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_MEMSX 0x80 /* load with sign extension */
23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */
27 #define BPF_MOV 0xb0 /* mov reg to reg */
28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
30 /* change endianness of a register */
31 #define BPF_END 0xd0 /* flags for endianness conversion: */
32 #define BPF_TO_LE 0x00 /* convert to little-endian */
33 #define BPF_TO_BE 0x08 /* convert to big-endian */
34 #define BPF_FROM_LE BPF_TO_LE
35 #define BPF_FROM_BE BPF_TO_BE
38 #define BPF_JNE 0x50 /* jump != */
39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
45 #define BPF_JCOND 0xe0 /* conditional pseudo jumps: may_goto, goto_or_nop */
46 #define BPF_CALL 0x80 /* function call */
47 #define BPF_EXIT 0x90 /* function return */
49 /* atomic op type fields (stored in immediate) */
50 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
51 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
52 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
54 enum bpf_cond_pseudo_jmp
{
58 /* Register numbers */
74 /* BPF has 10 general purpose 64-bit registers and stack frame. */
75 #define MAX_BPF_REG __MAX_BPF_REG
78 __u8 code
; /* opcode */
79 __u8 dst_reg
:4; /* dest register */
80 __u8 src_reg
:4; /* source register */
81 __s16 off
; /* signed offset */
82 __s32 imm
; /* signed immediate constant */
85 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
86 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
87 * the trailing flexible array member) instead.
89 struct bpf_lpm_trie_key
{
90 __u32 prefixlen
; /* up to 32 for AF_INET, 128 for AF_INET6 */
91 __u8 data
[0]; /* Arbitrary size */
94 /* Header for bpf_lpm_trie_key structs */
95 struct bpf_lpm_trie_key_hdr
{
99 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
100 struct bpf_lpm_trie_key_u8
{
102 struct bpf_lpm_trie_key_hdr hdr
;
105 __u8 data
[]; /* Arbitrary size */
108 struct bpf_cgroup_storage_key
{
109 __u64 cgroup_inode_id
; /* cgroup inode id */
110 __u32 attach_type
; /* program attach type (enum bpf_attach_type) */
113 enum bpf_cgroup_iter_order
{
114 BPF_CGROUP_ITER_ORDER_UNSPEC
= 0,
115 BPF_CGROUP_ITER_SELF_ONLY
, /* process only a single object. */
116 BPF_CGROUP_ITER_DESCENDANTS_PRE
, /* walk descendants in pre-order. */
117 BPF_CGROUP_ITER_DESCENDANTS_POST
, /* walk descendants in post-order. */
118 BPF_CGROUP_ITER_ANCESTORS_UP
, /* walk ancestors upward. */
121 union bpf_iter_link_info
{
126 enum bpf_cgroup_iter_order order
;
128 /* At most one of cgroup_fd and cgroup_id can be non-zero. If
129 * both are zero, the walk starts from the default cgroup v2
130 * root. For walking v1 hierarchy, one should always explicitly
136 /* Parameters of task iterators. */
144 /* BPF syscall commands, see bpf(2) man-page for more details. */
146 * DOC: eBPF Syscall Preamble
148 * The operation to be performed by the **bpf**\ () system call is determined
149 * by the *cmd* argument. Each operation takes an accompanying argument,
150 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
151 * below). The size argument is the size of the union pointed to by *attr*.
154 * DOC: eBPF Syscall Commands
158 * Create a map and return a file descriptor that refers to the
159 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
160 * is automatically enabled for the new file descriptor.
162 * Applying **close**\ (2) to the file descriptor returned by
163 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
166 * A new file descriptor (a nonnegative integer), or -1 if an
167 * error occurred (in which case, *errno* is set appropriately).
169 * BPF_MAP_LOOKUP_ELEM
171 * Look up an element with a given *key* in the map referred to
172 * by the file descriptor *map_fd*.
174 * The *flags* argument may be specified as one of the
178 * Look up the value of a spin-locked map without
179 * returning the lock. This must be specified if the
180 * elements contain a spinlock.
183 * Returns zero on success. On error, -1 is returned and *errno*
184 * is set appropriately.
186 * BPF_MAP_UPDATE_ELEM
188 * Create or update an element (key/value pair) in a specified map.
190 * The *flags* argument should be specified as one of the
194 * Create a new element or update an existing element.
196 * Create a new element only if it did not exist.
198 * Update an existing element.
200 * Update a spin_lock-ed map element.
203 * Returns zero on success. On error, -1 is returned and *errno*
204 * is set appropriately.
206 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
207 * **E2BIG**, **EEXIST**, or **ENOENT**.
210 * The number of elements in the map reached the
211 * *max_entries* limit specified at map creation time.
213 * If *flags* specifies **BPF_NOEXIST** and the element
214 * with *key* already exists in the map.
216 * If *flags* specifies **BPF_EXIST** and the element with
217 * *key* does not exist in the map.
219 * BPF_MAP_DELETE_ELEM
221 * Look up and delete an element by key in a specified map.
224 * Returns zero on success. On error, -1 is returned and *errno*
225 * is set appropriately.
227 * BPF_MAP_GET_NEXT_KEY
229 * Look up an element by key in a specified map and return the key
230 * of the next element. Can be used to iterate over all elements
234 * Returns zero on success. On error, -1 is returned and *errno*
235 * is set appropriately.
237 * The following cases can be used to iterate over all elements of
240 * * If *key* is not found, the operation returns zero and sets
241 * the *next_key* pointer to the key of the first element.
242 * * If *key* is found, the operation returns zero and sets the
243 * *next_key* pointer to the key of the next element.
244 * * If *key* is the last element, returns -1 and *errno* is set
247 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
248 * **EINVAL** on error.
252 * Verify and load an eBPF program, returning a new file
253 * descriptor associated with the program.
255 * Applying **close**\ (2) to the file descriptor returned by
256 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
258 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
259 * automatically enabled for the new file descriptor.
262 * A new file descriptor (a nonnegative integer), or -1 if an
263 * error occurred (in which case, *errno* is set appropriately).
267 * Pin an eBPF program or map referred by the specified *bpf_fd*
268 * to the provided *pathname* on the filesystem.
270 * The *pathname* argument must not contain a dot (".").
272 * On success, *pathname* retains a reference to the eBPF object,
273 * preventing deallocation of the object when the original
274 * *bpf_fd* is closed. This allow the eBPF object to live beyond
275 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
278 * Applying **unlink**\ (2) or similar calls to the *pathname*
279 * unpins the object from the filesystem, removing the reference.
280 * If no other file descriptors or filesystem nodes refer to the
281 * same object, it will be deallocated (see NOTES).
283 * The filesystem type for the parent directory of *pathname* must
284 * be **BPF_FS_MAGIC**.
287 * Returns zero on success. On error, -1 is returned and *errno*
288 * is set appropriately.
292 * Open a file descriptor for the eBPF object pinned to the
293 * specified *pathname*.
296 * A new file descriptor (a nonnegative integer), or -1 if an
297 * error occurred (in which case, *errno* is set appropriately).
301 * Attach an eBPF program to a *target_fd* at the specified
302 * *attach_type* hook.
304 * The *attach_type* specifies the eBPF attachment point to
305 * attach the program to, and must be one of *bpf_attach_type*
308 * The *attach_bpf_fd* must be a valid file descriptor for a
309 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
310 * or sock_ops type corresponding to the specified *attach_type*.
312 * The *target_fd* must be a valid file descriptor for a kernel
313 * object which depends on the attach type of *attach_bpf_fd*:
315 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
316 * **BPF_PROG_TYPE_CGROUP_SKB**,
317 * **BPF_PROG_TYPE_CGROUP_SOCK**,
318 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
319 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
320 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
321 * **BPF_PROG_TYPE_SOCK_OPS**
323 * Control Group v2 hierarchy with the eBPF controller
324 * enabled. Requires the kernel to be compiled with
325 * **CONFIG_CGROUP_BPF**.
327 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
329 * Network namespace (eg /proc/self/ns/net).
331 * **BPF_PROG_TYPE_LIRC_MODE2**
333 * LIRC device path (eg /dev/lircN). Requires the kernel
334 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
336 * **BPF_PROG_TYPE_SK_SKB**,
337 * **BPF_PROG_TYPE_SK_MSG**
339 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
342 * Returns zero on success. On error, -1 is returned and *errno*
343 * is set appropriately.
347 * Detach the eBPF program associated with the *target_fd* at the
348 * hook specified by *attach_type*. The program must have been
349 * previously attached using **BPF_PROG_ATTACH**.
352 * Returns zero on success. On error, -1 is returned and *errno*
353 * is set appropriately.
357 * Run the eBPF program associated with the *prog_fd* a *repeat*
358 * number of times against a provided program context *ctx_in* and
359 * data *data_in*, and return the modified program context
360 * *ctx_out*, *data_out* (for example, packet data), result of the
361 * execution *retval*, and *duration* of the test run.
363 * The sizes of the buffers provided as input and output
364 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
365 * be provided in the corresponding variables *ctx_size_in*,
366 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
367 * of these parameters are not provided (ie set to NULL), the
368 * corresponding size field must be zero.
370 * Some program types have particular requirements:
372 * **BPF_PROG_TYPE_SK_LOOKUP**
373 * *data_in* and *data_out* must be NULL.
375 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
376 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
378 * *ctx_out*, *data_in* and *data_out* must be NULL.
379 * *repeat* must be zero.
381 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
384 * Returns zero on success. On error, -1 is returned and *errno*
385 * is set appropriately.
388 * Either *data_size_out* or *ctx_size_out* is too small.
390 * This command is not supported by the program type of
391 * the program referred to by *prog_fd*.
393 * BPF_PROG_GET_NEXT_ID
395 * Fetch the next eBPF program currently loaded into the kernel.
397 * Looks for the eBPF program with an id greater than *start_id*
398 * and updates *next_id* on success. If no other eBPF programs
399 * remain with ids higher than *start_id*, returns -1 and sets
400 * *errno* to **ENOENT**.
403 * Returns zero on success. On error, or when no id remains, -1
404 * is returned and *errno* is set appropriately.
406 * BPF_MAP_GET_NEXT_ID
408 * Fetch the next eBPF map currently loaded into the kernel.
410 * Looks for the eBPF map with an id greater than *start_id*
411 * and updates *next_id* on success. If no other eBPF maps
412 * remain with ids higher than *start_id*, returns -1 and sets
413 * *errno* to **ENOENT**.
416 * Returns zero on success. On error, or when no id remains, -1
417 * is returned and *errno* is set appropriately.
419 * BPF_PROG_GET_FD_BY_ID
421 * Open a file descriptor for the eBPF program corresponding to
425 * A new file descriptor (a nonnegative integer), or -1 if an
426 * error occurred (in which case, *errno* is set appropriately).
428 * BPF_MAP_GET_FD_BY_ID
430 * Open a file descriptor for the eBPF map corresponding to
434 * A new file descriptor (a nonnegative integer), or -1 if an
435 * error occurred (in which case, *errno* is set appropriately).
437 * BPF_OBJ_GET_INFO_BY_FD
439 * Obtain information about the eBPF object corresponding to
442 * Populates up to *info_len* bytes of *info*, which will be in
443 * one of the following formats depending on the eBPF object type
446 * * **struct bpf_prog_info**
447 * * **struct bpf_map_info**
448 * * **struct bpf_btf_info**
449 * * **struct bpf_link_info**
452 * Returns zero on success. On error, -1 is returned and *errno*
453 * is set appropriately.
457 * Obtain information about eBPF programs associated with the
458 * specified *attach_type* hook.
460 * The *target_fd* must be a valid file descriptor for a kernel
461 * object which depends on the attach type of *attach_bpf_fd*:
463 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
464 * **BPF_PROG_TYPE_CGROUP_SKB**,
465 * **BPF_PROG_TYPE_CGROUP_SOCK**,
466 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
467 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
468 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
469 * **BPF_PROG_TYPE_SOCK_OPS**
471 * Control Group v2 hierarchy with the eBPF controller
472 * enabled. Requires the kernel to be compiled with
473 * **CONFIG_CGROUP_BPF**.
475 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
477 * Network namespace (eg /proc/self/ns/net).
479 * **BPF_PROG_TYPE_LIRC_MODE2**
481 * LIRC device path (eg /dev/lircN). Requires the kernel
482 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
484 * **BPF_PROG_QUERY** always fetches the number of programs
485 * attached and the *attach_flags* which were used to attach those
486 * programs. Additionally, if *prog_ids* is nonzero and the number
487 * of attached programs is less than *prog_cnt*, populates
488 * *prog_ids* with the eBPF program ids of the programs attached
491 * The following flags may alter the result:
493 * **BPF_F_QUERY_EFFECTIVE**
494 * Only return information regarding programs which are
495 * currently effective at the specified *target_fd*.
498 * Returns zero on success. On error, -1 is returned and *errno*
499 * is set appropriately.
501 * BPF_RAW_TRACEPOINT_OPEN
503 * Attach an eBPF program to a tracepoint *name* to access kernel
504 * internal arguments of the tracepoint in their raw form.
506 * The *prog_fd* must be a valid file descriptor associated with
507 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
509 * No ABI guarantees are made about the content of tracepoint
510 * arguments exposed to the corresponding eBPF program.
512 * Applying **close**\ (2) to the file descriptor returned by
513 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
516 * A new file descriptor (a nonnegative integer), or -1 if an
517 * error occurred (in which case, *errno* is set appropriately).
521 * Verify and load BPF Type Format (BTF) metadata into the kernel,
522 * returning a new file descriptor associated with the metadata.
523 * BTF is described in more detail at
524 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
526 * The *btf* parameter must point to valid memory providing
527 * *btf_size* bytes of BTF binary metadata.
529 * The returned file descriptor can be passed to other **bpf**\ ()
530 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
531 * associate the BTF with those objects.
533 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
534 * parameters to specify a *btf_log_buf*, *btf_log_size* and
535 * *btf_log_level* which allow the kernel to return freeform log
536 * output regarding the BTF verification process.
539 * A new file descriptor (a nonnegative integer), or -1 if an
540 * error occurred (in which case, *errno* is set appropriately).
542 * BPF_BTF_GET_FD_BY_ID
544 * Open a file descriptor for the BPF Type Format (BTF)
545 * corresponding to *btf_id*.
548 * A new file descriptor (a nonnegative integer), or -1 if an
549 * error occurred (in which case, *errno* is set appropriately).
553 * Obtain information about eBPF programs associated with the
554 * target process identified by *pid* and *fd*.
556 * If the *pid* and *fd* are associated with a tracepoint, kprobe
557 * or uprobe perf event, then the *prog_id* and *fd_type* will
558 * be populated with the eBPF program id and file descriptor type
559 * of type **bpf_task_fd_type**. If associated with a kprobe or
560 * uprobe, the *probe_offset* and *probe_addr* will also be
561 * populated. Optionally, if *buf* is provided, then up to
562 * *buf_len* bytes of *buf* will be populated with the name of
563 * the tracepoint, kprobe or uprobe.
565 * The resulting *prog_id* may be introspected in deeper detail
566 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
569 * Returns zero on success. On error, -1 is returned and *errno*
570 * is set appropriately.
572 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
574 * Look up an element with the given *key* in the map referred to
575 * by the file descriptor *fd*, and if found, delete the element.
577 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
578 * types, the *flags* argument needs to be set to 0, but for other
579 * map types, it may be specified as:
582 * Look up and delete the value of a spin-locked map
583 * without returning the lock. This must be specified if
584 * the elements contain a spinlock.
586 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
587 * implement this command as a "pop" operation, deleting the top
588 * element rather than one corresponding to *key*.
589 * The *key* and *key_len* parameters should be zeroed when
590 * issuing this operation for these map types.
592 * This command is only valid for the following map types:
593 * * **BPF_MAP_TYPE_QUEUE**
594 * * **BPF_MAP_TYPE_STACK**
595 * * **BPF_MAP_TYPE_HASH**
596 * * **BPF_MAP_TYPE_PERCPU_HASH**
597 * * **BPF_MAP_TYPE_LRU_HASH**
598 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
601 * Returns zero on success. On error, -1 is returned and *errno*
602 * is set appropriately.
606 * Freeze the permissions of the specified map.
608 * Write permissions may be frozen by passing zero *flags*.
609 * Upon success, no future syscall invocations may alter the
610 * map state of *map_fd*. Write operations from eBPF programs
611 * are still possible for a frozen map.
613 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
616 * Returns zero on success. On error, -1 is returned and *errno*
617 * is set appropriately.
619 * BPF_BTF_GET_NEXT_ID
621 * Fetch the next BPF Type Format (BTF) object currently loaded
624 * Looks for the BTF object with an id greater than *start_id*
625 * and updates *next_id* on success. If no other BTF objects
626 * remain with ids higher than *start_id*, returns -1 and sets
627 * *errno* to **ENOENT**.
630 * Returns zero on success. On error, or when no id remains, -1
631 * is returned and *errno* is set appropriately.
633 * BPF_MAP_LOOKUP_BATCH
635 * Iterate and fetch multiple elements in a map.
637 * Two opaque values are used to manage batch operations,
638 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
639 * to NULL to begin the batched operation. After each subsequent
640 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
641 * *out_batch* as the *in_batch* for the next operation to
642 * continue iteration from the current point. Both *in_batch* and
643 * *out_batch* must point to memory large enough to hold a key,
644 * except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,
645 * LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters
646 * must be at least 4 bytes wide regardless of key size.
648 * The *keys* and *values* are output parameters which must point
649 * to memory large enough to hold *count* items based on the key
650 * and value size of the map *map_fd*. The *keys* buffer must be
651 * of *key_size* * *count*. The *values* buffer must be of
652 * *value_size* * *count*.
654 * The *elem_flags* argument may be specified as one of the
658 * Look up the value of a spin-locked map without
659 * returning the lock. This must be specified if the
660 * elements contain a spinlock.
662 * On success, *count* elements from the map are copied into the
663 * user buffer, with the keys copied into *keys* and the values
664 * copied into the corresponding indices in *values*.
666 * If an error is returned and *errno* is not **EFAULT**, *count*
667 * is set to the number of successfully processed elements.
670 * Returns zero on success. On error, -1 is returned and *errno*
671 * is set appropriately.
673 * May set *errno* to **ENOSPC** to indicate that *keys* or
674 * *values* is too small to dump an entire bucket during
675 * iteration of a hash-based map type.
677 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
679 * Iterate and delete all elements in a map.
681 * This operation has the same behavior as
682 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
684 * * Every element that is successfully returned is also deleted
685 * from the map. This is at least *count* elements. Note that
686 * *count* is both an input and an output parameter.
687 * * Upon returning with *errno* set to **EFAULT**, up to
688 * *count* elements may be deleted without returning the keys
689 * and values of the deleted elements.
692 * Returns zero on success. On error, -1 is returned and *errno*
693 * is set appropriately.
695 * BPF_MAP_UPDATE_BATCH
697 * Update multiple elements in a map by *key*.
699 * The *keys* and *values* are input parameters which must point
700 * to memory large enough to hold *count* items based on the key
701 * and value size of the map *map_fd*. The *keys* buffer must be
702 * of *key_size* * *count*. The *values* buffer must be of
703 * *value_size* * *count*.
705 * Each element specified in *keys* is sequentially updated to the
706 * value in the corresponding index in *values*. The *in_batch*
707 * and *out_batch* parameters are ignored and should be zeroed.
709 * The *elem_flags* argument should be specified as one of the
713 * Create new elements or update a existing elements.
715 * Create new elements only if they do not exist.
717 * Update existing elements.
719 * Update spin_lock-ed map elements. This must be
720 * specified if the map value contains a spinlock.
722 * On success, *count* elements from the map are updated.
724 * If an error is returned and *errno* is not **EFAULT**, *count*
725 * is set to the number of successfully processed elements.
728 * Returns zero on success. On error, -1 is returned and *errno*
729 * is set appropriately.
731 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
732 * **E2BIG**. **E2BIG** indicates that the number of elements in
733 * the map reached the *max_entries* limit specified at map
736 * May set *errno* to one of the following error codes under
737 * specific circumstances:
740 * If *flags* specifies **BPF_NOEXIST** and the element
741 * with *key* already exists in the map.
743 * If *flags* specifies **BPF_EXIST** and the element with
744 * *key* does not exist in the map.
746 * BPF_MAP_DELETE_BATCH
748 * Delete multiple elements in a map by *key*.
750 * The *keys* parameter is an input parameter which must point
751 * to memory large enough to hold *count* items based on the key
752 * size of the map *map_fd*, that is, *key_size* * *count*.
754 * Each element specified in *keys* is sequentially deleted. The
755 * *in_batch*, *out_batch*, and *values* parameters are ignored
756 * and should be zeroed.
758 * The *elem_flags* argument may be specified as one of the
762 * Look up the value of a spin-locked map without
763 * returning the lock. This must be specified if the
764 * elements contain a spinlock.
766 * On success, *count* elements from the map are updated.
768 * If an error is returned and *errno* is not **EFAULT**, *count*
769 * is set to the number of successfully processed elements. If
770 * *errno* is **EFAULT**, up to *count* elements may be been
774 * Returns zero on success. On error, -1 is returned and *errno*
775 * is set appropriately.
779 * Attach an eBPF program to a *target_fd* at the specified
780 * *attach_type* hook and return a file descriptor handle for
784 * A new file descriptor (a nonnegative integer), or -1 if an
785 * error occurred (in which case, *errno* is set appropriately).
789 * Update the eBPF program in the specified *link_fd* to
793 * Returns zero on success. On error, -1 is returned and *errno*
794 * is set appropriately.
796 * BPF_LINK_GET_FD_BY_ID
798 * Open a file descriptor for the eBPF Link corresponding to
802 * A new file descriptor (a nonnegative integer), or -1 if an
803 * error occurred (in which case, *errno* is set appropriately).
805 * BPF_LINK_GET_NEXT_ID
807 * Fetch the next eBPF link currently loaded into the kernel.
809 * Looks for the eBPF link with an id greater than *start_id*
810 * and updates *next_id* on success. If no other eBPF links
811 * remain with ids higher than *start_id*, returns -1 and sets
812 * *errno* to **ENOENT**.
815 * Returns zero on success. On error, or when no id remains, -1
816 * is returned and *errno* is set appropriately.
820 * Enable eBPF runtime statistics gathering.
822 * Runtime statistics gathering for the eBPF runtime is disabled
823 * by default to minimize the corresponding performance overhead.
824 * This command enables statistics globally.
826 * Multiple programs may independently enable statistics.
827 * After gathering the desired statistics, eBPF runtime statistics
828 * may be disabled again by calling **close**\ (2) for the file
829 * descriptor returned by this function. Statistics will only be
830 * disabled system-wide when all outstanding file descriptors
831 * returned by prior calls for this subcommand are closed.
834 * A new file descriptor (a nonnegative integer), or -1 if an
835 * error occurred (in which case, *errno* is set appropriately).
839 * Create an iterator on top of the specified *link_fd* (as
840 * previously created using **BPF_LINK_CREATE**) and return a
841 * file descriptor that can be used to trigger the iteration.
843 * If the resulting file descriptor is pinned to the filesystem
844 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
845 * for that path will trigger the iterator to read kernel state
846 * using the eBPF program attached to *link_fd*.
849 * A new file descriptor (a nonnegative integer), or -1 if an
850 * error occurred (in which case, *errno* is set appropriately).
854 * Forcefully detach the specified *link_fd* from its
855 * corresponding attachment point.
858 * Returns zero on success. On error, -1 is returned and *errno*
859 * is set appropriately.
863 * Bind a map to the lifetime of an eBPF program.
865 * The map identified by *map_fd* is bound to the program
866 * identified by *prog_fd* and only released when *prog_fd* is
867 * released. This may be used in cases where metadata should be
868 * associated with a program which otherwise does not contain any
869 * references to the map (for example, embedded in the eBPF
870 * program instructions).
873 * Returns zero on success. On error, -1 is returned and *errno*
874 * is set appropriately.
878 * Create BPF token with embedded information about what
879 * BPF-related functionality it allows:
880 * - a set of allowed bpf() syscall commands;
881 * - a set of allowed BPF map types to be created with
882 * BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;
883 * - a set of allowed BPF program types and BPF program attach
884 * types to be loaded with BPF_PROG_LOAD command, if
885 * BPF_PROG_LOAD itself is allowed.
887 * BPF token is created (derived) from an instance of BPF FS,
888 * assuming it has necessary delegation mount options specified.
889 * This BPF token can be passed as an extra parameter to various
890 * bpf() syscall commands to grant BPF subsystem functionality to
891 * unprivileged processes.
893 * When created, BPF token is "associated" with the owning
894 * user namespace of BPF FS instance (super block) that it was
895 * derived from, and subsequent BPF operations performed with
896 * BPF token would be performing capabilities checks (i.e.,
897 * CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within
898 * that user namespace. Without BPF token, such capabilities
899 * have to be granted in init user namespace, making bpf()
900 * syscall incompatible with user namespace, for the most part.
903 * A new file descriptor (a nonnegative integer), or -1 if an
904 * error occurred (in which case, *errno* is set appropriately).
907 * eBPF objects (maps and programs) can be shared between processes.
909 * * After **fork**\ (2), the child inherits file descriptors
910 * referring to the same eBPF objects.
911 * * File descriptors referring to eBPF objects can be transferred over
912 * **unix**\ (7) domain sockets.
913 * * File descriptors referring to eBPF objects can be duplicated in the
914 * usual way, using **dup**\ (2) and similar calls.
915 * * File descriptors referring to eBPF objects can be pinned to the
916 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
918 * An eBPF object is deallocated only after all file descriptors referring
919 * to the object have been closed and no references remain pinned to the
920 * filesystem or attached (for example, bound to a program or device).
927 BPF_MAP_GET_NEXT_KEY
,
934 BPF_PROG_RUN
= BPF_PROG_TEST_RUN
,
935 BPF_PROG_GET_NEXT_ID
,
937 BPF_PROG_GET_FD_BY_ID
,
938 BPF_MAP_GET_FD_BY_ID
,
939 BPF_OBJ_GET_INFO_BY_FD
,
941 BPF_RAW_TRACEPOINT_OPEN
,
943 BPF_BTF_GET_FD_BY_ID
,
945 BPF_MAP_LOOKUP_AND_DELETE_ELEM
,
948 BPF_MAP_LOOKUP_BATCH
,
949 BPF_MAP_LOOKUP_AND_DELETE_BATCH
,
950 BPF_MAP_UPDATE_BATCH
,
951 BPF_MAP_DELETE_BATCH
,
954 BPF_LINK_GET_FD_BY_ID
,
955 BPF_LINK_GET_NEXT_ID
,
968 BPF_MAP_TYPE_PROG_ARRAY
,
969 BPF_MAP_TYPE_PERF_EVENT_ARRAY
,
970 BPF_MAP_TYPE_PERCPU_HASH
,
971 BPF_MAP_TYPE_PERCPU_ARRAY
,
972 BPF_MAP_TYPE_STACK_TRACE
,
973 BPF_MAP_TYPE_CGROUP_ARRAY
,
974 BPF_MAP_TYPE_LRU_HASH
,
975 BPF_MAP_TYPE_LRU_PERCPU_HASH
,
976 BPF_MAP_TYPE_LPM_TRIE
,
977 BPF_MAP_TYPE_ARRAY_OF_MAPS
,
978 BPF_MAP_TYPE_HASH_OF_MAPS
,
980 BPF_MAP_TYPE_SOCKMAP
,
983 BPF_MAP_TYPE_SOCKHASH
,
984 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED
,
985 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
986 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
987 * both cgroup-attached and other progs and supports all functionality
988 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
989 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
991 BPF_MAP_TYPE_CGROUP_STORAGE
= BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED
,
992 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY
,
993 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED
,
994 /* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
995 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
996 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
997 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
1000 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
= BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED
,
1003 BPF_MAP_TYPE_SK_STORAGE
,
1004 BPF_MAP_TYPE_DEVMAP_HASH
,
1005 BPF_MAP_TYPE_STRUCT_OPS
,
1006 BPF_MAP_TYPE_RINGBUF
,
1007 BPF_MAP_TYPE_INODE_STORAGE
,
1008 BPF_MAP_TYPE_TASK_STORAGE
,
1009 BPF_MAP_TYPE_BLOOM_FILTER
,
1010 BPF_MAP_TYPE_USER_RINGBUF
,
1011 BPF_MAP_TYPE_CGRP_STORAGE
,
1016 /* Note that tracing related programs such as
1017 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
1018 * are not subject to a stable API since kernel internal data
1019 * structures can change from release to release and may
1020 * therefore break existing tracing BPF programs. Tracing BPF
1021 * programs correspond to /a/ specific kernel which is to be
1022 * analyzed, and not /a/ specific kernel /and/ all future ones.
