1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */
26 #define BPF_MOV 0xb0 /* mov reg to reg */
27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
29 /* change endianness of a register */
30 #define BPF_END 0xd0 /* flags for endianness conversion: */
31 #define BPF_TO_LE 0x00 /* convert to little-endian */
32 #define BPF_TO_BE 0x08 /* convert to big-endian */
33 #define BPF_FROM_LE BPF_TO_LE
34 #define BPF_FROM_BE BPF_TO_BE
37 #define BPF_JNE 0x50 /* jump != */
38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
44 #define BPF_CALL 0x80 /* function call */
45 #define BPF_EXIT 0x90 /* function return */
47 /* atomic op type fields (stored in immediate) */
48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
52 /* Register numbers */
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG __MAX_BPF_REG
72 __u8 code; /* opcode */
73 __u8 dst_reg:4; /* dest register */
74 __u8 src_reg:4; /* source register */
75 __s16 off; /* signed offset */
76 __s32 imm; /* signed immediate constant */
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
82 __u8 data[]; /* Arbitrary size */
85 struct bpf_cgroup_storage_key {
86 __u64 cgroup_inode_id; /* cgroup inode id */
87 __u32 attach_type; /* program attach type (enum bpf_attach_type) */
90 union bpf_iter_link_info {
96 /* BPF syscall commands, see bpf(2) man-page for more details. */
98 * DOC: eBPF Syscall Preamble
100 * The operation to be performed by the **bpf**\ () system call is determined
101 * by the *cmd* argument. Each operation takes an accompanying argument,
102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
103 * below). The size argument is the size of the union pointed to by *attr*.
106 * DOC: eBPF Syscall Commands
110 * Create a map and return a file descriptor that refers to the
111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
112 * is automatically enabled for the new file descriptor.
114 * Applying **close**\ (2) to the file descriptor returned by
115 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
118 * A new file descriptor (a nonnegative integer), or -1 if an
119 * error occurred (in which case, *errno* is set appropriately).
121 * BPF_MAP_LOOKUP_ELEM
123 * Look up an element with a given *key* in the map referred to
124 * by the file descriptor *map_fd*.
126 * The *flags* argument may be specified as one of the
130 * Look up the value of a spin-locked map without
131 * returning the lock. This must be specified if the
132 * elements contain a spinlock.
135 * Returns zero on success. On error, -1 is returned and *errno*
136 * is set appropriately.
138 * BPF_MAP_UPDATE_ELEM
140 * Create or update an element (key/value pair) in a specified map.
142 * The *flags* argument should be specified as one of the
146 * Create a new element or update an existing element.
148 * Create a new element only if it did not exist.
150 * Update an existing element.
152 * Update a spin_lock-ed map element.
155 * Returns zero on success. On error, -1 is returned and *errno*
156 * is set appropriately.
158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
159 * **E2BIG**, **EEXIST**, or **ENOENT**.
162 * The number of elements in the map reached the
163 * *max_entries* limit specified at map creation time.
165 * If *flags* specifies **BPF_NOEXIST** and the element
166 * with *key* already exists in the map.
168 * If *flags* specifies **BPF_EXIST** and the element with
169 * *key* does not exist in the map.
171 * BPF_MAP_DELETE_ELEM
173 * Look up and delete an element by key in a specified map.
176 * Returns zero on success. On error, -1 is returned and *errno*
177 * is set appropriately.
179 * BPF_MAP_GET_NEXT_KEY
181 * Look up an element by key in a specified map and return the key
182 * of the next element. Can be used to iterate over all elements
186 * Returns zero on success. On error, -1 is returned and *errno*
187 * is set appropriately.
189 * The following cases can be used to iterate over all elements of
192 * * If *key* is not found, the operation returns zero and sets
193 * the *next_key* pointer to the key of the first element.
194 * * If *key* is found, the operation returns zero and sets the
195 * *next_key* pointer to the key of the next element.
196 * * If *key* is the last element, returns -1 and *errno* is set
199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
200 * **EINVAL** on error.
204 * Verify and load an eBPF program, returning a new file
205 * descriptor associated with the program.
207 * Applying **close**\ (2) to the file descriptor returned by
208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
211 * automatically enabled for the new file descriptor.
214 * A new file descriptor (a nonnegative integer), or -1 if an
215 * error occurred (in which case, *errno* is set appropriately).
219 * Pin an eBPF program or map referred by the specified *bpf_fd*
220 * to the provided *pathname* on the filesystem.
222 * The *pathname* argument must not contain a dot (".").
224 * On success, *pathname* retains a reference to the eBPF object,
225 * preventing deallocation of the object when the original
226 * *bpf_fd* is closed. This allow the eBPF object to live beyond
227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
230 * Applying **unlink**\ (2) or similar calls to the *pathname*
231 * unpins the object from the filesystem, removing the reference.
232 * If no other file descriptors or filesystem nodes refer to the
233 * same object, it will be deallocated (see NOTES).
235 * The filesystem type for the parent directory of *pathname* must
236 * be **BPF_FS_MAGIC**.
239 * Returns zero on success. On error, -1 is returned and *errno*
240 * is set appropriately.
244 * Open a file descriptor for the eBPF object pinned to the
245 * specified *pathname*.
248 * A new file descriptor (a nonnegative integer), or -1 if an
249 * error occurred (in which case, *errno* is set appropriately).
253 * Attach an eBPF program to a *target_fd* at the specified
254 * *attach_type* hook.
256 * The *attach_type* specifies the eBPF attachment point to
257 * attach the program to, and must be one of *bpf_attach_type*
260 * The *attach_bpf_fd* must be a valid file descriptor for a
261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
262 * or sock_ops type corresponding to the specified *attach_type*.
264 * The *target_fd* must be a valid file descriptor for a kernel
265 * object which depends on the attach type of *attach_bpf_fd*:
267 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
268 * **BPF_PROG_TYPE_CGROUP_SKB**,
269 * **BPF_PROG_TYPE_CGROUP_SOCK**,
270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
273 * **BPF_PROG_TYPE_SOCK_OPS**
275 * Control Group v2 hierarchy with the eBPF controller
276 * enabled. Requires the kernel to be compiled with
277 * **CONFIG_CGROUP_BPF**.
279 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
281 * Network namespace (eg /proc/self/ns/net).
283 * **BPF_PROG_TYPE_LIRC_MODE2**
285 * LIRC device path (eg /dev/lircN). Requires the kernel
286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
288 * **BPF_PROG_TYPE_SK_SKB**,
289 * **BPF_PROG_TYPE_SK_MSG**
291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
294 * Returns zero on success. On error, -1 is returned and *errno*
295 * is set appropriately.
299 * Detach the eBPF program associated with the *target_fd* at the
300 * hook specified by *attach_type*. The program must have been
301 * previously attached using **BPF_PROG_ATTACH**.
304 * Returns zero on success. On error, -1 is returned and *errno*
305 * is set appropriately.
309 * Run the eBPF program associated with the *prog_fd* a *repeat*
310 * number of times against a provided program context *ctx_in* and
311 * data *data_in*, and return the modified program context
312 * *ctx_out*, *data_out* (for example, packet data), result of the
313 * execution *retval*, and *duration* of the test run.
315 * The sizes of the buffers provided as input and output
316 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
317 * be provided in the corresponding variables *ctx_size_in*,
318 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
319 * of these parameters are not provided (ie set to NULL), the
320 * corresponding size field must be zero.
322 * Some program types have particular requirements:
324 * **BPF_PROG_TYPE_SK_LOOKUP**
325 * *data_in* and *data_out* must be NULL.
327 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
328 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
330 * *ctx_out*, *data_in* and *data_out* must be NULL.
331 * *repeat* must be zero.
333 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
336 * Returns zero on success. On error, -1 is returned and *errno*
337 * is set appropriately.
340 * Either *data_size_out* or *ctx_size_out* is too small.
342 * This command is not supported by the program type of
343 * the program referred to by *prog_fd*.
345 * BPF_PROG_GET_NEXT_ID
347 * Fetch the next eBPF program currently loaded into the kernel.
349 * Looks for the eBPF program with an id greater than *start_id*
350 * and updates *next_id* on success. If no other eBPF programs
351 * remain with ids higher than *start_id*, returns -1 and sets
352 * *errno* to **ENOENT**.
355 * Returns zero on success. On error, or when no id remains, -1
356 * is returned and *errno* is set appropriately.
358 * BPF_MAP_GET_NEXT_ID
360 * Fetch the next eBPF map currently loaded into the kernel.
362 * Looks for the eBPF map with an id greater than *start_id*
363 * and updates *next_id* on success. If no other eBPF maps
364 * remain with ids higher than *start_id*, returns -1 and sets
365 * *errno* to **ENOENT**.
368 * Returns zero on success. On error, or when no id remains, -1
369 * is returned and *errno* is set appropriately.
371 * BPF_PROG_GET_FD_BY_ID
373 * Open a file descriptor for the eBPF program corresponding to
377 * A new file descriptor (a nonnegative integer), or -1 if an
378 * error occurred (in which case, *errno* is set appropriately).
380 * BPF_MAP_GET_FD_BY_ID
382 * Open a file descriptor for the eBPF map corresponding to
386 * A new file descriptor (a nonnegative integer), or -1 if an
387 * error occurred (in which case, *errno* is set appropriately).
389 * BPF_OBJ_GET_INFO_BY_FD
391 * Obtain information about the eBPF object corresponding to
394 * Populates up to *info_len* bytes of *info*, which will be in
395 * one of the following formats depending on the eBPF object type
398 * * **struct bpf_prog_info**
399 * * **struct bpf_map_info**
400 * * **struct bpf_btf_info**
401 * * **struct bpf_link_info**
404 * Returns zero on success. On error, -1 is returned and *errno*
405 * is set appropriately.
409 * Obtain information about eBPF programs associated with the
410 * specified *attach_type* hook.
412 * The *target_fd* must be a valid file descriptor for a kernel
413 * object which depends on the attach type of *attach_bpf_fd*:
415 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
416 * **BPF_PROG_TYPE_CGROUP_SKB**,
417 * **BPF_PROG_TYPE_CGROUP_SOCK**,
418 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
419 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
420 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
421 * **BPF_PROG_TYPE_SOCK_OPS**
423 * Control Group v2 hierarchy with the eBPF controller
424 * enabled. Requires the kernel to be compiled with
425 * **CONFIG_CGROUP_BPF**.
427 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
429 * Network namespace (eg /proc/self/ns/net).
431 * **BPF_PROG_TYPE_LIRC_MODE2**
433 * LIRC device path (eg /dev/lircN). Requires the kernel
434 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
436 * **BPF_PROG_QUERY** always fetches the number of programs
437 * attached and the *attach_flags* which were used to attach those
438 * programs. Additionally, if *prog_ids* is nonzero and the number
439 * of attached programs is less than *prog_cnt*, populates
440 * *prog_ids* with the eBPF program ids of the programs attached
443 * The following flags may alter the result:
445 * **BPF_F_QUERY_EFFECTIVE**
446 * Only return information regarding programs which are
447 * currently effective at the specified *target_fd*.
450 * Returns zero on success. On error, -1 is returned and *errno*
451 * is set appropriately.
453 * BPF_RAW_TRACEPOINT_OPEN
455 * Attach an eBPF program to a tracepoint *name* to access kernel
456 * internal arguments of the tracepoint in their raw form.
458 * The *prog_fd* must be a valid file descriptor associated with
459 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
461 * No ABI guarantees are made about the content of tracepoint
462 * arguments exposed to the corresponding eBPF program.
464 * Applying **close**\ (2) to the file descriptor returned by
465 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
468 * A new file descriptor (a nonnegative integer), or -1 if an
469 * error occurred (in which case, *errno* is set appropriately).
473 * Verify and load BPF Type Format (BTF) metadata into the kernel,
474 * returning a new file descriptor associated with the metadata.
475 * BTF is described in more detail at
476 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
478 * The *btf* parameter must point to valid memory providing
479 * *btf_size* bytes of BTF binary metadata.
481 * The returned file descriptor can be passed to other **bpf**\ ()
482 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
483 * associate the BTF with those objects.
485 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
486 * parameters to specify a *btf_log_buf*, *btf_log_size* and
487 * *btf_log_level* which allow the kernel to return freeform log
488 * output regarding the BTF verification process.
491 * A new file descriptor (a nonnegative integer), or -1 if an
492 * error occurred (in which case, *errno* is set appropriately).
494 * BPF_BTF_GET_FD_BY_ID
496 * Open a file descriptor for the BPF Type Format (BTF)
497 * corresponding to *btf_id*.
500 * A new file descriptor (a nonnegative integer), or -1 if an
501 * error occurred (in which case, *errno* is set appropriately).
505 * Obtain information about eBPF programs associated with the
506 * target process identified by *pid* and *fd*.
508 * If the *pid* and *fd* are associated with a tracepoint, kprobe
509 * or uprobe perf event, then the *prog_id* and *fd_type* will
510 * be populated with the eBPF program id and file descriptor type
511 * of type **bpf_task_fd_type**. If associated with a kprobe or
512 * uprobe, the *probe_offset* and *probe_addr* will also be
513 * populated. Optionally, if *buf* is provided, then up to
514 * *buf_len* bytes of *buf* will be populated with the name of
515 * the tracepoint, kprobe or uprobe.
517 * The resulting *prog_id* may be introspected in deeper detail
518 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
521 * Returns zero on success. On error, -1 is returned and *errno*
522 * is set appropriately.
524 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
526 * Look up an element with the given *key* in the map referred to
527 * by the file descriptor *fd*, and if found, delete the element.
529 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
530 * types, the *flags* argument needs to be set to 0, but for other
531 * map types, it may be specified as:
534 * Look up and delete the value of a spin-locked map
535 * without returning the lock. This must be specified if
536 * the elements contain a spinlock.
538 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
539 * implement this command as a "pop" operation, deleting the top
540 * element rather than one corresponding to *key*.
541 * The *key* and *key_len* parameters should be zeroed when
542 * issuing this operation for these map types.
544 * This command is only valid for the following map types:
545 * * **BPF_MAP_TYPE_QUEUE**
546 * * **BPF_MAP_TYPE_STACK**
547 * * **BPF_MAP_TYPE_HASH**
548 * * **BPF_MAP_TYPE_PERCPU_HASH**
549 * * **BPF_MAP_TYPE_LRU_HASH**
550 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
553 * Returns zero on success. On error, -1 is returned and *errno*
554 * is set appropriately.
558 * Freeze the permissions of the specified map.
560 * Write permissions may be frozen by passing zero *flags*.
561 * Upon success, no future syscall invocations may alter the
562 * map state of *map_fd*. Write operations from eBPF programs
563 * are still possible for a frozen map.
565 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
568 * Returns zero on success. On error, -1 is returned and *errno*
569 * is set appropriately.
571 * BPF_BTF_GET_NEXT_ID
573 * Fetch the next BPF Type Format (BTF) object currently loaded
576 * Looks for the BTF object with an id greater than *start_id*
577 * and updates *next_id* on success. If no other BTF objects
578 * remain with ids higher than *start_id*, returns -1 and sets
579 * *errno* to **ENOENT**.
582 * Returns zero on success. On error, or when no id remains, -1
583 * is returned and *errno* is set appropriately.
585 * BPF_MAP_LOOKUP_BATCH
587 * Iterate and fetch multiple elements in a map.
589 * Two opaque values are used to manage batch operations,
590 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
591 * to NULL to begin the batched operation. After each subsequent
592 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
593 * *out_batch* as the *in_batch* for the next operation to
594 * continue iteration from the current point.
596 * The *keys* and *values* are output parameters which must point
597 * to memory large enough to hold *count* items based on the key
598 * and value size of the map *map_fd*. The *keys* buffer must be
599 * of *key_size* * *count*. The *values* buffer must be of
600 * *value_size* * *count*.
602 * The *elem_flags* argument may be specified as one of the
606 * Look up the value of a spin-locked map without
607 * returning the lock. This must be specified if the
608 * elements contain a spinlock.
610 * On success, *count* elements from the map are copied into the
611 * user buffer, with the keys copied into *keys* and the values
612 * copied into the corresponding indices in *values*.
614 * If an error is returned and *errno* is not **EFAULT**, *count*
615 * is set to the number of successfully processed elements.
618 * Returns zero on success. On error, -1 is returned and *errno*
619 * is set appropriately.
621 * May set *errno* to **ENOSPC** to indicate that *keys* or
622 * *values* is too small to dump an entire bucket during
623 * iteration of a hash-based map type.
625 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
627 * Iterate and delete all elements in a map.
629 * This operation has the same behavior as
630 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
632 * * Every element that is successfully returned is also deleted
633 * from the map. This is at least *count* elements. Note that
634 * *count* is both an input and an output parameter.
635 * * Upon returning with *errno* set to **EFAULT**, up to
636 * *count* elements may be deleted without returning the keys
637 * and values of the deleted elements.
640 * Returns zero on success. On error, -1 is returned and *errno*
641 * is set appropriately.
643 * BPF_MAP_UPDATE_BATCH
645 * Update multiple elements in a map by *key*.
647 * The *keys* and *values* are input parameters which must point
648 * to memory large enough to hold *count* items based on the key
649 * and value size of the map *map_fd*. The *keys* buffer must be
650 * of *key_size* * *count*. The *values* buffer must be of
651 * *value_size* * *count*.
653 * Each element specified in *keys* is sequentially updated to the
654 * value in the corresponding index in *values*. The *in_batch*
655 * and *out_batch* parameters are ignored and should be zeroed.
657 * The *elem_flags* argument should be specified as one of the
661 * Create new elements or update a existing elements.
663 * Create new elements only if they do not exist.
665 * Update existing elements.
667 * Update spin_lock-ed map elements. This must be
668 * specified if the map value contains a spinlock.
670 * On success, *count* elements from the map are updated.
672 * If an error is returned and *errno* is not **EFAULT**, *count*
673 * is set to the number of successfully processed elements.
676 * Returns zero on success. On error, -1 is returned and *errno*
677 * is set appropriately.
679 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
680 * **E2BIG**. **E2BIG** indicates that the number of elements in
681 * the map reached the *max_entries* limit specified at map
684 * May set *errno* to one of the following error codes under
685 * specific circumstances:
688 * If *flags* specifies **BPF_NOEXIST** and the element
689 * with *key* already exists in the map.
691 * If *flags* specifies **BPF_EXIST** and the element with
692 * *key* does not exist in the map.
694 * BPF_MAP_DELETE_BATCH
696 * Delete multiple elements in a map by *key*.
698 * The *keys* parameter is an input parameter which must point
699 * to memory large enough to hold *count* items based on the key
700 * size of the map *map_fd*, that is, *key_size* * *count*.
702 * Each element specified in *keys* is sequentially deleted. The
703 * *in_batch*, *out_batch*, and *values* parameters are ignored
704 * and should be zeroed.
706 * The *elem_flags* argument may be specified as one of the
710 * Look up the value of a spin-locked map without
711 * returning the lock. This must be specified if the
712 * elements contain a spinlock.
714 * On success, *count* elements from the map are updated.
716 * If an error is returned and *errno* is not **EFAULT**, *count*
717 * is set to the number of successfully processed elements. If
718 * *errno* is **EFAULT**, up to *count* elements may be been
722 * Returns zero on success. On error, -1 is returned and *errno*
723 * is set appropriately.
727 * Attach an eBPF program to a *target_fd* at the specified
728 * *attach_type* hook and return a file descriptor handle for
732 * A new file descriptor (a nonnegative integer), or -1 if an
733 * error occurred (in which case, *errno* is set appropriately).
737 * Update the eBPF program in the specified *link_fd* to
741 * Returns zero on success. On error, -1 is returned and *errno*
742 * is set appropriately.
744 * BPF_LINK_GET_FD_BY_ID
746 * Open a file descriptor for the eBPF Link corresponding to
750 * A new file descriptor (a nonnegative integer), or -1 if an
751 * error occurred (in which case, *errno* is set appropriately).
753 * BPF_LINK_GET_NEXT_ID
755 * Fetch the next eBPF link currently loaded into the kernel.
757 * Looks for the eBPF link with an id greater than *start_id*
758 * and updates *next_id* on success. If no other eBPF links
759 * remain with ids higher than *start_id*, returns -1 and sets
760 * *errno* to **ENOENT**.
763 * Returns zero on success. On error, or when no id remains, -1
764 * is returned and *errno* is set appropriately.
768 * Enable eBPF runtime statistics gathering.
770 * Runtime statistics gathering for the eBPF runtime is disabled
771 * by default to minimize the corresponding performance overhead.
772 * This command enables statistics globally.
774 * Multiple programs may independently enable statistics.
775 * After gathering the desired statistics, eBPF runtime statistics
776 * may be disabled again by calling **close**\ (2) for the file
777 * descriptor returned by this function. Statistics will only be
778 * disabled system-wide when all outstanding file descriptors
779 * returned by prior calls for this subcommand are closed.
782 * A new file descriptor (a nonnegative integer), or -1 if an
783 * error occurred (in which case, *errno* is set appropriately).
787 * Create an iterator on top of the specified *link_fd* (as
788 * previously created using **BPF_LINK_CREATE**) and return a
789 * file descriptor that can be used to trigger the iteration.
791 * If the resulting file descriptor is pinned to the filesystem
792 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
793 * for that path will trigger the iterator to read kernel state
794 * using the eBPF program attached to *link_fd*.
797 * A new file descriptor (a nonnegative integer), or -1 if an
798 * error occurred (in which case, *errno* is set appropriately).
802 * Forcefully detach the specified *link_fd* from its
803 * corresponding attachment point.
806 * Returns zero on success. On error, -1 is returned and *errno*
807 * is set appropriately.
811 * Bind a map to the lifetime of an eBPF program.
813 * The map identified by *map_fd* is bound to the program
814 * identified by *prog_fd* and only released when *prog_fd* is
815 * released. This may be used in cases where metadata should be
816 * associated with a program which otherwise does not contain any
817 * references to the map (for example, embedded in the eBPF
818 * program instructions).
821 * Returns zero on success. On error, -1 is returned and *errno*
822 * is set appropriately.
825 * eBPF objects (maps and programs) can be shared between processes.
827 * * After **fork**\ (2), the child inherits file descriptors
828 * referring to the same eBPF objects.
829 * * File descriptors referring to eBPF objects can be transferred over
830 * **unix**\ (7) domain sockets.
831 * * File descriptors referring to eBPF objects can be duplicated in the
832 * usual way, using **dup**\ (2) and similar calls.
833 * * File descriptors referring to eBPF objects can be pinned to the
834 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
836 * An eBPF object is deallocated only after all file descriptors referring
837 * to the object have been closed and no references remain pinned to the
838 * filesystem or attached (for example, bound to a program or device).
