * of the map value must be a struct and have **struct
3474 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3475 * Nested lock inside another struct is not allowed.
3476 * * The **struct bpf_spin_lock** *lock* field in a map value must
3477 * be aligned on a multiple of 4 bytes in that value.
3478 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3479 * the **bpf_spin_lock** field to user space.
3480 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3481 * a BPF program, do not update the **bpf_spin_lock** field.
3482 * * **bpf_spin_lock** cannot be on the stack or inside a
3483 * networking packet (it can only be inside of a map values).
3484 * * **bpf_spin_lock** is available to root only.
3485 * * Tracing programs and socket filter programs cannot use
3486 * **bpf_spin_lock**\ () due to insufficient preemption checks
3487 * (but this may change in the future).
3488 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3489 * Return
3490 * 0
3491 *
3492 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3493 * Description
3494 * Release the *lock* previously locked by a call to
3495 * **bpf_spin_lock**\ (\ *lock*\ ).
3496 * Return
3497 * 0
3498 *
3499 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3500 * Description
3501 * This helper gets a **struct bpf_sock** pointer such
3502 * that all the fields in this **bpf_sock** can be accessed.
3503 * Return
3504 * A **struct bpf_sock** pointer on success, or **NULL** in
3505 * case of failure.
3506 *
3507 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3508 * Description
3509 * This helper gets a **struct bpf_tcp_sock** pointer from a
3510 * **struct bpf_sock** pointer.
3511 * Return
3512 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3513 * case of failure.
3514 *
3515 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3516 * Description
3517 * Set ECN (Explicit Congestion Notification) field of IP header
3518 * to **CE** (Congestion Encountered) if current value is **ECT**
3519 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3520 * and IPv4.
3521 * Return
3522 * 1 if the **CE** flag is set (either by the current helper call
3523 * or because it was already present), 0 if it is not set.
3524 *
3525 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3526 * Description
3527 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3528 * **bpf_sk_release**\ () is unnecessary and not allowed.
3529 * Return
3530 * A **struct bpf_sock** pointer on success, or **NULL** in
3531 * case of failure.
3532 *
3533 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3534 * Description
3535 * Look for TCP socket matching *tuple*, optionally in a child
3536 * network namespace *netns*. The return value must be checked,
3537 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3538 *
3539 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3540 * that it also returns timewait or request sockets. Use
3541 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3542 * full structure.
3543 *
3544 * This helper is available only if the kernel was compiled with
3545 * **CONFIG_NET** configuration option.
3546 * Return
3547 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3548 * For sockets with reuseport option, the **struct bpf_sock**
3549 * result is from *reuse*\ **->socks**\ [] using the hash of the
3550 * tuple.
3551 *
3552 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3553 * Description
3554 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3555 * the listening socket in *sk*.
3556 *
3557 * *iph* points to the start of the IPv4 or IPv6 header, while
3558 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3559 * **sizeof**\ (**struct ip6hdr**).
3560 *
3561 * *th* points to the start of the TCP header, while *th_len*
3562 * contains **sizeof**\ (**struct tcphdr**).
3563 * Return
3564 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3565 * error otherwise.
3566 *
3567 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3568 * Description
3569 * Get name of sysctl in /proc/sys/ and copy it into provided by
3570 * program buffer *buf* of size *buf_len*.
3571 *
3572 * The buffer is always NUL terminated, unless it's zero-sized.
3573 *
3574 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3575 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3576 * only (e.g. "tcp_mem").
3577 * Return
3578 * Number of character copied (not including the trailing NUL).
3579 *
3580 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3581 * truncated name in this case).
3582 *
3583 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3584 * Description
3585 * Get current value of sysctl as it is presented in /proc/sys
3586 * (incl. newline, etc), and copy it as a string into provided
3587 * by program buffer *buf* of size *buf_len*.
3588 *
3589 * The whole value is copied, no matter what file position user
3590 * space issued e.g. sys_read at.
3591 *
3592 * The buffer is always NUL terminated, unless it's zero-sized.
3593 * Return
3594 * Number of character copied (not including the trailing NUL).
3595 *
3596 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3597 * truncated name in this case).
3598 *
3599 * **-EINVAL** if current value was unavailable, e.g. because
3600 * sysctl is uninitialized and read returns -EIO for it.
3601 *
3602 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3603 * Description
3604 * Get new value being written by user space to sysctl (before
3605 * the actual write happens) and copy it as a string into
3606 * provided by program buffer *buf* of size *buf_len*.
3607 *
3608 * User space may write new value at file position > 0.
3609 *
3610 * The buffer is always NUL terminated, unless it's zero-sized.
3611 * Return
3612 * Number of character copied (not including the trailing NUL).
3613 *
3614 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3615 * truncated name in this case).
3616 *
3617 * **-EINVAL** if sysctl is being read.
3618 *
3619 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3620 * Description
3621 * Override new value being written by user space to sysctl with
3622 * value provided by program in buffer *buf* of size *buf_len*.
3623 *
3624 * *buf* should contain a string in same form as provided by user
3625 * space on sysctl write.
3626 *
3627 * User space may write new value at file position > 0. To override
3628 * the whole sysctl value file position should be set to zero.
3629 * Return
3630 * 0 on success.
3631 *
3632 * **-E2BIG** if the *buf_len* is too big.
3633 *
3634 * **-EINVAL** if sysctl is being read.
3635 *
3636 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3637 * Description
3638 * Convert the initial part of the string from buffer *buf* of
3639 * size *buf_len* to a long integer according to the given base
3640 * and save the result in *res*.
3641 *
3642 * The string may begin with an arbitrary amount of white space
3643 * (as determined by **isspace**\ (3)) followed by a single
3644 * optional '**-**' sign.
3645 *
3646 * Five least significant bits of *flags* encode base, other bits
3647 * are currently unused.
3648 *
3649 * Base must be either 8, 10, 16 or 0 to detect it automatically
3650 * similar to user space **strtol**\ (3).
3651 * Return
3652 * Number of characters consumed on success. Must be positive but
3653 * no more than *buf_len*.
3654 *
3655 * **-EINVAL** if no valid digits were found or unsupported base
3656 * was provided.
3657 *
3658 * **-ERANGE** if resulting value was out of range.
3659 *
3660 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3661 * Description
3662 * Convert the initial part of the string from buffer *buf* of
3663 * size *buf_len* to an unsigned long integer according to the
3664 * given base and save the result in *res*.
3665 *
3666 * The string may begin with an arbitrary amount of white space
3667 * (as determined by **isspace**\ (3)).
3668 *
3669 * Five least significant bits of *flags* encode base, other bits
3670 * are currently unused.
3671 *
3672 * Base must be either 8, 10, 16 or 0 to detect it automatically
3673 * similar to user space **strtoul**\ (3).
3674 * Return
3675 * Number of characters consumed on success. Must be positive but
3676 * no more than *buf_len*.
3677 *
3678 * **-EINVAL** if no valid digits were found or unsupported base
3679 * was provided.
3680 *
3681 * **-ERANGE** if resulting value was out of range.
3682 *
3683 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3684 * Description
3685 * Get a bpf-local-storage from a *sk*.
3686 *
3687 * Logically, it could be thought of getting the value from
3688 * a *map* with *sk* as the **key**. From this
3689 * perspective, the usage is not much different from
3690 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3691 * helper enforces the key must be a full socket and the map must
3692 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
3693 *
3694 * Underneath, the value is stored locally at *sk* instead of
3695 * the *map*. The *map* is used as the bpf-local-storage
3696 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3697 * searched against all bpf-local-storages residing at *sk*.
3698 *
3699 * *sk* is a kernel **struct sock** pointer for LSM program.
3700 * *sk* is a **struct bpf_sock** pointer for other program types.
3701 *
3702 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3703 * used such that a new bpf-local-storage will be
3704 * created if one does not exist. *value* can be used
3705 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3706 * the initial value of a bpf-local-storage. If *value* is
3707 * **NULL**, the new bpf-local-storage will be zero initialized.
3708 * Return
3709 * A bpf-local-storage pointer is returned on success.
3710 *
3711 * **NULL** if not found or there was an error in adding
3712 * a new bpf-local-storage.
3713 *
3714 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3715 * Description
3716 * Delete a bpf-local-storage from a *sk*.
3717 * Return
3718 * 0 on success.
3719 *
3720 * **-ENOENT** if the bpf-local-storage cannot be found.
3721 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3722 *
3723 * long bpf_send_signal(u32 sig)
3724 * Description
3725 * Send signal *sig* to the process of the current task.
3726 * The signal may be delivered to any of this process's threads.
3727 * Return
3728 * 0 on success or successfully queued.
3729 *
3730 * **-EBUSY** if work queue under nmi is full.
3731 *
3732 * **-EINVAL** if *sig* is invalid.
3733 *
3734 * **-EPERM** if no permission to send the *sig*.
3735 *
3736 * **-EAGAIN** if bpf program can try again.
3737 *
3738 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3739 * Description
3740 * Try to issue a SYN cookie for the packet with corresponding
3741 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3742 *
3743 * *iph* points to the start of the IPv4 or IPv6 header, while
3744 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3745 * **sizeof**\ (**struct ip6hdr**).
3746 *
3747 * *th* points to the start of the TCP header, while *th_len*
3748 * contains the length of the TCP header.
3749 * Return
3750 * On success, lower 32 bits hold the generated SYN cookie in
3751 * followed by 16 bits which hold the MSS value for that cookie,
3752 * and the top 16 bits are unused.
3753 *
3754 * On failure, the returned value is one of the following:
3755 *
3756 * **-EINVAL** SYN cookie cannot be issued due to error
3757 *
3758 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
3759 *
3760 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3761 *
3762 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
3763 *
3764 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3765 * Description
3766 * Write raw *data* blob into a special BPF perf event held by
3767 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3768 * event must have the following attributes: **PERF_SAMPLE_RAW**
3769 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3770 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3771 *
3772 * The *flags* are used to indicate the index in *map* for which
3773 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3774 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3775 * to indicate that the index of the current CPU core should be
3776 * used.
3777 *
3778 * The value to write, of *size*, is passed through eBPF stack and
3779 * pointed by *data*.
3780 *
3781 * *ctx* is a pointer to in-kernel struct sk_buff.
3782 *
3783 * This helper is similar to **bpf_perf_event_output**\ () but
3784 * restricted to raw_tracepoint bpf programs.
3785 * Return
3786 * 0 on success, or a negative error in case of failure.
3787 *
3788 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3789 * Description
3790 * Safely attempt to read *size* bytes from user space address
3791 * *unsafe_ptr* and store the data in *dst*.
3792 * Return
3793 * 0 on success, or a negative error in case of failure.
3794 *
3795 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3796 * Description
3797 * Safely attempt to read *size* bytes from kernel space address
3798 * *unsafe_ptr* and store the data in *dst*.
3799 * Return
3800 * 0 on success, or a negative error in case of failure.
3801 *
3802 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3803 * Description
3804 * Copy a NUL terminated string from an unsafe user address
3805 * *unsafe_ptr* to *dst*. The *size* should include the
3806 * terminating NUL byte. In case the string length is smaller than
3807 * *size*, the target is not padded with further NUL bytes. If the
3808 * string length is larger than *size*, just *size*-1 bytes are
3809 * copied and the last byte is set to NUL.
3810 *
3811 * On success, returns the number of bytes that were written,
3812 * including the terminal NUL. This makes this helper useful in
3813 * tracing programs for reading strings, and more importantly to
3814 * get its length at runtime. See the following snippet:
3815 *
3816 * ::
3817 *
3818 * SEC("kprobe/sys_open")
3819 * void bpf_sys_open(struct pt_regs *ctx)
3820 * {
3821 * char buf[PATHLEN]; // PATHLEN is defined to 256
3822 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3823 * ctx->di);
3824 *
3825 * // Consume buf, for example push it to
3826 * // userspace via bpf_perf_event_output(); we
3827 * // can use res (the string length) as event
3828 * // size, after checking its boundaries.