1024 enum bpf_prog_type
{
1025 BPF_PROG_TYPE_UNSPEC
,
1026 BPF_PROG_TYPE_SOCKET_FILTER
,
1027 BPF_PROG_TYPE_KPROBE
,
1028 BPF_PROG_TYPE_SCHED_CLS
,
1029 BPF_PROG_TYPE_SCHED_ACT
,
1030 BPF_PROG_TYPE_TRACEPOINT
,
1032 BPF_PROG_TYPE_PERF_EVENT
,
1033 BPF_PROG_TYPE_CGROUP_SKB
,
1034 BPF_PROG_TYPE_CGROUP_SOCK
,
1035 BPF_PROG_TYPE_LWT_IN
,
1036 BPF_PROG_TYPE_LWT_OUT
,
1037 BPF_PROG_TYPE_LWT_XMIT
,
1038 BPF_PROG_TYPE_SOCK_OPS
,
1039 BPF_PROG_TYPE_SK_SKB
,
1040 BPF_PROG_TYPE_CGROUP_DEVICE
,
1041 BPF_PROG_TYPE_SK_MSG
,
1042 BPF_PROG_TYPE_RAW_TRACEPOINT
,
1043 BPF_PROG_TYPE_CGROUP_SOCK_ADDR
,
1044 BPF_PROG_TYPE_LWT_SEG6LOCAL
,
1045 BPF_PROG_TYPE_LIRC_MODE2
,
1046 BPF_PROG_TYPE_SK_REUSEPORT
,
1047 BPF_PROG_TYPE_FLOW_DISSECTOR
,
1048 BPF_PROG_TYPE_CGROUP_SYSCTL
,
1049 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE
,
1050 BPF_PROG_TYPE_CGROUP_SOCKOPT
,
1051 BPF_PROG_TYPE_TRACING
,
1052 BPF_PROG_TYPE_STRUCT_OPS
,
1055 BPF_PROG_TYPE_SK_LOOKUP
,
1056 BPF_PROG_TYPE_SYSCALL
, /* a program that can execute syscalls */
1057 BPF_PROG_TYPE_NETFILTER
,
1061 enum bpf_attach_type
{
1062 BPF_CGROUP_INET_INGRESS
,
1063 BPF_CGROUP_INET_EGRESS
,
1064 BPF_CGROUP_INET_SOCK_CREATE
,
1065 BPF_CGROUP_SOCK_OPS
,
1066 BPF_SK_SKB_STREAM_PARSER
,
1067 BPF_SK_SKB_STREAM_VERDICT
,
1070 BPF_CGROUP_INET4_BIND
,
1071 BPF_CGROUP_INET6_BIND
,
1072 BPF_CGROUP_INET4_CONNECT
,
1073 BPF_CGROUP_INET6_CONNECT
,
1074 BPF_CGROUP_INET4_POST_BIND
,
1075 BPF_CGROUP_INET6_POST_BIND
,
1076 BPF_CGROUP_UDP4_SENDMSG
,
1077 BPF_CGROUP_UDP6_SENDMSG
,
1081 BPF_CGROUP_UDP4_RECVMSG
,
1082 BPF_CGROUP_UDP6_RECVMSG
,
1083 BPF_CGROUP_GETSOCKOPT
,
1084 BPF_CGROUP_SETSOCKOPT
,
1091 BPF_CGROUP_INET4_GETPEERNAME
,
1092 BPF_CGROUP_INET6_GETPEERNAME
,
1093 BPF_CGROUP_INET4_GETSOCKNAME
,
1094 BPF_CGROUP_INET6_GETSOCKNAME
,
1096 BPF_CGROUP_INET_SOCK_RELEASE
,
1101 BPF_SK_REUSEPORT_SELECT
,
1102 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE
,
1104 BPF_TRACE_KPROBE_MULTI
,
1110 BPF_TRACE_UPROBE_MULTI
,
1111 BPF_CGROUP_UNIX_CONNECT
,
1112 BPF_CGROUP_UNIX_SENDMSG
,
1113 BPF_CGROUP_UNIX_RECVMSG
,
1114 BPF_CGROUP_UNIX_GETPEERNAME
,
1115 BPF_CGROUP_UNIX_GETSOCKNAME
,
1118 BPF_TRACE_KPROBE_SESSION
,
1119 BPF_TRACE_UPROBE_SESSION
,
1120 __MAX_BPF_ATTACH_TYPE
1123 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1125 /* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]
1126 * in sync with the definitions below.
1128 enum bpf_link_type
{
1129 BPF_LINK_TYPE_UNSPEC
= 0,
1130 BPF_LINK_TYPE_RAW_TRACEPOINT
= 1,
1131 BPF_LINK_TYPE_TRACING
= 2,
1132 BPF_LINK_TYPE_CGROUP
= 3,
1133 BPF_LINK_TYPE_ITER
= 4,
1134 BPF_LINK_TYPE_NETNS
= 5,
1135 BPF_LINK_TYPE_XDP
= 6,
1136 BPF_LINK_TYPE_PERF_EVENT
= 7,
1137 BPF_LINK_TYPE_KPROBE_MULTI
= 8,
1138 BPF_LINK_TYPE_STRUCT_OPS
= 9,
1139 BPF_LINK_TYPE_NETFILTER
= 10,
1140 BPF_LINK_TYPE_TCX
= 11,
1141 BPF_LINK_TYPE_UPROBE_MULTI
= 12,
1142 BPF_LINK_TYPE_NETKIT
= 13,
1143 BPF_LINK_TYPE_SOCKMAP
= 14,
1144 __MAX_BPF_LINK_TYPE
,
1147 #define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
1149 enum bpf_perf_event_type
{
1150 BPF_PERF_EVENT_UNSPEC
= 0,
1151 BPF_PERF_EVENT_UPROBE
= 1,
1152 BPF_PERF_EVENT_URETPROBE
= 2,
1153 BPF_PERF_EVENT_KPROBE
= 3,
1154 BPF_PERF_EVENT_KRETPROBE
= 4,
1155 BPF_PERF_EVENT_TRACEPOINT
= 5,
1156 BPF_PERF_EVENT_EVENT
= 6,
1159 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1161 * NONE(default): No further bpf programs allowed in the subtree.
1163 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1164 * the program in this cgroup yields to sub-cgroup program.
1166 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1167 * that cgroup program gets run in addition to the program in this cgroup.
1169 * Only one program is allowed to be attached to a cgroup with
1170 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1171 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1172 * release old program and attach the new one. Attach flags has to match.
1174 * Multiple programs are allowed to be attached to a cgroup with
1175 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1176 * (those that were attached first, run first)
1177 * The programs of sub-cgroup are executed first, then programs of
1178 * this cgroup and then programs of parent cgroup.
1179 * When children program makes decision (like picking TCP CA or sock bind)
1180 * parent program has a chance to override it.
1182 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1183 * programs for a cgroup. Though it's possible to replace an old program at
1184 * any position by also specifying BPF_F_REPLACE flag and position itself in
1185 * replace_bpf_fd attribute. Old program at this position will be released.
1187 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1188 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1190 * cgrp1 (MULTI progs A, B) ->
1191 * cgrp2 (OVERRIDE prog C) ->
1192 * cgrp3 (MULTI prog D) ->
1193 * cgrp4 (OVERRIDE prog E) ->
1194 * cgrp5 (NONE prog F)
1195 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1196 * if prog F is detached, the execution is E,D,A,B
1197 * if prog F and D are detached, the execution is E,A,B
1198 * if prog F, E and D are detached, the execution is C,A,B
1200 * All eligible programs are executed regardless of return code from
1203 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
1204 #define BPF_F_ALLOW_MULTI (1U << 1)
1205 /* Generic attachment flags. */
1206 #define BPF_F_REPLACE (1U << 2)
1207 #define BPF_F_BEFORE (1U << 3)
1208 #define BPF_F_AFTER (1U << 4)
1209 #define BPF_F_ID (1U << 5)
1210 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */
1212 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1213 * verifier will perform strict alignment checking as if the kernel
1214 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1215 * and NET_IP_ALIGN defined to 2.
1217 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
1219 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1220 * verifier will allow any alignment whatsoever. On platforms
1221 * with strict alignment requirements for loads ands stores (such
1222 * as sparc and mips) the verifier validates that all loads and
1223 * stores provably follow this requirement. This flag turns that
1224 * checking and enforcement off.
1226 * It is mostly used for testing when we want to validate the
1227 * context and memory access aspects of the verifier, but because
1228 * of an unaligned access the alignment check would trigger before
1229 * the one we are interested in.
1231 #define BPF_F_ANY_ALIGNMENT (1U << 1)
1233 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1234 * Verifier does sub-register def/use analysis and identifies instructions whose
1235 * def only matters for low 32-bit, high 32-bit is never referenced later
1236 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1237 * that it is safe to ignore clearing high 32-bit for these instructions. This
1238 * saves some back-ends a lot of code-gen. However such optimization is not
1239 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1240 * hence hasn't used verifier's analysis result. But, we really want to have a
1241 * way to be able to verify the correctness of the described optimization on
1242 * x86_64 on which testsuites are frequently exercised.
1244 * So, this flag is introduced. Once it is set, verifier will randomize high
1245 * 32-bit for those instructions who has been identified as safe to ignore them.
1246 * Then, if verifier is not doing correct analysis, such randomization will
1247 * regress tests to expose bugs.
1249 #define BPF_F_TEST_RND_HI32 (1U << 2)
1251 /* The verifier internal test flag. Behavior is undefined */
1252 #define BPF_F_TEST_STATE_FREQ (1U << 3)
1254 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1255 * restrict map and helper usage for such programs. Sleepable BPF programs can
1256 * only be attached to hooks where kernel execution context allows sleeping.
1257 * Such programs are allowed to use helpers that may sleep like
1258 * bpf_copy_from_user().
1260 #define BPF_F_SLEEPABLE (1U << 4)
1262 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1263 * fully support xdp frags.
1265 #define BPF_F_XDP_HAS_FRAGS (1U << 5)
1267 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1268 * program becomes device-bound but can access XDP metadata.
1270 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6)
1272 /* The verifier internal test flag. Behavior is undefined */
1273 #define BPF_F_TEST_REG_INVARIANTS (1U << 7)
1275 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1276 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1279 BPF_F_KPROBE_MULTI_RETURN
= (1U << 0)
1282 /* link_create.uprobe_multi.flags used in LINK_CREATE command for
1283 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1286 BPF_F_UPROBE_MULTI_RETURN
= (1U << 0)
1289 /* link_create.netfilter.flags used in LINK_CREATE command for
1290 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1292 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1294 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1295 * the following extensions:
1297 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1298 * insn[0].imm: map fd or fd_idx
1302 * ldimm64 rewrite: address of map
1303 * verifier type: CONST_PTR_TO_MAP
1305 #define BPF_PSEUDO_MAP_FD 1
1306 #define BPF_PSEUDO_MAP_IDX 5
1308 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1309 * insn[0].imm: map fd or fd_idx
1310 * insn[1].imm: offset into value
1313 * ldimm64 rewrite: address of map[0]+offset
1314 * verifier type: PTR_TO_MAP_VALUE
1316 #define BPF_PSEUDO_MAP_VALUE 2
1317 #define BPF_PSEUDO_MAP_IDX_VALUE 6
1319 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1320 * insn[0].imm: kernel btd id of VAR
1324 * ldimm64 rewrite: address of the kernel variable
1325 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1328 #define BPF_PSEUDO_BTF_ID 3
1329 /* insn[0].src_reg: BPF_PSEUDO_FUNC
1330 * insn[0].imm: insn offset to the func
1334 * ldimm64 rewrite: address of the function
1335 * verifier type: PTR_TO_FUNC.
1337 #define BPF_PSEUDO_FUNC 4
1339 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1340 * offset to another bpf function
1342 #define BPF_PSEUDO_CALL 1
1343 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1344 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1346 #define BPF_PSEUDO_KFUNC_CALL 2
1348 enum bpf_addr_space_cast
{
1349 BPF_ADDR_SPACE_CAST
= 1,
1352 /* flags for BPF_MAP_UPDATE_ELEM command */
1354 BPF_ANY
= 0, /* create new element or update existing */
1355 BPF_NOEXIST
= 1, /* create new element if it didn't exist */
1356 BPF_EXIST
= 2, /* update existing element */
1357 BPF_F_LOCK
= 4, /* spin_lock-ed map_lookup/map_update */
1360 /* flags for BPF_MAP_CREATE command */
1362 BPF_F_NO_PREALLOC
= (1U << 0),
1363 /* Instead of having one common LRU list in the
1364 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1365 * which can scale and perform better.
1366 * Note, the LRU nodes (including free nodes) cannot be moved
1367 * across different LRU lists.
1369 BPF_F_NO_COMMON_LRU
= (1U << 1),
1370 /* Specify numa node during map creation */
1371 BPF_F_NUMA_NODE
= (1U << 2),
1373 /* Flags for accessing BPF object from syscall side. */
1374 BPF_F_RDONLY
= (1U << 3),
1375 BPF_F_WRONLY
= (1U << 4),
1377 /* Flag for stack_map, store build_id+offset instead of pointer */
1378 BPF_F_STACK_BUILD_ID
= (1U << 5),
1380 /* Zero-initialize hash function seed. This should only be used for testing. */
1381 BPF_F_ZERO_SEED
= (1U << 6),
1383 /* Flags for accessing BPF object from program side. */
1384 BPF_F_RDONLY_PROG
= (1U << 7),
1385 BPF_F_WRONLY_PROG
= (1U << 8),
1387 /* Clone map from listener for newly accepted socket */
1388 BPF_F_CLONE
= (1U << 9),
1390 /* Enable memory-mapping BPF map */
1391 BPF_F_MMAPABLE
= (1U << 10),
1393 /* Share perf_event among processes */
1394 BPF_F_PRESERVE_ELEMS
= (1U << 11),
1396 /* Create a map that is suitable to be an inner map with dynamic max entries */
1397 BPF_F_INNER_MAP
= (1U << 12),
1399 /* Create a map that will be registered/unregesitered by the backed bpf_link */
1400 BPF_F_LINK
= (1U << 13),
1402 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1403 BPF_F_PATH_FD
= (1U << 14),
1405 /* Flag for value_type_btf_obj_fd, the fd is available */
1406 BPF_F_VTYPE_BTF_OBJ_FD
= (1U << 15),
1408 /* BPF token FD is passed in a corresponding command's token_fd field */
1409 BPF_F_TOKEN_FD
= (1U << 16),
1411 /* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */
1412 BPF_F_SEGV_ON_FAULT
= (1U << 17),
1414 /* Do not translate kernel bpf_arena pointers to user pointers */
1415 BPF_F_NO_USER_CONV
= (1U << 18),
1418 /* Flags for BPF_PROG_QUERY. */
1420 /* Query effective (directly attached + inherited from ancestor cgroups)
1421 * programs that will be executed for events within a cgroup.
1422 * attach_flags with this flag are always returned 0.
1424 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
1426 /* Flags for BPF_PROG_TEST_RUN */
1428 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1429 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1430 /* If set, XDP frames will be transmitted after processing */
1431 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
1432 /* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */
1433 #define BPF_F_TEST_SKB_CHECKSUM_COMPLETE (1U << 2)
1435 /* type for BPF_ENABLE_STATS */
1436 enum bpf_stats_type
{
1437 /* enabled run_time_ns and run_cnt */
1438 BPF_STATS_RUN_TIME
= 0,
1441 enum bpf_stack_build_id_status
{
1442 /* user space need an empty entry to identify end of a trace */
1443 BPF_STACK_BUILD_ID_EMPTY
= 0,
1444 /* with valid build_id and offset */
1445 BPF_STACK_BUILD_ID_VALID
= 1,
1446 /* couldn't get build_id, fallback to ip */
1447 BPF_STACK_BUILD_ID_IP
= 2,
1450 #define BPF_BUILD_ID_SIZE 20
1451 struct bpf_stack_build_id
{
1453 unsigned char build_id
[BPF_BUILD_ID_SIZE
];
1460 #define BPF_OBJ_NAME_LEN 16U
1463 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1464 __u32 map_type
; /* one of enum bpf_map_type */
1465 __u32 key_size
; /* size of key in bytes */
1466 __u32 value_size
; /* size of value in bytes */
1467 __u32 max_entries
; /* max number of entries in a map */
1468 __u32 map_flags
; /* BPF_MAP_CREATE related
1469 * flags defined above.
1471 __u32 inner_map_fd
; /* fd pointing to the inner map */
1472 __u32 numa_node
; /* numa node (effective only if
1473 * BPF_F_NUMA_NODE is set).
1475 char map_name
[BPF_OBJ_NAME_LEN
];
1476 __u32 map_ifindex
; /* ifindex of netdev to create on */
1477 __u32 btf_fd
; /* fd pointing to a BTF type data */
1478 __u32 btf_key_type_id
; /* BTF type_id of the key */
1479 __u32 btf_value_type_id
; /* BTF type_id of the value */
1480 __u32 btf_vmlinux_value_type_id
;/* BTF type_id of a kernel-
1481 * struct stored as the
1484 /* Any per-map-type extra fields
1486 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1487 * number of hash functions (if 0, the bloom filter will default
1488 * to using 5 hash functions).
1490 * BPF_MAP_TYPE_ARENA - contains the address where user space
1491 * is going to mmap() the arena. It has to be page aligned.
1495 __s32 value_type_btf_obj_fd
; /* fd pointing to a BTF
1497 * btf_vmlinux_value_type_id.
1499 /* BPF token FD to use with BPF_MAP_CREATE operation.
1500 * If provided, map_flags should have BPF_F_TOKEN_FD flag set.
1505 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1509 __aligned_u64 value
;
1510 __aligned_u64 next_key
;
1515 struct { /* struct used by BPF_MAP_*_BATCH commands */
1516 __aligned_u64 in_batch
; /* start batch,
1517 * NULL to start from beginning
1519 __aligned_u64 out_batch
; /* output: next start batch */
1521 __aligned_u64 values
;
1522 __u32 count
; /* input/output:
1523 * input: # of key/value
1525 * output: # of filled elements
1532 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1533 __u32 prog_type
; /* one of enum bpf_prog_type */
1535 __aligned_u64 insns
;
1536 __aligned_u64 license
;
1537 __u32 log_level
; /* verbosity level of verifier */
1538 __u32 log_size
; /* size of user buffer */
1539 __aligned_u64 log_buf
; /* user supplied buffer */
1540 __u32 kern_version
; /* not used */
1542 char prog_name
[BPF_OBJ_NAME_LEN
];
1543 __u32 prog_ifindex
; /* ifindex of netdev to prep for */
1544 /* For some prog types expected attach type must be known at
1545 * load time to verify attach type specific parts of prog
1546 * (context accesses, allowed helpers, etc).
1548 __u32 expected_attach_type
;
1549 __u32 prog_btf_fd
; /* fd pointing to BTF type data */
1550 __u32 func_info_rec_size
; /* userspace bpf_func_info size */
1551 __aligned_u64 func_info
; /* func info */
1552 __u32 func_info_cnt
; /* number of bpf_func_info records */
1553 __u32 line_info_rec_size
; /* userspace bpf_line_info size */
1554 __aligned_u64 line_info
; /* line info */
1555 __u32 line_info_cnt
; /* number of bpf_line_info records */
1556 __u32 attach_btf_id
; /* in-kernel BTF type id to attach to */
1558 /* valid prog_fd to attach to bpf prog */
1559 __u32 attach_prog_fd
;
1560 /* or valid module BTF object fd or 0 to attach to vmlinux */
1561 __u32 attach_btf_obj_fd
;
1563 __u32 core_relo_cnt
; /* number of bpf_core_relo */
1564 __aligned_u64 fd_array
; /* array of FDs */
1565 __aligned_u64 core_relos
;
1566 __u32 core_relo_rec_size
; /* sizeof(struct bpf_core_relo) */
1567 /* output: actual total log contents size (including termintaing zero).
1568 * It could be both larger than original log_size (if log was
1569 * truncated), or smaller (if log buffer wasn't filled completely).
1571 __u32 log_true_size
;
1572 /* BPF token FD to use with BPF_PROG_LOAD operation.
1573 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.
1575 __s32 prog_token_fd
;
1578 struct { /* anonymous struct used by BPF_OBJ_* commands */
1579 __aligned_u64 pathname
;
1582 /* Same as dirfd in openat() syscall; see openat(2)
1583 * manpage for details of path FD and pathname semantics;
1584 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1585 * file_flags field, otherwise it should be set to zero;
1586 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1591 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1593 __u32 target_fd
; /* target object to attach to or ... */
1594 __u32 target_ifindex
; /* target ifindex */
1596 __u32 attach_bpf_fd
;
1599 __u32 replace_bpf_fd
;
1604 __u64 expected_revision
;
1607 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1610 __u32 data_size_in
; /* input: len of data_in */
1611 __u32 data_size_out
; /* input/output: len of data_out
1612 * returns ENOSPC if data_out
1615 __aligned_u64 data_in
;
1616 __aligned_u64 data_out
;
1619 __u32 ctx_size_in
; /* input: len of ctx_in */
1620 __u32 ctx_size_out
; /* input/output: len of ctx_out
1621 * returns ENOSPC if ctx_out
1624 __aligned_u64 ctx_in
;
1625 __aligned_u64 ctx_out
;
1631 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1643 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1649 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1651 __u32 target_fd
; /* target object to query or ... */
1652 __u32 target_ifindex
; /* target ifindex */
1657 __aligned_u64 prog_ids
;
1663 /* output: per-program attach_flags.
1664 * not allowed to be set during effective query.
1666 __aligned_u64 prog_attach_flags
;
1667 __aligned_u64 link_ids
;
1668 __aligned_u64 link_attach_flags
;
1672 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1676 __aligned_u64 cookie
;
1679 struct { /* anonymous struct for BPF_BTF_LOAD */
1681 __aligned_u64 btf_log_buf
;
1684 __u32 btf_log_level
;
1685 /* output: actual total log contents size (including termintaing zero).
1686 * It could be both larger than original log_size (if log was
1687 * truncated), or smaller (if log buffer wasn't filled completely).
1689 __u32 btf_log_true_size
;
1691 /* BPF token FD to use with BPF_BTF_LOAD operation.
1692 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.
1698 __u32 pid
; /* input: pid */
1699 __u32 fd
; /* input: fd */
1700 __u32 flags
; /* input: flags */
1701 __u32 buf_len
; /* input/output: buf len */
1702 __aligned_u64 buf
; /* input/output:
1703 * tp_name for tracepoint
1705 * filename for uprobe
1707 __u32 prog_id
; /* output: prod_id */
1708 __u32 fd_type
; /* output: BPF_FD_TYPE_* */
1709 __u64 probe_offset
; /* output: probe_offset */
1710 __u64 probe_addr
; /* output: probe_addr */
1713 struct { /* struct used by BPF_LINK_CREATE command */
1715 __u32 prog_fd
; /* eBPF program to attach */
1716 __u32 map_fd
; /* struct_ops to attach */
1719 __u32 target_fd
; /* target object to attach to or ... */
1720 __u32 target_ifindex
; /* target ifindex */
1722 __u32 attach_type
; /* attach type */
1723 __u32 flags
; /* extra flags */
1725 __u32 target_btf_id
; /* btf_id of target to attach to */
1727 __aligned_u64 iter_info
; /* extra bpf_iter_link_info */
1728 __u32 iter_info_len
; /* iter_info length */
1731 /* black box user-provided value passed through
1732 * to BPF program at the execution time and
1733 * accessible through bpf_get_attach_cookie() BPF helper
1741 __aligned_u64 addrs
;
1742 __aligned_u64 cookies
;
1745 /* this is overlaid with the target_btf_id above. */
1746 __u32 target_btf_id
;
1747 /* black box user-provided value passed through
1748 * to BPF program at the execution time and
1749 * accessible through bpf_get_attach_cookie() BPF helper
1764 __u64 expected_revision
;
1768 __aligned_u64 offsets
;
1769 __aligned_u64 ref_ctr_offsets
;
1770 __aligned_u64 cookies
;
1780 __u64 expected_revision
;
1785 struct { /* struct used by BPF_LINK_UPDATE command */
1786 __u32 link_fd
; /* link fd */
1788 /* new program fd to update link with */
1790 /* new struct_ops map fd to update link with */
1793 __u32 flags
; /* extra flags */
1795 /* expected link's program fd; is specified only if
1796 * BPF_F_REPLACE flag is set in flags.
1799 /* expected link's map fd; is specified only
1800 * if BPF_F_REPLACE flag is set.
1810 struct { /* struct used by BPF_ENABLE_STATS command */
1814 struct { /* struct used by BPF_ITER_CREATE command */
1819 struct { /* struct used by BPF_PROG_BIND_MAP command */
1822 __u32 flags
; /* extra flags */
1825 struct { /* struct used by BPF_TOKEN_CREATE command */
1830 } __attribute__((aligned(8)));
1832 /* The description below is an attempt at providing documentation to eBPF
1833 * developers about the multiple available eBPF helper functions. It can be
1834 * parsed and used to produce a manual page. The workflow is the following,
1835 * and requires the rst2man utility:
1837 * $ ./scripts/bpf_doc.py \
1838 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1839 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1840 * $ man /tmp/bpf-helpers.7
1842 * Note that in order to produce this external documentation, some RST
1843 * formatting is used in the descriptions to get "bold" and "italics" in
1844 * manual pages. Also note that the few trailing white spaces are
1845 * intentional, removing them would break paragraphs for rst2man.
1847 * Start of BPF helper function descriptions:
1849 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1851 * Perform a lookup in *map* for an entry associated to *key*.
1853 * Map value associated to *key*, or **NULL** if no entry was
1856 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1858 * Add or update the value of the entry associated to *key* in
1859 * *map* with *value*. *flags* is one of:
1862 * The entry for *key* must not exist in the map.
1864 * The entry for *key* must already exist in the map.
1866 * No condition on the existence of the entry for *key*.
1868 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1869 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1870 * elements always exist), the helper would return an error.
1872 * 0 on success, or a negative error in case of failure.
1874 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1876 * Delete entry with *key* from *map*.
1878 * 0 on success, or a negative error in case of failure.
1880 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1882 * For tracing programs, safely attempt to read *size* bytes from
1883 * kernel space address *unsafe_ptr* and store the data in *dst*.
1885 * Generally, use **bpf_probe_read_user**\ () or
1886 * **bpf_probe_read_kernel**\ () instead.
1888 * 0 on success, or a negative error in case of failure.
1890 * u64 bpf_ktime_get_ns(void)
1892 * Return the time elapsed since system boot, in nanoseconds.
1893 * Does not include time the system was suspended.
1894 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1898 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1900 * This helper is a "printk()-like" facility for debugging. It
1901 * prints a message defined by format *fmt* (of size *fmt_size*)
1902 * to file *\/sys/kernel/tracing/trace* from TraceFS, if
1903 * available. It can take up to three additional **u64**
1904 * arguments (as an eBPF helpers, the total number of arguments is
1907 * Each time the helper is called, it appends a line to the trace.
1908 * Lines are discarded while *\/sys/kernel/tracing/trace* is
1909 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
1910 * The format of the trace is customizable, and the exact output
1911 * one will get depends on the options set in
1912 * *\/sys/kernel/tracing/trace_options* (see also the
1913 * *README* file under the same directory). However, it usually
1914 * defaults to something like:
1918 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1922 * * ``telnet`` is the name of the current task.
1923 * * ``470`` is the PID of the current task.
1924 * * ``001`` is the CPU number on which the task is
1926 * * In ``.N..``, each character refers to a set of
1927 * options (whether irqs are enabled, scheduling
1928 * options, whether hard/softirqs are running, level of
1929 * preempt_disabled respectively). **N** means that
1930 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1932 * * ``419421.045894`` is a timestamp.
1933 * * ``0x00000001`` is a fake value used by BPF for the
1934 * instruction pointer register.
1935 * * ``<formatted msg>`` is the message formatted with
1938 * The conversion specifiers supported by *fmt* are similar, but
1939 * more limited than for printk(). They are **%d**, **%i**,
1940 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1941 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1942 * of field, padding with zeroes, etc.) is available, and the
1943 * helper will return **-EINVAL** (but print nothing) if it
1944 * encounters an unknown specifier.
1946 * Also, note that **bpf_trace_printk**\ () is slow, and should
1947 * only be used for debugging purposes. For this reason, a notice
1948 * block (spanning several lines) is printed to kernel logs and
1949 * states that the helper should not be used "for production use"
1950 * the first time this helper is used (or more precisely, when
1951 * **trace_printk**\ () buffers are allocated). For passing values
1952 * to user space, perf events should be preferred.
1954 * The number of bytes written to the buffer, or a negative error
1955 * in case of failure.
1957 * u32 bpf_get_prandom_u32(void)
1959 * Get a pseudo-random number.
1961 * From a security point of view, this helper uses its own
1962 * pseudo-random internal state, and cannot be used to infer the
1963 * seed of other random functions in the kernel. However, it is
1964 * essential to note that the generator used by the helper is not
1965 * cryptographically secure.
1967 * A random 32-bit unsigned value.
