1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
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.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
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 */
26 #define BPF_MOV 0xb0 /* mov reg to reg */
27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
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
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 */
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 */
52 /* Register numbers */
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG __MAX_BPF_REG
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 */
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
82 __u8 data[0]; /* Arbitrary size */
85 struct bpf_cgroup_storage_key {
86 __u64 cgroup_inode_id; /* cgroup inode id */
87 __u32 attach_type; /* program attach type */
90 union bpf_iter_link_info {
96 /* BPF syscall commands, see bpf(2) man-page for more details. */
98 * DOC: eBPF Syscall Preamble
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*.
106 * DOC: eBPF Syscall Commands
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.
114 * Applying **close**\ (2) to the file descriptor returned by
115 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
118 * A new file descriptor (a nonnegative integer), or -1 if an
119 * error occurred (in which case, *errno* is set appropriately).
121 * BPF_MAP_LOOKUP_ELEM
123 * Look up an element with a given *key* in the map referred to
124 * by the file descriptor *map_fd*.
126 * The *flags* argument may be specified as one of the
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.
135 * Returns zero on success. On error, -1 is returned and *errno*
136 * is set appropriately.
138 * BPF_MAP_UPDATE_ELEM
140 * Create or update an element (key/value pair) in a specified map.
142 * The *flags* argument should be specified as one of the
146 * Create a new element or update an existing element.
148 * Create a new element only if it did not exist.
150 * Update an existing element.
152 * Update a spin_lock-ed map element.
155 * Returns zero on success. On error, -1 is returned and *errno*
156 * is set appropriately.
158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
159 * **E2BIG**, **EEXIST**, or **ENOENT**.
162 * The number of elements in the map reached the
163 * *max_entries* limit specified at map creation time.
165 * If *flags* specifies **BPF_NOEXIST** and the element
166 * with *key* already exists in the map.
168 * If *flags* specifies **BPF_EXIST** and the element with
169 * *key* does not exist in the map.
171 * BPF_MAP_DELETE_ELEM
173 * Look up and delete an element by key in a specified map.
176 * Returns zero on success. On error, -1 is returned and *errno*
177 * is set appropriately.
179 * BPF_MAP_GET_NEXT_KEY
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
186 * Returns zero on success. On error, -1 is returned and *errno*
187 * is set appropriately.
189 * The following cases can be used to iterate over all elements of
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
199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
200 * **EINVAL** on error.
204 * Verify and load an eBPF program, returning a new file
205 * descriptor associated with the program.
207 * Applying **close**\ (2) to the file descriptor returned by
208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
211 * automatically enabled for the new file descriptor.
214 * A new file descriptor (a nonnegative integer), or -1 if an
215 * error occurred (in which case, *errno* is set appropriately).
219 * Pin an eBPF program or map referred by the specified *bpf_fd*
220 * to the provided *pathname* on the filesystem.
222 * The *pathname* argument must not contain a dot (".").
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
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).
235 * The filesystem type for the parent directory of *pathname* must
236 * be **BPF_FS_MAGIC**.
239 * Returns zero on success. On error, -1 is returned and *errno*
240 * is set appropriately.
244 * Open a file descriptor for the eBPF object pinned to the
245 * specified *pathname*.
248 * A new file descriptor (a nonnegative integer), or -1 if an
249 * error occurred (in which case, *errno* is set appropriately).
253 * Attach an eBPF program to a *target_fd* at the specified
254 * *attach_type* hook.
256 * The *attach_type* specifies the eBPF attachment point to
257 * attach the program to, and must be one of *bpf_attach_type*
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*.
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*:
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**
275 * Control Group v2 hierarchy with the eBPF controller
276 * enabled. Requires the kernel to be compiled with
277 * **CONFIG_CGROUP_BPF**.
279 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
281 * Network namespace (eg /proc/self/ns/net).
283 * **BPF_PROG_TYPE_LIRC_MODE2**
285 * LIRC device path (eg /dev/lircN). Requires the kernel
286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
288 * **BPF_PROG_TYPE_SK_SKB**,
289 * **BPF_PROG_TYPE_SK_MSG**
291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
294 * Returns zero on success. On error, -1 is returned and *errno*
295 * is set appropriately.
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**.
304 * Returns zero on success. On error, -1 is returned and *errno*
305 * is set appropriately.
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.
316 * Returns zero on success. On error, -1 is returned and *errno*
317 * is set appropriately.
320 * Either *data_size_out* or *ctx_size_out* is too small.
322 * This command is not supported by the program type of
323 * the program referred to by *prog_fd*.
325 * BPF_PROG_GET_NEXT_ID
327 * Fetch the next eBPF program currently loaded into the kernel.
329 * Looks for the eBPF program with an id greater than *start_id*
330 * and updates *next_id* on success. If no other eBPF programs
331 * remain with ids higher than *start_id*, returns -1 and sets
332 * *errno* to **ENOENT**.
335 * Returns zero on success. On error, or when no id remains, -1
336 * is returned and *errno* is set appropriately.
338 * BPF_MAP_GET_NEXT_ID
340 * Fetch the next eBPF map currently loaded into the kernel.
342 * Looks for the eBPF map with an id greater than *start_id*
343 * and updates *next_id* on success. If no other eBPF maps
344 * remain with ids higher than *start_id*, returns -1 and sets
345 * *errno* to **ENOENT**.
348 * Returns zero on success. On error, or when no id remains, -1
349 * is returned and *errno* is set appropriately.
351 * BPF_PROG_GET_FD_BY_ID
353 * Open a file descriptor for the eBPF program corresponding to
357 * A new file descriptor (a nonnegative integer), or -1 if an
358 * error occurred (in which case, *errno* is set appropriately).
360 * BPF_MAP_GET_FD_BY_ID
362 * Open a file descriptor for the eBPF map corresponding to
366 * A new file descriptor (a nonnegative integer), or -1 if an
367 * error occurred (in which case, *errno* is set appropriately).
369 * BPF_OBJ_GET_INFO_BY_FD
371 * Obtain information about the eBPF object corresponding to
374 * Populates up to *info_len* bytes of *info*, which will be in
375 * one of the following formats depending on the eBPF object type
378 * * **struct bpf_prog_info**
379 * * **struct bpf_map_info**
380 * * **struct bpf_btf_info**
381 * * **struct bpf_link_info**
384 * Returns zero on success. On error, -1 is returned and *errno*
385 * is set appropriately.
389 * Obtain information about eBPF programs associated with the
390 * specified *attach_type* hook.
392 * The *target_fd* must be a valid file descriptor for a kernel
393 * object which depends on the attach type of *attach_bpf_fd*:
395 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
396 * **BPF_PROG_TYPE_CGROUP_SKB**,
397 * **BPF_PROG_TYPE_CGROUP_SOCK**,
398 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
399 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
400 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
401 * **BPF_PROG_TYPE_SOCK_OPS**
403 * Control Group v2 hierarchy with the eBPF controller
404 * enabled. Requires the kernel to be compiled with
405 * **CONFIG_CGROUP_BPF**.
407 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
409 * Network namespace (eg /proc/self/ns/net).
411 * **BPF_PROG_TYPE_LIRC_MODE2**
413 * LIRC device path (eg /dev/lircN). Requires the kernel
414 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
416 * **BPF_PROG_QUERY** always fetches the number of programs
417 * attached and the *attach_flags* which were used to attach those
418 * programs. Additionally, if *prog_ids* is nonzero and the number
419 * of attached programs is less than *prog_cnt*, populates
420 * *prog_ids* with the eBPF program ids of the programs attached
423 * The following flags may alter the result:
425 * **BPF_F_QUERY_EFFECTIVE**
426 * Only return information regarding programs which are
427 * currently effective at the specified *target_fd*.
430 * Returns zero on success. On error, -1 is returned and *errno*
431 * is set appropriately.
433 * BPF_RAW_TRACEPOINT_OPEN
435 * Attach an eBPF program to a tracepoint *name* to access kernel
436 * internal arguments of the tracepoint in their raw form.
438 * The *prog_fd* must be a valid file descriptor associated with
439 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
441 * No ABI guarantees are made about the content of tracepoint
442 * arguments exposed to the corresponding eBPF program.
444 * Applying **close**\ (2) to the file descriptor returned by
445 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
448 * A new file descriptor (a nonnegative integer), or -1 if an
449 * error occurred (in which case, *errno* is set appropriately).
453 * Verify and load BPF Type Format (BTF) metadata into the kernel,
454 * returning a new file descriptor associated with the metadata.
455 * BTF is described in more detail at
456 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
458 * The *btf* parameter must point to valid memory providing
459 * *btf_size* bytes of BTF binary metadata.
461 * The returned file descriptor can be passed to other **bpf**\ ()
462 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
463 * associate the BTF with those objects.
465 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
466 * parameters to specify a *btf_log_buf*, *btf_log_size* and
467 * *btf_log_level* which allow the kernel to return freeform log
468 * output regarding the BTF verification process.
471 * A new file descriptor (a nonnegative integer), or -1 if an
472 * error occurred (in which case, *errno* is set appropriately).
474 * BPF_BTF_GET_FD_BY_ID
476 * Open a file descriptor for the BPF Type Format (BTF)
477 * corresponding to *btf_id*.
480 * A new file descriptor (a nonnegative integer), or -1 if an
481 * error occurred (in which case, *errno* is set appropriately).
485 * Obtain information about eBPF programs associated with the
486 * target process identified by *pid* and *fd*.
488 * If the *pid* and *fd* are associated with a tracepoint, kprobe
489 * or uprobe perf event, then the *prog_id* and *fd_type* will
490 * be populated with the eBPF program id and file descriptor type
491 * of type **bpf_task_fd_type**. If associated with a kprobe or
492 * uprobe, the *probe_offset* and *probe_addr* will also be
493 * populated. Optionally, if *buf* is provided, then up to
494 * *buf_len* bytes of *buf* will be populated with the name of
495 * the tracepoint, kprobe or uprobe.
497 * The resulting *prog_id* may be introspected in deeper detail
498 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
501 * Returns zero on success. On error, -1 is returned and *errno*
502 * is set appropriately.
504 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
506 * Look up an element with the given *key* in the map referred to
507 * by the file descriptor *fd*, and if found, delete the element.
509 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
510 * implement this command as a "pop" operation, deleting the top
511 * element rather than one corresponding to *key*.
512 * The *key* and *key_len* parameters should be zeroed when
513 * issuing this operation for these map types.
515 * This command is only valid for the following map types:
516 * * **BPF_MAP_TYPE_QUEUE**
517 * * **BPF_MAP_TYPE_STACK**
520 * Returns zero on success. On error, -1 is returned and *errno*
521 * is set appropriately.
525 * Freeze the permissions of the specified map.
527 * Write permissions may be frozen by passing zero *flags*.
528 * Upon success, no future syscall invocations may alter the
529 * map state of *map_fd*. Write operations from eBPF programs
530 * are still possible for a frozen map.
532 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
535 * Returns zero on success. On error, -1 is returned and *errno*
536 * is set appropriately.
538 * BPF_BTF_GET_NEXT_ID
540 * Fetch the next BPF Type Format (BTF) object currently loaded
543 * Looks for the BTF object with an id greater than *start_id*
544 * and updates *next_id* on success. If no other BTF objects
545 * remain with ids higher than *start_id*, returns -1 and sets
546 * *errno* to **ENOENT**.
549 * Returns zero on success. On error, or when no id remains, -1
550 * is returned and *errno* is set appropriately.
552 * BPF_MAP_LOOKUP_BATCH
554 * Iterate and fetch multiple elements in a map.
556 * Two opaque values are used to manage batch operations,
557 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
558 * to NULL to begin the batched operation. After each subsequent
559 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
560 * *out_batch* as the *in_batch* for the next operation to
561 * continue iteration from the current point.
563 * The *keys* and *values* are output parameters which must point
564 * to memory large enough to hold *count* items based on the key
565 * and value size of the map *map_fd*. The *keys* buffer must be
566 * of *key_size* * *count*. The *values* buffer must be of
567 * *value_size* * *count*.
569 * The *elem_flags* argument may be specified as one of the
573 * Look up the value of a spin-locked map without
574 * returning the lock. This must be specified if the
575 * elements contain a spinlock.
577 * On success, *count* elements from the map are copied into the
578 * user buffer, with the keys copied into *keys* and the values
579 * copied into the corresponding indices in *values*.
581 * If an error is returned and *errno* is not **EFAULT**, *count*
582 * is set to the number of successfully processed elements.
585 * Returns zero on success. On error, -1 is returned and *errno*
586 * is set appropriately.
588 * May set *errno* to **ENOSPC** to indicate that *keys* or
589 * *values* is too small to dump an entire bucket during
590 * iteration of a hash-based map type.
592 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
594 * Iterate and delete all elements in a map.
596 * This operation has the same behavior as
597 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
599 * * Every element that is successfully returned is also deleted
600 * from the map. This is at least *count* elements. Note that
601 * *count* is both an input and an output parameter.
602 * * Upon returning with *errno* set to **EFAULT**, up to
603 * *count* elements may be deleted without returning the keys
604 * and values of the deleted elements.
607 * Returns zero on success. On error, -1 is returned and *errno*
608 * is set appropriately.
610 * BPF_MAP_UPDATE_BATCH
612 * Update multiple elements in a map by *key*.
614 * The *keys* and *values* are input parameters which must point
615 * to memory large enough to hold *count* items based on the key
616 * and value size of the map *map_fd*. The *keys* buffer must be
617 * of *key_size* * *count*. The *values* buffer must be of
618 * *value_size* * *count*.
620 * Each element specified in *keys* is sequentially updated to the
621 * value in the corresponding index in *values*. The *in_batch*
622 * and *out_batch* parameters are ignored and should be zeroed.
624 * The *elem_flags* argument should be specified as one of the
628 * Create new elements or update a existing elements.
630 * Create new elements only if they do not exist.
632 * Update existing elements.
634 * Update spin_lock-ed map elements. This must be
635 * specified if the map value contains a spinlock.
637 * On success, *count* elements from the map are updated.
639 * If an error is returned and *errno* is not **EFAULT**, *count*
640 * is set to the number of successfully processed elements.
643 * Returns zero on success. On error, -1 is returned and *errno*
644 * is set appropriately.
646 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
647 * **E2BIG**. **E2BIG** indicates that the number of elements in
648 * the map reached the *max_entries* limit specified at map
651 * May set *errno* to one of the following error codes under
652 * specific circumstances:
655 * If *flags* specifies **BPF_NOEXIST** and the element
656 * with *key* already exists in the map.
658 * If *flags* specifies **BPF_EXIST** and the element with
659 * *key* does not exist in the map.
661 * BPF_MAP_DELETE_BATCH
663 * Delete multiple elements in a map by *key*.
665 * The *keys* parameter is an input parameter which must point
666 * to memory large enough to hold *count* items based on the key
667 * size of the map *map_fd*, that is, *key_size* * *count*.
669 * Each element specified in *keys* is sequentially deleted. The
670 * *in_batch*, *out_batch*, and *values* parameters are ignored
671 * and should be zeroed.
673 * The *elem_flags* argument may be specified as one of the
677 * Look up the value of a spin-locked map without
678 * returning the lock. This must be specified if the
679 * elements contain a spinlock.
681 * On success, *count* elements from the map are updated.
683 * If an error is returned and *errno* is not **EFAULT**, *count*
684 * is set to the number of successfully processed elements. If
685 * *errno* is **EFAULT**, up to *count* elements may be been
689 * Returns zero on success. On error, -1 is returned and *errno*
690 * is set appropriately.
694 * Attach an eBPF program to a *target_fd* at the specified
695 * *attach_type* hook and return a file descriptor handle for
699 * A new file descriptor (a nonnegative integer), or -1 if an
700 * error occurred (in which case, *errno* is set appropriately).
704 * Update the eBPF program in the specified *link_fd* to
708 * Returns zero on success. On error, -1 is returned and *errno*
709 * is set appropriately.
711 * BPF_LINK_GET_FD_BY_ID
713 * Open a file descriptor for the eBPF Link corresponding to
717 * A new file descriptor (a nonnegative integer), or -1 if an
718 * error occurred (in which case, *errno* is set appropriately).
720 * BPF_LINK_GET_NEXT_ID
722 * Fetch the next eBPF link currently loaded into the kernel.
724 * Looks for the eBPF link with an id greater than *start_id*
725 * and updates *next_id* on success. If no other eBPF links
726 * remain with ids higher than *start_id*, returns -1 and sets
727 * *errno* to **ENOENT**.
730 * Returns zero on success. On error, or when no id remains, -1
731 * is returned and *errno* is set appropriately.
735 * Enable eBPF runtime statistics gathering.
737 * Runtime statistics gathering for the eBPF runtime is disabled
738 * by default to minimize the corresponding performance overhead.
739 * This command enables statistics globally.
741 * Multiple programs may independently enable statistics.
742 * After gathering the desired statistics, eBPF runtime statistics
743 * may be disabled again by calling **close**\ (2) for the file
744 * descriptor returned by this function. Statistics will only be
745 * disabled system-wide when all outstanding file descriptors
746 * returned by prior calls for this subcommand are closed.
749 * A new file descriptor (a nonnegative integer), or -1 if an
750 * error occurred (in which case, *errno* is set appropriately).
754 * Create an iterator on top of the specified *link_fd* (as
755 * previously created using **BPF_LINK_CREATE**) and return a
756 * file descriptor that can be used to trigger the iteration.
758 * If the resulting file descriptor is pinned to the filesystem
759 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
760 * for that path will trigger the iterator to read kernel state
761 * using the eBPF program attached to *link_fd*.
764 * A new file descriptor (a nonnegative integer), or -1 if an
765 * error occurred (in which case, *errno* is set appropriately).
769 * Forcefully detach the specified *link_fd* from its
770 * corresponding attachment point.
773 * Returns zero on success. On error, -1 is returned and *errno*
774 * is set appropriately.
778 * Bind a map to the lifetime of an eBPF program.
780 * The map identified by *map_fd* is bound to the program
781 * identified by *prog_fd* and only released when *prog_fd* is
782 * released. This may be used in cases where metadata should be
783 * associated with a program which otherwise does not contain any
784 * references to the map (for example, embedded in the eBPF
785 * program instructions).
788 * Returns zero on success. On error, -1 is returned and *errno*
789 * is set appropriately.
792 * eBPF objects (maps and programs) can be shared between processes.
794 * * After **fork**\ (2), the child inherits file descriptors
795 * referring to the same eBPF objects.
796 * * File descriptors referring to eBPF objects can be transferred over
797 * **unix**\ (7) domain sockets.
798 * * File descriptors referring to eBPF objects can be duplicated in the
799 * usual way, using **dup**\ (2) and similar calls.
800 * * File descriptors referring to eBPF objects can be pinned to the
801 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
803 * An eBPF object is deallocated only after all file descriptors referring
804 * to the object have been closed and no references remain pinned to the
805 * filesystem or attached (for example, bound to a program or device).
812 BPF_MAP_GET_NEXT_KEY,
819 BPF_PROG_GET_NEXT_ID,
821 BPF_PROG_GET_FD_BY_ID,
822 BPF_MAP_GET_FD_BY_ID,
823 BPF_OBJ_GET_INFO_BY_FD,
825 BPF_RAW_TRACEPOINT_OPEN,
827 BPF_BTF_GET_FD_BY_ID,
829 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
832 BPF_MAP_LOOKUP_BATCH,
833 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
834 BPF_MAP_UPDATE_BATCH,
835 BPF_MAP_DELETE_BATCH,
838 BPF_LINK_GET_FD_BY_ID,
839 BPF_LINK_GET_NEXT_ID,
850 BPF_MAP_TYPE_PROG_ARRAY,
851 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
852 BPF_MAP_TYPE_PERCPU_HASH,
853 BPF_MAP_TYPE_PERCPU_ARRAY,
854 BPF_MAP_TYPE_STACK_TRACE,
855 BPF_MAP_TYPE_CGROUP_ARRAY,
856 BPF_MAP_TYPE_LRU_HASH,
857 BPF_MAP_TYPE_LRU_PERCPU_HASH,
858 BPF_MAP_TYPE_LPM_TRIE,
859 BPF_MAP_TYPE_ARRAY_OF_MAPS,
860 BPF_MAP_TYPE_HASH_OF_MAPS,
862 BPF_MAP_TYPE_SOCKMAP,
865 BPF_MAP_TYPE_SOCKHASH,
866 BPF_MAP_TYPE_CGROUP_STORAGE,
867 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
868 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
871 BPF_MAP_TYPE_SK_STORAGE,
872 BPF_MAP_TYPE_DEVMAP_HASH,
873 BPF_MAP_TYPE_STRUCT_OPS,
874 BPF_MAP_TYPE_RINGBUF,
875 BPF_MAP_TYPE_INODE_STORAGE,
876 BPF_MAP_TYPE_TASK_STORAGE,
879 /* Note that tracing related programs such as
880 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
881 * are not subject to a stable API since kernel internal data
882 * structures can change from release to release and may
883 * therefore break existing tracing BPF programs. Tracing BPF
884 * programs correspond to /a/ specific kernel which is to be
885 * analyzed, and not /a/ specific kernel /and/ all future ones.