3829 * }
3830 *
3831 * In comparison, using **bpf_probe_read_user**\ () helper here
3832 * instead to read the string would require to estimate the length
3833 * at compile time, and would often result in copying more memory
3834 * than necessary.
3835 *
3836 * Another useful use case is when parsing individual process
3837 * arguments or individual environment variables navigating
3838 * *current*\ **->mm->arg_start** and *current*\
3839 * **->mm->env_start**: using this helper and the return value,
3840 * one can quickly iterate at the right offset of the memory area.
3841 * Return
3842 * On success, the strictly positive length of the output string,
3843 * including the trailing NUL character. On error, a negative
3844 * value.
3845 *
3846 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3847 * Description
3848 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3849 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3850 * Return
3851 * On success, the strictly positive length of the string, including
3852 * the trailing NUL character. On error, a negative value.
3853 *
3854 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3855 * Description
3856 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3857 * *rcv_nxt* is the ack_seq to be sent out.
3858 * Return
3859 * 0 on success, or a negative error in case of failure.
3860 *
3861 * long bpf_send_signal_thread(u32 sig)
3862 * Description
3863 * Send signal *sig* to the thread corresponding to the current task.
3864 * Return
3865 * 0 on success or successfully queued.
3866 *
3867 * **-EBUSY** if work queue under nmi is full.
3868 *
3869 * **-EINVAL** if *sig* is invalid.
3870 *
3871 * **-EPERM** if no permission to send the *sig*.
3872 *
3873 * **-EAGAIN** if bpf program can try again.
3874 *
3875 * u64 bpf_jiffies64(void)
3876 * Description
3877 * Obtain the 64bit jiffies
3878 * Return
3879 * The 64 bit jiffies
3880 *
3881 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3882 * Description
3883 * For an eBPF program attached to a perf event, retrieve the
3884 * branch records (**struct perf_branch_entry**) associated to *ctx*
3885 * and store it in the buffer pointed by *buf* up to size
3886 * *size* bytes.
3887 * Return
3888 * On success, number of bytes written to *buf*. On error, a
3889 * negative value.
3890 *
3891 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3892 * instead return the number of bytes required to store all the
3893 * branch entries. If this flag is set, *buf* may be NULL.
3894 *
3895 * **-EINVAL** if arguments invalid or **size** not a multiple
3896 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3897 *
3898 * **-ENOENT** if architecture does not support branch records.
3899 *
3900 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3901 * Description
3902 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3903 * *namespace* will be returned in *nsdata*.
3904 * Return
3905 * 0 on success, or one of the following in case of failure:
3906 *
3907 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3908 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3909 *
3910 * **-ENOENT** if pidns does not exists for the current task.
3911 *
3912 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3913 * Description
3914 * Write raw *data* blob into a special BPF perf event held by
3915 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3916 * event must have the following attributes: **PERF_SAMPLE_RAW**
3917 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3918 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3919 *
3920 * The *flags* are used to indicate the index in *map* for which
3921 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3922 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3923 * to indicate that the index of the current CPU core should be
3924 * used.
3925 *
3926 * The value to write, of *size*, is passed through eBPF stack and
3927 * pointed by *data*.
3928 *
3929 * *ctx* is a pointer to in-kernel struct xdp_buff.
3930 *
3931 * This helper is similar to **bpf_perf_eventoutput**\ () but
3932 * restricted to raw_tracepoint bpf programs.
3933 * Return
3934 * 0 on success, or a negative error in case of failure.
3935 *
3936 * u64 bpf_get_netns_cookie(void *ctx)
3937 * Description
3938 * Retrieve the cookie (generated by the kernel) of the network
3939 * namespace the input *ctx* is associated with. The network
3940 * namespace cookie remains stable for its lifetime and provides
3941 * a global identifier that can be assumed unique. If *ctx* is
3942 * NULL, then the helper returns the cookie for the initial
3943 * network namespace. The cookie itself is very similar to that
3944 * of **bpf_get_socket_cookie**\ () helper, but for network
3945 * namespaces instead of sockets.
3946 * Return
3947 * A 8-byte long opaque number.
3948 *
3949 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3950 * Description
3951 * Return id of cgroup v2 that is ancestor of the cgroup associated
3952 * with the current task at the *ancestor_level*. The root cgroup
3953 * is at *ancestor_level* zero and each step down the hierarchy
3954 * increments the level. If *ancestor_level* == level of cgroup
3955 * associated with the current task, then return value will be the
3956 * same as that of **bpf_get_current_cgroup_id**\ ().
3957 *
3958 * The helper is useful to implement policies based on cgroups
3959 * that are upper in hierarchy than immediate cgroup associated
3960 * with the current task.
3961 *
3962 * The format of returned id and helper limitations are same as in
3963 * **bpf_get_current_cgroup_id**\ ().
3964 * Return
3965 * The id is returned or 0 in case the id could not be retrieved.
3966 *
3967 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3968 * Description
3969 * Helper is overloaded depending on BPF program type. This
3970 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3971 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3972 *
3973 * Assign the *sk* to the *skb*. When combined with appropriate
3974 * routing configuration to receive the packet towards the socket,
3975 * will cause *skb* to be delivered to the specified socket.
3976 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3977 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3978 * interfere with successful delivery to the socket.
3979 *
3980 * This operation is only valid from TC ingress path.
3981 *
3982 * The *flags* argument must be zero.
3983 * Return
3984 * 0 on success, or a negative error in case of failure:
3985 *
3986 * **-EINVAL** if specified *flags* are not supported.
3987 *
3988 * **-ENOENT** if the socket is unavailable for assignment.
3989 *
3990 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3991 *
3992 * **-EOPNOTSUPP** if the operation is not supported, for example
3993 * a call from outside of TC ingress.
3994 *
3995 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3996 * (reuseport).
3997 *
3998 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3999 * Description
4000 * Helper is overloaded depending on BPF program type. This
4001 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4002 *
4003 * Select the *sk* as a result of a socket lookup.
4004 *
4005 * For the operation to succeed passed socket must be compatible
4006 * with the packet description provided by the *ctx* object.
4007 *
4008 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4009 * be an exact match. While IP family (**AF_INET** or
4010 * **AF_INET6**) must be compatible, that is IPv6 sockets
4011 * that are not v6-only can be selected for IPv4 packets.
4012 *
4013 * Only TCP listeners and UDP unconnected sockets can be
4014 * selected. *sk* can also be NULL to reset any previous
4015 * selection.
4016 *
4017 * *flags* argument can combination of following values:
4018 *
4019 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4020 * socket selection, potentially done by a BPF program
4021 * that ran before us.
4022 *
4023 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4024 * load-balancing within reuseport group for the socket
4025 * being selected.
4026 *
4027 * On success *ctx->sk* will point to the selected socket.
4028 *
4029 * Return
4030 * 0 on success, or a negative errno in case of failure.
4031 *
4032 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4033 * not compatible with packet family (*ctx->family*).
4034 *
4035 * * **-EEXIST** if socket has been already selected,
4036 * potentially by another program, and
4037 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4038 *
4039 * * **-EINVAL** if unsupported flags were specified.
4040 *
4041 * * **-EPROTOTYPE** if socket L4 protocol
4042 * (*sk->protocol*) doesn't match packet protocol
4043 * (*ctx->protocol*).
4044 *
4045 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4046 * state (TCP listening or UDP unconnected).
4047 *
4048 * u64 bpf_ktime_get_boot_ns(void)
4049 * Description
4050 * Return the time elapsed since system boot, in nanoseconds.
4051 * Does include the time the system was suspended.
4052 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4053 * Return
4054 * Current *ktime*.
4055 *
4056 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4057 * Description
4058 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4059 * out the format string.
4060 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4061 * the format string itself. The *data* and *data_len* are format string
4062 * arguments. The *data* are a **u64** array and corresponding format string
4063 * values are stored in the array. For strings and pointers where pointees
4064 * are accessed, only the pointer values are stored in the *data* array.
4065 * The *data_len* is the size of *data* in bytes.
4066 *
4067 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4068 * Reading kernel memory may fail due to either invalid address or
4069 * valid address but requiring a major memory fault. If reading kernel memory
4070 * fails, the string for **%s** will be an empty string, and the ip
4071 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4072 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4073 * Return
4074 * 0 on success, or a negative error in case of failure:
4075 *
4076 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4077 * by returning 1 from bpf program.
4078 *
4079 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4080 *
4081 * **-E2BIG** if *fmt* contains too many format specifiers.
4082 *
4083 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4084 *
4085 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4086 * Description
4087 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4088 * The *m* represents the seq_file. The *data* and *len* represent the
4089 * data to write in bytes.
4090 * Return
4091 * 0 on success, or a negative error in case of failure:
4092 *
4093 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4094 *
4095 * u64 bpf_sk_cgroup_id(void *sk)
4096 * Description
4097 * Return the cgroup v2 id of the socket *sk*.
4098 *
4099 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4100 * returned from **bpf_sk_lookup_xxx**\ (),
4101 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4102 * same as in **bpf_skb_cgroup_id**\ ().
4103 *
4104 * This helper is available only if the kernel was compiled with
4105 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4106 * Return
4107 * The id is returned or 0 in case the id could not be retrieved.
4108 *
4109 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4110 * Description
4111 * Return id of cgroup v2 that is ancestor of cgroup associated
4112 * with the *sk* at the *ancestor_level*. The root cgroup is at
4113 * *ancestor_level* zero and each step down the hierarchy
4114 * increments the level. If *ancestor_level* == level of cgroup
4115 * associated with *sk*, then return value will be same as that
4116 * of **bpf_sk_cgroup_id**\ ().
4117 *
4118 * The helper is useful to implement policies based on cgroups
4119 * that are upper in hierarchy than immediate cgroup associated
4120 * with *sk*.
4121 *
4122 * The format of returned id and helper limitations are same as in
4123 * **bpf_sk_cgroup_id**\ ().
4124 * Return
4125 * The id is returned or 0 in case the id could not be retrieved.
4126 *
4127 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4128 * Description
4129 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4130 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4131 * of new data availability is sent.
4132 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4133 * of new data availability is sent unconditionally.
4134 * If **0** is specified in *flags*, an adaptive notification
4135 * of new data availability is sent.
4136 *
4137 * An adaptive notification is a notification sent whenever the user-space
4138 * process has caught up and consumed all available payloads. In case the user-space
4139 * process is still processing a previous payload, then no notification is needed
4140 * as it will process the newly added payload automatically.
4141 * Return
4142 * 0 on success, or a negative error in case of failure.
4143 *
4144 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4145 * Description
4146 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4147 * *flags* must be 0.
4148 * Return
4149 * Valid pointer with *size* bytes of memory available; NULL,
4150 * otherwise.
4151 *
4152 * void bpf_ringbuf_submit(void *data, u64 flags)
4153 * Description
4154 * Submit reserved ring buffer sample, pointed to by *data*.
4155 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4156 * of new data availability is sent.
4157 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4158 * of new data availability is sent unconditionally.
4159 * If **0** is specified in *flags*, an adaptive notification
4160 * of new data availability is sent.
4161 *
4162 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4163 * Return
4164 * Nothing. Always succeeds.
4165 *
4166 * void bpf_ringbuf_discard(void *data, u64 flags)
4167 * Description
4168 * Discard reserved ring buffer sample, pointed to by *data*.
4169 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4170 * of new data availability is sent.
4171 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4172 * of new data availability is sent unconditionally.
4173 * If **0** is specified in *flags*, an adaptive notification
4174 * of new data availability is sent.