845 BPF_MAP_GET_NEXT_KEY,
852 BPF_PROG_RUN = BPF_PROG_TEST_RUN,
853 BPF_PROG_GET_NEXT_ID,
855 BPF_PROG_GET_FD_BY_ID,
856 BPF_MAP_GET_FD_BY_ID,
857 BPF_OBJ_GET_INFO_BY_FD,
859 BPF_RAW_TRACEPOINT_OPEN,
861 BPF_BTF_GET_FD_BY_ID,
863 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
866 BPF_MAP_LOOKUP_BATCH,
867 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
868 BPF_MAP_UPDATE_BATCH,
869 BPF_MAP_DELETE_BATCH,
872 BPF_LINK_GET_FD_BY_ID,
873 BPF_LINK_GET_NEXT_ID,
884 BPF_MAP_TYPE_PROG_ARRAY,
885 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
886 BPF_MAP_TYPE_PERCPU_HASH,
887 BPF_MAP_TYPE_PERCPU_ARRAY,
888 BPF_MAP_TYPE_STACK_TRACE,
889 BPF_MAP_TYPE_CGROUP_ARRAY,
890 BPF_MAP_TYPE_LRU_HASH,
891 BPF_MAP_TYPE_LRU_PERCPU_HASH,
892 BPF_MAP_TYPE_LPM_TRIE,
893 BPF_MAP_TYPE_ARRAY_OF_MAPS,
894 BPF_MAP_TYPE_HASH_OF_MAPS,
896 BPF_MAP_TYPE_SOCKMAP,
899 BPF_MAP_TYPE_SOCKHASH,
900 BPF_MAP_TYPE_CGROUP_STORAGE,
901 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
902 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
905 BPF_MAP_TYPE_SK_STORAGE,
906 BPF_MAP_TYPE_DEVMAP_HASH,
907 BPF_MAP_TYPE_STRUCT_OPS,
908 BPF_MAP_TYPE_RINGBUF,
909 BPF_MAP_TYPE_INODE_STORAGE,
910 BPF_MAP_TYPE_TASK_STORAGE,
911 BPF_MAP_TYPE_BLOOM_FILTER,
914 /* Note that tracing related programs such as
915 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
916 * are not subject to a stable API since kernel internal data
917 * structures can change from release to release and may
918 * therefore break existing tracing BPF programs. Tracing BPF
919 * programs correspond to /a/ specific kernel which is to be
920 * analyzed, and not /a/ specific kernel /and/ all future ones.
923 BPF_PROG_TYPE_UNSPEC,
924 BPF_PROG_TYPE_SOCKET_FILTER,
925 BPF_PROG_TYPE_KPROBE,
926 BPF_PROG_TYPE_SCHED_CLS,
927 BPF_PROG_TYPE_SCHED_ACT,
928 BPF_PROG_TYPE_TRACEPOINT,
930 BPF_PROG_TYPE_PERF_EVENT,
931 BPF_PROG_TYPE_CGROUP_SKB,
932 BPF_PROG_TYPE_CGROUP_SOCK,
933 BPF_PROG_TYPE_LWT_IN,
934 BPF_PROG_TYPE_LWT_OUT,
935 BPF_PROG_TYPE_LWT_XMIT,
936 BPF_PROG_TYPE_SOCK_OPS,
937 BPF_PROG_TYPE_SK_SKB,
938 BPF_PROG_TYPE_CGROUP_DEVICE,
939 BPF_PROG_TYPE_SK_MSG,
940 BPF_PROG_TYPE_RAW_TRACEPOINT,
941 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
942 BPF_PROG_TYPE_LWT_SEG6LOCAL,
943 BPF_PROG_TYPE_LIRC_MODE2,
944 BPF_PROG_TYPE_SK_REUSEPORT,
945 BPF_PROG_TYPE_FLOW_DISSECTOR,
946 BPF_PROG_TYPE_CGROUP_SYSCTL,
947 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
948 BPF_PROG_TYPE_CGROUP_SOCKOPT,
949 BPF_PROG_TYPE_TRACING,
950 BPF_PROG_TYPE_STRUCT_OPS,
953 BPF_PROG_TYPE_SK_LOOKUP,
954 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
957 enum bpf_attach_type {
958 BPF_CGROUP_INET_INGRESS,
959 BPF_CGROUP_INET_EGRESS,
960 BPF_CGROUP_INET_SOCK_CREATE,
962 BPF_SK_SKB_STREAM_PARSER,
963 BPF_SK_SKB_STREAM_VERDICT,
966 BPF_CGROUP_INET4_BIND,
967 BPF_CGROUP_INET6_BIND,
968 BPF_CGROUP_INET4_CONNECT,
969 BPF_CGROUP_INET6_CONNECT,
970 BPF_CGROUP_INET4_POST_BIND,
971 BPF_CGROUP_INET6_POST_BIND,
972 BPF_CGROUP_UDP4_SENDMSG,
973 BPF_CGROUP_UDP6_SENDMSG,
977 BPF_CGROUP_UDP4_RECVMSG,
978 BPF_CGROUP_UDP6_RECVMSG,
979 BPF_CGROUP_GETSOCKOPT,
980 BPF_CGROUP_SETSOCKOPT,
987 BPF_CGROUP_INET4_GETPEERNAME,
988 BPF_CGROUP_INET6_GETPEERNAME,
989 BPF_CGROUP_INET4_GETSOCKNAME,
990 BPF_CGROUP_INET6_GETSOCKNAME,
992 BPF_CGROUP_INET_SOCK_RELEASE,
997 BPF_SK_REUSEPORT_SELECT,
998 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1000 BPF_TRACE_KPROBE_MULTI,
1002 __MAX_BPF_ATTACH_TYPE
1005 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1007 enum bpf_link_type {
1008 BPF_LINK_TYPE_UNSPEC = 0,
1009 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1010 BPF_LINK_TYPE_TRACING = 2,
1011 BPF_LINK_TYPE_CGROUP = 3,
1012 BPF_LINK_TYPE_ITER = 4,
1013 BPF_LINK_TYPE_NETNS = 5,
1014 BPF_LINK_TYPE_XDP = 6,
1015 BPF_LINK_TYPE_PERF_EVENT = 7,
1016 BPF_LINK_TYPE_KPROBE_MULTI = 8,
1017 BPF_LINK_TYPE_STRUCT_OPS = 9,
1022 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1024 * NONE(default): No further bpf programs allowed in the subtree.
1026 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1027 * the program in this cgroup yields to sub-cgroup program.
1029 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1030 * that cgroup program gets run in addition to the program in this cgroup.
1032 * Only one program is allowed to be attached to a cgroup with
1033 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1034 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1035 * release old program and attach the new one. Attach flags has to match.
1037 * Multiple programs are allowed to be attached to a cgroup with
1038 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1039 * (those that were attached first, run first)
1040 * The programs of sub-cgroup are executed first, then programs of
1041 * this cgroup and then programs of parent cgroup.
1042 * When children program makes decision (like picking TCP CA or sock bind)
1043 * parent program has a chance to override it.
1045 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1046 * programs for a cgroup. Though it's possible to replace an old program at
1047 * any position by also specifying BPF_F_REPLACE flag and position itself in
1048 * replace_bpf_fd attribute. Old program at this position will be released.
1050 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1051 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1053 * cgrp1 (MULTI progs A, B) ->
1054 * cgrp2 (OVERRIDE prog C) ->
1055 * cgrp3 (MULTI prog D) ->
1056 * cgrp4 (OVERRIDE prog E) ->
1057 * cgrp5 (NONE prog F)
1058 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1059 * if prog F is detached, the execution is E,D,A,B
1060 * if prog F and D are detached, the execution is E,A,B
1061 * if prog F, E and D are detached, the execution is C,A,B
1063 * All eligible programs are executed regardless of return code from
1066 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
1067 #define BPF_F_ALLOW_MULTI (1U << 1)
1068 #define BPF_F_REPLACE (1U << 2)
1070 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1071 * verifier will perform strict alignment checking as if the kernel
1072 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1073 * and NET_IP_ALIGN defined to 2.
1075 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
1077 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1078 * verifier will allow any alignment whatsoever. On platforms
1079 * with strict alignment requirements for loads ands stores (such
1080 * as sparc and mips) the verifier validates that all loads and
1081 * stores provably follow this requirement. This flag turns that
1082 * checking and enforcement off.
1084 * It is mostly used for testing when we want to validate the
1085 * context and memory access aspects of the verifier, but because
1086 * of an unaligned access the alignment check would trigger before
1087 * the one we are interested in.
1089 #define BPF_F_ANY_ALIGNMENT (1U << 1)
1091 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1092 * Verifier does sub-register def/use analysis and identifies instructions whose
1093 * def only matters for low 32-bit, high 32-bit is never referenced later
1094 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1095 * that it is safe to ignore clearing high 32-bit for these instructions. This
1096 * saves some back-ends a lot of code-gen. However such optimization is not
1097 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1098 * hence hasn't used verifier's analysis result. But, we really want to have a
1099 * way to be able to verify the correctness of the described optimization on
1100 * x86_64 on which testsuites are frequently exercised.
1102 * So, this flag is introduced. Once it is set, verifier will randomize high
1103 * 32-bit for those instructions who has been identified as safe to ignore them.
1104 * Then, if verifier is not doing correct analysis, such randomization will
1105 * regress tests to expose bugs.
1107 #define BPF_F_TEST_RND_HI32 (1U << 2)
1109 /* The verifier internal test flag. Behavior is undefined */
1110 #define BPF_F_TEST_STATE_FREQ (1U << 3)
1112 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1113 * restrict map and helper usage for such programs. Sleepable BPF programs can
1114 * only be attached to hooks where kernel execution context allows sleeping.
1115 * Such programs are allowed to use helpers that may sleep like
1116 * bpf_copy_from_user().
1118 #define BPF_F_SLEEPABLE (1U << 4)
1120 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1121 * fully support xdp frags.
1123 #define BPF_F_XDP_HAS_FRAGS (1U << 5)
1125 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1126 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1128 #define BPF_F_KPROBE_MULTI_RETURN (1U << 0)
1130 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1131 * the following extensions:
1133 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1134 * insn[0].imm: map fd or fd_idx
1138 * ldimm64 rewrite: address of map
1139 * verifier type: CONST_PTR_TO_MAP
1141 #define BPF_PSEUDO_MAP_FD 1
1142 #define BPF_PSEUDO_MAP_IDX 5
1144 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1145 * insn[0].imm: map fd or fd_idx
1146 * insn[1].imm: offset into value
1149 * ldimm64 rewrite: address of map[0]+offset
1150 * verifier type: PTR_TO_MAP_VALUE
1152 #define BPF_PSEUDO_MAP_VALUE 2
1153 #define BPF_PSEUDO_MAP_IDX_VALUE 6
1155 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1156 * insn[0].imm: kernel btd id of VAR
1160 * ldimm64 rewrite: address of the kernel variable
1161 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1164 #define BPF_PSEUDO_BTF_ID 3
1165 /* insn[0].src_reg: BPF_PSEUDO_FUNC
1166 * insn[0].imm: insn offset to the func
1170 * ldimm64 rewrite: address of the function
1171 * verifier type: PTR_TO_FUNC.
1173 #define BPF_PSEUDO_FUNC 4
1175 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1176 * offset to another bpf function
1178 #define BPF_PSEUDO_CALL 1
1179 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1180 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1182 #define BPF_PSEUDO_KFUNC_CALL 2
1184 /* flags for BPF_MAP_UPDATE_ELEM command */
1186 BPF_ANY = 0, /* create new element or update existing */
1187 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1188 BPF_EXIST = 2, /* update existing element */
1189 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1192 /* flags for BPF_MAP_CREATE command */
1194 BPF_F_NO_PREALLOC = (1U << 0),
1195 /* Instead of having one common LRU list in the
1196 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1197 * which can scale and perform better.
1198 * Note, the LRU nodes (including free nodes) cannot be moved
1199 * across different LRU lists.
1201 BPF_F_NO_COMMON_LRU = (1U << 1),
1202 /* Specify numa node during map creation */
1203 BPF_F_NUMA_NODE = (1U << 2),
1205 /* Flags for accessing BPF object from syscall side. */
1206 BPF_F_RDONLY = (1U << 3),
1207 BPF_F_WRONLY = (1U << 4),
1209 /* Flag for stack_map, store build_id+offset instead of pointer */
1210 BPF_F_STACK_BUILD_ID = (1U << 5),
1212 /* Zero-initialize hash function seed. This should only be used for testing. */
1213 BPF_F_ZERO_SEED = (1U << 6),
1215 /* Flags for accessing BPF object from program side. */
1216 BPF_F_RDONLY_PROG = (1U << 7),
1217 BPF_F_WRONLY_PROG = (1U << 8),
1219 /* Clone map from listener for newly accepted socket */
1220 BPF_F_CLONE = (1U << 9),
1222 /* Enable memory-mapping BPF map */
1223 BPF_F_MMAPABLE = (1U << 10),
1225 /* Share perf_event among processes */
1226 BPF_F_PRESERVE_ELEMS = (1U << 11),
1228 /* Create a map that is suitable to be an inner map with dynamic max entries */
1229 BPF_F_INNER_MAP = (1U << 12),
1232 /* Flags for BPF_PROG_QUERY. */
1234 /* Query effective (directly attached + inherited from ancestor cgroups)
1235 * programs that will be executed for events within a cgroup.
1236 * attach_flags with this flag are returned only for directly attached programs.
1238 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
1240 /* Flags for BPF_PROG_TEST_RUN */
1242 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1243 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1244 /* If set, XDP frames will be transmitted after processing */
1245 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
1247 /* type for BPF_ENABLE_STATS */
1248 enum bpf_stats_type {
1249 /* enabled run_time_ns and run_cnt */
1250 BPF_STATS_RUN_TIME = 0,
1253 enum bpf_stack_build_id_status {
1254 /* user space need an empty entry to identify end of a trace */
1255 BPF_STACK_BUILD_ID_EMPTY = 0,
1256 /* with valid build_id and offset */
1257 BPF_STACK_BUILD_ID_VALID = 1,
1258 /* couldn't get build_id, fallback to ip */
1259 BPF_STACK_BUILD_ID_IP = 2,
1262 #define BPF_BUILD_ID_SIZE 20
1263 struct bpf_stack_build_id {
1265 unsigned char build_id[BPF_BUILD_ID_SIZE];
1272 #define BPF_OBJ_NAME_LEN 16U
1275 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1276 __u32 map_type; /* one of enum bpf_map_type */
1277 __u32 key_size; /* size of key in bytes */
1278 __u32 value_size; /* size of value in bytes */
1279 __u32 max_entries; /* max number of entries in a map */
1280 __u32 map_flags; /* BPF_MAP_CREATE related
1281 * flags defined above.
1283 __u32 inner_map_fd; /* fd pointing to the inner map */
1284 __u32 numa_node; /* numa node (effective only if
1285 * BPF_F_NUMA_NODE is set).
1287 char map_name[BPF_OBJ_NAME_LEN];
1288 __u32 map_ifindex; /* ifindex of netdev to create on */
1289 __u32 btf_fd; /* fd pointing to a BTF type data */
1290 __u32 btf_key_type_id; /* BTF type_id of the key */
1291 __u32 btf_value_type_id; /* BTF type_id of the value */
1292 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1293 * struct stored as the
1296 /* Any per-map-type extra fields
1298 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1299 * number of hash functions (if 0, the bloom filter will default
1300 * to using 5 hash functions).
1305 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1309 __aligned_u64 value;
1310 __aligned_u64 next_key;
1315 struct { /* struct used by BPF_MAP_*_BATCH commands */
1316 __aligned_u64 in_batch; /* start batch,
1317 * NULL to start from beginning
1319 __aligned_u64 out_batch; /* output: next start batch */
1321 __aligned_u64 values;
1322 __u32 count; /* input/output:
1323 * input: # of key/value
1325 * output: # of filled elements
1332 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1333 __u32 prog_type; /* one of enum bpf_prog_type */
1335 __aligned_u64 insns;
1336 __aligned_u64 license;
1337 __u32 log_level; /* verbosity level of verifier */
1338 __u32 log_size; /* size of user buffer */
1339 __aligned_u64 log_buf; /* user supplied buffer */
1340 __u32 kern_version; /* not used */
1342 char prog_name[BPF_OBJ_NAME_LEN];
1343 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1344 /* For some prog types expected attach type must be known at
1345 * load time to verify attach type specific parts of prog
1346 * (context accesses, allowed helpers, etc).
1348 __u32 expected_attach_type;
1349 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1350 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1351 __aligned_u64 func_info; /* func info */
1352 __u32 func_info_cnt; /* number of bpf_func_info records */
1353 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1354 __aligned_u64 line_info; /* line info */
1355 __u32 line_info_cnt; /* number of bpf_line_info records */
1356 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1358 /* valid prog_fd to attach to bpf prog */
1359 __u32 attach_prog_fd;
1360 /* or valid module BTF object fd or 0 to attach to vmlinux */
1361 __u32 attach_btf_obj_fd;
1363 __u32 core_relo_cnt; /* number of bpf_core_relo */
1364 __aligned_u64 fd_array; /* array of FDs */
1365 __aligned_u64 core_relos;
1366 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1369 struct { /* anonymous struct used by BPF_OBJ_* commands */
1370 __aligned_u64 pathname;
1375 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1376 __u32 target_fd; /* container object to attach to */
1377 __u32 attach_bpf_fd; /* eBPF program to attach */
1380 __u32 replace_bpf_fd; /* previously attached eBPF
1381 * program to replace if
1382 * BPF_F_REPLACE is used
1386 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1389 __u32 data_size_in; /* input: len of data_in */
1390 __u32 data_size_out; /* input/output: len of data_out
1391 * returns ENOSPC if data_out
1394 __aligned_u64 data_in;
1395 __aligned_u64 data_out;
1398 __u32 ctx_size_in; /* input: len of ctx_in */
1399 __u32 ctx_size_out; /* input/output: len of ctx_out
1400 * returns ENOSPC if ctx_out
1403 __aligned_u64 ctx_in;
1404 __aligned_u64 ctx_out;
1410 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1422 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1428 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1429 __u32 target_fd; /* container object to query */
1433 __aligned_u64 prog_ids;
1435 __aligned_u64 prog_attach_flags; /* output: per-program attach_flags */
1438 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1443 struct { /* anonymous struct for BPF_BTF_LOAD */
1445 __aligned_u64 btf_log_buf;
1448 __u32 btf_log_level;
1452 __u32 pid; /* input: pid */
1453 __u32 fd; /* input: fd */
1454 __u32 flags; /* input: flags */
1455 __u32 buf_len; /* input/output: buf len */
1456 __aligned_u64 buf; /* input/output:
1457 * tp_name for tracepoint
1459 * filename for uprobe
1461 __u32 prog_id; /* output: prod_id */
1462 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1463 __u64 probe_offset; /* output: probe_offset */
1464 __u64 probe_addr; /* output: probe_addr */
1467 struct { /* struct used by BPF_LINK_CREATE command */
1468 __u32 prog_fd; /* eBPF program to attach */
1470 __u32 target_fd; /* object to attach to */
1471 __u32 target_ifindex; /* target ifindex */
1473 __u32 attach_type; /* attach type */
1474 __u32 flags; /* extra flags */
1476 __u32 target_btf_id; /* btf_id of target to attach to */
1478 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1479 __u32 iter_info_len; /* iter_info length */
1482 /* black box user-provided value passed through
1483 * to BPF program at the execution time and
1484 * accessible through bpf_get_attach_cookie() BPF helper
1492 __aligned_u64 addrs;
1493 __aligned_u64 cookies;
1496 /* this is overlaid with the target_btf_id above. */
1497 __u32 target_btf_id;
1498 /* black box user-provided value passed through
1499 * to BPF program at the execution time and
1500 * accessible through bpf_get_attach_cookie() BPF helper
1507 struct { /* struct used by BPF_LINK_UPDATE command */
1508 __u32 link_fd; /* link fd */
1509 /* new program fd to update link with */
1511 __u32 flags; /* extra flags */
1512 /* expected link's program fd; is specified only if
1513 * BPF_F_REPLACE flag is set in flags */
1521 struct { /* struct used by BPF_ENABLE_STATS command */
1525 struct { /* struct used by BPF_ITER_CREATE command */
1530 struct { /* struct used by BPF_PROG_BIND_MAP command */
1533 __u32 flags; /* extra flags */
1536 } __attribute__((aligned(8)));
1538 /* The description below is an attempt at providing documentation to eBPF
1539 * developers about the multiple available eBPF helper functions. It can be
1540 * parsed and used to produce a manual page. The workflow is the following,
1541 * and requires the rst2man utility:
1543 * $ ./scripts/bpf_doc.py \
1544 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1545 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1546 * $ man /tmp/bpf-helpers.7
1548 * Note that in order to produce this external documentation, some RST
1549 * formatting is used in the descriptions to get "bold" and "italics" in
1550 * manual pages. Also note that the few trailing white spaces are
1551 * intentional, removing them would break paragraphs for rst2man.
1553 * Start of BPF helper function descriptions:
1555 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1557 * Perform a lookup in *map* for an entry associated to *key*.
1559 * Map value associated to *key*, or **NULL** if no entry was
1562 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1564 * Add or update the value of the entry associated to *key* in
1565 * *map* with *value*. *flags* is one of:
1568 * The entry for *key* must not exist in the map.
1570 * The entry for *key* must already exist in the map.
1572 * No condition on the existence of the entry for *key*.
1574 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1575 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1576 * elements always exist), the helper would return an error.
1578 * 0 on success, or a negative error in case of failure.
1580 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1582 * Delete entry with *key* from *map*.
1584 * 0 on success, or a negative error in case of failure.
1586 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1588 * For tracing programs, safely attempt to read *size* bytes from
1589 * kernel space address *unsafe_ptr* and store the data in *dst*.
1591 * Generally, use **bpf_probe_read_user**\ () or
1592 * **bpf_probe_read_kernel**\ () instead.
1594 * 0 on success, or a negative error in case of failure.
1596 * u64 bpf_ktime_get_ns(void)
1598 * Return the time elapsed since system boot, in nanoseconds.
1599 * Does not include time the system was suspended.
1600 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1604 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1606 * This helper is a "printk()-like" facility for debugging. It
1607 * prints a message defined by format *fmt* (of size *fmt_size*)
1608 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1609 * available. It can take up to three additional **u64**
1610 * arguments (as an eBPF helpers, the total number of arguments is
1613 * Each time the helper is called, it appends a line to the trace.
1614 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1615 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1616 * The format of the trace is customizable, and the exact output
1617 * one will get depends on the options set in
1618 * *\/sys/kernel/debug/tracing/trace_options* (see also the
1619 * *README* file under the same directory). However, it usually
1620 * defaults to something like:
1624 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1628 * * ``telnet`` is the name of the current task.
1629 * * ``470`` is the PID of the current task.
1630 * * ``001`` is the CPU number on which the task is
1632 * * In ``.N..``, each character refers to a set of
1633 * options (whether irqs are enabled, scheduling
1634 * options, whether hard/softirqs are running, level of
1635 * preempt_disabled respectively). **N** means that
1636 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1638 * * ``419421.045894`` is a timestamp.
1639 * * ``0x00000001`` is a fake value used by BPF for the
1640 * instruction pointer register.
1641 * * ``<formatted msg>`` is the message formatted with
1644 * The conversion specifiers supported by *fmt* are similar, but
1645 * more limited than for printk(). They are **%d**, **%i**,
1646 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1647 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1648 * of field, padding with zeroes, etc.) is available, and the
1649 * helper will return **-EINVAL** (but print nothing) if it
1650 * encounters an unknown specifier.
1652 * Also, note that **bpf_trace_printk**\ () is slow, and should
1653 * only be used for debugging purposes. For this reason, a notice
1654 * block (spanning several lines) is printed to kernel logs and
1655 * states that the helper should not be used "for production use"
1656 * the first time this helper is used (or more precisely, when
1657 * **trace_printk**\ () buffers are allocated). For passing values
1658 * to user space, perf events should be preferred.
1660 * The number of bytes written to the buffer, or a negative error
1661 * in case of failure.
1663 * u32 bpf_get_prandom_u32(void)
1665 * Get a pseudo-random number.
1667 * From a security point of view, this helper uses its own
1668 * pseudo-random internal state, and cannot be used to infer the
1669 * seed of other random functions in the kernel. However, it is
1670 * essential to note that the generator used by the helper is not
1671 * cryptographically secure.
1673 * A random 32-bit unsigned value.
1675 * u32 bpf_get_smp_processor_id(void)
1677 * Get the SMP (symmetric multiprocessing) processor id. Note that
1678 * all programs run with migration disabled, which means that the
1679 * SMP processor id is stable during all the execution of the
1682 * The SMP id of the processor running the program.