888 BPF_PROG_TYPE_UNSPEC,
889 BPF_PROG_TYPE_SOCKET_FILTER,
890 BPF_PROG_TYPE_KPROBE,
891 BPF_PROG_TYPE_SCHED_CLS,
892 BPF_PROG_TYPE_SCHED_ACT,
893 BPF_PROG_TYPE_TRACEPOINT,
895 BPF_PROG_TYPE_PERF_EVENT,
896 BPF_PROG_TYPE_CGROUP_SKB,
897 BPF_PROG_TYPE_CGROUP_SOCK,
898 BPF_PROG_TYPE_LWT_IN,
899 BPF_PROG_TYPE_LWT_OUT,
900 BPF_PROG_TYPE_LWT_XMIT,
901 BPF_PROG_TYPE_SOCK_OPS,
902 BPF_PROG_TYPE_SK_SKB,
903 BPF_PROG_TYPE_CGROUP_DEVICE,
904 BPF_PROG_TYPE_SK_MSG,
905 BPF_PROG_TYPE_RAW_TRACEPOINT,
906 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
907 BPF_PROG_TYPE_LWT_SEG6LOCAL,
908 BPF_PROG_TYPE_LIRC_MODE2,
909 BPF_PROG_TYPE_SK_REUSEPORT,
910 BPF_PROG_TYPE_FLOW_DISSECTOR,
911 BPF_PROG_TYPE_CGROUP_SYSCTL,
912 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
913 BPF_PROG_TYPE_CGROUP_SOCKOPT,
914 BPF_PROG_TYPE_TRACING,
915 BPF_PROG_TYPE_STRUCT_OPS,
918 BPF_PROG_TYPE_SK_LOOKUP,
921 enum bpf_attach_type {
922 BPF_CGROUP_INET_INGRESS,
923 BPF_CGROUP_INET_EGRESS,
924 BPF_CGROUP_INET_SOCK_CREATE,
926 BPF_SK_SKB_STREAM_PARSER,
927 BPF_SK_SKB_STREAM_VERDICT,
930 BPF_CGROUP_INET4_BIND,
931 BPF_CGROUP_INET6_BIND,
932 BPF_CGROUP_INET4_CONNECT,
933 BPF_CGROUP_INET6_CONNECT,
934 BPF_CGROUP_INET4_POST_BIND,
935 BPF_CGROUP_INET6_POST_BIND,
936 BPF_CGROUP_UDP4_SENDMSG,
937 BPF_CGROUP_UDP6_SENDMSG,
941 BPF_CGROUP_UDP4_RECVMSG,
942 BPF_CGROUP_UDP6_RECVMSG,
943 BPF_CGROUP_GETSOCKOPT,
944 BPF_CGROUP_SETSOCKOPT,
951 BPF_CGROUP_INET4_GETPEERNAME,
952 BPF_CGROUP_INET6_GETPEERNAME,
953 BPF_CGROUP_INET4_GETSOCKNAME,
954 BPF_CGROUP_INET6_GETSOCKNAME,
956 BPF_CGROUP_INET_SOCK_RELEASE,
960 __MAX_BPF_ATTACH_TYPE
963 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
966 BPF_LINK_TYPE_UNSPEC = 0,
967 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
968 BPF_LINK_TYPE_TRACING = 2,
969 BPF_LINK_TYPE_CGROUP = 3,
970 BPF_LINK_TYPE_ITER = 4,
971 BPF_LINK_TYPE_NETNS = 5,
972 BPF_LINK_TYPE_XDP = 6,
977 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
979 * NONE(default): No further bpf programs allowed in the subtree.
981 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
982 * the program in this cgroup yields to sub-cgroup program.
984 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
985 * that cgroup program gets run in addition to the program in this cgroup.
987 * Only one program is allowed to be attached to a cgroup with
988 * NONE or BPF_F_ALLOW_OVERRIDE flag.
989 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
990 * release old program and attach the new one. Attach flags has to match.
992 * Multiple programs are allowed to be attached to a cgroup with
993 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
994 * (those that were attached first, run first)
995 * The programs of sub-cgroup are executed first, then programs of
996 * this cgroup and then programs of parent cgroup.
997 * When children program makes decision (like picking TCP CA or sock bind)
998 * parent program has a chance to override it.
1000 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1001 * programs for a cgroup. Though it's possible to replace an old program at
1002 * any position by also specifying BPF_F_REPLACE flag and position itself in
1003 * replace_bpf_fd attribute. Old program at this position will be released.
1005 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1006 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1008 * cgrp1 (MULTI progs A, B) ->
1009 * cgrp2 (OVERRIDE prog C) ->
1010 * cgrp3 (MULTI prog D) ->
1011 * cgrp4 (OVERRIDE prog E) ->
1012 * cgrp5 (NONE prog F)
1013 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1014 * if prog F is detached, the execution is E,D,A,B
1015 * if prog F and D are detached, the execution is E,A,B
1016 * if prog F, E and D are detached, the execution is C,A,B
1018 * All eligible programs are executed regardless of return code from
1021 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
1022 #define BPF_F_ALLOW_MULTI (1U << 1)
1023 #define BPF_F_REPLACE (1U << 2)
1025 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1026 * verifier will perform strict alignment checking as if the kernel
1027 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1028 * and NET_IP_ALIGN defined to 2.
1030 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
1032 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1033 * verifier will allow any alignment whatsoever. On platforms
1034 * with strict alignment requirements for loads ands stores (such
1035 * as sparc and mips) the verifier validates that all loads and
1036 * stores provably follow this requirement. This flag turns that
1037 * checking and enforcement off.
1039 * It is mostly used for testing when we want to validate the
1040 * context and memory access aspects of the verifier, but because
1041 * of an unaligned access the alignment check would trigger before
1042 * the one we are interested in.
1044 #define BPF_F_ANY_ALIGNMENT (1U << 1)
1046 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1047 * Verifier does sub-register def/use analysis and identifies instructions whose
1048 * def only matters for low 32-bit, high 32-bit is never referenced later
1049 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1050 * that it is safe to ignore clearing high 32-bit for these instructions. This
1051 * saves some back-ends a lot of code-gen. However such optimization is not
1052 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1053 * hence hasn't used verifier's analysis result. But, we really want to have a
1054 * way to be able to verify the correctness of the described optimization on
1055 * x86_64 on which testsuites are frequently exercised.
1057 * So, this flag is introduced. Once it is set, verifier will randomize high
1058 * 32-bit for those instructions who has been identified as safe to ignore them.
1059 * Then, if verifier is not doing correct analysis, such randomization will
1060 * regress tests to expose bugs.
1062 #define BPF_F_TEST_RND_HI32 (1U << 2)
1064 /* The verifier internal test flag. Behavior is undefined */
1065 #define BPF_F_TEST_STATE_FREQ (1U << 3)
1067 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1068 * restrict map and helper usage for such programs. Sleepable BPF programs can
1069 * only be attached to hooks where kernel execution context allows sleeping.
1070 * Such programs are allowed to use helpers that may sleep like
1071 * bpf_copy_from_user().
1073 #define BPF_F_SLEEPABLE (1U << 4)
1075 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1076 * the following extensions:
1078 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
1079 * insn[0].imm: map fd
1083 * ldimm64 rewrite: address of map
1084 * verifier type: CONST_PTR_TO_MAP
1086 #define BPF_PSEUDO_MAP_FD 1
1087 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
1088 * insn[0].imm: map fd
1089 * insn[1].imm: offset into value
1092 * ldimm64 rewrite: address of map[0]+offset
1093 * verifier type: PTR_TO_MAP_VALUE
1095 #define BPF_PSEUDO_MAP_VALUE 2
1096 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1097 * insn[0].imm: kernel btd id of VAR
1101 * ldimm64 rewrite: address of the kernel variable
1102 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1105 #define BPF_PSEUDO_BTF_ID 3
1106 /* insn[0].src_reg: BPF_PSEUDO_FUNC
1107 * insn[0].imm: insn offset to the func
1111 * ldimm64 rewrite: address of the function
1112 * verifier type: PTR_TO_FUNC.
1114 #define BPF_PSEUDO_FUNC 4
1116 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1117 * offset to another bpf function
1119 #define BPF_PSEUDO_CALL 1
1120 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1121 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1123 #define BPF_PSEUDO_KFUNC_CALL 2
1125 /* flags for BPF_MAP_UPDATE_ELEM command */
1127 BPF_ANY = 0, /* create new element or update existing */
1128 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1129 BPF_EXIST = 2, /* update existing element */
1130 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1133 /* flags for BPF_MAP_CREATE command */
1135 BPF_F_NO_PREALLOC = (1U << 0),
1136 /* Instead of having one common LRU list in the
1137 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1138 * which can scale and perform better.
1139 * Note, the LRU nodes (including free nodes) cannot be moved
1140 * across different LRU lists.
1142 BPF_F_NO_COMMON_LRU = (1U << 1),
1143 /* Specify numa node during map creation */
1144 BPF_F_NUMA_NODE = (1U << 2),
1146 /* Flags for accessing BPF object from syscall side. */
1147 BPF_F_RDONLY = (1U << 3),
1148 BPF_F_WRONLY = (1U << 4),
1150 /* Flag for stack_map, store build_id+offset instead of pointer */
1151 BPF_F_STACK_BUILD_ID = (1U << 5),
1153 /* Zero-initialize hash function seed. This should only be used for testing. */
1154 BPF_F_ZERO_SEED = (1U << 6),
1156 /* Flags for accessing BPF object from program side. */
1157 BPF_F_RDONLY_PROG = (1U << 7),
1158 BPF_F_WRONLY_PROG = (1U << 8),
1160 /* Clone map from listener for newly accepted socket */
1161 BPF_F_CLONE = (1U << 9),
1163 /* Enable memory-mapping BPF map */
1164 BPF_F_MMAPABLE = (1U << 10),
1166 /* Share perf_event among processes */
1167 BPF_F_PRESERVE_ELEMS = (1U << 11),
1169 /* Create a map that is suitable to be an inner map with dynamic max entries */
1170 BPF_F_INNER_MAP = (1U << 12),
1173 /* Flags for BPF_PROG_QUERY. */
1175 /* Query effective (directly attached + inherited from ancestor cgroups)
1176 * programs that will be executed for events within a cgroup.
1177 * attach_flags with this flag are returned only for directly attached programs.
1179 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
1181 /* Flags for BPF_PROG_TEST_RUN */
1183 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1184 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1186 /* type for BPF_ENABLE_STATS */
1187 enum bpf_stats_type {
1188 /* enabled run_time_ns and run_cnt */
1189 BPF_STATS_RUN_TIME = 0,
1192 enum bpf_stack_build_id_status {
1193 /* user space need an empty entry to identify end of a trace */
1194 BPF_STACK_BUILD_ID_EMPTY = 0,
1195 /* with valid build_id and offset */
1196 BPF_STACK_BUILD_ID_VALID = 1,
1197 /* couldn't get build_id, fallback to ip */
1198 BPF_STACK_BUILD_ID_IP = 2,
1201 #define BPF_BUILD_ID_SIZE 20
1202 struct bpf_stack_build_id {
1204 unsigned char build_id[BPF_BUILD_ID_SIZE];
1211 #define BPF_OBJ_NAME_LEN 16U
1214 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1215 __u32 map_type; /* one of enum bpf_map_type */
1216 __u32 key_size; /* size of key in bytes */
1217 __u32 value_size; /* size of value in bytes */
1218 __u32 max_entries; /* max number of entries in a map */
1219 __u32 map_flags; /* BPF_MAP_CREATE related
1220 * flags defined above.
1222 __u32 inner_map_fd; /* fd pointing to the inner map */
1223 __u32 numa_node; /* numa node (effective only if
1224 * BPF_F_NUMA_NODE is set).
1226 char map_name[BPF_OBJ_NAME_LEN];
1227 __u32 map_ifindex; /* ifindex of netdev to create on */
1228 __u32 btf_fd; /* fd pointing to a BTF type data */
1229 __u32 btf_key_type_id; /* BTF type_id of the key */
1230 __u32 btf_value_type_id; /* BTF type_id of the value */
1231 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1232 * struct stored as the
1237 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1241 __aligned_u64 value;
1242 __aligned_u64 next_key;
1247 struct { /* struct used by BPF_MAP_*_BATCH commands */
1248 __aligned_u64 in_batch; /* start batch,
1249 * NULL to start from beginning
1251 __aligned_u64 out_batch; /* output: next start batch */
1253 __aligned_u64 values;
1254 __u32 count; /* input/output:
1255 * input: # of key/value
1257 * output: # of filled elements
1264 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1265 __u32 prog_type; /* one of enum bpf_prog_type */
1267 __aligned_u64 insns;
1268 __aligned_u64 license;
1269 __u32 log_level; /* verbosity level of verifier */
1270 __u32 log_size; /* size of user buffer */
1271 __aligned_u64 log_buf; /* user supplied buffer */
1272 __u32 kern_version; /* not used */
1274 char prog_name[BPF_OBJ_NAME_LEN];
1275 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1276 /* For some prog types expected attach type must be known at
1277 * load time to verify attach type specific parts of prog
1278 * (context accesses, allowed helpers, etc).
1280 __u32 expected_attach_type;
1281 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1282 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1283 __aligned_u64 func_info; /* func info */
1284 __u32 func_info_cnt; /* number of bpf_func_info records */
1285 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1286 __aligned_u64 line_info; /* line info */
1287 __u32 line_info_cnt; /* number of bpf_line_info records */
1288 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1290 /* valid prog_fd to attach to bpf prog */
1291 __u32 attach_prog_fd;
1292 /* or valid module BTF object fd or 0 to attach to vmlinux */
1293 __u32 attach_btf_obj_fd;
1297 struct { /* anonymous struct used by BPF_OBJ_* commands */
1298 __aligned_u64 pathname;
1303 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1304 __u32 target_fd; /* container object to attach to */
1305 __u32 attach_bpf_fd; /* eBPF program to attach */
1308 __u32 replace_bpf_fd; /* previously attached eBPF
1309 * program to replace if
1310 * BPF_F_REPLACE is used
1314 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1317 __u32 data_size_in; /* input: len of data_in */
1318 __u32 data_size_out; /* input/output: len of data_out
1319 * returns ENOSPC if data_out
1322 __aligned_u64 data_in;
1323 __aligned_u64 data_out;
1326 __u32 ctx_size_in; /* input: len of ctx_in */
1327 __u32 ctx_size_out; /* input/output: len of ctx_out
1328 * returns ENOSPC if ctx_out
1331 __aligned_u64 ctx_in;
1332 __aligned_u64 ctx_out;
1337 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1349 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1355 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1356 __u32 target_fd; /* container object to query */
1360 __aligned_u64 prog_ids;
1364 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1369 struct { /* anonymous struct for BPF_BTF_LOAD */
1371 __aligned_u64 btf_log_buf;
1374 __u32 btf_log_level;
1378 __u32 pid; /* input: pid */
1379 __u32 fd; /* input: fd */
1380 __u32 flags; /* input: flags */
1381 __u32 buf_len; /* input/output: buf len */
1382 __aligned_u64 buf; /* input/output:
1383 * tp_name for tracepoint
1385 * filename for uprobe
1387 __u32 prog_id; /* output: prod_id */
1388 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1389 __u64 probe_offset; /* output: probe_offset */
1390 __u64 probe_addr; /* output: probe_addr */
1393 struct { /* struct used by BPF_LINK_CREATE command */
1394 __u32 prog_fd; /* eBPF program to attach */
1396 __u32 target_fd; /* object to attach to */
1397 __u32 target_ifindex; /* target ifindex */
1399 __u32 attach_type; /* attach type */
1400 __u32 flags; /* extra flags */
1402 __u32 target_btf_id; /* btf_id of target to attach to */
1404 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1405 __u32 iter_info_len; /* iter_info length */
1410 struct { /* struct used by BPF_LINK_UPDATE command */
1411 __u32 link_fd; /* link fd */
1412 /* new program fd to update link with */
1414 __u32 flags; /* extra flags */
1415 /* expected link's program fd; is specified only if
1416 * BPF_F_REPLACE flag is set in flags */
1424 struct { /* struct used by BPF_ENABLE_STATS command */
1428 struct { /* struct used by BPF_ITER_CREATE command */
1433 struct { /* struct used by BPF_PROG_BIND_MAP command */
1436 __u32 flags; /* extra flags */
1439 } __attribute__((aligned(8)));
1441 /* The description below is an attempt at providing documentation to eBPF
1442 * developers about the multiple available eBPF helper functions. It can be
1443 * parsed and used to produce a manual page. The workflow is the following,
1444 * and requires the rst2man utility:
1446 * $ ./scripts/bpf_doc.py \
1447 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1448 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1449 * $ man /tmp/bpf-helpers.7
1451 * Note that in order to produce this external documentation, some RST
1452 * formatting is used in the descriptions to get "bold" and "italics" in
1453 * manual pages. Also note that the few trailing white spaces are
1454 * intentional, removing them would break paragraphs for rst2man.
1456 * Start of BPF helper function descriptions:
1458 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1460 * Perform a lookup in *map* for an entry associated to *key*.
1462 * Map value associated to *key*, or **NULL** if no entry was
1465 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1467 * Add or update the value of the entry associated to *key* in
1468 * *map* with *value*. *flags* is one of:
1471 * The entry for *key* must not exist in the map.
1473 * The entry for *key* must already exist in the map.
1475 * No condition on the existence of the entry for *key*.
1477 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1478 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1479 * elements always exist), the helper would return an error.
1481 * 0 on success, or a negative error in case of failure.
1483 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1485 * Delete entry with *key* from *map*.
1487 * 0 on success, or a negative error in case of failure.
1489 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1491 * For tracing programs, safely attempt to read *size* bytes from
1492 * kernel space address *unsafe_ptr* and store the data in *dst*.
1494 * Generally, use **bpf_probe_read_user**\ () or
1495 * **bpf_probe_read_kernel**\ () instead.
1497 * 0 on success, or a negative error in case of failure.
1499 * u64 bpf_ktime_get_ns(void)
1501 * Return the time elapsed since system boot, in nanoseconds.
1502 * Does not include time the system was suspended.
1503 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1507 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1509 * This helper is a "printk()-like" facility for debugging. It
1510 * prints a message defined by format *fmt* (of size *fmt_size*)
1511 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1512 * available. It can take up to three additional **u64**
1513 * arguments (as an eBPF helpers, the total number of arguments is
1516 * Each time the helper is called, it appends a line to the trace.
1517 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1518 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1519 * The format of the trace is customizable, and the exact output
1520 * one will get depends on the options set in
1521 * *\/sys/kernel/debug/tracing/trace_options* (see also the
1522 * *README* file under the same directory). However, it usually
1523 * defaults to something like:
1527 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1531 * * ``telnet`` is the name of the current task.
1532 * * ``470`` is the PID of the current task.
1533 * * ``001`` is the CPU number on which the task is
1535 * * In ``.N..``, each character refers to a set of
1536 * options (whether irqs are enabled, scheduling
1537 * options, whether hard/softirqs are running, level of
1538 * preempt_disabled respectively). **N** means that
1539 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1541 * * ``419421.045894`` is a timestamp.
1542 * * ``0x00000001`` is a fake value used by BPF for the
1543 * instruction pointer register.
1544 * * ``<formatted msg>`` is the message formatted with
1547 * The conversion specifiers supported by *fmt* are similar, but
1548 * more limited than for printk(). They are **%d**, **%i**,
1549 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1550 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1551 * of field, padding with zeroes, etc.) is available, and the
1552 * helper will return **-EINVAL** (but print nothing) if it
1553 * encounters an unknown specifier.
1555 * Also, note that **bpf_trace_printk**\ () is slow, and should
1556 * only be used for debugging purposes. For this reason, a notice
1557 * block (spanning several lines) is printed to kernel logs and
1558 * states that the helper should not be used "for production use"
1559 * the first time this helper is used (or more precisely, when
1560 * **trace_printk**\ () buffers are allocated). For passing values
1561 * to user space, perf events should be preferred.
1563 * The number of bytes written to the buffer, or a negative error
1564 * in case of failure.
1566 * u32 bpf_get_prandom_u32(void)
1568 * Get a pseudo-random number.
1570 * From a security point of view, this helper uses its own
1571 * pseudo-random internal state, and cannot be used to infer the
1572 * seed of other random functions in the kernel. However, it is
1573 * essential to note that the generator used by the helper is not
1574 * cryptographically secure.
1576 * A random 32-bit unsigned value.