4175 *
4176 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4177 * Return
4178 * Nothing. Always succeeds.
4179 *
4180 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4181 * Description
4182 * Query various characteristics of provided ring buffer. What
4183 * exactly is queries is determined by *flags*:
4184 *
4185 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4186 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4187 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4188 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4189 *
4190 * Data returned is just a momentary snapshot of actual values
4191 * and could be inaccurate, so this facility should be used to
4192 * power heuristics and for reporting, not to make 100% correct
4193 * calculation.
4194 * Return
4195 * Requested value, or 0, if *flags* are not recognized.
4196 *
4197 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4198 * Description
4199 * Change the skbs checksum level by one layer up or down, or
4200 * reset it entirely to none in order to have the stack perform
4201 * checksum validation. The level is applicable to the following
4202 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4203 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4204 * through **bpf_skb_adjust_room**\ () helper with passing in
4205 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4206 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4207 * the UDP header is removed. Similarly, an encap of the latter
4208 * into the former could be accompanied by a helper call to
4209 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4210 * skb is still intended to be processed in higher layers of the
4211 * stack instead of just egressing at tc.
4212 *
4213 * There are three supported level settings at this time:
4214 *
4215 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4216 * with CHECKSUM_UNNECESSARY.
4217 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4218 * with CHECKSUM_UNNECESSARY.
4219 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4220 * sets CHECKSUM_NONE to force checksum validation by the stack.
4221 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4222 * skb->csum_level.
4223 * Return
4224 * 0 on success, or a negative error in case of failure. In the
4225 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4226 * is returned or the error code -EACCES in case the skb is not
4227 * subject to CHECKSUM_UNNECESSARY.
4228 *
4229 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4230 * Description
4231 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4232 * Return
4233 * *sk* if casting is valid, or **NULL** otherwise.
4234 *
4235 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4236 * Description
4237 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4238 * Return
4239 * *sk* if casting is valid, or **NULL** otherwise.
4240 *
4241 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4242 * Description
4243 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4244 * Return
4245 * *sk* if casting is valid, or **NULL** otherwise.
4246 *
4247 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4248 * Description
4249 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4250 * Return
4251 * *sk* if casting is valid, or **NULL** otherwise.
4252 *
4253 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4254 * Description
4255 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4256 * Return
4257 * *sk* if casting is valid, or **NULL** otherwise.
4258 *
4259 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4260 * Description
4261 * Return a user or a kernel stack in bpf program provided buffer.
4262 * Note: the user stack will only be populated if the *task* is
4263 * the current task; all other tasks will return -EOPNOTSUPP.
4264 * To achieve this, the helper needs *task*, which is a valid
4265 * pointer to **struct task_struct**. To store the stacktrace, the
4266 * bpf program provides *buf* with a nonnegative *size*.
4267 *
4268 * The last argument, *flags*, holds the number of stack frames to
4269 * skip (from 0 to 255), masked with
4270 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4271 * the following flags:
4272 *
4273 * **BPF_F_USER_STACK**
4274 * Collect a user space stack instead of a kernel stack.
4275 * The *task* must be the current task.
4276 * **BPF_F_USER_BUILD_ID**
4277 * Collect buildid+offset instead of ips for user stack,
4278 * only valid if **BPF_F_USER_STACK** is also specified.
4279 *
4280 * **bpf_get_task_stack**\ () can collect up to
4281 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4282 * to sufficient large buffer size. Note that
4283 * this limit can be controlled with the **sysctl** program, and
4284 * that it should be manually increased in order to profile long
4285 * user stacks (such as stacks for Java programs). To do so, use:
4286 *
4287 * ::
4288 *
4289 * # sysctl kernel.perf_event_max_stack=<new value>
4290 * Return
4291 * The non-negative copied *buf* length equal to or less than
4292 * *size* on success, or a negative error in case of failure.
4293 *
4294 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4295 * Description
4296 * Load header option. Support reading a particular TCP header
4297 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4298 *
4299 * If *flags* is 0, it will search the option from the
4300 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4301 * has details on what skb_data contains under different
4302 * *skops*\ **->op**.
4303 *
4304 * The first byte of the *searchby_res* specifies the
4305 * kind that it wants to search.
4306 *
4307 * If the searching kind is an experimental kind
4308 * (i.e. 253 or 254 according to RFC6994). It also
4309 * needs to specify the "magic" which is either
4310 * 2 bytes or 4 bytes. It then also needs to
4311 * specify the size of the magic by using
4312 * the 2nd byte which is "kind-length" of a TCP
4313 * header option and the "kind-length" also
4314 * includes the first 2 bytes "kind" and "kind-length"
4315 * itself as a normal TCP header option also does.
4316 *
4317 * For example, to search experimental kind 254 with
4318 * 2 byte magic 0xeB9F, the searchby_res should be
4319 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4320 *
4321 * To search for the standard window scale option (3),
4322 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4323 * Note, kind-length must be 0 for regular option.
4324 *
4325 * Searching for No-Op (0) and End-of-Option-List (1) are
4326 * not supported.
4327 *
4328 * *len* must be at least 2 bytes which is the minimal size
4329 * of a header option.
4330 *
4331 * Supported flags:
4332 *
4333 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4334 * saved_syn packet or the just-received syn packet.
4335 *
4336 * Return
4337 * > 0 when found, the header option is copied to *searchby_res*.
4338 * The return value is the total length copied. On failure, a
4339 * negative error code is returned:
4340 *
4341 * **-EINVAL** if a parameter is invalid.
4342 *
4343 * **-ENOMSG** if the option is not found.
4344 *
4345 * **-ENOENT** if no syn packet is available when
4346 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4347 *
4348 * **-ENOSPC** if there is not enough space. Only *len* number of
4349 * bytes are copied.
4350 *
4351 * **-EFAULT** on failure to parse the header options in the
4352 * packet.
4353 *
4354 * **-EPERM** if the helper cannot be used under the current
4355 * *skops*\ **->op**.
4356 *
4357 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4358 * Description
4359 * Store header option. The data will be copied
4360 * from buffer *from* with length *len* to the TCP header.
4361 *
4362 * The buffer *from* should have the whole option that
4363 * includes the kind, kind-length, and the actual
4364 * option data. The *len* must be at least kind-length
4365 * long. The kind-length does not have to be 4 byte
4366 * aligned. The kernel will take care of the padding
4367 * and setting the 4 bytes aligned value to th->doff.
4368 *
4369 * This helper will check for duplicated option
4370 * by searching the same option in the outgoing skb.
4371 *
4372 * This helper can only be called during
4373 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4374 *
4375 * Return
4376 * 0 on success, or negative error in case of failure:
4377 *
4378 * **-EINVAL** If param is invalid.
4379 *
4380 * **-ENOSPC** if there is not enough space in the header.
4381 * Nothing has been written
4382 *
4383 * **-EEXIST** if the option already exists.
4384 *
4385 * **-EFAULT** on failrue to parse the existing header options.
4386 *
4387 * **-EPERM** if the helper cannot be used under the current
4388 * *skops*\ **->op**.
4389 *
4390 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4391 * Description
4392 * Reserve *len* bytes for the bpf header option. The
4393 * space will be used by **bpf_store_hdr_opt**\ () later in
4394 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4395 *
4396 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4397 * the total number of bytes will be reserved.
4398 *
4399 * This helper can only be called during
4400 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4401 *
4402 * Return
4403 * 0 on success, or negative error in case of failure:
4404 *
4405 * **-EINVAL** if a parameter is invalid.
4406 *
4407 * **-ENOSPC** if there is not enough space in the header.
4408 *
4409 * **-EPERM** if the helper cannot be used under the current
4410 * *skops*\ **->op**.
4411 *
4412 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4413 * Description
4414 * Get a bpf_local_storage from an *inode*.
4415 *
4416 * Logically, it could be thought of as getting the value from
4417 * a *map* with *inode* as the **key**. From this
4418 * perspective, the usage is not much different from
4419 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4420 * helper enforces the key must be an inode and the map must also
4421 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4422 *
4423 * Underneath, the value is stored locally at *inode* instead of
4424 * the *map*. The *map* is used as the bpf-local-storage
4425 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4426 * searched against all bpf_local_storage residing at *inode*.
4427 *
4428 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4429 * used such that a new bpf_local_storage will be
4430 * created if one does not exist. *value* can be used
4431 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4432 * the initial value of a bpf_local_storage. If *value* is
4433 * **NULL**, the new bpf_local_storage will be zero initialized.
4434 * Return
4435 * A bpf_local_storage pointer is returned on success.
4436 *
4437 * **NULL** if not found or there was an error in adding
4438 * a new bpf_local_storage.
4439 *
4440 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4441 * Description
4442 * Delete a bpf_local_storage from an *inode*.
4443 * Return
4444 * 0 on success.
4445 *
4446 * **-ENOENT** if the bpf_local_storage cannot be found.
4447 *
4448 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4449 * Description
4450 * Return full path for given **struct path** object, which
4451 * needs to be the kernel BTF *path* object. The path is
4452 * returned in the provided buffer *buf* of size *sz* and
4453 * is zero terminated.
4454 *
4455 * Return
4456 * On success, the strictly positive length of the string,
4457 * including the trailing NUL character. On error, a negative
4458 * value.
4459 *
4460 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4461 * Description
4462 * Read *size* bytes from user space address *user_ptr* and store
4463 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4464 * Return
4465 * 0 on success, or a negative error in case of failure.
4466 *
4467 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4468 * Description
4469 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4470 * using *ptr*->type_id. This value should specify the type
4471 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4472 * can be used to look up vmlinux BTF type ids. Traversing the
4473 * data structure using BTF, the type information and values are
4474 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4475 * the pointer data is carried out to avoid kernel crashes during
4476 * operation. Smaller types can use string space on the stack;
4477 * larger programs can use map data to store the string
4478 * representation.
4479 *
4480 * The string can be subsequently shared with userspace via
4481 * bpf_perf_event_output() or ring buffer interfaces.
4482 * bpf_trace_printk() is to be avoided as it places too small
4483 * a limit on string size to be useful.
4484 *
4485 * *flags* is a combination of
4486 *
4487 * **BTF_F_COMPACT**
4488 * no formatting around type information
4489 * **BTF_F_NONAME**
4490 * no struct/union member names/types
4491 * **BTF_F_PTR_RAW**
4492 * show raw (unobfuscated) pointer values;
4493 * equivalent to printk specifier %px.
4494 * **BTF_F_ZERO**
4495 * show zero-valued struct/union members; they
4496 * are not displayed by default
4497 *
4498 * Return
4499 * The number of bytes that were written (or would have been
4500 * written if output had to be truncated due to string size),
4501 * or a negative error in cases of failure.
4502 *
4503 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4504 * Description
4505 * Use BTF to write to seq_write a string representation of
4506 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4507 * *flags* are identical to those used for bpf_snprintf_btf.
4508 * Return
4509 * 0 on success or a negative error in case of failure.
4510 *
4511 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4512 * Description
4513 * See **bpf_get_cgroup_classid**\ () for the main description.
4514 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4515 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4516 * associated socket instead of the current process.
4517 * Return
4518 * The id is returned or 0 in case the id could not be retrieved.
4519 *
4520 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4521 * Description
4522 * Redirect the packet to another net device of index *ifindex*
4523 * and fill in L2 addresses from neighboring subsystem. This helper
4524 * is somewhat similar to **bpf_redirect**\ (), except that it
4525 * populates L2 addresses as well, meaning, internally, the helper
4526 * relies on the neighbor lookup for the L2 address of the nexthop.
4527 *
4528 * The helper will perform a FIB lookup based on the skb's
4529 * networking header to get the address of the next hop, unless
4530 * this is supplied by the caller in the *params* argument. The
4531 * *plen* argument indicates the len of *params* and should be set
4532 * to 0 if *params* is NULL.