1684 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1686 * Store *len* bytes from address *from* into the packet
1687 * associated to *skb*, at *offset*. *flags* are a combination of
1688 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1689 * checksum for the packet after storing the bytes) and
1690 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1691 * **->swhash** and *skb*\ **->l4hash** to 0).
1693 * A call to this helper is susceptible to change the underlying
1694 * packet buffer. Therefore, at load time, all checks on pointers
1695 * previously done by the verifier are invalidated and must be
1696 * performed again, if the helper is used in combination with
1697 * direct packet access.
1699 * 0 on success, or a negative error in case of failure.
1701 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1703 * Recompute the layer 3 (e.g. IP) checksum for the packet
1704 * associated to *skb*. Computation is incremental, so the helper
1705 * must know the former value of the header field that was
1706 * modified (*from*), the new value of this field (*to*), and the
1707 * number of bytes (2 or 4) for this field, stored in *size*.
1708 * Alternatively, it is possible to store the difference between
1709 * the previous and the new values of the header field in *to*, by
1710 * setting *from* and *size* to 0. For both methods, *offset*
1711 * indicates the location of the IP checksum within the packet.
1713 * This helper works in combination with **bpf_csum_diff**\ (),
1714 * which does not update the checksum in-place, but offers more
1715 * flexibility and can handle sizes larger than 2 or 4 for the
1716 * checksum to update.
1718 * A call to this helper is susceptible to change the underlying
1719 * packet buffer. Therefore, at load time, all checks on pointers
1720 * previously done by the verifier are invalidated and must be
1721 * performed again, if the helper is used in combination with
1722 * direct packet access.
1724 * 0 on success, or a negative error in case of failure.
1726 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1728 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1729 * packet associated to *skb*. Computation is incremental, so the
1730 * helper must know the former value of the header field that was
1731 * modified (*from*), the new value of this field (*to*), and the
1732 * number of bytes (2 or 4) for this field, stored on the lowest
1733 * four bits of *flags*. Alternatively, it is possible to store
1734 * the difference between the previous and the new values of the
1735 * header field in *to*, by setting *from* and the four lowest
1736 * bits of *flags* to 0. For both methods, *offset* indicates the
1737 * location of the IP checksum within the packet. In addition to
1738 * the size of the field, *flags* can be added (bitwise OR) actual
1739 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1740 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1741 * for updates resulting in a null checksum the value is set to
1742 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1743 * the checksum is to be computed against a pseudo-header.
1745 * This helper works in combination with **bpf_csum_diff**\ (),
1746 * which does not update the checksum in-place, but offers more
1747 * flexibility and can handle sizes larger than 2 or 4 for the
1748 * checksum to update.
1750 * A call to this helper is susceptible to change the underlying
1751 * packet buffer. Therefore, at load time, all checks on pointers
1752 * previously done by the verifier are invalidated and must be
1753 * performed again, if the helper is used in combination with
1754 * direct packet access.
1756 * 0 on success, or a negative error in case of failure.
1758 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1760 * This special helper is used to trigger a "tail call", or in
1761 * other words, to jump into another eBPF program. The same stack
1762 * frame is used (but values on stack and in registers for the
1763 * caller are not accessible to the callee). This mechanism allows
1764 * for program chaining, either for raising the maximum number of
1765 * available eBPF instructions, or to execute given programs in
1766 * conditional blocks. For security reasons, there is an upper
1767 * limit to the number of successive tail calls that can be
1770 * Upon call of this helper, the program attempts to jump into a
1771 * program referenced at index *index* in *prog_array_map*, a
1772 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1773 * *ctx*, a pointer to the context.
1775 * If the call succeeds, the kernel immediately runs the first
1776 * instruction of the new program. This is not a function call,
1777 * and it never returns to the previous program. If the call
1778 * fails, then the helper has no effect, and the caller continues
1779 * to run its subsequent instructions. A call can fail if the
1780 * destination program for the jump does not exist (i.e. *index*
1781 * is superior to the number of entries in *prog_array_map*), or
1782 * if the maximum number of tail calls has been reached for this
1783 * chain of programs. This limit is defined in the kernel by the
1784 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1785 * which is currently set to 33.
1787 * 0 on success, or a negative error in case of failure.
1789 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1791 * Clone and redirect the packet associated to *skb* to another
1792 * net device of index *ifindex*. Both ingress and egress
1793 * interfaces can be used for redirection. The **BPF_F_INGRESS**
1794 * value in *flags* is used to make the distinction (ingress path
1795 * is selected if the flag is present, egress path otherwise).
1796 * This is the only flag supported for now.
1798 * In comparison with **bpf_redirect**\ () helper,
1799 * **bpf_clone_redirect**\ () has the associated cost of
1800 * duplicating the packet buffer, but this can be executed out of
1801 * the eBPF program. Conversely, **bpf_redirect**\ () is more
1802 * efficient, but it is handled through an action code where the
1803 * redirection happens only after the eBPF program has returned.
1805 * A call to this helper is susceptible to change the underlying
1806 * packet buffer. Therefore, at load time, all checks on pointers
1807 * previously done by the verifier are invalidated and must be
1808 * performed again, if the helper is used in combination with
1809 * direct packet access.
1811 * 0 on success, or a negative error in case of failure.
1813 * u64 bpf_get_current_pid_tgid(void)
1815 * Get the current pid and tgid.
1817 * A 64-bit integer containing the current tgid and pid, and
1819 * *current_task*\ **->tgid << 32 \|**
1820 * *current_task*\ **->pid**.
1822 * u64 bpf_get_current_uid_gid(void)
1824 * Get the current uid and gid.
1826 * A 64-bit integer containing the current GID and UID, and
1827 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1829 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1831 * Copy the **comm** attribute of the current task into *buf* of
1832 * *size_of_buf*. The **comm** attribute contains the name of
1833 * the executable (excluding the path) for the current task. The
1834 * *size_of_buf* must be strictly positive. On success, the
1835 * helper makes sure that the *buf* is NUL-terminated. On failure,
1836 * it is filled with zeroes.
1838 * 0 on success, or a negative error in case of failure.
1840 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1842 * Retrieve the classid for the current task, i.e. for the net_cls
1843 * cgroup to which *skb* belongs.
1845 * This helper can be used on TC egress path, but not on ingress.
1847 * The net_cls cgroup provides an interface to tag network packets
1848 * based on a user-provided identifier for all traffic coming from
1849 * the tasks belonging to the related cgroup. See also the related
1850 * kernel documentation, available from the Linux sources in file
1851 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1853 * The Linux kernel has two versions for cgroups: there are
1854 * cgroups v1 and cgroups v2. Both are available to users, who can
1855 * use a mixture of them, but note that the net_cls cgroup is for
1856 * cgroup v1 only. This makes it incompatible with BPF programs
1857 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1858 * only hold data for one version of cgroups at a time).
1860 * This helper is only available is the kernel was compiled with
1861 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1862 * "**y**" or to "**m**".
1864 * The classid, or 0 for the default unconfigured classid.
1866 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1868 * Push a *vlan_tci* (VLAN tag control information) of protocol
1869 * *vlan_proto* to the packet associated to *skb*, then update
1870 * the checksum. Note that if *vlan_proto* is different from
1871 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1872 * be **ETH_P_8021Q**.
1874 * A call to this helper is susceptible to change the underlying
1875 * packet buffer. Therefore, at load time, all checks on pointers
1876 * previously done by the verifier are invalidated and must be
1877 * performed again, if the helper is used in combination with
1878 * direct packet access.
1880 * 0 on success, or a negative error in case of failure.
1882 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1884 * Pop a VLAN header from the packet associated to *skb*.
1886 * A call to this helper is susceptible to change the underlying
1887 * packet buffer. Therefore, at load time, all checks on pointers
1888 * previously done by the verifier are invalidated and must be
1889 * performed again, if the helper is used in combination with
1890 * direct packet access.
1892 * 0 on success, or a negative error in case of failure.
1894 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1896 * Get tunnel metadata. This helper takes a pointer *key* to an
1897 * empty **struct bpf_tunnel_key** of **size**, that will be
1898 * filled with tunnel metadata for the packet associated to *skb*.
1899 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1900 * indicates that the tunnel is based on IPv6 protocol instead of
1903 * The **struct bpf_tunnel_key** is an object that generalizes the
1904 * principal parameters used by various tunneling protocols into a
1905 * single struct. This way, it can be used to easily make a
1906 * decision based on the contents of the encapsulation header,
1907 * "summarized" in this struct. In particular, it holds the IP
1908 * address of the remote end (IPv4 or IPv6, depending on the case)
1909 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1910 * this struct exposes the *key*\ **->tunnel_id**, which is
1911 * generally mapped to a VNI (Virtual Network Identifier), making
1912 * it programmable together with the **bpf_skb_set_tunnel_key**\
1915 * Let's imagine that the following code is part of a program
1916 * attached to the TC ingress interface, on one end of a GRE
1917 * tunnel, and is supposed to filter out all messages coming from
1918 * remote ends with IPv4 address other than 10.0.0.1:
1923 * struct bpf_tunnel_key key = {};
1925 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1927 * return TC_ACT_SHOT; // drop packet
1929 * if (key.remote_ipv4 != 0x0a000001)
1930 * return TC_ACT_SHOT; // drop packet
1932 * return TC_ACT_OK; // accept packet
1934 * This interface can also be used with all encapsulation devices
1935 * that can operate in "collect metadata" mode: instead of having
1936 * one network device per specific configuration, the "collect
1937 * metadata" mode only requires a single device where the
1938 * configuration can be extracted from this helper.
1940 * This can be used together with various tunnels such as VXLan,
1941 * Geneve, GRE or IP in IP (IPIP).
1943 * 0 on success, or a negative error in case of failure.
1945 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1947 * Populate tunnel metadata for packet associated to *skb.* The
1948 * tunnel metadata is set to the contents of *key*, of *size*. The
1949 * *flags* can be set to a combination of the following values:
1951 * **BPF_F_TUNINFO_IPV6**
1952 * Indicate that the tunnel is based on IPv6 protocol
1954 * **BPF_F_ZERO_CSUM_TX**
1955 * For IPv4 packets, add a flag to tunnel metadata
1956 * indicating that checksum computation should be skipped
1957 * and checksum set to zeroes.
1958 * **BPF_F_DONT_FRAGMENT**
1959 * Add a flag to tunnel metadata indicating that the
1960 * packet should not be fragmented.
1961 * **BPF_F_SEQ_NUMBER**
1962 * Add a flag to tunnel metadata indicating that a
1963 * sequence number should be added to tunnel header before
1964 * sending the packet. This flag was added for GRE
1965 * encapsulation, but might be used with other protocols
1966 * as well in the future.
1968 * Here is a typical usage on the transmit path:
1972 * struct bpf_tunnel_key key;
1974 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1975 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1977 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1978 * helper for additional information.
1980 * 0 on success, or a negative error in case of failure.
1982 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1984 * Read the value of a perf event counter. This helper relies on a
1985 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1986 * the perf event counter is selected when *map* is updated with
1987 * perf event file descriptors. The *map* is an array whose size
1988 * is the number of available CPUs, and each cell contains a value
1989 * relative to one CPU. The value to retrieve is indicated by
1990 * *flags*, that contains the index of the CPU to look up, masked
1991 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1992 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1993 * current CPU should be retrieved.
1995 * Note that before Linux 4.13, only hardware perf event can be
1998 * Also, be aware that the newer helper
1999 * **bpf_perf_event_read_value**\ () is recommended over
2000 * **bpf_perf_event_read**\ () in general. The latter has some ABI
2001 * quirks where error and counter value are used as a return code
2002 * (which is wrong to do since ranges may overlap). This issue is
2003 * fixed with **bpf_perf_event_read_value**\ (), which at the same
2004 * time provides more features over the **bpf_perf_event_read**\
2005 * () interface. Please refer to the description of
2006 * **bpf_perf_event_read_value**\ () for details.
2008 * The value of the perf event counter read from the map, or a
2009 * negative error code in case of failure.
2011 * long bpf_redirect(u32 ifindex, u64 flags)
2013 * Redirect the packet to another net device of index *ifindex*.
2014 * This helper is somewhat similar to **bpf_clone_redirect**\
2015 * (), except that the packet is not cloned, which provides
2016 * increased performance.
2018 * Except for XDP, both ingress and egress interfaces can be used
2019 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
2020 * to make the distinction (ingress path is selected if the flag
2021 * is present, egress path otherwise). Currently, XDP only
2022 * supports redirection to the egress interface, and accepts no
2025 * The same effect can also be attained with the more generic
2026 * **bpf_redirect_map**\ (), which uses a BPF map to store the
2027 * redirect target instead of providing it directly to the helper.
2029 * For XDP, the helper returns **XDP_REDIRECT** on success or
2030 * **XDP_ABORTED** on error. For other program types, the values
2031 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2034 * u32 bpf_get_route_realm(struct sk_buff *skb)
2036 * Retrieve the realm or the route, that is to say the
2037 * **tclassid** field of the destination for the *skb*. The
2038 * identifier retrieved is a user-provided tag, similar to the
2039 * one used with the net_cls cgroup (see description for
2040 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
2041 * held by a route (a destination entry), not by a task.
2043 * Retrieving this identifier works with the clsact TC egress hook
2044 * (see also **tc-bpf(8)**), or alternatively on conventional
2045 * classful egress qdiscs, but not on TC ingress path. In case of
2046 * clsact TC egress hook, this has the advantage that, internally,
2047 * the destination entry has not been dropped yet in the transmit
2048 * path. Therefore, the destination entry does not need to be
2049 * artificially held via **netif_keep_dst**\ () for a classful
2050 * qdisc until the *skb* is freed.
2052 * This helper is available only if the kernel was compiled with
2053 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2055 * The realm of the route for the packet associated to *skb*, or 0
2056 * if none was found.
2058 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2060 * Write raw *data* blob into a special BPF perf event held by
2061 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2062 * event must have the following attributes: **PERF_SAMPLE_RAW**
2063 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2064 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2066 * The *flags* are used to indicate the index in *map* for which
2067 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2068 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2069 * to indicate that the index of the current CPU core should be
2072 * The value to write, of *size*, is passed through eBPF stack and
2073 * pointed by *data*.
2075 * The context of the program *ctx* needs also be passed to the
2078 * On user space, a program willing to read the values needs to
2079 * call **perf_event_open**\ () on the perf event (either for
2080 * one or for all CPUs) and to store the file descriptor into the
2081 * *map*. This must be done before the eBPF program can send data
2082 * into it. An example is available in file
2083 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2084 * tree (the eBPF program counterpart is in
2085 * *samples/bpf/trace_output_kern.c*).
2087 * **bpf_perf_event_output**\ () achieves better performance
2088 * than **bpf_trace_printk**\ () for sharing data with user
2089 * space, and is much better suitable for streaming data from eBPF
2092 * Note that this helper is not restricted to tracing use cases
2093 * and can be used with programs attached to TC or XDP as well,
2094 * where it allows for passing data to user space listeners. Data
2097 * * Only custom structs,
2098 * * Only the packet payload, or
2099 * * A combination of both.
2101 * 0 on success, or a negative error in case of failure.
2103 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2105 * This helper was provided as an easy way to load data from a
2106 * packet. It can be used to load *len* bytes from *offset* from
2107 * the packet associated to *skb*, into the buffer pointed by
2110 * Since Linux 4.7, usage of this helper has mostly been replaced
2111 * by "direct packet access", enabling packet data to be
2112 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2113 * pointing respectively to the first byte of packet data and to
2114 * the byte after the last byte of packet data. However, it
2115 * remains useful if one wishes to read large quantities of data
2116 * at once from a packet into the eBPF stack.
2118 * 0 on success, or a negative error in case of failure.
2120 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2122 * Walk a user or a kernel stack and return its id. To achieve
2123 * this, the helper needs *ctx*, which is a pointer to the context
2124 * on which the tracing program is executed, and a pointer to a
2125 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2127 * The last argument, *flags*, holds the number of stack frames to
2128 * skip (from 0 to 255), masked with
2129 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2130 * a combination of the following flags:
2132 * **BPF_F_USER_STACK**
2133 * Collect a user space stack instead of a kernel stack.
2134 * **BPF_F_FAST_STACK_CMP**
2135 * Compare stacks by hash only.
2136 * **BPF_F_REUSE_STACKID**
2137 * If two different stacks hash into the same *stackid*,
2138 * discard the old one.
2140 * The stack id retrieved is a 32 bit long integer handle which
2141 * can be further combined with other data (including other stack
2142 * ids) and used as a key into maps. This can be useful for
2143 * generating a variety of graphs (such as flame graphs or off-cpu
2146 * For walking a stack, this helper is an improvement over
2147 * **bpf_probe_read**\ (), which can be used with unrolled loops
2148 * but is not efficient and consumes a lot of eBPF instructions.
2149 * Instead, **bpf_get_stackid**\ () can collect up to
2150 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2151 * this limit can be controlled with the **sysctl** program, and
2152 * that it should be manually increased in order to profile long
2153 * user stacks (such as stacks for Java programs). To do so, use:
2157 * # sysctl kernel.perf_event_max_stack=<new value>
2159 * The positive or null stack id on success, or a negative error
2160 * in case of failure.
2162 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2164 * Compute a checksum difference, from the raw buffer pointed by
2165 * *from*, of length *from_size* (that must be a multiple of 4),
2166 * towards the raw buffer pointed by *to*, of size *to_size*
2167 * (same remark). An optional *seed* can be added to the value
2168 * (this can be cascaded, the seed may come from a previous call
2171 * This is flexible enough to be used in several ways:
2173 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2174 * checksum, it can be used when pushing new data.
2175 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2176 * checksum, it can be used when removing data from a packet.
2177 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2178 * can be used to compute a diff. Note that *from_size* and
2179 * *to_size* do not need to be equal.
2181 * This helper can be used in combination with
2182 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2183 * which one can feed in the difference computed with
2184 * **bpf_csum_diff**\ ().
2186 * The checksum result, or a negative error code in case of
2189 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2191 * Retrieve tunnel options metadata for the packet associated to
2192 * *skb*, and store the raw tunnel option data to the buffer *opt*
2195 * This helper can be used with encapsulation devices that can
2196 * operate in "collect metadata" mode (please refer to the related
2197 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2198 * more details). A particular example where this can be used is
2199 * in combination with the Geneve encapsulation protocol, where it
2200 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2201 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2202 * the eBPF program. This allows for full customization of these
2205 * The size of the option data retrieved.
2207 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2209 * Set tunnel options metadata for the packet associated to *skb*
2210 * to the option data contained in the raw buffer *opt* of *size*.
2212 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2213 * helper for additional information.
2215 * 0 on success, or a negative error in case of failure.
2217 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2219 * Change the protocol of the *skb* to *proto*. Currently
2220 * supported are transition from IPv4 to IPv6, and from IPv6 to
2221 * IPv4. The helper takes care of the groundwork for the
2222 * transition, including resizing the socket buffer. The eBPF
2223 * program is expected to fill the new headers, if any, via
2224 * **skb_store_bytes**\ () and to recompute the checksums with
2225 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2226 * (). The main case for this helper is to perform NAT64
2227 * operations out of an eBPF program.
2229 * Internally, the GSO type is marked as dodgy so that headers are
2230 * checked and segments are recalculated by the GSO/GRO engine.
2231 * The size for GSO target is adapted as well.
2233 * All values for *flags* are reserved for future usage, and must
2236 * A call to this helper is susceptible to change the underlying
2237 * packet buffer. Therefore, at load time, all checks on pointers
2238 * previously done by the verifier are invalidated and must be
2239 * performed again, if the helper is used in combination with
2240 * direct packet access.
2242 * 0 on success, or a negative error in case of failure.
2244 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2246 * Change the packet type for the packet associated to *skb*. This
2247 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2248 * the eBPF program does not have a write access to *skb*\
2249 * **->pkt_type** beside this helper. Using a helper here allows
2250 * for graceful handling of errors.
2252 * The major use case is to change incoming *skb*s to
2253 * **PACKET_HOST** in a programmatic way instead of having to
2254 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2257 * Note that *type* only allows certain values. At this time, they
2262 * **PACKET_BROADCAST**
2263 * Send packet to all.
2264 * **PACKET_MULTICAST**
2265 * Send packet to group.
2266 * **PACKET_OTHERHOST**
2267 * Send packet to someone else.
2269 * 0 on success, or a negative error in case of failure.
2271 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2273 * Check whether *skb* is a descendant of the cgroup2 held by
2274 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2276 * The return value depends on the result of the test, and can be:
2278 * * 0, if the *skb* failed the cgroup2 descendant test.
2279 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2280 * * A negative error code, if an error occurred.
2282 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2284 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2285 * not set, in particular if the hash was cleared due to mangling,
2286 * recompute this hash. Later accesses to the hash can be done
2287 * directly with *skb*\ **->hash**.
2289 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2290 * prototype with **bpf_skb_change_proto**\ (), or calling
2291 * **bpf_skb_store_bytes**\ () with the
2292 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2293 * the hash and to trigger a new computation for the next call to
2294 * **bpf_get_hash_recalc**\ ().
2298 * u64 bpf_get_current_task(void)
2300 * Get the current task.
2302 * A pointer to the current task struct.
2304 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2306 * Attempt in a safe way to write *len* bytes from the buffer
2307 * *src* to *dst* in memory. It only works for threads that are in
2308 * user context, and *dst* must be a valid user space address.
2310 * This helper should not be used to implement any kind of
2311 * security mechanism because of TOC-TOU attacks, but rather to
2312 * debug, divert, and manipulate execution of semi-cooperative
2315 * Keep in mind that this feature is meant for experiments, and it
2316 * has a risk of crashing the system and running programs.
2317 * Therefore, when an eBPF program using this helper is attached,
2318 * a warning including PID and process name is printed to kernel
2321 * 0 on success, or a negative error in case of failure.
2323 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2325 * Check whether the probe is being run is the context of a given
2326 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2327 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2329 * The return value depends on the result of the test, and can be:
2331 * * 1, if current task belongs to the cgroup2.
2332 * * 0, if current task does not belong to the cgroup2.
2333 * * A negative error code, if an error occurred.
2335 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2337 * Resize (trim or grow) the packet associated to *skb* to the
2338 * new *len*. The *flags* are reserved for future usage, and must
2341 * The basic idea is that the helper performs the needed work to
2342 * change the size of the packet, then the eBPF program rewrites
2343 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2344 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2345 * and others. This helper is a slow path utility intended for
2346 * replies with control messages. And because it is targeted for
2347 * slow path, the helper itself can afford to be slow: it
2348 * implicitly linearizes, unclones and drops offloads from the
2351 * A call to this helper is susceptible to change the underlying
2352 * packet buffer. Therefore, at load time, all checks on pointers
2353 * previously done by the verifier are invalidated and must be
2354 * performed again, if the helper is used in combination with
2355 * direct packet access.
2357 * 0 on success, or a negative error in case of failure.
2359 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2361 * Pull in non-linear data in case the *skb* is non-linear and not
2362 * all of *len* are part of the linear section. Make *len* bytes
2363 * from *skb* readable and writable. If a zero value is passed for
2364 * *len*, then all bytes in the linear part of *skb* will be made
2365 * readable and writable.
2367 * This helper is only needed for reading and writing with direct
2370 * For direct packet access, testing that offsets to access
2371 * are within packet boundaries (test on *skb*\ **->data_end**) is
2372 * susceptible to fail if offsets are invalid, or if the requested
2373 * data is in non-linear parts of the *skb*. On failure the
2374 * program can just bail out, or in the case of a non-linear
2375 * buffer, use a helper to make the data available. The
2376 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2377 * the data. Another one consists in using **bpf_skb_pull_data**
2378 * to pull in once the non-linear parts, then retesting and
2379 * eventually access the data.