1969 * u32 bpf_get_smp_processor_id(void)
1971 * Get the SMP (symmetric multiprocessing) processor id. Note that
1972 * all programs run with migration disabled, which means that the
1973 * SMP processor id is stable during all the execution of the
1976 * The SMP id of the processor running the program.
1980 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1982 * Store *len* bytes from address *from* into the packet
1983 * associated to *skb*, at *offset*. *flags* are a combination of
1984 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1985 * checksum for the packet after storing the bytes) and
1986 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1987 * **->swhash** and *skb*\ **->l4hash** to 0).
1989 * A call to this helper is susceptible to change the underlying
1990 * packet buffer. Therefore, at load time, all checks on pointers
1991 * previously done by the verifier are invalidated and must be
1992 * performed again, if the helper is used in combination with
1993 * direct packet access.
1995 * 0 on success, or a negative error in case of failure.
1997 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1999 * Recompute the layer 3 (e.g. IP) checksum for the packet
2000 * associated to *skb*. Computation is incremental, so the helper
2001 * must know the former value of the header field that was
2002 * modified (*from*), the new value of this field (*to*), and the
2003 * number of bytes (2 or 4) for this field, stored in *size*.
2004 * Alternatively, it is possible to store the difference between
2005 * the previous and the new values of the header field in *to*, by
2006 * setting *from* and *size* to 0. For both methods, *offset*
2007 * indicates the location of the IP checksum within the packet.
2009 * This helper works in combination with **bpf_csum_diff**\ (),
2010 * which does not update the checksum in-place, but offers more
2011 * flexibility and can handle sizes larger than 2 or 4 for the
2012 * checksum to update.
2014 * A call to this helper is susceptible to change the underlying
2015 * packet buffer. Therefore, at load time, all checks on pointers
2016 * previously done by the verifier are invalidated and must be
2017 * performed again, if the helper is used in combination with
2018 * direct packet access.
2020 * 0 on success, or a negative error in case of failure.
2022 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
2024 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
2025 * packet associated to *skb*. Computation is incremental, so the
2026 * helper must know the former value of the header field that was
2027 * modified (*from*), the new value of this field (*to*), and the
2028 * number of bytes (2 or 4) for this field, stored on the lowest
2029 * four bits of *flags*. Alternatively, it is possible to store
2030 * the difference between the previous and the new values of the
2031 * header field in *to*, by setting *from* and the four lowest
2032 * bits of *flags* to 0. For both methods, *offset* indicates the
2033 * location of the IP checksum within the packet. In addition to
2034 * the size of the field, *flags* can be added (bitwise OR) actual
2035 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
2036 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
2037 * for updates resulting in a null checksum the value is set to
2038 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
2039 * the checksum is to be computed against a pseudo-header.
2041 * This helper works in combination with **bpf_csum_diff**\ (),
2042 * which does not update the checksum in-place, but offers more
2043 * flexibility and can handle sizes larger than 2 or 4 for the
2044 * checksum to update.
2046 * A call to this helper is susceptible to change the underlying
2047 * packet buffer. Therefore, at load time, all checks on pointers
2048 * previously done by the verifier are invalidated and must be
2049 * performed again, if the helper is used in combination with
2050 * direct packet access.
2052 * 0 on success, or a negative error in case of failure.
2054 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
2056 * This special helper is used to trigger a "tail call", or in
2057 * other words, to jump into another eBPF program. The same stack
2058 * frame is used (but values on stack and in registers for the
2059 * caller are not accessible to the callee). This mechanism allows
2060 * for program chaining, either for raising the maximum number of
2061 * available eBPF instructions, or to execute given programs in
2062 * conditional blocks. For security reasons, there is an upper
2063 * limit to the number of successive tail calls that can be
2066 * Upon call of this helper, the program attempts to jump into a
2067 * program referenced at index *index* in *prog_array_map*, a
2068 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
2069 * *ctx*, a pointer to the context.
2071 * If the call succeeds, the kernel immediately runs the first
2072 * instruction of the new program. This is not a function call,
2073 * and it never returns to the previous program. If the call
2074 * fails, then the helper has no effect, and the caller continues
2075 * to run its subsequent instructions. A call can fail if the
2076 * destination program for the jump does not exist (i.e. *index*
2077 * is superior to the number of entries in *prog_array_map*), or
2078 * if the maximum number of tail calls has been reached for this
2079 * chain of programs. This limit is defined in the kernel by the
2080 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
2081 * which is currently set to 33.
2083 * 0 on success, or a negative error in case of failure.
2085 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
2087 * Clone and redirect the packet associated to *skb* to another
2088 * net device of index *ifindex*. Both ingress and egress
2089 * interfaces can be used for redirection. The **BPF_F_INGRESS**
2090 * value in *flags* is used to make the distinction (ingress path
2091 * is selected if the flag is present, egress path otherwise).
2092 * This is the only flag supported for now.
2094 * In comparison with **bpf_redirect**\ () helper,
2095 * **bpf_clone_redirect**\ () has the associated cost of
2096 * duplicating the packet buffer, but this can be executed out of
2097 * the eBPF program. Conversely, **bpf_redirect**\ () is more
2098 * efficient, but it is handled through an action code where the
2099 * redirection happens only after the eBPF program has returned.
2101 * A call to this helper is susceptible to change the underlying
2102 * packet buffer. Therefore, at load time, all checks on pointers
2103 * previously done by the verifier are invalidated and must be
2104 * performed again, if the helper is used in combination with
2105 * direct packet access.
2107 * 0 on success, or a negative error in case of failure. Positive
2108 * error indicates a potential drop or congestion in the target
2109 * device. The particular positive error codes are not defined.
2111 * u64 bpf_get_current_pid_tgid(void)
2113 * Get the current pid and tgid.
2115 * A 64-bit integer containing the current tgid and pid, and
2117 * *current_task*\ **->tgid << 32 \|**
2118 * *current_task*\ **->pid**.
2120 * u64 bpf_get_current_uid_gid(void)
2122 * Get the current uid and gid.
2124 * A 64-bit integer containing the current GID and UID, and
2125 * created as such: *current_gid* **<< 32 \|** *current_uid*.
2127 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
2129 * Copy the **comm** attribute of the current task into *buf* of
2130 * *size_of_buf*. The **comm** attribute contains the name of
2131 * the executable (excluding the path) for the current task. The
2132 * *size_of_buf* must be strictly positive. On success, the
2133 * helper makes sure that the *buf* is NUL-terminated. On failure,
2134 * it is filled with zeroes.
2136 * 0 on success, or a negative error in case of failure.
2138 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
2140 * Retrieve the classid for the current task, i.e. for the net_cls
2141 * cgroup to which *skb* belongs.
2143 * This helper can be used on TC egress path, but not on ingress.
2145 * The net_cls cgroup provides an interface to tag network packets
2146 * based on a user-provided identifier for all traffic coming from
2147 * the tasks belonging to the related cgroup. See also the related
2148 * kernel documentation, available from the Linux sources in file
2149 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2151 * The Linux kernel has two versions for cgroups: there are
2152 * cgroups v1 and cgroups v2. Both are available to users, who can
2153 * use a mixture of them, but note that the net_cls cgroup is for
2154 * cgroup v1 only. This makes it incompatible with BPF programs
2155 * run on cgroups, which is a cgroup-v2-only feature (a socket can
2156 * only hold data for one version of cgroups at a time).
2158 * This helper is only available is the kernel was compiled with
2159 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2160 * "**y**" or to "**m**".
2162 * The classid, or 0 for the default unconfigured classid.
2164 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2166 * Push a *vlan_tci* (VLAN tag control information) of protocol
2167 * *vlan_proto* to the packet associated to *skb*, then update
2168 * the checksum. Note that if *vlan_proto* is different from
2169 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2170 * be **ETH_P_8021Q**.
2172 * A call to this helper is susceptible to change the underlying
2173 * packet buffer. Therefore, at load time, all checks on pointers
2174 * previously done by the verifier are invalidated and must be
2175 * performed again, if the helper is used in combination with
2176 * direct packet access.
2178 * 0 on success, or a negative error in case of failure.
2180 * long bpf_skb_vlan_pop(struct sk_buff *skb)
2182 * Pop a VLAN header from the packet associated to *skb*.
2184 * A call to this helper is susceptible to change the underlying
2185 * packet buffer. Therefore, at load time, all checks on pointers
2186 * previously done by the verifier are invalidated and must be
2187 * performed again, if the helper is used in combination with
2188 * direct packet access.
2190 * 0 on success, or a negative error in case of failure.
2192 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2194 * Get tunnel metadata. This helper takes a pointer *key* to an
2195 * empty **struct bpf_tunnel_key** of **size**, that will be
2196 * filled with tunnel metadata for the packet associated to *skb*.
2197 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2198 * indicates that the tunnel is based on IPv6 protocol instead of
2201 * The **struct bpf_tunnel_key** is an object that generalizes the
2202 * principal parameters used by various tunneling protocols into a
2203 * single struct. This way, it can be used to easily make a
2204 * decision based on the contents of the encapsulation header,
2205 * "summarized" in this struct. In particular, it holds the IP
2206 * address of the remote end (IPv4 or IPv6, depending on the case)
2207 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2208 * this struct exposes the *key*\ **->tunnel_id**, which is
2209 * generally mapped to a VNI (Virtual Network Identifier), making
2210 * it programmable together with the **bpf_skb_set_tunnel_key**\
2213 * Let's imagine that the following code is part of a program
2214 * attached to the TC ingress interface, on one end of a GRE
2215 * tunnel, and is supposed to filter out all messages coming from
2216 * remote ends with IPv4 address other than 10.0.0.1:
2221 * struct bpf_tunnel_key key = {};
2223 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2225 * return TC_ACT_SHOT; // drop packet
2227 * if (key.remote_ipv4 != 0x0a000001)
2228 * return TC_ACT_SHOT; // drop packet
2230 * return TC_ACT_OK; // accept packet
2232 * This interface can also be used with all encapsulation devices
2233 * that can operate in "collect metadata" mode: instead of having
2234 * one network device per specific configuration, the "collect
2235 * metadata" mode only requires a single device where the
2236 * configuration can be extracted from this helper.
2238 * This can be used together with various tunnels such as VXLan,
2239 * Geneve, GRE or IP in IP (IPIP).
2241 * 0 on success, or a negative error in case of failure.
2243 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2245 * Populate tunnel metadata for packet associated to *skb.* The
2246 * tunnel metadata is set to the contents of *key*, of *size*. The
2247 * *flags* can be set to a combination of the following values:
2249 * **BPF_F_TUNINFO_IPV6**
2250 * Indicate that the tunnel is based on IPv6 protocol
2252 * **BPF_F_ZERO_CSUM_TX**
2253 * For IPv4 packets, add a flag to tunnel metadata
2254 * indicating that checksum computation should be skipped
2255 * and checksum set to zeroes.
2256 * **BPF_F_DONT_FRAGMENT**
2257 * Add a flag to tunnel metadata indicating that the
2258 * packet should not be fragmented.
2259 * **BPF_F_SEQ_NUMBER**
2260 * Add a flag to tunnel metadata indicating that a
2261 * sequence number should be added to tunnel header before
2262 * sending the packet. This flag was added for GRE
2263 * encapsulation, but might be used with other protocols
2264 * as well in the future.
2265 * **BPF_F_NO_TUNNEL_KEY**
2266 * Add a flag to tunnel metadata indicating that no tunnel
2267 * key should be set in the resulting tunnel header.
2269 * Here is a typical usage on the transmit path:
2273 * struct bpf_tunnel_key key;
2275 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2276 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2278 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
2279 * helper for additional information.
2281 * 0 on success, or a negative error in case of failure.
2283 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2285 * Read the value of a perf event counter. This helper relies on a
2286 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2287 * the perf event counter is selected when *map* is updated with
2288 * perf event file descriptors. The *map* is an array whose size
2289 * is the number of available CPUs, and each cell contains a value
2290 * relative to one CPU. The value to retrieve is indicated by
2291 * *flags*, that contains the index of the CPU to look up, masked
2292 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2293 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2294 * current CPU should be retrieved.
2296 * Note that before Linux 4.13, only hardware perf event can be
2299 * Also, be aware that the newer helper
2300 * **bpf_perf_event_read_value**\ () is recommended over
2301 * **bpf_perf_event_read**\ () in general. The latter has some ABI
2302 * quirks where error and counter value are used as a return code
2303 * (which is wrong to do since ranges may overlap). This issue is
2304 * fixed with **bpf_perf_event_read_value**\ (), which at the same
2305 * time provides more features over the **bpf_perf_event_read**\
2306 * () interface. Please refer to the description of
2307 * **bpf_perf_event_read_value**\ () for details.
2309 * The value of the perf event counter read from the map, or a
2310 * negative error code in case of failure.
2312 * long bpf_redirect(u32 ifindex, u64 flags)
2314 * Redirect the packet to another net device of index *ifindex*.
2315 * This helper is somewhat similar to **bpf_clone_redirect**\
2316 * (), except that the packet is not cloned, which provides
2317 * increased performance.
2319 * Except for XDP, both ingress and egress interfaces can be used
2320 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
2321 * to make the distinction (ingress path is selected if the flag
2322 * is present, egress path otherwise). Currently, XDP only
2323 * supports redirection to the egress interface, and accepts no
2326 * The same effect can also be attained with the more generic
2327 * **bpf_redirect_map**\ (), which uses a BPF map to store the
2328 * redirect target instead of providing it directly to the helper.
2330 * For XDP, the helper returns **XDP_REDIRECT** on success or
2331 * **XDP_ABORTED** on error. For other program types, the values
2332 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2335 * u32 bpf_get_route_realm(struct sk_buff *skb)
2337 * Retrieve the realm or the route, that is to say the
2338 * **tclassid** field of the destination for the *skb*. The
2339 * identifier retrieved is a user-provided tag, similar to the
2340 * one used with the net_cls cgroup (see description for
2341 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
2342 * held by a route (a destination entry), not by a task.
2344 * Retrieving this identifier works with the clsact TC egress hook
2345 * (see also **tc-bpf(8)**), or alternatively on conventional
2346 * classful egress qdiscs, but not on TC ingress path. In case of
2347 * clsact TC egress hook, this has the advantage that, internally,
2348 * the destination entry has not been dropped yet in the transmit
2349 * path. Therefore, the destination entry does not need to be
2350 * artificially held via **netif_keep_dst**\ () for a classful
2351 * qdisc until the *skb* is freed.
2353 * This helper is available only if the kernel was compiled with
2354 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2356 * The realm of the route for the packet associated to *skb*, or 0
2357 * if none was found.
2359 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2361 * Write raw *data* blob into a special BPF perf event held by
2362 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2363 * event must have the following attributes: **PERF_SAMPLE_RAW**
2364 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2365 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2367 * The *flags* are used to indicate the index in *map* for which
2368 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2369 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2370 * to indicate that the index of the current CPU core should be
2373 * The value to write, of *size*, is passed through eBPF stack and
2374 * pointed by *data*.
2376 * The context of the program *ctx* needs also be passed to the
2379 * On user space, a program willing to read the values needs to
2380 * call **perf_event_open**\ () on the perf event (either for
2381 * one or for all CPUs) and to store the file descriptor into the
2382 * *map*. This must be done before the eBPF program can send data
2383 * into it. An example is available in file
2384 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2385 * tree (the eBPF program counterpart is in
2386 * *samples/bpf/trace_output_kern.c*).
2388 * **bpf_perf_event_output**\ () achieves better performance
2389 * than **bpf_trace_printk**\ () for sharing data with user
2390 * space, and is much better suitable for streaming data from eBPF
2393 * Note that this helper is not restricted to tracing use cases
2394 * and can be used with programs attached to TC or XDP as well,
2395 * where it allows for passing data to user space listeners. Data
2398 * * Only custom structs,
2399 * * Only the packet payload, or
2400 * * A combination of both.
2402 * 0 on success, or a negative error in case of failure.
2404 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2406 * This helper was provided as an easy way to load data from a
2407 * packet. It can be used to load *len* bytes from *offset* from
2408 * the packet associated to *skb*, into the buffer pointed by
2411 * Since Linux 4.7, usage of this helper has mostly been replaced
2412 * by "direct packet access", enabling packet data to be
2413 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2414 * pointing respectively to the first byte of packet data and to
2415 * the byte after the last byte of packet data. However, it
2416 * remains useful if one wishes to read large quantities of data
2417 * at once from a packet into the eBPF stack.
2419 * 0 on success, or a negative error in case of failure.
2421 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2423 * Walk a user or a kernel stack and return its id. To achieve
2424 * this, the helper needs *ctx*, which is a pointer to the context
2425 * on which the tracing program is executed, and a pointer to a
2426 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2428 * The last argument, *flags*, holds the number of stack frames to
2429 * skip (from 0 to 255), masked with
2430 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2431 * a combination of the following flags:
2433 * **BPF_F_USER_STACK**
2434 * Collect a user space stack instead of a kernel stack.
2435 * **BPF_F_FAST_STACK_CMP**
2436 * Compare stacks by hash only.
2437 * **BPF_F_REUSE_STACKID**
2438 * If two different stacks hash into the same *stackid*,
2439 * discard the old one.
2441 * The stack id retrieved is a 32 bit long integer handle which
2442 * can be further combined with other data (including other stack
2443 * ids) and used as a key into maps. This can be useful for
2444 * generating a variety of graphs (such as flame graphs or off-cpu
2447 * For walking a stack, this helper is an improvement over
2448 * **bpf_probe_read**\ (), which can be used with unrolled loops
2449 * but is not efficient and consumes a lot of eBPF instructions.
2450 * Instead, **bpf_get_stackid**\ () can collect up to
2451 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2452 * this limit can be controlled with the **sysctl** program, and
2453 * that it should be manually increased in order to profile long
2454 * user stacks (such as stacks for Java programs). To do so, use:
2458 * # sysctl kernel.perf_event_max_stack=<new value>
2460 * The positive or null stack id on success, or a negative error
2461 * in case of failure.
2463 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2465 * Compute a checksum difference, from the raw buffer pointed by
2466 * *from*, of length *from_size* (that must be a multiple of 4),
2467 * towards the raw buffer pointed by *to*, of size *to_size*
2468 * (same remark). An optional *seed* can be added to the value
2469 * (this can be cascaded, the seed may come from a previous call
2472 * This is flexible enough to be used in several ways:
2474 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2475 * checksum, it can be used when pushing new data.
2476 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2477 * checksum, it can be used when removing data from a packet.
2478 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2479 * can be used to compute a diff. Note that *from_size* and
2480 * *to_size* do not need to be equal.
2482 * This helper can be used in combination with
2483 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2484 * which one can feed in the difference computed with
2485 * **bpf_csum_diff**\ ().
2487 * The checksum result, or a negative error code in case of
2490 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2492 * Retrieve tunnel options metadata for the packet associated to
2493 * *skb*, and store the raw tunnel option data to the buffer *opt*
2496 * This helper can be used with encapsulation devices that can
2497 * operate in "collect metadata" mode (please refer to the related
2498 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2499 * more details). A particular example where this can be used is
2500 * in combination with the Geneve encapsulation protocol, where it
2501 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2502 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2503 * the eBPF program. This allows for full customization of these
2506 * The size of the option data retrieved.
2508 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2510 * Set tunnel options metadata for the packet associated to *skb*
2511 * to the option data contained in the raw buffer *opt* of *size*.
2513 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2514 * helper for additional information.
2516 * 0 on success, or a negative error in case of failure.
2518 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2520 * Change the protocol of the *skb* to *proto*. Currently
2521 * supported are transition from IPv4 to IPv6, and from IPv6 to
2522 * IPv4. The helper takes care of the groundwork for the
2523 * transition, including resizing the socket buffer. The eBPF
2524 * program is expected to fill the new headers, if any, via
2525 * **skb_store_bytes**\ () and to recompute the checksums with
2526 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2527 * (). The main case for this helper is to perform NAT64
2528 * operations out of an eBPF program.
2530 * Internally, the GSO type is marked as dodgy so that headers are
2531 * checked and segments are recalculated by the GSO/GRO engine.
2532 * The size for GSO target is adapted as well.
2534 * All values for *flags* are reserved for future usage, and must
2537 * A call to this helper is susceptible to change the underlying
2538 * packet buffer. Therefore, at load time, all checks on pointers
2539 * previously done by the verifier are invalidated and must be
2540 * performed again, if the helper is used in combination with
2541 * direct packet access.
2543 * 0 on success, or a negative error in case of failure.
2545 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2547 * Change the packet type for the packet associated to *skb*. This
2548 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2549 * the eBPF program does not have a write access to *skb*\
2550 * **->pkt_type** beside this helper. Using a helper here allows
2551 * for graceful handling of errors.
2553 * The major use case is to change incoming *skb*s to
2554 * **PACKET_HOST** in a programmatic way instead of having to
2555 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2558 * Note that *type* only allows certain values. At this time, they
2563 * **PACKET_BROADCAST**
2564 * Send packet to all.
2565 * **PACKET_MULTICAST**
2566 * Send packet to group.
2567 * **PACKET_OTHERHOST**
2568 * Send packet to someone else.
2570 * 0 on success, or a negative error in case of failure.
2572 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2574 * Check whether *skb* is a descendant of the cgroup2 held by
2575 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2577 * The return value depends on the result of the test, and can be:
2579 * * 0, if the *skb* failed the cgroup2 descendant test.
2580 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2581 * * A negative error code, if an error occurred.
2583 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2585 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2586 * not set, in particular if the hash was cleared due to mangling,
2587 * recompute this hash. Later accesses to the hash can be done
2588 * directly with *skb*\ **->hash**.
2590 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2591 * prototype with **bpf_skb_change_proto**\ (), or calling
2592 * **bpf_skb_store_bytes**\ () with the
2593 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2594 * the hash and to trigger a new computation for the next call to
2595 * **bpf_get_hash_recalc**\ ().
2599 * u64 bpf_get_current_task(void)
2601 * Get the current task.
2603 * A pointer to the current task struct.
2605 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2607 * Attempt in a safe way to write *len* bytes from the buffer
2608 * *src* to *dst* in memory. It only works for threads that are in
2609 * user context, and *dst* must be a valid user space address.
2611 * This helper should not be used to implement any kind of
2612 * security mechanism because of TOC-TOU attacks, but rather to
2613 * debug, divert, and manipulate execution of semi-cooperative
2616 * Keep in mind that this feature is meant for experiments, and it
2617 * has a risk of crashing the system and running programs.
2618 * Therefore, when an eBPF program using this helper is attached,
2619 * a warning including PID and process name is printed to kernel
2622 * 0 on success, or a negative error in case of failure.
2624 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2626 * Check whether the probe is being run is the context of a given
2627 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2628 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2630 * The return value depends on the result of the test, and can be:
2632 * * 1, if current task belongs to the cgroup2.
2633 * * 0, if current task does not belong to the cgroup2.
2634 * * A negative error code, if an error occurred.
2636 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2638 * Resize (trim or grow) the packet associated to *skb* to the
2639 * new *len*. The *flags* are reserved for future usage, and must
2642 * The basic idea is that the helper performs the needed work to
2643 * change the size of the packet, then the eBPF program rewrites
2644 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2645 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2646 * and others. This helper is a slow path utility intended for
2647 * replies with control messages. And because it is targeted for
2648 * slow path, the helper itself can afford to be slow: it
2649 * implicitly linearizes, unclones and drops offloads from the
2652 * A call to this helper is susceptible to change the underlying
2653 * packet buffer. Therefore, at load time, all checks on pointers
2654 * previously done by the verifier are invalidated and must be
2655 * performed again, if the helper is used in combination with
2656 * direct packet access.
2658 * 0 on success, or a negative error in case of failure.
2660 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2662 * Pull in non-linear data in case the *skb* is non-linear and not
2663 * all of *len* are part of the linear section. Make *len* bytes
2664 * from *skb* readable and writable. If a zero value is passed for
2665 * *len*, then all bytes in the linear part of *skb* will be made
2666 * readable and writable.
2668 * This helper is only needed for reading and writing with direct
2671 * For direct packet access, testing that offsets to access
2672 * are within packet boundaries (test on *skb*\ **->data_end**) is
2673 * susceptible to fail if offsets are invalid, or if the requested
2674 * data is in non-linear parts of the *skb*. On failure the
2675 * program can just bail out, or in the case of a non-linear
2676 * buffer, use a helper to make the data available. The
2677 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2678 * the data. Another one consists in using **bpf_skb_pull_data**
2679 * to pull in once the non-linear parts, then retesting and
2680 * eventually access the data.
2682 * At the same time, this also makes sure the *skb* is uncloned,
2683 * which is a necessary condition for direct write. As this needs
2684 * to be an invariant for the write part only, the verifier
2685 * detects writes and adds a prologue that is calling
2686 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2687 * the very beginning in case it is indeed cloned.
2689 * A call to this helper is susceptible to change the underlying
2690 * packet buffer. Therefore, at load time, all checks on pointers
2691 * previously done by the verifier are invalidated and must be
2692 * performed again, if the helper is used in combination with
2693 * direct packet access.
2695 * 0 on success, or a negative error in case of failure.
2697 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2699 * Add the checksum *csum* into *skb*\ **->csum** in case the
2700 * driver has supplied a checksum for the entire packet into that
2701 * field. Return an error otherwise. This helper is intended to be
2702 * used in combination with **bpf_csum_diff**\ (), in particular
2703 * when the checksum needs to be updated after data has been
2704 * written into the packet through direct packet access.
2706 * The checksum on success, or a negative error code in case of
2709 * void bpf_set_hash_invalid(struct sk_buff *skb)
2711 * Invalidate the current *skb*\ **->hash**. It can be used after
2712 * mangling on headers through direct packet access, in order to
2713 * indicate that the hash is outdated and to trigger a
2714 * recalculation the next time the kernel tries to access this
2715 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2719 * long bpf_get_numa_node_id(void)
2721 * Return the id of the current NUMA node. The primary use case
2722 * for this helper is the selection of sockets for the local NUMA
2723 * node, when the program is attached to sockets using the
2724 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2725 * but the helper is also available to other eBPF program types,
2726 * similarly to **bpf_get_smp_processor_id**\ ().
2728 * The id of current NUMA node.
2730 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2732 * Grows headroom of packet associated to *skb* and adjusts the
2733 * offset of the MAC header accordingly, adding *len* bytes of
2734 * space. It automatically extends and reallocates memory as
2737 * This helper can be used on a layer 3 *skb* to push a MAC header
2738 * for redirection into a layer 2 device.
2740 * All values for *flags* are reserved for future usage, and must
2743 * A call to this helper is susceptible to change the underlying
2744 * packet buffer. Therefore, at load time, all checks on pointers
2745 * previously done by the verifier are invalidated and must be
2746 * performed again, if the helper is used in combination with
2747 * direct packet access.
2749 * 0 on success, or a negative error in case of failure.
2751 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2753 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2754 * it is possible to use a negative value for *delta*. This helper
2755 * can be used to prepare the packet for pushing or popping
2758 * A call to this helper is susceptible to change the underlying
2759 * packet buffer. Therefore, at load time, all checks on pointers
2760 * previously done by the verifier are invalidated and must be
2761 * performed again, if the helper is used in combination with
2762 * direct packet access.
2764 * 0 on success, or a negative error in case of failure.
2766 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2768 * Copy a NUL terminated string from an unsafe kernel address
2769 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2772 * Generally, use **bpf_probe_read_user_str**\ () or
2773 * **bpf_probe_read_kernel_str**\ () instead.
2775 * On success, the strictly positive length of the string,
2776 * including the trailing NUL character. On error, a negative
2779 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2781 * If the **struct sk_buff** pointed by *skb* has a known socket,
2782 * retrieve the cookie (generated by the kernel) of this socket.
2783 * If no cookie has been set yet, generate a new cookie. Once
2784 * generated, the socket cookie remains stable for the life of the
2785 * socket. This helper can be useful for monitoring per socket
2786 * networking traffic statistics as it provides a global socket
2787 * identifier that can be assumed unique.
2789 * A 8-byte long unique number on success, or 0 if the socket
2790 * field is missing inside *skb*.
2792 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2794 * Equivalent to bpf_get_socket_cookie() helper that accepts
2795 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2797 * A 8-byte long unique number.
2799 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2801 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2802 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2804 * A 8-byte long unique number.
2806 * u64 bpf_get_socket_cookie(struct sock *sk)
2808 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2809 * *sk*, but gets socket from a BTF **struct sock**. This helper
2810 * also works for sleepable programs.
2812 * A 8-byte long unique number or 0 if *sk* is NULL.
2814 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2816 * Get the owner UID of the socked associated to *skb*.
2818 * The owner UID of the socket associated to *skb*. If the socket
2819 * is **NULL**, or if it is not a full socket (i.e. if it is a
2820 * time-wait or a request socket instead), **overflowuid** value
2821 * is returned (note that **overflowuid** might also be the actual
2822 * UID value for the socket).