4533 *
4534 * The *flags* argument is reserved and must be 0. The helper is
4535 * currently only supported for tc BPF program types, and enabled
4536 * for IPv4 and IPv6 protocols.
4537 * Return
4538 * The helper returns **TC_ACT_REDIRECT** on success or
4539 * **TC_ACT_SHOT** on error.
4540 *
4541 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4542 * Description
4543 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4544 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4545 * extern variable decorated with '__ksym'. For ksym, there is a
4546 * global var (either static or global) defined of the same name
4547 * in the kernel. The ksym is percpu if the global var is percpu.
4548 * The returned pointer points to the global percpu var on *cpu*.
4549 *
4550 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4551 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4552 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4553 * bpf_per_cpu_ptr() must check the returned value.
4554 * Return
4555 * A pointer pointing to the kernel percpu variable on *cpu*, or
4556 * NULL, if *cpu* is invalid.
4557 *
4558 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4559 * Description
4560 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4561 * pointer to the percpu kernel variable on this cpu. See the
4562 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4563 *
4564 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4565 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4566 * never return NULL.
4567 * Return
4568 * A pointer pointing to the kernel percpu variable on this cpu.
4569 *
4570 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4571 * Description
4572 * Redirect the packet to another net device of index *ifindex*.
4573 * This helper is somewhat similar to **bpf_redirect**\ (), except
4574 * that the redirection happens to the *ifindex*' peer device and
4575 * the netns switch takes place from ingress to ingress without
4576 * going through the CPU's backlog queue.
4577 *
4578 * The *flags* argument is reserved and must be 0. The helper is
4579 * currently only supported for tc BPF program types at the ingress
4580 * hook and for veth device types. The peer device must reside in a
4581 * different network namespace.
4582 * Return
4583 * The helper returns **TC_ACT_REDIRECT** on success or
4584 * **TC_ACT_SHOT** on error.
4585 *
4586 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4587 * Description
4588 * Get a bpf_local_storage from the *task*.
4589 *
4590 * Logically, it could be thought of as getting the value from
4591 * a *map* with *task* as the **key**. From this
4592 * perspective, the usage is not much different from
4593 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4594 * helper enforces the key must be an task_struct and the map must also
4595 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4596 *
4597 * Underneath, the value is stored locally at *task* instead of
4598 * the *map*. The *map* is used as the bpf-local-storage
4599 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4600 * searched against all bpf_local_storage residing at *task*.
4601 *
4602 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4603 * used such that a new bpf_local_storage will be
4604 * created if one does not exist. *value* can be used
4605 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4606 * the initial value of a bpf_local_storage. If *value* is
4607 * **NULL**, the new bpf_local_storage will be zero initialized.
4608 * Return
4609 * A bpf_local_storage pointer is returned on success.
4610 *
4611 * **NULL** if not found or there was an error in adding
4612 * a new bpf_local_storage.
4613 *
4614 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4615 * Description
4616 * Delete a bpf_local_storage from a *task*.
4617 * Return
4618 * 0 on success.
4619 *
4620 * **-ENOENT** if the bpf_local_storage cannot be found.
4621 *
4622 * struct task_struct *bpf_get_current_task_btf(void)
4623 * Description
4624 * Return a BTF pointer to the "current" task.
4625 * This pointer can also be used in helpers that accept an
4626 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
4627 * Return
4628 * Pointer to the current task.
4629 *
4630 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4631 * Description
4632 * Set or clear certain options on *bprm*:
4633 *
4634 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4635 * which sets the **AT_SECURE** auxv for glibc. The bit
4636 * is cleared if the flag is not specified.
4637 * Return
4638 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
4639 *
4640 * u64 bpf_ktime_get_coarse_ns(void)
4641 * Description
4642 * Return a coarse-grained version of the time elapsed since
4643 * system boot, in nanoseconds. Does not include time the system
4644 * was suspended.
4645 *
4646 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4647 * Return
4648 * Current *ktime*.
4649 *
4650 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4651 * Description
4652 * Returns the stored IMA hash of the *inode* (if it's avaialable).
4653 * If the hash is larger than *size*, then only *size*
4654 * bytes will be copied to *dst*
4655 * Return
4656 * The **hash_algo** is returned on success,
4657 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4658 * invalid arguments are passed.
4659 *
4660 * struct socket *bpf_sock_from_file(struct file *file)
4661 * Description
4662 * If the given file represents a socket, returns the associated
4663 * socket.
4664 * Return
4665 * A pointer to a struct socket on success or NULL if the file is
4666 * not a socket.
4667 *
4668 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4669 * Description
4670 * Check packet size against exceeding MTU of net device (based
4671 * on *ifindex*). This helper will likely be used in combination
4672 * with helpers that adjust/change the packet size.
4673 *
4674 * The argument *len_diff* can be used for querying with a planned
4675 * size change. This allows to check MTU prior to changing packet
4676 * ctx. Providing an *len_diff* adjustment that is larger than the
4677 * actual packet size (resulting in negative packet size) will in
4678 * principle not exceed the MTU, why it is not considered a
4679 * failure. Other BPF-helpers are needed for performing the
4680 * planned size change, why the responsability for catch a negative
4681 * packet size belong in those helpers.
4682 *
4683 * Specifying *ifindex* zero means the MTU check is performed
4684 * against the current net device. This is practical if this isn't
4685 * used prior to redirect.
4686 *
4687 * On input *mtu_len* must be a valid pointer, else verifier will
4688 * reject BPF program. If the value *mtu_len* is initialized to
4689 * zero then the ctx packet size is use. When value *mtu_len* is
4690 * provided as input this specify the L3 length that the MTU check
4691 * is done against. Remember XDP and TC length operate at L2, but
4692 * this value is L3 as this correlate to MTU and IP-header tot_len
4693 * values which are L3 (similar behavior as bpf_fib_lookup).
4694 *
4695 * The Linux kernel route table can configure MTUs on a more
4696 * specific per route level, which is not provided by this helper.
4697 * For route level MTU checks use the **bpf_fib_lookup**\ ()
4698 * helper.
4699 *
4700 * *ctx* is either **struct xdp_md** for XDP programs or
4701 * **struct sk_buff** for tc cls_act programs.
4702 *
4703 * The *flags* argument can be a combination of one or more of the
4704 * following values:
4705 *
4706 * **BPF_MTU_CHK_SEGS**
4707 * This flag will only works for *ctx* **struct sk_buff**.
4708 * If packet context contains extra packet segment buffers
4709 * (often knows as GSO skb), then MTU check is harder to
4710 * check at this point, because in transmit path it is
4711 * possible for the skb packet to get re-segmented
4712 * (depending on net device features). This could still be
4713 * a MTU violation, so this flag enables performing MTU
4714 * check against segments, with a different violation
4715 * return code to tell it apart. Check cannot use len_diff.
4716 *
4717 * On return *mtu_len* pointer contains the MTU value of the net
4718 * device. Remember the net device configured MTU is the L3 size,
4719 * which is returned here and XDP and TC length operate at L2.
4720 * Helper take this into account for you, but remember when using
4721 * MTU value in your BPF-code.
4722 *
4723 * Return
4724 * * 0 on success, and populate MTU value in *mtu_len* pointer.
4725 *
4726 * * < 0 if any input argument is invalid (*mtu_len* not updated)
4727 *
4728 * MTU violations return positive values, but also populate MTU
4729 * value in *mtu_len* pointer, as this can be needed for
4730 * implementing PMTU handing:
4731 *
4732 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
4733 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4734 *
4735 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4736 * Description
4737 * For each element in **map**, call **callback_fn** function with
4738 * **map**, **callback_ctx** and other map-specific parameters.
4739 * The **callback_fn** should be a static function and
4740 * the **callback_ctx** should be a pointer to the stack.
4741 * The **flags** is used to control certain aspects of the helper.
4742 * Currently, the **flags** must be 0.
4743 *
4744 * The following are a list of supported map types and their
4745 * respective expected callback signatures:
4746 *
4747 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4748 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4749 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4750 *
4751 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4752 *
4753 * For per_cpu maps, the map_value is the value on the cpu where the
4754 * bpf_prog is running.
4755 *
4756 * If **callback_fn** return 0, the helper will continue to the next
4757 * element. If return value is 1, the helper will skip the rest of
4758 * elements and return. Other return values are not used now.
4759 *
4760 * Return
4761 * The number of traversed map elements for success, **-EINVAL** for
4762 * invalid **flags**.
4763 *
4764 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4765 * Description
4766 * Outputs a string into the **str** buffer of size **str_size**
4767 * based on a format string stored in a read-only map pointed by
4768 * **fmt**.
4769 *
4770 * Each format specifier in **fmt** corresponds to one u64 element
4771 * in the **data** array. For strings and pointers where pointees
4772 * are accessed, only the pointer values are stored in the *data*
4773 * array. The *data_len* is the size of *data* in bytes.
4774 *
4775 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4776 * memory. Reading kernel memory may fail due to either invalid
4777 * address or valid address but requiring a major memory fault. If
4778 * reading kernel memory fails, the string for **%s** will be an
4779 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4780 * Not returning error to bpf program is consistent with what
4781 * **bpf_trace_printk**\ () does for now.
4782 *
4783 * Return
4784 * The strictly positive length of the formatted string, including
4785 * the trailing zero character. If the return value is greater than
4786 * **str_size**, **str** contains a truncated string, guaranteed to
4787 * be zero-terminated except when **str_size** is 0.
4788 *
4789 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4790 *
4791 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
4792 * Description
4793 * Execute bpf syscall with given arguments.
4794 * Return
4795 * A syscall result.
4796 *
4797 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
4798 * Description
4799 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
4800 * Return
4801 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
4802 *
4803 * long bpf_sys_close(u32 fd)
4804 * Description
4805 * Execute close syscall for given FD.
4806 * Return
4807 * A syscall result.
4808 *
4809 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
4810 * Description
4811 * Initialize the timer.
4812 * First 4 bits of *flags* specify clockid.
4813 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
4814 * All other bits of *flags* are reserved.
4815 * The verifier will reject the program if *timer* is not from
4816 * the same *map*.
4817 * Return
4818 * 0 on success.
4819 * **-EBUSY** if *timer* is already initialized.
4820 * **-EINVAL** if invalid *flags* are passed.
4821 * **-EPERM** if *timer* is in a map that doesn't have any user references.
4822 * The user space should either hold a file descriptor to a map with timers
4823 * or pin such map in bpffs. When map is unpinned or file descriptor is
4824 * closed all timers in the map will be cancelled and freed.
4825 *
4826 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
4827 * Description
4828 * Configure the timer to call *callback_fn* static function.
4829 * Return
4830 * 0 on success.
4831 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4832 * **-EPERM** if *timer* is in a map that doesn't have any user references.
4833 * The user space should either hold a file descriptor to a map with timers
4834 * or pin such map in bpffs. When map is unpinned or file descriptor is
4835 * closed all timers in the map will be cancelled and freed.
4836 *
4837 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
4838 * Description
4839 * Set timer expiration N nanoseconds from the current time. The
4840 * configured callback will be invoked in soft irq context on some cpu
4841 * and will not repeat unless another bpf_timer_start() is made.
4842 * In such case the next invocation can migrate to a different cpu.
4843 * Since struct bpf_timer is a field inside map element the map
4844 * owns the timer. The bpf_timer_set_callback() will increment refcnt
4845 * of BPF program to make sure that callback_fn code stays valid.
4846 * When user space reference to a map reaches zero all timers
4847 * in a map are cancelled and corresponding program's refcnts are
4848 * decremented. This is done to make sure that Ctrl-C of a user
4849 * process doesn't leave any timers running. If map is pinned in
4850 * bpffs the callback_fn can re-arm itself indefinitely.