2381 * At the same time, this also makes sure the *skb* is uncloned,
2382 * which is a necessary condition for direct write. As this needs
2383 * to be an invariant for the write part only, the verifier
2384 * detects writes and adds a prologue that is calling
2385 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2386 * the very beginning in case it is indeed cloned.
2388 * A call to this helper is susceptible to change the underlying
2389 * packet buffer. Therefore, at load time, all checks on pointers
2390 * previously done by the verifier are invalidated and must be
2391 * performed again, if the helper is used in combination with
2392 * direct packet access.
2394 * 0 on success, or a negative error in case of failure.
2396 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2398 * Add the checksum *csum* into *skb*\ **->csum** in case the
2399 * driver has supplied a checksum for the entire packet into that
2400 * field. Return an error otherwise. This helper is intended to be
2401 * used in combination with **bpf_csum_diff**\ (), in particular
2402 * when the checksum needs to be updated after data has been
2403 * written into the packet through direct packet access.
2405 * The checksum on success, or a negative error code in case of
2408 * void bpf_set_hash_invalid(struct sk_buff *skb)
2410 * Invalidate the current *skb*\ **->hash**. It can be used after
2411 * mangling on headers through direct packet access, in order to
2412 * indicate that the hash is outdated and to trigger a
2413 * recalculation the next time the kernel tries to access this
2414 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2418 * long bpf_get_numa_node_id(void)
2420 * Return the id of the current NUMA node. The primary use case
2421 * for this helper is the selection of sockets for the local NUMA
2422 * node, when the program is attached to sockets using the
2423 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2424 * but the helper is also available to other eBPF program types,
2425 * similarly to **bpf_get_smp_processor_id**\ ().
2427 * The id of current NUMA node.
2429 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2431 * Grows headroom of packet associated to *skb* and adjusts the
2432 * offset of the MAC header accordingly, adding *len* bytes of
2433 * space. It automatically extends and reallocates memory as
2436 * This helper can be used on a layer 3 *skb* to push a MAC header
2437 * for redirection into a layer 2 device.
2439 * All values for *flags* are reserved for future usage, and must
2442 * A call to this helper is susceptible to change the underlying
2443 * packet buffer. Therefore, at load time, all checks on pointers
2444 * previously done by the verifier are invalidated and must be
2445 * performed again, if the helper is used in combination with
2446 * direct packet access.
2448 * 0 on success, or a negative error in case of failure.
2450 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2452 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2453 * it is possible to use a negative value for *delta*. This helper
2454 * can be used to prepare the packet for pushing or popping
2457 * A call to this helper is susceptible to change the underlying
2458 * packet buffer. Therefore, at load time, all checks on pointers
2459 * previously done by the verifier are invalidated and must be
2460 * performed again, if the helper is used in combination with
2461 * direct packet access.
2463 * 0 on success, or a negative error in case of failure.
2465 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2467 * Copy a NUL terminated string from an unsafe kernel address
2468 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2471 * Generally, use **bpf_probe_read_user_str**\ () or
2472 * **bpf_probe_read_kernel_str**\ () instead.
2474 * On success, the strictly positive length of the string,
2475 * including the trailing NUL character. On error, a negative
2478 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2480 * If the **struct sk_buff** pointed by *skb* has a known socket,
2481 * retrieve the cookie (generated by the kernel) of this socket.
2482 * If no cookie has been set yet, generate a new cookie. Once
2483 * generated, the socket cookie remains stable for the life of the
2484 * socket. This helper can be useful for monitoring per socket
2485 * networking traffic statistics as it provides a global socket
2486 * identifier that can be assumed unique.
2488 * A 8-byte long unique number on success, or 0 if the socket
2489 * field is missing inside *skb*.
2491 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2493 * Equivalent to bpf_get_socket_cookie() helper that accepts
2494 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2496 * A 8-byte long unique number.
2498 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2500 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2501 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2503 * A 8-byte long unique number.
2505 * u64 bpf_get_socket_cookie(struct sock *sk)
2507 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2508 * *sk*, but gets socket from a BTF **struct sock**. This helper
2509 * also works for sleepable programs.
2511 * A 8-byte long unique number or 0 if *sk* is NULL.
2513 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2515 * Get the owner UID of the socked associated to *skb*.
2517 * The owner UID of the socket associated to *skb*. If the socket
2518 * is **NULL**, or if it is not a full socket (i.e. if it is a
2519 * time-wait or a request socket instead), **overflowuid** value
2520 * is returned (note that **overflowuid** might also be the actual
2521 * UID value for the socket).
2523 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2525 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2530 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2532 * Emulate a call to **setsockopt()** on the socket associated to
2533 * *bpf_socket*, which must be a full socket. The *level* at
2534 * which the option resides and the name *optname* of the option
2535 * must be specified, see **setsockopt(2)** for more information.
2536 * The option value of length *optlen* is pointed by *optval*.
2538 * *bpf_socket* should be one of the following:
2540 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2541 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2542 * and **BPF_CGROUP_INET6_CONNECT**.
2544 * This helper actually implements a subset of **setsockopt()**.
2545 * It supports the following *level*\ s:
2547 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2548 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2549 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2550 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2551 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2552 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2553 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2554 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2555 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2556 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2557 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2559 * 0 on success, or a negative error in case of failure.
2561 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2563 * Grow or shrink the room for data in the packet associated to
2564 * *skb* by *len_diff*, and according to the selected *mode*.
2566 * By default, the helper will reset any offloaded checksum
2567 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2568 * by the following flag:
2570 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2571 * checksum data of the skb to CHECKSUM_NONE.
2573 * There are two supported modes at this time:
2575 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2576 * (room space is added or removed below the layer 2 header).
2578 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2579 * (room space is added or removed below the layer 3 header).
2581 * The following flags are supported at this time:
2583 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2584 * Adjusting mss in this way is not allowed for datagrams.
2586 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2587 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2588 * Any new space is reserved to hold a tunnel header.
2589 * Configure skb offsets and other fields accordingly.
2591 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2592 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2593 * Use with ENCAP_L3 flags to further specify the tunnel type.
2595 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2596 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2597 * type; *len* is the length of the inner MAC header.
2599 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2600 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2601 * L2 type as Ethernet.
2603 * A call to this helper is susceptible to change the underlying
2604 * packet buffer. Therefore, at load time, all checks on pointers
2605 * previously done by the verifier are invalidated and must be
2606 * performed again, if the helper is used in combination with
2607 * direct packet access.
2609 * 0 on success, or a negative error in case of failure.
2611 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2613 * Redirect the packet to the endpoint referenced by *map* at
2614 * index *key*. Depending on its type, this *map* can contain
2615 * references to net devices (for forwarding packets through other
2616 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2617 * but this is only implemented for native XDP (with driver
2618 * support) as of this writing).
2620 * The lower two bits of *flags* are used as the return code if
2621 * the map lookup fails. This is so that the return value can be
2622 * one of the XDP program return codes up to **XDP_TX**, as chosen
2623 * by the caller. The higher bits of *flags* can be set to
2624 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2626 * With BPF_F_BROADCAST the packet will be broadcasted to all the
2627 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2628 * interface will be excluded when do broadcasting.
2630 * See also **bpf_redirect**\ (), which only supports redirecting
2631 * to an ifindex, but doesn't require a map to do so.
2633 * **XDP_REDIRECT** on success, or the value of the two lower bits
2634 * of the *flags* argument on error.
2636 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2638 * Redirect the packet to the socket referenced by *map* (of type
2639 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2640 * egress interfaces can be used for redirection. The
2641 * **BPF_F_INGRESS** value in *flags* is used to make the
2642 * distinction (ingress path is selected if the flag is present,
2643 * egress path otherwise). This is the only flag supported for now.
2645 * **SK_PASS** on success, or **SK_DROP** on error.
2647 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2649 * Add an entry to, or update a *map* referencing sockets. The
2650 * *skops* is used as a new value for the entry associated to
2651 * *key*. *flags* is one of:
2654 * The entry for *key* must not exist in the map.
2656 * The entry for *key* must already exist in the map.
2658 * No condition on the existence of the entry for *key*.
2660 * If the *map* has eBPF programs (parser and verdict), those will
2661 * be inherited by the socket being added. If the socket is
2662 * already attached to eBPF programs, this results in an error.
2664 * 0 on success, or a negative error in case of failure.
2666 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2668 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2669 * *delta* (which can be positive or negative). Note that this
2670 * operation modifies the address stored in *xdp_md*\ **->data**,
2671 * so the latter must be loaded only after the helper has been
2674 * The use of *xdp_md*\ **->data_meta** is optional and programs
2675 * are not required to use it. The rationale is that when the
2676 * packet is processed with XDP (e.g. as DoS filter), it is
2677 * possible to push further meta data along with it before passing
2678 * to the stack, and to give the guarantee that an ingress eBPF
2679 * program attached as a TC classifier on the same device can pick
2680 * this up for further post-processing. Since TC works with socket
2681 * buffers, it remains possible to set from XDP the **mark** or
2682 * **priority** pointers, or other pointers for the socket buffer.
2683 * Having this scratch space generic and programmable allows for
2684 * more flexibility as the user is free to store whatever meta
2687 * A call to this helper is susceptible to change the underlying
2688 * packet buffer. Therefore, at load time, all checks on pointers
2689 * previously done by the verifier are invalidated and must be
2690 * performed again, if the helper is used in combination with
2691 * direct packet access.
2693 * 0 on success, or a negative error in case of failure.
2695 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2697 * Read the value of a perf event counter, and store it into *buf*
2698 * of size *buf_size*. This helper relies on a *map* of type
2699 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2700 * counter is selected when *map* is updated with perf event file
2701 * descriptors. The *map* is an array whose size is the number of
2702 * available CPUs, and each cell contains a value relative to one
2703 * CPU. The value to retrieve is indicated by *flags*, that
2704 * contains the index of the CPU to look up, masked with
2705 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2706 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2707 * current CPU should be retrieved.
2709 * This helper behaves in a way close to
2710 * **bpf_perf_event_read**\ () helper, save that instead of
2711 * just returning the value observed, it fills the *buf*
2712 * structure. This allows for additional data to be retrieved: in
2713 * particular, the enabled and running times (in *buf*\
2714 * **->enabled** and *buf*\ **->running**, respectively) are
2715 * copied. In general, **bpf_perf_event_read_value**\ () is
2716 * recommended over **bpf_perf_event_read**\ (), which has some
2717 * ABI issues and provides fewer functionalities.
2719 * These values are interesting, because hardware PMU (Performance
2720 * Monitoring Unit) counters are limited resources. When there are
2721 * more PMU based perf events opened than available counters,
2722 * kernel will multiplex these events so each event gets certain
2723 * percentage (but not all) of the PMU time. In case that
2724 * multiplexing happens, the number of samples or counter value
2725 * will not reflect the case compared to when no multiplexing
2726 * occurs. This makes comparison between different runs difficult.
2727 * Typically, the counter value should be normalized before
2728 * comparing to other experiments. The usual normalization is done
2733 * normalized_counter = counter * t_enabled / t_running
2735 * Where t_enabled is the time enabled for event and t_running is
2736 * the time running for event since last normalization. The
2737 * enabled and running times are accumulated since the perf event
2738 * open. To achieve scaling factor between two invocations of an
2739 * eBPF program, users can use CPU id as the key (which is
2740 * typical for perf array usage model) to remember the previous
2741 * value and do the calculation inside the eBPF program.
2743 * 0 on success, or a negative error in case of failure.
2745 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2747 * For en eBPF program attached to a perf event, retrieve the
2748 * value of the event counter associated to *ctx* and store it in
2749 * the structure pointed by *buf* and of size *buf_size*. Enabled
2750 * and running times are also stored in the structure (see
2751 * description of helper **bpf_perf_event_read_value**\ () for
2754 * 0 on success, or a negative error in case of failure.
2756 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2758 * Emulate a call to **getsockopt()** on the socket associated to
2759 * *bpf_socket*, which must be a full socket. The *level* at
2760 * which the option resides and the name *optname* of the option
2761 * must be specified, see **getsockopt(2)** for more information.
2762 * The retrieved value is stored in the structure pointed by
2763 * *opval* and of length *optlen*.
2765 * *bpf_socket* should be one of the following:
2767 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2768 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2769 * and **BPF_CGROUP_INET6_CONNECT**.
2771 * This helper actually implements a subset of **getsockopt()**.
2772 * It supports the following *level*\ s:
2774 * * **IPPROTO_TCP**, which supports *optname*
2775 * **TCP_CONGESTION**.
2776 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2777 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2779 * 0 on success, or a negative error in case of failure.
2781 * long bpf_override_return(struct pt_regs *regs, u64 rc)
2783 * Used for error injection, this helper uses kprobes to override
2784 * the return value of the probed function, and to set it to *rc*.
2785 * The first argument is the context *regs* on which the kprobe
2788 * This helper works by setting the PC (program counter)
2789 * to an override function which is run in place of the original
2790 * probed function. This means the probed function is not run at
2791 * all. The replacement function just returns with the required
2794 * This helper has security implications, and thus is subject to
2795 * restrictions. It is only available if the kernel was compiled
2796 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2797 * option, and in this case it only works on functions tagged with
2798 * **ALLOW_ERROR_INJECTION** in the kernel code.
2800 * Also, the helper is only available for the architectures having
2801 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2802 * x86 architecture is the only one to support this feature.
2806 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2808 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2809 * for the full TCP socket associated to *bpf_sock_ops* to
2812 * The primary use of this field is to determine if there should
2813 * be calls to eBPF programs of type
2814 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2815 * code. A program of the same type can change its value, per
2816 * connection and as necessary, when the connection is
2817 * established. This field is directly accessible for reading, but
2818 * this helper must be used for updates in order to return an
2819 * error if an eBPF program tries to set a callback that is not
2820 * supported in the current kernel.
2822 * *argval* is a flag array which can combine these flags:
2824 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2825 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2826 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2827 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2829 * Therefore, this function can be used to clear a callback flag by
2830 * setting the appropriate bit to zero. e.g. to disable the RTO
2833 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
2834 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2836 * Here are some examples of where one could call such eBPF
2840 * * When a packet is retransmitted.
2841 * * When the connection terminates.
2842 * * When a packet is sent.
2843 * * When a packet is received.
2845 * Code **-EINVAL** if the socket is not a full TCP socket;
2846 * otherwise, a positive number containing the bits that could not
2847 * be set is returned (which comes down to 0 if all bits were set
2850 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2852 * This helper is used in programs implementing policies at the
2853 * socket level. If the message *msg* is allowed to pass (i.e. if
2854 * the verdict eBPF program returns **SK_PASS**), redirect it to
2855 * the socket referenced by *map* (of type
2856 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2857 * egress interfaces can be used for redirection. The
2858 * **BPF_F_INGRESS** value in *flags* is used to make the
2859 * distinction (ingress path is selected if the flag is present,
2860 * egress path otherwise). This is the only flag supported for now.
2862 * **SK_PASS** on success, or **SK_DROP** on error.
2864 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2866 * For socket policies, apply the verdict of the eBPF program to
2867 * the next *bytes* (number of bytes) of message *msg*.
2869 * For example, this helper can be used in the following cases:
2871 * * A single **sendmsg**\ () or **sendfile**\ () system call
2872 * contains multiple logical messages that the eBPF program is
2873 * supposed to read and for which it should apply a verdict.
2874 * * An eBPF program only cares to read the first *bytes* of a
2875 * *msg*. If the message has a large payload, then setting up
2876 * and calling the eBPF program repeatedly for all bytes, even
2877 * though the verdict is already known, would create unnecessary
2880 * When called from within an eBPF program, the helper sets a
2881 * counter internal to the BPF infrastructure, that is used to
2882 * apply the last verdict to the next *bytes*. If *bytes* is
2883 * smaller than the current data being processed from a
2884 * **sendmsg**\ () or **sendfile**\ () system call, the first
2885 * *bytes* will be sent and the eBPF program will be re-run with
2886 * the pointer for start of data pointing to byte number *bytes*
2887 * **+ 1**. If *bytes* is larger than the current data being
2888 * processed, then the eBPF verdict will be applied to multiple
2889 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2892 * Note that if a socket closes with the internal counter holding
2893 * a non-zero value, this is not a problem because data is not
2894 * being buffered for *bytes* and is sent as it is received.
2898 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2900 * For socket policies, prevent the execution of the verdict eBPF
2901 * program for message *msg* until *bytes* (byte number) have been
2904 * This can be used when one needs a specific number of bytes
2905 * before a verdict can be assigned, even if the data spans
2906 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2907 * case would be a user calling **sendmsg**\ () repeatedly with
2908 * 1-byte long message segments. Obviously, this is bad for
2909 * performance, but it is still valid. If the eBPF program needs
2910 * *bytes* bytes to validate a header, this helper can be used to
2911 * prevent the eBPF program to be called again until *bytes* have
2916 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2918 * For socket policies, pull in non-linear data from user space
2919 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2920 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2923 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2924 * *msg* it can only parse data that the (**data**, **data_end**)
2925 * pointers have already consumed. For **sendmsg**\ () hooks this
2926 * is likely the first scatterlist element. But for calls relying
2927 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2928 * be the range (**0**, **0**) because the data is shared with
2929 * user space and by default the objective is to avoid allowing
2930 * user space to modify data while (or after) eBPF verdict is
2931 * being decided. This helper can be used to pull in data and to
2932 * set the start and end pointer to given values. Data will be
2933 * copied if necessary (i.e. if data was not linear and if start
2934 * and end pointers do not point to the same chunk).
2936 * A call to this helper is susceptible to change the underlying
2937 * packet buffer. Therefore, at load time, all checks on pointers
2938 * previously done by the verifier are invalidated and must be
2939 * performed again, if the helper is used in combination with
2940 * direct packet access.
2942 * All values for *flags* are reserved for future usage, and must
2945 * 0 on success, or a negative error in case of failure.
2947 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2949 * Bind the socket associated to *ctx* to the address pointed by
2950 * *addr*, of length *addr_len*. This allows for making outgoing
2951 * connection from the desired IP address, which can be useful for
2952 * example when all processes inside a cgroup should use one
2953 * single IP address on a host that has multiple IP configured.
2955 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2956 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2957 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2958 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2959 * behavior and lets the kernel efficiently pick up an unused
2960 * port as long as 4-tuple is unique. Passing non-zero port might
2961 * lead to degraded performance.
2963 * 0 on success, or a negative error in case of failure.
2965 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2967 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2968 * possible to both shrink and grow the packet tail.
2969 * Shrink done via *delta* being a negative integer.
2971 * A call to this helper is susceptible to change the underlying
2972 * packet buffer. Therefore, at load time, all checks on pointers
2973 * previously done by the verifier are invalidated and must be
2974 * performed again, if the helper is used in combination with
2975 * direct packet access.
2977 * 0 on success, or a negative error in case of failure.
2979 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2981 * Retrieve the XFRM state (IP transform framework, see also
2982 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2984 * The retrieved value is stored in the **struct bpf_xfrm_state**
2985 * pointed by *xfrm_state* and of length *size*.
2987 * All values for *flags* are reserved for future usage, and must
2990 * This helper is available only if the kernel was compiled with
2991 * **CONFIG_XFRM** configuration option.
2993 * 0 on success, or a negative error in case of failure.
2995 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2997 * Return a user or a kernel stack in bpf program provided buffer.
2998 * To achieve this, the helper needs *ctx*, which is a pointer
2999 * to the context on which the tracing program is executed.
3000 * To store the stacktrace, the bpf program provides *buf* with
3001 * a nonnegative *size*.
3003 * The last argument, *flags*, holds the number of stack frames to
3004 * skip (from 0 to 255), masked with
3005 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3006 * the following flags:
3008 * **BPF_F_USER_STACK**
3009 * Collect a user space stack instead of a kernel stack.
3010 * **BPF_F_USER_BUILD_ID**
3011 * Collect buildid+offset instead of ips for user stack,
3012 * only valid if **BPF_F_USER_STACK** is also specified.
3014 * **bpf_get_stack**\ () can collect up to
3015 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3016 * to sufficient large buffer size. Note that
3017 * this limit can be controlled with the **sysctl** program, and
3018 * that it should be manually increased in order to profile long
3019 * user stacks (such as stacks for Java programs). To do so, use:
3023 * # sysctl kernel.perf_event_max_stack=<new value>
3025 * The non-negative copied *buf* length equal to or less than
3026 * *size* on success, or a negative error in case of failure.
3028 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3030 * This helper is similar to **bpf_skb_load_bytes**\ () in that
3031 * it provides an easy way to load *len* bytes from *offset*
3032 * from the packet associated to *skb*, into the buffer pointed
3033 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3034 * a fifth argument *start_header* exists in order to select a
3035 * base offset to start from. *start_header* can be one of:
3037 * **BPF_HDR_START_MAC**
3038 * Base offset to load data from is *skb*'s mac header.
3039 * **BPF_HDR_START_NET**
3040 * Base offset to load data from is *skb*'s network header.
3042 * In general, "direct packet access" is the preferred method to
3043 * access packet data, however, this helper is in particular useful
3044 * in socket filters where *skb*\ **->data** does not always point
3045 * to the start of the mac header and where "direct packet access"
3048 * 0 on success, or a negative error in case of failure.
3050 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3052 * Do FIB lookup in kernel tables using parameters in *params*.
3053 * If lookup is successful and result shows packet is to be
3054 * forwarded, the neighbor tables are searched for the nexthop.
3055 * If successful (ie., FIB lookup shows forwarding and nexthop
3056 * is resolved), the nexthop address is returned in ipv4_dst
3057 * or ipv6_dst based on family, smac is set to mac address of
3058 * egress device, dmac is set to nexthop mac address, rt_metric
3059 * is set to metric from route (IPv4/IPv6 only), and ifindex
3060 * is set to the device index of the nexthop from the FIB lookup.
3062 * *plen* argument is the size of the passed in struct.
3063 * *flags* argument can be a combination of one or more of the
3066 * **BPF_FIB_LOOKUP_DIRECT**
3067 * Do a direct table lookup vs full lookup using FIB
3069 * **BPF_FIB_LOOKUP_OUTPUT**
3070 * Perform lookup from an egress perspective (default is
3073 * *ctx* is either **struct xdp_md** for XDP programs or
3074 * **struct sk_buff** tc cls_act programs.
3076 * * < 0 if any input argument is invalid
3077 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3078 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3079 * packet is not forwarded or needs assist from full stack
3081 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3082 * was exceeded and output params->mtu_result contains the MTU.
3084 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3086 * Add an entry to, or update a sockhash *map* referencing sockets.
3087 * The *skops* is used as a new value for the entry associated to
3088 * *key*. *flags* is one of:
3091 * The entry for *key* must not exist in the map.
3093 * The entry for *key* must already exist in the map.
3095 * No condition on the existence of the entry for *key*.
3097 * If the *map* has eBPF programs (parser and verdict), those will
3098 * be inherited by the socket being added. If the socket is
3099 * already attached to eBPF programs, this results in an error.
3101 * 0 on success, or a negative error in case of failure.
3103 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3105 * This helper is used in programs implementing policies at the
3106 * socket level. If the message *msg* is allowed to pass (i.e. if
3107 * the verdict eBPF program returns **SK_PASS**), redirect it to
3108 * the socket referenced by *map* (of type
3109 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3110 * egress interfaces can be used for redirection. The
3111 * **BPF_F_INGRESS** value in *flags* is used to make the
3112 * distinction (ingress path is selected if the flag is present,
3113 * egress path otherwise). This is the only flag supported for now.