2824 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2826 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2831 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2833 * Emulate a call to **setsockopt()** on the socket associated to
2834 * *bpf_socket*, which must be a full socket. The *level* at
2835 * which the option resides and the name *optname* of the option
2836 * must be specified, see **setsockopt(2)** for more information.
2837 * The option value of length *optlen* is pointed by *optval*.
2839 * *bpf_socket* should be one of the following:
2841 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2842 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
2843 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
2845 * This helper actually implements a subset of **setsockopt()**.
2846 * It supports the following *level*\ s:
2848 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2849 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2850 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2851 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2852 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2853 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2854 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2855 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2856 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2857 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2858 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2859 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2860 * **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.
2861 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2862 * * **IPPROTO_IPV6**, which supports the following *optname*\ s:
2863 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2865 * 0 on success, or a negative error in case of failure.
2867 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2869 * Grow or shrink the room for data in the packet associated to
2870 * *skb* by *len_diff*, and according to the selected *mode*.
2872 * By default, the helper will reset any offloaded checksum
2873 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2874 * by the following flag:
2876 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2877 * checksum data of the skb to CHECKSUM_NONE.
2879 * There are two supported modes at this time:
2881 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2882 * (room space is added or removed between the layer 2 and
2885 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2886 * (room space is added or removed between the layer 3 and
2889 * The following flags are supported at this time:
2891 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2892 * Adjusting mss in this way is not allowed for datagrams.
2894 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2895 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2896 * Any new space is reserved to hold a tunnel header.
2897 * Configure skb offsets and other fields accordingly.
2899 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2900 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2901 * Use with ENCAP_L3 flags to further specify the tunnel type.
2903 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2904 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2905 * type; *len* is the length of the inner MAC header.
2907 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2908 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2909 * L2 type as Ethernet.
2911 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
2912 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
2913 * Indicate the new IP header version after decapsulating the outer
2914 * IP header. Used when the inner and outer IP versions are different.
2916 * A call to this helper is susceptible to change the underlying
2917 * packet buffer. Therefore, at load time, all checks on pointers
2918 * previously done by the verifier are invalidated and must be
2919 * performed again, if the helper is used in combination with
2920 * direct packet access.
2922 * 0 on success, or a negative error in case of failure.
2924 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2926 * Redirect the packet to the endpoint referenced by *map* at
2927 * index *key*. Depending on its type, this *map* can contain
2928 * references to net devices (for forwarding packets through other
2929 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2930 * but this is only implemented for native XDP (with driver
2931 * support) as of this writing).
2933 * The lower two bits of *flags* are used as the return code if
2934 * the map lookup fails. This is so that the return value can be
2935 * one of the XDP program return codes up to **XDP_TX**, as chosen
2936 * by the caller. The higher bits of *flags* can be set to
2937 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2939 * With BPF_F_BROADCAST the packet will be broadcasted to all the
2940 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2941 * interface will be excluded when do broadcasting.
2943 * See also **bpf_redirect**\ (), which only supports redirecting
2944 * to an ifindex, but doesn't require a map to do so.
2946 * **XDP_REDIRECT** on success, or the value of the two lower bits
2947 * of the *flags* argument on error.
2949 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2951 * Redirect the packet to the socket referenced by *map* (of type
2952 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2953 * egress interfaces can be used for redirection. The
2954 * **BPF_F_INGRESS** value in *flags* is used to make the
2955 * distinction (ingress path is selected if the flag is present,
2956 * egress path otherwise). This is the only flag supported for now.
2958 * **SK_PASS** on success, or **SK_DROP** on error.
2960 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2962 * Add an entry to, or update a *map* referencing sockets. The
2963 * *skops* is used as a new value for the entry associated to
2964 * *key*. *flags* is one of:
2967 * The entry for *key* must not exist in the map.
2969 * The entry for *key* must already exist in the map.
2971 * No condition on the existence of the entry for *key*.
2973 * If the *map* has eBPF programs (parser and verdict), those will
2974 * be inherited by the socket being added. If the socket is
2975 * already attached to eBPF programs, this results in an error.
2977 * 0 on success, or a negative error in case of failure.
2979 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2981 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2982 * *delta* (which can be positive or negative). Note that this
2983 * operation modifies the address stored in *xdp_md*\ **->data**,
2984 * so the latter must be loaded only after the helper has been
2987 * The use of *xdp_md*\ **->data_meta** is optional and programs
2988 * are not required to use it. The rationale is that when the
2989 * packet is processed with XDP (e.g. as DoS filter), it is
2990 * possible to push further meta data along with it before passing
2991 * to the stack, and to give the guarantee that an ingress eBPF
2992 * program attached as a TC classifier on the same device can pick
2993 * this up for further post-processing. Since TC works with socket
2994 * buffers, it remains possible to set from XDP the **mark** or
2995 * **priority** pointers, or other pointers for the socket buffer.
2996 * Having this scratch space generic and programmable allows for
2997 * more flexibility as the user is free to store whatever meta
3000 * A call to this helper is susceptible to change the underlying
3001 * packet buffer. Therefore, at load time, all checks on pointers
3002 * previously done by the verifier are invalidated and must be
3003 * performed again, if the helper is used in combination with
3004 * direct packet access.
3006 * 0 on success, or a negative error in case of failure.
3008 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
3010 * Read the value of a perf event counter, and store it into *buf*
3011 * of size *buf_size*. This helper relies on a *map* of type
3012 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
3013 * counter is selected when *map* is updated with perf event file
3014 * descriptors. The *map* is an array whose size is the number of
3015 * available CPUs, and each cell contains a value relative to one
3016 * CPU. The value to retrieve is indicated by *flags*, that
3017 * contains the index of the CPU to look up, masked with
3018 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
3019 * **BPF_F_CURRENT_CPU** to indicate that the value for the
3020 * current CPU should be retrieved.
3022 * This helper behaves in a way close to
3023 * **bpf_perf_event_read**\ () helper, save that instead of
3024 * just returning the value observed, it fills the *buf*
3025 * structure. This allows for additional data to be retrieved: in
3026 * particular, the enabled and running times (in *buf*\
3027 * **->enabled** and *buf*\ **->running**, respectively) are
3028 * copied. In general, **bpf_perf_event_read_value**\ () is
3029 * recommended over **bpf_perf_event_read**\ (), which has some
3030 * ABI issues and provides fewer functionalities.
3032 * These values are interesting, because hardware PMU (Performance
3033 * Monitoring Unit) counters are limited resources. When there are
3034 * more PMU based perf events opened than available counters,
3035 * kernel will multiplex these events so each event gets certain
3036 * percentage (but not all) of the PMU time. In case that
3037 * multiplexing happens, the number of samples or counter value
3038 * will not reflect the case compared to when no multiplexing
3039 * occurs. This makes comparison between different runs difficult.
3040 * Typically, the counter value should be normalized before
3041 * comparing to other experiments. The usual normalization is done
3046 * normalized_counter = counter * t_enabled / t_running
3048 * Where t_enabled is the time enabled for event and t_running is
3049 * the time running for event since last normalization. The
3050 * enabled and running times are accumulated since the perf event
3051 * open. To achieve scaling factor between two invocations of an
3052 * eBPF program, users can use CPU id as the key (which is
3053 * typical for perf array usage model) to remember the previous
3054 * value and do the calculation inside the eBPF program.
3056 * 0 on success, or a negative error in case of failure.
3058 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
3060 * For an eBPF program attached to a perf event, retrieve the
3061 * value of the event counter associated to *ctx* and store it in
3062 * the structure pointed by *buf* and of size *buf_size*. Enabled
3063 * and running times are also stored in the structure (see
3064 * description of helper **bpf_perf_event_read_value**\ () for
3067 * 0 on success, or a negative error in case of failure.
3069 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
3071 * Emulate a call to **getsockopt()** on the socket associated to
3072 * *bpf_socket*, which must be a full socket. The *level* at
3073 * which the option resides and the name *optname* of the option
3074 * must be specified, see **getsockopt(2)** for more information.
3075 * The retrieved value is stored in the structure pointed by
3076 * *opval* and of length *optlen*.
3078 * *bpf_socket* should be one of the following:
3080 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
3081 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
3082 * **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
3084 * This helper actually implements a subset of **getsockopt()**.
3085 * It supports the same set of *optname*\ s that is supported by
3086 * the **bpf_setsockopt**\ () helper. The exceptions are
3087 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and
3088 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
3090 * 0 on success, or a negative error in case of failure.
3092 * long bpf_override_return(struct pt_regs *regs, u64 rc)
3094 * Used for error injection, this helper uses kprobes to override
3095 * the return value of the probed function, and to set it to *rc*.
3096 * The first argument is the context *regs* on which the kprobe
3099 * This helper works by setting the PC (program counter)
3100 * to an override function which is run in place of the original
3101 * probed function. This means the probed function is not run at
3102 * all. The replacement function just returns with the required
3105 * This helper has security implications, and thus is subject to
3106 * restrictions. It is only available if the kernel was compiled
3107 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
3108 * option, and in this case it only works on functions tagged with
3109 * **ALLOW_ERROR_INJECTION** in the kernel code.
3113 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
3115 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
3116 * for the full TCP socket associated to *bpf_sock_ops* to
3119 * The primary use of this field is to determine if there should
3120 * be calls to eBPF programs of type
3121 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
3122 * code. A program of the same type can change its value, per
3123 * connection and as necessary, when the connection is
3124 * established. This field is directly accessible for reading, but
3125 * this helper must be used for updates in order to return an
3126 * error if an eBPF program tries to set a callback that is not
3127 * supported in the current kernel.
3129 * *argval* is a flag array which can combine these flags:
3131 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
3132 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
3133 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
3134 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
3136 * Therefore, this function can be used to clear a callback flag by
3137 * setting the appropriate bit to zero. e.g. to disable the RTO
3140 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
3141 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3143 * Here are some examples of where one could call such eBPF
3147 * * When a packet is retransmitted.
3148 * * When the connection terminates.
3149 * * When a packet is sent.
3150 * * When a packet is received.
3152 * Code **-EINVAL** if the socket is not a full TCP socket;
3153 * otherwise, a positive number containing the bits that could not
3154 * be set is returned (which comes down to 0 if all bits were set
3157 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3159 * This helper is used in programs implementing policies at the
3160 * socket level. If the message *msg* is allowed to pass (i.e. if
3161 * the verdict eBPF program returns **SK_PASS**), redirect it to
3162 * the socket referenced by *map* (of type
3163 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3164 * egress interfaces can be used for redirection. The
3165 * **BPF_F_INGRESS** value in *flags* is used to make the
3166 * distinction (ingress path is selected if the flag is present,
3167 * egress path otherwise). This is the only flag supported for now.
3169 * **SK_PASS** on success, or **SK_DROP** on error.
3171 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3173 * For socket policies, apply the verdict of the eBPF program to
3174 * the next *bytes* (number of bytes) of message *msg*.
3176 * For example, this helper can be used in the following cases:
3178 * * A single **sendmsg**\ () or **sendfile**\ () system call
3179 * contains multiple logical messages that the eBPF program is
3180 * supposed to read and for which it should apply a verdict.
3181 * * An eBPF program only cares to read the first *bytes* of a
3182 * *msg*. If the message has a large payload, then setting up
3183 * and calling the eBPF program repeatedly for all bytes, even
3184 * though the verdict is already known, would create unnecessary
3187 * When called from within an eBPF program, the helper sets a
3188 * counter internal to the BPF infrastructure, that is used to
3189 * apply the last verdict to the next *bytes*. If *bytes* is
3190 * smaller than the current data being processed from a
3191 * **sendmsg**\ () or **sendfile**\ () system call, the first
3192 * *bytes* will be sent and the eBPF program will be re-run with
3193 * the pointer for start of data pointing to byte number *bytes*
3194 * **+ 1**. If *bytes* is larger than the current data being
3195 * processed, then the eBPF verdict will be applied to multiple
3196 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3199 * Note that if a socket closes with the internal counter holding
3200 * a non-zero value, this is not a problem because data is not
3201 * being buffered for *bytes* and is sent as it is received.
3205 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3207 * For socket policies, prevent the execution of the verdict eBPF
3208 * program for message *msg* until *bytes* (byte number) have been
3211 * This can be used when one needs a specific number of bytes
3212 * before a verdict can be assigned, even if the data spans
3213 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3214 * case would be a user calling **sendmsg**\ () repeatedly with
3215 * 1-byte long message segments. Obviously, this is bad for
3216 * performance, but it is still valid. If the eBPF program needs
3217 * *bytes* bytes to validate a header, this helper can be used to
3218 * prevent the eBPF program to be called again until *bytes* have
3223 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3225 * For socket policies, pull in non-linear data from user space
3226 * for *msg* and set pointers *msg*\ **->data** and *msg*\
3227 * **->data_end** to *start* and *end* bytes offsets into *msg*,
3230 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3231 * *msg* it can only parse data that the (**data**, **data_end**)
3232 * pointers have already consumed. For **sendmsg**\ () hooks this
3233 * is likely the first scatterlist element. But for calls relying
3234 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3235 * be the range (**0**, **0**) because the data is shared with
3236 * user space and by default the objective is to avoid allowing
3237 * user space to modify data while (or after) eBPF verdict is
3238 * being decided. This helper can be used to pull in data and to
3239 * set the start and end pointer to given values. Data will be
3240 * copied if necessary (i.e. if data was not linear and if start
3241 * and end pointers do not point to the same chunk).
3243 * A call to this helper is susceptible to change the underlying
3244 * packet buffer. Therefore, at load time, all checks on pointers
3245 * previously done by the verifier are invalidated and must be
3246 * performed again, if the helper is used in combination with
3247 * direct packet access.
3249 * All values for *flags* are reserved for future usage, and must
3252 * 0 on success, or a negative error in case of failure.
3254 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3256 * Bind the socket associated to *ctx* to the address pointed by
3257 * *addr*, of length *addr_len*. This allows for making outgoing
3258 * connection from the desired IP address, which can be useful for
3259 * example when all processes inside a cgroup should use one
3260 * single IP address on a host that has multiple IP configured.
3262 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3263 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3264 * **AF_INET6**). It's advised to pass zero port (**sin_port**
3265 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3266 * behavior and lets the kernel efficiently pick up an unused
3267 * port as long as 4-tuple is unique. Passing non-zero port might
3268 * lead to degraded performance.
3270 * 0 on success, or a negative error in case of failure.
3272 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3274 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3275 * possible to both shrink and grow the packet tail.
3276 * Shrink done via *delta* being a negative integer.
3278 * A call to this helper is susceptible to change the underlying
3279 * packet buffer. Therefore, at load time, all checks on pointers
3280 * previously done by the verifier are invalidated and must be
3281 * performed again, if the helper is used in combination with
3282 * direct packet access.
3284 * 0 on success, or a negative error in case of failure.
3286 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3288 * Retrieve the XFRM state (IP transform framework, see also
3289 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3291 * The retrieved value is stored in the **struct bpf_xfrm_state**
3292 * pointed by *xfrm_state* and of length *size*.
3294 * All values for *flags* are reserved for future usage, and must
3297 * This helper is available only if the kernel was compiled with
3298 * **CONFIG_XFRM** configuration option.
3300 * 0 on success, or a negative error in case of failure.
3302 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3304 * Return a user or a kernel stack in bpf program provided buffer.
3305 * To achieve this, the helper needs *ctx*, which is a pointer
3306 * to the context on which the tracing program is executed.
3307 * To store the stacktrace, the bpf program provides *buf* with
3308 * a nonnegative *size*.
3310 * The last argument, *flags*, holds the number of stack frames to
3311 * skip (from 0 to 255), masked with
3312 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3313 * the following flags:
3315 * **BPF_F_USER_STACK**
3316 * Collect a user space stack instead of a kernel stack.
3317 * **BPF_F_USER_BUILD_ID**
3318 * Collect (build_id, file_offset) instead of ips for user
3319 * stack, only valid if **BPF_F_USER_STACK** is also
3322 * *file_offset* is an offset relative to the beginning
3323 * of the executable or shared object file backing the vma
3324 * which the *ip* falls in. It is *not* an offset relative
3325 * to that object's base address. Accordingly, it must be
3326 * adjusted by adding (sh_addr - sh_offset), where
3327 * sh_{addr,offset} correspond to the executable section
3328 * containing *file_offset* in the object, for comparisons
3329 * to symbols' st_value to be valid.
3331 * **bpf_get_stack**\ () can collect up to
3332 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3333 * to sufficient large buffer size. Note that
3334 * this limit can be controlled with the **sysctl** program, and
3335 * that it should be manually increased in order to profile long
3336 * user stacks (such as stacks for Java programs). To do so, use:
3340 * # sysctl kernel.perf_event_max_stack=<new value>
3342 * The non-negative copied *buf* length equal to or less than
3343 * *size* on success, or a negative error in case of failure.
3345 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3347 * This helper is similar to **bpf_skb_load_bytes**\ () in that
3348 * it provides an easy way to load *len* bytes from *offset*
3349 * from the packet associated to *skb*, into the buffer pointed
3350 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3351 * a fifth argument *start_header* exists in order to select a
3352 * base offset to start from. *start_header* can be one of:
3354 * **BPF_HDR_START_MAC**
3355 * Base offset to load data from is *skb*'s mac header.
3356 * **BPF_HDR_START_NET**
3357 * Base offset to load data from is *skb*'s network header.
3359 * In general, "direct packet access" is the preferred method to
3360 * access packet data, however, this helper is in particular useful
3361 * in socket filters where *skb*\ **->data** does not always point
3362 * to the start of the mac header and where "direct packet access"
3365 * 0 on success, or a negative error in case of failure.
3367 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3369 * Do FIB lookup in kernel tables using parameters in *params*.
3370 * If lookup is successful and result shows packet is to be
3371 * forwarded, the neighbor tables are searched for the nexthop.
3372 * If successful (ie., FIB lookup shows forwarding and nexthop
3373 * is resolved), the nexthop address is returned in ipv4_dst
3374 * or ipv6_dst based on family, smac is set to mac address of
3375 * egress device, dmac is set to nexthop mac address, rt_metric
3376 * is set to metric from route (IPv4/IPv6 only), and ifindex
3377 * is set to the device index of the nexthop from the FIB lookup.
3379 * *plen* argument is the size of the passed in struct.
3380 * *flags* argument can be a combination of one or more of the
3383 * **BPF_FIB_LOOKUP_DIRECT**
3384 * Do a direct table lookup vs full lookup using FIB
3386 * **BPF_FIB_LOOKUP_TBID**
3387 * Used with BPF_FIB_LOOKUP_DIRECT.
3388 * Use the routing table ID present in *params*->tbid
3389 * for the fib lookup.
3390 * **BPF_FIB_LOOKUP_OUTPUT**
3391 * Perform lookup from an egress perspective (default is
3393 * **BPF_FIB_LOOKUP_SKIP_NEIGH**
3394 * Skip the neighbour table lookup. *params*->dmac
3395 * and *params*->smac will not be set as output. A common
3396 * use case is to call **bpf_redirect_neigh**\ () after
3397 * doing **bpf_fib_lookup**\ ().
3398 * **BPF_FIB_LOOKUP_SRC**
3399 * Derive and set source IP addr in *params*->ipv{4,6}_src
3400 * for the nexthop. If the src addr cannot be derived,
3401 * **BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3402 * case, *params*->dmac and *params*->smac are not set either.
3403 * **BPF_FIB_LOOKUP_MARK**
3404 * Use the mark present in *params*->mark for the fib lookup.
3405 * This option should not be used with BPF_FIB_LOOKUP_DIRECT,
3406 * as it only has meaning for full lookups.
3408 * *ctx* is either **struct xdp_md** for XDP programs or
3409 * **struct sk_buff** tc cls_act programs.
3411 * * < 0 if any input argument is invalid
3412 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3413 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3414 * packet is not forwarded or needs assist from full stack
3416 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3417 * was exceeded and output params->mtu_result contains the MTU.
3419 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3421 * Add an entry to, or update a sockhash *map* referencing sockets.
3422 * The *skops* is used as a new value for the entry associated to
3423 * *key*. *flags* is one of:
3426 * The entry for *key* must not exist in the map.
3428 * The entry for *key* must already exist in the map.
3430 * No condition on the existence of the entry for *key*.
3432 * If the *map* has eBPF programs (parser and verdict), those will
3433 * be inherited by the socket being added. If the socket is
3434 * already attached to eBPF programs, this results in an error.
3436 * 0 on success, or a negative error in case of failure.
3438 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3440 * This helper is used in programs implementing policies at the
3441 * socket level. If the message *msg* is allowed to pass (i.e. if
3442 * the verdict eBPF program returns **SK_PASS**), redirect it to
3443 * the socket referenced by *map* (of type
3444 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3445 * egress interfaces can be used for redirection. The
3446 * **BPF_F_INGRESS** value in *flags* is used to make the
3447 * distinction (ingress path is selected if the flag is present,
3448 * egress path otherwise). This is the only flag supported for now.
3450 * **SK_PASS** on success, or **SK_DROP** on error.
3452 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3454 * This helper is used in programs implementing policies at the
3455 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3456 * if the verdict eBPF program returns **SK_PASS**), redirect it
3457 * to the socket referenced by *map* (of type
3458 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3459 * egress interfaces can be used for redirection. The
3460 * **BPF_F_INGRESS** value in *flags* is used to make the
3461 * distinction (ingress path is selected if the flag is present,
3462 * egress otherwise). This is the only flag supported for now.
3464 * **SK_PASS** on success, or **SK_DROP** on error.
3466 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3468 * Encapsulate the packet associated to *skb* within a Layer 3
3469 * protocol header. This header is provided in the buffer at
3470 * address *hdr*, with *len* its size in bytes. *type* indicates
3471 * the protocol of the header and can be one of:
3473 * **BPF_LWT_ENCAP_SEG6**
3474 * IPv6 encapsulation with Segment Routing Header
3475 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3476 * the IPv6 header is computed by the kernel.
3477 * **BPF_LWT_ENCAP_SEG6_INLINE**
3478 * Only works if *skb* contains an IPv6 packet. Insert a
3479 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3481 * **BPF_LWT_ENCAP_IP**
3482 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3483 * must be IPv4 or IPv6, followed by zero or more
3484 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3485 * total bytes in all prepended headers. Please note that
3486 * if **skb_is_gso**\ (*skb*) is true, no more than two
3487 * headers can be prepended, and the inner header, if
3488 * present, should be either GRE or UDP/GUE.
3490 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3491 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3492 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3493 * **BPF_PROG_TYPE_LWT_XMIT**.
3495 * A call to this helper is susceptible to change the underlying
3496 * packet buffer. Therefore, at load time, all checks on pointers
3497 * previously done by the verifier are invalidated and must be
3498 * performed again, if the helper is used in combination with
3499 * direct packet access.
3501 * 0 on success, or a negative error in case of failure.
3503 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3505 * Store *len* bytes from address *from* into the packet
3506 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3507 * inside the outermost IPv6 Segment Routing Header can be
3508 * modified through this helper.
3510 * A call to this helper is susceptible to change the underlying
3511 * packet buffer. Therefore, at load time, all checks on pointers
3512 * previously done by the verifier are invalidated and must be
3513 * performed again, if the helper is used in combination with
3514 * direct packet access.
3516 * 0 on success, or a negative error in case of failure.
3518 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3520 * Adjust the size allocated to TLVs in the outermost IPv6
3521 * Segment Routing Header contained in the packet associated to
3522 * *skb*, at position *offset* by *delta* bytes. Only offsets
3523 * after the segments are accepted. *delta* can be as well
3524 * positive (growing) as negative (shrinking).
3526 * A call to this helper is susceptible to change the underlying
3527 * packet buffer. Therefore, at load time, all checks on pointers
3528 * previously done by the verifier are invalidated and must be
3529 * performed again, if the helper is used in combination with
3530 * direct packet access.
3532 * 0 on success, or a negative error in case of failure.
3534 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3536 * Apply an IPv6 Segment Routing action of type *action* to the
3537 * packet associated to *skb*. Each action takes a parameter
3538 * contained at address *param*, and of length *param_len* bytes.
3539 * *action* can be one of:
3541 * **SEG6_LOCAL_ACTION_END_X**
3542 * End.X action: Endpoint with Layer-3 cross-connect.
3543 * Type of *param*: **struct in6_addr**.
3544 * **SEG6_LOCAL_ACTION_END_T**
3545 * End.T action: Endpoint with specific IPv6 table lookup.
3546 * Type of *param*: **int**.
3547 * **SEG6_LOCAL_ACTION_END_B6**
3548 * End.B6 action: Endpoint bound to an SRv6 policy.
3549 * Type of *param*: **struct ipv6_sr_hdr**.
3550 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3551 * End.B6.Encap action: Endpoint bound to an SRv6
3552 * encapsulation policy.
3553 * Type of *param*: **struct ipv6_sr_hdr**.
3555 * A call to this helper is susceptible to change the underlying
3556 * packet buffer. Therefore, at load time, all checks on pointers
3557 * previously done by the verifier are invalidated and must be
3558 * performed again, if the helper is used in combination with
3559 * direct packet access.
3561 * 0 on success, or a negative error in case of failure.
3563 * long bpf_rc_repeat(void *ctx)
3565 * This helper is used in programs implementing IR decoding, to
3566 * report a successfully decoded repeat key message. This delays
3567 * the generation of a key up event for previously generated
3570 * Some IR protocols like NEC have a special IR message for
3571 * repeating last button, for when a button is held down.
3573 * The *ctx* should point to the lirc sample as passed into
3576 * This helper is only available is the kernel was compiled with
3577 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3582 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3584 * This helper is used in programs implementing IR decoding, to
3585 * report a successfully decoded key press with *scancode*,
3586 * *toggle* value in the given *protocol*. The scancode will be
3587 * translated to a keycode using the rc keymap, and reported as
3588 * an input key down event. After a period a key up event is
3589 * generated. This period can be extended by calling either
3590 * **bpf_rc_keydown**\ () again with the same values, or calling
3591 * **bpf_rc_repeat**\ ().
3593 * Some protocols include a toggle bit, in case the button was
3594 * released and pressed again between consecutive scancodes.
3596 * The *ctx* should point to the lirc sample as passed into
3599 * The *protocol* is the decoded protocol number (see
3600 * **enum rc_proto** for some predefined values).
3602 * This helper is only available is the kernel was compiled with
3603 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3608 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3610 * Return the cgroup v2 id of the socket associated with the *skb*.
3611 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3612 * helper for cgroup v1 by providing a tag resp. identifier that
3613 * can be matched on or used for map lookups e.g. to implement
3614 * policy. The cgroup v2 id of a given path in the hierarchy is
3615 * exposed in user space through the f_handle API in order to get
3616 * to the same 64-bit id.
3618 * This helper can be used on TC egress path, but not on ingress,
3619 * and is available only if the kernel was compiled with the
3620 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3622 * The id is returned or 0 in case the id could not be retrieved.
3624 * u64 bpf_get_current_cgroup_id(void)
3626 * Get the current cgroup id based on the cgroup within which
3627 * the current task is running.
3629 * A 64-bit integer containing the current cgroup id based
3630 * on the cgroup within which the current task is running.
3632 * void *bpf_get_local_storage(void *map, u64 flags)
3634 * Get the pointer to the local storage area.
3635 * The type and the size of the local storage is defined
3636 * by the *map* argument.
3637 * The *flags* meaning is specific for each map type,
3638 * and has to be 0 for cgroup local storage.
3640 * Depending on the BPF program type, a local storage area
3641 * can be shared between multiple instances of the BPF program,
3642 * running simultaneously.
3644 * A user should care about the synchronization by himself.
3645 * For example, by using the **BPF_ATOMIC** instructions to alter
3648 * A pointer to the local storage area.
3650 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3652 * Select a **SO_REUSEPORT** socket from a
3653 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3654 * It checks the selected socket is matching the incoming
3655 * request in the socket buffer.
3657 * 0 on success, or a negative error in case of failure.
3659 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3661 * Return id of cgroup v2 that is ancestor of cgroup associated
3662 * with the *skb* at the *ancestor_level*. The root cgroup is at
3663 * *ancestor_level* zero and each step down the hierarchy
3664 * increments the level. If *ancestor_level* == level of cgroup
3665 * associated with *skb*, then return value will be same as that
3666 * of **bpf_skb_cgroup_id**\ ().
3668 * The helper is useful to implement policies based on cgroups
3669 * that are upper in hierarchy than immediate cgroup associated
3672 * The format of returned id and helper limitations are same as in
3673 * **bpf_skb_cgroup_id**\ ().
3675 * The id is returned or 0 in case the id could not be retrieved.
3677 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3679 * Look for TCP socket matching *tuple*, optionally in a child
3680 * network namespace *netns*. The return value must be checked,
3681 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3683 * The *ctx* should point to the context of the program, such as
3684 * the skb or socket (depending on the hook in use). This is used
3685 * to determine the base network namespace for the lookup.