4851 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
4852 * cancel and free the timer in the given map element.
4853 * The map can contain timers that invoke callback_fn-s from different
4854 * programs. The same callback_fn can serve different timers from
4855 * different maps if key/value layout matches across maps.
4856 * Every bpf_timer_set_callback() can have different callback_fn.
4857 *
4858 * Return
4859 * 0 on success.
4860 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
4861 * or invalid *flags* are passed.
4862 *
4863 * long bpf_timer_cancel(struct bpf_timer *timer)
4864 * Description
4865 * Cancel the timer and wait for callback_fn to finish if it was running.
4866 * Return
4867 * 0 if the timer was not active.
4868 * 1 if the timer was active.
4869 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4870 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
4871 * own timer which would have led to a deadlock otherwise.
4872 *
4873 * u64 bpf_get_func_ip(void *ctx)
4874 * Description
4875 * Get address of the traced function (for tracing and kprobe programs).
4876 * Return
4877 * Address of the traced function.
4878 *
4879 * u64 bpf_get_attach_cookie(void *ctx)
4880 * Description
4881 * Get bpf_cookie value provided (optionally) during the program
4882 * attachment. It might be different for each individual
4883 * attachment, even if BPF program itself is the same.
4884 * Expects BPF program context *ctx* as a first argument.
4885 *
4886 * Supported for the following program types:
4887 * - kprobe/uprobe;
4888 * - tracepoint;
4889 * - perf_event.
4890 * Return
4891 * Value specified by user at BPF link creation/attachment time
4892 * or 0, if it was not specified.
4893 *
4894 * long bpf_task_pt_regs(struct task_struct *task)
4895 * Description
4896 * Get the struct pt_regs associated with **task**.
4897 * Return
4898 * A pointer to struct pt_regs.
4899 */
4900#define __BPF_FUNC_MAPPER(FN) \
4901 FN(unspec), \
4902 FN(map_lookup_elem), \
4903 FN(map_update_elem), \
4904 FN(map_delete_elem), \
4905 FN(probe_read), \
4906 FN(ktime_get_ns), \
4907 FN(trace_printk), \
4908 FN(get_prandom_u32), \
4909 FN(get_smp_processor_id), \
4910 FN(skb_store_bytes), \
4911 FN(l3_csum_replace), \
4912 FN(l4_csum_replace), \
4913 FN(tail_call), \
4914 FN(clone_redirect), \
4915 FN(get_current_pid_tgid), \
4916 FN(get_current_uid_gid), \
4917 FN(get_current_comm), \
4918 FN(get_cgroup_classid), \
4919 FN(skb_vlan_push), \
4920 FN(skb_vlan_pop), \
4921 FN(skb_get_tunnel_key), \
4922 FN(skb_set_tunnel_key), \
4923 FN(perf_event_read), \
4924 FN(redirect), \
4925 FN(get_route_realm), \
4926 FN(perf_event_output), \
4927 FN(skb_load_bytes), \
4928 FN(get_stackid), \
4929 FN(csum_diff), \
4930 FN(skb_get_tunnel_opt), \
4931 FN(skb_set_tunnel_opt), \
4932 FN(skb_change_proto), \
4933 FN(skb_change_type), \
4934 FN(skb_under_cgroup), \
4935 FN(get_hash_recalc), \
4936 FN(get_current_task), \
4937 FN(probe_write_user), \
4938 FN(current_task_under_cgroup), \
4939 FN(skb_change_tail), \
4940 FN(skb_pull_data), \
4941 FN(csum_update), \
4942 FN(set_hash_invalid), \
4943 FN(get_numa_node_id), \
4944 FN(skb_change_head), \
4945 FN(xdp_adjust_head), \
4946 FN(probe_read_str), \
4947 FN(get_socket_cookie), \
4948 FN(get_socket_uid), \
4949 FN(set_hash), \
4950 FN(setsockopt), \
4951 FN(skb_adjust_room), \
4952 FN(redirect_map), \
4953 FN(sk_redirect_map), \
4954 FN(sock_map_update), \
4955 FN(xdp_adjust_meta), \
4956 FN(perf_event_read_value), \
4957 FN(perf_prog_read_value), \
4958 FN(getsockopt), \
4959 FN(override_return), \
4960 FN(sock_ops_cb_flags_set), \
4961 FN(msg_redirect_map), \
4962 FN(msg_apply_bytes), \
4963 FN(msg_cork_bytes), \
4964 FN(msg_pull_data), \
4965 FN(bind), \
4966 FN(xdp_adjust_tail), \
4967 FN(skb_get_xfrm_state), \
4968 FN(get_stack), \
4969 FN(skb_load_bytes_relative), \
4970 FN(fib_lookup), \
4971 FN(sock_hash_update), \
4972 FN(msg_redirect_hash), \
4973 FN(sk_redirect_hash), \
4974 FN(lwt_push_encap), \
4975 FN(lwt_seg6_store_bytes), \
4976 FN(lwt_seg6_adjust_srh), \
4977 FN(lwt_seg6_action), \
4978 FN(rc_repeat), \
4979 FN(rc_keydown), \
4980 FN(skb_cgroup_id), \
4981 FN(get_current_cgroup_id), \
4982 FN(get_local_storage), \
4983 FN(sk_select_reuseport), \
4984 FN(skb_ancestor_cgroup_id), \
4985 FN(sk_lookup_tcp), \
4986 FN(sk_lookup_udp), \
4987 FN(sk_release), \
4988 FN(map_push_elem), \
4989 FN(map_pop_elem), \
4990 FN(map_peek_elem), \
4991 FN(msg_push_data), \
4992 FN(msg_pop_data), \
4993 FN(rc_pointer_rel), \
4994 FN(spin_lock), \
4995 FN(spin_unlock), \
4996 FN(sk_fullsock), \
4997 FN(tcp_sock), \
4998 FN(skb_ecn_set_ce), \
4999 FN(get_listener_sock), \
5000 FN(skc_lookup_tcp), \
5001 FN(tcp_check_syncookie), \
5002 FN(sysctl_get_name), \
5003 FN(sysctl_get_current_value), \
5004 FN(sysctl_get_new_value), \
5005 FN(sysctl_set_new_value), \
5006 FN(strtol), \
5007 FN(strtoul), \
5008 FN(sk_storage_get), \
5009 FN(sk_storage_delete), \
5010 FN(send_signal), \
5011 FN(tcp_gen_syncookie), \
5012 FN(skb_output), \
5013 FN(probe_read_user), \
5014 FN(probe_read_kernel), \
5015 FN(probe_read_user_str), \
5016 FN(probe_read_kernel_str), \
5017 FN(tcp_send_ack), \
5018 FN(send_signal_thread), \
5019 FN(jiffies64), \
5020 FN(read_branch_records), \
5021 FN(get_ns_current_pid_tgid), \
5022 FN(xdp_output), \
5023 FN(get_netns_cookie), \
5024 FN(get_current_ancestor_cgroup_id), \
5025 FN(sk_assign), \
5026 FN(ktime_get_boot_ns), \
5027 FN(seq_printf), \
5028 FN(seq_write), \
5029 FN(sk_cgroup_id), \
5030 FN(sk_ancestor_cgroup_id), \
5031 FN(ringbuf_output), \
5032 FN(ringbuf_reserve), \
5033 FN(ringbuf_submit), \
5034 FN(ringbuf_discard), \
5035 FN(ringbuf_query), \
5036 FN(csum_level), \
5037 FN(skc_to_tcp6_sock), \
5038 FN(skc_to_tcp_sock), \
5039 FN(skc_to_tcp_timewait_sock), \
5040 FN(skc_to_tcp_request_sock), \
5041 FN(skc_to_udp6_sock), \
5042 FN(get_task_stack), \
5043 FN(load_hdr_opt), \
5044 FN(store_hdr_opt), \
5045 FN(reserve_hdr_opt), \
5046 FN(inode_storage_get), \
5047 FN(inode_storage_delete), \
5048 FN(d_path), \
5049 FN(copy_from_user), \
5050 FN(snprintf_btf), \
5051 FN(seq_printf_btf), \
5052 FN(skb_cgroup_classid), \
5053 FN(redirect_neigh), \
5054 FN(per_cpu_ptr), \
5055 FN(this_cpu_ptr), \
5056 FN(redirect_peer), \
5057 FN(task_storage_get), \
5058 FN(task_storage_delete), \
5059 FN(get_current_task_btf), \
5060 FN(bprm_opts_set), \
5061 FN(ktime_get_coarse_ns), \
5062 FN(ima_inode_hash), \
5063 FN(sock_from_file), \
5064 FN(check_mtu), \
5065 FN(for_each_map_elem), \
5066 FN(snprintf), \
5067 FN(sys_bpf), \
5068 FN(btf_find_by_name_kind), \
5069 FN(sys_close), \
5070 FN(timer_init), \
5071 FN(timer_set_callback), \
5072 FN(timer_start), \
5073 FN(timer_cancel), \
5074 FN(get_func_ip), \
5075 FN(get_attach_cookie), \
5076 FN(task_pt_regs), \
5077 /* */
5078
5079/* integer value in 'imm' field of BPF_CALL instruction selects which helper
5080 * function eBPF program intends to call
5081 */
5082#define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
5083enum bpf_func_id {
5084 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5085 __BPF_FUNC_MAX_ID,
5086};
5087#undef __BPF_ENUM_FN
5088
5089/* All flags used by eBPF helper functions, placed here. */
5090
5091/* BPF_FUNC_skb_store_bytes flags. */
5092enum {
5093 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
5094 BPF_F_INVALIDATE_HASH = (1ULL << 1),
5095};
5096
5097/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5098 * First 4 bits are for passing the header field size.
5099 */
5100enum {
5101 BPF_F_HDR_FIELD_MASK = 0xfULL,
5102};
5103
5104/* BPF_FUNC_l4_csum_replace flags. */
5105enum {
5106 BPF_F_PSEUDO_HDR = (1ULL << 4),
5107 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
5108 BPF_F_MARK_ENFORCE = (1ULL << 6),
5109};
5110
5111/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5112enum {
5113 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
5114};
5115
5116/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5117enum {
5118 BPF_F_SKIP_FIELD_MASK = 0xffULL,
5119 BPF_F_USER_STACK = (1ULL << 8),
5120/* flags used by BPF_FUNC_get_stackid only. */
5121 BPF_F_FAST_STACK_CMP = (1ULL << 9),
5122 BPF_F_REUSE_STACKID = (1ULL << 10),
5123/* flags used by BPF_FUNC_get_stack only. */
5124 BPF_F_USER_BUILD_ID = (1ULL << 11),
5125};
5126
5127/* BPF_FUNC_skb_set_tunnel_key flags. */
5128enum {
5129 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
5130 BPF_F_DONT_FRAGMENT = (1ULL << 2),
5131 BPF_F_SEQ_NUMBER = (1ULL << 3),
5132};
5133
5134/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5135 * BPF_FUNC_perf_event_read_value flags.
5136 */
5137enum {
5138 BPF_F_INDEX_MASK = 0xffffffffULL,
5139 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
5140/* BPF_FUNC_perf_event_output for sk_buff input context. */
5141 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
5142};
5143
5144/* Current network namespace */
5145enum {
5146 BPF_F_CURRENT_NETNS = (-1L),
5147};
5148
5149/* BPF_FUNC_csum_level level values. */
5150enum {
5151 BPF_CSUM_LEVEL_QUERY,
5152 BPF_CSUM_LEVEL_INC,
5153 BPF_CSUM_LEVEL_DEC,
5154 BPF_CSUM_LEVEL_RESET,
5155};
5156
5157/* BPF_FUNC_skb_adjust_room flags. */
5158enum {
5159 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
5160 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
5161 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
5162 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
5163 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
5164 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
5165 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
5166};
5167
5168enum {
5169 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
5170 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
5171};
5172
5173#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
5174 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5175 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5176
5177/* BPF_FUNC_sysctl_get_name flags. */
5178enum {
5179 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
5180};
5181
5182/* BPF_FUNC_<kernel_obj>_storage_get flags */
5183enum {
5184 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
5185 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5186 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5187 */
5188 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
5189};
5190
5191/* BPF_FUNC_read_branch_records flags. */
5192enum {
5193 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
5194};
5195
5196/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5197 * BPF_FUNC_bpf_ringbuf_output flags.