1578 * u32 bpf_get_smp_processor_id(void)
1580 * Get the SMP (symmetric multiprocessing) processor id. Note that
1581 * all programs run with preemption disabled, which means that the
1582 * SMP processor id is stable during all the execution of the
1585 * The SMP id of the processor running the program.
1587 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1589 * Store *len* bytes from address *from* into the packet
1590 * associated to *skb*, at *offset*. *flags* are a combination of
1591 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1592 * checksum for the packet after storing the bytes) and
1593 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1594 * **->swhash** and *skb*\ **->l4hash** to 0).
1596 * A call to this helper is susceptible to change the underlying
1597 * packet buffer. Therefore, at load time, all checks on pointers
1598 * previously done by the verifier are invalidated and must be
1599 * performed again, if the helper is used in combination with
1600 * direct packet access.
1602 * 0 on success, or a negative error in case of failure.
1604 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1606 * Recompute the layer 3 (e.g. IP) checksum for the packet
1607 * associated to *skb*. Computation is incremental, so the helper
1608 * must know the former value of the header field that was
1609 * modified (*from*), the new value of this field (*to*), and the
1610 * number of bytes (2 or 4) for this field, stored in *size*.
1611 * Alternatively, it is possible to store the difference between
1612 * the previous and the new values of the header field in *to*, by
1613 * setting *from* and *size* to 0. For both methods, *offset*
1614 * indicates the location of the IP checksum within the packet.
1616 * This helper works in combination with **bpf_csum_diff**\ (),
1617 * which does not update the checksum in-place, but offers more
1618 * flexibility and can handle sizes larger than 2 or 4 for the
1619 * checksum to update.
1621 * A call to this helper is susceptible to change the underlying
1622 * packet buffer. Therefore, at load time, all checks on pointers
1623 * previously done by the verifier are invalidated and must be
1624 * performed again, if the helper is used in combination with
1625 * direct packet access.
1627 * 0 on success, or a negative error in case of failure.
1629 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1631 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1632 * packet associated to *skb*. Computation is incremental, so the
1633 * helper must know the former value of the header field that was
1634 * modified (*from*), the new value of this field (*to*), and the
1635 * number of bytes (2 or 4) for this field, stored on the lowest
1636 * four bits of *flags*. Alternatively, it is possible to store
1637 * the difference between the previous and the new values of the
1638 * header field in *to*, by setting *from* and the four lowest
1639 * bits of *flags* to 0. For both methods, *offset* indicates the
1640 * location of the IP checksum within the packet. In addition to
1641 * the size of the field, *flags* can be added (bitwise OR) actual
1642 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1643 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1644 * for updates resulting in a null checksum the value is set to
1645 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1646 * the checksum is to be computed against a pseudo-header.
1648 * This helper works in combination with **bpf_csum_diff**\ (),
1649 * which does not update the checksum in-place, but offers more
1650 * flexibility and can handle sizes larger than 2 or 4 for the
1651 * checksum to update.
1653 * A call to this helper is susceptible to change the underlying
1654 * packet buffer. Therefore, at load time, all checks on pointers
1655 * previously done by the verifier are invalidated and must be
1656 * performed again, if the helper is used in combination with
1657 * direct packet access.
1659 * 0 on success, or a negative error in case of failure.
1661 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1663 * This special helper is used to trigger a "tail call", or in
1664 * other words, to jump into another eBPF program. The same stack
1665 * frame is used (but values on stack and in registers for the
1666 * caller are not accessible to the callee). This mechanism allows
1667 * for program chaining, either for raising the maximum number of
1668 * available eBPF instructions, or to execute given programs in
1669 * conditional blocks. For security reasons, there is an upper
1670 * limit to the number of successive tail calls that can be
1673 * Upon call of this helper, the program attempts to jump into a
1674 * program referenced at index *index* in *prog_array_map*, a
1675 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1676 * *ctx*, a pointer to the context.
1678 * If the call succeeds, the kernel immediately runs the first
1679 * instruction of the new program. This is not a function call,
1680 * and it never returns to the previous program. If the call
1681 * fails, then the helper has no effect, and the caller continues
1682 * to run its subsequent instructions. A call can fail if the
1683 * destination program for the jump does not exist (i.e. *index*
1684 * is superior to the number of entries in *prog_array_map*), or
1685 * if the maximum number of tail calls has been reached for this
1686 * chain of programs. This limit is defined in the kernel by the
1687 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1688 * which is currently set to 32.
1690 * 0 on success, or a negative error in case of failure.
1692 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1694 * Clone and redirect the packet associated to *skb* to another
1695 * net device of index *ifindex*. Both ingress and egress
1696 * interfaces can be used for redirection. The **BPF_F_INGRESS**
1697 * value in *flags* is used to make the distinction (ingress path
1698 * is selected if the flag is present, egress path otherwise).
1699 * This is the only flag supported for now.
1701 * In comparison with **bpf_redirect**\ () helper,
1702 * **bpf_clone_redirect**\ () has the associated cost of
1703 * duplicating the packet buffer, but this can be executed out of
1704 * the eBPF program. Conversely, **bpf_redirect**\ () is more
1705 * efficient, but it is handled through an action code where the
1706 * redirection happens only after the eBPF program has returned.
1708 * A call to this helper is susceptible to change the underlying
1709 * packet buffer. Therefore, at load time, all checks on pointers
1710 * previously done by the verifier are invalidated and must be
1711 * performed again, if the helper is used in combination with
1712 * direct packet access.
1714 * 0 on success, or a negative error in case of failure.
1716 * u64 bpf_get_current_pid_tgid(void)
1718 * A 64-bit integer containing the current tgid and pid, and
1720 * *current_task*\ **->tgid << 32 \|**
1721 * *current_task*\ **->pid**.
1723 * u64 bpf_get_current_uid_gid(void)
1725 * A 64-bit integer containing the current GID and UID, and
1726 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1728 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1730 * Copy the **comm** attribute of the current task into *buf* of
1731 * *size_of_buf*. The **comm** attribute contains the name of
1732 * the executable (excluding the path) for the current task. The
1733 * *size_of_buf* must be strictly positive. On success, the
1734 * helper makes sure that the *buf* is NUL-terminated. On failure,
1735 * it is filled with zeroes.
1737 * 0 on success, or a negative error in case of failure.
1739 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1741 * Retrieve the classid for the current task, i.e. for the net_cls
1742 * cgroup to which *skb* belongs.
1744 * This helper can be used on TC egress path, but not on ingress.
1746 * The net_cls cgroup provides an interface to tag network packets
1747 * based on a user-provided identifier for all traffic coming from
1748 * the tasks belonging to the related cgroup. See also the related
1749 * kernel documentation, available from the Linux sources in file
1750 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1752 * The Linux kernel has two versions for cgroups: there are
1753 * cgroups v1 and cgroups v2. Both are available to users, who can
1754 * use a mixture of them, but note that the net_cls cgroup is for
1755 * cgroup v1 only. This makes it incompatible with BPF programs
1756 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1757 * only hold data for one version of cgroups at a time).
1759 * This helper is only available is the kernel was compiled with
1760 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1761 * "**y**" or to "**m**".
1763 * The classid, or 0 for the default unconfigured classid.
1765 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1767 * Push a *vlan_tci* (VLAN tag control information) of protocol
1768 * *vlan_proto* to the packet associated to *skb*, then update
1769 * the checksum. Note that if *vlan_proto* is different from
1770 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1771 * be **ETH_P_8021Q**.
1773 * A call to this helper is susceptible to change the underlying
1774 * packet buffer. Therefore, at load time, all checks on pointers
1775 * previously done by the verifier are invalidated and must be
1776 * performed again, if the helper is used in combination with
1777 * direct packet access.
1779 * 0 on success, or a negative error in case of failure.
1781 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1783 * Pop a VLAN header from the packet associated to *skb*.
1785 * A call to this helper is susceptible to change the underlying
1786 * packet buffer. Therefore, at load time, all checks on pointers
1787 * previously done by the verifier are invalidated and must be
1788 * performed again, if the helper is used in combination with
1789 * direct packet access.
1791 * 0 on success, or a negative error in case of failure.
1793 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1795 * Get tunnel metadata. This helper takes a pointer *key* to an
1796 * empty **struct bpf_tunnel_key** of **size**, that will be
1797 * filled with tunnel metadata for the packet associated to *skb*.
1798 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1799 * indicates that the tunnel is based on IPv6 protocol instead of
1802 * The **struct bpf_tunnel_key** is an object that generalizes the
1803 * principal parameters used by various tunneling protocols into a
1804 * single struct. This way, it can be used to easily make a
1805 * decision based on the contents of the encapsulation header,
1806 * "summarized" in this struct. In particular, it holds the IP
1807 * address of the remote end (IPv4 or IPv6, depending on the case)
1808 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1809 * this struct exposes the *key*\ **->tunnel_id**, which is
1810 * generally mapped to a VNI (Virtual Network Identifier), making
1811 * it programmable together with the **bpf_skb_set_tunnel_key**\
1814 * Let's imagine that the following code is part of a program
1815 * attached to the TC ingress interface, on one end of a GRE
1816 * tunnel, and is supposed to filter out all messages coming from
1817 * remote ends with IPv4 address other than 10.0.0.1:
1822 * struct bpf_tunnel_key key = {};
1824 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1826 * return TC_ACT_SHOT; // drop packet
1828 * if (key.remote_ipv4 != 0x0a000001)
1829 * return TC_ACT_SHOT; // drop packet
1831 * return TC_ACT_OK; // accept packet
1833 * This interface can also be used with all encapsulation devices
1834 * that can operate in "collect metadata" mode: instead of having
1835 * one network device per specific configuration, the "collect
1836 * metadata" mode only requires a single device where the
1837 * configuration can be extracted from this helper.
1839 * This can be used together with various tunnels such as VXLan,
1840 * Geneve, GRE or IP in IP (IPIP).
1842 * 0 on success, or a negative error in case of failure.
1844 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1846 * Populate tunnel metadata for packet associated to *skb.* The
1847 * tunnel metadata is set to the contents of *key*, of *size*. The
1848 * *flags* can be set to a combination of the following values:
1850 * **BPF_F_TUNINFO_IPV6**
1851 * Indicate that the tunnel is based on IPv6 protocol
1853 * **BPF_F_ZERO_CSUM_TX**
1854 * For IPv4 packets, add a flag to tunnel metadata
1855 * indicating that checksum computation should be skipped
1856 * and checksum set to zeroes.
1857 * **BPF_F_DONT_FRAGMENT**
1858 * Add a flag to tunnel metadata indicating that the
1859 * packet should not be fragmented.
1860 * **BPF_F_SEQ_NUMBER**
1861 * Add a flag to tunnel metadata indicating that a
1862 * sequence number should be added to tunnel header before
1863 * sending the packet. This flag was added for GRE
1864 * encapsulation, but might be used with other protocols
1865 * as well in the future.
1867 * Here is a typical usage on the transmit path:
1871 * struct bpf_tunnel_key key;
1873 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1874 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1876 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1877 * helper for additional information.
1879 * 0 on success, or a negative error in case of failure.
1881 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1883 * Read the value of a perf event counter. This helper relies on a
1884 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1885 * the perf event counter is selected when *map* is updated with
1886 * perf event file descriptors. The *map* is an array whose size
1887 * is the number of available CPUs, and each cell contains a value
1888 * relative to one CPU. The value to retrieve is indicated by
1889 * *flags*, that contains the index of the CPU to look up, masked
1890 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1891 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1892 * current CPU should be retrieved.
1894 * Note that before Linux 4.13, only hardware perf event can be
1897 * Also, be aware that the newer helper
1898 * **bpf_perf_event_read_value**\ () is recommended over
1899 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1900 * quirks where error and counter value are used as a return code
1901 * (which is wrong to do since ranges may overlap). This issue is
1902 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1903 * time provides more features over the **bpf_perf_event_read**\
1904 * () interface. Please refer to the description of
1905 * **bpf_perf_event_read_value**\ () for details.
1907 * The value of the perf event counter read from the map, or a
1908 * negative error code in case of failure.
1910 * long bpf_redirect(u32 ifindex, u64 flags)
1912 * Redirect the packet to another net device of index *ifindex*.
1913 * This helper is somewhat similar to **bpf_clone_redirect**\
1914 * (), except that the packet is not cloned, which provides
1915 * increased performance.
1917 * Except for XDP, both ingress and egress interfaces can be used
1918 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1919 * to make the distinction (ingress path is selected if the flag
1920 * is present, egress path otherwise). Currently, XDP only
1921 * supports redirection to the egress interface, and accepts no
1924 * The same effect can also be attained with the more generic
1925 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1926 * redirect target instead of providing it directly to the helper.
1928 * For XDP, the helper returns **XDP_REDIRECT** on success or
1929 * **XDP_ABORTED** on error. For other program types, the values
1930 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1933 * u32 bpf_get_route_realm(struct sk_buff *skb)
1935 * Retrieve the realm or the route, that is to say the
1936 * **tclassid** field of the destination for the *skb*. The
1937 * identifier retrieved is a user-provided tag, similar to the
1938 * one used with the net_cls cgroup (see description for
1939 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1940 * held by a route (a destination entry), not by a task.
1942 * Retrieving this identifier works with the clsact TC egress hook
1943 * (see also **tc-bpf(8)**), or alternatively on conventional
1944 * classful egress qdiscs, but not on TC ingress path. In case of
1945 * clsact TC egress hook, this has the advantage that, internally,
1946 * the destination entry has not been dropped yet in the transmit
1947 * path. Therefore, the destination entry does not need to be
1948 * artificially held via **netif_keep_dst**\ () for a classful
1949 * qdisc until the *skb* is freed.
1951 * This helper is available only if the kernel was compiled with
1952 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1954 * The realm of the route for the packet associated to *skb*, or 0
1955 * if none was found.
1957 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1959 * Write raw *data* blob into a special BPF perf event held by
1960 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1961 * event must have the following attributes: **PERF_SAMPLE_RAW**
1962 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1963 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1965 * The *flags* are used to indicate the index in *map* for which
1966 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1967 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1968 * to indicate that the index of the current CPU core should be
1971 * The value to write, of *size*, is passed through eBPF stack and
1972 * pointed by *data*.
1974 * The context of the program *ctx* needs also be passed to the
1977 * On user space, a program willing to read the values needs to
1978 * call **perf_event_open**\ () on the perf event (either for
1979 * one or for all CPUs) and to store the file descriptor into the
1980 * *map*. This must be done before the eBPF program can send data
1981 * into it. An example is available in file
1982 * *samples/bpf/trace_output_user.c* in the Linux kernel source
1983 * tree (the eBPF program counterpart is in
1984 * *samples/bpf/trace_output_kern.c*).
1986 * **bpf_perf_event_output**\ () achieves better performance
1987 * than **bpf_trace_printk**\ () for sharing data with user
1988 * space, and is much better suitable for streaming data from eBPF
1991 * Note that this helper is not restricted to tracing use cases
1992 * and can be used with programs attached to TC or XDP as well,
1993 * where it allows for passing data to user space listeners. Data
1996 * * Only custom structs,
1997 * * Only the packet payload, or
1998 * * A combination of both.
2000 * 0 on success, or a negative error in case of failure.
2002 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2004 * This helper was provided as an easy way to load data from a
2005 * packet. It can be used to load *len* bytes from *offset* from
2006 * the packet associated to *skb*, into the buffer pointed by
2009 * Since Linux 4.7, usage of this helper has mostly been replaced
2010 * by "direct packet access", enabling packet data to be
2011 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2012 * pointing respectively to the first byte of packet data and to
2013 * the byte after the last byte of packet data. However, it
2014 * remains useful if one wishes to read large quantities of data
2015 * at once from a packet into the eBPF stack.
2017 * 0 on success, or a negative error in case of failure.
2019 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2021 * Walk a user or a kernel stack and return its id. To achieve
2022 * this, the helper needs *ctx*, which is a pointer to the context
2023 * on which the tracing program is executed, and a pointer to a
2024 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2026 * The last argument, *flags*, holds the number of stack frames to
2027 * skip (from 0 to 255), masked with
2028 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2029 * a combination of the following flags:
2031 * **BPF_F_USER_STACK**
2032 * Collect a user space stack instead of a kernel stack.
2033 * **BPF_F_FAST_STACK_CMP**
2034 * Compare stacks by hash only.
2035 * **BPF_F_REUSE_STACKID**
2036 * If two different stacks hash into the same *stackid*,
2037 * discard the old one.
2039 * The stack id retrieved is a 32 bit long integer handle which
2040 * can be further combined with other data (including other stack
2041 * ids) and used as a key into maps. This can be useful for
2042 * generating a variety of graphs (such as flame graphs or off-cpu
2045 * For walking a stack, this helper is an improvement over
2046 * **bpf_probe_read**\ (), which can be used with unrolled loops
2047 * but is not efficient and consumes a lot of eBPF instructions.
2048 * Instead, **bpf_get_stackid**\ () can collect up to
2049 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2050 * this limit can be controlled with the **sysctl** program, and
2051 * that it should be manually increased in order to profile long
2052 * user stacks (such as stacks for Java programs). To do so, use:
2056 * # sysctl kernel.perf_event_max_stack=<new value>
2058 * The positive or null stack id on success, or a negative error
2059 * in case of failure.
2061 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2063 * Compute a checksum difference, from the raw buffer pointed by
2064 * *from*, of length *from_size* (that must be a multiple of 4),
2065 * towards the raw buffer pointed by *to*, of size *to_size*
2066 * (same remark). An optional *seed* can be added to the value
2067 * (this can be cascaded, the seed may come from a previous call
2070 * This is flexible enough to be used in several ways:
2072 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2073 * checksum, it can be used when pushing new data.
2074 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2075 * checksum, it can be used when removing data from a packet.
2076 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2077 * can be used to compute a diff. Note that *from_size* and
2078 * *to_size* do not need to be equal.
2080 * This helper can be used in combination with
2081 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2082 * which one can feed in the difference computed with
2083 * **bpf_csum_diff**\ ().
2085 * The checksum result, or a negative error code in case of
2088 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2090 * Retrieve tunnel options metadata for the packet associated to
2091 * *skb*, and store the raw tunnel option data to the buffer *opt*
2094 * This helper can be used with encapsulation devices that can
2095 * operate in "collect metadata" mode (please refer to the related
2096 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2097 * more details). A particular example where this can be used is
2098 * in combination with the Geneve encapsulation protocol, where it
2099 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2100 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2101 * the eBPF program. This allows for full customization of these
2104 * The size of the option data retrieved.
2106 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2108 * Set tunnel options metadata for the packet associated to *skb*
2109 * to the option data contained in the raw buffer *opt* of *size*.
2111 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2112 * helper for additional information.
2114 * 0 on success, or a negative error in case of failure.
2116 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2118 * Change the protocol of the *skb* to *proto*. Currently
2119 * supported are transition from IPv4 to IPv6, and from IPv6 to
2120 * IPv4. The helper takes care of the groundwork for the
2121 * transition, including resizing the socket buffer. The eBPF
2122 * program is expected to fill the new headers, if any, via
2123 * **skb_store_bytes**\ () and to recompute the checksums with
2124 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2125 * (). The main case for this helper is to perform NAT64
2126 * operations out of an eBPF program.
2128 * Internally, the GSO type is marked as dodgy so that headers are
2129 * checked and segments are recalculated by the GSO/GRO engine.
2130 * The size for GSO target is adapted as well.
2132 * All values for *flags* are reserved for future usage, and must
2135 * A call to this helper is susceptible to change the underlying
2136 * packet buffer. Therefore, at load time, all checks on pointers
2137 * previously done by the verifier are invalidated and must be
2138 * performed again, if the helper is used in combination with
2139 * direct packet access.
2141 * 0 on success, or a negative error in case of failure.
2143 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2145 * Change the packet type for the packet associated to *skb*. This
2146 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2147 * the eBPF program does not have a write access to *skb*\
2148 * **->pkt_type** beside this helper. Using a helper here allows
2149 * for graceful handling of errors.
2151 * The major use case is to change incoming *skb*s to
2152 * **PACKET_HOST** in a programmatic way instead of having to
2153 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2156 * Note that *type* only allows certain values. At this time, they
2161 * **PACKET_BROADCAST**
2162 * Send packet to all.
2163 * **PACKET_MULTICAST**
2164 * Send packet to group.
2165 * **PACKET_OTHERHOST**
2166 * Send packet to someone else.
2168 * 0 on success, or a negative error in case of failure.
2170 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2172 * Check whether *skb* is a descendant of the cgroup2 held by
2173 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2175 * The return value depends on the result of the test, and can be:
2177 * * 0, if the *skb* failed the cgroup2 descendant test.
2178 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2179 * * A negative error code, if an error occurred.
2181 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2183 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2184 * not set, in particular if the hash was cleared due to mangling,
2185 * recompute this hash. Later accesses to the hash can be done
2186 * directly with *skb*\ **->hash**.