3687 * *tuple_size* must be one of:
3689 * **sizeof**\ (*tuple*\ **->ipv4**)
3690 * Look for an IPv4 socket.
3691 * **sizeof**\ (*tuple*\ **->ipv6**)
3692 * Look for an IPv6 socket.
3694 * If the *netns* is a negative signed 32-bit integer, then the
3695 * socket lookup table in the netns associated with the *ctx*
3696 * will be used. For the TC hooks, this is the netns of the device
3697 * in the skb. For socket hooks, this is the netns of the socket.
3698 * If *netns* is any other signed 32-bit value greater than or
3699 * equal to zero then it specifies the ID of the netns relative to
3700 * the netns associated with the *ctx*. *netns* values beyond the
3701 * range of 32-bit integers are reserved for future use.
3703 * All values for *flags* are reserved for future usage, and must
3706 * This helper is available only if the kernel was compiled with
3707 * **CONFIG_NET** configuration option.
3709 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3710 * For sockets with reuseport option, the **struct bpf_sock**
3711 * result is from *reuse*\ **->socks**\ [] using the hash of the
3714 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3716 * Look for UDP socket matching *tuple*, optionally in a child
3717 * network namespace *netns*. The return value must be checked,
3718 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3720 * The *ctx* should point to the context of the program, such as
3721 * the skb or socket (depending on the hook in use). This is used
3722 * to determine the base network namespace for the lookup.
3724 * *tuple_size* must be one of:
3726 * **sizeof**\ (*tuple*\ **->ipv4**)
3727 * Look for an IPv4 socket.
3728 * **sizeof**\ (*tuple*\ **->ipv6**)
3729 * Look for an IPv6 socket.
3731 * If the *netns* is a negative signed 32-bit integer, then the
3732 * socket lookup table in the netns associated with the *ctx*
3733 * will be used. For the TC hooks, this is the netns of the device
3734 * in the skb. For socket hooks, this is the netns of the socket.
3735 * If *netns* is any other signed 32-bit value greater than or
3736 * equal to zero then it specifies the ID of the netns relative to
3737 * the netns associated with the *ctx*. *netns* values beyond the
3738 * range of 32-bit integers are reserved for future use.
3740 * All values for *flags* are reserved for future usage, and must
3743 * This helper is available only if the kernel was compiled with
3744 * **CONFIG_NET** configuration option.
3746 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3747 * For sockets with reuseport option, the **struct bpf_sock**
3748 * result is from *reuse*\ **->socks**\ [] using the hash of the
3751 * long bpf_sk_release(void *sock)
3753 * Release the reference held by *sock*. *sock* must be a
3754 * non-**NULL** pointer that was returned from
3755 * **bpf_sk_lookup_xxx**\ ().
3757 * 0 on success, or a negative error in case of failure.
3759 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3761 * Push an element *value* in *map*. *flags* is one of:
3764 * If the queue/stack is full, the oldest element is
3765 * removed to make room for this.
3767 * 0 on success, or a negative error in case of failure.
3769 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3771 * Pop an element from *map*.
3773 * 0 on success, or a negative error in case of failure.
3775 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3777 * Get an element from *map* without removing it.
3779 * 0 on success, or a negative error in case of failure.
3781 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3783 * For socket policies, insert *len* bytes into *msg* at offset
3786 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3787 * *msg* it may want to insert metadata or options into the *msg*.
3788 * This can later be read and used by any of the lower layer BPF
3791 * This helper may fail if under memory pressure (a malloc
3792 * fails) in these cases BPF programs will get an appropriate
3793 * error and BPF programs will need to handle them.
3795 * 0 on success, or a negative error in case of failure.
3797 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3799 * Will remove *len* bytes from a *msg* starting at byte *start*.
3800 * This may result in **ENOMEM** errors under certain situations if
3801 * an allocation and copy are required due to a full ring buffer.
3802 * However, the helper will try to avoid doing the allocation
3803 * if possible. Other errors can occur if input parameters are
3804 * invalid either due to *start* byte not being valid part of *msg*
3805 * payload and/or *pop* value being to large.
3807 * 0 on success, or a negative error in case of failure.
3809 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3811 * This helper is used in programs implementing IR decoding, to
3812 * report a successfully decoded pointer movement.
3814 * The *ctx* should point to the lirc sample as passed into
3817 * This helper is only available is the kernel was compiled with
3818 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3823 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3825 * Acquire a spinlock represented by the pointer *lock*, which is
3826 * stored as part of a value of a map. Taking the lock allows to
3827 * safely update the rest of the fields in that value. The
3828 * spinlock can (and must) later be released with a call to
3829 * **bpf_spin_unlock**\ (\ *lock*\ ).
3831 * Spinlocks in BPF programs come with a number of restrictions
3834 * * **bpf_spin_lock** objects are only allowed inside maps of
3835 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3836 * list could be extended in the future).
3837 * * BTF description of the map is mandatory.
3838 * * The BPF program can take ONE lock at a time, since taking two
3839 * or more could cause dead locks.
3840 * * Only one **struct bpf_spin_lock** is allowed per map element.
3841 * * When the lock is taken, calls (either BPF to BPF or helpers)
3843 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3844 * allowed inside a spinlock-ed region.
3845 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3846 * the lock, on all execution paths, before it returns.
3847 * * The BPF program can access **struct bpf_spin_lock** only via
3848 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3849 * helpers. Loading or storing data into the **struct
3850 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3851 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3852 * of the map value must be a struct and have **struct
3853 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3854 * Nested lock inside another struct is not allowed.
3855 * * The **struct bpf_spin_lock** *lock* field in a map value must
3856 * be aligned on a multiple of 4 bytes in that value.
3857 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3858 * the **bpf_spin_lock** field to user space.
3859 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3860 * a BPF program, do not update the **bpf_spin_lock** field.
3861 * * **bpf_spin_lock** cannot be on the stack or inside a
3862 * networking packet (it can only be inside of a map values).
3863 * * **bpf_spin_lock** is available to root only.
3864 * * Tracing programs and socket filter programs cannot use
3865 * **bpf_spin_lock**\ () due to insufficient preemption checks
3866 * (but this may change in the future).
3867 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3871 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3873 * Release the *lock* previously locked by a call to
3874 * **bpf_spin_lock**\ (\ *lock*\ ).
3878 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3880 * This helper gets a **struct bpf_sock** pointer such
3881 * that all the fields in this **bpf_sock** can be accessed.
3883 * A **struct bpf_sock** pointer on success, or **NULL** in
3886 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3888 * This helper gets a **struct bpf_tcp_sock** pointer from a
3889 * **struct bpf_sock** pointer.
3891 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3894 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3896 * Set ECN (Explicit Congestion Notification) field of IP header
3897 * to **CE** (Congestion Encountered) if current value is **ECT**
3898 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3901 * 1 if the **CE** flag is set (either by the current helper call
3902 * or because it was already present), 0 if it is not set.
3904 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3906 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3907 * **bpf_sk_release**\ () is unnecessary and not allowed.
3909 * A **struct bpf_sock** pointer on success, or **NULL** in
3912 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3914 * Look for TCP socket matching *tuple*, optionally in a child
3915 * network namespace *netns*. The return value must be checked,
3916 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3918 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3919 * that it also returns timewait or request sockets. Use
3920 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3923 * This helper is available only if the kernel was compiled with
3924 * **CONFIG_NET** configuration option.
3926 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3927 * For sockets with reuseport option, the **struct bpf_sock**
3928 * result is from *reuse*\ **->socks**\ [] using the hash of the
3931 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3933 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3934 * the listening socket in *sk*.
3936 * *iph* points to the start of the IPv4 or IPv6 header, while
3937 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3938 * **sizeof**\ (**struct ipv6hdr**).
3940 * *th* points to the start of the TCP header, while *th_len*
3941 * contains the length of the TCP header (at least
3942 * **sizeof**\ (**struct tcphdr**)).
3944 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3947 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3949 * Get name of sysctl in /proc/sys/ and copy it into provided by
3950 * program buffer *buf* of size *buf_len*.
3952 * The buffer is always NUL terminated, unless it's zero-sized.
3954 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3955 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3956 * only (e.g. "tcp_mem").
3958 * Number of character copied (not including the trailing NUL).
3960 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3961 * truncated name in this case).
3963 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3965 * Get current value of sysctl as it is presented in /proc/sys
3966 * (incl. newline, etc), and copy it as a string into provided
3967 * by program buffer *buf* of size *buf_len*.
3969 * The whole value is copied, no matter what file position user
3970 * space issued e.g. sys_read at.
3972 * The buffer is always NUL terminated, unless it's zero-sized.
3974 * Number of character copied (not including the trailing NUL).
3976 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3977 * truncated name in this case).
3979 * **-EINVAL** if current value was unavailable, e.g. because
3980 * sysctl is uninitialized and read returns -EIO for it.
3982 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3984 * Get new value being written by user space to sysctl (before
3985 * the actual write happens) and copy it as a string into
3986 * provided by program buffer *buf* of size *buf_len*.
3988 * User space may write new value at file position > 0.
3990 * The buffer is always NUL terminated, unless it's zero-sized.
3992 * Number of character copied (not including the trailing NUL).
3994 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3995 * truncated name in this case).
3997 * **-EINVAL** if sysctl is being read.
3999 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
4001 * Override new value being written by user space to sysctl with
4002 * value provided by program in buffer *buf* of size *buf_len*.
4004 * *buf* should contain a string in same form as provided by user
4005 * space on sysctl write.
4007 * User space may write new value at file position > 0. To override
4008 * the whole sysctl value file position should be set to zero.
4012 * **-E2BIG** if the *buf_len* is too big.
4014 * **-EINVAL** if sysctl is being read.
4016 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
4018 * Convert the initial part of the string from buffer *buf* of
4019 * size *buf_len* to a long integer according to the given base
4020 * and save the result in *res*.
4022 * The string may begin with an arbitrary amount of white space
4023 * (as determined by **isspace**\ (3)) followed by a single
4024 * optional '**-**' sign.
4026 * Five least significant bits of *flags* encode base, other bits
4027 * are currently unused.
4029 * Base must be either 8, 10, 16 or 0 to detect it automatically
4030 * similar to user space **strtol**\ (3).
4032 * Number of characters consumed on success. Must be positive but
4033 * no more than *buf_len*.
4035 * **-EINVAL** if no valid digits were found or unsupported base
4038 * **-ERANGE** if resulting value was out of range.
4040 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
4042 * Convert the initial part of the string from buffer *buf* of
4043 * size *buf_len* to an unsigned long integer according to the
4044 * given base and save the result in *res*.
4046 * The string may begin with an arbitrary amount of white space
4047 * (as determined by **isspace**\ (3)).
4049 * Five least significant bits of *flags* encode base, other bits
4050 * are currently unused.
4052 * Base must be either 8, 10, 16 or 0 to detect it automatically
4053 * similar to user space **strtoul**\ (3).
4055 * Number of characters consumed on success. Must be positive but
4056 * no more than *buf_len*.
4058 * **-EINVAL** if no valid digits were found or unsupported base
4061 * **-ERANGE** if resulting value was out of range.
4063 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
4065 * Get a bpf-local-storage from a *sk*.
4067 * Logically, it could be thought of getting the value from
4068 * a *map* with *sk* as the **key**. From this
4069 * perspective, the usage is not much different from
4070 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
4071 * helper enforces the key must be a full socket and the map must
4072 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
4074 * Underneath, the value is stored locally at *sk* instead of
4075 * the *map*. The *map* is used as the bpf-local-storage
4076 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4077 * searched against all bpf-local-storages residing at *sk*.
4079 * *sk* is a kernel **struct sock** pointer for LSM program.
4080 * *sk* is a **struct bpf_sock** pointer for other program types.
4082 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
4083 * used such that a new bpf-local-storage will be
4084 * created if one does not exist. *value* can be used
4085 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
4086 * the initial value of a bpf-local-storage. If *value* is
4087 * **NULL**, the new bpf-local-storage will be zero initialized.
4089 * A bpf-local-storage pointer is returned on success.
4091 * **NULL** if not found or there was an error in adding
4092 * a new bpf-local-storage.
4094 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
4096 * Delete a bpf-local-storage from a *sk*.
4100 * **-ENOENT** if the bpf-local-storage cannot be found.
4101 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
4103 * long bpf_send_signal(u32 sig)
4105 * Send signal *sig* to the process of the current task.
4106 * The signal may be delivered to any of this process's threads.
4108 * 0 on success or successfully queued.
4110 * **-EBUSY** if work queue under nmi is full.
4112 * **-EINVAL** if *sig* is invalid.
4114 * **-EPERM** if no permission to send the *sig*.
4116 * **-EAGAIN** if bpf program can try again.
4118 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
4120 * Try to issue a SYN cookie for the packet with corresponding
4121 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
4123 * *iph* points to the start of the IPv4 or IPv6 header, while
4124 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
4125 * **sizeof**\ (**struct ipv6hdr**).
4127 * *th* points to the start of the TCP header, while *th_len*
4128 * contains the length of the TCP header with options (at least
4129 * **sizeof**\ (**struct tcphdr**)).
4131 * On success, lower 32 bits hold the generated SYN cookie in
4132 * followed by 16 bits which hold the MSS value for that cookie,
4133 * and the top 16 bits are unused.
4135 * On failure, the returned value is one of the following:
4137 * **-EINVAL** SYN cookie cannot be issued due to error
4139 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
4141 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
4143 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
4145 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4147 * Write raw *data* blob into a special BPF perf event held by
4148 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4149 * event must have the following attributes: **PERF_SAMPLE_RAW**
4150 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4151 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4153 * The *flags* are used to indicate the index in *map* for which
4154 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4155 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4156 * to indicate that the index of the current CPU core should be
4159 * The value to write, of *size*, is passed through eBPF stack and
4160 * pointed by *data*.
4162 * *ctx* is a pointer to in-kernel struct sk_buff.
4164 * This helper is similar to **bpf_perf_event_output**\ () but
4165 * restricted to raw_tracepoint bpf programs.
4167 * 0 on success, or a negative error in case of failure.
4169 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4171 * Safely attempt to read *size* bytes from user space address
4172 * *unsafe_ptr* and store the data in *dst*.
4174 * 0 on success, or a negative error in case of failure.
4176 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4178 * Safely attempt to read *size* bytes from kernel space address
4179 * *unsafe_ptr* and store the data in *dst*.
4181 * 0 on success, or a negative error in case of failure.
4183 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4185 * Copy a NUL terminated string from an unsafe user address
4186 * *unsafe_ptr* to *dst*. The *size* should include the
4187 * terminating NUL byte. In case the string length is smaller than
4188 * *size*, the target is not padded with further NUL bytes. If the
4189 * string length is larger than *size*, just *size*-1 bytes are
4190 * copied and the last byte is set to NUL.
4192 * On success, returns the number of bytes that were written,
4193 * including the terminal NUL. This makes this helper useful in
4194 * tracing programs for reading strings, and more importantly to
4195 * get its length at runtime. See the following snippet:
4199 * SEC("kprobe/sys_open")
4200 * void bpf_sys_open(struct pt_regs *ctx)
4202 * char buf[PATHLEN]; // PATHLEN is defined to 256
4203 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
4206 * // Consume buf, for example push it to
4207 * // userspace via bpf_perf_event_output(); we
4208 * // can use res (the string length) as event
4209 * // size, after checking its boundaries.
4212 * In comparison, using **bpf_probe_read_user**\ () helper here
4213 * instead to read the string would require to estimate the length
4214 * at compile time, and would often result in copying more memory
4217 * Another useful use case is when parsing individual process
4218 * arguments or individual environment variables navigating
4219 * *current*\ **->mm->arg_start** and *current*\
4220 * **->mm->env_start**: using this helper and the return value,
4221 * one can quickly iterate at the right offset of the memory area.
4223 * On success, the strictly positive length of the output string,
4224 * including the trailing NUL character. On error, a negative
4227 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4229 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4230 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4232 * On success, the strictly positive length of the string, including
4233 * the trailing NUL character. On error, a negative value.
4235 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4237 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4238 * *rcv_nxt* is the ack_seq to be sent out.
4240 * 0 on success, or a negative error in case of failure.
4242 * long bpf_send_signal_thread(u32 sig)
4244 * Send signal *sig* to the thread corresponding to the current task.
4246 * 0 on success or successfully queued.
4248 * **-EBUSY** if work queue under nmi is full.
4250 * **-EINVAL** if *sig* is invalid.
4252 * **-EPERM** if no permission to send the *sig*.
4254 * **-EAGAIN** if bpf program can try again.
4256 * u64 bpf_jiffies64(void)
4258 * Obtain the 64bit jiffies
4260 * The 64 bit jiffies
4262 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4264 * For an eBPF program attached to a perf event, retrieve the
4265 * branch records (**struct perf_branch_entry**) associated to *ctx*
4266 * and store it in the buffer pointed by *buf* up to size
4269 * On success, number of bytes written to *buf*. On error, a
4272 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4273 * instead return the number of bytes required to store all the
4274 * branch entries. If this flag is set, *buf* may be NULL.
4276 * **-EINVAL** if arguments invalid or **size** not a multiple
4277 * of **sizeof**\ (**struct perf_branch_entry**\ ).
4279 * **-ENOENT** if architecture does not support branch records.
4281 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4283 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
4284 * *namespace* will be returned in *nsdata*.
4286 * 0 on success, or one of the following in case of failure:
4288 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
4289 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
4291 * **-ENOENT** if pidns does not exists for the current task.
4293 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4295 * Write raw *data* blob into a special BPF perf event held by
4296 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4297 * event must have the following attributes: **PERF_SAMPLE_RAW**
4298 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4299 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4301 * The *flags* are used to indicate the index in *map* for which
4302 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4303 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4304 * to indicate that the index of the current CPU core should be
4307 * The value to write, of *size*, is passed through eBPF stack and
4308 * pointed by *data*.
4310 * *ctx* is a pointer to in-kernel struct xdp_buff.
4312 * This helper is similar to **bpf_perf_eventoutput**\ () but
4313 * restricted to raw_tracepoint bpf programs.
4315 * 0 on success, or a negative error in case of failure.
4317 * u64 bpf_get_netns_cookie(void *ctx)
4319 * Retrieve the cookie (generated by the kernel) of the network
4320 * namespace the input *ctx* is associated with. The network
4321 * namespace cookie remains stable for its lifetime and provides
4322 * a global identifier that can be assumed unique. If *ctx* is
4323 * NULL, then the helper returns the cookie for the initial
4324 * network namespace. The cookie itself is very similar to that
4325 * of **bpf_get_socket_cookie**\ () helper, but for network
4326 * namespaces instead of sockets.
4328 * A 8-byte long opaque number.
4330 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4332 * Return id of cgroup v2 that is ancestor of the cgroup associated
4333 * with the current task at the *ancestor_level*. The root cgroup
4334 * is at *ancestor_level* zero and each step down the hierarchy
4335 * increments the level. If *ancestor_level* == level of cgroup
4336 * associated with the current task, then return value will be the
4337 * same as that of **bpf_get_current_cgroup_id**\ ().
4339 * The helper is useful to implement policies based on cgroups
4340 * that are upper in hierarchy than immediate cgroup associated
4341 * with the current task.
4343 * The format of returned id and helper limitations are same as in
4344 * **bpf_get_current_cgroup_id**\ ().
4346 * The id is returned or 0 in case the id could not be retrieved.
4348 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4350 * Helper is overloaded depending on BPF program type. This
4351 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4352 * **BPF_PROG_TYPE_SCHED_ACT** programs.
4354 * Assign the *sk* to the *skb*. When combined with appropriate
4355 * routing configuration to receive the packet towards the socket,
4356 * will cause *skb* to be delivered to the specified socket.
4357 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
4358 * **bpf_clone_redirect**\ () or other methods outside of BPF may
4359 * interfere with successful delivery to the socket.
4361 * This operation is only valid from TC ingress path.
4363 * The *flags* argument must be zero.
4365 * 0 on success, or a negative error in case of failure:
4367 * **-EINVAL** if specified *flags* are not supported.
4369 * **-ENOENT** if the socket is unavailable for assignment.
4371 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
4373 * **-EOPNOTSUPP** if the operation is not supported, for example
4374 * a call from outside of TC ingress.
4376 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4378 * Helper is overloaded depending on BPF program type. This
4379 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4381 * Select the *sk* as a result of a socket lookup.
4383 * For the operation to succeed passed socket must be compatible
4384 * with the packet description provided by the *ctx* object.
4386 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4387 * be an exact match. While IP family (**AF_INET** or
4388 * **AF_INET6**) must be compatible, that is IPv6 sockets
4389 * that are not v6-only can be selected for IPv4 packets.
4391 * Only TCP listeners and UDP unconnected sockets can be
4392 * selected. *sk* can also be NULL to reset any previous
4395 * *flags* argument can combination of following values:
4397 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4398 * socket selection, potentially done by a BPF program
4399 * that ran before us.
4401 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4402 * load-balancing within reuseport group for the socket
4405 * On success *ctx->sk* will point to the selected socket.
4408 * 0 on success, or a negative errno in case of failure.
4410 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4411 * not compatible with packet family (*ctx->family*).
4413 * * **-EEXIST** if socket has been already selected,
4414 * potentially by another program, and
4415 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4417 * * **-EINVAL** if unsupported flags were specified.
4419 * * **-EPROTOTYPE** if socket L4 protocol
4420 * (*sk->protocol*) doesn't match packet protocol
4421 * (*ctx->protocol*).
4423 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4424 * state (TCP listening or UDP unconnected).
4426 * u64 bpf_ktime_get_boot_ns(void)
4428 * Return the time elapsed since system boot, in nanoseconds.
4429 * Does include the time the system was suspended.
4430 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4434 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4436 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4437 * out the format string.
4438 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4439 * the format string itself. The *data* and *data_len* are format string
4440 * arguments. The *data* are a **u64** array and corresponding format string
4441 * values are stored in the array. For strings and pointers where pointees
4442 * are accessed, only the pointer values are stored in the *data* array.
4443 * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4445 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4446 * Reading kernel memory may fail due to either invalid address or
4447 * valid address but requiring a major memory fault. If reading kernel memory
4448 * fails, the string for **%s** will be an empty string, and the ip
4449 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4450 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4452 * 0 on success, or a negative error in case of failure:
4454 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4455 * by returning 1 from bpf program.
4457 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4459 * **-E2BIG** if *fmt* contains too many format specifiers.
4461 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4463 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4465 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4466 * The *m* represents the seq_file. The *data* and *len* represent the
4467 * data to write in bytes.
4469 * 0 on success, or a negative error in case of failure:
4471 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4473 * u64 bpf_sk_cgroup_id(void *sk)
4475 * Return the cgroup v2 id of the socket *sk*.
4477 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4478 * returned from **bpf_sk_lookup_xxx**\ (),
4479 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4480 * same as in **bpf_skb_cgroup_id**\ ().
4482 * This helper is available only if the kernel was compiled with
4483 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4485 * The id is returned or 0 in case the id could not be retrieved.
4487 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4489 * Return id of cgroup v2 that is ancestor of cgroup associated
4490 * with the *sk* at the *ancestor_level*. The root cgroup is at
4491 * *ancestor_level* zero and each step down the hierarchy
4492 * increments the level. If *ancestor_level* == level of cgroup
4493 * associated with *sk*, then return value will be same as that
4494 * of **bpf_sk_cgroup_id**\ ().
4496 * The helper is useful to implement policies based on cgroups
4497 * that are upper in hierarchy than immediate cgroup associated
4500 * The format of returned id and helper limitations are same as in
4501 * **bpf_sk_cgroup_id**\ ().
4503 * The id is returned or 0 in case the id could not be retrieved.
4505 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4507 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4508 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4509 * of new data availability is sent.
4510 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4511 * of new data availability is sent unconditionally.
4512 * If **0** is specified in *flags*, an adaptive notification
4513 * of new data availability is sent.
4515 * An adaptive notification is a notification sent whenever the user-space
4516 * process has caught up and consumed all available payloads. In case the user-space
4517 * process is still processing a previous payload, then no notification is needed
4518 * as it will process the newly added payload automatically.
4520 * 0 on success, or a negative error in case of failure.
4522 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4524 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4525 * *flags* must be 0.
4527 * Valid pointer with *size* bytes of memory available; NULL,
4530 * void bpf_ringbuf_submit(void *data, u64 flags)
4532 * Submit reserved ring buffer sample, pointed to by *data*.
4533 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4534 * of new data availability is sent.
4535 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4536 * of new data availability is sent unconditionally.
4537 * If **0** is specified in *flags*, an adaptive notification
4538 * of new data availability is sent.
4540 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4542 * Nothing. Always succeeds.
4544 * void bpf_ringbuf_discard(void *data, u64 flags)
4546 * Discard reserved ring buffer sample, pointed to by *data*.
4547 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4548 * of new data availability is sent.
4549 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4550 * of new data availability is sent unconditionally.
4551 * If **0** is specified in *flags*, an adaptive notification
4552 * of new data availability is sent.
4554 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4556 * Nothing. Always succeeds.
4558 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4560 * Query various characteristics of provided ring buffer. What
4561 * exactly is queries is determined by *flags*:
4563 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4564 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4565 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4566 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4568 * Data returned is just a momentary snapshot of actual values
4569 * and could be inaccurate, so this facility should be used to
4570 * power heuristics and for reporting, not to make 100% correct
4573 * Requested value, or 0, if *flags* are not recognized.
4575 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4577 * Change the skbs checksum level by one layer up or down, or
4578 * reset it entirely to none in order to have the stack perform
4579 * checksum validation. The level is applicable to the following
4580 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4581 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4582 * through **bpf_skb_adjust_room**\ () helper with passing in
4583 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4584 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4585 * the UDP header is removed. Similarly, an encap of the latter
4586 * into the former could be accompanied by a helper call to
4587 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4588 * skb is still intended to be processed in higher layers of the
4589 * stack instead of just egressing at tc.
4591 * There are three supported level settings at this time:
4593 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4594 * with CHECKSUM_UNNECESSARY.
4595 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4596 * with CHECKSUM_UNNECESSARY.
4597 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4598 * sets CHECKSUM_NONE to force checksum validation by the stack.
4599 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4602 * 0 on success, or a negative error in case of failure. In the
4603 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4604 * is returned or the error code -EACCES in case the skb is not
4605 * subject to CHECKSUM_UNNECESSARY.
4607 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4609 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4611 * *sk* if casting is valid, or **NULL** otherwise.
4613 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4615 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4617 * *sk* if casting is valid, or **NULL** otherwise.
4619 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4621 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4623 * *sk* if casting is valid, or **NULL** otherwise.
4625 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4627 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4629 * *sk* if casting is valid, or **NULL** otherwise.
4631 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4633 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4635 * *sk* if casting is valid, or **NULL** otherwise.
4637 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4639 * Return a user or a kernel stack in bpf program provided buffer.
4640 * Note: the user stack will only be populated if the *task* is
4641 * the current task; all other tasks will return -EOPNOTSUPP.
4642 * To achieve this, the helper needs *task*, which is a valid
4643 * pointer to **struct task_struct**. To store the stacktrace, the
4644 * bpf program provides *buf* with a nonnegative *size*.
4646 * The last argument, *flags*, holds the number of stack frames to
4647 * skip (from 0 to 255), masked with
4648 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4649 * the following flags:
4651 * **BPF_F_USER_STACK**
4652 * Collect a user space stack instead of a kernel stack.
4653 * The *task* must be the current task.
4654 * **BPF_F_USER_BUILD_ID**
4655 * Collect buildid+offset instead of ips for user stack,
4656 * only valid if **BPF_F_USER_STACK** is also specified.
4658 * **bpf_get_task_stack**\ () can collect up to
4659 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4660 * to sufficient large buffer size. Note that
4661 * this limit can be controlled with the **sysctl** program, and
4662 * that it should be manually increased in order to profile long
4663 * user stacks (such as stacks for Java programs). To do so, use:
4667 * # sysctl kernel.perf_event_max_stack=<new value>
4669 * The non-negative copied *buf* length equal to or less than
4670 * *size* on success, or a negative error in case of failure.
4672 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4674 * Load header option. Support reading a particular TCP header
4675 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4677 * If *flags* is 0, it will search the option from the
4678 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4679 * has details on what skb_data contains under different
4680 * *skops*\ **->op**.
4682 * The first byte of the *searchby_res* specifies the
4683 * kind that it wants to search.
4685 * If the searching kind is an experimental kind
4686 * (i.e. 253 or 254 according to RFC6994). It also
4687 * needs to specify the "magic" which is either
4688 * 2 bytes or 4 bytes. It then also needs to
4689 * specify the size of the magic by using
4690 * the 2nd byte which is "kind-length" of a TCP
4691 * header option and the "kind-length" also
4692 * includes the first 2 bytes "kind" and "kind-length"
4693 * itself as a normal TCP header option also does.
4695 * For example, to search experimental kind 254 with
4696 * 2 byte magic 0xeB9F, the searchby_res should be
4697 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4699 * To search for the standard window scale option (3),
4700 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4701 * Note, kind-length must be 0 for regular option.