3115 * **SK_PASS** on success, or **SK_DROP** on error.
3117 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3119 * This helper is used in programs implementing policies at the
3120 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3121 * if the verdict eBPF program returns **SK_PASS**), redirect it
3122 * to the socket referenced by *map* (of type
3123 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3124 * egress interfaces can be used for redirection. The
3125 * **BPF_F_INGRESS** value in *flags* is used to make the
3126 * distinction (ingress path is selected if the flag is present,
3127 * egress otherwise). This is the only flag supported for now.
3129 * **SK_PASS** on success, or **SK_DROP** on error.
3131 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3133 * Encapsulate the packet associated to *skb* within a Layer 3
3134 * protocol header. This header is provided in the buffer at
3135 * address *hdr*, with *len* its size in bytes. *type* indicates
3136 * the protocol of the header and can be one of:
3138 * **BPF_LWT_ENCAP_SEG6**
3139 * IPv6 encapsulation with Segment Routing Header
3140 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3141 * the IPv6 header is computed by the kernel.
3142 * **BPF_LWT_ENCAP_SEG6_INLINE**
3143 * Only works if *skb* contains an IPv6 packet. Insert a
3144 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3146 * **BPF_LWT_ENCAP_IP**
3147 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3148 * must be IPv4 or IPv6, followed by zero or more
3149 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3150 * total bytes in all prepended headers. Please note that
3151 * if **skb_is_gso**\ (*skb*) is true, no more than two
3152 * headers can be prepended, and the inner header, if
3153 * present, should be either GRE or UDP/GUE.
3155 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3156 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3157 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3158 * **BPF_PROG_TYPE_LWT_XMIT**.
3160 * A call to this helper is susceptible to change the underlying
3161 * packet buffer. Therefore, at load time, all checks on pointers
3162 * previously done by the verifier are invalidated and must be
3163 * performed again, if the helper is used in combination with
3164 * direct packet access.
3166 * 0 on success, or a negative error in case of failure.
3168 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3170 * Store *len* bytes from address *from* into the packet
3171 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3172 * inside the outermost IPv6 Segment Routing Header can be
3173 * modified through this helper.
3175 * A call to this helper is susceptible to change the underlying
3176 * packet buffer. Therefore, at load time, all checks on pointers
3177 * previously done by the verifier are invalidated and must be
3178 * performed again, if the helper is used in combination with
3179 * direct packet access.
3181 * 0 on success, or a negative error in case of failure.
3183 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3185 * Adjust the size allocated to TLVs in the outermost IPv6
3186 * Segment Routing Header contained in the packet associated to
3187 * *skb*, at position *offset* by *delta* bytes. Only offsets
3188 * after the segments are accepted. *delta* can be as well
3189 * positive (growing) as negative (shrinking).
3191 * A call to this helper is susceptible to change the underlying
3192 * packet buffer. Therefore, at load time, all checks on pointers
3193 * previously done by the verifier are invalidated and must be
3194 * performed again, if the helper is used in combination with
3195 * direct packet access.
3197 * 0 on success, or a negative error in case of failure.
3199 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3201 * Apply an IPv6 Segment Routing action of type *action* to the
3202 * packet associated to *skb*. Each action takes a parameter
3203 * contained at address *param*, and of length *param_len* bytes.
3204 * *action* can be one of:
3206 * **SEG6_LOCAL_ACTION_END_X**
3207 * End.X action: Endpoint with Layer-3 cross-connect.
3208 * Type of *param*: **struct in6_addr**.
3209 * **SEG6_LOCAL_ACTION_END_T**
3210 * End.T action: Endpoint with specific IPv6 table lookup.
3211 * Type of *param*: **int**.
3212 * **SEG6_LOCAL_ACTION_END_B6**
3213 * End.B6 action: Endpoint bound to an SRv6 policy.
3214 * Type of *param*: **struct ipv6_sr_hdr**.
3215 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3216 * End.B6.Encap action: Endpoint bound to an SRv6
3217 * encapsulation policy.
3218 * Type of *param*: **struct ipv6_sr_hdr**.
3220 * A call to this helper is susceptible to change the underlying
3221 * packet buffer. Therefore, at load time, all checks on pointers
3222 * previously done by the verifier are invalidated and must be
3223 * performed again, if the helper is used in combination with
3224 * direct packet access.
3226 * 0 on success, or a negative error in case of failure.
3228 * long bpf_rc_repeat(void *ctx)
3230 * This helper is used in programs implementing IR decoding, to
3231 * report a successfully decoded repeat key message. This delays
3232 * the generation of a key up event for previously generated
3235 * Some IR protocols like NEC have a special IR message for
3236 * repeating last button, for when a button is held down.
3238 * The *ctx* should point to the lirc sample as passed into
3241 * This helper is only available is the kernel was compiled with
3242 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3247 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3249 * This helper is used in programs implementing IR decoding, to
3250 * report a successfully decoded key press with *scancode*,
3251 * *toggle* value in the given *protocol*. The scancode will be
3252 * translated to a keycode using the rc keymap, and reported as
3253 * an input key down event. After a period a key up event is
3254 * generated. This period can be extended by calling either
3255 * **bpf_rc_keydown**\ () again with the same values, or calling
3256 * **bpf_rc_repeat**\ ().
3258 * Some protocols include a toggle bit, in case the button was
3259 * released and pressed again between consecutive scancodes.
3261 * The *ctx* should point to the lirc sample as passed into
3264 * The *protocol* is the decoded protocol number (see
3265 * **enum rc_proto** for some predefined values).
3267 * This helper is only available is the kernel was compiled with
3268 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3273 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3275 * Return the cgroup v2 id of the socket associated with the *skb*.
3276 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3277 * helper for cgroup v1 by providing a tag resp. identifier that
3278 * can be matched on or used for map lookups e.g. to implement
3279 * policy. The cgroup v2 id of a given path in the hierarchy is
3280 * exposed in user space through the f_handle API in order to get
3281 * to the same 64-bit id.
3283 * This helper can be used on TC egress path, but not on ingress,
3284 * and is available only if the kernel was compiled with the
3285 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3287 * The id is returned or 0 in case the id could not be retrieved.
3289 * u64 bpf_get_current_cgroup_id(void)
3291 * Get the current cgroup id based on the cgroup within which
3292 * the current task is running.
3294 * A 64-bit integer containing the current cgroup id based
3295 * on the cgroup within which the current task is running.
3297 * void *bpf_get_local_storage(void *map, u64 flags)
3299 * Get the pointer to the local storage area.
3300 * The type and the size of the local storage is defined
3301 * by the *map* argument.
3302 * The *flags* meaning is specific for each map type,
3303 * and has to be 0 for cgroup local storage.
3305 * Depending on the BPF program type, a local storage area
3306 * can be shared between multiple instances of the BPF program,
3307 * running simultaneously.
3309 * A user should care about the synchronization by himself.
3310 * For example, by using the **BPF_ATOMIC** instructions to alter
3313 * A pointer to the local storage area.
3315 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3317 * Select a **SO_REUSEPORT** socket from a
3318 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3319 * It checks the selected socket is matching the incoming
3320 * request in the socket buffer.
3322 * 0 on success, or a negative error in case of failure.
3324 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3326 * Return id of cgroup v2 that is ancestor of cgroup associated
3327 * with the *skb* at the *ancestor_level*. The root cgroup is at
3328 * *ancestor_level* zero and each step down the hierarchy
3329 * increments the level. If *ancestor_level* == level of cgroup
3330 * associated with *skb*, then return value will be same as that
3331 * of **bpf_skb_cgroup_id**\ ().
3333 * The helper is useful to implement policies based on cgroups
3334 * that are upper in hierarchy than immediate cgroup associated
3337 * The format of returned id and helper limitations are same as in
3338 * **bpf_skb_cgroup_id**\ ().
3340 * The id is returned or 0 in case the id could not be retrieved.
3342 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3344 * Look for TCP socket matching *tuple*, optionally in a child
3345 * network namespace *netns*. The return value must be checked,
3346 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3348 * The *ctx* should point to the context of the program, such as
3349 * the skb or socket (depending on the hook in use). This is used
3350 * to determine the base network namespace for the lookup.
3352 * *tuple_size* must be one of:
3354 * **sizeof**\ (*tuple*\ **->ipv4**)
3355 * Look for an IPv4 socket.
3356 * **sizeof**\ (*tuple*\ **->ipv6**)
3357 * Look for an IPv6 socket.
3359 * If the *netns* is a negative signed 32-bit integer, then the
3360 * socket lookup table in the netns associated with the *ctx*
3361 * will be used. For the TC hooks, this is the netns of the device
3362 * in the skb. For socket hooks, this is the netns of the socket.
3363 * If *netns* is any other signed 32-bit value greater than or
3364 * equal to zero then it specifies the ID of the netns relative to
3365 * the netns associated with the *ctx*. *netns* values beyond the
3366 * range of 32-bit integers are reserved for future use.
3368 * All values for *flags* are reserved for future usage, and must
3371 * This helper is available only if the kernel was compiled with
3372 * **CONFIG_NET** configuration option.
3374 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3375 * For sockets with reuseport option, the **struct bpf_sock**
3376 * result is from *reuse*\ **->socks**\ [] using the hash of the
3379 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3381 * Look for UDP socket matching *tuple*, optionally in a child
3382 * network namespace *netns*. The return value must be checked,
3383 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3385 * The *ctx* should point to the context of the program, such as
3386 * the skb or socket (depending on the hook in use). This is used
3387 * to determine the base network namespace for the lookup.
3389 * *tuple_size* must be one of:
3391 * **sizeof**\ (*tuple*\ **->ipv4**)
3392 * Look for an IPv4 socket.
3393 * **sizeof**\ (*tuple*\ **->ipv6**)
3394 * Look for an IPv6 socket.
3396 * If the *netns* is a negative signed 32-bit integer, then the
3397 * socket lookup table in the netns associated with the *ctx*
3398 * will be used. For the TC hooks, this is the netns of the device
3399 * in the skb. For socket hooks, this is the netns of the socket.
3400 * If *netns* is any other signed 32-bit value greater than or
3401 * equal to zero then it specifies the ID of the netns relative to
3402 * the netns associated with the *ctx*. *netns* values beyond the
3403 * range of 32-bit integers are reserved for future use.
3405 * All values for *flags* are reserved for future usage, and must
3408 * This helper is available only if the kernel was compiled with
3409 * **CONFIG_NET** configuration option.
3411 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3412 * For sockets with reuseport option, the **struct bpf_sock**
3413 * result is from *reuse*\ **->socks**\ [] using the hash of the
3416 * long bpf_sk_release(void *sock)
3418 * Release the reference held by *sock*. *sock* must be a
3419 * non-**NULL** pointer that was returned from
3420 * **bpf_sk_lookup_xxx**\ ().
3422 * 0 on success, or a negative error in case of failure.
3424 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3426 * Push an element *value* in *map*. *flags* is one of:
3429 * If the queue/stack is full, the oldest element is
3430 * removed to make room for this.
3432 * 0 on success, or a negative error in case of failure.
3434 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3436 * Pop an element from *map*.
3438 * 0 on success, or a negative error in case of failure.
3440 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3442 * Get an element from *map* without removing it.
3444 * 0 on success, or a negative error in case of failure.
3446 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3448 * For socket policies, insert *len* bytes into *msg* at offset
3451 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3452 * *msg* it may want to insert metadata or options into the *msg*.
3453 * This can later be read and used by any of the lower layer BPF
3456 * This helper may fail if under memory pressure (a malloc
3457 * fails) in these cases BPF programs will get an appropriate
3458 * error and BPF programs will need to handle them.
3460 * 0 on success, or a negative error in case of failure.
3462 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3464 * Will remove *len* bytes from a *msg* starting at byte *start*.
3465 * This may result in **ENOMEM** errors under certain situations if
3466 * an allocation and copy are required due to a full ring buffer.
3467 * However, the helper will try to avoid doing the allocation
3468 * if possible. Other errors can occur if input parameters are
3469 * invalid either due to *start* byte not being valid part of *msg*
3470 * payload and/or *pop* value being to large.
3472 * 0 on success, or a negative error in case of failure.
3474 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3476 * This helper is used in programs implementing IR decoding, to
3477 * report a successfully decoded pointer movement.
3479 * The *ctx* should point to the lirc sample as passed into
3482 * This helper is only available is the kernel was compiled with
3483 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3488 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3490 * Acquire a spinlock represented by the pointer *lock*, which is
3491 * stored as part of a value of a map. Taking the lock allows to
3492 * safely update the rest of the fields in that value. The
3493 * spinlock can (and must) later be released with a call to
3494 * **bpf_spin_unlock**\ (\ *lock*\ ).
3496 * Spinlocks in BPF programs come with a number of restrictions
3499 * * **bpf_spin_lock** objects are only allowed inside maps of
3500 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3501 * list could be extended in the future).
3502 * * BTF description of the map is mandatory.
3503 * * The BPF program can take ONE lock at a time, since taking two
3504 * or more could cause dead locks.
3505 * * Only one **struct bpf_spin_lock** is allowed per map element.
3506 * * When the lock is taken, calls (either BPF to BPF or helpers)
3508 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3509 * allowed inside a spinlock-ed region.
3510 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3511 * the lock, on all execution paths, before it returns.
3512 * * The BPF program can access **struct bpf_spin_lock** only via
3513 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3514 * helpers. Loading or storing data into the **struct
3515 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3516 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3517 * of the map value must be a struct and have **struct
3518 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3519 * Nested lock inside another struct is not allowed.
3520 * * The **struct bpf_spin_lock** *lock* field in a map value must
3521 * be aligned on a multiple of 4 bytes in that value.
3522 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3523 * the **bpf_spin_lock** field to user space.
3524 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3525 * a BPF program, do not update the **bpf_spin_lock** field.
3526 * * **bpf_spin_lock** cannot be on the stack or inside a
3527 * networking packet (it can only be inside of a map values).
3528 * * **bpf_spin_lock** is available to root only.
3529 * * Tracing programs and socket filter programs cannot use
3530 * **bpf_spin_lock**\ () due to insufficient preemption checks
3531 * (but this may change in the future).
3532 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3536 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3538 * Release the *lock* previously locked by a call to
3539 * **bpf_spin_lock**\ (\ *lock*\ ).
3543 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3545 * This helper gets a **struct bpf_sock** pointer such
3546 * that all the fields in this **bpf_sock** can be accessed.
3548 * A **struct bpf_sock** pointer on success, or **NULL** in
3551 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3553 * This helper gets a **struct bpf_tcp_sock** pointer from a
3554 * **struct bpf_sock** pointer.
3556 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3559 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3561 * Set ECN (Explicit Congestion Notification) field of IP header
3562 * to **CE** (Congestion Encountered) if current value is **ECT**
3563 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3566 * 1 if the **CE** flag is set (either by the current helper call
3567 * or because it was already present), 0 if it is not set.
3569 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3571 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3572 * **bpf_sk_release**\ () is unnecessary and not allowed.
3574 * A **struct bpf_sock** pointer on success, or **NULL** in
3577 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3579 * Look for TCP socket matching *tuple*, optionally in a child
3580 * network namespace *netns*. The return value must be checked,
3581 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3583 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3584 * that it also returns timewait or request sockets. Use
3585 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3588 * This helper is available only if the kernel was compiled with
3589 * **CONFIG_NET** configuration option.
3591 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3592 * For sockets with reuseport option, the **struct bpf_sock**
3593 * result is from *reuse*\ **->socks**\ [] using the hash of the
3596 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3598 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3599 * the listening socket in *sk*.
3601 * *iph* points to the start of the IPv4 or IPv6 header, while
3602 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3603 * **sizeof**\ (**struct ipv6hdr**).
3605 * *th* points to the start of the TCP header, while *th_len*
3606 * contains the length of the TCP header (at least
3607 * **sizeof**\ (**struct tcphdr**)).
3609 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3612 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3614 * Get name of sysctl in /proc/sys/ and copy it into provided by
3615 * program buffer *buf* of size *buf_len*.
3617 * The buffer is always NUL terminated, unless it's zero-sized.
3619 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3620 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3621 * only (e.g. "tcp_mem").
3623 * Number of character copied (not including the trailing NUL).
3625 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3626 * truncated name in this case).
3628 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3630 * Get current value of sysctl as it is presented in /proc/sys
3631 * (incl. newline, etc), and copy it as a string into provided
3632 * by program buffer *buf* of size *buf_len*.
3634 * The whole value is copied, no matter what file position user
3635 * space issued e.g. sys_read at.
3637 * The buffer is always NUL terminated, unless it's zero-sized.
3639 * Number of character copied (not including the trailing NUL).
3641 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3642 * truncated name in this case).
3644 * **-EINVAL** if current value was unavailable, e.g. because
3645 * sysctl is uninitialized and read returns -EIO for it.
3647 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3649 * Get new value being written by user space to sysctl (before
3650 * the actual write happens) and copy it as a string into
3651 * provided by program buffer *buf* of size *buf_len*.
3653 * User space may write new value at file position > 0.
3655 * The buffer is always NUL terminated, unless it's zero-sized.
3657 * Number of character copied (not including the trailing NUL).
3659 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3660 * truncated name in this case).
3662 * **-EINVAL** if sysctl is being read.
3664 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3666 * Override new value being written by user space to sysctl with
3667 * value provided by program in buffer *buf* of size *buf_len*.
3669 * *buf* should contain a string in same form as provided by user
3670 * space on sysctl write.
3672 * User space may write new value at file position > 0. To override
3673 * the whole sysctl value file position should be set to zero.
3677 * **-E2BIG** if the *buf_len* is too big.
3679 * **-EINVAL** if sysctl is being read.
3681 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3683 * Convert the initial part of the string from buffer *buf* of
3684 * size *buf_len* to a long integer according to the given base
3685 * and save the result in *res*.
3687 * The string may begin with an arbitrary amount of white space
3688 * (as determined by **isspace**\ (3)) followed by a single
3689 * optional '**-**' sign.
3691 * Five least significant bits of *flags* encode base, other bits
3692 * are currently unused.
3694 * Base must be either 8, 10, 16 or 0 to detect it automatically
3695 * similar to user space **strtol**\ (3).
3697 * Number of characters consumed on success. Must be positive but
3698 * no more than *buf_len*.
3700 * **-EINVAL** if no valid digits were found or unsupported base
3703 * **-ERANGE** if resulting value was out of range.
3705 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3707 * Convert the initial part of the string from buffer *buf* of
3708 * size *buf_len* to an unsigned long integer according to the
3709 * given base and save the result in *res*.
3711 * The string may begin with an arbitrary amount of white space
3712 * (as determined by **isspace**\ (3)).
3714 * Five least significant bits of *flags* encode base, other bits
3715 * are currently unused.
3717 * Base must be either 8, 10, 16 or 0 to detect it automatically
3718 * similar to user space **strtoul**\ (3).
3720 * Number of characters consumed on success. Must be positive but
3721 * no more than *buf_len*.
3723 * **-EINVAL** if no valid digits were found or unsupported base
3726 * **-ERANGE** if resulting value was out of range.
3728 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3730 * Get a bpf-local-storage from a *sk*.
3732 * Logically, it could be thought of getting the value from
3733 * a *map* with *sk* as the **key**. From this
3734 * perspective, the usage is not much different from
3735 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3736 * helper enforces the key must be a full socket and the map must
3737 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
3739 * Underneath, the value is stored locally at *sk* instead of
3740 * the *map*. The *map* is used as the bpf-local-storage
3741 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3742 * searched against all bpf-local-storages residing at *sk*.
3744 * *sk* is a kernel **struct sock** pointer for LSM program.
3745 * *sk* is a **struct bpf_sock** pointer for other program types.
3747 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3748 * used such that a new bpf-local-storage will be
3749 * created if one does not exist. *value* can be used
3750 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3751 * the initial value of a bpf-local-storage. If *value* is
3752 * **NULL**, the new bpf-local-storage will be zero initialized.
3754 * A bpf-local-storage pointer is returned on success.
3756 * **NULL** if not found or there was an error in adding
3757 * a new bpf-local-storage.
3759 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3761 * Delete a bpf-local-storage from a *sk*.
3765 * **-ENOENT** if the bpf-local-storage cannot be found.
3766 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3768 * long bpf_send_signal(u32 sig)
3770 * Send signal *sig* to the process of the current task.
3771 * The signal may be delivered to any of this process's threads.
3773 * 0 on success or successfully queued.
3775 * **-EBUSY** if work queue under nmi is full.
3777 * **-EINVAL** if *sig* is invalid.
3779 * **-EPERM** if no permission to send the *sig*.
3781 * **-EAGAIN** if bpf program can try again.
3783 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3785 * Try to issue a SYN cookie for the packet with corresponding
3786 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3788 * *iph* points to the start of the IPv4 or IPv6 header, while
3789 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3790 * **sizeof**\ (**struct ipv6hdr**).
3792 * *th* points to the start of the TCP header, while *th_len*
3793 * contains the length of the TCP header with options (at least
3794 * **sizeof**\ (**struct tcphdr**)).
3796 * On success, lower 32 bits hold the generated SYN cookie in
3797 * followed by 16 bits which hold the MSS value for that cookie,
3798 * and the top 16 bits are unused.
3800 * On failure, the returned value is one of the following:
3802 * **-EINVAL** SYN cookie cannot be issued due to error
3804 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
3806 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3808 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
3810 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3812 * Write raw *data* blob into a special BPF perf event held by
3813 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3814 * event must have the following attributes: **PERF_SAMPLE_RAW**
3815 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3816 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3818 * The *flags* are used to indicate the index in *map* for which
3819 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3820 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3821 * to indicate that the index of the current CPU core should be
3824 * The value to write, of *size*, is passed through eBPF stack and
3825 * pointed by *data*.
3827 * *ctx* is a pointer to in-kernel struct sk_buff.
3829 * This helper is similar to **bpf_perf_event_output**\ () but
3830 * restricted to raw_tracepoint bpf programs.
3832 * 0 on success, or a negative error in case of failure.
3834 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3836 * Safely attempt to read *size* bytes from user space address
3837 * *unsafe_ptr* and store the data in *dst*.
3839 * 0 on success, or a negative error in case of failure.
3841 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3843 * Safely attempt to read *size* bytes from kernel space address
3844 * *unsafe_ptr* and store the data in *dst*.
3846 * 0 on success, or a negative error in case of failure.
3848 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3850 * Copy a NUL terminated string from an unsafe user address
3851 * *unsafe_ptr* to *dst*. The *size* should include the
3852 * terminating NUL byte. In case the string length is smaller than
3853 * *size*, the target is not padded with further NUL bytes. If the
3854 * string length is larger than *size*, just *size*-1 bytes are
3855 * copied and the last byte is set to NUL.
3857 * On success, returns the number of bytes that were written,
3858 * including the terminal NUL. This makes this helper useful in
3859 * tracing programs for reading strings, and more importantly to
3860 * get its length at runtime. See the following snippet:
3864 * SEC("kprobe/sys_open")
3865 * void bpf_sys_open(struct pt_regs *ctx)
3867 * char buf[PATHLEN]; // PATHLEN is defined to 256
3868 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3871 * // Consume buf, for example push it to
3872 * // userspace via bpf_perf_event_output(); we
3873 * // can use res (the string length) as event
3874 * // size, after checking its boundaries.
3877 * In comparison, using **bpf_probe_read_user**\ () helper here
3878 * instead to read the string would require to estimate the length
3879 * at compile time, and would often result in copying more memory
3882 * Another useful use case is when parsing individual process
3883 * arguments or individual environment variables navigating
3884 * *current*\ **->mm->arg_start** and *current*\
3885 * **->mm->env_start**: using this helper and the return value,
3886 * one can quickly iterate at the right offset of the memory area.