2188 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2189 * prototype with **bpf_skb_change_proto**\ (), or calling
2190 * **bpf_skb_store_bytes**\ () with the
2191 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2192 * the hash and to trigger a new computation for the next call to
2193 * **bpf_get_hash_recalc**\ ().
2197 * u64 bpf_get_current_task(void)
2199 * A pointer to the current task struct.
2201 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2203 * Attempt in a safe way to write *len* bytes from the buffer
2204 * *src* to *dst* in memory. It only works for threads that are in
2205 * user context, and *dst* must be a valid user space address.
2207 * This helper should not be used to implement any kind of
2208 * security mechanism because of TOC-TOU attacks, but rather to
2209 * debug, divert, and manipulate execution of semi-cooperative
2212 * Keep in mind that this feature is meant for experiments, and it
2213 * has a risk of crashing the system and running programs.
2214 * Therefore, when an eBPF program using this helper is attached,
2215 * a warning including PID and process name is printed to kernel
2218 * 0 on success, or a negative error in case of failure.
2220 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2222 * Check whether the probe is being run is the context of a given
2223 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2224 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2226 * The return value depends on the result of the test, and can be:
2228 * * 0, if current task belongs to the cgroup2.
2229 * * 1, if current task does not belong to the cgroup2.
2230 * * A negative error code, if an error occurred.
2232 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2234 * Resize (trim or grow) the packet associated to *skb* to the
2235 * new *len*. The *flags* are reserved for future usage, and must
2238 * The basic idea is that the helper performs the needed work to
2239 * change the size of the packet, then the eBPF program rewrites
2240 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2241 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2242 * and others. This helper is a slow path utility intended for
2243 * replies with control messages. And because it is targeted for
2244 * slow path, the helper itself can afford to be slow: it
2245 * implicitly linearizes, unclones and drops offloads from the
2248 * A call to this helper is susceptible to change the underlying
2249 * packet buffer. Therefore, at load time, all checks on pointers
2250 * previously done by the verifier are invalidated and must be
2251 * performed again, if the helper is used in combination with
2252 * direct packet access.
2254 * 0 on success, or a negative error in case of failure.
2256 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2258 * Pull in non-linear data in case the *skb* is non-linear and not
2259 * all of *len* are part of the linear section. Make *len* bytes
2260 * from *skb* readable and writable. If a zero value is passed for
2261 * *len*, then the whole length of the *skb* is pulled.
2263 * This helper is only needed for reading and writing with direct
2266 * For direct packet access, testing that offsets to access
2267 * are within packet boundaries (test on *skb*\ **->data_end**) is
2268 * susceptible to fail if offsets are invalid, or if the requested
2269 * data is in non-linear parts of the *skb*. On failure the
2270 * program can just bail out, or in the case of a non-linear
2271 * buffer, use a helper to make the data available. The
2272 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2273 * the data. Another one consists in using **bpf_skb_pull_data**
2274 * to pull in once the non-linear parts, then retesting and
2275 * eventually access the data.
2277 * At the same time, this also makes sure the *skb* is uncloned,
2278 * which is a necessary condition for direct write. As this needs
2279 * to be an invariant for the write part only, the verifier
2280 * detects writes and adds a prologue that is calling
2281 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2282 * the very beginning in case it is indeed cloned.
2284 * A call to this helper is susceptible to change the underlying
2285 * packet buffer. Therefore, at load time, all checks on pointers
2286 * previously done by the verifier are invalidated and must be
2287 * performed again, if the helper is used in combination with
2288 * direct packet access.
2290 * 0 on success, or a negative error in case of failure.
2292 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2294 * Add the checksum *csum* into *skb*\ **->csum** in case the
2295 * driver has supplied a checksum for the entire packet into that
2296 * field. Return an error otherwise. This helper is intended to be
2297 * used in combination with **bpf_csum_diff**\ (), in particular
2298 * when the checksum needs to be updated after data has been
2299 * written into the packet through direct packet access.
2301 * The checksum on success, or a negative error code in case of
2304 * void bpf_set_hash_invalid(struct sk_buff *skb)
2306 * Invalidate the current *skb*\ **->hash**. It can be used after
2307 * mangling on headers through direct packet access, in order to
2308 * indicate that the hash is outdated and to trigger a
2309 * recalculation the next time the kernel tries to access this
2310 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2312 * long bpf_get_numa_node_id(void)
2314 * Return the id of the current NUMA node. The primary use case
2315 * for this helper is the selection of sockets for the local NUMA
2316 * node, when the program is attached to sockets using the
2317 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2318 * but the helper is also available to other eBPF program types,
2319 * similarly to **bpf_get_smp_processor_id**\ ().
2321 * The id of current NUMA node.
2323 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2325 * Grows headroom of packet associated to *skb* and adjusts the
2326 * offset of the MAC header accordingly, adding *len* bytes of
2327 * space. It automatically extends and reallocates memory as
2330 * This helper can be used on a layer 3 *skb* to push a MAC header
2331 * for redirection into a layer 2 device.
2333 * All values for *flags* are reserved for future usage, and must
2336 * A call to this helper is susceptible to change the underlying
2337 * packet buffer. Therefore, at load time, all checks on pointers
2338 * previously done by the verifier are invalidated and must be
2339 * performed again, if the helper is used in combination with
2340 * direct packet access.
2342 * 0 on success, or a negative error in case of failure.
2344 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2346 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2347 * it is possible to use a negative value for *delta*. This helper
2348 * can be used to prepare the packet for pushing or popping
2351 * A call to this helper is susceptible to change the underlying
2352 * packet buffer. Therefore, at load time, all checks on pointers
2353 * previously done by the verifier are invalidated and must be
2354 * performed again, if the helper is used in combination with
2355 * direct packet access.
2357 * 0 on success, or a negative error in case of failure.
2359 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2361 * Copy a NUL terminated string from an unsafe kernel address
2362 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2365 * Generally, use **bpf_probe_read_user_str**\ () or
2366 * **bpf_probe_read_kernel_str**\ () instead.
2368 * On success, the strictly positive length of the string,
2369 * including the trailing NUL character. On error, a negative
2372 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2374 * If the **struct sk_buff** pointed by *skb* has a known socket,
2375 * retrieve the cookie (generated by the kernel) of this socket.
2376 * If no cookie has been set yet, generate a new cookie. Once
2377 * generated, the socket cookie remains stable for the life of the
2378 * socket. This helper can be useful for monitoring per socket
2379 * networking traffic statistics as it provides a global socket
2380 * identifier that can be assumed unique.
2382 * A 8-byte long unique number on success, or 0 if the socket
2383 * field is missing inside *skb*.
2385 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2387 * Equivalent to bpf_get_socket_cookie() helper that accepts
2388 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2390 * A 8-byte long unique number.
2392 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2394 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2395 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2397 * A 8-byte long unique number.
2399 * u64 bpf_get_socket_cookie(struct sock *sk)
2401 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2402 * *sk*, but gets socket from a BTF **struct sock**. This helper
2403 * also works for sleepable programs.
2405 * A 8-byte long unique number or 0 if *sk* is NULL.
2407 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2409 * The owner UID of the socket associated to *skb*. If the socket
2410 * is **NULL**, or if it is not a full socket (i.e. if it is a
2411 * time-wait or a request socket instead), **overflowuid** value
2412 * is returned (note that **overflowuid** might also be the actual
2413 * UID value for the socket).
2415 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2417 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2422 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2424 * Emulate a call to **setsockopt()** on the socket associated to
2425 * *bpf_socket*, which must be a full socket. The *level* at
2426 * which the option resides and the name *optname* of the option
2427 * must be specified, see **setsockopt(2)** for more information.
2428 * The option value of length *optlen* is pointed by *optval*.
2430 * *bpf_socket* should be one of the following:
2432 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2433 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2434 * and **BPF_CGROUP_INET6_CONNECT**.
2436 * This helper actually implements a subset of **setsockopt()**.
2437 * It supports the following *level*\ s:
2439 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2440 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2441 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2442 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2443 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2444 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2445 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2446 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2447 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2448 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2449 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2451 * 0 on success, or a negative error in case of failure.
2453 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2455 * Grow or shrink the room for data in the packet associated to
2456 * *skb* by *len_diff*, and according to the selected *mode*.
2458 * By default, the helper will reset any offloaded checksum
2459 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2460 * by the following flag:
2462 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2463 * checksum data of the skb to CHECKSUM_NONE.
2465 * There are two supported modes at this time:
2467 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2468 * (room space is added or removed below the layer 2 header).
2470 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2471 * (room space is added or removed below the layer 3 header).
2473 * The following flags are supported at this time:
2475 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2476 * Adjusting mss in this way is not allowed for datagrams.
2478 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2479 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2480 * Any new space is reserved to hold a tunnel header.
2481 * Configure skb offsets and other fields accordingly.
2483 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2484 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2485 * Use with ENCAP_L3 flags to further specify the tunnel type.
2487 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2488 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2489 * type; *len* is the length of the inner MAC header.
2491 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2492 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2493 * L2 type as Ethernet.
2495 * A call to this helper is susceptible to change the underlying
2496 * packet buffer. Therefore, at load time, all checks on pointers
2497 * previously done by the verifier are invalidated and must be
2498 * performed again, if the helper is used in combination with
2499 * direct packet access.
2501 * 0 on success, or a negative error in case of failure.
2503 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2505 * Redirect the packet to the endpoint referenced by *map* at
2506 * index *key*. Depending on its type, this *map* can contain
2507 * references to net devices (for forwarding packets through other
2508 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2509 * but this is only implemented for native XDP (with driver
2510 * support) as of this writing).
2512 * The lower two bits of *flags* are used as the return code if
2513 * the map lookup fails. This is so that the return value can be
2514 * one of the XDP program return codes up to **XDP_TX**, as chosen
2515 * by the caller. Any higher bits in the *flags* argument must be
2518 * See also **bpf_redirect**\ (), which only supports redirecting
2519 * to an ifindex, but doesn't require a map to do so.
2521 * **XDP_REDIRECT** on success, or the value of the two lower bits
2522 * of the *flags* argument on error.
2524 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2526 * Redirect the packet to the socket referenced by *map* (of type
2527 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2528 * egress interfaces can be used for redirection. The
2529 * **BPF_F_INGRESS** value in *flags* is used to make the
2530 * distinction (ingress path is selected if the flag is present,
2531 * egress path otherwise). This is the only flag supported for now.
2533 * **SK_PASS** on success, or **SK_DROP** on error.
2535 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2537 * Add an entry to, or update a *map* referencing sockets. The
2538 * *skops* is used as a new value for the entry associated to
2539 * *key*. *flags* is one of:
2542 * The entry for *key* must not exist in the map.
2544 * The entry for *key* must already exist in the map.
2546 * No condition on the existence of the entry for *key*.
2548 * If the *map* has eBPF programs (parser and verdict), those will
2549 * be inherited by the socket being added. If the socket is
2550 * already attached to eBPF programs, this results in an error.
2552 * 0 on success, or a negative error in case of failure.
2554 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2556 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2557 * *delta* (which can be positive or negative). Note that this
2558 * operation modifies the address stored in *xdp_md*\ **->data**,
2559 * so the latter must be loaded only after the helper has been
2562 * The use of *xdp_md*\ **->data_meta** is optional and programs
2563 * are not required to use it. The rationale is that when the
2564 * packet is processed with XDP (e.g. as DoS filter), it is
2565 * possible to push further meta data along with it before passing
2566 * to the stack, and to give the guarantee that an ingress eBPF
2567 * program attached as a TC classifier on the same device can pick
2568 * this up for further post-processing. Since TC works with socket
2569 * buffers, it remains possible to set from XDP the **mark** or
2570 * **priority** pointers, or other pointers for the socket buffer.
2571 * Having this scratch space generic and programmable allows for
2572 * more flexibility as the user is free to store whatever meta
2575 * A call to this helper is susceptible to change the underlying
2576 * packet buffer. Therefore, at load time, all checks on pointers
2577 * previously done by the verifier are invalidated and must be
2578 * performed again, if the helper is used in combination with
2579 * direct packet access.
2581 * 0 on success, or a negative error in case of failure.
2583 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2585 * Read the value of a perf event counter, and store it into *buf*
2586 * of size *buf_size*. This helper relies on a *map* of type
2587 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2588 * counter is selected when *map* is updated with perf event file
2589 * descriptors. The *map* is an array whose size is the number of
2590 * available CPUs, and each cell contains a value relative to one
2591 * CPU. The value to retrieve is indicated by *flags*, that
2592 * contains the index of the CPU to look up, masked with
2593 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2594 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2595 * current CPU should be retrieved.
2597 * This helper behaves in a way close to
2598 * **bpf_perf_event_read**\ () helper, save that instead of
2599 * just returning the value observed, it fills the *buf*
2600 * structure. This allows for additional data to be retrieved: in
2601 * particular, the enabled and running times (in *buf*\
2602 * **->enabled** and *buf*\ **->running**, respectively) are
2603 * copied. In general, **bpf_perf_event_read_value**\ () is
2604 * recommended over **bpf_perf_event_read**\ (), which has some
2605 * ABI issues and provides fewer functionalities.
2607 * These values are interesting, because hardware PMU (Performance
2608 * Monitoring Unit) counters are limited resources. When there are
2609 * more PMU based perf events opened than available counters,
2610 * kernel will multiplex these events so each event gets certain
2611 * percentage (but not all) of the PMU time. In case that
2612 * multiplexing happens, the number of samples or counter value
2613 * will not reflect the case compared to when no multiplexing
2614 * occurs. This makes comparison between different runs difficult.
2615 * Typically, the counter value should be normalized before
2616 * comparing to other experiments. The usual normalization is done
2621 * normalized_counter = counter * t_enabled / t_running
2623 * Where t_enabled is the time enabled for event and t_running is
2624 * the time running for event since last normalization. The
2625 * enabled and running times are accumulated since the perf event
2626 * open. To achieve scaling factor between two invocations of an
2627 * eBPF program, users can use CPU id as the key (which is
2628 * typical for perf array usage model) to remember the previous
2629 * value and do the calculation inside the eBPF program.
2631 * 0 on success, or a negative error in case of failure.
2633 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2635 * For en eBPF program attached to a perf event, retrieve the
2636 * value of the event counter associated to *ctx* and store it in
2637 * the structure pointed by *buf* and of size *buf_size*. Enabled
2638 * and running times are also stored in the structure (see
2639 * description of helper **bpf_perf_event_read_value**\ () for
2642 * 0 on success, or a negative error in case of failure.
2644 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2646 * Emulate a call to **getsockopt()** on the socket associated to
2647 * *bpf_socket*, which must be a full socket. The *level* at
2648 * which the option resides and the name *optname* of the option
2649 * must be specified, see **getsockopt(2)** for more information.
2650 * The retrieved value is stored in the structure pointed by
2651 * *opval* and of length *optlen*.
2653 * *bpf_socket* should be one of the following:
2655 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2656 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2657 * and **BPF_CGROUP_INET6_CONNECT**.
2659 * This helper actually implements a subset of **getsockopt()**.
2660 * It supports the following *level*\ s:
2662 * * **IPPROTO_TCP**, which supports *optname*
2663 * **TCP_CONGESTION**.
2664 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2665 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2667 * 0 on success, or a negative error in case of failure.
2669 * long bpf_override_return(struct pt_regs *regs, u64 rc)
2671 * Used for error injection, this helper uses kprobes to override
2672 * the return value of the probed function, and to set it to *rc*.
2673 * The first argument is the context *regs* on which the kprobe
2676 * This helper works by setting the PC (program counter)
2677 * to an override function which is run in place of the original
2678 * probed function. This means the probed function is not run at
2679 * all. The replacement function just returns with the required
2682 * This helper has security implications, and thus is subject to
2683 * restrictions. It is only available if the kernel was compiled
2684 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2685 * option, and in this case it only works on functions tagged with
2686 * **ALLOW_ERROR_INJECTION** in the kernel code.
2688 * Also, the helper is only available for the architectures having
2689 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2690 * x86 architecture is the only one to support this feature.
2694 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2696 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2697 * for the full TCP socket associated to *bpf_sock_ops* to
2700 * The primary use of this field is to determine if there should
2701 * be calls to eBPF programs of type
2702 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2703 * code. A program of the same type can change its value, per
2704 * connection and as necessary, when the connection is
2705 * established. This field is directly accessible for reading, but
2706 * this helper must be used for updates in order to return an
2707 * error if an eBPF program tries to set a callback that is not
2708 * supported in the current kernel.
2710 * *argval* is a flag array which can combine these flags:
2712 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2713 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2714 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2715 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2717 * Therefore, this function can be used to clear a callback flag by
2718 * setting the appropriate bit to zero. e.g. to disable the RTO
2721 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
2722 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2724 * Here are some examples of where one could call such eBPF
2728 * * When a packet is retransmitted.
2729 * * When the connection terminates.
2730 * * When a packet is sent.
2731 * * When a packet is received.
2733 * Code **-EINVAL** if the socket is not a full TCP socket;
2734 * otherwise, a positive number containing the bits that could not
2735 * be set is returned (which comes down to 0 if all bits were set
2738 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2740 * This helper is used in programs implementing policies at the
2741 * socket level. If the message *msg* is allowed to pass (i.e. if
2742 * the verdict eBPF program returns **SK_PASS**), redirect it to
2743 * the socket referenced by *map* (of type
2744 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2745 * egress interfaces can be used for redirection. The
2746 * **BPF_F_INGRESS** value in *flags* is used to make the
2747 * distinction (ingress path is selected if the flag is present,
2748 * egress path otherwise). This is the only flag supported for now.
2750 * **SK_PASS** on success, or **SK_DROP** on error.
2752 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2754 * For socket policies, apply the verdict of the eBPF program to
2755 * the next *bytes* (number of bytes) of message *msg*.
2757 * For example, this helper can be used in the following cases:
2759 * * A single **sendmsg**\ () or **sendfile**\ () system call
2760 * contains multiple logical messages that the eBPF program is
2761 * supposed to read and for which it should apply a verdict.
2762 * * An eBPF program only cares to read the first *bytes* of a
2763 * *msg*. If the message has a large payload, then setting up
2764 * and calling the eBPF program repeatedly for all bytes, even
2765 * though the verdict is already known, would create unnecessary
2768 * When called from within an eBPF program, the helper sets a
2769 * counter internal to the BPF infrastructure, that is used to
2770 * apply the last verdict to the next *bytes*. If *bytes* is
2771 * smaller than the current data being processed from a
2772 * **sendmsg**\ () or **sendfile**\ () system call, the first
2773 * *bytes* will be sent and the eBPF program will be re-run with
2774 * the pointer for start of data pointing to byte number *bytes*
2775 * **+ 1**. If *bytes* is larger than the current data being
2776 * processed, then the eBPF verdict will be applied to multiple
2777 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2780 * Note that if a socket closes with the internal counter holding
2781 * a non-zero value, this is not a problem because data is not
2782 * being buffered for *bytes* and is sent as it is received.
2786 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2788 * For socket policies, prevent the execution of the verdict eBPF
2789 * program for message *msg* until *bytes* (byte number) have been
2792 * This can be used when one needs a specific number of bytes
2793 * before a verdict can be assigned, even if the data spans
2794 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2795 * case would be a user calling **sendmsg**\ () repeatedly with
2796 * 1-byte long message segments. Obviously, this is bad for
2797 * performance, but it is still valid. If the eBPF program needs
2798 * *bytes* bytes to validate a header, this helper can be used to
2799 * prevent the eBPF program to be called again until *bytes* have
2804 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2806 * For socket policies, pull in non-linear data from user space
2807 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2808 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2811 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2812 * *msg* it can only parse data that the (**data**, **data_end**)
2813 * pointers have already consumed. For **sendmsg**\ () hooks this
2814 * is likely the first scatterlist element. But for calls relying
2815 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2816 * be the range (**0**, **0**) because the data is shared with
2817 * user space and by default the objective is to avoid allowing
2818 * user space to modify data while (or after) eBPF verdict is
2819 * being decided. This helper can be used to pull in data and to
2820 * set the start and end pointer to given values. Data will be
2821 * copied if necessary (i.e. if data was not linear and if start
2822 * and end pointers do not point to the same chunk).
2824 * A call to this helper is susceptible to change the underlying
2825 * packet buffer. Therefore, at load time, all checks on pointers
2826 * previously done by the verifier are invalidated and must be
2827 * performed again, if the helper is used in combination with
2828 * direct packet access.
2830 * All values for *flags* are reserved for future usage, and must
2833 * 0 on success, or a negative error in case of failure.
2835 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2837 * Bind the socket associated to *ctx* to the address pointed by
2838 * *addr*, of length *addr_len*. This allows for making outgoing
2839 * connection from the desired IP address, which can be useful for
2840 * example when all processes inside a cgroup should use one
2841 * single IP address on a host that has multiple IP configured.