5198 */
5199enum {
5200 BPF_RB_NO_WAKEUP = (1ULL << 0),
5201 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
5202};
5203
5204/* BPF_FUNC_bpf_ringbuf_query flags */
5205enum {
5206 BPF_RB_AVAIL_DATA = 0,
5207 BPF_RB_RING_SIZE = 1,
5208 BPF_RB_CONS_POS = 2,
5209 BPF_RB_PROD_POS = 3,
5210};
5211
5212/* BPF ring buffer constants */
5213enum {
5214 BPF_RINGBUF_BUSY_BIT = (1U << 31),
5215 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
5216 BPF_RINGBUF_HDR_SZ = 8,
5217};
5218
5219/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5220enum {
5221 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
5222 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
5223};
5224
5225/* Mode for BPF_FUNC_skb_adjust_room helper. */
5226enum bpf_adj_room_mode {
5227 BPF_ADJ_ROOM_NET,
5228 BPF_ADJ_ROOM_MAC,
5229};
5230
5231/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5232enum bpf_hdr_start_off {
5233 BPF_HDR_START_MAC,
5234 BPF_HDR_START_NET,
5235};
5236
5237/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5238enum bpf_lwt_encap_mode {
5239 BPF_LWT_ENCAP_SEG6,
5240 BPF_LWT_ENCAP_SEG6_INLINE,
5241 BPF_LWT_ENCAP_IP,
5242};
5243
5244/* Flags for bpf_bprm_opts_set helper */
5245enum {
5246 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
5247};
5248
5249/* Flags for bpf_redirect and bpf_redirect_map helpers */
5250enum {
5251 BPF_F_INGRESS = (1ULL << 0), /* used for skb path */
5252 BPF_F_BROADCAST = (1ULL << 3), /* used for XDP path */
5253 BPF_F_EXCLUDE_INGRESS = (1ULL << 4), /* used for XDP path */
5254#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
5255};
5256
5257#define __bpf_md_ptr(type, name) \
5258union { \
5259 type name; \
5260 __u64 :64; \
5261} __attribute__((aligned(8)))
5262
5263/* user accessible mirror of in-kernel sk_buff.
5264 * new fields can only be added to the end of this structure
5265 */
5266struct __sk_buff {
5267 __u32 len;
5268 __u32 pkt_type;
5269 __u32 mark;
5270 __u32 queue_mapping;
5271 __u32 protocol;
5272 __u32 vlan_present;
5273 __u32 vlan_tci;
5274 __u32 vlan_proto;
5275 __u32 priority;
5276 __u32 ingress_ifindex;
5277 __u32 ifindex;
5278 __u32 tc_index;
5279 __u32 cb[5];
5280 __u32 hash;
5281 __u32 tc_classid;
5282 __u32 data;
5283 __u32 data_end;
5284 __u32 napi_id;
5285
5286 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5287 __u32 family;
5288 __u32 remote_ip4; /* Stored in network byte order */
5289 __u32 local_ip4; /* Stored in network byte order */
5290 __u32 remote_ip6[4]; /* Stored in network byte order */
5291 __u32 local_ip6[4]; /* Stored in network byte order */
5292 __u32 remote_port; /* Stored in network byte order */
5293 __u32 local_port; /* stored in host byte order */
5294 /* ... here. */
5295
5296 __u32 data_meta;
5297 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5298 __u64 tstamp;
5299 __u32 wire_len;
5300 __u32 gso_segs;
5301 __bpf_md_ptr(struct bpf_sock *, sk);
5302 __u32 gso_size;
5303};
5304
5305struct bpf_tunnel_key {
5306 __u32 tunnel_id;
5307 union {
5308 __u32 remote_ipv4;
5309 __u32 remote_ipv6[4];
5310 };
5311 __u8 tunnel_tos;
5312 __u8 tunnel_ttl;
5313 __u16 tunnel_ext; /* Padding, future use. */
5314 __u32 tunnel_label;
5315};
5316
5317/* user accessible mirror of in-kernel xfrm_state.
5318 * new fields can only be added to the end of this structure
5319 */
5320struct bpf_xfrm_state {
5321 __u32 reqid;
5322 __u32 spi; /* Stored in network byte order */
5323 __u16 family;
5324 __u16 ext; /* Padding, future use. */
5325 union {
5326 __u32 remote_ipv4; /* Stored in network byte order */
5327 __u32 remote_ipv6[4]; /* Stored in network byte order */
5328 };
5329};
5330
5331/* Generic BPF return codes which all BPF program types may support.
5332 * The values are binary compatible with their TC_ACT_* counter-part to
5333 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5334 * programs.
5335 *
5336 * XDP is handled seprately, see XDP_*.
5337 */
5338enum bpf_ret_code {
5339 BPF_OK = 0,
5340 /* 1 reserved */
5341 BPF_DROP = 2,
5342 /* 3-6 reserved */
5343 BPF_REDIRECT = 7,
5344 /* >127 are reserved for prog type specific return codes.
5345 *
5346 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5347 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5348 * changed and should be routed based on its new L3 header.
5349 * (This is an L3 redirect, as opposed to L2 redirect
5350 * represented by BPF_REDIRECT above).
5351 */
5352 BPF_LWT_REROUTE = 128,
5353};
5354
5355struct bpf_sock {
5356 __u32 bound_dev_if;
5357 __u32 family;
5358 __u32 type;
5359 __u32 protocol;
5360 __u32 mark;
5361 __u32 priority;
5362 /* IP address also allows 1 and 2 bytes access */
5363 __u32 src_ip4;
5364 __u32 src_ip6[4];
5365 __u32 src_port; /* host byte order */
5366 __be16 dst_port; /* network byte order */
5367 __u16 :16; /* zero padding */
5368 __u32 dst_ip4;
5369 __u32 dst_ip6[4];
5370 __u32 state;
5371 __s32 rx_queue_mapping;
5372};
5373
5374struct bpf_tcp_sock {
5375 __u32 snd_cwnd; /* Sending congestion window */
5376 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
5377 __u32 rtt_min;
5378 __u32 snd_ssthresh; /* Slow start size threshold */
5379 __u32 rcv_nxt; /* What we want to receive next */
5380 __u32 snd_nxt; /* Next sequence we send */
5381 __u32 snd_una; /* First byte we want an ack for */
5382 __u32 mss_cache; /* Cached effective mss, not including SACKS */
5383 __u32 ecn_flags; /* ECN status bits. */
5384 __u32 rate_delivered; /* saved rate sample: packets delivered */
5385 __u32 rate_interval_us; /* saved rate sample: time elapsed */
5386 __u32 packets_out; /* Packets which are "in flight" */
5387 __u32 retrans_out; /* Retransmitted packets out */
5388 __u32 total_retrans; /* Total retransmits for entire connection */
5389 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
5390 * total number of segments in.
5391 */
5392 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
5393 * total number of data segments in.
5394 */
5395 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
5396 * The total number of segments sent.
5397 */
5398 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
5399 * total number of data segments sent.
5400 */
5401 __u32 lost_out; /* Lost packets */
5402 __u32 sacked_out; /* SACK'd packets */
5403 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
5404 * sum(delta(rcv_nxt)), or how many bytes
5405 * were acked.
5406 */
5407 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
5408 * sum(delta(snd_una)), or how many bytes
5409 * were acked.
5410 */
5411 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
5412 * total number of DSACK blocks received
5413 */
5414 __u32 delivered; /* Total data packets delivered incl. rexmits */
5415 __u32 delivered_ce; /* Like the above but only ECE marked packets */
5416 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
5417};
5418
5419struct bpf_sock_tuple {
5420 union {
5421 struct {
5422 __be32 saddr;
5423 __be32 daddr;
5424 __be16 sport;
5425 __be16 dport;
5426 } ipv4;
5427 struct {
5428 __be32 saddr[4];
5429 __be32 daddr[4];
5430 __be16 sport;
5431 __be16 dport;
5432 } ipv6;
5433 };
5434};
5435
5436struct bpf_xdp_sock {
5437 __u32 queue_id;
5438};
5439
5440#define XDP_PACKET_HEADROOM 256
5441
5442/* User return codes for XDP prog type.
5443 * A valid XDP program must return one of these defined values. All other
5444 * return codes are reserved for future use. Unknown return codes will
5445 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5446 */
5447enum xdp_action {
5448 XDP_ABORTED = 0,
5449 XDP_DROP,
5450 XDP_PASS,
5451 XDP_TX,
5452 XDP_REDIRECT,
5453};
5454
5455/* user accessible metadata for XDP packet hook
5456 * new fields must be added to the end of this structure
5457 */
5458struct xdp_md {
5459 __u32 data;
5460 __u32 data_end;
5461 __u32 data_meta;
5462 /* Below access go through struct xdp_rxq_info */
5463 __u32 ingress_ifindex; /* rxq->dev->ifindex */
5464 __u32 rx_queue_index; /* rxq->queue_index */
5465
5466 __u32 egress_ifindex; /* txq->dev->ifindex */
5467};
5468
5469/* DEVMAP map-value layout
5470 *
5471 * The struct data-layout of map-value is a configuration interface.
5472 * New members can only be added to the end of this structure.
5473 */
5474struct bpf_devmap_val {
5475 __u32 ifindex; /* device index */
5476 union {
5477 int fd; /* prog fd on map write */
5478 __u32 id; /* prog id on map read */
5479 } bpf_prog;
5480};
5481
5482/* CPUMAP map-value layout
5483 *
5484 * The struct data-layout of map-value is a configuration interface.
5485 * New members can only be added to the end of this structure.
5486 */
5487struct bpf_cpumap_val {
5488 __u32 qsize; /* queue size to remote target CPU */
5489 union {
5490 int fd; /* prog fd on map write */
5491 __u32 id; /* prog id on map read */
5492 } bpf_prog;
5493};
5494
5495enum sk_action {
5496 SK_DROP = 0,
5497 SK_PASS,
5498};
5499
5500/* user accessible metadata for SK_MSG packet hook, new fields must
5501 * be added to the end of this structure
5502 */
5503struct sk_msg_md {
5504 __bpf_md_ptr(void *, data);
5505 __bpf_md_ptr(void *, data_end);
5506
5507 __u32 family;
5508 __u32 remote_ip4; /* Stored in network byte order */
5509 __u32 local_ip4; /* Stored in network byte order */
5510 __u32 remote_ip6[4]; /* Stored in network byte order */
5511 __u32 local_ip6[4]; /* Stored in network byte order */
5512 __u32 remote_port; /* Stored in network byte order */
5513 __u32 local_port; /* stored in host byte order */
5514 __u32 size; /* Total size of sk_msg */
5515
5516 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
5517};
5518
5519struct sk_reuseport_md {
5520 /*
5521 * Start of directly accessible data. It begins from
5522 * the tcp/udp header.
5523 */
5524 __bpf_md_ptr(void *, data);
5525 /* End of directly accessible data */
5526 __bpf_md_ptr(void *, data_end);
5527 /*
5528 * Total length of packet (starting from the tcp/udp header).
5529 * Note that the directly accessible bytes (data_end - data)
5530 * could be less than this "len". Those bytes could be
5531 * indirectly read by a helper "bpf_skb_load_bytes()".