3888 * On success, the strictly positive length of the output string,
3889 * including the trailing NUL character. On error, a negative
3892 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3894 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3895 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3897 * On success, the strictly positive length of the string, including
3898 * the trailing NUL character. On error, a negative value.
3900 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3902 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3903 * *rcv_nxt* is the ack_seq to be sent out.
3905 * 0 on success, or a negative error in case of failure.
3907 * long bpf_send_signal_thread(u32 sig)
3909 * Send signal *sig* to the thread corresponding to the current task.
3911 * 0 on success or successfully queued.
3913 * **-EBUSY** if work queue under nmi is full.
3915 * **-EINVAL** if *sig* is invalid.
3917 * **-EPERM** if no permission to send the *sig*.
3919 * **-EAGAIN** if bpf program can try again.
3921 * u64 bpf_jiffies64(void)
3923 * Obtain the 64bit jiffies
3925 * The 64 bit jiffies
3927 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3929 * For an eBPF program attached to a perf event, retrieve the
3930 * branch records (**struct perf_branch_entry**) associated to *ctx*
3931 * and store it in the buffer pointed by *buf* up to size
3934 * On success, number of bytes written to *buf*. On error, a
3937 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3938 * instead return the number of bytes required to store all the
3939 * branch entries. If this flag is set, *buf* may be NULL.
3941 * **-EINVAL** if arguments invalid or **size** not a multiple
3942 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3944 * **-ENOENT** if architecture does not support branch records.
3946 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3948 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3949 * *namespace* will be returned in *nsdata*.
3951 * 0 on success, or one of the following in case of failure:
3953 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3954 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3956 * **-ENOENT** if pidns does not exists for the current task.
3958 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3960 * Write raw *data* blob into a special BPF perf event held by
3961 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3962 * event must have the following attributes: **PERF_SAMPLE_RAW**
3963 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3964 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3966 * The *flags* are used to indicate the index in *map* for which
3967 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3968 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3969 * to indicate that the index of the current CPU core should be
3972 * The value to write, of *size*, is passed through eBPF stack and
3973 * pointed by *data*.
3975 * *ctx* is a pointer to in-kernel struct xdp_buff.
3977 * This helper is similar to **bpf_perf_eventoutput**\ () but
3978 * restricted to raw_tracepoint bpf programs.
3980 * 0 on success, or a negative error in case of failure.
3982 * u64 bpf_get_netns_cookie(void *ctx)
3984 * Retrieve the cookie (generated by the kernel) of the network
3985 * namespace the input *ctx* is associated with. The network
3986 * namespace cookie remains stable for its lifetime and provides
3987 * a global identifier that can be assumed unique. If *ctx* is
3988 * NULL, then the helper returns the cookie for the initial
3989 * network namespace. The cookie itself is very similar to that
3990 * of **bpf_get_socket_cookie**\ () helper, but for network
3991 * namespaces instead of sockets.
3993 * A 8-byte long opaque number.
3995 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3997 * Return id of cgroup v2 that is ancestor of the cgroup associated
3998 * with the current task at the *ancestor_level*. The root cgroup
3999 * is at *ancestor_level* zero and each step down the hierarchy
4000 * increments the level. If *ancestor_level* == level of cgroup
4001 * associated with the current task, then return value will be the
4002 * same as that of **bpf_get_current_cgroup_id**\ ().
4004 * The helper is useful to implement policies based on cgroups
4005 * that are upper in hierarchy than immediate cgroup associated
4006 * with the current task.
4008 * The format of returned id and helper limitations are same as in
4009 * **bpf_get_current_cgroup_id**\ ().
4011 * The id is returned or 0 in case the id could not be retrieved.
4013 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4015 * Helper is overloaded depending on BPF program type. This
4016 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4017 * **BPF_PROG_TYPE_SCHED_ACT** programs.
4019 * Assign the *sk* to the *skb*. When combined with appropriate
4020 * routing configuration to receive the packet towards the socket,
4021 * will cause *skb* to be delivered to the specified socket.
4022 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
4023 * **bpf_clone_redirect**\ () or other methods outside of BPF may
4024 * interfere with successful delivery to the socket.
4026 * This operation is only valid from TC ingress path.
4028 * The *flags* argument must be zero.
4030 * 0 on success, or a negative error in case of failure:
4032 * **-EINVAL** if specified *flags* are not supported.
4034 * **-ENOENT** if the socket is unavailable for assignment.
4036 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
4038 * **-EOPNOTSUPP** if the operation is not supported, for example
4039 * a call from outside of TC ingress.
4041 * **-ESOCKTNOSUPPORT** if the socket type is not supported
4044 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4046 * Helper is overloaded depending on BPF program type. This
4047 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4049 * Select the *sk* as a result of a socket lookup.
4051 * For the operation to succeed passed socket must be compatible
4052 * with the packet description provided by the *ctx* object.
4054 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4055 * be an exact match. While IP family (**AF_INET** or
4056 * **AF_INET6**) must be compatible, that is IPv6 sockets
4057 * that are not v6-only can be selected for IPv4 packets.
4059 * Only TCP listeners and UDP unconnected sockets can be
4060 * selected. *sk* can also be NULL to reset any previous
4063 * *flags* argument can combination of following values:
4065 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4066 * socket selection, potentially done by a BPF program
4067 * that ran before us.
4069 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4070 * load-balancing within reuseport group for the socket
4073 * On success *ctx->sk* will point to the selected socket.
4076 * 0 on success, or a negative errno in case of failure.
4078 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4079 * not compatible with packet family (*ctx->family*).
4081 * * **-EEXIST** if socket has been already selected,
4082 * potentially by another program, and
4083 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4085 * * **-EINVAL** if unsupported flags were specified.
4087 * * **-EPROTOTYPE** if socket L4 protocol
4088 * (*sk->protocol*) doesn't match packet protocol
4089 * (*ctx->protocol*).
4091 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4092 * state (TCP listening or UDP unconnected).
4094 * u64 bpf_ktime_get_boot_ns(void)
4096 * Return the time elapsed since system boot, in nanoseconds.
4097 * Does include the time the system was suspended.
4098 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4102 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4104 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4105 * out the format string.
4106 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4107 * the format string itself. The *data* and *data_len* are format string
4108 * arguments. The *data* are a **u64** array and corresponding format string
4109 * values are stored in the array. For strings and pointers where pointees
4110 * are accessed, only the pointer values are stored in the *data* array.
4111 * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4113 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4114 * Reading kernel memory may fail due to either invalid address or
4115 * valid address but requiring a major memory fault. If reading kernel memory
4116 * fails, the string for **%s** will be an empty string, and the ip
4117 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4118 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4120 * 0 on success, or a negative error in case of failure:
4122 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4123 * by returning 1 from bpf program.
4125 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4127 * **-E2BIG** if *fmt* contains too many format specifiers.
4129 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4131 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4133 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4134 * The *m* represents the seq_file. The *data* and *len* represent the
4135 * data to write in bytes.
4137 * 0 on success, or a negative error in case of failure:
4139 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4141 * u64 bpf_sk_cgroup_id(void *sk)
4143 * Return the cgroup v2 id of the socket *sk*.
4145 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4146 * returned from **bpf_sk_lookup_xxx**\ (),
4147 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4148 * same as in **bpf_skb_cgroup_id**\ ().
4150 * This helper is available only if the kernel was compiled with
4151 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4153 * The id is returned or 0 in case the id could not be retrieved.
4155 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4157 * Return id of cgroup v2 that is ancestor of cgroup associated
4158 * with the *sk* at the *ancestor_level*. The root cgroup is at
4159 * *ancestor_level* zero and each step down the hierarchy
4160 * increments the level. If *ancestor_level* == level of cgroup
4161 * associated with *sk*, then return value will be same as that
4162 * of **bpf_sk_cgroup_id**\ ().
4164 * The helper is useful to implement policies based on cgroups
4165 * that are upper in hierarchy than immediate cgroup associated
4168 * The format of returned id and helper limitations are same as in
4169 * **bpf_sk_cgroup_id**\ ().
4171 * The id is returned or 0 in case the id could not be retrieved.
4173 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4175 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4176 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4177 * of new data availability is sent.
4178 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4179 * of new data availability is sent unconditionally.
4180 * If **0** is specified in *flags*, an adaptive notification
4181 * of new data availability is sent.
4183 * An adaptive notification is a notification sent whenever the user-space
4184 * process has caught up and consumed all available payloads. In case the user-space
4185 * process is still processing a previous payload, then no notification is needed
4186 * as it will process the newly added payload automatically.
4188 * 0 on success, or a negative error in case of failure.
4190 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4192 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4193 * *flags* must be 0.
4195 * Valid pointer with *size* bytes of memory available; NULL,
4198 * void bpf_ringbuf_submit(void *data, u64 flags)
4200 * Submit reserved ring buffer sample, pointed to by *data*.
4201 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4202 * of new data availability is sent.
4203 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4204 * of new data availability is sent unconditionally.
4205 * If **0** is specified in *flags*, an adaptive notification
4206 * of new data availability is sent.
4208 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4210 * Nothing. Always succeeds.
4212 * void bpf_ringbuf_discard(void *data, u64 flags)
4214 * Discard reserved ring buffer sample, pointed to by *data*.
4215 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4216 * of new data availability is sent.
4217 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4218 * of new data availability is sent unconditionally.
4219 * If **0** is specified in *flags*, an adaptive notification
4220 * of new data availability is sent.
4222 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4224 * Nothing. Always succeeds.
4226 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4228 * Query various characteristics of provided ring buffer. What
4229 * exactly is queries is determined by *flags*:
4231 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4232 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4233 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4234 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4236 * Data returned is just a momentary snapshot of actual values
4237 * and could be inaccurate, so this facility should be used to
4238 * power heuristics and for reporting, not to make 100% correct
4241 * Requested value, or 0, if *flags* are not recognized.
4243 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4245 * Change the skbs checksum level by one layer up or down, or
4246 * reset it entirely to none in order to have the stack perform
4247 * checksum validation. The level is applicable to the following
4248 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4249 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4250 * through **bpf_skb_adjust_room**\ () helper with passing in
4251 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4252 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4253 * the UDP header is removed. Similarly, an encap of the latter
4254 * into the former could be accompanied by a helper call to
4255 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4256 * skb is still intended to be processed in higher layers of the
4257 * stack instead of just egressing at tc.
4259 * There are three supported level settings at this time:
4261 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4262 * with CHECKSUM_UNNECESSARY.
4263 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4264 * with CHECKSUM_UNNECESSARY.
4265 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4266 * sets CHECKSUM_NONE to force checksum validation by the stack.
4267 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4270 * 0 on success, or a negative error in case of failure. In the
4271 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4272 * is returned or the error code -EACCES in case the skb is not
4273 * subject to CHECKSUM_UNNECESSARY.
4275 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4277 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4279 * *sk* if casting is valid, or **NULL** otherwise.
4281 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4283 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4285 * *sk* if casting is valid, or **NULL** otherwise.
4287 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4289 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4291 * *sk* if casting is valid, or **NULL** otherwise.
4293 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4295 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4297 * *sk* if casting is valid, or **NULL** otherwise.
4299 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4301 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4303 * *sk* if casting is valid, or **NULL** otherwise.
4305 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4307 * Return a user or a kernel stack in bpf program provided buffer.
4308 * To achieve this, the helper needs *task*, which is a valid
4309 * pointer to **struct task_struct**. To store the stacktrace, the
4310 * bpf program provides *buf* with a nonnegative *size*.
4312 * The last argument, *flags*, holds the number of stack frames to
4313 * skip (from 0 to 255), masked with
4314 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4315 * the following flags:
4317 * **BPF_F_USER_STACK**
4318 * Collect a user space stack instead of a kernel stack.
4319 * **BPF_F_USER_BUILD_ID**
4320 * Collect buildid+offset instead of ips for user stack,
4321 * only valid if **BPF_F_USER_STACK** is also specified.
4323 * **bpf_get_task_stack**\ () can collect up to
4324 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4325 * to sufficient large buffer size. Note that
4326 * this limit can be controlled with the **sysctl** program, and
4327 * that it should be manually increased in order to profile long
4328 * user stacks (such as stacks for Java programs). To do so, use:
4332 * # sysctl kernel.perf_event_max_stack=<new value>
4334 * The non-negative copied *buf* length equal to or less than
4335 * *size* on success, or a negative error in case of failure.
4337 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4339 * Load header option. Support reading a particular TCP header
4340 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4342 * If *flags* is 0, it will search the option from the
4343 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4344 * has details on what skb_data contains under different
4345 * *skops*\ **->op**.
4347 * The first byte of the *searchby_res* specifies the
4348 * kind that it wants to search.
4350 * If the searching kind is an experimental kind
4351 * (i.e. 253 or 254 according to RFC6994). It also
4352 * needs to specify the "magic" which is either
4353 * 2 bytes or 4 bytes. It then also needs to
4354 * specify the size of the magic by using
4355 * the 2nd byte which is "kind-length" of a TCP
4356 * header option and the "kind-length" also
4357 * includes the first 2 bytes "kind" and "kind-length"
4358 * itself as a normal TCP header option also does.
4360 * For example, to search experimental kind 254 with
4361 * 2 byte magic 0xeB9F, the searchby_res should be
4362 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4364 * To search for the standard window scale option (3),
4365 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4366 * Note, kind-length must be 0 for regular option.
4368 * Searching for No-Op (0) and End-of-Option-List (1) are
4371 * *len* must be at least 2 bytes which is the minimal size
4372 * of a header option.
4376 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4377 * saved_syn packet or the just-received syn packet.
4380 * > 0 when found, the header option is copied to *searchby_res*.
4381 * The return value is the total length copied. On failure, a
4382 * negative error code is returned:
4384 * **-EINVAL** if a parameter is invalid.
4386 * **-ENOMSG** if the option is not found.
4388 * **-ENOENT** if no syn packet is available when
4389 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4391 * **-ENOSPC** if there is not enough space. Only *len* number of
4394 * **-EFAULT** on failure to parse the header options in the
4397 * **-EPERM** if the helper cannot be used under the current
4398 * *skops*\ **->op**.
4400 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4402 * Store header option. The data will be copied
4403 * from buffer *from* with length *len* to the TCP header.
4405 * The buffer *from* should have the whole option that
4406 * includes the kind, kind-length, and the actual
4407 * option data. The *len* must be at least kind-length
4408 * long. The kind-length does not have to be 4 byte
4409 * aligned. The kernel will take care of the padding
4410 * and setting the 4 bytes aligned value to th->doff.
4412 * This helper will check for duplicated option
4413 * by searching the same option in the outgoing skb.
4415 * This helper can only be called during
4416 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4419 * 0 on success, or negative error in case of failure:
4421 * **-EINVAL** If param is invalid.
4423 * **-ENOSPC** if there is not enough space in the header.
4424 * Nothing has been written
4426 * **-EEXIST** if the option already exists.
4428 * **-EFAULT** on failrue to parse the existing header options.
4430 * **-EPERM** if the helper cannot be used under the current
4431 * *skops*\ **->op**.
4433 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4435 * Reserve *len* bytes for the bpf header option. The
4436 * space will be used by **bpf_store_hdr_opt**\ () later in
4437 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4439 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4440 * the total number of bytes will be reserved.
4442 * This helper can only be called during
4443 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4446 * 0 on success, or negative error in case of failure:
4448 * **-EINVAL** if a parameter is invalid.
4450 * **-ENOSPC** if there is not enough space in the header.
4452 * **-EPERM** if the helper cannot be used under the current
4453 * *skops*\ **->op**.
4455 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4457 * Get a bpf_local_storage from an *inode*.
4459 * Logically, it could be thought of as getting the value from
4460 * a *map* with *inode* as the **key**. From this
4461 * perspective, the usage is not much different from
4462 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4463 * helper enforces the key must be an inode and the map must also
4464 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4466 * Underneath, the value is stored locally at *inode* instead of
4467 * the *map*. The *map* is used as the bpf-local-storage
4468 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4469 * searched against all bpf_local_storage residing at *inode*.
4471 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4472 * used such that a new bpf_local_storage will be
4473 * created if one does not exist. *value* can be used
4474 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4475 * the initial value of a bpf_local_storage. If *value* is
4476 * **NULL**, the new bpf_local_storage will be zero initialized.
4478 * A bpf_local_storage pointer is returned on success.
4480 * **NULL** if not found or there was an error in adding
4481 * a new bpf_local_storage.
4483 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4485 * Delete a bpf_local_storage from an *inode*.
4489 * **-ENOENT** if the bpf_local_storage cannot be found.
4491 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4493 * Return full path for given **struct path** object, which
4494 * needs to be the kernel BTF *path* object. The path is
4495 * returned in the provided buffer *buf* of size *sz* and
4496 * is zero terminated.
4499 * On success, the strictly positive length of the string,
4500 * including the trailing NUL character. On error, a negative
4503 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4505 * Read *size* bytes from user space address *user_ptr* and store
4506 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4508 * 0 on success, or a negative error in case of failure.
4510 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4512 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4513 * using *ptr*->type_id. This value should specify the type
4514 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4515 * can be used to look up vmlinux BTF type ids. Traversing the
4516 * data structure using BTF, the type information and values are
4517 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4518 * the pointer data is carried out to avoid kernel crashes during
4519 * operation. Smaller types can use string space on the stack;
4520 * larger programs can use map data to store the string
4523 * The string can be subsequently shared with userspace via
4524 * bpf_perf_event_output() or ring buffer interfaces.
4525 * bpf_trace_printk() is to be avoided as it places too small
4526 * a limit on string size to be useful.
4528 * *flags* is a combination of
4531 * no formatting around type information
4533 * no struct/union member names/types
4535 * show raw (unobfuscated) pointer values;
4536 * equivalent to printk specifier %px.
4538 * show zero-valued struct/union members; they
4539 * are not displayed by default
4542 * The number of bytes that were written (or would have been
4543 * written if output had to be truncated due to string size),
4544 * or a negative error in cases of failure.
4546 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4548 * Use BTF to write to seq_write a string representation of
4549 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4550 * *flags* are identical to those used for bpf_snprintf_btf.
4552 * 0 on success or a negative error in case of failure.
4554 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4556 * See **bpf_get_cgroup_classid**\ () for the main description.
4557 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4558 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4559 * associated socket instead of the current process.
4561 * The id is returned or 0 in case the id could not be retrieved.
4563 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4565 * Redirect the packet to another net device of index *ifindex*
4566 * and fill in L2 addresses from neighboring subsystem. This helper
4567 * is somewhat similar to **bpf_redirect**\ (), except that it
4568 * populates L2 addresses as well, meaning, internally, the helper
4569 * relies on the neighbor lookup for the L2 address of the nexthop.
4571 * The helper will perform a FIB lookup based on the skb's
4572 * networking header to get the address of the next hop, unless
4573 * this is supplied by the caller in the *params* argument. The
4574 * *plen* argument indicates the len of *params* and should be set
4575 * to 0 if *params* is NULL.
4577 * The *flags* argument is reserved and must be 0. The helper is
4578 * currently only supported for tc BPF program types, and enabled
4579 * for IPv4 and IPv6 protocols.
4581 * The helper returns **TC_ACT_REDIRECT** on success or
4582 * **TC_ACT_SHOT** on error.
4584 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4586 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4587 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4588 * extern variable decorated with '__ksym'. For ksym, there is a
4589 * global var (either static or global) defined of the same name
4590 * in the kernel. The ksym is percpu if the global var is percpu.
4591 * The returned pointer points to the global percpu var on *cpu*.
4593 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4594 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4595 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4596 * bpf_per_cpu_ptr() must check the returned value.
4598 * A pointer pointing to the kernel percpu variable on *cpu*, or
4599 * NULL, if *cpu* is invalid.
4601 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4603 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4604 * pointer to the percpu kernel variable on this cpu. See the
4605 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4607 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4608 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4609 * never return NULL.
4611 * A pointer pointing to the kernel percpu variable on this cpu.
4613 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4615 * Redirect the packet to another net device of index *ifindex*.
4616 * This helper is somewhat similar to **bpf_redirect**\ (), except
4617 * that the redirection happens to the *ifindex*' peer device and
4618 * the netns switch takes place from ingress to ingress without
4619 * going through the CPU's backlog queue.
4621 * The *flags* argument is reserved and must be 0. The helper is
4622 * currently only supported for tc BPF program types at the ingress
4623 * hook and for veth device types. The peer device must reside in a
4624 * different network namespace.
4626 * The helper returns **TC_ACT_REDIRECT** on success or
4627 * **TC_ACT_SHOT** on error.
4629 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4631 * Get a bpf_local_storage from the *task*.
4633 * Logically, it could be thought of as getting the value from
4634 * a *map* with *task* as the **key**. From this
4635 * perspective, the usage is not much different from
4636 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4637 * helper enforces the key must be an task_struct and the map must also
4638 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4640 * Underneath, the value is stored locally at *task* instead of
4641 * the *map*. The *map* is used as the bpf-local-storage
4642 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4643 * searched against all bpf_local_storage residing at *task*.
4645 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4646 * used such that a new bpf_local_storage will be
4647 * created if one does not exist. *value* can be used
4648 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4649 * the initial value of a bpf_local_storage. If *value* is
4650 * **NULL**, the new bpf_local_storage will be zero initialized.
4652 * A bpf_local_storage pointer is returned on success.
4654 * **NULL** if not found or there was an error in adding
4655 * a new bpf_local_storage.
4657 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4659 * Delete a bpf_local_storage from a *task*.
4663 * **-ENOENT** if the bpf_local_storage cannot be found.
4665 * struct task_struct *bpf_get_current_task_btf(void)
4667 * Return a BTF pointer to the "current" task.
4668 * This pointer can also be used in helpers that accept an
4669 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
4671 * Pointer to the current task.
4673 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4675 * Set or clear certain options on *bprm*:
4677 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4678 * which sets the **AT_SECURE** auxv for glibc. The bit
4679 * is cleared if the flag is not specified.
4681 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
4683 * u64 bpf_ktime_get_coarse_ns(void)
4685 * Return a coarse-grained version of the time elapsed since
4686 * system boot, in nanoseconds. Does not include time the system
4689 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4693 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4695 * Returns the stored IMA hash of the *inode* (if it's avaialable).
4696 * If the hash is larger than *size*, then only *size*
4697 * bytes will be copied to *dst*
4699 * The **hash_algo** is returned on success,
4700 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4701 * invalid arguments are passed.
4703 * struct socket *bpf_sock_from_file(struct file *file)
4705 * If the given file represents a socket, returns the associated
4708 * A pointer to a struct socket on success or NULL if the file is
4711 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4713 * Check packet size against exceeding MTU of net device (based
4714 * on *ifindex*). This helper will likely be used in combination
4715 * with helpers that adjust/change the packet size.
4717 * The argument *len_diff* can be used for querying with a planned
4718 * size change. This allows to check MTU prior to changing packet
4719 * ctx. Providing an *len_diff* adjustment that is larger than the
4720 * actual packet size (resulting in negative packet size) will in
4721 * principle not exceed the MTU, why it is not considered a
4722 * failure. Other BPF-helpers are needed for performing the
4723 * planned size change, why the responsability for catch a negative
4724 * packet size belong in those helpers.
4726 * Specifying *ifindex* zero means the MTU check is performed
4727 * against the current net device. This is practical if this isn't
4728 * used prior to redirect.
4730 * On input *mtu_len* must be a valid pointer, else verifier will
4731 * reject BPF program. If the value *mtu_len* is initialized to
4732 * zero then the ctx packet size is use. When value *mtu_len* is
4733 * provided as input this specify the L3 length that the MTU check
4734 * is done against. Remember XDP and TC length operate at L2, but
4735 * this value is L3 as this correlate to MTU and IP-header tot_len
4736 * values which are L3 (similar behavior as bpf_fib_lookup).