2843 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2844 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2845 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2846 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2847 * behavior and lets the kernel efficiently pick up an unused
2848 * port as long as 4-tuple is unique. Passing non-zero port might
2849 * lead to degraded performance.
2851 * 0 on success, or a negative error in case of failure.
2853 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2855 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2856 * possible to both shrink and grow the packet tail.
2857 * Shrink done via *delta* being a negative integer.
2859 * A call to this helper is susceptible to change the underlying
2860 * packet buffer. Therefore, at load time, all checks on pointers
2861 * previously done by the verifier are invalidated and must be
2862 * performed again, if the helper is used in combination with
2863 * direct packet access.
2865 * 0 on success, or a negative error in case of failure.
2867 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2869 * Retrieve the XFRM state (IP transform framework, see also
2870 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2872 * The retrieved value is stored in the **struct bpf_xfrm_state**
2873 * pointed by *xfrm_state* and of length *size*.
2875 * All values for *flags* are reserved for future usage, and must
2878 * This helper is available only if the kernel was compiled with
2879 * **CONFIG_XFRM** configuration option.
2881 * 0 on success, or a negative error in case of failure.
2883 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2885 * Return a user or a kernel stack in bpf program provided buffer.
2886 * To achieve this, the helper needs *ctx*, which is a pointer
2887 * to the context on which the tracing program is executed.
2888 * To store the stacktrace, the bpf program provides *buf* with
2889 * a nonnegative *size*.
2891 * The last argument, *flags*, holds the number of stack frames to
2892 * skip (from 0 to 255), masked with
2893 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2894 * the following flags:
2896 * **BPF_F_USER_STACK**
2897 * Collect a user space stack instead of a kernel stack.
2898 * **BPF_F_USER_BUILD_ID**
2899 * Collect buildid+offset instead of ips for user stack,
2900 * only valid if **BPF_F_USER_STACK** is also specified.
2902 * **bpf_get_stack**\ () can collect up to
2903 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2904 * to sufficient large buffer size. Note that
2905 * this limit can be controlled with the **sysctl** program, and
2906 * that it should be manually increased in order to profile long
2907 * user stacks (such as stacks for Java programs). To do so, use:
2911 * # sysctl kernel.perf_event_max_stack=<new value>
2913 * A non-negative value equal to or less than *size* on success,
2914 * or a negative error in case of failure.
2916 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2918 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2919 * it provides an easy way to load *len* bytes from *offset*
2920 * from the packet associated to *skb*, into the buffer pointed
2921 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2922 * a fifth argument *start_header* exists in order to select a
2923 * base offset to start from. *start_header* can be one of:
2925 * **BPF_HDR_START_MAC**
2926 * Base offset to load data from is *skb*'s mac header.
2927 * **BPF_HDR_START_NET**
2928 * Base offset to load data from is *skb*'s network header.
2930 * In general, "direct packet access" is the preferred method to
2931 * access packet data, however, this helper is in particular useful
2932 * in socket filters where *skb*\ **->data** does not always point
2933 * to the start of the mac header and where "direct packet access"
2936 * 0 on success, or a negative error in case of failure.
2938 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2940 * Do FIB lookup in kernel tables using parameters in *params*.
2941 * If lookup is successful and result shows packet is to be
2942 * forwarded, the neighbor tables are searched for the nexthop.
2943 * If successful (ie., FIB lookup shows forwarding and nexthop
2944 * is resolved), the nexthop address is returned in ipv4_dst
2945 * or ipv6_dst based on family, smac is set to mac address of
2946 * egress device, dmac is set to nexthop mac address, rt_metric
2947 * is set to metric from route (IPv4/IPv6 only), and ifindex
2948 * is set to the device index of the nexthop from the FIB lookup.
2950 * *plen* argument is the size of the passed in struct.
2951 * *flags* argument can be a combination of one or more of the
2954 * **BPF_FIB_LOOKUP_DIRECT**
2955 * Do a direct table lookup vs full lookup using FIB
2957 * **BPF_FIB_LOOKUP_OUTPUT**
2958 * Perform lookup from an egress perspective (default is
2961 * *ctx* is either **struct xdp_md** for XDP programs or
2962 * **struct sk_buff** tc cls_act programs.
2964 * * < 0 if any input argument is invalid
2965 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2966 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2967 * packet is not forwarded or needs assist from full stack
2969 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
2970 * was exceeded and output params->mtu_result contains the MTU.
2972 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2974 * Add an entry to, or update a sockhash *map* referencing sockets.
2975 * The *skops* is used as a new value for the entry associated to
2976 * *key*. *flags* is one of:
2979 * The entry for *key* must not exist in the map.
2981 * The entry for *key* must already exist in the map.
2983 * No condition on the existence of the entry for *key*.
2985 * If the *map* has eBPF programs (parser and verdict), those will
2986 * be inherited by the socket being added. If the socket is
2987 * already attached to eBPF programs, this results in an error.
2989 * 0 on success, or a negative error in case of failure.
2991 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
2993 * This helper is used in programs implementing policies at the
2994 * socket level. If the message *msg* is allowed to pass (i.e. if
2995 * the verdict eBPF program returns **SK_PASS**), redirect it to
2996 * the socket referenced by *map* (of type
2997 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
2998 * egress interfaces can be used for redirection. The
2999 * **BPF_F_INGRESS** value in *flags* is used to make the
3000 * distinction (ingress path is selected if the flag is present,
3001 * egress path otherwise). This is the only flag supported for now.
3003 * **SK_PASS** on success, or **SK_DROP** on error.
3005 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3007 * This helper is used in programs implementing policies at the
3008 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3009 * if the verdict eBPF program returns **SK_PASS**), redirect it
3010 * to the socket referenced by *map* (of type
3011 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3012 * egress interfaces can be used for redirection. The
3013 * **BPF_F_INGRESS** value in *flags* is used to make the
3014 * distinction (ingress path is selected if the flag is present,
3015 * egress otherwise). This is the only flag supported for now.
3017 * **SK_PASS** on success, or **SK_DROP** on error.
3019 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3021 * Encapsulate the packet associated to *skb* within a Layer 3
3022 * protocol header. This header is provided in the buffer at
3023 * address *hdr*, with *len* its size in bytes. *type* indicates
3024 * the protocol of the header and can be one of:
3026 * **BPF_LWT_ENCAP_SEG6**
3027 * IPv6 encapsulation with Segment Routing Header
3028 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3029 * the IPv6 header is computed by the kernel.
3030 * **BPF_LWT_ENCAP_SEG6_INLINE**
3031 * Only works if *skb* contains an IPv6 packet. Insert a
3032 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3034 * **BPF_LWT_ENCAP_IP**
3035 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3036 * must be IPv4 or IPv6, followed by zero or more
3037 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3038 * total bytes in all prepended headers. Please note that
3039 * if **skb_is_gso**\ (*skb*) is true, no more than two
3040 * headers can be prepended, and the inner header, if
3041 * present, should be either GRE or UDP/GUE.
3043 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3044 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3045 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3046 * **BPF_PROG_TYPE_LWT_XMIT**.
3048 * A call to this helper is susceptible to change the underlying
3049 * packet buffer. Therefore, at load time, all checks on pointers
3050 * previously done by the verifier are invalidated and must be
3051 * performed again, if the helper is used in combination with
3052 * direct packet access.
3054 * 0 on success, or a negative error in case of failure.
3056 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3058 * Store *len* bytes from address *from* into the packet
3059 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3060 * inside the outermost IPv6 Segment Routing Header can be
3061 * modified through this helper.
3063 * A call to this helper is susceptible to change the underlying
3064 * packet buffer. Therefore, at load time, all checks on pointers
3065 * previously done by the verifier are invalidated and must be
3066 * performed again, if the helper is used in combination with
3067 * direct packet access.
3069 * 0 on success, or a negative error in case of failure.
3071 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3073 * Adjust the size allocated to TLVs in the outermost IPv6
3074 * Segment Routing Header contained in the packet associated to
3075 * *skb*, at position *offset* by *delta* bytes. Only offsets
3076 * after the segments are accepted. *delta* can be as well
3077 * positive (growing) as negative (shrinking).
3079 * A call to this helper is susceptible to change the underlying
3080 * packet buffer. Therefore, at load time, all checks on pointers
3081 * previously done by the verifier are invalidated and must be
3082 * performed again, if the helper is used in combination with
3083 * direct packet access.
3085 * 0 on success, or a negative error in case of failure.
3087 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3089 * Apply an IPv6 Segment Routing action of type *action* to the
3090 * packet associated to *skb*. Each action takes a parameter
3091 * contained at address *param*, and of length *param_len* bytes.
3092 * *action* can be one of:
3094 * **SEG6_LOCAL_ACTION_END_X**
3095 * End.X action: Endpoint with Layer-3 cross-connect.
3096 * Type of *param*: **struct in6_addr**.
3097 * **SEG6_LOCAL_ACTION_END_T**
3098 * End.T action: Endpoint with specific IPv6 table lookup.
3099 * Type of *param*: **int**.
3100 * **SEG6_LOCAL_ACTION_END_B6**
3101 * End.B6 action: Endpoint bound to an SRv6 policy.
3102 * Type of *param*: **struct ipv6_sr_hdr**.
3103 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3104 * End.B6.Encap action: Endpoint bound to an SRv6
3105 * encapsulation policy.
3106 * Type of *param*: **struct ipv6_sr_hdr**.
3108 * A call to this helper is susceptible to change the underlying
3109 * packet buffer. Therefore, at load time, all checks on pointers
3110 * previously done by the verifier are invalidated and must be
3111 * performed again, if the helper is used in combination with
3112 * direct packet access.
3114 * 0 on success, or a negative error in case of failure.
3116 * long bpf_rc_repeat(void *ctx)
3118 * This helper is used in programs implementing IR decoding, to
3119 * report a successfully decoded repeat key message. This delays
3120 * the generation of a key up event for previously generated
3123 * Some IR protocols like NEC have a special IR message for
3124 * repeating last button, for when a button is held down.
3126 * The *ctx* should point to the lirc sample as passed into
3129 * This helper is only available is the kernel was compiled with
3130 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3135 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3137 * This helper is used in programs implementing IR decoding, to
3138 * report a successfully decoded key press with *scancode*,
3139 * *toggle* value in the given *protocol*. The scancode will be
3140 * translated to a keycode using the rc keymap, and reported as
3141 * an input key down event. After a period a key up event is
3142 * generated. This period can be extended by calling either
3143 * **bpf_rc_keydown**\ () again with the same values, or calling
3144 * **bpf_rc_repeat**\ ().
3146 * Some protocols include a toggle bit, in case the button was
3147 * released and pressed again between consecutive scancodes.
3149 * The *ctx* should point to the lirc sample as passed into
3152 * The *protocol* is the decoded protocol number (see
3153 * **enum rc_proto** for some predefined values).
3155 * This helper is only available is the kernel was compiled with
3156 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3161 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3163 * Return the cgroup v2 id of the socket associated with the *skb*.
3164 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3165 * helper for cgroup v1 by providing a tag resp. identifier that
3166 * can be matched on or used for map lookups e.g. to implement
3167 * policy. The cgroup v2 id of a given path in the hierarchy is
3168 * exposed in user space through the f_handle API in order to get
3169 * to the same 64-bit id.
3171 * This helper can be used on TC egress path, but not on ingress,
3172 * and is available only if the kernel was compiled with the
3173 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3175 * The id is returned or 0 in case the id could not be retrieved.
3177 * u64 bpf_get_current_cgroup_id(void)
3179 * A 64-bit integer containing the current cgroup id based
3180 * on the cgroup within which the current task is running.
3182 * void *bpf_get_local_storage(void *map, u64 flags)
3184 * Get the pointer to the local storage area.
3185 * The type and the size of the local storage is defined
3186 * by the *map* argument.
3187 * The *flags* meaning is specific for each map type,
3188 * and has to be 0 for cgroup local storage.
3190 * Depending on the BPF program type, a local storage area
3191 * can be shared between multiple instances of the BPF program,
3192 * running simultaneously.
3194 * A user should care about the synchronization by himself.
3195 * For example, by using the **BPF_ATOMIC** instructions to alter
3198 * A pointer to the local storage area.
3200 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3202 * Select a **SO_REUSEPORT** socket from a
3203 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
3204 * It checks the selected socket is matching the incoming
3205 * request in the socket buffer.
3207 * 0 on success, or a negative error in case of failure.
3209 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3211 * Return id of cgroup v2 that is ancestor of cgroup associated
3212 * with the *skb* at the *ancestor_level*. The root cgroup is at
3213 * *ancestor_level* zero and each step down the hierarchy
3214 * increments the level. If *ancestor_level* == level of cgroup
3215 * associated with *skb*, then return value will be same as that
3216 * of **bpf_skb_cgroup_id**\ ().
3218 * The helper is useful to implement policies based on cgroups
3219 * that are upper in hierarchy than immediate cgroup associated
3222 * The format of returned id and helper limitations are same as in
3223 * **bpf_skb_cgroup_id**\ ().
3225 * The id is returned or 0 in case the id could not be retrieved.
3227 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3229 * Look for TCP socket matching *tuple*, optionally in a child
3230 * network namespace *netns*. The return value must be checked,
3231 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3233 * The *ctx* should point to the context of the program, such as
3234 * the skb or socket (depending on the hook in use). This is used
3235 * to determine the base network namespace for the lookup.
3237 * *tuple_size* must be one of:
3239 * **sizeof**\ (*tuple*\ **->ipv4**)
3240 * Look for an IPv4 socket.
3241 * **sizeof**\ (*tuple*\ **->ipv6**)
3242 * Look for an IPv6 socket.
3244 * If the *netns* is a negative signed 32-bit integer, then the
3245 * socket lookup table in the netns associated with the *ctx*
3246 * will be used. For the TC hooks, this is the netns of the device
3247 * in the skb. For socket hooks, this is the netns of the socket.
3248 * If *netns* is any other signed 32-bit value greater than or
3249 * equal to zero then it specifies the ID of the netns relative to
3250 * the netns associated with the *ctx*. *netns* values beyond the
3251 * range of 32-bit integers are reserved for future use.
3253 * All values for *flags* are reserved for future usage, and must
3256 * This helper is available only if the kernel was compiled with
3257 * **CONFIG_NET** configuration option.
3259 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3260 * For sockets with reuseport option, the **struct bpf_sock**
3261 * result is from *reuse*\ **->socks**\ [] using the hash of the
3264 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3266 * Look for UDP socket matching *tuple*, optionally in a child
3267 * network namespace *netns*. The return value must be checked,
3268 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3270 * The *ctx* should point to the context of the program, such as
3271 * the skb or socket (depending on the hook in use). This is used
3272 * to determine the base network namespace for the lookup.
3274 * *tuple_size* must be one of:
3276 * **sizeof**\ (*tuple*\ **->ipv4**)
3277 * Look for an IPv4 socket.
3278 * **sizeof**\ (*tuple*\ **->ipv6**)
3279 * Look for an IPv6 socket.
3281 * If the *netns* is a negative signed 32-bit integer, then the
3282 * socket lookup table in the netns associated with the *ctx*
3283 * will be used. For the TC hooks, this is the netns of the device
3284 * in the skb. For socket hooks, this is the netns of the socket.
3285 * If *netns* is any other signed 32-bit value greater than or
3286 * equal to zero then it specifies the ID of the netns relative to
3287 * the netns associated with the *ctx*. *netns* values beyond the
3288 * range of 32-bit integers are reserved for future use.
3290 * All values for *flags* are reserved for future usage, and must
3293 * This helper is available only if the kernel was compiled with
3294 * **CONFIG_NET** configuration option.
3296 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3297 * For sockets with reuseport option, the **struct bpf_sock**
3298 * result is from *reuse*\ **->socks**\ [] using the hash of the
3301 * long bpf_sk_release(void *sock)
3303 * Release the reference held by *sock*. *sock* must be a
3304 * non-**NULL** pointer that was returned from
3305 * **bpf_sk_lookup_xxx**\ ().
3307 * 0 on success, or a negative error in case of failure.
3309 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3311 * Push an element *value* in *map*. *flags* is one of:
3314 * If the queue/stack is full, the oldest element is
3315 * removed to make room for this.
3317 * 0 on success, or a negative error in case of failure.
3319 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3321 * Pop an element from *map*.
3323 * 0 on success, or a negative error in case of failure.
3325 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3327 * Get an element from *map* without removing it.
3329 * 0 on success, or a negative error in case of failure.
3331 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3333 * For socket policies, insert *len* bytes into *msg* at offset
3336 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3337 * *msg* it may want to insert metadata or options into the *msg*.
3338 * This can later be read and used by any of the lower layer BPF
3341 * This helper may fail if under memory pressure (a malloc
3342 * fails) in these cases BPF programs will get an appropriate
3343 * error and BPF programs will need to handle them.
3345 * 0 on success, or a negative error in case of failure.
3347 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3349 * Will remove *len* bytes from a *msg* starting at byte *start*.
3350 * This may result in **ENOMEM** errors under certain situations if
3351 * an allocation and copy are required due to a full ring buffer.
3352 * However, the helper will try to avoid doing the allocation
3353 * if possible. Other errors can occur if input parameters are
3354 * invalid either due to *start* byte not being valid part of *msg*
3355 * payload and/or *pop* value being to large.
3357 * 0 on success, or a negative error in case of failure.
3359 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3361 * This helper is used in programs implementing IR decoding, to
3362 * report a successfully decoded pointer movement.
3364 * The *ctx* should point to the lirc sample as passed into
3367 * This helper is only available is the kernel was compiled with
3368 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3373 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3375 * Acquire a spinlock represented by the pointer *lock*, which is
3376 * stored as part of a value of a map. Taking the lock allows to
3377 * safely update the rest of the fields in that value. The
3378 * spinlock can (and must) later be released with a call to
3379 * **bpf_spin_unlock**\ (\ *lock*\ ).
3381 * Spinlocks in BPF programs come with a number of restrictions
3384 * * **bpf_spin_lock** objects are only allowed inside maps of
3385 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3386 * list could be extended in the future).
3387 * * BTF description of the map is mandatory.
3388 * * The BPF program can take ONE lock at a time, since taking two
3389 * or more could cause dead locks.
3390 * * Only one **struct bpf_spin_lock** is allowed per map element.
3391 * * When the lock is taken, calls (either BPF to BPF or helpers)
3393 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3394 * allowed inside a spinlock-ed region.
3395 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3396 * the lock, on all execution paths, before it returns.
3397 * * The BPF program can access **struct bpf_spin_lock** only via
3398 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3399 * helpers. Loading or storing data into the **struct
3400 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3401 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3402 * of the map value must be a struct and have **struct
3403 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3404 * Nested lock inside another struct is not allowed.
3405 * * The **struct bpf_spin_lock** *lock* field in a map value must
3406 * be aligned on a multiple of 4 bytes in that value.
3407 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3408 * the **bpf_spin_lock** field to user space.
3409 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3410 * a BPF program, do not update the **bpf_spin_lock** field.
3411 * * **bpf_spin_lock** cannot be on the stack or inside a
3412 * networking packet (it can only be inside of a map values).
3413 * * **bpf_spin_lock** is available to root only.
3414 * * Tracing programs and socket filter programs cannot use
3415 * **bpf_spin_lock**\ () due to insufficient preemption checks
3416 * (but this may change in the future).
3417 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3421 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3423 * Release the *lock* previously locked by a call to
3424 * **bpf_spin_lock**\ (\ *lock*\ ).
3428 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3430 * This helper gets a **struct bpf_sock** pointer such
3431 * that all the fields in this **bpf_sock** can be accessed.
3433 * A **struct bpf_sock** pointer on success, or **NULL** in
3436 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3438 * This helper gets a **struct bpf_tcp_sock** pointer from a
3439 * **struct bpf_sock** pointer.
3441 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3444 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3446 * Set ECN (Explicit Congestion Notification) field of IP header
3447 * to **CE** (Congestion Encountered) if current value is **ECT**
3448 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3451 * 1 if the **CE** flag is set (either by the current helper call
3452 * or because it was already present), 0 if it is not set.
3454 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3456 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3457 * **bpf_sk_release**\ () is unnecessary and not allowed.
3459 * A **struct bpf_sock** pointer on success, or **NULL** in
3462 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3464 * Look for TCP socket matching *tuple*, optionally in a child
3465 * network namespace *netns*. The return value must be checked,
3466 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3468 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3469 * that it also returns timewait or request sockets. Use
3470 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3473 * This helper is available only if the kernel was compiled with
3474 * **CONFIG_NET** configuration option.
3476 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3477 * For sockets with reuseport option, the **struct bpf_sock**
3478 * result is from *reuse*\ **->socks**\ [] using the hash of the
3481 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3483 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3484 * the listening socket in *sk*.
3486 * *iph* points to the start of the IPv4 or IPv6 header, while
3487 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3488 * **sizeof**\ (**struct ip6hdr**).
3490 * *th* points to the start of the TCP header, while *th_len*
3491 * contains **sizeof**\ (**struct tcphdr**).