4703 * Searching for No-Op (0) and End-of-Option-List (1) are
4706 * *len* must be at least 2 bytes which is the minimal size
4707 * of a header option.
4711 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4712 * saved_syn packet or the just-received syn packet.
4715 * > 0 when found, the header option is copied to *searchby_res*.
4716 * The return value is the total length copied. On failure, a
4717 * negative error code is returned:
4719 * **-EINVAL** if a parameter is invalid.
4721 * **-ENOMSG** if the option is not found.
4723 * **-ENOENT** if no syn packet is available when
4724 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4726 * **-ENOSPC** if there is not enough space. Only *len* number of
4729 * **-EFAULT** on failure to parse the header options in the
4732 * **-EPERM** if the helper cannot be used under the current
4733 * *skops*\ **->op**.
4735 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4737 * Store header option. The data will be copied
4738 * from buffer *from* with length *len* to the TCP header.
4740 * The buffer *from* should have the whole option that
4741 * includes the kind, kind-length, and the actual
4742 * option data. The *len* must be at least kind-length
4743 * long. The kind-length does not have to be 4 byte
4744 * aligned. The kernel will take care of the padding
4745 * and setting the 4 bytes aligned value to th->doff.
4747 * This helper will check for duplicated option
4748 * by searching the same option in the outgoing skb.
4750 * This helper can only be called during
4751 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4754 * 0 on success, or negative error in case of failure:
4756 * **-EINVAL** If param is invalid.
4758 * **-ENOSPC** if there is not enough space in the header.
4759 * Nothing has been written
4761 * **-EEXIST** if the option already exists.
4763 * **-EFAULT** on failure to parse the existing header options.
4765 * **-EPERM** if the helper cannot be used under the current
4766 * *skops*\ **->op**.
4768 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4770 * Reserve *len* bytes for the bpf header option. The
4771 * space will be used by **bpf_store_hdr_opt**\ () later in
4772 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4774 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4775 * the total number of bytes will be reserved.
4777 * This helper can only be called during
4778 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4781 * 0 on success, or negative error in case of failure:
4783 * **-EINVAL** if a parameter is invalid.
4785 * **-ENOSPC** if there is not enough space in the header.
4787 * **-EPERM** if the helper cannot be used under the current
4788 * *skops*\ **->op**.
4790 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4792 * Get a bpf_local_storage from an *inode*.
4794 * Logically, it could be thought of as getting the value from
4795 * a *map* with *inode* as the **key**. From this
4796 * perspective, the usage is not much different from
4797 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4798 * helper enforces the key must be an inode and the map must also
4799 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4801 * Underneath, the value is stored locally at *inode* instead of
4802 * the *map*. The *map* is used as the bpf-local-storage
4803 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4804 * searched against all bpf_local_storage residing at *inode*.
4806 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4807 * used such that a new bpf_local_storage will be
4808 * created if one does not exist. *value* can be used
4809 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4810 * the initial value of a bpf_local_storage. If *value* is
4811 * **NULL**, the new bpf_local_storage will be zero initialized.
4813 * A bpf_local_storage pointer is returned on success.
4815 * **NULL** if not found or there was an error in adding
4816 * a new bpf_local_storage.
4818 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4820 * Delete a bpf_local_storage from an *inode*.
4824 * **-ENOENT** if the bpf_local_storage cannot be found.
4826 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4828 * Return full path for given **struct path** object, which
4829 * needs to be the kernel BTF *path* object. The path is
4830 * returned in the provided buffer *buf* of size *sz* and
4831 * is zero terminated.
4834 * On success, the strictly positive length of the string,
4835 * including the trailing NUL character. On error, a negative
4838 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4840 * Read *size* bytes from user space address *user_ptr* and store
4841 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4843 * 0 on success, or a negative error in case of failure.
4845 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4847 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4848 * using *ptr*->type_id. This value should specify the type
4849 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4850 * can be used to look up vmlinux BTF type ids. Traversing the
4851 * data structure using BTF, the type information and values are
4852 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4853 * the pointer data is carried out to avoid kernel crashes during
4854 * operation. Smaller types can use string space on the stack;
4855 * larger programs can use map data to store the string
4858 * The string can be subsequently shared with userspace via
4859 * bpf_perf_event_output() or ring buffer interfaces.
4860 * bpf_trace_printk() is to be avoided as it places too small
4861 * a limit on string size to be useful.
4863 * *flags* is a combination of
4866 * no formatting around type information
4868 * no struct/union member names/types
4870 * show raw (unobfuscated) pointer values;
4871 * equivalent to printk specifier %px.
4873 * show zero-valued struct/union members; they
4874 * are not displayed by default
4877 * The number of bytes that were written (or would have been
4878 * written if output had to be truncated due to string size),
4879 * or a negative error in cases of failure.
4881 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4883 * Use BTF to write to seq_write a string representation of
4884 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4885 * *flags* are identical to those used for bpf_snprintf_btf.
4887 * 0 on success or a negative error in case of failure.
4889 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4891 * See **bpf_get_cgroup_classid**\ () for the main description.
4892 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4893 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4894 * associated socket instead of the current process.
4896 * The id is returned or 0 in case the id could not be retrieved.
4898 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4900 * Redirect the packet to another net device of index *ifindex*
4901 * and fill in L2 addresses from neighboring subsystem. This helper
4902 * is somewhat similar to **bpf_redirect**\ (), except that it
4903 * populates L2 addresses as well, meaning, internally, the helper
4904 * relies on the neighbor lookup for the L2 address of the nexthop.
4906 * The helper will perform a FIB lookup based on the skb's
4907 * networking header to get the address of the next hop, unless
4908 * this is supplied by the caller in the *params* argument. The
4909 * *plen* argument indicates the len of *params* and should be set
4910 * to 0 if *params* is NULL.
4912 * The *flags* argument is reserved and must be 0. The helper is
4913 * currently only supported for tc BPF program types, and enabled
4914 * for IPv4 and IPv6 protocols.
4916 * The helper returns **TC_ACT_REDIRECT** on success or
4917 * **TC_ACT_SHOT** on error.
4919 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4921 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4922 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4923 * extern variable decorated with '__ksym'. For ksym, there is a
4924 * global var (either static or global) defined of the same name
4925 * in the kernel. The ksym is percpu if the global var is percpu.
4926 * The returned pointer points to the global percpu var on *cpu*.
4928 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4929 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4930 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4931 * bpf_per_cpu_ptr() must check the returned value.
4933 * A pointer pointing to the kernel percpu variable on *cpu*, or
4934 * NULL, if *cpu* is invalid.
4936 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4938 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4939 * pointer to the percpu kernel variable on this cpu. See the
4940 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4942 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4943 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4944 * never return NULL.
4946 * A pointer pointing to the kernel percpu variable on this cpu.
4948 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4950 * Redirect the packet to another net device of index *ifindex*.
4951 * This helper is somewhat similar to **bpf_redirect**\ (), except
4952 * that the redirection happens to the *ifindex*' peer device and
4953 * the netns switch takes place from ingress to ingress without
4954 * going through the CPU's backlog queue.
4956 * The *flags* argument is reserved and must be 0. The helper is
4957 * currently only supported for tc BPF program types at the
4958 * ingress hook and for veth and netkit target device types. The
4959 * peer device must reside in a different network namespace.
4961 * The helper returns **TC_ACT_REDIRECT** on success or
4962 * **TC_ACT_SHOT** on error.
4964 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4966 * Get a bpf_local_storage from the *task*.
4968 * Logically, it could be thought of as getting the value from
4969 * a *map* with *task* as the **key**. From this
4970 * perspective, the usage is not much different from
4971 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4972 * helper enforces the key must be a task_struct and the map must also
4973 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4975 * Underneath, the value is stored locally at *task* instead of
4976 * the *map*. The *map* is used as the bpf-local-storage
4977 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4978 * searched against all bpf_local_storage residing at *task*.
4980 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4981 * used such that a new bpf_local_storage will be
4982 * created if one does not exist. *value* can be used
4983 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4984 * the initial value of a bpf_local_storage. If *value* is
4985 * **NULL**, the new bpf_local_storage will be zero initialized.
4987 * A bpf_local_storage pointer is returned on success.
4989 * **NULL** if not found or there was an error in adding
4990 * a new bpf_local_storage.
4992 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4994 * Delete a bpf_local_storage from a *task*.
4998 * **-ENOENT** if the bpf_local_storage cannot be found.
5000 * struct task_struct *bpf_get_current_task_btf(void)
5002 * Return a BTF pointer to the "current" task.
5003 * This pointer can also be used in helpers that accept an
5004 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
5006 * Pointer to the current task.
5008 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
5010 * Set or clear certain options on *bprm*:
5012 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
5013 * which sets the **AT_SECURE** auxv for glibc. The bit
5014 * is cleared if the flag is not specified.
5016 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
5018 * u64 bpf_ktime_get_coarse_ns(void)
5020 * Return a coarse-grained version of the time elapsed since
5021 * system boot, in nanoseconds. Does not include time the system
5024 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
5028 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
5030 * Returns the stored IMA hash of the *inode* (if it's available).
5031 * If the hash is larger than *size*, then only *size*
5032 * bytes will be copied to *dst*
5034 * The **hash_algo** is returned on success,
5035 * **-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if
5036 * invalid arguments are passed.
5038 * struct socket *bpf_sock_from_file(struct file *file)
5040 * If the given file represents a socket, returns the associated
5043 * A pointer to a struct socket on success or NULL if the file is
5046 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
5048 * Check packet size against exceeding MTU of net device (based
5049 * on *ifindex*). This helper will likely be used in combination
5050 * with helpers that adjust/change the packet size.
5052 * The argument *len_diff* can be used for querying with a planned
5053 * size change. This allows to check MTU prior to changing packet
5054 * ctx. Providing a *len_diff* adjustment that is larger than the
5055 * actual packet size (resulting in negative packet size) will in
5056 * principle not exceed the MTU, which is why it is not considered
5057 * a failure. Other BPF helpers are needed for performing the
5058 * planned size change; therefore the responsibility for catching
5059 * a negative packet size belongs in those helpers.
5061 * Specifying *ifindex* zero means the MTU check is performed
5062 * against the current net device. This is practical if this isn't
5063 * used prior to redirect.
5065 * On input *mtu_len* must be a valid pointer, else verifier will
5066 * reject BPF program. If the value *mtu_len* is initialized to
5067 * zero then the ctx packet size is use. When value *mtu_len* is
5068 * provided as input this specify the L3 length that the MTU check
5069 * is done against. Remember XDP and TC length operate at L2, but
5070 * this value is L3 as this correlate to MTU and IP-header tot_len
5071 * values which are L3 (similar behavior as bpf_fib_lookup).
5073 * The Linux kernel route table can configure MTUs on a more
5074 * specific per route level, which is not provided by this helper.
5075 * For route level MTU checks use the **bpf_fib_lookup**\ ()
5078 * *ctx* is either **struct xdp_md** for XDP programs or
5079 * **struct sk_buff** for tc cls_act programs.
5081 * The *flags* argument can be a combination of one or more of the
5084 * **BPF_MTU_CHK_SEGS**
5085 * This flag will only works for *ctx* **struct sk_buff**.
5086 * If packet context contains extra packet segment buffers
5087 * (often knows as GSO skb), then MTU check is harder to
5088 * check at this point, because in transmit path it is
5089 * possible for the skb packet to get re-segmented
5090 * (depending on net device features). This could still be
5091 * a MTU violation, so this flag enables performing MTU
5092 * check against segments, with a different violation
5093 * return code to tell it apart. Check cannot use len_diff.
5095 * On return *mtu_len* pointer contains the MTU value of the net
5096 * device. Remember the net device configured MTU is the L3 size,
5097 * which is returned here and XDP and TC length operate at L2.
5098 * Helper take this into account for you, but remember when using
5099 * MTU value in your BPF-code.
5102 * * 0 on success, and populate MTU value in *mtu_len* pointer.
5104 * * < 0 if any input argument is invalid (*mtu_len* not updated)
5106 * MTU violations return positive values, but also populate MTU
5107 * value in *mtu_len* pointer, as this can be needed for
5108 * implementing PMTU handing:
5110 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
5111 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
5113 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
5115 * For each element in **map**, call **callback_fn** function with
5116 * **map**, **callback_ctx** and other map-specific parameters.
5117 * The **callback_fn** should be a static function and
5118 * the **callback_ctx** should be a pointer to the stack.
5119 * The **flags** is used to control certain aspects of the helper.
5120 * Currently, the **flags** must be 0.
5122 * The following are a list of supported map types and their
5123 * respective expected callback signatures:
5125 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
5126 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
5127 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
5129 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
5131 * For per_cpu maps, the map_value is the value on the cpu where the
5132 * bpf_prog is running.
5134 * If **callback_fn** return 0, the helper will continue to the next
5135 * element. If return value is 1, the helper will skip the rest of
5136 * elements and return. Other return values are not used now.
5139 * The number of traversed map elements for success, **-EINVAL** for
5140 * invalid **flags**.
5142 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
5144 * Outputs a string into the **str** buffer of size **str_size**
5145 * based on a format string stored in a read-only map pointed by
5148 * Each format specifier in **fmt** corresponds to one u64 element
5149 * in the **data** array. For strings and pointers where pointees
5150 * are accessed, only the pointer values are stored in the *data*
5151 * array. The *data_len* is the size of *data* in bytes - must be
5154 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5155 * memory. Reading kernel memory may fail due to either invalid
5156 * address or valid address but requiring a major memory fault. If
5157 * reading kernel memory fails, the string for **%s** will be an
5158 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5159 * Not returning error to bpf program is consistent with what
5160 * **bpf_trace_printk**\ () does for now.
5163 * The strictly positive length of the formatted string, including
5164 * the trailing zero character. If the return value is greater than
5165 * **str_size**, **str** contains a truncated string, guaranteed to
5166 * be zero-terminated except when **str_size** is 0.
5168 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5170 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5172 * Execute bpf syscall with given arguments.
5176 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5178 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5180 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5182 * long bpf_sys_close(u32 fd)
5184 * Execute close syscall for given FD.
5188 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5190 * Initialize the timer.
5191 * First 4 bits of *flags* specify clockid.
5192 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5193 * All other bits of *flags* are reserved.
5194 * The verifier will reject the program if *timer* is not from
5198 * **-EBUSY** if *timer* is already initialized.
5199 * **-EINVAL** if invalid *flags* are passed.
5200 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5201 * The user space should either hold a file descriptor to a map with timers
5202 * or pin such map in bpffs. When map is unpinned or file descriptor is
5203 * closed all timers in the map will be cancelled and freed.
5205 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5207 * Configure the timer to call *callback_fn* static function.
5210 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5211 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5212 * The user space should either hold a file descriptor to a map with timers
5213 * or pin such map in bpffs. When map is unpinned or file descriptor is
5214 * closed all timers in the map will be cancelled and freed.
5216 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5218 * Set timer expiration N nanoseconds from the current time. The
5219 * configured callback will be invoked in soft irq context on some cpu
5220 * and will not repeat unless another bpf_timer_start() is made.
5221 * In such case the next invocation can migrate to a different cpu.
5222 * Since struct bpf_timer is a field inside map element the map
5223 * owns the timer. The bpf_timer_set_callback() will increment refcnt
5224 * of BPF program to make sure that callback_fn code stays valid.
5225 * When user space reference to a map reaches zero all timers
5226 * in a map are cancelled and corresponding program's refcnts are
5227 * decremented. This is done to make sure that Ctrl-C of a user
5228 * process doesn't leave any timers running. If map is pinned in
5229 * bpffs the callback_fn can re-arm itself indefinitely.
5230 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5231 * cancel and free the timer in the given map element.
5232 * The map can contain timers that invoke callback_fn-s from different
5233 * programs. The same callback_fn can serve different timers from
5234 * different maps if key/value layout matches across maps.
5235 * Every bpf_timer_set_callback() can have different callback_fn.
5237 * *flags* can be one of:
5239 * **BPF_F_TIMER_ABS**
5240 * Start the timer in absolute expire value instead of the
5241 * default relative one.
5242 * **BPF_F_TIMER_CPU_PIN**
5243 * Timer will be pinned to the CPU of the caller.
5247 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5248 * or invalid *flags* are passed.
5250 * long bpf_timer_cancel(struct bpf_timer *timer)
5252 * Cancel the timer and wait for callback_fn to finish if it was running.
5254 * 0 if the timer was not active.
5255 * 1 if the timer was active.
5256 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5257 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5258 * own timer which would have led to a deadlock otherwise.
5260 * u64 bpf_get_func_ip(void *ctx)
5262 * Get address of the traced function (for tracing and kprobe programs).
5264 * When called for kprobe program attached as uprobe it returns
5265 * probe address for both entry and return uprobe.
5268 * Address of the traced function for kprobe.
5269 * 0 for kprobes placed within the function (not at the entry).
5270 * Address of the probe for uprobe and return uprobe.
5272 * u64 bpf_get_attach_cookie(void *ctx)
5274 * Get bpf_cookie value provided (optionally) during the program
5275 * attachment. It might be different for each individual
5276 * attachment, even if BPF program itself is the same.
5277 * Expects BPF program context *ctx* as a first argument.
5279 * Supported for the following program types:
5284 * Value specified by user at BPF link creation/attachment time
5285 * or 0, if it was not specified.
5287 * long bpf_task_pt_regs(struct task_struct *task)
5289 * Get the struct pt_regs associated with **task**.
5291 * A pointer to struct pt_regs.
5293 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5295 * Get branch trace from hardware engines like Intel LBR. The
5296 * hardware engine is stopped shortly after the helper is
5297 * called. Therefore, the user need to filter branch entries
5298 * based on the actual use case. To capture branch trace
5299 * before the trigger point of the BPF program, the helper
5300 * should be called at the beginning of the BPF program.
5302 * The data is stored as struct perf_branch_entry into output
5303 * buffer *entries*. *size* is the size of *entries* in bytes.
5304 * *flags* is reserved for now and must be zero.
5307 * On success, number of bytes written to *buf*. On error, a
5310 * **-EINVAL** if *flags* is not zero.
5312 * **-ENOENT** if architecture does not support branch records.
5314 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5316 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5317 * to format and can handle more format args as a result.
5319 * Arguments are to be used as in **bpf_seq_printf**\ () helper.
5321 * The number of bytes written to the buffer, or a negative error
5322 * in case of failure.
5324 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5326 * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5328 * *sk* if casting is valid, or **NULL** otherwise.
5330 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5332 * Get the address of a kernel symbol, returned in *res*. *res* is
5333 * set to 0 if the symbol is not found.
5335 * On success, zero. On error, a negative value.
5337 * **-EINVAL** if *flags* is not zero.
5339 * **-EINVAL** if string *name* is not the same size as *name_sz*.
5341 * **-ENOENT** if symbol is not found.
5343 * **-EPERM** if caller does not have permission to obtain kernel address.
5345 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5347 * Find vma of *task* that contains *addr*, call *callback_fn*
5348 * function with *task*, *vma*, and *callback_ctx*.
5349 * The *callback_fn* should be a static function and
5350 * the *callback_ctx* should be a pointer to the stack.
5351 * The *flags* is used to control certain aspects of the helper.
5352 * Currently, the *flags* must be 0.
5354 * The expected callback signature is
5356 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5360 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5361 * **-EBUSY** if failed to try lock mmap_lock.
5362 * **-EINVAL** for invalid **flags**.
5364 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5366 * For **nr_loops**, call **callback_fn** function
5367 * with **callback_ctx** as the context parameter.
5368 * The **callback_fn** should be a static function and
5369 * the **callback_ctx** should be a pointer to the stack.
5370 * The **flags** is used to control certain aspects of the helper.
5371 * Currently, the **flags** must be 0. Currently, nr_loops is
5372 * limited to 1 << 23 (~8 million) loops.
5374 * long (\*callback_fn)(u64 index, void \*ctx);
5376 * where **index** is the current index in the loop. The index
5379 * If **callback_fn** returns 0, the helper will continue to the next
5380 * loop. If return value is 1, the helper will skip the rest of
5381 * the loops and return. Other return values are not used now,
5382 * and will be rejected by the verifier.
5385 * The number of loops performed, **-EINVAL** for invalid **flags**,
5386 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5388 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5390 * Do strncmp() between **s1** and **s2**. **s1** doesn't need
5391 * to be null-terminated and **s1_sz** is the maximum storage
5392 * size of **s1**. **s2** must be a read-only string.
5394 * An integer less than, equal to, or greater than zero
5395 * if the first **s1_sz** bytes of **s1** is found to be
5396 * less than, to match, or be greater than **s2**.
5398 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5400 * Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5401 * returned in **value**.
5405 * **-EINVAL** if n >= argument register count of traced function.
5407 * long bpf_get_func_ret(void *ctx, u64 *value)
5409 * Get return value of the traced function (for tracing programs)
5414 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5416 * long bpf_get_func_arg_cnt(void *ctx)
5418 * Get number of registers of the traced function (for tracing programs) where
5419 * function arguments are stored in these registers.
5422 * The number of argument registers of the traced function.
5424 * int bpf_get_retval(void)
5426 * Get the BPF program's return value that will be returned to the upper layers.
5428 * This helper is currently supported by cgroup programs and only by the hooks
5429 * where BPF program's return value is returned to the userspace via errno.
5431 * The BPF program's return value.
5433 * int bpf_set_retval(int retval)
5435 * Set the BPF program's return value that will be returned to the upper layers.
5437 * This helper is currently supported by cgroup programs and only by the hooks
5438 * where BPF program's return value is returned to the userspace via errno.
5440 * Note that there is the following corner case where the program exports an error
5441 * via bpf_set_retval but signals success via 'return 1':
5443 * bpf_set_retval(-EPERM);
5446 * In this case, the BPF program's return value will use helper's -EPERM. This
5447 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5450 * 0 on success, or a negative error in case of failure.
5452 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5454 * Get the total size of a given xdp buff (linear and paged area)
5456 * The total size of a given xdp buffer.
5458 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5460 * This helper is provided as an easy way to load data from a
5461 * xdp buffer. It can be used to load *len* bytes from *offset* from
5462 * the frame associated to *xdp_md*, into the buffer pointed by
5465 * 0 on success, or a negative error in case of failure.
5467 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5469 * Store *len* bytes from buffer *buf* into the frame
5470 * associated to *xdp_md*, at *offset*.
5472 * 0 on success, or a negative error in case of failure.
5474 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5476 * Read *size* bytes from user space address *user_ptr* in *tsk*'s
5477 * address space, and stores the data in *dst*. *flags* is not
5478 * used yet and is provided for future extensibility. This helper
5479 * can only be used by sleepable programs.
5481 * 0 on success, or a negative error in case of failure. On error
5482 * *dst* buffer is zeroed out.
5484 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5486 * Change the __sk_buff->tstamp_type to *tstamp_type*
5487 * and set *tstamp* to the __sk_buff->tstamp together.
5489 * If there is no need to change the __sk_buff->tstamp_type,
5490 * the tstamp value can be directly written to __sk_buff->tstamp
5493 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5494 * will be kept during bpf_redirect_*(). A non zero
5495 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5498 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5499 * with a zero *tstamp*.
5501 * Only IPv4 and IPv6 skb->protocol are supported.
5503 * This function is most useful when it needs to set a
5504 * mono delivery time to __sk_buff->tstamp and then
5505 * bpf_redirect_*() to the egress of an iface. For example,
5506 * changing the (rcv) timestamp in __sk_buff->tstamp at
5507 * ingress to a mono delivery time and then bpf_redirect_*()
5508 * to sch_fq@phy-dev.
5511 * **-EINVAL** for invalid input
5512 * **-EOPNOTSUPP** for unsupported protocol
5514 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5516 * Returns a calculated IMA hash of the *file*.
5517 * If the hash is larger than *size*, then only *size*
5518 * bytes will be copied to *dst*
5520 * The **hash_algo** is returned on success,
5521 * **-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if
5522 * invalid arguments are passed.
5524 * void *bpf_kptr_xchg(void *dst, void *ptr)
5526 * Exchange kptr at pointer *dst* with *ptr*, and return the old value.
5527 * *dst* can be map value or local kptr. *ptr* can be NULL, otherwise
5528 * it must be a referenced pointer which will be released when this helper
5531 * The old value of kptr (which can be NULL). The returned pointer
5532 * if not NULL, is a reference which must be released using its
5533 * corresponding release function, or moved into a BPF map before
5536 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5538 * Perform a lookup in *percpu map* for an entry associated to
5541 * Map value associated to *key* on *cpu*, or **NULL** if no entry
5542 * was found or *cpu* is invalid.
5544 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5546 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5548 * *sk* if casting is valid, or **NULL** otherwise.
5550 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5552 * Get a dynptr to local memory *data*.
5554 * *data* must be a ptr to a map value.
5555 * The maximum *size* supported is DYNPTR_MAX_SIZE.
5556 * *flags* is currently unused.
5558 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5559 * -EINVAL if flags is not 0.
5561 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5563 * Reserve *size* bytes of payload in a ring buffer *ringbuf*
5564 * through the dynptr interface. *flags* must be 0.
5566 * Please note that a corresponding bpf_ringbuf_submit_dynptr or
5567 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5568 * reservation fails. This is enforced by the verifier.
5570 * 0 on success, or a negative error in case of failure.
5572 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5574 * Submit reserved ring buffer sample, pointed to by *data*,
5575 * through the dynptr interface. This is a no-op if the dynptr is
5578 * For more information on *flags*, please see
5579 * 'bpf_ringbuf_submit'.
5581 * Nothing. Always succeeds.
5583 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5585 * Discard reserved ring buffer sample through the dynptr
5586 * interface. This is a no-op if the dynptr is invalid/null.
5588 * For more information on *flags*, please see
5589 * 'bpf_ringbuf_discard'.
5591 * Nothing. Always succeeds.
5593 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
5595 * Read *len* bytes from *src* into *dst*, starting from *offset*
5597 * *flags* is currently unused.
5599 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5600 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5603 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5605 * Write *len* bytes from *src* into *dst*, starting from *offset*
5608 * *flags* must be 0 except for skb-type dynptrs.
5610 * For skb-type dynptrs:
5611 * * All data slices of the dynptr are automatically
5612 * invalidated after **bpf_dynptr_write**\ (). This is
5613 * because writing may pull the skb and change the
5614 * underlying packet buffer.
5616 * * For *flags*, please see the flags accepted by
5617 * **bpf_skb_store_bytes**\ ().
5619 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5620 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5621 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5622 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5624 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
5626 * Get a pointer to the underlying dynptr data.
5628 * *len* must be a statically known value. The returned data slice
5629 * is invalidated whenever the dynptr is invalidated.
5631 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5632 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5634 * Pointer to the underlying dynptr data, NULL if the dynptr is
5635 * read-only, if the dynptr is invalid, or if the offset and length
5638 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5640 * Try to issue a SYN cookie for the packet with corresponding
5641 * IPv4/TCP headers, *iph* and *th*, without depending on a
5644 * *iph* points to the IPv4 header.
5646 * *th* points to the start of the TCP header, while *th_len*
5647 * contains the length of the TCP header (at least
5648 * **sizeof**\ (**struct tcphdr**)).
5650 * On success, lower 32 bits hold the generated SYN cookie in
5651 * followed by 16 bits which hold the MSS value for that cookie,
5652 * and the top 16 bits are unused.
5654 * On failure, the returned value is one of the following:
5656 * **-EINVAL** if *th_len* is invalid.
5658 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5660 * Try to issue a SYN cookie for the packet with corresponding
5661 * IPv6/TCP headers, *iph* and *th*, without depending on a
5664 * *iph* points to the IPv6 header.
5666 * *th* points to the start of the TCP header, while *th_len*
5667 * contains the length of the TCP header (at least
5668 * **sizeof**\ (**struct tcphdr**)).
5670 * On success, lower 32 bits hold the generated SYN cookie in
5671 * followed by 16 bits which hold the MSS value for that cookie,
5672 * and the top 16 bits are unused.
5674 * On failure, the returned value is one of the following:
5676 * **-EINVAL** if *th_len* is invalid.
5678 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5680 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5682 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5683 * without depending on a listening socket.
5685 * *iph* points to the IPv4 header.
5687 * *th* points to the TCP header.
5689 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5691 * On failure, the returned value is one of the following:
5693 * **-EACCES** if the SYN cookie is not valid.
5695 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5697 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5698 * without depending on a listening socket.
5700 * *iph* points to the IPv6 header.
5702 * *th* points to the TCP header.
5704 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5706 * On failure, the returned value is one of the following:
5708 * **-EACCES** if the SYN cookie is not valid.
5710 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5712 * u64 bpf_ktime_get_tai_ns(void)
5714 * A nonsettable system-wide clock derived from wall-clock time but
5715 * ignoring leap seconds. This clock does not experience
5716 * discontinuities and backwards jumps caused by NTP inserting leap
5717 * seconds as CLOCK_REALTIME does.