5532 */
5533 __u32 len;
5534 /*
5535 * Eth protocol in the mac header (network byte order). e.g.
5536 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
5537 */
5538 __u32 eth_protocol;
5539 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
5540 __u32 bind_inany; /* Is sock bound to an INANY address? */
5541 __u32 hash; /* A hash of the packet 4 tuples */
5542 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
5543 * new incoming connection request (e.g. selecting a listen sk for
5544 * the received SYN in the TCP case). reuse->sk is one of the sk
5545 * in the reuseport group. The bpf prog can use reuse->sk to learn
5546 * the local listening ip/port without looking into the skb.
5547 *
5548 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
5549 * reuse->migrating_sk is the socket that needs to be migrated
5550 * to another listening socket. migrating_sk could be a fullsock
5551 * sk that is fully established or a reqsk that is in-the-middle
5552 * of 3-way handshake.
5553 */
5554 __bpf_md_ptr(struct bpf_sock *, sk);
5555 __bpf_md_ptr(struct bpf_sock *, migrating_sk);
5556};
5557
5558#define BPF_TAG_SIZE 8
5559
5560struct bpf_prog_info {
5561 __u32 type;
5562 __u32 id;
5563 __u8 tag[BPF_TAG_SIZE];
5564 __u32 jited_prog_len;
5565 __u32 xlated_prog_len;
5566 __aligned_u64 jited_prog_insns;
5567 __aligned_u64 xlated_prog_insns;
5568 __u64 load_time; /* ns since boottime */
5569 __u32 created_by_uid;
5570 __u32 nr_map_ids;
5571 __aligned_u64 map_ids;
5572 char name[BPF_OBJ_NAME_LEN];
5573 __u32 ifindex;
5574 __u32 gpl_compatible:1;
5575 __u32 :31; /* alignment pad */
5576 __u64 netns_dev;
5577 __u64 netns_ino;
5578 __u32 nr_jited_ksyms;
5579 __u32 nr_jited_func_lens;
5580 __aligned_u64 jited_ksyms;
5581 __aligned_u64 jited_func_lens;
5582 __u32 btf_id;
5583 __u32 func_info_rec_size;
5584 __aligned_u64 func_info;
5585 __u32 nr_func_info;
5586 __u32 nr_line_info;
5587 __aligned_u64 line_info;
5588 __aligned_u64 jited_line_info;
5589 __u32 nr_jited_line_info;
5590 __u32 line_info_rec_size;
5591 __u32 jited_line_info_rec_size;
5592 __u32 nr_prog_tags;
5593 __aligned_u64 prog_tags;
5594 __u64 run_time_ns;
5595 __u64 run_cnt;
5596 __u64 recursion_misses;
5597} __attribute__((aligned(8)));
5598
5599struct bpf_map_info {
5600 __u32 type;
5601 __u32 id;
5602 __u32 key_size;
5603 __u32 value_size;
5604 __u32 max_entries;
5605 __u32 map_flags;
5606 char name[BPF_OBJ_NAME_LEN];
5607 __u32 ifindex;
5608 __u32 btf_vmlinux_value_type_id;
5609 __u64 netns_dev;
5610 __u64 netns_ino;
5611 __u32 btf_id;
5612 __u32 btf_key_type_id;
5613 __u32 btf_value_type_id;
5614} __attribute__((aligned(8)));
5615
5616struct bpf_btf_info {
5617 __aligned_u64 btf;
5618 __u32 btf_size;
5619 __u32 id;
5620 __aligned_u64 name;
5621 __u32 name_len;
5622 __u32 kernel_btf;
5623} __attribute__((aligned(8)));
5624
5625struct bpf_link_info {
5626 __u32 type;
5627 __u32 id;
5628 __u32 prog_id;
5629 union {
5630 struct {
5631 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
5632 __u32 tp_name_len; /* in/out: tp_name buffer len */
5633 } raw_tracepoint;
5634 struct {
5635 __u32 attach_type;
5636 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
5637 __u32 target_btf_id; /* BTF type id inside the object */
5638 } tracing;
5639 struct {
5640 __u64 cgroup_id;
5641 __u32 attach_type;
5642 } cgroup;
5643 struct {
5644 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
5645 __u32 target_name_len; /* in/out: target_name buffer len */
5646 union {
5647 struct {
5648 __u32 map_id;
5649 } map;
5650 };
5651 } iter;
5652 struct {
5653 __u32 netns_ino;
5654 __u32 attach_type;
5655 } netns;
5656 struct {
5657 __u32 ifindex;
5658 } xdp;
5659 };
5660} __attribute__((aligned(8)));
5661
5662/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
5663 * by user and intended to be used by socket (e.g. to bind to, depends on
5664 * attach type).
5665 */
5666struct bpf_sock_addr {
5667 __u32 user_family; /* Allows 4-byte read, but no write. */
5668 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5669 * Stored in network byte order.
5670 */
5671 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5672 * Stored in network byte order.
5673 */
5674 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
5675 * Stored in network byte order
5676 */
5677 __u32 family; /* Allows 4-byte read, but no write */
5678 __u32 type; /* Allows 4-byte read, but no write */
5679 __u32 protocol; /* Allows 4-byte read, but no write */
5680 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5681 * Stored in network byte order.
5682 */
5683 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5684 * Stored in network byte order.
5685 */
5686 __bpf_md_ptr(struct bpf_sock *, sk);
5687};
5688
5689/* User bpf_sock_ops struct to access socket values and specify request ops
5690 * and their replies.
5691 * Some of this fields are in network (bigendian) byte order and may need
5692 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
5693 * New fields can only be added at the end of this structure
5694 */
5695struct bpf_sock_ops {
5696 __u32 op;
5697 union {
5698 __u32 args[4]; /* Optionally passed to bpf program */
5699 __u32 reply; /* Returned by bpf program */
5700 __u32 replylong[4]; /* Optionally returned by bpf prog */
5701 };
5702 __u32 family;
5703 __u32 remote_ip4; /* Stored in network byte order */
5704 __u32 local_ip4; /* Stored in network byte order */
5705 __u32 remote_ip6[4]; /* Stored in network byte order */
5706 __u32 local_ip6[4]; /* Stored in network byte order */
5707 __u32 remote_port; /* Stored in network byte order */
5708 __u32 local_port; /* stored in host byte order */
5709 __u32 is_fullsock; /* Some TCP fields are only valid if
5710 * there is a full socket. If not, the
5711 * fields read as zero.
5712 */
5713 __u32 snd_cwnd;
5714 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
5715 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
5716 __u32 state;
5717 __u32 rtt_min;
5718 __u32 snd_ssthresh;
5719 __u32 rcv_nxt;
5720 __u32 snd_nxt;
5721 __u32 snd_una;
5722 __u32 mss_cache;
5723 __u32 ecn_flags;
5724 __u32 rate_delivered;
5725 __u32 rate_interval_us;
5726 __u32 packets_out;
5727 __u32 retrans_out;
5728 __u32 total_retrans;
5729 __u32 segs_in;
5730 __u32 data_segs_in;
5731 __u32 segs_out;
5732 __u32 data_segs_out;
5733 __u32 lost_out;
5734 __u32 sacked_out;
5735 __u32 sk_txhash;
5736 __u64 bytes_received;
5737 __u64 bytes_acked;
5738 __bpf_md_ptr(struct bpf_sock *, sk);
5739 /* [skb_data, skb_data_end) covers the whole TCP header.
5740 *
5741 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
5742 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
5743 * header has not been written.
5744 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
5745 * been written so far.
5746 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
5747 * the 3WHS.
5748 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
5749 * the 3WHS.
5750 *
5751 * bpf_load_hdr_opt() can also be used to read a particular option.
5752 */
5753 __bpf_md_ptr(void *, skb_data);
5754 __bpf_md_ptr(void *, skb_data_end);
5755 __u32 skb_len; /* The total length of a packet.
5756 * It includes the header, options,
5757 * and payload.
5758 */
5759 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
5760 * an easy way to check for tcp_flags
5761 * without parsing skb_data.
5762 *
5763 * In particular, the skb_tcp_flags
5764 * will still be available in
5765 * BPF_SOCK_OPS_HDR_OPT_LEN even though
5766 * the outgoing header has not
5767 * been written yet.
5768 */
5769};
5770
5771/* Definitions for bpf_sock_ops_cb_flags */
5772enum {
5773 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
5774 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
5775 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
5776 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
5777 /* Call bpf for all received TCP headers. The bpf prog will be
5778 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5779 *
5780 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5781 * for the header option related helpers that will be useful
5782 * to the bpf programs.
5783 *
5784 * It could be used at the client/active side (i.e. connect() side)
5785 * when the server told it that the server was in syncookie
5786 * mode and required the active side to resend the bpf-written
5787 * options. The active side can keep writing the bpf-options until
5788 * it received a valid packet from the server side to confirm
5789 * the earlier packet (and options) has been received. The later
5790 * example patch is using it like this at the active side when the
5791 * server is in syncookie mode.
5792 *
5793 * The bpf prog will usually turn this off in the common cases.
5794 */
5795 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
5796 /* Call bpf when kernel has received a header option that
5797 * the kernel cannot handle. The bpf prog will be called under
5798 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
5799 *
5800 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5801 * for the header option related helpers that will be useful
5802 * to the bpf programs.
5803 */
5804 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
5805 /* Call bpf when the kernel is writing header options for the
5806 * outgoing packet. The bpf prog will first be called
5807 * to reserve space in a skb under
5808 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
5809 * the bpf prog will be called to write the header option(s)
5810 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5811 *
5812 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
5813 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
5814 * related helpers that will be useful to the bpf programs.
5815 *
5816 * The kernel gets its chance to reserve space and write
5817 * options first before the BPF program does.
5818 */
5819 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
5820/* Mask of all currently supported cb flags */
5821 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
5822};
5823
5824/* List of known BPF sock_ops operators.
5825 * New entries can only be added at the end
5826 */
5827enum {
5828 BPF_SOCK_OPS_VOID,
5829 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
5830 * -1 if default value should be used
5831 */
5832 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
5833 * window (in packets) or -1 if default
5834 * value should be used
5835 */
5836 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
5837 * active connection is initialized
5838 */
5839 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
5840 * active connection is
5841 * established
5842 */
5843 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
5844 * passive connection is
5845 * established
5846 */
5847 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
5848 * needs ECN
5849 */
5850 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
5851 * based on the path and may be
5852 * dependent on the congestion control
5853 * algorithm. In general it indicates
5854 * a congestion threshold. RTTs above
5855 * this indicate congestion
5856 */
5857 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
5858 * Arg1: value of icsk_retransmits
5859 * Arg2: value of icsk_rto
5860 * Arg3: whether RTO has expired
5861 */
5862 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
5863 * Arg1: sequence number of 1st byte
5864 * Arg2: # segments
5865 * Arg3: return value of
5866 * tcp_transmit_skb (0 => success)
5867 */
5868 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
5869 * Arg1: old_state
5870 * Arg2: new_state
5871 */
5872 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
5873 * socket transition to LISTEN state.
5874 */
5875 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
5876 */
5877 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
5878 * It will be called to handle
5879 * the packets received at
5880 * an already established
5881 * connection.
5882 *
5883 * sock_ops->skb_data:
5884 * Referring to the received skb.
5885 * It covers the TCP header only.
5886 *
5887 * bpf_load_hdr_opt() can also
5888 * be used to search for a
5889 * particular option.
5890 */
5891 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
5892 * header option later in
5893 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5894 * Arg1: bool want_cookie. (in
5895 * writing SYNACK only)
5896 *
5897 * sock_ops->skb_data:
5898 * Not available because no header has
5899 * been written yet.