3493 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3496 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3498 * Get name of sysctl in /proc/sys/ and copy it into provided by
3499 * program buffer *buf* of size *buf_len*.
3501 * The buffer is always NUL terminated, unless it's zero-sized.
3503 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3504 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3505 * only (e.g. "tcp_mem").
3507 * Number of character copied (not including the trailing NUL).
3509 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3510 * truncated name in this case).
3512 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3514 * Get current value of sysctl as it is presented in /proc/sys
3515 * (incl. newline, etc), and copy it as a string into provided
3516 * by program buffer *buf* of size *buf_len*.
3518 * The whole value is copied, no matter what file position user
3519 * space issued e.g. sys_read at.
3521 * The buffer is always NUL terminated, unless it's zero-sized.
3523 * Number of character copied (not including the trailing NUL).
3525 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3526 * truncated name in this case).
3528 * **-EINVAL** if current value was unavailable, e.g. because
3529 * sysctl is uninitialized and read returns -EIO for it.
3531 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3533 * Get new value being written by user space to sysctl (before
3534 * the actual write happens) and copy it as a string into
3535 * provided by program buffer *buf* of size *buf_len*.
3537 * User space may write new value at file position > 0.
3539 * The buffer is always NUL terminated, unless it's zero-sized.
3541 * Number of character copied (not including the trailing NUL).
3543 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3544 * truncated name in this case).
3546 * **-EINVAL** if sysctl is being read.
3548 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3550 * Override new value being written by user space to sysctl with
3551 * value provided by program in buffer *buf* of size *buf_len*.
3553 * *buf* should contain a string in same form as provided by user
3554 * space on sysctl write.
3556 * User space may write new value at file position > 0. To override
3557 * the whole sysctl value file position should be set to zero.
3561 * **-E2BIG** if the *buf_len* is too big.
3563 * **-EINVAL** if sysctl is being read.
3565 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3567 * Convert the initial part of the string from buffer *buf* of
3568 * size *buf_len* to a long integer according to the given base
3569 * and save the result in *res*.
3571 * The string may begin with an arbitrary amount of white space
3572 * (as determined by **isspace**\ (3)) followed by a single
3573 * optional '**-**' sign.
3575 * Five least significant bits of *flags* encode base, other bits
3576 * are currently unused.
3578 * Base must be either 8, 10, 16 or 0 to detect it automatically
3579 * similar to user space **strtol**\ (3).
3581 * Number of characters consumed on success. Must be positive but
3582 * no more than *buf_len*.
3584 * **-EINVAL** if no valid digits were found or unsupported base
3587 * **-ERANGE** if resulting value was out of range.
3589 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3591 * Convert the initial part of the string from buffer *buf* of
3592 * size *buf_len* to an unsigned long integer according to the
3593 * given base and save the result in *res*.
3595 * The string may begin with an arbitrary amount of white space
3596 * (as determined by **isspace**\ (3)).
3598 * Five least significant bits of *flags* encode base, other bits
3599 * are currently unused.
3601 * Base must be either 8, 10, 16 or 0 to detect it automatically
3602 * similar to user space **strtoul**\ (3).
3604 * Number of characters consumed on success. Must be positive but
3605 * no more than *buf_len*.
3607 * **-EINVAL** if no valid digits were found or unsupported base
3610 * **-ERANGE** if resulting value was out of range.
3612 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3614 * Get a bpf-local-storage from a *sk*.
3616 * Logically, it could be thought of getting the value from
3617 * a *map* with *sk* as the **key**. From this
3618 * perspective, the usage is not much different from
3619 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3620 * helper enforces the key must be a full socket and the map must
3621 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
3623 * Underneath, the value is stored locally at *sk* instead of
3624 * the *map*. The *map* is used as the bpf-local-storage
3625 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3626 * searched against all bpf-local-storages residing at *sk*.
3628 * *sk* is a kernel **struct sock** pointer for LSM program.
3629 * *sk* is a **struct bpf_sock** pointer for other program types.
3631 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3632 * used such that a new bpf-local-storage will be
3633 * created if one does not exist. *value* can be used
3634 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3635 * the initial value of a bpf-local-storage. If *value* is
3636 * **NULL**, the new bpf-local-storage will be zero initialized.
3638 * A bpf-local-storage pointer is returned on success.
3640 * **NULL** if not found or there was an error in adding
3641 * a new bpf-local-storage.
3643 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3645 * Delete a bpf-local-storage from a *sk*.
3649 * **-ENOENT** if the bpf-local-storage cannot be found.
3650 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3652 * long bpf_send_signal(u32 sig)
3654 * Send signal *sig* to the process of the current task.
3655 * The signal may be delivered to any of this process's threads.
3657 * 0 on success or successfully queued.
3659 * **-EBUSY** if work queue under nmi is full.
3661 * **-EINVAL** if *sig* is invalid.
3663 * **-EPERM** if no permission to send the *sig*.
3665 * **-EAGAIN** if bpf program can try again.
3667 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3669 * Try to issue a SYN cookie for the packet with corresponding
3670 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3672 * *iph* points to the start of the IPv4 or IPv6 header, while
3673 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3674 * **sizeof**\ (**struct ip6hdr**).
3676 * *th* points to the start of the TCP header, while *th_len*
3677 * contains the length of the TCP header.
3679 * On success, lower 32 bits hold the generated SYN cookie in
3680 * followed by 16 bits which hold the MSS value for that cookie,
3681 * and the top 16 bits are unused.
3683 * On failure, the returned value is one of the following:
3685 * **-EINVAL** SYN cookie cannot be issued due to error
3687 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
3689 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3691 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
3693 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3695 * Write raw *data* blob into a special BPF perf event held by
3696 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3697 * event must have the following attributes: **PERF_SAMPLE_RAW**
3698 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3699 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3701 * The *flags* are used to indicate the index in *map* for which
3702 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3703 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3704 * to indicate that the index of the current CPU core should be
3707 * The value to write, of *size*, is passed through eBPF stack and
3708 * pointed by *data*.
3710 * *ctx* is a pointer to in-kernel struct sk_buff.
3712 * This helper is similar to **bpf_perf_event_output**\ () but
3713 * restricted to raw_tracepoint bpf programs.
3715 * 0 on success, or a negative error in case of failure.
3717 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3719 * Safely attempt to read *size* bytes from user space address
3720 * *unsafe_ptr* and store the data in *dst*.
3722 * 0 on success, or a negative error in case of failure.
3724 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3726 * Safely attempt to read *size* bytes from kernel space address
3727 * *unsafe_ptr* and store the data in *dst*.
3729 * 0 on success, or a negative error in case of failure.
3731 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3733 * Copy a NUL terminated string from an unsafe user address
3734 * *unsafe_ptr* to *dst*. The *size* should include the
3735 * terminating NUL byte. In case the string length is smaller than
3736 * *size*, the target is not padded with further NUL bytes. If the
3737 * string length is larger than *size*, just *size*-1 bytes are
3738 * copied and the last byte is set to NUL.
3740 * On success, returns the number of bytes that were written,
3741 * including the terminal NUL. This makes this helper useful in
3742 * tracing programs for reading strings, and more importantly to
3743 * get its length at runtime. See the following snippet:
3747 * SEC("kprobe/sys_open")
3748 * void bpf_sys_open(struct pt_regs *ctx)
3750 * char buf[PATHLEN]; // PATHLEN is defined to 256
3751 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3754 * // Consume buf, for example push it to
3755 * // userspace via bpf_perf_event_output(); we
3756 * // can use res (the string length) as event
3757 * // size, after checking its boundaries.
3760 * In comparison, using **bpf_probe_read_user**\ () helper here
3761 * instead to read the string would require to estimate the length
3762 * at compile time, and would often result in copying more memory
3765 * Another useful use case is when parsing individual process
3766 * arguments or individual environment variables navigating
3767 * *current*\ **->mm->arg_start** and *current*\
3768 * **->mm->env_start**: using this helper and the return value,
3769 * one can quickly iterate at the right offset of the memory area.
3771 * On success, the strictly positive length of the output string,
3772 * including the trailing NUL character. On error, a negative
3775 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3777 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3778 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3780 * On success, the strictly positive length of the string, including
3781 * the trailing NUL character. On error, a negative value.
3783 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3785 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3786 * *rcv_nxt* is the ack_seq to be sent out.
3788 * 0 on success, or a negative error in case of failure.
3790 * long bpf_send_signal_thread(u32 sig)
3792 * Send signal *sig* to the thread corresponding to the current task.
3794 * 0 on success or successfully queued.
3796 * **-EBUSY** if work queue under nmi is full.
3798 * **-EINVAL** if *sig* is invalid.
3800 * **-EPERM** if no permission to send the *sig*.
3802 * **-EAGAIN** if bpf program can try again.
3804 * u64 bpf_jiffies64(void)
3806 * Obtain the 64bit jiffies
3808 * The 64 bit jiffies
3810 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3812 * For an eBPF program attached to a perf event, retrieve the
3813 * branch records (**struct perf_branch_entry**) associated to *ctx*
3814 * and store it in the buffer pointed by *buf* up to size
3817 * On success, number of bytes written to *buf*. On error, a
3820 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3821 * instead return the number of bytes required to store all the
3822 * branch entries. If this flag is set, *buf* may be NULL.
3824 * **-EINVAL** if arguments invalid or **size** not a multiple
3825 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3827 * **-ENOENT** if architecture does not support branch records.
3829 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3831 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3832 * *namespace* will be returned in *nsdata*.
3834 * 0 on success, or one of the following in case of failure:
3836 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3837 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3839 * **-ENOENT** if pidns does not exists for the current task.
3841 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3843 * Write raw *data* blob into a special BPF perf event held by
3844 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3845 * event must have the following attributes: **PERF_SAMPLE_RAW**
3846 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3847 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3849 * The *flags* are used to indicate the index in *map* for which
3850 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3851 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3852 * to indicate that the index of the current CPU core should be
3855 * The value to write, of *size*, is passed through eBPF stack and
3856 * pointed by *data*.
3858 * *ctx* is a pointer to in-kernel struct xdp_buff.
3860 * This helper is similar to **bpf_perf_eventoutput**\ () but
3861 * restricted to raw_tracepoint bpf programs.
3863 * 0 on success, or a negative error in case of failure.
3865 * u64 bpf_get_netns_cookie(void *ctx)
3867 * Retrieve the cookie (generated by the kernel) of the network
3868 * namespace the input *ctx* is associated with. The network
3869 * namespace cookie remains stable for its lifetime and provides
3870 * a global identifier that can be assumed unique. If *ctx* is
3871 * NULL, then the helper returns the cookie for the initial
3872 * network namespace. The cookie itself is very similar to that
3873 * of **bpf_get_socket_cookie**\ () helper, but for network
3874 * namespaces instead of sockets.
3876 * A 8-byte long opaque number.
3878 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3880 * Return id of cgroup v2 that is ancestor of the cgroup associated
3881 * with the current task at the *ancestor_level*. The root cgroup
3882 * is at *ancestor_level* zero and each step down the hierarchy
3883 * increments the level. If *ancestor_level* == level of cgroup
3884 * associated with the current task, then return value will be the
3885 * same as that of **bpf_get_current_cgroup_id**\ ().
3887 * The helper is useful to implement policies based on cgroups
3888 * that are upper in hierarchy than immediate cgroup associated
3889 * with the current task.
3891 * The format of returned id and helper limitations are same as in
3892 * **bpf_get_current_cgroup_id**\ ().
3894 * The id is returned or 0 in case the id could not be retrieved.
3896 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3898 * Helper is overloaded depending on BPF program type. This
3899 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3900 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3902 * Assign the *sk* to the *skb*. When combined with appropriate
3903 * routing configuration to receive the packet towards the socket,
3904 * will cause *skb* to be delivered to the specified socket.
3905 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3906 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3907 * interfere with successful delivery to the socket.
3909 * This operation is only valid from TC ingress path.
3911 * The *flags* argument must be zero.
3913 * 0 on success, or a negative error in case of failure:
3915 * **-EINVAL** if specified *flags* are not supported.
3917 * **-ENOENT** if the socket is unavailable for assignment.
3919 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3921 * **-EOPNOTSUPP** if the operation is not supported, for example
3922 * a call from outside of TC ingress.
3924 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3927 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3929 * Helper is overloaded depending on BPF program type. This
3930 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3932 * Select the *sk* as a result of a socket lookup.
3934 * For the operation to succeed passed socket must be compatible
3935 * with the packet description provided by the *ctx* object.
3937 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3938 * be an exact match. While IP family (**AF_INET** or
3939 * **AF_INET6**) must be compatible, that is IPv6 sockets
3940 * that are not v6-only can be selected for IPv4 packets.
3942 * Only TCP listeners and UDP unconnected sockets can be
3943 * selected. *sk* can also be NULL to reset any previous
3946 * *flags* argument can combination of following values:
3948 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3949 * socket selection, potentially done by a BPF program
3950 * that ran before us.
3952 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3953 * load-balancing within reuseport group for the socket
3956 * On success *ctx->sk* will point to the selected socket.
3959 * 0 on success, or a negative errno in case of failure.
3961 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3962 * not compatible with packet family (*ctx->family*).
3964 * * **-EEXIST** if socket has been already selected,
3965 * potentially by another program, and
3966 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3968 * * **-EINVAL** if unsupported flags were specified.
3970 * * **-EPROTOTYPE** if socket L4 protocol
3971 * (*sk->protocol*) doesn't match packet protocol
3972 * (*ctx->protocol*).
3974 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3975 * state (TCP listening or UDP unconnected).
3977 * u64 bpf_ktime_get_boot_ns(void)
3979 * Return the time elapsed since system boot, in nanoseconds.
3980 * Does include the time the system was suspended.
3981 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
3985 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
3987 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
3988 * out the format string.
3989 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
3990 * the format string itself. The *data* and *data_len* are format string
3991 * arguments. The *data* are a **u64** array and corresponding format string
3992 * values are stored in the array. For strings and pointers where pointees
3993 * are accessed, only the pointer values are stored in the *data* array.
3994 * The *data_len* is the size of *data* in bytes.
3996 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
3997 * Reading kernel memory may fail due to either invalid address or
3998 * valid address but requiring a major memory fault. If reading kernel memory
3999 * fails, the string for **%s** will be an empty string, and the ip
4000 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4001 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4003 * 0 on success, or a negative error in case of failure:
4005 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4006 * by returning 1 from bpf program.
4008 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4010 * **-E2BIG** if *fmt* contains too many format specifiers.
4012 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4014 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4016 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4017 * The *m* represents the seq_file. The *data* and *len* represent the
4018 * data to write in bytes.
4020 * 0 on success, or a negative error in case of failure:
4022 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4024 * u64 bpf_sk_cgroup_id(void *sk)
4026 * Return the cgroup v2 id of the socket *sk*.
4028 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4029 * returned from **bpf_sk_lookup_xxx**\ (),
4030 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4031 * same as in **bpf_skb_cgroup_id**\ ().
4033 * This helper is available only if the kernel was compiled with
4034 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4036 * The id is returned or 0 in case the id could not be retrieved.
4038 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4040 * Return id of cgroup v2 that is ancestor of cgroup associated
4041 * with the *sk* at the *ancestor_level*. The root cgroup is at
4042 * *ancestor_level* zero and each step down the hierarchy
4043 * increments the level. If *ancestor_level* == level of cgroup
4044 * associated with *sk*, then return value will be same as that
4045 * of **bpf_sk_cgroup_id**\ ().
4047 * The helper is useful to implement policies based on cgroups
4048 * that are upper in hierarchy than immediate cgroup associated
4051 * The format of returned id and helper limitations are same as in
4052 * **bpf_sk_cgroup_id**\ ().
4054 * The id is returned or 0 in case the id could not be retrieved.
4056 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4058 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4059 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4060 * of new data availability is sent.
4061 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4062 * of new data availability is sent unconditionally.
4064 * 0 on success, or a negative error in case of failure.
4066 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4068 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4070 * Valid pointer with *size* bytes of memory available; NULL,
4073 * void bpf_ringbuf_submit(void *data, u64 flags)
4075 * Submit reserved ring buffer sample, pointed to by *data*.
4076 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4077 * of new data availability is sent.
4078 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4079 * of new data availability is sent unconditionally.
4081 * Nothing. Always succeeds.
4083 * void bpf_ringbuf_discard(void *data, u64 flags)
4085 * Discard reserved ring buffer sample, pointed to by *data*.
4086 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4087 * of new data availability is sent.
4088 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4089 * of new data availability is sent unconditionally.
4091 * Nothing. Always succeeds.
4093 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4095 * Query various characteristics of provided ring buffer. What
4096 * exactly is queries is determined by *flags*:
4098 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4099 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4100 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4101 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4103 * Data returned is just a momentary snapshot of actual values
4104 * and could be inaccurate, so this facility should be used to
4105 * power heuristics and for reporting, not to make 100% correct
4108 * Requested value, or 0, if *flags* are not recognized.
4110 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4112 * Change the skbs checksum level by one layer up or down, or
4113 * reset it entirely to none in order to have the stack perform
4114 * checksum validation. The level is applicable to the following
4115 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4116 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4117 * through **bpf_skb_adjust_room**\ () helper with passing in
4118 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4119 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4120 * the UDP header is removed. Similarly, an encap of the latter
4121 * into the former could be accompanied by a helper call to
4122 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4123 * skb is still intended to be processed in higher layers of the
4124 * stack instead of just egressing at tc.
4126 * There are three supported level settings at this time:
4128 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4129 * with CHECKSUM_UNNECESSARY.
4130 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4131 * with CHECKSUM_UNNECESSARY.
4132 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4133 * sets CHECKSUM_NONE to force checksum validation by the stack.
4134 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4137 * 0 on success, or a negative error in case of failure. In the
4138 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4139 * is returned or the error code -EACCES in case the skb is not
4140 * subject to CHECKSUM_UNNECESSARY.
4142 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4144 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4146 * *sk* if casting is valid, or **NULL** otherwise.
4148 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4150 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4152 * *sk* if casting is valid, or **NULL** otherwise.
4154 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4156 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4158 * *sk* if casting is valid, or **NULL** otherwise.
4160 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4162 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4164 * *sk* if casting is valid, or **NULL** otherwise.
4166 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4168 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4170 * *sk* if casting is valid, or **NULL** otherwise.
4172 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4174 * Return a user or a kernel stack in bpf program provided buffer.
4175 * To achieve this, the helper needs *task*, which is a valid
4176 * pointer to **struct task_struct**. To store the stacktrace, the
4177 * bpf program provides *buf* with a nonnegative *size*.
4179 * The last argument, *flags*, holds the number of stack frames to
4180 * skip (from 0 to 255), masked with
4181 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4182 * the following flags:
4184 * **BPF_F_USER_STACK**
4185 * Collect a user space stack instead of a kernel stack.
4186 * **BPF_F_USER_BUILD_ID**
4187 * Collect buildid+offset instead of ips for user stack,
4188 * only valid if **BPF_F_USER_STACK** is also specified.
4190 * **bpf_get_task_stack**\ () can collect up to
4191 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4192 * to sufficient large buffer size. Note that
4193 * this limit can be controlled with the **sysctl** program, and
4194 * that it should be manually increased in order to profile long
4195 * user stacks (such as stacks for Java programs). To do so, use:
4199 * # sysctl kernel.perf_event_max_stack=<new value>
4201 * A non-negative value equal to or less than *size* on success,
4202 * or a negative error in case of failure.
4204 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4206 * Load header option. Support reading a particular TCP header
4207 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4209 * If *flags* is 0, it will search the option from the
4210 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4211 * has details on what skb_data contains under different
4212 * *skops*\ **->op**.
4214 * The first byte of the *searchby_res* specifies the
4215 * kind that it wants to search.
4217 * If the searching kind is an experimental kind
4218 * (i.e. 253 or 254 according to RFC6994). It also
4219 * needs to specify the "magic" which is either
4220 * 2 bytes or 4 bytes. It then also needs to
4221 * specify the size of the magic by using
4222 * the 2nd byte which is "kind-length" of a TCP
4223 * header option and the "kind-length" also
4224 * includes the first 2 bytes "kind" and "kind-length"
4225 * itself as a normal TCP header option also does.
4227 * For example, to search experimental kind 254 with
4228 * 2 byte magic 0xeB9F, the searchby_res should be
4229 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4231 * To search for the standard window scale option (3),
4232 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4233 * Note, kind-length must be 0 for regular option.
4235 * Searching for No-Op (0) and End-of-Option-List (1) are
4238 * *len* must be at least 2 bytes which is the minimal size
4239 * of a header option.
4243 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4244 * saved_syn packet or the just-received syn packet.
4247 * > 0 when found, the header option is copied to *searchby_res*.
4248 * The return value is the total length copied. On failure, a
4249 * negative error code is returned:
4251 * **-EINVAL** if a parameter is invalid.