4738 * The Linux kernel route table can configure MTUs on a more
4739 * specific per route level, which is not provided by this helper.
4740 * For route level MTU checks use the **bpf_fib_lookup**\ ()
4743 * *ctx* is either **struct xdp_md** for XDP programs or
4744 * **struct sk_buff** for tc cls_act programs.
4746 * The *flags* argument can be a combination of one or more of the
4749 * **BPF_MTU_CHK_SEGS**
4750 * This flag will only works for *ctx* **struct sk_buff**.
4751 * If packet context contains extra packet segment buffers
4752 * (often knows as GSO skb), then MTU check is harder to
4753 * check at this point, because in transmit path it is
4754 * possible for the skb packet to get re-segmented
4755 * (depending on net device features). This could still be
4756 * a MTU violation, so this flag enables performing MTU
4757 * check against segments, with a different violation
4758 * return code to tell it apart. Check cannot use len_diff.
4760 * On return *mtu_len* pointer contains the MTU value of the net
4761 * device. Remember the net device configured MTU is the L3 size,
4762 * which is returned here and XDP and TC length operate at L2.
4763 * Helper take this into account for you, but remember when using
4764 * MTU value in your BPF-code.
4767 * * 0 on success, and populate MTU value in *mtu_len* pointer.
4769 * * < 0 if any input argument is invalid (*mtu_len* not updated)
4771 * MTU violations return positive values, but also populate MTU
4772 * value in *mtu_len* pointer, as this can be needed for
4773 * implementing PMTU handing:
4775 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
4776 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4778 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4780 * For each element in **map**, call **callback_fn** function with
4781 * **map**, **callback_ctx** and other map-specific parameters.
4782 * The **callback_fn** should be a static function and
4783 * the **callback_ctx** should be a pointer to the stack.
4784 * The **flags** is used to control certain aspects of the helper.
4785 * Currently, the **flags** must be 0.
4787 * The following are a list of supported map types and their
4788 * respective expected callback signatures:
4790 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4791 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4792 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4794 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4796 * For per_cpu maps, the map_value is the value on the cpu where the
4797 * bpf_prog is running.
4799 * If **callback_fn** return 0, the helper will continue to the next
4800 * element. If return value is 1, the helper will skip the rest of
4801 * elements and return. Other return values are not used now.
4804 * The number of traversed map elements for success, **-EINVAL** for
4805 * invalid **flags**.
4807 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4809 * Outputs a string into the **str** buffer of size **str_size**
4810 * based on a format string stored in a read-only map pointed by
4813 * Each format specifier in **fmt** corresponds to one u64 element
4814 * in the **data** array. For strings and pointers where pointees
4815 * are accessed, only the pointer values are stored in the *data*
4816 * array. The *data_len* is the size of *data* in bytes - must be
4819 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4820 * memory. Reading kernel memory may fail due to either invalid
4821 * address or valid address but requiring a major memory fault. If
4822 * reading kernel memory fails, the string for **%s** will be an
4823 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4824 * Not returning error to bpf program is consistent with what
4825 * **bpf_trace_printk**\ () does for now.
4828 * The strictly positive length of the formatted string, including
4829 * the trailing zero character. If the return value is greater than
4830 * **str_size**, **str** contains a truncated string, guaranteed to
4831 * be zero-terminated except when **str_size** is 0.
4833 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4835 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
4837 * Execute bpf syscall with given arguments.
4841 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
4843 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
4845 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
4847 * long bpf_sys_close(u32 fd)
4849 * Execute close syscall for given FD.
4853 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
4855 * Initialize the timer.
4856 * First 4 bits of *flags* specify clockid.
4857 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
4858 * All other bits of *flags* are reserved.
4859 * The verifier will reject the program if *timer* is not from
4863 * **-EBUSY** if *timer* is already initialized.
4864 * **-EINVAL** if invalid *flags* are passed.
4865 * **-EPERM** if *timer* is in a map that doesn't have any user references.
4866 * The user space should either hold a file descriptor to a map with timers
4867 * or pin such map in bpffs. When map is unpinned or file descriptor is
4868 * closed all timers in the map will be cancelled and freed.
4870 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
4872 * Configure the timer to call *callback_fn* static function.
4875 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4876 * **-EPERM** if *timer* is in a map that doesn't have any user references.
4877 * The user space should either hold a file descriptor to a map with timers
4878 * or pin such map in bpffs. When map is unpinned or file descriptor is
4879 * closed all timers in the map will be cancelled and freed.
4881 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
4883 * Set timer expiration N nanoseconds from the current time. The
4884 * configured callback will be invoked in soft irq context on some cpu
4885 * and will not repeat unless another bpf_timer_start() is made.
4886 * In such case the next invocation can migrate to a different cpu.
4887 * Since struct bpf_timer is a field inside map element the map
4888 * owns the timer. The bpf_timer_set_callback() will increment refcnt
4889 * of BPF program to make sure that callback_fn code stays valid.
4890 * When user space reference to a map reaches zero all timers
4891 * in a map are cancelled and corresponding program's refcnts are
4892 * decremented. This is done to make sure that Ctrl-C of a user
4893 * process doesn't leave any timers running. If map is pinned in
4894 * bpffs the callback_fn can re-arm itself indefinitely.
4895 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
4896 * cancel and free the timer in the given map element.
4897 * The map can contain timers that invoke callback_fn-s from different
4898 * programs. The same callback_fn can serve different timers from
4899 * different maps if key/value layout matches across maps.
4900 * Every bpf_timer_set_callback() can have different callback_fn.
4904 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
4905 * or invalid *flags* are passed.
4907 * long bpf_timer_cancel(struct bpf_timer *timer)
4909 * Cancel the timer and wait for callback_fn to finish if it was running.
4911 * 0 if the timer was not active.
4912 * 1 if the timer was active.
4913 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4914 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
4915 * own timer which would have led to a deadlock otherwise.
4917 * u64 bpf_get_func_ip(void *ctx)
4919 * Get address of the traced function (for tracing and kprobe programs).
4921 * Address of the traced function.
4923 * u64 bpf_get_attach_cookie(void *ctx)
4925 * Get bpf_cookie value provided (optionally) during the program
4926 * attachment. It might be different for each individual
4927 * attachment, even if BPF program itself is the same.
4928 * Expects BPF program context *ctx* as a first argument.
4930 * Supported for the following program types:
4935 * Value specified by user at BPF link creation/attachment time
4936 * or 0, if it was not specified.
4938 * long bpf_task_pt_regs(struct task_struct *task)
4940 * Get the struct pt_regs associated with **task**.
4942 * A pointer to struct pt_regs.
4944 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
4946 * Get branch trace from hardware engines like Intel LBR. The
4947 * hardware engine is stopped shortly after the helper is
4948 * called. Therefore, the user need to filter branch entries
4949 * based on the actual use case. To capture branch trace
4950 * before the trigger point of the BPF program, the helper
4951 * should be called at the beginning of the BPF program.
4953 * The data is stored as struct perf_branch_entry into output
4954 * buffer *entries*. *size* is the size of *entries* in bytes.
4955 * *flags* is reserved for now and must be zero.
4958 * On success, number of bytes written to *buf*. On error, a
4961 * **-EINVAL** if *flags* is not zero.
4963 * **-ENOENT** if architecture does not support branch records.
4965 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4967 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
4968 * to format and can handle more format args as a result.
4970 * Arguments are to be used as in **bpf_seq_printf**\ () helper.
4972 * The number of bytes written to the buffer, or a negative error
4973 * in case of failure.
4975 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
4977 * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
4979 * *sk* if casting is valid, or **NULL** otherwise.
4981 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
4983 * Get the address of a kernel symbol, returned in *res*. *res* is
4984 * set to 0 if the symbol is not found.
4986 * On success, zero. On error, a negative value.
4988 * **-EINVAL** if *flags* is not zero.
4990 * **-EINVAL** if string *name* is not the same size as *name_sz*.
4992 * **-ENOENT** if symbol is not found.
4994 * **-EPERM** if caller does not have permission to obtain kernel address.
4996 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
4998 * Find vma of *task* that contains *addr*, call *callback_fn*
4999 * function with *task*, *vma*, and *callback_ctx*.
5000 * The *callback_fn* should be a static function and
5001 * the *callback_ctx* should be a pointer to the stack.
5002 * The *flags* is used to control certain aspects of the helper.
5003 * Currently, the *flags* must be 0.
5005 * The expected callback signature is
5007 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5011 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5012 * **-EBUSY** if failed to try lock mmap_lock.
5013 * **-EINVAL** for invalid **flags**.
5015 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5017 * For **nr_loops**, call **callback_fn** function
5018 * with **callback_ctx** as the context parameter.
5019 * The **callback_fn** should be a static function and
5020 * the **callback_ctx** should be a pointer to the stack.
5021 * The **flags** is used to control certain aspects of the helper.
5022 * Currently, the **flags** must be 0. Currently, nr_loops is
5023 * limited to 1 << 23 (~8 million) loops.
5025 * long (\*callback_fn)(u32 index, void \*ctx);
5027 * where **index** is the current index in the loop. The index
5030 * If **callback_fn** returns 0, the helper will continue to the next
5031 * loop. If return value is 1, the helper will skip the rest of
5032 * the loops and return. Other return values are not used now,
5033 * and will be rejected by the verifier.
5036 * The number of loops performed, **-EINVAL** for invalid **flags**,
5037 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5039 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5041 * Do strncmp() between **s1** and **s2**. **s1** doesn't need
5042 * to be null-terminated and **s1_sz** is the maximum storage
5043 * size of **s1**. **s2** must be a read-only string.
5045 * An integer less than, equal to, or greater than zero
5046 * if the first **s1_sz** bytes of **s1** is found to be
5047 * less than, to match, or be greater than **s2**.
5049 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5051 * Get **n**-th argument (zero based) of the traced function (for tracing programs)
5052 * returned in **value**.
5056 * **-EINVAL** if n >= arguments count of traced function.
5058 * long bpf_get_func_ret(void *ctx, u64 *value)
5060 * Get return value of the traced function (for tracing programs)
5065 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5067 * long bpf_get_func_arg_cnt(void *ctx)
5069 * Get number of arguments of the traced function (for tracing programs).
5072 * The number of arguments of the traced function.
5074 * int bpf_get_retval(void)
5076 * Get the syscall's return value that will be returned to userspace.
5078 * This helper is currently supported by cgroup programs only.
5080 * The syscall's return value.
5082 * int bpf_set_retval(int retval)
5084 * Set the syscall's return value that will be returned to userspace.
5086 * This helper is currently supported by cgroup programs only.
5088 * 0 on success, or a negative error in case of failure.
5090 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5092 * Get the total size of a given xdp buff (linear and paged area)
5094 * The total size of a given xdp buffer.
5096 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5098 * This helper is provided as an easy way to load data from a
5099 * xdp buffer. It can be used to load *len* bytes from *offset* from
5100 * the frame associated to *xdp_md*, into the buffer pointed by
5103 * 0 on success, or a negative error in case of failure.
5105 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5107 * Store *len* bytes from buffer *buf* into the frame
5108 * associated to *xdp_md*, at *offset*.
5110 * 0 on success, or a negative error in case of failure.
5112 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5114 * Read *size* bytes from user space address *user_ptr* in *tsk*'s
5115 * address space, and stores the data in *dst*. *flags* is not
5116 * used yet and is provided for future extensibility. This helper
5117 * can only be used by sleepable programs.
5119 * 0 on success, or a negative error in case of failure. On error
5120 * *dst* buffer is zeroed out.
5122 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5124 * Change the __sk_buff->tstamp_type to *tstamp_type*
5125 * and set *tstamp* to the __sk_buff->tstamp together.
5127 * If there is no need to change the __sk_buff->tstamp_type,
5128 * the tstamp value can be directly written to __sk_buff->tstamp
5131 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5132 * will be kept during bpf_redirect_*(). A non zero
5133 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5136 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5137 * with a zero *tstamp*.
5139 * Only IPv4 and IPv6 skb->protocol are supported.
5141 * This function is most useful when it needs to set a
5142 * mono delivery time to __sk_buff->tstamp and then
5143 * bpf_redirect_*() to the egress of an iface. For example,
5144 * changing the (rcv) timestamp in __sk_buff->tstamp at
5145 * ingress to a mono delivery time and then bpf_redirect_*()
5146 * to sch_fq@phy-dev.
5149 * **-EINVAL** for invalid input
5150 * **-EOPNOTSUPP** for unsupported protocol
5152 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5154 * Returns a calculated IMA hash of the *file*.
5155 * If the hash is larger than *size*, then only *size*
5156 * bytes will be copied to *dst*
5158 * The **hash_algo** is returned on success,
5159 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
5160 * invalid arguments are passed.
5162 * void *bpf_kptr_xchg(void *map_value, void *ptr)
5164 * Exchange kptr at pointer *map_value* with *ptr*, and return the
5165 * old value. *ptr* can be NULL, otherwise it must be a referenced
5166 * pointer which will be released when this helper is called.
5168 * The old value of kptr (which can be NULL). The returned pointer
5169 * if not NULL, is a reference which must be released using its
5170 * corresponding release function, or moved into a BPF map before
5173 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5175 * Perform a lookup in *percpu map* for an entry associated to
5178 * Map value associated to *key* on *cpu*, or **NULL** if no entry
5179 * was found or *cpu* is invalid.
5181 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5183 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5185 * *sk* if casting is valid, or **NULL** otherwise.
5187 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5189 * Get a dynptr to local memory *data*.
5191 * *data* must be a ptr to a map value.
5192 * The maximum *size* supported is DYNPTR_MAX_SIZE.
5193 * *flags* is currently unused.
5195 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5196 * -EINVAL if flags is not 0.
5198 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5200 * Reserve *size* bytes of payload in a ring buffer *ringbuf*
5201 * through the dynptr interface. *flags* must be 0.
5203 * Please note that a corresponding bpf_ringbuf_submit_dynptr or
5204 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5205 * reservation fails. This is enforced by the verifier.
5207 * 0 on success, or a negative error in case of failure.
5209 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5211 * Submit reserved ring buffer sample, pointed to by *data*,
5212 * through the dynptr interface. This is a no-op if the dynptr is
5215 * For more information on *flags*, please see
5216 * 'bpf_ringbuf_submit'.
5218 * Nothing. Always succeeds.
5220 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5222 * Discard reserved ring buffer sample through the dynptr
5223 * interface. This is a no-op if the dynptr is invalid/null.
5225 * For more information on *flags*, please see
5226 * 'bpf_ringbuf_discard'.
5228 * Nothing. Always succeeds.
5230 * long bpf_dynptr_read(void *dst, u32 len, struct bpf_dynptr *src, u32 offset, u64 flags)
5232 * Read *len* bytes from *src* into *dst*, starting from *offset*
5234 * *flags* is currently unused.
5236 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5237 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5240 * long bpf_dynptr_write(struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5242 * Write *len* bytes from *src* into *dst*, starting from *offset*
5244 * *flags* is currently unused.
5246 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5247 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5248 * is a read-only dynptr or if *flags* is not 0.
5250 * void *bpf_dynptr_data(struct bpf_dynptr *ptr, u32 offset, u32 len)
5252 * Get a pointer to the underlying dynptr data.
5254 * *len* must be a statically known value. The returned data slice
5255 * is invalidated whenever the dynptr is invalidated.
5257 * Pointer to the underlying dynptr data, NULL if the dynptr is
5258 * read-only, if the dynptr is invalid, or if the offset and length
5261 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5263 * Try to issue a SYN cookie for the packet with corresponding
5264 * IPv4/TCP headers, *iph* and *th*, without depending on a
5267 * *iph* points to the IPv4 header.
5269 * *th* points to the start of the TCP header, while *th_len*
5270 * contains the length of the TCP header (at least
5271 * **sizeof**\ (**struct tcphdr**)).
5273 * On success, lower 32 bits hold the generated SYN cookie in
5274 * followed by 16 bits which hold the MSS value for that cookie,
5275 * and the top 16 bits are unused.
5277 * On failure, the returned value is one of the following:
5279 * **-EINVAL** if *th_len* is invalid.
5281 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5283 * Try to issue a SYN cookie for the packet with corresponding
5284 * IPv6/TCP headers, *iph* and *th*, without depending on a
5287 * *iph* points to the IPv6 header.
5289 * *th* points to the start of the TCP header, while *th_len*
5290 * contains the length of the TCP header (at least
5291 * **sizeof**\ (**struct tcphdr**)).
5293 * On success, lower 32 bits hold the generated SYN cookie in
5294 * followed by 16 bits which hold the MSS value for that cookie,
5295 * and the top 16 bits are unused.
5297 * On failure, the returned value is one of the following:
5299 * **-EINVAL** if *th_len* is invalid.
5301 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5303 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5305 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5306 * without depending on a listening socket.
5308 * *iph* points to the IPv4 header.
5310 * *th* points to the TCP header.
5312 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5314 * On failure, the returned value is one of the following:
5316 * **-EACCES** if the SYN cookie is not valid.
5318 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5320 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5321 * without depending on a listening socket.
5323 * *iph* points to the IPv6 header.
5325 * *th* points to the TCP header.
5327 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5329 * On failure, the returned value is one of the following:
5331 * **-EACCES** if the SYN cookie is not valid.
5333 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5335 #define __BPF_FUNC_MAPPER(FN) \
5337 FN(map_lookup_elem), \
5338 FN(map_update_elem), \
5339 FN(map_delete_elem), \
5343 FN(get_prandom_u32), \
5344 FN(get_smp_processor_id), \
5345 FN(skb_store_bytes), \
5346 FN(l3_csum_replace), \
5347 FN(l4_csum_replace), \
5349 FN(clone_redirect), \
5350 FN(get_current_pid_tgid), \
5351 FN(get_current_uid_gid), \
5352 FN(get_current_comm), \
5353 FN(get_cgroup_classid), \
5354 FN(skb_vlan_push), \
5356 FN(skb_get_tunnel_key), \
5357 FN(skb_set_tunnel_key), \
5358 FN(perf_event_read), \
5360 FN(get_route_realm), \
5361 FN(perf_event_output), \
5362 FN(skb_load_bytes), \
5365 FN(skb_get_tunnel_opt), \
5366 FN(skb_set_tunnel_opt), \
5367 FN(skb_change_proto), \
5368 FN(skb_change_type), \
5369 FN(skb_under_cgroup), \
5370 FN(get_hash_recalc), \
5371 FN(get_current_task), \
5372 FN(probe_write_user), \
5373 FN(current_task_under_cgroup), \
5374 FN(skb_change_tail), \
5375 FN(skb_pull_data), \
5377 FN(set_hash_invalid), \
5378 FN(get_numa_node_id), \
5379 FN(skb_change_head), \
5380 FN(xdp_adjust_head), \
5381 FN(probe_read_str), \
5382 FN(get_socket_cookie), \
5383 FN(get_socket_uid), \
5386 FN(skb_adjust_room), \
5388 FN(sk_redirect_map), \
5389 FN(sock_map_update), \
5390 FN(xdp_adjust_meta), \
5391 FN(perf_event_read_value), \
5392 FN(perf_prog_read_value), \
5394 FN(override_return), \
5395 FN(sock_ops_cb_flags_set), \
5396 FN(msg_redirect_map), \
5397 FN(msg_apply_bytes), \
5398 FN(msg_cork_bytes), \
5399 FN(msg_pull_data), \
5401 FN(xdp_adjust_tail), \
5402 FN(skb_get_xfrm_state), \
5404 FN(skb_load_bytes_relative), \
5406 FN(sock_hash_update), \
5407 FN(msg_redirect_hash), \
5408 FN(sk_redirect_hash), \
5409 FN(lwt_push_encap), \
5410 FN(lwt_seg6_store_bytes), \
5411 FN(lwt_seg6_adjust_srh), \
5412 FN(lwt_seg6_action), \
5415 FN(skb_cgroup_id), \
5416 FN(get_current_cgroup_id), \
5417 FN(get_local_storage), \
5418 FN(sk_select_reuseport), \
5419 FN(skb_ancestor_cgroup_id), \
5420 FN(sk_lookup_tcp), \
5421 FN(sk_lookup_udp), \
5423 FN(map_push_elem), \
5425 FN(map_peek_elem), \
5426 FN(msg_push_data), \
5428 FN(rc_pointer_rel), \
5433 FN(skb_ecn_set_ce), \
5434 FN(get_listener_sock), \
5435 FN(skc_lookup_tcp), \
5436 FN(tcp_check_syncookie), \
5437 FN(sysctl_get_name), \
5438 FN(sysctl_get_current_value), \
5439 FN(sysctl_get_new_value), \
5440 FN(sysctl_set_new_value), \
5443 FN(sk_storage_get), \
5444 FN(sk_storage_delete), \
5446 FN(tcp_gen_syncookie), \
5448 FN(probe_read_user), \
5449 FN(probe_read_kernel), \
5450 FN(probe_read_user_str), \
5451 FN(probe_read_kernel_str), \
5453 FN(send_signal_thread), \
5455 FN(read_branch_records), \
5456 FN(get_ns_current_pid_tgid), \
5458 FN(get_netns_cookie), \
5459 FN(get_current_ancestor_cgroup_id), \
5461 FN(ktime_get_boot_ns), \
5465 FN(sk_ancestor_cgroup_id), \
5466 FN(ringbuf_output), \
5467 FN(ringbuf_reserve), \
5468 FN(ringbuf_submit), \
5469 FN(ringbuf_discard), \
5470 FN(ringbuf_query), \
5472 FN(skc_to_tcp6_sock), \
5473 FN(skc_to_tcp_sock), \
5474 FN(skc_to_tcp_timewait_sock), \
5475 FN(skc_to_tcp_request_sock), \
5476 FN(skc_to_udp6_sock), \
5477 FN(get_task_stack), \
5479 FN(store_hdr_opt), \
5480 FN(reserve_hdr_opt), \
5481 FN(inode_storage_get), \
5482 FN(inode_storage_delete), \
5484 FN(copy_from_user), \
5486 FN(seq_printf_btf), \
5487 FN(skb_cgroup_classid), \
5488 FN(redirect_neigh), \
5491 FN(redirect_peer), \
5492 FN(task_storage_get), \
5493 FN(task_storage_delete), \
5494 FN(get_current_task_btf), \
5495 FN(bprm_opts_set), \
5496 FN(ktime_get_coarse_ns), \
5497 FN(ima_inode_hash), \
5498 FN(sock_from_file), \
5500 FN(for_each_map_elem), \
5503 FN(btf_find_by_name_kind), \
5506 FN(timer_set_callback), \
5510 FN(get_attach_cookie), \
5512 FN(get_branch_snapshot), \
5513 FN(trace_vprintk), \
5514 FN(skc_to_unix_sock), \
5515 FN(kallsyms_lookup_name), \
5521 FN(get_func_arg_cnt), \
5524 FN(xdp_get_buff_len), \
5525 FN(xdp_load_bytes), \
5526 FN(xdp_store_bytes), \
5527 FN(copy_from_user_task), \
5528 FN(skb_set_tstamp), \
5529 FN(ima_file_hash), \
5531 FN(map_lookup_percpu_elem), \
5532 FN(skc_to_mptcp_sock), \
5533 FN(dynptr_from_mem), \
5534 FN(ringbuf_reserve_dynptr), \
5535 FN(ringbuf_submit_dynptr), \
5536 FN(ringbuf_discard_dynptr), \
5540 FN(tcp_raw_gen_syncookie_ipv4), \
5541 FN(tcp_raw_gen_syncookie_ipv6), \
5542 FN(tcp_raw_check_syncookie_ipv4), \
5543 FN(tcp_raw_check_syncookie_ipv6), \
5546 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5547 * function eBPF program intends to call
5549 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
5551 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5554 #undef __BPF_ENUM_FN
5556 /* All flags used by eBPF helper functions, placed here. */
5558 /* BPF_FUNC_skb_store_bytes flags. */
5560 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
5561 BPF_F_INVALIDATE_HASH = (1ULL << 1),
5564 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5565 * First 4 bits are for passing the header field size.