5719 * See: **clock_gettime**\ (**CLOCK_TAI**)
5723 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5725 * Drain samples from the specified user ring buffer, and invoke
5726 * the provided callback for each such sample:
5728 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5730 * If **callback_fn** returns 0, the helper will continue to try
5731 * and drain the next sample, up to a maximum of
5732 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5733 * the helper will skip the rest of the samples and return. Other
5734 * return values are not used now, and will be rejected by the
5737 * The number of drained samples if no error was encountered while
5738 * draining samples, or 0 if no samples were present in the ring
5739 * buffer. If a user-space producer was epoll-waiting on this map,
5740 * and at least one sample was drained, they will receive an event
5741 * notification notifying them of available space in the ring
5742 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5743 * function, no wakeup notification will be sent. If the
5744 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5745 * be sent even if no sample was drained.
5747 * On failure, the returned value is one of the following:
5749 * **-EBUSY** if the ring buffer is contended, and another calling
5750 * context was concurrently draining the ring buffer.
5752 * **-EINVAL** if user-space is not properly tracking the ring
5753 * buffer due to the producer position not being aligned to 8
5754 * bytes, a sample not being aligned to 8 bytes, or the producer
5755 * position not matching the advertised length of a sample.
5757 * **-E2BIG** if user-space has tried to publish a sample which is
5758 * larger than the size of the ring buffer, or which cannot fit
5759 * within a struct bpf_dynptr.
5761 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5763 * Get a bpf_local_storage from the *cgroup*.
5765 * Logically, it could be thought of as getting the value from
5766 * a *map* with *cgroup* as the **key**. From this
5767 * perspective, the usage is not much different from
5768 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5769 * helper enforces the key must be a cgroup struct and the map must also
5770 * be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5772 * In reality, the local-storage value is embedded directly inside of the
5773 * *cgroup* object itself, rather than being located in the
5774 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5775 * queried for some *map* on a *cgroup* object, the kernel will perform an
5776 * O(n) iteration over all of the live local-storage values for that
5777 * *cgroup* object until the local-storage value for the *map* is found.
5779 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5780 * used such that a new bpf_local_storage will be
5781 * created if one does not exist. *value* can be used
5782 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5783 * the initial value of a bpf_local_storage. If *value* is
5784 * **NULL**, the new bpf_local_storage will be zero initialized.
5786 * A bpf_local_storage pointer is returned on success.
5788 * **NULL** if not found or there was an error in adding
5789 * a new bpf_local_storage.
5791 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5793 * Delete a bpf_local_storage from a *cgroup*.
5797 * **-ENOENT** if the bpf_local_storage cannot be found.
5799 #define ___BPF_FUNC_MAPPER(FN, ctx...) \
5800 FN(unspec, 0, ##ctx) \
5801 FN(map_lookup_elem, 1, ##ctx) \
5802 FN(map_update_elem, 2, ##ctx) \
5803 FN(map_delete_elem, 3, ##ctx) \
5804 FN(probe_read, 4, ##ctx) \
5805 FN(ktime_get_ns, 5, ##ctx) \
5806 FN(trace_printk, 6, ##ctx) \
5807 FN(get_prandom_u32, 7, ##ctx) \
5808 FN(get_smp_processor_id, 8, ##ctx) \
5809 FN(skb_store_bytes, 9, ##ctx) \
5810 FN(l3_csum_replace, 10, ##ctx) \
5811 FN(l4_csum_replace, 11, ##ctx) \
5812 FN(tail_call, 12, ##ctx) \
5813 FN(clone_redirect, 13, ##ctx) \
5814 FN(get_current_pid_tgid, 14, ##ctx) \
5815 FN(get_current_uid_gid, 15, ##ctx) \
5816 FN(get_current_comm, 16, ##ctx) \
5817 FN(get_cgroup_classid, 17, ##ctx) \
5818 FN(skb_vlan_push, 18, ##ctx) \
5819 FN(skb_vlan_pop, 19, ##ctx) \
5820 FN(skb_get_tunnel_key, 20, ##ctx) \
5821 FN(skb_set_tunnel_key, 21, ##ctx) \
5822 FN(perf_event_read, 22, ##ctx) \
5823 FN(redirect, 23, ##ctx) \
5824 FN(get_route_realm, 24, ##ctx) \
5825 FN(perf_event_output, 25, ##ctx) \
5826 FN(skb_load_bytes, 26, ##ctx) \
5827 FN(get_stackid, 27, ##ctx) \
5828 FN(csum_diff, 28, ##ctx) \
5829 FN(skb_get_tunnel_opt, 29, ##ctx) \
5830 FN(skb_set_tunnel_opt, 30, ##ctx) \
5831 FN(skb_change_proto, 31, ##ctx) \
5832 FN(skb_change_type, 32, ##ctx) \
5833 FN(skb_under_cgroup, 33, ##ctx) \
5834 FN(get_hash_recalc, 34, ##ctx) \
5835 FN(get_current_task, 35, ##ctx) \
5836 FN(probe_write_user, 36, ##ctx) \
5837 FN(current_task_under_cgroup, 37, ##ctx) \
5838 FN(skb_change_tail, 38, ##ctx) \
5839 FN(skb_pull_data, 39, ##ctx) \
5840 FN(csum_update, 40, ##ctx) \
5841 FN(set_hash_invalid, 41, ##ctx) \
5842 FN(get_numa_node_id, 42, ##ctx) \
5843 FN(skb_change_head, 43, ##ctx) \
5844 FN(xdp_adjust_head, 44, ##ctx) \
5845 FN(probe_read_str, 45, ##ctx) \
5846 FN(get_socket_cookie, 46, ##ctx) \
5847 FN(get_socket_uid, 47, ##ctx) \
5848 FN(set_hash, 48, ##ctx) \
5849 FN(setsockopt, 49, ##ctx) \
5850 FN(skb_adjust_room, 50, ##ctx) \
5851 FN(redirect_map, 51, ##ctx) \
5852 FN(sk_redirect_map, 52, ##ctx) \
5853 FN(sock_map_update, 53, ##ctx) \
5854 FN(xdp_adjust_meta, 54, ##ctx) \
5855 FN(perf_event_read_value, 55, ##ctx) \
5856 FN(perf_prog_read_value, 56, ##ctx) \
5857 FN(getsockopt, 57, ##ctx) \
5858 FN(override_return, 58, ##ctx) \
5859 FN(sock_ops_cb_flags_set, 59, ##ctx) \
5860 FN(msg_redirect_map, 60, ##ctx) \
5861 FN(msg_apply_bytes, 61, ##ctx) \
5862 FN(msg_cork_bytes, 62, ##ctx) \
5863 FN(msg_pull_data, 63, ##ctx) \
5864 FN(bind, 64, ##ctx) \
5865 FN(xdp_adjust_tail, 65, ##ctx) \
5866 FN(skb_get_xfrm_state, 66, ##ctx) \
5867 FN(get_stack, 67, ##ctx) \
5868 FN(skb_load_bytes_relative, 68, ##ctx) \
5869 FN(fib_lookup, 69, ##ctx) \
5870 FN(sock_hash_update, 70, ##ctx) \
5871 FN(msg_redirect_hash, 71, ##ctx) \
5872 FN(sk_redirect_hash, 72, ##ctx) \
5873 FN(lwt_push_encap, 73, ##ctx) \
5874 FN(lwt_seg6_store_bytes, 74, ##ctx) \
5875 FN(lwt_seg6_adjust_srh, 75, ##ctx) \
5876 FN(lwt_seg6_action, 76, ##ctx) \
5877 FN(rc_repeat, 77, ##ctx) \
5878 FN(rc_keydown, 78, ##ctx) \
5879 FN(skb_cgroup_id, 79, ##ctx) \
5880 FN(get_current_cgroup_id, 80, ##ctx) \
5881 FN(get_local_storage, 81, ##ctx) \
5882 FN(sk_select_reuseport, 82, ##ctx) \
5883 FN(skb_ancestor_cgroup_id, 83, ##ctx) \
5884 FN(sk_lookup_tcp, 84, ##ctx) \
5885 FN(sk_lookup_udp, 85, ##ctx) \
5886 FN(sk_release, 86, ##ctx) \
5887 FN(map_push_elem, 87, ##ctx) \
5888 FN(map_pop_elem, 88, ##ctx) \
5889 FN(map_peek_elem, 89, ##ctx) \
5890 FN(msg_push_data, 90, ##ctx) \
5891 FN(msg_pop_data, 91, ##ctx) \
5892 FN(rc_pointer_rel, 92, ##ctx) \
5893 FN(spin_lock, 93, ##ctx) \
5894 FN(spin_unlock, 94, ##ctx) \
5895 FN(sk_fullsock, 95, ##ctx) \
5896 FN(tcp_sock, 96, ##ctx) \
5897 FN(skb_ecn_set_ce, 97, ##ctx) \
5898 FN(get_listener_sock, 98, ##ctx) \
5899 FN(skc_lookup_tcp, 99, ##ctx) \
5900 FN(tcp_check_syncookie, 100, ##ctx) \
5901 FN(sysctl_get_name, 101, ##ctx) \
5902 FN(sysctl_get_current_value, 102, ##ctx) \
5903 FN(sysctl_get_new_value, 103, ##ctx) \
5904 FN(sysctl_set_new_value, 104, ##ctx) \
5905 FN(strtol, 105, ##ctx) \
5906 FN(strtoul, 106, ##ctx) \
5907 FN(sk_storage_get, 107, ##ctx) \
5908 FN(sk_storage_delete, 108, ##ctx) \
5909 FN(send_signal, 109, ##ctx) \
5910 FN(tcp_gen_syncookie, 110, ##ctx) \
5911 FN(skb_output, 111, ##ctx) \
5912 FN(probe_read_user, 112, ##ctx) \
5913 FN(probe_read_kernel, 113, ##ctx) \
5914 FN(probe_read_user_str, 114, ##ctx) \
5915 FN(probe_read_kernel_str, 115, ##ctx) \
5916 FN(tcp_send_ack, 116, ##ctx) \
5917 FN(send_signal_thread, 117, ##ctx) \
5918 FN(jiffies64, 118, ##ctx) \
5919 FN(read_branch_records, 119, ##ctx) \
5920 FN(get_ns_current_pid_tgid, 120, ##ctx) \
5921 FN(xdp_output, 121, ##ctx) \
5922 FN(get_netns_cookie, 122, ##ctx) \
5923 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \
5924 FN(sk_assign, 124, ##ctx) \
5925 FN(ktime_get_boot_ns, 125, ##ctx) \
5926 FN(seq_printf, 126, ##ctx) \
5927 FN(seq_write, 127, ##ctx) \
5928 FN(sk_cgroup_id, 128, ##ctx) \
5929 FN(sk_ancestor_cgroup_id, 129, ##ctx) \
5930 FN(ringbuf_output, 130, ##ctx) \
5931 FN(ringbuf_reserve, 131, ##ctx) \
5932 FN(ringbuf_submit, 132, ##ctx) \
5933 FN(ringbuf_discard, 133, ##ctx) \
5934 FN(ringbuf_query, 134, ##ctx) \
5935 FN(csum_level, 135, ##ctx) \
5936 FN(skc_to_tcp6_sock, 136, ##ctx) \
5937 FN(skc_to_tcp_sock, 137, ##ctx) \
5938 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \
5939 FN(skc_to_tcp_request_sock, 139, ##ctx) \
5940 FN(skc_to_udp6_sock, 140, ##ctx) \
5941 FN(get_task_stack, 141, ##ctx) \
5942 FN(load_hdr_opt, 142, ##ctx) \
5943 FN(store_hdr_opt, 143, ##ctx) \
5944 FN(reserve_hdr_opt, 144, ##ctx) \
5945 FN(inode_storage_get, 145, ##ctx) \
5946 FN(inode_storage_delete, 146, ##ctx) \
5947 FN(d_path, 147, ##ctx) \
5948 FN(copy_from_user, 148, ##ctx) \
5949 FN(snprintf_btf, 149, ##ctx) \
5950 FN(seq_printf_btf, 150, ##ctx) \
5951 FN(skb_cgroup_classid, 151, ##ctx) \
5952 FN(redirect_neigh, 152, ##ctx) \
5953 FN(per_cpu_ptr, 153, ##ctx) \
5954 FN(this_cpu_ptr, 154, ##ctx) \
5955 FN(redirect_peer, 155, ##ctx) \
5956 FN(task_storage_get, 156, ##ctx) \
5957 FN(task_storage_delete, 157, ##ctx) \
5958 FN(get_current_task_btf, 158, ##ctx) \
5959 FN(bprm_opts_set, 159, ##ctx) \
5960 FN(ktime_get_coarse_ns, 160, ##ctx) \
5961 FN(ima_inode_hash, 161, ##ctx) \
5962 FN(sock_from_file, 162, ##ctx) \
5963 FN(check_mtu, 163, ##ctx) \
5964 FN(for_each_map_elem, 164, ##ctx) \
5965 FN(snprintf, 165, ##ctx) \
5966 FN(sys_bpf, 166, ##ctx) \
5967 FN(btf_find_by_name_kind, 167, ##ctx) \
5968 FN(sys_close, 168, ##ctx) \
5969 FN(timer_init, 169, ##ctx) \
5970 FN(timer_set_callback, 170, ##ctx) \
5971 FN(timer_start, 171, ##ctx) \
5972 FN(timer_cancel, 172, ##ctx) \
5973 FN(get_func_ip, 173, ##ctx) \
5974 FN(get_attach_cookie, 174, ##ctx) \
5975 FN(task_pt_regs, 175, ##ctx) \
5976 FN(get_branch_snapshot, 176, ##ctx) \
5977 FN(trace_vprintk, 177, ##ctx) \
5978 FN(skc_to_unix_sock, 178, ##ctx) \
5979 FN(kallsyms_lookup_name, 179, ##ctx) \
5980 FN(find_vma, 180, ##ctx) \
5981 FN(loop, 181, ##ctx) \
5982 FN(strncmp, 182, ##ctx) \
5983 FN(get_func_arg, 183, ##ctx) \
5984 FN(get_func_ret, 184, ##ctx) \
5985 FN(get_func_arg_cnt, 185, ##ctx) \
5986 FN(get_retval, 186, ##ctx) \
5987 FN(set_retval, 187, ##ctx) \
5988 FN(xdp_get_buff_len, 188, ##ctx) \
5989 FN(xdp_load_bytes, 189, ##ctx) \
5990 FN(xdp_store_bytes, 190, ##ctx) \
5991 FN(copy_from_user_task, 191, ##ctx) \
5992 FN(skb_set_tstamp, 192, ##ctx) \
5993 FN(ima_file_hash, 193, ##ctx) \
5994 FN(kptr_xchg, 194, ##ctx) \
5995 FN(map_lookup_percpu_elem, 195, ##ctx) \
5996 FN(skc_to_mptcp_sock, 196, ##ctx) \
5997 FN(dynptr_from_mem, 197, ##ctx) \
5998 FN(ringbuf_reserve_dynptr, 198, ##ctx) \
5999 FN(ringbuf_submit_dynptr, 199, ##ctx) \
6000 FN(ringbuf_discard_dynptr, 200, ##ctx) \
6001 FN(dynptr_read, 201, ##ctx) \
6002 FN(dynptr_write, 202, ##ctx) \
6003 FN(dynptr_data, 203, ##ctx) \
6004 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \
6005 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \
6006 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \
6007 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \
6008 FN(ktime_get_tai_ns, 208, ##ctx) \
6009 FN(user_ringbuf_drain, 209, ##ctx) \
6010 FN(cgrp_storage_get, 210, ##ctx) \
6011 FN(cgrp_storage_delete, 211, ##ctx) \
6014 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
6015 * know or care about integer value that is now passed as second argument
6017 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
6018 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
6020 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
6021 * function eBPF program intends to call
6023 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
6025 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN
)
6028 #undef __BPF_ENUM_FN
6030 /* All flags used by eBPF helper functions, placed here. */
6032 /* BPF_FUNC_skb_store_bytes flags. */
6034 BPF_F_RECOMPUTE_CSUM
= (1ULL << 0),
6035 BPF_F_INVALIDATE_HASH
= (1ULL << 1),
6038 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
6039 * First 4 bits are for passing the header field size.
6042 BPF_F_HDR_FIELD_MASK
= 0xfULL
,
6045 /* BPF_FUNC_l4_csum_replace flags. */
6047 BPF_F_PSEUDO_HDR
= (1ULL << 4),
6048 BPF_F_MARK_MANGLED_0
= (1ULL << 5),
6049 BPF_F_MARK_ENFORCE
= (1ULL << 6),
6052 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
6054 BPF_F_TUNINFO_IPV6
= (1ULL << 0),
6057 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
6059 BPF_F_SKIP_FIELD_MASK
= 0xffULL
,
6060 BPF_F_USER_STACK
= (1ULL << 8),
6061 /* flags used by BPF_FUNC_get_stackid only. */
6062 BPF_F_FAST_STACK_CMP
= (1ULL << 9),
6063 BPF_F_REUSE_STACKID
= (1ULL << 10),
6064 /* flags used by BPF_FUNC_get_stack only. */
6065 BPF_F_USER_BUILD_ID
= (1ULL << 11),
6068 /* BPF_FUNC_skb_set_tunnel_key flags. */
6070 BPF_F_ZERO_CSUM_TX
= (1ULL << 1),
6071 BPF_F_DONT_FRAGMENT
= (1ULL << 2),
6072 BPF_F_SEQ_NUMBER
= (1ULL << 3),
6073 BPF_F_NO_TUNNEL_KEY
= (1ULL << 4),
6076 /* BPF_FUNC_skb_get_tunnel_key flags. */
6078 BPF_F_TUNINFO_FLAGS
= (1ULL << 4),
6081 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
6082 * BPF_FUNC_perf_event_read_value flags.
6085 BPF_F_INDEX_MASK
= 0xffffffffULL
,
6086 BPF_F_CURRENT_CPU
= BPF_F_INDEX_MASK
,
6087 /* BPF_FUNC_perf_event_output for sk_buff input context. */
6088 BPF_F_CTXLEN_MASK
= (0xfffffULL
<< 32),
6091 /* Current network namespace */
6093 BPF_F_CURRENT_NETNS
= (-1L),
6096 /* BPF_FUNC_csum_level level values. */
6098 BPF_CSUM_LEVEL_QUERY
,
6101 BPF_CSUM_LEVEL_RESET
,
6104 /* BPF_FUNC_skb_adjust_room flags. */
6106 BPF_F_ADJ_ROOM_FIXED_GSO
= (1ULL << 0),
6107 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4
= (1ULL << 1),
6108 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6
= (1ULL << 2),
6109 BPF_F_ADJ_ROOM_ENCAP_L4_GRE
= (1ULL << 3),
6110 BPF_F_ADJ_ROOM_ENCAP_L4_UDP
= (1ULL << 4),
6111 BPF_F_ADJ_ROOM_NO_CSUM_RESET
= (1ULL << 5),
6112 BPF_F_ADJ_ROOM_ENCAP_L2_ETH
= (1ULL << 6),
6113 BPF_F_ADJ_ROOM_DECAP_L3_IPV4
= (1ULL << 7),
6114 BPF_F_ADJ_ROOM_DECAP_L3_IPV6
= (1ULL << 8),
6118 BPF_ADJ_ROOM_ENCAP_L2_MASK
= 0xff,
6119 BPF_ADJ_ROOM_ENCAP_L2_SHIFT
= 56,
6122 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
6123 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
6124 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6126 /* BPF_FUNC_sysctl_get_name flags. */
6128 BPF_F_SYSCTL_BASE_NAME
= (1ULL << 0),
6131 /* BPF_FUNC_<kernel_obj>_storage_get flags */
6133 BPF_LOCAL_STORAGE_GET_F_CREATE
= (1ULL << 0),
6134 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
6135 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
6137 BPF_SK_STORAGE_GET_F_CREATE
= BPF_LOCAL_STORAGE_GET_F_CREATE
,
6140 /* BPF_FUNC_read_branch_records flags. */
6142 BPF_F_GET_BRANCH_RECORDS_SIZE
= (1ULL << 0),
6145 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
6146 * BPF_FUNC_bpf_ringbuf_output flags.
6149 BPF_RB_NO_WAKEUP
= (1ULL << 0),
6150 BPF_RB_FORCE_WAKEUP
= (1ULL << 1),
6153 /* BPF_FUNC_bpf_ringbuf_query flags */
6155 BPF_RB_AVAIL_DATA
= 0,
6156 BPF_RB_RING_SIZE
= 1,
6157 BPF_RB_CONS_POS
= 2,
6158 BPF_RB_PROD_POS
= 3,
6161 /* BPF ring buffer constants */
6163 BPF_RINGBUF_BUSY_BIT
= (1U << 31),
6164 BPF_RINGBUF_DISCARD_BIT
= (1U << 30),
6165 BPF_RINGBUF_HDR_SZ
= 8,
6168 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6170 BPF_SK_LOOKUP_F_REPLACE
= (1ULL << 0),
6171 BPF_SK_LOOKUP_F_NO_REUSEPORT
= (1ULL << 1),
6174 /* Mode for BPF_FUNC_skb_adjust_room helper. */
6175 enum bpf_adj_room_mode
{
6180 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6181 enum bpf_hdr_start_off
{
6186 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6187 enum bpf_lwt_encap_mode
{
6189 BPF_LWT_ENCAP_SEG6_INLINE
,
6193 /* Flags for bpf_bprm_opts_set helper */
6195 BPF_F_BPRM_SECUREEXEC
= (1ULL << 0),
6198 /* Flags for bpf_redirect and bpf_redirect_map helpers */
6200 BPF_F_INGRESS
= (1ULL << 0), /* used for skb path */
6201 BPF_F_BROADCAST
= (1ULL << 3), /* used for XDP path */
6202 BPF_F_EXCLUDE_INGRESS
= (1ULL << 4), /* used for XDP path */
6203 #define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
6206 #define __bpf_md_ptr(type, name) \
6210 } __attribute__((aligned(8)))
6212 /* The enum used in skb->tstamp_type. It specifies the clock type
6213 * of the time stored in the skb->tstamp.
6216 BPF_SKB_TSTAMP_UNSPEC
= 0, /* DEPRECATED */
6217 BPF_SKB_TSTAMP_DELIVERY_MONO
= 1, /* DEPRECATED */
6218 BPF_SKB_CLOCK_REALTIME
= 0,
6219 BPF_SKB_CLOCK_MONOTONIC
= 1,
6220 BPF_SKB_CLOCK_TAI
= 2,
6221 /* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,
6222 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.
6226 /* user accessible mirror of in-kernel sk_buff.
6227 * new fields can only be added to the end of this structure
6233 __u32 queue_mapping
;
6239 __u32 ingress_ifindex
;
6249 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6251 __u32 remote_ip4
; /* Stored in network byte order */
6252 __u32 local_ip4
; /* Stored in network byte order */
6253 __u32 remote_ip6
[4]; /* Stored in network byte order */
6254 __u32 local_ip6
[4]; /* Stored in network byte order */
6255 __u32 remote_port
; /* Stored in network byte order */
6256 __u32 local_port
; /* stored in host byte order */
6260 __bpf_md_ptr(struct bpf_flow_keys
*, flow_keys
);
6264 __bpf_md_ptr(struct bpf_sock
*, sk
);
6267 __u32
:24; /* Padding, future use. */
6271 struct bpf_tunnel_key
{
6275 __u32 remote_ipv6
[4];
6280 __u16 tunnel_ext
; /* compat */
6281 __be16 tunnel_flags
;
6286 __u32 local_ipv6
[4];
6290 /* user accessible mirror of in-kernel xfrm_state.
6291 * new fields can only be added to the end of this structure
6293 struct bpf_xfrm_state
{
6295 __u32 spi
; /* Stored in network byte order */
6297 __u16 ext
; /* Padding, future use. */
6299 __u32 remote_ipv4
; /* Stored in network byte order */
6300 __u32 remote_ipv6
[4]; /* Stored in network byte order */
6304 /* Generic BPF return codes which all BPF program types may support.
6305 * The values are binary compatible with their TC_ACT_* counter-part to
6306 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6309 * XDP is handled seprately, see XDP_*.
6317 /* >127 are reserved for prog type specific return codes.
6319 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6320 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6321 * changed and should be routed based on its new L3 header.
6322 * (This is an L3 redirect, as opposed to L2 redirect
6323 * represented by BPF_REDIRECT above).
6325 BPF_LWT_REROUTE
= 128,
6326 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6327 * to indicate that no custom dissection was performed, and
6328 * fallback to standard dissector is requested.
6330 BPF_FLOW_DISSECTOR_CONTINUE
= 129,
6340 /* IP address also allows 1 and 2 bytes access */
6343 __u32 src_port
; /* host byte order */
6344 __be16 dst_port
; /* network byte order */
6345 __u16
:16; /* zero padding */
6349 __s32 rx_queue_mapping
;
6352 struct bpf_tcp_sock
{
6353 __u32 snd_cwnd
; /* Sending congestion window */
6354 __u32 srtt_us
; /* smoothed round trip time << 3 in usecs */
6356 __u32 snd_ssthresh
; /* Slow start size threshold */
6357 __u32 rcv_nxt
; /* What we want to receive next */
6358 __u32 snd_nxt
; /* Next sequence we send */
6359 __u32 snd_una
; /* First byte we want an ack for */
6360 __u32 mss_cache
; /* Cached effective mss, not including SACKS */
6361 __u32 ecn_flags
; /* ECN status bits. */
6362 __u32 rate_delivered
; /* saved rate sample: packets delivered */
6363 __u32 rate_interval_us
; /* saved rate sample: time elapsed */
6364 __u32 packets_out
; /* Packets which are "in flight" */
6365 __u32 retrans_out
; /* Retransmitted packets out */
6366 __u32 total_retrans
; /* Total retransmits for entire connection */
6367 __u32 segs_in
; /* RFC4898 tcpEStatsPerfSegsIn
6368 * total number of segments in.
6370 __u32 data_segs_in
; /* RFC4898 tcpEStatsPerfDataSegsIn
6371 * total number of data segments in.
6373 __u32 segs_out
; /* RFC4898 tcpEStatsPerfSegsOut
6374 * The total number of segments sent.
6376 __u32 data_segs_out
; /* RFC4898 tcpEStatsPerfDataSegsOut
6377 * total number of data segments sent.
6379 __u32 lost_out
; /* Lost packets */
6380 __u32 sacked_out
; /* SACK'd packets */
6381 __u64 bytes_received
; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
6382 * sum(delta(rcv_nxt)), or how many bytes
6385 __u64 bytes_acked
; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
6386 * sum(delta(snd_una)), or how many bytes
6389 __u32 dsack_dups
; /* RFC4898 tcpEStatsStackDSACKDups
6390 * total number of DSACK blocks received
6392 __u32 delivered
; /* Total data packets delivered incl. rexmits */
6393 __u32 delivered_ce
; /* Like the above but only ECE marked packets */
6394 __u32 icsk_retransmits
; /* Number of unrecovered [RTO] timeouts */
6397 struct bpf_sock_tuple
{
6414 /* (Simplified) user return codes for tcx prog type.
6415 * A valid tcx program must return one of these defined values. All other
6416 * return codes are reserved for future use. Must remain compatible with
6417 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6418 * return codes are mapped to TCX_NEXT.
6420 enum tcx_action_base
{
6427 struct bpf_xdp_sock
{
6431 #define XDP_PACKET_HEADROOM 256
6433 /* User return codes for XDP prog type.
6434 * A valid XDP program must return one of these defined values. All other
6435 * return codes are reserved for future use. Unknown return codes will
6436 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6446 /* user accessible metadata for XDP packet hook
6447 * new fields must be added to the end of this structure
6453 /* Below access go through struct xdp_rxq_info */
6454 __u32 ingress_ifindex
; /* rxq->dev->ifindex */
6455 __u32 rx_queue_index
; /* rxq->queue_index */
6457 __u32 egress_ifindex
; /* txq->dev->ifindex */
6460 /* DEVMAP map-value layout
6462 * The struct data-layout of map-value is a configuration interface.
6463 * New members can only be added to the end of this structure.
6465 struct bpf_devmap_val
{
6466 __u32 ifindex
; /* device index */
6468 int fd
; /* prog fd on map write */
6469 __u32 id
; /* prog id on map read */
6473 /* CPUMAP map-value layout
6475 * The struct data-layout of map-value is a configuration interface.
6476 * New members can only be added to the end of this structure.