4253 * **-ENOMSG** if the option is not found.
4255 * **-ENOENT** if no syn packet is available when
4256 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4258 * **-ENOSPC** if there is not enough space. Only *len* number of
4261 * **-EFAULT** on failure to parse the header options in the
4264 * **-EPERM** if the helper cannot be used under the current
4265 * *skops*\ **->op**.
4267 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4269 * Store header option. The data will be copied
4270 * from buffer *from* with length *len* to the TCP header.
4272 * The buffer *from* should have the whole option that
4273 * includes the kind, kind-length, and the actual
4274 * option data. The *len* must be at least kind-length
4275 * long. The kind-length does not have to be 4 byte
4276 * aligned. The kernel will take care of the padding
4277 * and setting the 4 bytes aligned value to th->doff.
4279 * This helper will check for duplicated option
4280 * by searching the same option in the outgoing skb.
4282 * This helper can only be called during
4283 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4286 * 0 on success, or negative error in case of failure:
4288 * **-EINVAL** If param is invalid.
4290 * **-ENOSPC** if there is not enough space in the header.
4291 * Nothing has been written
4293 * **-EEXIST** if the option already exists.
4295 * **-EFAULT** on failrue to parse the existing header options.
4297 * **-EPERM** if the helper cannot be used under the current
4298 * *skops*\ **->op**.
4300 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4302 * Reserve *len* bytes for the bpf header option. The
4303 * space will be used by **bpf_store_hdr_opt**\ () later in
4304 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4306 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4307 * the total number of bytes will be reserved.
4309 * This helper can only be called during
4310 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4313 * 0 on success, or negative error in case of failure:
4315 * **-EINVAL** if a parameter is invalid.
4317 * **-ENOSPC** if there is not enough space in the header.
4319 * **-EPERM** if the helper cannot be used under the current
4320 * *skops*\ **->op**.
4322 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4324 * Get a bpf_local_storage from an *inode*.
4326 * Logically, it could be thought of as getting the value from
4327 * a *map* with *inode* as the **key**. From this
4328 * perspective, the usage is not much different from
4329 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4330 * helper enforces the key must be an inode and the map must also
4331 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4333 * Underneath, the value is stored locally at *inode* instead of
4334 * the *map*. The *map* is used as the bpf-local-storage
4335 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4336 * searched against all bpf_local_storage residing at *inode*.
4338 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4339 * used such that a new bpf_local_storage will be
4340 * created if one does not exist. *value* can be used
4341 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4342 * the initial value of a bpf_local_storage. If *value* is
4343 * **NULL**, the new bpf_local_storage will be zero initialized.
4345 * A bpf_local_storage pointer is returned on success.
4347 * **NULL** if not found or there was an error in adding
4348 * a new bpf_local_storage.
4350 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4352 * Delete a bpf_local_storage from an *inode*.
4356 * **-ENOENT** if the bpf_local_storage cannot be found.
4358 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4360 * Return full path for given **struct path** object, which
4361 * needs to be the kernel BTF *path* object. The path is
4362 * returned in the provided buffer *buf* of size *sz* and
4363 * is zero terminated.
4366 * On success, the strictly positive length of the string,
4367 * including the trailing NUL character. On error, a negative
4370 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4372 * Read *size* bytes from user space address *user_ptr* and store
4373 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4375 * 0 on success, or a negative error in case of failure.
4377 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4379 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4380 * using *ptr*->type_id. This value should specify the type
4381 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4382 * can be used to look up vmlinux BTF type ids. Traversing the
4383 * data structure using BTF, the type information and values are
4384 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4385 * the pointer data is carried out to avoid kernel crashes during
4386 * operation. Smaller types can use string space on the stack;
4387 * larger programs can use map data to store the string
4390 * The string can be subsequently shared with userspace via
4391 * bpf_perf_event_output() or ring buffer interfaces.
4392 * bpf_trace_printk() is to be avoided as it places too small
4393 * a limit on string size to be useful.
4395 * *flags* is a combination of
4398 * no formatting around type information
4400 * no struct/union member names/types
4402 * show raw (unobfuscated) pointer values;
4403 * equivalent to printk specifier %px.
4405 * show zero-valued struct/union members; they
4406 * are not displayed by default
4409 * The number of bytes that were written (or would have been
4410 * written if output had to be truncated due to string size),
4411 * or a negative error in cases of failure.
4413 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4415 * Use BTF to write to seq_write a string representation of
4416 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4417 * *flags* are identical to those used for bpf_snprintf_btf.
4419 * 0 on success or a negative error in case of failure.
4421 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4423 * See **bpf_get_cgroup_classid**\ () for the main description.
4424 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4425 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4426 * associated socket instead of the current process.
4428 * The id is returned or 0 in case the id could not be retrieved.
4430 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4432 * Redirect the packet to another net device of index *ifindex*
4433 * and fill in L2 addresses from neighboring subsystem. This helper
4434 * is somewhat similar to **bpf_redirect**\ (), except that it
4435 * populates L2 addresses as well, meaning, internally, the helper
4436 * relies on the neighbor lookup for the L2 address of the nexthop.
4438 * The helper will perform a FIB lookup based on the skb's
4439 * networking header to get the address of the next hop, unless
4440 * this is supplied by the caller in the *params* argument. The
4441 * *plen* argument indicates the len of *params* and should be set
4442 * to 0 if *params* is NULL.
4444 * The *flags* argument is reserved and must be 0. The helper is
4445 * currently only supported for tc BPF program types, and enabled
4446 * for IPv4 and IPv6 protocols.
4448 * The helper returns **TC_ACT_REDIRECT** on success or
4449 * **TC_ACT_SHOT** on error.
4451 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4453 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4454 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4455 * extern variable decorated with '__ksym'. For ksym, there is a
4456 * global var (either static or global) defined of the same name
4457 * in the kernel. The ksym is percpu if the global var is percpu.
4458 * The returned pointer points to the global percpu var on *cpu*.
4460 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4461 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4462 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4463 * bpf_per_cpu_ptr() must check the returned value.
4465 * A pointer pointing to the kernel percpu variable on *cpu*, or
4466 * NULL, if *cpu* is invalid.
4468 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4470 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4471 * pointer to the percpu kernel variable on this cpu. See the
4472 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4474 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4475 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4476 * never return NULL.
4478 * A pointer pointing to the kernel percpu variable on this cpu.
4480 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4482 * Redirect the packet to another net device of index *ifindex*.
4483 * This helper is somewhat similar to **bpf_redirect**\ (), except
4484 * that the redirection happens to the *ifindex*' peer device and
4485 * the netns switch takes place from ingress to ingress without
4486 * going through the CPU's backlog queue.
4488 * The *flags* argument is reserved and must be 0. The helper is
4489 * currently only supported for tc BPF program types at the ingress
4490 * hook and for veth device types. The peer device must reside in a
4491 * different network namespace.
4493 * The helper returns **TC_ACT_REDIRECT** on success or
4494 * **TC_ACT_SHOT** on error.
4496 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4498 * Get a bpf_local_storage from the *task*.
4500 * Logically, it could be thought of as getting the value from
4501 * a *map* with *task* as the **key**. From this
4502 * perspective, the usage is not much different from
4503 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4504 * helper enforces the key must be an task_struct and the map must also
4505 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4507 * Underneath, the value is stored locally at *task* instead of
4508 * the *map*. The *map* is used as the bpf-local-storage
4509 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4510 * searched against all bpf_local_storage residing at *task*.
4512 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4513 * used such that a new bpf_local_storage will be
4514 * created if one does not exist. *value* can be used
4515 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4516 * the initial value of a bpf_local_storage. If *value* is
4517 * **NULL**, the new bpf_local_storage will be zero initialized.
4519 * A bpf_local_storage pointer is returned on success.
4521 * **NULL** if not found or there was an error in adding
4522 * a new bpf_local_storage.
4524 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4526 * Delete a bpf_local_storage from a *task*.
4530 * **-ENOENT** if the bpf_local_storage cannot be found.
4532 * struct task_struct *bpf_get_current_task_btf(void)
4534 * Return a BTF pointer to the "current" task.
4535 * This pointer can also be used in helpers that accept an
4536 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
4538 * Pointer to the current task.
4540 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4542 * Set or clear certain options on *bprm*:
4544 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4545 * which sets the **AT_SECURE** auxv for glibc. The bit
4546 * is cleared if the flag is not specified.
4548 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
4550 * u64 bpf_ktime_get_coarse_ns(void)
4552 * Return a coarse-grained version of the time elapsed since
4553 * system boot, in nanoseconds. Does not include time the system
4556 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4560 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4562 * Returns the stored IMA hash of the *inode* (if it's avaialable).
4563 * If the hash is larger than *size*, then only *size*
4564 * bytes will be copied to *dst*
4566 * The **hash_algo** is returned on success,
4567 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4568 * invalid arguments are passed.
4570 * struct socket *bpf_sock_from_file(struct file *file)
4572 * If the given file represents a socket, returns the associated
4575 * A pointer to a struct socket on success or NULL if the file is
4578 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4580 * Check ctx packet size against exceeding MTU of net device (based
4581 * on *ifindex*). This helper will likely be used in combination
4582 * with helpers that adjust/change the packet size.
4584 * The argument *len_diff* can be used for querying with a planned
4585 * size change. This allows to check MTU prior to changing packet
4586 * ctx. Providing an *len_diff* adjustment that is larger than the
4587 * actual packet size (resulting in negative packet size) will in
4588 * principle not exceed the MTU, why it is not considered a
4589 * failure. Other BPF-helpers are needed for performing the
4590 * planned size change, why the responsability for catch a negative
4591 * packet size belong in those helpers.
4593 * Specifying *ifindex* zero means the MTU check is performed
4594 * against the current net device. This is practical if this isn't
4595 * used prior to redirect.
4597 * The Linux kernel route table can configure MTUs on a more
4598 * specific per route level, which is not provided by this helper.
4599 * For route level MTU checks use the **bpf_fib_lookup**\ ()
4602 * *ctx* is either **struct xdp_md** for XDP programs or
4603 * **struct sk_buff** for tc cls_act programs.
4605 * The *flags* argument can be a combination of one or more of the
4608 * **BPF_MTU_CHK_SEGS**
4609 * This flag will only works for *ctx* **struct sk_buff**.
4610 * If packet context contains extra packet segment buffers
4611 * (often knows as GSO skb), then MTU check is harder to
4612 * check at this point, because in transmit path it is
4613 * possible for the skb packet to get re-segmented
4614 * (depending on net device features). This could still be
4615 * a MTU violation, so this flag enables performing MTU
4616 * check against segments, with a different violation
4617 * return code to tell it apart. Check cannot use len_diff.
4619 * On return *mtu_len* pointer contains the MTU value of the net
4620 * device. Remember the net device configured MTU is the L3 size,
4621 * which is returned here and XDP and TX length operate at L2.
4622 * Helper take this into account for you, but remember when using
4623 * MTU value in your BPF-code. On input *mtu_len* must be a valid
4624 * pointer and be initialized (to zero), else verifier will reject
4628 * * 0 on success, and populate MTU value in *mtu_len* pointer.
4630 * * < 0 if any input argument is invalid (*mtu_len* not updated)
4632 * MTU violations return positive values, but also populate MTU
4633 * value in *mtu_len* pointer, as this can be needed for
4634 * implementing PMTU handing:
4636 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
4637 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4639 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4641 * For each element in **map**, call **callback_fn** function with
4642 * **map**, **callback_ctx** and other map-specific parameters.
4643 * The **callback_fn** should be a static function and
4644 * the **callback_ctx** should be a pointer to the stack.
4645 * The **flags** is used to control certain aspects of the helper.
4646 * Currently, the **flags** must be 0.
4648 * The following are a list of supported map types and their
4649 * respective expected callback signatures:
4651 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4652 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4653 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4655 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4657 * For per_cpu maps, the map_value is the value on the cpu where the
4658 * bpf_prog is running.
4660 * If **callback_fn** return 0, the helper will continue to the next
4661 * element. If return value is 1, the helper will skip the rest of
4662 * elements and return. Other return values are not used now.
4665 * The number of traversed map elements for success, **-EINVAL** for
4666 * invalid **flags**.
4668 #define __BPF_FUNC_MAPPER(FN) \
4670 FN(map_lookup_elem), \
4671 FN(map_update_elem), \
4672 FN(map_delete_elem), \
4676 FN(get_prandom_u32), \
4677 FN(get_smp_processor_id), \
4678 FN(skb_store_bytes), \
4679 FN(l3_csum_replace), \
4680 FN(l4_csum_replace), \
4682 FN(clone_redirect), \
4683 FN(get_current_pid_tgid), \
4684 FN(get_current_uid_gid), \
4685 FN(get_current_comm), \
4686 FN(get_cgroup_classid), \
4687 FN(skb_vlan_push), \
4689 FN(skb_get_tunnel_key), \
4690 FN(skb_set_tunnel_key), \
4691 FN(perf_event_read), \
4693 FN(get_route_realm), \
4694 FN(perf_event_output), \
4695 FN(skb_load_bytes), \
4698 FN(skb_get_tunnel_opt), \
4699 FN(skb_set_tunnel_opt), \
4700 FN(skb_change_proto), \
4701 FN(skb_change_type), \
4702 FN(skb_under_cgroup), \
4703 FN(get_hash_recalc), \
4704 FN(get_current_task), \
4705 FN(probe_write_user), \
4706 FN(current_task_under_cgroup), \
4707 FN(skb_change_tail), \
4708 FN(skb_pull_data), \
4710 FN(set_hash_invalid), \
4711 FN(get_numa_node_id), \
4712 FN(skb_change_head), \
4713 FN(xdp_adjust_head), \
4714 FN(probe_read_str), \
4715 FN(get_socket_cookie), \
4716 FN(get_socket_uid), \
4719 FN(skb_adjust_room), \
4721 FN(sk_redirect_map), \
4722 FN(sock_map_update), \
4723 FN(xdp_adjust_meta), \
4724 FN(perf_event_read_value), \
4725 FN(perf_prog_read_value), \
4727 FN(override_return), \
4728 FN(sock_ops_cb_flags_set), \
4729 FN(msg_redirect_map), \
4730 FN(msg_apply_bytes), \
4731 FN(msg_cork_bytes), \
4732 FN(msg_pull_data), \
4734 FN(xdp_adjust_tail), \
4735 FN(skb_get_xfrm_state), \
4737 FN(skb_load_bytes_relative), \
4739 FN(sock_hash_update), \
4740 FN(msg_redirect_hash), \
4741 FN(sk_redirect_hash), \
4742 FN(lwt_push_encap), \
4743 FN(lwt_seg6_store_bytes), \
4744 FN(lwt_seg6_adjust_srh), \
4745 FN(lwt_seg6_action), \
4748 FN(skb_cgroup_id), \
4749 FN(get_current_cgroup_id), \
4750 FN(get_local_storage), \
4751 FN(sk_select_reuseport), \
4752 FN(skb_ancestor_cgroup_id), \
4753 FN(sk_lookup_tcp), \
4754 FN(sk_lookup_udp), \
4756 FN(map_push_elem), \
4758 FN(map_peek_elem), \
4759 FN(msg_push_data), \
4761 FN(rc_pointer_rel), \
4766 FN(skb_ecn_set_ce), \
4767 FN(get_listener_sock), \
4768 FN(skc_lookup_tcp), \
4769 FN(tcp_check_syncookie), \
4770 FN(sysctl_get_name), \
4771 FN(sysctl_get_current_value), \
4772 FN(sysctl_get_new_value), \
4773 FN(sysctl_set_new_value), \
4776 FN(sk_storage_get), \
4777 FN(sk_storage_delete), \
4779 FN(tcp_gen_syncookie), \
4781 FN(probe_read_user), \
4782 FN(probe_read_kernel), \
4783 FN(probe_read_user_str), \
4784 FN(probe_read_kernel_str), \
4786 FN(send_signal_thread), \
4788 FN(read_branch_records), \
4789 FN(get_ns_current_pid_tgid), \
4791 FN(get_netns_cookie), \
4792 FN(get_current_ancestor_cgroup_id), \
4794 FN(ktime_get_boot_ns), \
4798 FN(sk_ancestor_cgroup_id), \
4799 FN(ringbuf_output), \
4800 FN(ringbuf_reserve), \
4801 FN(ringbuf_submit), \
4802 FN(ringbuf_discard), \
4803 FN(ringbuf_query), \
4805 FN(skc_to_tcp6_sock), \
4806 FN(skc_to_tcp_sock), \
4807 FN(skc_to_tcp_timewait_sock), \
4808 FN(skc_to_tcp_request_sock), \
4809 FN(skc_to_udp6_sock), \
4810 FN(get_task_stack), \
4812 FN(store_hdr_opt), \
4813 FN(reserve_hdr_opt), \
4814 FN(inode_storage_get), \
4815 FN(inode_storage_delete), \
4817 FN(copy_from_user), \
4819 FN(seq_printf_btf), \
4820 FN(skb_cgroup_classid), \
4821 FN(redirect_neigh), \
4824 FN(redirect_peer), \
4825 FN(task_storage_get), \
4826 FN(task_storage_delete), \
4827 FN(get_current_task_btf), \
4828 FN(bprm_opts_set), \
4829 FN(ktime_get_coarse_ns), \
4830 FN(ima_inode_hash), \
4831 FN(sock_from_file), \
4833 FN(for_each_map_elem), \
4836 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
4837 * function eBPF program intends to call
4839 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
4841 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
4844 #undef __BPF_ENUM_FN
4846 /* All flags used by eBPF helper functions, placed here. */
4848 /* BPF_FUNC_skb_store_bytes flags. */
4850 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
4851 BPF_F_INVALIDATE_HASH = (1ULL << 1),
4854 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
4855 * First 4 bits are for passing the header field size.
4858 BPF_F_HDR_FIELD_MASK = 0xfULL,
4861 /* BPF_FUNC_l4_csum_replace flags. */
4863 BPF_F_PSEUDO_HDR = (1ULL << 4),
4864 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
4865 BPF_F_MARK_ENFORCE = (1ULL << 6),
4868 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
4870 BPF_F_INGRESS = (1ULL << 0),
4873 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
4875 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
4878 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
4880 BPF_F_SKIP_FIELD_MASK = 0xffULL,
4881 BPF_F_USER_STACK = (1ULL << 8),
4882 /* flags used by BPF_FUNC_get_stackid only. */
4883 BPF_F_FAST_STACK_CMP = (1ULL << 9),
4884 BPF_F_REUSE_STACKID = (1ULL << 10),
4885 /* flags used by BPF_FUNC_get_stack only. */
4886 BPF_F_USER_BUILD_ID = (1ULL << 11),
4889 /* BPF_FUNC_skb_set_tunnel_key flags. */
4891 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
4892 BPF_F_DONT_FRAGMENT = (1ULL << 2),
4893 BPF_F_SEQ_NUMBER = (1ULL << 3),
4896 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
4897 * BPF_FUNC_perf_event_read_value flags.
4900 BPF_F_INDEX_MASK = 0xffffffffULL,
4901 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
4902 /* BPF_FUNC_perf_event_output for sk_buff input context. */
4903 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
4906 /* Current network namespace */
4908 BPF_F_CURRENT_NETNS = (-1L),
4911 /* BPF_FUNC_csum_level level values. */
4913 BPF_CSUM_LEVEL_QUERY,
4916 BPF_CSUM_LEVEL_RESET,
4919 /* BPF_FUNC_skb_adjust_room flags. */
4921 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
4922 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
4923 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
4924 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
4925 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
4926 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
4927 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
4931 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
4932 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
4935 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
4936 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
4937 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
4939 /* BPF_FUNC_sysctl_get_name flags. */
4941 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
4944 /* BPF_FUNC_<kernel_obj>_storage_get flags */
4946 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
4947 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
4948 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
4950 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
4953 /* BPF_FUNC_read_branch_records flags. */
4955 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
4958 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
4959 * BPF_FUNC_bpf_ringbuf_output flags.
4962 BPF_RB_NO_WAKEUP = (1ULL << 0),
4963 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
4966 /* BPF_FUNC_bpf_ringbuf_query flags */
4968 BPF_RB_AVAIL_DATA = 0,
4969 BPF_RB_RING_SIZE = 1,
4970 BPF_RB_CONS_POS = 2,
4971 BPF_RB_PROD_POS = 3,
4974 /* BPF ring buffer constants */
4976 BPF_RINGBUF_BUSY_BIT = (1U << 31),
4977 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
4978 BPF_RINGBUF_HDR_SZ = 8,
4981 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
4983 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
4984 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
4987 /* Mode for BPF_FUNC_skb_adjust_room helper. */
4988 enum bpf_adj_room_mode {
4993 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
4994 enum bpf_hdr_start_off {
4999 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5000 enum bpf_lwt_encap_mode {
5002 BPF_LWT_ENCAP_SEG6_INLINE,
5006 /* Flags for bpf_bprm_opts_set helper */
5008 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
5011 #define __bpf_md_ptr(type, name) \
5015 } __attribute__((aligned(8)))
5017 /* user accessible mirror of in-kernel sk_buff.