5900 *
5901 * sock_ops->skb_tcp_flags:
5902 * The tcp_flags of the
5903 * outgoing skb. (e.g. SYN, ACK, FIN).
5904 *
5905 * bpf_reserve_hdr_opt() should
5906 * be used to reserve space.
5907 */
5908 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
5909 * Arg1: bool want_cookie. (in
5910 * writing SYNACK only)
5911 *
5912 * sock_ops->skb_data:
5913 * Referring to the outgoing skb.
5914 * It covers the TCP header
5915 * that has already been written
5916 * by the kernel and the
5917 * earlier bpf-progs.
5918 *
5919 * sock_ops->skb_tcp_flags:
5920 * The tcp_flags of the outgoing
5921 * skb. (e.g. SYN, ACK, FIN).
5922 *
5923 * bpf_store_hdr_opt() should
5924 * be used to write the
5925 * option.
5926 *
5927 * bpf_load_hdr_opt() can also
5928 * be used to search for a
5929 * particular option that
5930 * has already been written
5931 * by the kernel or the
5932 * earlier bpf-progs.
5933 */
5934};
5935
5936/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
5937 * changes between the TCP and BPF versions. Ideally this should never happen.
5938 * If it does, we need to add code to convert them before calling
5939 * the BPF sock_ops function.
5940 */
5941enum {
5942 BPF_TCP_ESTABLISHED = 1,
5943 BPF_TCP_SYN_SENT,
5944 BPF_TCP_SYN_RECV,
5945 BPF_TCP_FIN_WAIT1,
5946 BPF_TCP_FIN_WAIT2,
5947 BPF_TCP_TIME_WAIT,
5948 BPF_TCP_CLOSE,
5949 BPF_TCP_CLOSE_WAIT,
5950 BPF_TCP_LAST_ACK,
5951 BPF_TCP_LISTEN,
5952 BPF_TCP_CLOSING, /* Now a valid state */
5953 BPF_TCP_NEW_SYN_RECV,
5954
5955 BPF_TCP_MAX_STATES /* Leave at the end! */
5956};
5957
5958enum {
5959 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
5960 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
5961 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
5962 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
5963 /* Copy the SYN pkt to optval
5964 *
5965 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
5966 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
5967 * to only getting from the saved_syn. It can either get the
5968 * syn packet from:
5969 *
5970 * 1. the just-received SYN packet (only available when writing the
5971 * SYNACK). It will be useful when it is not necessary to
5972 * save the SYN packet for latter use. It is also the only way
5973 * to get the SYN during syncookie mode because the syn
5974 * packet cannot be saved during syncookie.
5975 *
5976 * OR
5977 *
5978 * 2. the earlier saved syn which was done by
5979 * bpf_setsockopt(TCP_SAVE_SYN).
5980 *
5981 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
5982 * SYN packet is obtained.
5983 *
5984 * If the bpf-prog does not need the IP[46] header, the
5985 * bpf-prog can avoid parsing the IP header by using
5986 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
5987 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
5988 *
5989 * >0: Total number of bytes copied
5990 * -ENOSPC: Not enough space in optval. Only optlen number of
5991 * bytes is copied.
5992 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
5993 * is not saved by setsockopt(TCP_SAVE_SYN).
5994 */
5995 TCP_BPF_SYN = 1005, /* Copy the TCP header */
5996 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
5997 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
5998};
5999
6000enum {
6001 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6002};
6003
6004/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6005 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6006 */
6007enum {
6008 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
6009 * total option spaces
6010 * required for an established
6011 * sk in order to calculate the
6012 * MSS. No skb is actually
6013 * sent.
6014 */
6015 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
6016 * when sending a SYN.
6017 */
6018};
6019
6020struct bpf_perf_event_value {
6021 __u64 counter;
6022 __u64 enabled;
6023 __u64 running;
6024};
6025
6026enum {
6027 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
6028 BPF_DEVCG_ACC_READ = (1ULL << 1),
6029 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
6030};
6031
6032enum {
6033 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
6034 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
6035};
6036
6037struct bpf_cgroup_dev_ctx {
6038 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6039 __u32 access_type;
6040 __u32 major;
6041 __u32 minor;
6042};
6043
6044struct bpf_raw_tracepoint_args {
6045 __u64 args[0];
6046};
6047
6048/* DIRECT: Skip the FIB rules and go to FIB table associated with device
6049 * OUTPUT: Do lookup from egress perspective; default is ingress
6050 */
6051enum {
6052 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
6053 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
6054 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
6055 BPF_FIB_LOOKUP_TBID = (1U << 3),
6056 BPF_FIB_LOOKUP_SRC = (1U << 4),
6057};
6058
6059enum {
6060 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
6061 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
6062 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
6063 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
6064 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
6065 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6066 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
6067 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
6068 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
6069 BPF_FIB_LKUP_RET_NO_SRC_ADDR, /* failed to derive IP src addr */
6070};
6071
6072struct bpf_fib_lookup {
6073 /* input: network family for lookup (AF_INET, AF_INET6)
6074 * output: network family of egress nexthop
6075 */
6076 __u8 family;
6077
6078 /* set if lookup is to consider L4 data - e.g., FIB rules */
6079 __u8 l4_protocol;
6080 __be16 sport;
6081 __be16 dport;
6082
6083 union { /* used for MTU check */
6084 /* input to lookup */
6085 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
6086
6087 /* output: MTU value */
6088 __u16 mtu_result;
6089 } __attribute__((packed, aligned(2)));
6090 /* input: L3 device index for lookup
6091 * output: device index from FIB lookup
6092 */
6093 __u32 ifindex;
6094
6095 union {
6096 /* inputs to lookup */
6097 __u8 tos; /* AF_INET */
6098 __be32 flowinfo; /* AF_INET6, flow_label + priority */
6099
6100 /* output: metric of fib result (IPv4/IPv6 only) */
6101 __u32 rt_metric;
6102 };
6103
6104 /* input: source address to consider for lookup
6105 * output: source address result from lookup
6106 */
6107 union {
6108 __be32 ipv4_src;
6109 __u32 ipv6_src[4]; /* in6_addr; network order */
6110 };
6111
6112 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
6113 * network header. output: bpf_fib_lookup sets to gateway address
6114 * if FIB lookup returns gateway route
6115 */
6116 union {
6117 __be32 ipv4_dst;
6118 __u32 ipv6_dst[4]; /* in6_addr; network order */
6119 };
6120
6121 union {
6122 struct {
6123 /* output */
6124 __be16 h_vlan_proto;
6125 __be16 h_vlan_TCI;
6126 };
6127 /* input: when accompanied with the
6128 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
6129 * specific routing table to use for the fib lookup.
6130 */
6131 __u32 tbid;
6132 };
6133
6134 __u8 smac[6]; /* ETH_ALEN */
6135 __u8 dmac[6]; /* ETH_ALEN */
6136};
6137
6138struct bpf_redir_neigh {
6139 /* network family for lookup (AF_INET, AF_INET6) */
6140 __u32 nh_family;
6141 /* network address of nexthop; skips fib lookup to find gateway */
6142 union {
6143 __be32 ipv4_nh;
6144 __u32 ipv6_nh[4]; /* in6_addr; network order */
6145 };
6146};
6147
6148/* bpf_check_mtu flags*/
6149enum bpf_check_mtu_flags {
6150 BPF_MTU_CHK_SEGS = (1U << 0),
6151};
6152
6153enum bpf_check_mtu_ret {
6154 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
6155 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
6156 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
6157};
6158
6159enum bpf_task_fd_type {
6160 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
6161 BPF_FD_TYPE_TRACEPOINT, /* tp name */
6162 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
6163 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
6164 BPF_FD_TYPE_UPROBE, /* filename + offset */
6165 BPF_FD_TYPE_URETPROBE, /* filename + offset */
6166};
6167
6168enum {
6169 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
6170 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
6171 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
6172};
6173
6174struct bpf_flow_keys {
6175 __u16 nhoff;
6176 __u16 thoff;
6177 __u16 addr_proto; /* ETH_P_* of valid addrs */
6178 __u8 is_frag;
6179 __u8 is_first_frag;
6180 __u8 is_encap;
6181 __u8 ip_proto;
6182 __be16 n_proto;
6183 __be16 sport;
6184 __be16 dport;
6185 union {
6186 struct {
6187 __be32 ipv4_src;
6188 __be32 ipv4_dst;
6189 };
6190 struct {
6191 __u32 ipv6_src[4]; /* in6_addr; network order */
6192 __u32 ipv6_dst[4]; /* in6_addr; network order */
6193 };
6194 };
6195 __u32 flags;
6196 __be32 flow_label;
6197};
6198
6199struct bpf_func_info {
6200 __u32 insn_off;
6201 __u32 type_id;
6202};
6203
6204#define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
6205#define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
6206
6207struct bpf_line_info {
6208 __u32 insn_off;
6209 __u32 file_name_off;
6210 __u32 line_off;
6211 __u32 line_col;
6212};
6213
6214struct bpf_spin_lock {
6215 __u32 val;
6216};
6217
6218struct bpf_timer {
6219 __u64 :64;
6220 __u64 :64;
6221} __attribute__((aligned(8)));
6222
6223struct bpf_sysctl {
6224 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
6225 * Allows 1,2,4-byte read, but no write.
6226 */
6227 __u32 file_pos; /* Sysctl file position to read from, write to.
6228 * Allows 1,2,4-byte read an 4-byte write.
6229 */
6230};
6231
6232struct bpf_sockopt {
6233 __bpf_md_ptr(struct bpf_sock *, sk);
6234 __bpf_md_ptr(void *, optval);
6235 __bpf_md_ptr(void *, optval_end);
6236
6237 __s32 level;
6238 __s32 optname;
6239 __s32 optlen;
6240 __s32 retval;
6241};
6242
6243struct bpf_pidns_info {
6244 __u32 pid;
6245 __u32 tgid;
6246};
6247
6248/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6249struct bpf_sk_lookup {
6250 union {
6251 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6252 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6253 };
6254
6255 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
6256 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6257 __u32 remote_ip4; /* Network byte order */
6258 __u32 remote_ip6[4]; /* Network byte order */
6259 __be16 remote_port; /* Network byte order */
6260 __u16 :16; /* Zero padding */
6261 __u32 local_ip4; /* Network byte order */
6262 __u32 local_ip6[4]; /* Network byte order */
6263 __u32 local_port; /* Host byte order */
6264};
6265
6266/*
6267 * struct btf_ptr is used for typed pointer representation; the
6268 * type id is used to render the pointer data as the appropriate type
6269 * via the bpf_snprintf_btf() helper described above. A flags field -
6270 * potentially to specify additional details about the BTF pointer
6271 * (rather than its mode of display) - is included for future use.
6272 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6273 */
6274struct btf_ptr {
6275 void *ptr;
6276 __u32 type_id;
6277 __u32 flags; /* BTF ptr flags; unused at present. */
6278};
6279
6280/*
6281 * Flags to control bpf_snprintf_btf() behaviour.
6282 * - BTF_F_COMPACT: no formatting around type information
6283 * - BTF_F_NONAME: no struct/union member names/types
6284 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6285 * equivalent to %px.
6286 * - BTF_F_ZERO: show zero-valued struct/union members; they
6287 * are not displayed by default
6288 */
6289enum {
6290 BTF_F_COMPACT = (1ULL << 0),
6291 BTF_F_NONAME = (1ULL << 1),
6292 BTF_F_PTR_RAW = (1ULL << 2),
6293 BTF_F_ZERO = (1ULL << 3),
6294};
6295
6296#endif /* __LINUX_BPF_H__ */
6297

Provided by KDAB

Privacy Policy
Improve your Profiling and Debugging skills
Find out more

source code of include/linux/bpf.h