6478 struct bpf_cpumap_val
{
6479 __u32 qsize
; /* queue size to remote target CPU */
6481 int fd
; /* prog fd on map write */
6482 __u32 id
; /* prog id on map read */
6491 /* user accessible metadata for SK_MSG packet hook, new fields must
6492 * be added to the end of this structure
6495 __bpf_md_ptr(void *, data
);
6496 __bpf_md_ptr(void *, data_end
);
6499 __u32 remote_ip4
; /* Stored in network byte order */
6500 __u32 local_ip4
; /* Stored in network byte order */
6501 __u32 remote_ip6
[4]; /* Stored in network byte order */
6502 __u32 local_ip6
[4]; /* Stored in network byte order */
6503 __u32 remote_port
; /* Stored in network byte order */
6504 __u32 local_port
; /* stored in host byte order */
6505 __u32 size
; /* Total size of sk_msg */
6507 __bpf_md_ptr(struct bpf_sock
*, sk
); /* current socket */
6510 struct sk_reuseport_md
{
6512 * Start of directly accessible data. It begins from
6513 * the tcp/udp header.
6515 __bpf_md_ptr(void *, data
);
6516 /* End of directly accessible data */
6517 __bpf_md_ptr(void *, data_end
);
6519 * Total length of packet (starting from the tcp/udp header).
6520 * Note that the directly accessible bytes (data_end - data)
6521 * could be less than this "len". Those bytes could be
6522 * indirectly read by a helper "bpf_skb_load_bytes()".
6526 * Eth protocol in the mac header (network byte order). e.g.
6527 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6530 __u32 ip_protocol
; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6531 __u32 bind_inany
; /* Is sock bound to an INANY address? */
6532 __u32 hash
; /* A hash of the packet 4 tuples */
6533 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6534 * new incoming connection request (e.g. selecting a listen sk for
6535 * the received SYN in the TCP case). reuse->sk is one of the sk
6536 * in the reuseport group. The bpf prog can use reuse->sk to learn
6537 * the local listening ip/port without looking into the skb.
6539 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6540 * reuse->migrating_sk is the socket that needs to be migrated
6541 * to another listening socket. migrating_sk could be a fullsock
6542 * sk that is fully established or a reqsk that is in-the-middle
6543 * of 3-way handshake.
6545 __bpf_md_ptr(struct bpf_sock
*, sk
);
6546 __bpf_md_ptr(struct bpf_sock
*, migrating_sk
);
6549 #define BPF_TAG_SIZE 8
6551 struct bpf_prog_info
{
6554 __u8 tag
[BPF_TAG_SIZE
];
6555 __u32 jited_prog_len
;
6556 __u32 xlated_prog_len
;
6557 __aligned_u64 jited_prog_insns
;
6558 __aligned_u64 xlated_prog_insns
;
6559 __u64 load_time
; /* ns since boottime */
6560 __u32 created_by_uid
;
6562 __aligned_u64 map_ids
;
6563 char name
[BPF_OBJ_NAME_LEN
];
6565 __u32 gpl_compatible
:1;
6566 __u32
:31; /* alignment pad */
6569 __u32 nr_jited_ksyms
;
6570 __u32 nr_jited_func_lens
;
6571 __aligned_u64 jited_ksyms
;
6572 __aligned_u64 jited_func_lens
;
6574 __u32 func_info_rec_size
;
6575 __aligned_u64 func_info
;
6578 __aligned_u64 line_info
;
6579 __aligned_u64 jited_line_info
;
6580 __u32 nr_jited_line_info
;
6581 __u32 line_info_rec_size
;
6582 __u32 jited_line_info_rec_size
;
6584 __aligned_u64 prog_tags
;
6587 __u64 recursion_misses
;
6588 __u32 verified_insns
;
6589 __u32 attach_btf_obj_id
;
6590 __u32 attach_btf_id
;
6591 } __attribute__((aligned(8)));
6593 struct bpf_map_info
{
6600 char name
[BPF_OBJ_NAME_LEN
];
6602 __u32 btf_vmlinux_value_type_id
;
6606 __u32 btf_key_type_id
;
6607 __u32 btf_value_type_id
;
6608 __u32 btf_vmlinux_id
;
6610 } __attribute__((aligned(8)));
6612 struct bpf_btf_info
{
6619 } __attribute__((aligned(8)));
6621 struct bpf_link_info
{
6627 __aligned_u64 tp_name
; /* in/out: tp_name buffer ptr */
6628 __u32 tp_name_len
; /* in/out: tp_name buffer len */
6632 __u32 target_obj_id
; /* prog_id for PROG_EXT, otherwise btf object id */
6633 __u32 target_btf_id
; /* BTF type id inside the object */
6640 __aligned_u64 target_name
; /* in/out: target_name buffer ptr */
6641 __u32 target_name_len
; /* in/out: target_name buffer len */
6643 /* If the iter specific field is 32 bits, it can be put
6644 * in the first or second union. Otherwise it should be
6645 * put in the second union.
6680 __aligned_u64 addrs
;
6681 __u32 count
; /* in/out: kprobe_multi function count */
6684 __aligned_u64 cookies
;
6688 __aligned_u64 offsets
;
6689 __aligned_u64 ref_ctr_offsets
;
6690 __aligned_u64 cookies
;
6691 __u32 path_size
; /* in/out: real path size on success, including zero byte */
6692 __u32 count
; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
6697 __u32 type
; /* enum bpf_perf_event_type */
6701 __aligned_u64 file_name
; /* in/out */
6703 __u32 offset
; /* offset from file_name */
6705 } uprobe
; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6707 __aligned_u64 func_name
; /* in/out */
6709 __u32 offset
; /* offset from func_name */
6713 } kprobe
; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6715 __aligned_u64 tp_name
; /* in/out */
6719 } tracepoint
; /* BPF_PERF_EVENT_TRACEPOINT */
6725 } event
; /* BPF_PERF_EVENT_EVENT */
6741 } __attribute__((aligned(8)));
6743 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6744 * by user and intended to be used by socket (e.g. to bind to, depends on
6747 struct bpf_sock_addr
{
6748 __u32 user_family
; /* Allows 4-byte read, but no write. */
6749 __u32 user_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
6750 * Stored in network byte order.
6752 __u32 user_ip6
[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6753 * Stored in network byte order.
6755 __u32 user_port
; /* Allows 1,2,4-byte read and 4-byte write.
6756 * Stored in network byte order
6758 __u32 family
; /* Allows 4-byte read, but no write */
6759 __u32 type
; /* Allows 4-byte read, but no write */
6760 __u32 protocol
; /* Allows 4-byte read, but no write */
6761 __u32 msg_src_ip4
; /* Allows 1,2,4-byte read and 4-byte write.
6762 * Stored in network byte order.
6764 __u32 msg_src_ip6
[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6765 * Stored in network byte order.
6767 __bpf_md_ptr(struct bpf_sock
*, sk
);
6770 /* User bpf_sock_ops struct to access socket values and specify request ops
6771 * and their replies.
6772 * Some of this fields are in network (bigendian) byte order and may need
6773 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6774 * New fields can only be added at the end of this structure
6776 struct bpf_sock_ops
{
6779 __u32 args
[4]; /* Optionally passed to bpf program */
6780 __u32 reply
; /* Returned by bpf program */
6781 __u32 replylong
[4]; /* Optionally returned by bpf prog */
6784 __u32 remote_ip4
; /* Stored in network byte order */
6785 __u32 local_ip4
; /* Stored in network byte order */
6786 __u32 remote_ip6
[4]; /* Stored in network byte order */
6787 __u32 local_ip6
[4]; /* Stored in network byte order */
6788 __u32 remote_port
; /* Stored in network byte order */
6789 __u32 local_port
; /* stored in host byte order */
6790 __u32 is_fullsock
; /* Some TCP fields are only valid if
6791 * there is a full socket. If not, the
6792 * fields read as zero.
6795 __u32 srtt_us
; /* Averaged RTT << 3 in usecs */
6796 __u32 bpf_sock_ops_cb_flags
; /* flags defined in uapi/linux/tcp.h */
6805 __u32 rate_delivered
;
6806 __u32 rate_interval_us
;
6809 __u32 total_retrans
;
6813 __u32 data_segs_out
;
6817 __u64 bytes_received
;
6819 __bpf_md_ptr(struct bpf_sock
*, sk
);
6820 /* [skb_data, skb_data_end) covers the whole TCP header.
6822 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6823 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
6824 * header has not been written.
6825 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6826 * been written so far.
6827 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
6829 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6832 * bpf_load_hdr_opt() can also be used to read a particular option.
6834 __bpf_md_ptr(void *, skb_data
);
6835 __bpf_md_ptr(void *, skb_data_end
);
6836 __u32 skb_len
; /* The total length of a packet.
6837 * It includes the header, options,
6840 __u32 skb_tcp_flags
; /* tcp_flags of the header. It provides
6841 * an easy way to check for tcp_flags
6842 * without parsing skb_data.
6844 * In particular, the skb_tcp_flags
6845 * will still be available in
6846 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6847 * the outgoing header has not
6853 /* Definitions for bpf_sock_ops_cb_flags */
6855 BPF_SOCK_OPS_RTO_CB_FLAG
= (1<<0),
6856 BPF_SOCK_OPS_RETRANS_CB_FLAG
= (1<<1),
6857 BPF_SOCK_OPS_STATE_CB_FLAG
= (1<<2),
6858 BPF_SOCK_OPS_RTT_CB_FLAG
= (1<<3),
6859 /* Call bpf for all received TCP headers. The bpf prog will be
6860 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6862 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6863 * for the header option related helpers that will be useful
6864 * to the bpf programs.
6866 * It could be used at the client/active side (i.e. connect() side)
6867 * when the server told it that the server was in syncookie
6868 * mode and required the active side to resend the bpf-written
6869 * options. The active side can keep writing the bpf-options until
6870 * it received a valid packet from the server side to confirm
6871 * the earlier packet (and options) has been received. The later
6872 * example patch is using it like this at the active side when the
6873 * server is in syncookie mode.
6875 * The bpf prog will usually turn this off in the common cases.
6877 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG
= (1<<4),
6878 /* Call bpf when kernel has received a header option that
6879 * the kernel cannot handle. The bpf prog will be called under
6880 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6882 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6883 * for the header option related helpers that will be useful
6884 * to the bpf programs.
6886 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG
= (1<<5),
6887 /* Call bpf when the kernel is writing header options for the
6888 * outgoing packet. The bpf prog will first be called
6889 * to reserve space in a skb under
6890 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
6891 * the bpf prog will be called to write the header option(s)
6892 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6894 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6895 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6896 * related helpers that will be useful to the bpf programs.
6898 * The kernel gets its chance to reserve space and write
6899 * options first before the BPF program does.
6901 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG
= (1<<6),
6902 /* Mask of all currently supported cb flags */
6903 BPF_SOCK_OPS_ALL_CB_FLAGS
= 0x7F,
6906 /* List of known BPF sock_ops operators.
6907 * New entries can only be added at the end
6911 BPF_SOCK_OPS_TIMEOUT_INIT
, /* Should return SYN-RTO value to use or
6912 * -1 if default value should be used
6914 BPF_SOCK_OPS_RWND_INIT
, /* Should return initial advertized
6915 * window (in packets) or -1 if default
6916 * value should be used
6918 BPF_SOCK_OPS_TCP_CONNECT_CB
, /* Calls BPF program right before an
6919 * active connection is initialized
6921 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB
, /* Calls BPF program when an
6922 * active connection is
6925 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
, /* Calls BPF program when a
6926 * passive connection is
6929 BPF_SOCK_OPS_NEEDS_ECN
, /* If connection's congestion control
6932 BPF_SOCK_OPS_BASE_RTT
, /* Get base RTT. The correct value is
6933 * based on the path and may be
6934 * dependent on the congestion control
6935 * algorithm. In general it indicates
6936 * a congestion threshold. RTTs above
6937 * this indicate congestion
6939 BPF_SOCK_OPS_RTO_CB
, /* Called when an RTO has triggered.
6940 * Arg1: value of icsk_retransmits
6941 * Arg2: value of icsk_rto
6942 * Arg3: whether RTO has expired
6944 BPF_SOCK_OPS_RETRANS_CB
, /* Called when skb is retransmitted.
6945 * Arg1: sequence number of 1st byte
6947 * Arg3: return value of
6948 * tcp_transmit_skb (0 => success)
6950 BPF_SOCK_OPS_STATE_CB
, /* Called when TCP changes state.
6954 BPF_SOCK_OPS_TCP_LISTEN_CB
, /* Called on listen(2), right after
6955 * socket transition to LISTEN state.
6957 BPF_SOCK_OPS_RTT_CB
, /* Called on every RTT.
6958 * Arg1: measured RTT input (mrtt)
6959 * Arg2: updated srtt
6961 BPF_SOCK_OPS_PARSE_HDR_OPT_CB
, /* Parse the header option.
6962 * It will be called to handle
6963 * the packets received at
6964 * an already established
6967 * sock_ops->skb_data:
6968 * Referring to the received skb.
6969 * It covers the TCP header only.
6971 * bpf_load_hdr_opt() can also
6972 * be used to search for a
6973 * particular option.
6975 BPF_SOCK_OPS_HDR_OPT_LEN_CB
, /* Reserve space for writing the
6976 * header option later in
6977 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6978 * Arg1: bool want_cookie. (in
6979 * writing SYNACK only)
6981 * sock_ops->skb_data:
6982 * Not available because no header has
6985 * sock_ops->skb_tcp_flags:
6986 * The tcp_flags of the
6987 * outgoing skb. (e.g. SYN, ACK, FIN).
6989 * bpf_reserve_hdr_opt() should
6990 * be used to reserve space.
6992 BPF_SOCK_OPS_WRITE_HDR_OPT_CB
, /* Write the header options
6993 * Arg1: bool want_cookie. (in
6994 * writing SYNACK only)
6996 * sock_ops->skb_data:
6997 * Referring to the outgoing skb.
6998 * It covers the TCP header
6999 * that has already been written
7000 * by the kernel and the
7001 * earlier bpf-progs.
7003 * sock_ops->skb_tcp_flags:
7004 * The tcp_flags of the outgoing
7005 * skb. (e.g. SYN, ACK, FIN).
7007 * bpf_store_hdr_opt() should
7008 * be used to write the
7011 * bpf_load_hdr_opt() can also
7012 * be used to search for a
7013 * particular option that
7014 * has already been written
7015 * by the kernel or the
7016 * earlier bpf-progs.
7020 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
7021 * changes between the TCP and BPF versions. Ideally this should never happen.
7022 * If it does, we need to add code to convert them before calling
7023 * the BPF sock_ops function.
7026 BPF_TCP_ESTABLISHED
= 1,
7036 BPF_TCP_CLOSING
, /* Now a valid state */
7037 BPF_TCP_NEW_SYN_RECV
,
7038 BPF_TCP_BOUND_INACTIVE
,
7040 BPF_TCP_MAX_STATES
/* Leave at the end! */
7044 TCP_BPF_IW
= 1001, /* Set TCP initial congestion window */
7045 TCP_BPF_SNDCWND_CLAMP
= 1002, /* Set sndcwnd_clamp */
7046 TCP_BPF_DELACK_MAX
= 1003, /* Max delay ack in usecs */
7047 TCP_BPF_RTO_MIN
= 1004, /* Min delay ack in usecs */
7048 /* Copy the SYN pkt to optval
7050 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
7051 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
7052 * to only getting from the saved_syn. It can either get the
7055 * 1. the just-received SYN packet (only available when writing the
7056 * SYNACK). It will be useful when it is not necessary to
7057 * save the SYN packet for latter use. It is also the only way
7058 * to get the SYN during syncookie mode because the syn
7059 * packet cannot be saved during syncookie.
7063 * 2. the earlier saved syn which was done by
7064 * bpf_setsockopt(TCP_SAVE_SYN).
7066 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
7067 * SYN packet is obtained.
7069 * If the bpf-prog does not need the IP[46] header, the
7070 * bpf-prog can avoid parsing the IP header by using
7071 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
7072 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
7074 * >0: Total number of bytes copied
7075 * -ENOSPC: Not enough space in optval. Only optlen number of
7077 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
7078 * is not saved by setsockopt(TCP_SAVE_SYN).
7080 TCP_BPF_SYN
= 1005, /* Copy the TCP header */
7081 TCP_BPF_SYN_IP
= 1006, /* Copy the IP[46] and TCP header */
7082 TCP_BPF_SYN_MAC
= 1007, /* Copy the MAC, IP[46], and TCP header */
7083 TCP_BPF_SOCK_OPS_CB_FLAGS
= 1008, /* Get or Set TCP sock ops flags */
7087 BPF_LOAD_HDR_OPT_TCP_SYN
= (1ULL << 0),
7090 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
7091 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
7094 BPF_WRITE_HDR_TCP_CURRENT_MSS
= 1, /* Kernel is finding the
7095 * total option spaces
7096 * required for an established
7097 * sk in order to calculate the
7098 * MSS. No skb is actually
7101 BPF_WRITE_HDR_TCP_SYNACK_COOKIE
= 2, /* Kernel is in syncookie mode
7102 * when sending a SYN.
7106 struct bpf_perf_event_value
{
7113 BPF_DEVCG_ACC_MKNOD
= (1ULL << 0),
7114 BPF_DEVCG_ACC_READ
= (1ULL << 1),
7115 BPF_DEVCG_ACC_WRITE
= (1ULL << 2),
7119 BPF_DEVCG_DEV_BLOCK
= (1ULL << 0),
7120 BPF_DEVCG_DEV_CHAR
= (1ULL << 1),
7123 struct bpf_cgroup_dev_ctx
{
7124 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
7130 struct bpf_raw_tracepoint_args
{
7134 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
7135 * OUTPUT: Do lookup from egress perspective; default is ingress
7138 BPF_FIB_LOOKUP_DIRECT
= (1U << 0),
7139 BPF_FIB_LOOKUP_OUTPUT
= (1U << 1),
7140 BPF_FIB_LOOKUP_SKIP_NEIGH
= (1U << 2),
7141 BPF_FIB_LOOKUP_TBID
= (1U << 3),
7142 BPF_FIB_LOOKUP_SRC
= (1U << 4),
7143 BPF_FIB_LOOKUP_MARK
= (1U << 5),
7147 BPF_FIB_LKUP_RET_SUCCESS
, /* lookup successful */
7148 BPF_FIB_LKUP_RET_BLACKHOLE
, /* dest is blackholed; can be dropped */
7149 BPF_FIB_LKUP_RET_UNREACHABLE
, /* dest is unreachable; can be dropped */
7150 BPF_FIB_LKUP_RET_PROHIBIT
, /* dest not allowed; can be dropped */
7151 BPF_FIB_LKUP_RET_NOT_FWDED
, /* packet is not forwarded */
7152 BPF_FIB_LKUP_RET_FWD_DISABLED
, /* fwding is not enabled on ingress */
7153 BPF_FIB_LKUP_RET_UNSUPP_LWT
, /* fwd requires encapsulation */
7154 BPF_FIB_LKUP_RET_NO_NEIGH
, /* no neighbor entry for nh */
7155 BPF_FIB_LKUP_RET_FRAG_NEEDED
, /* fragmentation required to fwd */
7156 BPF_FIB_LKUP_RET_NO_SRC_ADDR
, /* failed to derive IP src addr */
7159 struct bpf_fib_lookup
{
7160 /* input: network family for lookup (AF_INET, AF_INET6)
7161 * output: network family of egress nexthop
7165 /* set if lookup is to consider L4 data - e.g., FIB rules */
7170 union { /* used for MTU check */
7171 /* input to lookup */
7172 __u16 tot_len
; /* L3 length from network hdr (iph->tot_len) */
7174 /* output: MTU value */
7176 } __attribute__((packed
, aligned(2)));
7177 /* input: L3 device index for lookup
7178 * output: device index from FIB lookup
7183 /* inputs to lookup */
7184 __u8 tos
; /* AF_INET */
7185 __be32 flowinfo
; /* AF_INET6, flow_label + priority */
7187 /* output: metric of fib result (IPv4/IPv6 only) */
7191 /* input: source address to consider for lookup
7192 * output: source address result from lookup
7196 __u32 ipv6_src
[4]; /* in6_addr; network order */
7199 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7200 * network header. output: bpf_fib_lookup sets to gateway address
7201 * if FIB lookup returns gateway route
7205 __u32 ipv6_dst
[4]; /* in6_addr; network order */
7211 __be16 h_vlan_proto
;
7214 /* input: when accompanied with the
7215 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7216 * specific routing table to use for the fib lookup.
7224 __u32 mark
; /* policy routing */
7225 /* 2 4-byte holes for input */
7228 /* output: source and dest mac */
7230 __u8 smac
[6]; /* ETH_ALEN */
7231 __u8 dmac
[6]; /* ETH_ALEN */
7236 struct bpf_redir_neigh
{
7237 /* network family for lookup (AF_INET, AF_INET6) */
7239 /* network address of nexthop; skips fib lookup to find gateway */
7242 __u32 ipv6_nh
[4]; /* in6_addr; network order */
7246 /* bpf_check_mtu flags*/
7247 enum bpf_check_mtu_flags
{
7248 BPF_MTU_CHK_SEGS
= (1U << 0),
7251 enum bpf_check_mtu_ret
{
7252 BPF_MTU_CHK_RET_SUCCESS
, /* check and lookup successful */
7253 BPF_MTU_CHK_RET_FRAG_NEEDED
, /* fragmentation required to fwd */
7254 BPF_MTU_CHK_RET_SEGS_TOOBIG
, /* GSO re-segmentation needed to fwd */
7257 enum bpf_task_fd_type
{
7258 BPF_FD_TYPE_RAW_TRACEPOINT
, /* tp name */
7259 BPF_FD_TYPE_TRACEPOINT
, /* tp name */
7260 BPF_FD_TYPE_KPROBE
, /* (symbol + offset) or addr */
7261 BPF_FD_TYPE_KRETPROBE
, /* (symbol + offset) or addr */
7262 BPF_FD_TYPE_UPROBE
, /* filename + offset */
7263 BPF_FD_TYPE_URETPROBE
, /* filename + offset */
7267 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG
= (1U << 0),
7268 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL
= (1U << 1),
7269 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP
= (1U << 2),
7272 struct bpf_flow_keys
{
7275 __u16 addr_proto
; /* ETH_P_* of valid addrs */
7289 __u32 ipv6_src
[4]; /* in6_addr; network order */
7290 __u32 ipv6_dst
[4]; /* in6_addr; network order */
7297 struct bpf_func_info
{
7302 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
7303 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
7305 struct bpf_line_info
{
7307 __u32 file_name_off
;
7312 struct bpf_spin_lock
{
7318 } __attribute__((aligned(8)));
7322 } __attribute__((aligned(8)));
7326 } __attribute__((aligned(8)));
7328 struct bpf_list_head
{
7330 } __attribute__((aligned(8)));
7332 struct bpf_list_node
{
7334 } __attribute__((aligned(8)));
7336 struct bpf_rb_root
{
7338 } __attribute__((aligned(8)));
7340 struct bpf_rb_node
{
7342 } __attribute__((aligned(8)));
7344 struct bpf_refcount
{
7346 } __attribute__((aligned(4)));
7349 __u32 write
; /* Sysctl is being read (= 0) or written (= 1).
7350 * Allows 1,2,4-byte read, but no write.
7352 __u32 file_pos
; /* Sysctl file position to read from, write to.
7353 * Allows 1,2,4-byte read an 4-byte write.
7357 struct bpf_sockopt
{
7358 __bpf_md_ptr(struct bpf_sock
*, sk
);
7359 __bpf_md_ptr(void *, optval
);
7360 __bpf_md_ptr(void *, optval_end
);
7368 struct bpf_pidns_info
{
7373 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7374 struct bpf_sk_lookup
{
7376 __bpf_md_ptr(struct bpf_sock
*, sk
); /* Selected socket */
7377 __u64 cookie
; /* Non-zero if socket was selected in PROG_TEST_RUN */
7380 __u32 family
; /* Protocol family (AF_INET, AF_INET6) */
7381 __u32 protocol
; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7382 __u32 remote_ip4
; /* Network byte order */
7383 __u32 remote_ip6
[4]; /* Network byte order */
7384 __be16 remote_port
; /* Network byte order */
7385 __u16
:16; /* Zero padding */
7386 __u32 local_ip4
; /* Network byte order */
7387 __u32 local_ip6
[4]; /* Network byte order */
7388 __u32 local_port
; /* Host byte order */
7389 __u32 ingress_ifindex
; /* The arriving interface. Determined by inet_iif. */
7393 * struct btf_ptr is used for typed pointer representation; the
7394 * type id is used to render the pointer data as the appropriate type
7395 * via the bpf_snprintf_btf() helper described above. A flags field -
7396 * potentially to specify additional details about the BTF pointer
7397 * (rather than its mode of display) - is included for future use.
7398 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7403 __u32 flags
; /* BTF ptr flags; unused at present. */
7407 * Flags to control bpf_snprintf_btf() behaviour.
7408 * - BTF_F_COMPACT: no formatting around type information
7409 * - BTF_F_NONAME: no struct/union member names/types
7410 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7411 * equivalent to %px.
7412 * - BTF_F_ZERO: show zero-valued struct/union members; they
7413 * are not displayed by default
7416 BTF_F_COMPACT
= (1ULL << 0),
7417 BTF_F_NONAME
= (1ULL << 1),
7418 BTF_F_PTR_RAW
= (1ULL << 2),
7419 BTF_F_ZERO
= (1ULL << 3),
7422 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7423 * has to be adjusted by relocations. It is emitted by llvm and passed to
7424 * libbpf and later to the kernel.
7426 enum bpf_core_relo_kind
{
7427 BPF_CORE_FIELD_BYTE_OFFSET
= 0, /* field byte offset */
7428 BPF_CORE_FIELD_BYTE_SIZE
= 1, /* field size in bytes */
7429 BPF_CORE_FIELD_EXISTS
= 2, /* field existence in target kernel */
7430 BPF_CORE_FIELD_SIGNED
= 3, /* field signedness (0 - unsigned, 1 - signed) */
7431 BPF_CORE_FIELD_LSHIFT_U64
= 4, /* bitfield-specific left bitshift */
7432 BPF_CORE_FIELD_RSHIFT_U64
= 5, /* bitfield-specific right bitshift */
7433 BPF_CORE_TYPE_ID_LOCAL
= 6, /* type ID in local BPF object */
7434 BPF_CORE_TYPE_ID_TARGET
= 7, /* type ID in target kernel */
7435 BPF_CORE_TYPE_EXISTS
= 8, /* type existence in target kernel */
7436 BPF_CORE_TYPE_SIZE
= 9, /* type size in bytes */
7437 BPF_CORE_ENUMVAL_EXISTS
= 10, /* enum value existence in target kernel */
7438 BPF_CORE_ENUMVAL_VALUE
= 11, /* enum value integer value */
7439 BPF_CORE_TYPE_MATCHES
= 12, /* type match in target kernel */
7443 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7444 * and from libbpf to the kernel.
7446 * CO-RE relocation captures the following data:
7447 * - insn_off - instruction offset (in bytes) within a BPF program that needs
7448 * its insn->imm field to be relocated with actual field info;
7449 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7451 * - access_str_off - offset into corresponding .BTF string section. String
7452 * interpretation depends on specific relocation kind:
7453 * - for field-based relocations, string encodes an accessed field using
7454 * a sequence of field and array indices, separated by colon (:). It's
7455 * conceptually very close to LLVM's getelementptr ([0]) instruction's
7456 * arguments for identifying offset to a field.
7457 * - for type-based relocations, strings is expected to be just "0";
7458 * - for enum value-based relocations, string contains an index of enum
7459 * value within its enum type;
7460 * - kind - one of enum bpf_core_relo_kind;
7470 * struct sample *s = ...;
7471 * int *x = &s->a; // encoded as "0:0" (a is field #0)
7472 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
7473 * // b is field #0 inside anon struct, accessing elem #5)
7474 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7476 * type_id for all relocs in this example will capture BTF type id of
7479 * Such relocation is emitted when using __builtin_preserve_access_index()
7480 * Clang built-in, passing expression that captures field address, e.g.:
7482 * bpf_probe_read(&dst, sizeof(dst),
7483 * __builtin_preserve_access_index(&src->a.b.c));
7485 * In this case Clang will emit field relocation recording necessary data to
7486 * be able to find offset of embedded `a.b.c` field within `src` struct.
7488 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7490 struct bpf_core_relo
{
7493 __u32 access_str_off
;
7494 enum bpf_core_relo_kind kind
;
7498 * Flags to control bpf_timer_start() behaviour.
7499 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7500 * relative to current time.
7501 * - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
7504 BPF_F_TIMER_ABS
= (1ULL << 0),
7505 BPF_F_TIMER_CPU_PIN
= (1ULL << 1),
7508 /* BPF numbers iterator state */
7509 struct bpf_iter_num
{
7510 /* opaque iterator state; having __u64 here allows to preserve correct
7511 * alignment requirements in vmlinux.h, generated from BTF
7514 } __attribute__((aligned(8)));
7517 * Flags to control BPF kfunc behaviour.
7518 * - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective
7519 * helper documentation for details.)
7521 enum bpf_kfunc_flags
{
7522 BPF_F_PAD_ZEROS
= (1ULL << 0),
7525 #endif /* _UAPI__LINUX_BPF_H__ */