5018 * new fields can only be added to the end of this structure
5024 __u32 queue_mapping;
5030 __u32 ingress_ifindex;
5040 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5042 __u32 remote_ip4; /* Stored in network byte order */
5043 __u32 local_ip4; /* Stored in network byte order */
5044 __u32 remote_ip6[4]; /* Stored in network byte order */
5045 __u32 local_ip6[4]; /* Stored in network byte order */
5046 __u32 remote_port; /* Stored in network byte order */
5047 __u32 local_port; /* stored in host byte order */
5051 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5055 __bpf_md_ptr(struct bpf_sock *, sk);
5059 struct bpf_tunnel_key {
5063 __u32 remote_ipv6[4];
5067 __u16 tunnel_ext; /* Padding, future use. */
5071 /* user accessible mirror of in-kernel xfrm_state.
5072 * new fields can only be added to the end of this structure
5074 struct bpf_xfrm_state {
5076 __u32 spi; /* Stored in network byte order */
5078 __u16 ext; /* Padding, future use. */
5080 __u32 remote_ipv4; /* Stored in network byte order */
5081 __u32 remote_ipv6[4]; /* Stored in network byte order */
5085 /* Generic BPF return codes which all BPF program types may support.
5086 * The values are binary compatible with their TC_ACT_* counter-part to
5087 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5090 * XDP is handled seprately, see XDP_*.
5098 /* >127 are reserved for prog type specific return codes.
5100 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5101 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5102 * changed and should be routed based on its new L3 header.
5103 * (This is an L3 redirect, as opposed to L2 redirect
5104 * represented by BPF_REDIRECT above).
5106 BPF_LWT_REROUTE = 128,
5116 /* IP address also allows 1 and 2 bytes access */
5119 __u32 src_port; /* host byte order */
5120 __u32 dst_port; /* network byte order */
5124 __s32 rx_queue_mapping;
5127 struct bpf_tcp_sock {
5128 __u32 snd_cwnd; /* Sending congestion window */
5129 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
5131 __u32 snd_ssthresh; /* Slow start size threshold */
5132 __u32 rcv_nxt; /* What we want to receive next */
5133 __u32 snd_nxt; /* Next sequence we send */
5134 __u32 snd_una; /* First byte we want an ack for */
5135 __u32 mss_cache; /* Cached effective mss, not including SACKS */
5136 __u32 ecn_flags; /* ECN status bits. */
5137 __u32 rate_delivered; /* saved rate sample: packets delivered */
5138 __u32 rate_interval_us; /* saved rate sample: time elapsed */
5139 __u32 packets_out; /* Packets which are "in flight" */
5140 __u32 retrans_out; /* Retransmitted packets out */
5141 __u32 total_retrans; /* Total retransmits for entire connection */
5142 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
5143 * total number of segments in.
5145 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
5146 * total number of data segments in.
5148 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
5149 * The total number of segments sent.
5151 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
5152 * total number of data segments sent.
5154 __u32 lost_out; /* Lost packets */
5155 __u32 sacked_out; /* SACK'd packets */
5156 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
5157 * sum(delta(rcv_nxt)), or how many bytes
5160 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
5161 * sum(delta(snd_una)), or how many bytes
5164 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
5165 * total number of DSACK blocks received
5167 __u32 delivered; /* Total data packets delivered incl. rexmits */
5168 __u32 delivered_ce; /* Like the above but only ECE marked packets */
5169 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
5172 struct bpf_sock_tuple {
5189 struct bpf_xdp_sock {
5193 #define XDP_PACKET_HEADROOM 256
5195 /* User return codes for XDP prog type.
5196 * A valid XDP program must return one of these defined values. All other
5197 * return codes are reserved for future use. Unknown return codes will
5198 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5208 /* user accessible metadata for XDP packet hook
5209 * new fields must be added to the end of this structure
5215 /* Below access go through struct xdp_rxq_info */
5216 __u32 ingress_ifindex; /* rxq->dev->ifindex */
5217 __u32 rx_queue_index; /* rxq->queue_index */
5219 __u32 egress_ifindex; /* txq->dev->ifindex */
5222 /* DEVMAP map-value layout
5224 * The struct data-layout of map-value is a configuration interface.
5225 * New members can only be added to the end of this structure.
5227 struct bpf_devmap_val {
5228 __u32 ifindex; /* device index */
5230 int fd; /* prog fd on map write */
5231 __u32 id; /* prog id on map read */
5235 /* CPUMAP map-value layout
5237 * The struct data-layout of map-value is a configuration interface.
5238 * New members can only be added to the end of this structure.
5240 struct bpf_cpumap_val {
5241 __u32 qsize; /* queue size to remote target CPU */
5243 int fd; /* prog fd on map write */
5244 __u32 id; /* prog id on map read */
5253 /* user accessible metadata for SK_MSG packet hook, new fields must
5254 * be added to the end of this structure
5257 __bpf_md_ptr(void *, data);
5258 __bpf_md_ptr(void *, data_end);
5261 __u32 remote_ip4; /* Stored in network byte order */
5262 __u32 local_ip4; /* Stored in network byte order */
5263 __u32 remote_ip6[4]; /* Stored in network byte order */
5264 __u32 local_ip6[4]; /* Stored in network byte order */
5265 __u32 remote_port; /* Stored in network byte order */
5266 __u32 local_port; /* stored in host byte order */
5267 __u32 size; /* Total size of sk_msg */
5269 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
5272 struct sk_reuseport_md {
5274 * Start of directly accessible data. It begins from
5275 * the tcp/udp header.
5277 __bpf_md_ptr(void *, data);
5278 /* End of directly accessible data */
5279 __bpf_md_ptr(void *, data_end);
5281 * Total length of packet (starting from the tcp/udp header).
5282 * Note that the directly accessible bytes (data_end - data)
5283 * could be less than this "len". Those bytes could be
5284 * indirectly read by a helper "bpf_skb_load_bytes()".
5288 * Eth protocol in the mac header (network byte order). e.g.
5289 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
5292 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
5293 __u32 bind_inany; /* Is sock bound to an INANY address? */
5294 __u32 hash; /* A hash of the packet 4 tuples */
5297 #define BPF_TAG_SIZE 8
5299 struct bpf_prog_info {
5302 __u8 tag[BPF_TAG_SIZE];
5303 __u32 jited_prog_len;
5304 __u32 xlated_prog_len;
5305 __aligned_u64 jited_prog_insns;
5306 __aligned_u64 xlated_prog_insns;
5307 __u64 load_time; /* ns since boottime */
5308 __u32 created_by_uid;
5310 __aligned_u64 map_ids;
5311 char name[BPF_OBJ_NAME_LEN];
5313 __u32 gpl_compatible:1;
5314 __u32 :31; /* alignment pad */
5317 __u32 nr_jited_ksyms;
5318 __u32 nr_jited_func_lens;
5319 __aligned_u64 jited_ksyms;
5320 __aligned_u64 jited_func_lens;
5322 __u32 func_info_rec_size;
5323 __aligned_u64 func_info;
5326 __aligned_u64 line_info;
5327 __aligned_u64 jited_line_info;
5328 __u32 nr_jited_line_info;
5329 __u32 line_info_rec_size;
5330 __u32 jited_line_info_rec_size;
5332 __aligned_u64 prog_tags;
5335 __u64 recursion_misses;
5336 } __attribute__((aligned(8)));
5338 struct bpf_map_info {
5345 char name[BPF_OBJ_NAME_LEN];
5347 __u32 btf_vmlinux_value_type_id;
5351 __u32 btf_key_type_id;
5352 __u32 btf_value_type_id;
5353 } __attribute__((aligned(8)));
5355 struct bpf_btf_info {
5362 } __attribute__((aligned(8)));
5364 struct bpf_link_info {
5370 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
5371 __u32 tp_name_len; /* in/out: tp_name buffer len */
5381 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
5382 __u32 target_name_len; /* in/out: target_name buffer len */
5397 } __attribute__((aligned(8)));
5399 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
5400 * by user and intended to be used by socket (e.g. to bind to, depends on
5403 struct bpf_sock_addr {
5404 __u32 user_family; /* Allows 4-byte read, but no write. */
5405 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5406 * Stored in network byte order.
5408 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5409 * Stored in network byte order.
5411 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
5412 * Stored in network byte order
5414 __u32 family; /* Allows 4-byte read, but no write */
5415 __u32 type; /* Allows 4-byte read, but no write */
5416 __u32 protocol; /* Allows 4-byte read, but no write */
5417 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5418 * Stored in network byte order.
5420 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5421 * Stored in network byte order.
5423 __bpf_md_ptr(struct bpf_sock *, sk);
5426 /* User bpf_sock_ops struct to access socket values and specify request ops
5427 * and their replies.
5428 * Some of this fields are in network (bigendian) byte order and may need
5429 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
5430 * New fields can only be added at the end of this structure
5432 struct bpf_sock_ops {
5435 __u32 args[4]; /* Optionally passed to bpf program */
5436 __u32 reply; /* Returned by bpf program */
5437 __u32 replylong[4]; /* Optionally returned by bpf prog */
5440 __u32 remote_ip4; /* Stored in network byte order */
5441 __u32 local_ip4; /* Stored in network byte order */
5442 __u32 remote_ip6[4]; /* Stored in network byte order */
5443 __u32 local_ip6[4]; /* Stored in network byte order */
5444 __u32 remote_port; /* Stored in network byte order */
5445 __u32 local_port; /* stored in host byte order */
5446 __u32 is_fullsock; /* Some TCP fields are only valid if
5447 * there is a full socket. If not, the
5448 * fields read as zero.
5451 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
5452 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
5461 __u32 rate_delivered;
5462 __u32 rate_interval_us;
5465 __u32 total_retrans;
5469 __u32 data_segs_out;
5473 __u64 bytes_received;
5475 __bpf_md_ptr(struct bpf_sock *, sk);
5476 /* [skb_data, skb_data_end) covers the whole TCP header.
5478 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
5479 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
5480 * header has not been written.
5481 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
5482 * been written so far.
5483 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
5485 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
5488 * bpf_load_hdr_opt() can also be used to read a particular option.
5490 __bpf_md_ptr(void *, skb_data);
5491 __bpf_md_ptr(void *, skb_data_end);
5492 __u32 skb_len; /* The total length of a packet.
5493 * It includes the header, options,
5496 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
5497 * an easy way to check for tcp_flags
5498 * without parsing skb_data.
5500 * In particular, the skb_tcp_flags
5501 * will still be available in
5502 * BPF_SOCK_OPS_HDR_OPT_LEN even though
5503 * the outgoing header has not
5508 /* Definitions for bpf_sock_ops_cb_flags */
5510 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
5511 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
5512 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
5513 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
5514 /* Call bpf for all received TCP headers. The bpf prog will be
5515 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5517 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5518 * for the header option related helpers that will be useful
5519 * to the bpf programs.
5521 * It could be used at the client/active side (i.e. connect() side)
5522 * when the server told it that the server was in syncookie
5523 * mode and required the active side to resend the bpf-written
5524 * options. The active side can keep writing the bpf-options until
5525 * it received a valid packet from the server side to confirm
5526 * the earlier packet (and options) has been received. The later
5527 * example patch is using it like this at the active side when the
5528 * server is in syncookie mode.
5530 * The bpf prog will usually turn this off in the common cases.
5532 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
5533 /* Call bpf when kernel has received a header option that
5534 * the kernel cannot handle. The bpf prog will be called under
5535 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
5537 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5538 * for the header option related helpers that will be useful
5539 * to the bpf programs.
5541 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
5542 /* Call bpf when the kernel is writing header options for the
5543 * outgoing packet. The bpf prog will first be called
5544 * to reserve space in a skb under
5545 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
5546 * the bpf prog will be called to write the header option(s)
5547 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5549 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
5550 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
5551 * related helpers that will be useful to the bpf programs.
5553 * The kernel gets its chance to reserve space and write
5554 * options first before the BPF program does.
5556 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
5557 /* Mask of all currently supported cb flags */
5558 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
5561 /* List of known BPF sock_ops operators.
5562 * New entries can only be added at the end
5566 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
5567 * -1 if default value should be used
5569 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
5570 * window (in packets) or -1 if default
5571 * value should be used
5573 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
5574 * active connection is initialized
5576 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
5577 * active connection is
5580 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
5581 * passive connection is
5584 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
5587 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
5588 * based on the path and may be
5589 * dependent on the congestion control
5590 * algorithm. In general it indicates
5591 * a congestion threshold. RTTs above
5592 * this indicate congestion
5594 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
5595 * Arg1: value of icsk_retransmits
5596 * Arg2: value of icsk_rto
5597 * Arg3: whether RTO has expired
5599 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
5600 * Arg1: sequence number of 1st byte
5602 * Arg3: return value of
5603 * tcp_transmit_skb (0 => success)
5605 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
5609 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
5610 * socket transition to LISTEN state.
5612 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
5614 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
5615 * It will be called to handle
5616 * the packets received at
5617 * an already established
5620 * sock_ops->skb_data:
5621 * Referring to the received skb.
5622 * It covers the TCP header only.
5624 * bpf_load_hdr_opt() can also
5625 * be used to search for a
5626 * particular option.
5628 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
5629 * header option later in
5630 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5631 * Arg1: bool want_cookie. (in
5632 * writing SYNACK only)
5634 * sock_ops->skb_data:
5635 * Not available because no header has
5638 * sock_ops->skb_tcp_flags:
5639 * The tcp_flags of the
5640 * outgoing skb. (e.g. SYN, ACK, FIN).
5642 * bpf_reserve_hdr_opt() should
5643 * be used to reserve space.
5645 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
5646 * Arg1: bool want_cookie. (in
5647 * writing SYNACK only)
5649 * sock_ops->skb_data:
5650 * Referring to the outgoing skb.
5651 * It covers the TCP header
5652 * that has already been written
5653 * by the kernel and the
5654 * earlier bpf-progs.
5656 * sock_ops->skb_tcp_flags:
5657 * The tcp_flags of the outgoing
5658 * skb. (e.g. SYN, ACK, FIN).
5660 * bpf_store_hdr_opt() should
5661 * be used to write the
5664 * bpf_load_hdr_opt() can also
5665 * be used to search for a
5666 * particular option that
5667 * has already been written
5668 * by the kernel or the
5669 * earlier bpf-progs.
5673 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
5674 * changes between the TCP and BPF versions. Ideally this should never happen.
5675 * If it does, we need to add code to convert them before calling
5676 * the BPF sock_ops function.
5679 BPF_TCP_ESTABLISHED = 1,
5689 BPF_TCP_CLOSING, /* Now a valid state */
5690 BPF_TCP_NEW_SYN_RECV,
5692 BPF_TCP_MAX_STATES /* Leave at the end! */
5696 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
5697 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
5698 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
5699 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
5700 /* Copy the SYN pkt to optval
5702 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
5703 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
5704 * to only getting from the saved_syn. It can either get the
5707 * 1. the just-received SYN packet (only available when writing the
5708 * SYNACK). It will be useful when it is not necessary to
5709 * save the SYN packet for latter use. It is also the only way
5710 * to get the SYN during syncookie mode because the syn
5711 * packet cannot be saved during syncookie.
5715 * 2. the earlier saved syn which was done by
5716 * bpf_setsockopt(TCP_SAVE_SYN).
5718 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
5719 * SYN packet is obtained.
5721 * If the bpf-prog does not need the IP[46] header, the
5722 * bpf-prog can avoid parsing the IP header by using
5723 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
5724 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
5726 * >0: Total number of bytes copied
5727 * -ENOSPC: Not enough space in optval. Only optlen number of
5729 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
5730 * is not saved by setsockopt(TCP_SAVE_SYN).
5732 TCP_BPF_SYN = 1005, /* Copy the TCP header */
5733 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
5734 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
5738 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
5741 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
5742 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5745 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
5746 * total option spaces
5747 * required for an established
5748 * sk in order to calculate the
5749 * MSS. No skb is actually
5752 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
5753 * when sending a SYN.
5757 struct bpf_perf_event_value {
5764 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
5765 BPF_DEVCG_ACC_READ = (1ULL << 1),
5766 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
5770 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
5771 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
5774 struct bpf_cgroup_dev_ctx {
5775 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
5781 struct bpf_raw_tracepoint_args {
5785 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
5786 * OUTPUT: Do lookup from egress perspective; default is ingress
5789 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
5790 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
5794 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
5795 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
5796 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
5797 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
5798 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
5799 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
5800 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
5801 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
5802 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5805 struct bpf_fib_lookup {
5806 /* input: network family for lookup (AF_INET, AF_INET6)
5807 * output: network family of egress nexthop
5811 /* set if lookup is to consider L4 data - e.g., FIB rules */
5816 union { /* used for MTU check */
5817 /* input to lookup */
5818 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
5820 /* output: MTU value */
5823 /* input: L3 device index for lookup
5824 * output: device index from FIB lookup
5829 /* inputs to lookup */
5830 __u8 tos; /* AF_INET */
5831 __be32 flowinfo; /* AF_INET6, flow_label + priority */
5833 /* output: metric of fib result (IPv4/IPv6 only) */
5839 __u32 ipv6_src[4]; /* in6_addr; network order */
5842 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
5843 * network header. output: bpf_fib_lookup sets to gateway address
5844 * if FIB lookup returns gateway route
5848 __u32 ipv6_dst[4]; /* in6_addr; network order */
5852 __be16 h_vlan_proto;
5854 __u8 smac[6]; /* ETH_ALEN */
5855 __u8 dmac[6]; /* ETH_ALEN */
5858 struct bpf_redir_neigh {
5859 /* network family for lookup (AF_INET, AF_INET6) */
5861 /* network address of nexthop; skips fib lookup to find gateway */
5864 __u32 ipv6_nh[4]; /* in6_addr; network order */
5868 /* bpf_check_mtu flags*/
5869 enum bpf_check_mtu_flags {
5870 BPF_MTU_CHK_SEGS = (1U << 0),
5873 enum bpf_check_mtu_ret {
5874 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
5875 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5876 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
5879 enum bpf_task_fd_type {
5880 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
5881 BPF_FD_TYPE_TRACEPOINT, /* tp name */
5882 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
5883 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
5884 BPF_FD_TYPE_UPROBE, /* filename + offset */
5885 BPF_FD_TYPE_URETPROBE, /* filename + offset */
5889 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
5890 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
5891 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
5894 struct bpf_flow_keys {
5897 __u16 addr_proto; /* ETH_P_* of valid addrs */
5911 __u32 ipv6_src[4]; /* in6_addr; network order */
5912 __u32 ipv6_dst[4]; /* in6_addr; network order */
5919 struct bpf_func_info {
5924 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
5925 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
5927 struct bpf_line_info {
5929 __u32 file_name_off;
5934 struct bpf_spin_lock {
5939 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
5940 * Allows 1,2,4-byte read, but no write.
5942 __u32 file_pos; /* Sysctl file position to read from, write to.
5943 * Allows 1,2,4-byte read an 4-byte write.
5947 struct bpf_sockopt {
5948 __bpf_md_ptr(struct bpf_sock *, sk);
5949 __bpf_md_ptr(void *, optval);
5950 __bpf_md_ptr(void *, optval_end);
5958 struct bpf_pidns_info {
5963 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
5964 struct bpf_sk_lookup {
5966 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
5967 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
5970 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
5971 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
5972 __u32 remote_ip4; /* Network byte order */
5973 __u32 remote_ip6[4]; /* Network byte order */
5974 __u32 remote_port; /* Network byte order */
5975 __u32 local_ip4; /* Network byte order */
5976 __u32 local_ip6[4]; /* Network byte order */
5977 __u32 local_port; /* Host byte order */
5981 * struct btf_ptr is used for typed pointer representation; the
5982 * type id is used to render the pointer data as the appropriate type
5983 * via the bpf_snprintf_btf() helper described above. A flags field -
5984 * potentially to specify additional details about the BTF pointer
5985 * (rather than its mode of display) - is included for future use.
5986 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
5991 __u32 flags; /* BTF ptr flags; unused at present. */
5995 * Flags to control bpf_snprintf_btf() behaviour.
5996 * - BTF_F_COMPACT: no formatting around type information
5997 * - BTF_F_NONAME: no struct/union member names/types
5998 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
5999 * equivalent to %px.
6000 * - BTF_F_ZERO: show zero-valued struct/union members; they
6001 * are not displayed by default
6004 BTF_F_COMPACT = (1ULL << 0),
6005 BTF_F_NONAME = (1ULL << 1),
6006 BTF_F_PTR_RAW = (1ULL << 2),
6007 BTF_F_ZERO = (1ULL << 3),
6010 #endif /* _UAPI__LINUX_BPF_H__ */