5568 BPF_F_HDR_FIELD_MASK = 0xfULL,
5571 /* BPF_FUNC_l4_csum_replace flags. */
5573 BPF_F_PSEUDO_HDR = (1ULL << 4),
5574 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
5575 BPF_F_MARK_ENFORCE = (1ULL << 6),
5578 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5580 BPF_F_INGRESS = (1ULL << 0),
5583 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5585 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
5588 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5590 BPF_F_SKIP_FIELD_MASK = 0xffULL,
5591 BPF_F_USER_STACK = (1ULL << 8),
5592 /* flags used by BPF_FUNC_get_stackid only. */
5593 BPF_F_FAST_STACK_CMP = (1ULL << 9),
5594 BPF_F_REUSE_STACKID = (1ULL << 10),
5595 /* flags used by BPF_FUNC_get_stack only. */
5596 BPF_F_USER_BUILD_ID = (1ULL << 11),
5599 /* BPF_FUNC_skb_set_tunnel_key flags. */
5601 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
5602 BPF_F_DONT_FRAGMENT = (1ULL << 2),
5603 BPF_F_SEQ_NUMBER = (1ULL << 3),
5606 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5607 * BPF_FUNC_perf_event_read_value flags.
5610 BPF_F_INDEX_MASK = 0xffffffffULL,
5611 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
5612 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5613 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
5616 /* Current network namespace */
5618 BPF_F_CURRENT_NETNS = (-1L),
5621 /* BPF_FUNC_csum_level level values. */
5623 BPF_CSUM_LEVEL_QUERY,
5626 BPF_CSUM_LEVEL_RESET,
5629 /* BPF_FUNC_skb_adjust_room flags. */
5631 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
5632 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
5633 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
5634 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
5635 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
5636 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
5637 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
5641 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
5642 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
5645 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
5646 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5647 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5649 /* BPF_FUNC_sysctl_get_name flags. */
5651 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
5654 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5656 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
5657 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5658 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5660 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
5663 /* BPF_FUNC_read_branch_records flags. */
5665 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
5668 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5669 * BPF_FUNC_bpf_ringbuf_output flags.
5672 BPF_RB_NO_WAKEUP = (1ULL << 0),
5673 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
5676 /* BPF_FUNC_bpf_ringbuf_query flags */
5678 BPF_RB_AVAIL_DATA = 0,
5679 BPF_RB_RING_SIZE = 1,
5680 BPF_RB_CONS_POS = 2,
5681 BPF_RB_PROD_POS = 3,
5684 /* BPF ring buffer constants */
5686 BPF_RINGBUF_BUSY_BIT = (1U << 31),
5687 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
5688 BPF_RINGBUF_HDR_SZ = 8,
5691 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5693 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
5694 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
5697 /* Mode for BPF_FUNC_skb_adjust_room helper. */
5698 enum bpf_adj_room_mode {
5703 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5704 enum bpf_hdr_start_off {
5709 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5710 enum bpf_lwt_encap_mode {
5712 BPF_LWT_ENCAP_SEG6_INLINE,
5716 /* Flags for bpf_bprm_opts_set helper */
5718 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
5721 /* Flags for bpf_redirect_map helper */
5723 BPF_F_BROADCAST = (1ULL << 3),
5724 BPF_F_EXCLUDE_INGRESS = (1ULL << 4),
5727 #define __bpf_md_ptr(type, name) \
5731 } __attribute__((aligned(8)))
5734 BPF_SKB_TSTAMP_UNSPEC,
5735 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */
5736 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
5737 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
5738 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
5742 /* user accessible mirror of in-kernel sk_buff.
5743 * new fields can only be added to the end of this structure
5749 __u32 queue_mapping;
5755 __u32 ingress_ifindex;
5765 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5767 __u32 remote_ip4; /* Stored in network byte order */
5768 __u32 local_ip4; /* Stored in network byte order */
5769 __u32 remote_ip6[4]; /* Stored in network byte order */
5770 __u32 local_ip6[4]; /* Stored in network byte order */
5771 __u32 remote_port; /* Stored in network byte order */
5772 __u32 local_port; /* stored in host byte order */
5776 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5780 __bpf_md_ptr(struct bpf_sock *, sk);
5783 __u32 :24; /* Padding, future use. */
5787 struct bpf_tunnel_key {
5791 __u32 remote_ipv6[4];
5795 __u16 tunnel_ext; /* Padding, future use. */
5799 __u32 local_ipv6[4];
5803 /* user accessible mirror of in-kernel xfrm_state.
5804 * new fields can only be added to the end of this structure
5806 struct bpf_xfrm_state {
5808 __u32 spi; /* Stored in network byte order */
5810 __u16 ext; /* Padding, future use. */
5812 __u32 remote_ipv4; /* Stored in network byte order */
5813 __u32 remote_ipv6[4]; /* Stored in network byte order */
5817 /* Generic BPF return codes which all BPF program types may support.
5818 * The values are binary compatible with their TC_ACT_* counter-part to
5819 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5822 * XDP is handled seprately, see XDP_*.
5830 /* >127 are reserved for prog type specific return codes.
5832 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5833 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5834 * changed and should be routed based on its new L3 header.
5835 * (This is an L3 redirect, as opposed to L2 redirect
5836 * represented by BPF_REDIRECT above).
5838 BPF_LWT_REROUTE = 128,
5848 /* IP address also allows 1 and 2 bytes access */
5851 __u32 src_port; /* host byte order */
5852 __be16 dst_port; /* network byte order */
5853 __u16 :16; /* zero padding */
5857 __s32 rx_queue_mapping;
5860 struct bpf_tcp_sock {
5861 __u32 snd_cwnd; /* Sending congestion window */
5862 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
5864 __u32 snd_ssthresh; /* Slow start size threshold */
5865 __u32 rcv_nxt; /* What we want to receive next */
5866 __u32 snd_nxt; /* Next sequence we send */
5867 __u32 snd_una; /* First byte we want an ack for */
5868 __u32 mss_cache; /* Cached effective mss, not including SACKS */
5869 __u32 ecn_flags; /* ECN status bits. */
5870 __u32 rate_delivered; /* saved rate sample: packets delivered */
5871 __u32 rate_interval_us; /* saved rate sample: time elapsed */
5872 __u32 packets_out; /* Packets which are "in flight" */
5873 __u32 retrans_out; /* Retransmitted packets out */
5874 __u32 total_retrans; /* Total retransmits for entire connection */
5875 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
5876 * total number of segments in.
5878 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
5879 * total number of data segments in.
5881 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
5882 * The total number of segments sent.
5884 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
5885 * total number of data segments sent.
5887 __u32 lost_out; /* Lost packets */
5888 __u32 sacked_out; /* SACK'd packets */
5889 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
5890 * sum(delta(rcv_nxt)), or how many bytes
5893 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
5894 * sum(delta(snd_una)), or how many bytes
5897 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
5898 * total number of DSACK blocks received
5900 __u32 delivered; /* Total data packets delivered incl. rexmits */
5901 __u32 delivered_ce; /* Like the above but only ECE marked packets */
5902 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
5905 struct bpf_sock_tuple {
5922 struct bpf_xdp_sock {
5926 #define XDP_PACKET_HEADROOM 256
5928 /* User return codes for XDP prog type.
5929 * A valid XDP program must return one of these defined values. All other
5930 * return codes are reserved for future use. Unknown return codes will
5931 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5941 /* user accessible metadata for XDP packet hook
5942 * new fields must be added to the end of this structure
5948 /* Below access go through struct xdp_rxq_info */
5949 __u32 ingress_ifindex; /* rxq->dev->ifindex */
5950 __u32 rx_queue_index; /* rxq->queue_index */
5952 __u32 egress_ifindex; /* txq->dev->ifindex */
5955 /* DEVMAP map-value layout
5957 * The struct data-layout of map-value is a configuration interface.
5958 * New members can only be added to the end of this structure.
5960 struct bpf_devmap_val {
5961 __u32 ifindex; /* device index */
5963 int fd; /* prog fd on map write */
5964 __u32 id; /* prog id on map read */
5968 /* CPUMAP map-value layout
5970 * The struct data-layout of map-value is a configuration interface.
5971 * New members can only be added to the end of this structure.
5973 struct bpf_cpumap_val {
5974 __u32 qsize; /* queue size to remote target CPU */
5976 int fd; /* prog fd on map write */
5977 __u32 id; /* prog id on map read */
5986 /* user accessible metadata for SK_MSG packet hook, new fields must
5987 * be added to the end of this structure
5990 __bpf_md_ptr(void *, data);
5991 __bpf_md_ptr(void *, data_end);
5994 __u32 remote_ip4; /* Stored in network byte order */
5995 __u32 local_ip4; /* Stored in network byte order */
5996 __u32 remote_ip6[4]; /* Stored in network byte order */
5997 __u32 local_ip6[4]; /* Stored in network byte order */
5998 __u32 remote_port; /* Stored in network byte order */
5999 __u32 local_port; /* stored in host byte order */
6000 __u32 size; /* Total size of sk_msg */
6002 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6005 struct sk_reuseport_md {
6007 * Start of directly accessible data. It begins from
6008 * the tcp/udp header.
6010 __bpf_md_ptr(void *, data);
6011 /* End of directly accessible data */
6012 __bpf_md_ptr(void *, data_end);
6014 * Total length of packet (starting from the tcp/udp header).
6015 * Note that the directly accessible bytes (data_end - data)
6016 * could be less than this "len". Those bytes could be
6017 * indirectly read by a helper "bpf_skb_load_bytes()".
6021 * Eth protocol in the mac header (network byte order). e.g.
6022 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6025 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6026 __u32 bind_inany; /* Is sock bound to an INANY address? */
6027 __u32 hash; /* A hash of the packet 4 tuples */
6028 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6029 * new incoming connection request (e.g. selecting a listen sk for
6030 * the received SYN in the TCP case). reuse->sk is one of the sk
6031 * in the reuseport group. The bpf prog can use reuse->sk to learn
6032 * the local listening ip/port without looking into the skb.
6034 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6035 * reuse->migrating_sk is the socket that needs to be migrated
6036 * to another listening socket. migrating_sk could be a fullsock
6037 * sk that is fully established or a reqsk that is in-the-middle
6038 * of 3-way handshake.
6040 __bpf_md_ptr(struct bpf_sock *, sk);
6041 __bpf_md_ptr(struct bpf_sock *, migrating_sk);
6044 #define BPF_TAG_SIZE 8
6046 struct bpf_prog_info {
6049 __u8 tag[BPF_TAG_SIZE];
6050 __u32 jited_prog_len;
6051 __u32 xlated_prog_len;
6052 __aligned_u64 jited_prog_insns;
6053 __aligned_u64 xlated_prog_insns;
6054 __u64 load_time; /* ns since boottime */
6055 __u32 created_by_uid;
6057 __aligned_u64 map_ids;
6058 char name[BPF_OBJ_NAME_LEN];
6060 __u32 gpl_compatible:1;
6061 __u32 :31; /* alignment pad */
6064 __u32 nr_jited_ksyms;
6065 __u32 nr_jited_func_lens;
6066 __aligned_u64 jited_ksyms;
6067 __aligned_u64 jited_func_lens;
6069 __u32 func_info_rec_size;
6070 __aligned_u64 func_info;
6073 __aligned_u64 line_info;
6074 __aligned_u64 jited_line_info;
6075 __u32 nr_jited_line_info;
6076 __u32 line_info_rec_size;
6077 __u32 jited_line_info_rec_size;
6079 __aligned_u64 prog_tags;
6082 __u64 recursion_misses;
6083 __u32 verified_insns;
6084 __u32 attach_btf_obj_id;
6085 __u32 attach_btf_id;
6086 } __attribute__((aligned(8)));
6088 struct bpf_map_info {
6095 char name[BPF_OBJ_NAME_LEN];
6097 __u32 btf_vmlinux_value_type_id;
6101 __u32 btf_key_type_id;
6102 __u32 btf_value_type_id;
6103 __u32 :32; /* alignment pad */
6105 } __attribute__((aligned(8)));
6107 struct bpf_btf_info {
6114 } __attribute__((aligned(8)));
6116 struct bpf_link_info {
6122 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6123 __u32 tp_name_len; /* in/out: tp_name buffer len */
6127 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6128 __u32 target_btf_id; /* BTF type id inside the object */
6135 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
6136 __u32 target_name_len; /* in/out: target_name buffer len */
6151 } __attribute__((aligned(8)));
6153 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6154 * by user and intended to be used by socket (e.g. to bind to, depends on
6157 struct bpf_sock_addr {
6158 __u32 user_family; /* Allows 4-byte read, but no write. */
6159 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6160 * Stored in network byte order.
6162 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6163 * Stored in network byte order.
6165 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
6166 * Stored in network byte order
6168 __u32 family; /* Allows 4-byte read, but no write */
6169 __u32 type; /* Allows 4-byte read, but no write */
6170 __u32 protocol; /* Allows 4-byte read, but no write */
6171 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6172 * Stored in network byte order.
6174 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6175 * Stored in network byte order.
6177 __bpf_md_ptr(struct bpf_sock *, sk);
6180 /* User bpf_sock_ops struct to access socket values and specify request ops
6181 * and their replies.
6182 * Some of this fields are in network (bigendian) byte order and may need
6183 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6184 * New fields can only be added at the end of this structure
6186 struct bpf_sock_ops {
6189 __u32 args[4]; /* Optionally passed to bpf program */
6190 __u32 reply; /* Returned by bpf program */
6191 __u32 replylong[4]; /* Optionally returned by bpf prog */
6194 __u32 remote_ip4; /* Stored in network byte order */
6195 __u32 local_ip4; /* Stored in network byte order */
6196 __u32 remote_ip6[4]; /* Stored in network byte order */
6197 __u32 local_ip6[4]; /* Stored in network byte order */
6198 __u32 remote_port; /* Stored in network byte order */
6199 __u32 local_port; /* stored in host byte order */
6200 __u32 is_fullsock; /* Some TCP fields are only valid if
6201 * there is a full socket. If not, the
6202 * fields read as zero.
6205 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
6206 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6215 __u32 rate_delivered;
6216 __u32 rate_interval_us;
6219 __u32 total_retrans;
6223 __u32 data_segs_out;
6227 __u64 bytes_received;
6229 __bpf_md_ptr(struct bpf_sock *, sk);
6230 /* [skb_data, skb_data_end) covers the whole TCP header.
6232 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6233 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
6234 * header has not been written.
6235 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6236 * been written so far.
6237 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
6239 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6242 * bpf_load_hdr_opt() can also be used to read a particular option.
6244 __bpf_md_ptr(void *, skb_data);
6245 __bpf_md_ptr(void *, skb_data_end);
6246 __u32 skb_len; /* The total length of a packet.
6247 * It includes the header, options,
6250 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
6251 * an easy way to check for tcp_flags
6252 * without parsing skb_data.
6254 * In particular, the skb_tcp_flags
6255 * will still be available in
6256 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6257 * the outgoing header has not
6262 /* Definitions for bpf_sock_ops_cb_flags */
6264 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
6265 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
6266 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
6267 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
6268 /* Call bpf for all received TCP headers. The bpf prog will be
6269 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6271 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6272 * for the header option related helpers that will be useful
6273 * to the bpf programs.
6275 * It could be used at the client/active side (i.e. connect() side)
6276 * when the server told it that the server was in syncookie
6277 * mode and required the active side to resend the bpf-written
6278 * options. The active side can keep writing the bpf-options until
6279 * it received a valid packet from the server side to confirm
6280 * the earlier packet (and options) has been received. The later
6281 * example patch is using it like this at the active side when the
6282 * server is in syncookie mode.
6284 * The bpf prog will usually turn this off in the common cases.
6286 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
6287 /* Call bpf when kernel has received a header option that
6288 * the kernel cannot handle. The bpf prog will be called under
6289 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6291 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6292 * for the header option related helpers that will be useful
6293 * to the bpf programs.
6295 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6296 /* Call bpf when the kernel is writing header options for the
6297 * outgoing packet. The bpf prog will first be called
6298 * to reserve space in a skb under
6299 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
6300 * the bpf prog will be called to write the header option(s)
6301 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6303 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6304 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6305 * related helpers that will be useful to the bpf programs.
6307 * The kernel gets its chance to reserve space and write
6308 * options first before the BPF program does.
6310 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6311 /* Mask of all currently supported cb flags */
6312 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
6315 /* List of known BPF sock_ops operators.
6316 * New entries can only be added at the end
6320 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
6321 * -1 if default value should be used
6323 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
6324 * window (in packets) or -1 if default
6325 * value should be used
6327 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
6328 * active connection is initialized
6330 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
6331 * active connection is
6334 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
6335 * passive connection is
6338 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
6341 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
6342 * based on the path and may be
6343 * dependent on the congestion control
6344 * algorithm. In general it indicates
6345 * a congestion threshold. RTTs above
6346 * this indicate congestion
6348 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
6349 * Arg1: value of icsk_retransmits
6350 * Arg2: value of icsk_rto
6351 * Arg3: whether RTO has expired
6353 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
6354 * Arg1: sequence number of 1st byte
6356 * Arg3: return value of
6357 * tcp_transmit_skb (0 => success)
6359 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
6363 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
6364 * socket transition to LISTEN state.
6366 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
6368 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
6369 * It will be called to handle
6370 * the packets received at
6371 * an already established
6374 * sock_ops->skb_data:
6375 * Referring to the received skb.
6376 * It covers the TCP header only.
6378 * bpf_load_hdr_opt() can also
6379 * be used to search for a
6380 * particular option.
6382 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
6383 * header option later in
6384 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6385 * Arg1: bool want_cookie. (in
6386 * writing SYNACK only)
6388 * sock_ops->skb_data:
6389 * Not available because no header has
6392 * sock_ops->skb_tcp_flags:
6393 * The tcp_flags of the
6394 * outgoing skb. (e.g. SYN, ACK, FIN).
6396 * bpf_reserve_hdr_opt() should
6397 * be used to reserve space.
6399 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
6400 * Arg1: bool want_cookie. (in
6401 * writing SYNACK only)
6403 * sock_ops->skb_data:
6404 * Referring to the outgoing skb.
6405 * It covers the TCP header
6406 * that has already been written
6407 * by the kernel and the
6408 * earlier bpf-progs.
6410 * sock_ops->skb_tcp_flags:
6411 * The tcp_flags of the outgoing
6412 * skb. (e.g. SYN, ACK, FIN).
6414 * bpf_store_hdr_opt() should
6415 * be used to write the
6418 * bpf_load_hdr_opt() can also
6419 * be used to search for a
6420 * particular option that
6421 * has already been written
6422 * by the kernel or the
6423 * earlier bpf-progs.
6427 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
6428 * changes between the TCP and BPF versions. Ideally this should never happen.
6429 * If it does, we need to add code to convert them before calling
6430 * the BPF sock_ops function.
6433 BPF_TCP_ESTABLISHED = 1,
6443 BPF_TCP_CLOSING, /* Now a valid state */
6444 BPF_TCP_NEW_SYN_RECV,
6446 BPF_TCP_MAX_STATES /* Leave at the end! */
6450 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
6451 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
6452 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
6453 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
6454 /* Copy the SYN pkt to optval
6456 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
6457 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6458 * to only getting from the saved_syn. It can either get the
6461 * 1. the just-received SYN packet (only available when writing the
6462 * SYNACK). It will be useful when it is not necessary to
6463 * save the SYN packet for latter use. It is also the only way
6464 * to get the SYN during syncookie mode because the syn
6465 * packet cannot be saved during syncookie.
6469 * 2. the earlier saved syn which was done by
6470 * bpf_setsockopt(TCP_SAVE_SYN).
6472 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6473 * SYN packet is obtained.
6475 * If the bpf-prog does not need the IP[46] header, the
6476 * bpf-prog can avoid parsing the IP header by using
6477 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
6478 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6480 * >0: Total number of bytes copied
6481 * -ENOSPC: Not enough space in optval. Only optlen number of
6483 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6484 * is not saved by setsockopt(TCP_SAVE_SYN).
6486 TCP_BPF_SYN = 1005, /* Copy the TCP header */
6487 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
6488 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
6492 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6495 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6496 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6499 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
6500 * total option spaces
6501 * required for an established
6502 * sk in order to calculate the
6503 * MSS. No skb is actually
6506 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
6507 * when sending a SYN.
6511 struct bpf_perf_event_value {
6518 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
6519 BPF_DEVCG_ACC_READ = (1ULL << 1),
6520 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
6524 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
6525 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
6528 struct bpf_cgroup_dev_ctx {
6529 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6535 struct bpf_raw_tracepoint_args {
6539 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
6540 * OUTPUT: Do lookup from egress perspective; default is ingress
6543 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
6544 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
6548 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
6549 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
6550 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
6551 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
6552 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
6553 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6554 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
6555 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
6556 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
6559 struct bpf_fib_lookup {
6560 /* input: network family for lookup (AF_INET, AF_INET6)
6561 * output: network family of egress nexthop
6565 /* set if lookup is to consider L4 data - e.g., FIB rules */
6570 union { /* used for MTU check */
6571 /* input to lookup */
6572 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
6574 /* output: MTU value */
6577 /* input: L3 device index for lookup
6578 * output: device index from FIB lookup
6583 /* inputs to lookup */
6584 __u8 tos; /* AF_INET */
6585 __be32 flowinfo; /* AF_INET6, flow_label + priority */
6587 /* output: metric of fib result (IPv4/IPv6 only) */
6593 __u32 ipv6_src[4]; /* in6_addr; network order */
6596 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
6597 * network header. output: bpf_fib_lookup sets to gateway address
6598 * if FIB lookup returns gateway route
6602 __u32 ipv6_dst[4]; /* in6_addr; network order */
6606 __be16 h_vlan_proto;
6608 __u8 smac[6]; /* ETH_ALEN */
6609 __u8 dmac[6]; /* ETH_ALEN */
6612 struct bpf_redir_neigh {
6613 /* network family for lookup (AF_INET, AF_INET6) */
6615 /* network address of nexthop; skips fib lookup to find gateway */
6618 __u32 ipv6_nh[4]; /* in6_addr; network order */
6622 /* bpf_check_mtu flags*/
6623 enum bpf_check_mtu_flags {
6624 BPF_MTU_CHK_SEGS = (1U << 0),
6627 enum bpf_check_mtu_ret {
6628 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
6629 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
6630 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
6633 enum bpf_task_fd_type {
6634 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
6635 BPF_FD_TYPE_TRACEPOINT, /* tp name */
6636 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
6637 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
6638 BPF_FD_TYPE_UPROBE, /* filename + offset */
6639 BPF_FD_TYPE_URETPROBE, /* filename + offset */
6643 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
6644 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
6645 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
6648 struct bpf_flow_keys {
6651 __u16 addr_proto; /* ETH_P_* of valid addrs */
6665 __u32 ipv6_src[4]; /* in6_addr; network order */
6666 __u32 ipv6_dst[4]; /* in6_addr; network order */
6673 struct bpf_func_info {
6678 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
6679 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
6681 struct bpf_line_info {
6683 __u32 file_name_off;
6688 struct bpf_spin_lock {
6695 } __attribute__((aligned(8)));
6700 } __attribute__((aligned(8)));
6703 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
6704 * Allows 1,2,4-byte read, but no write.
6706 __u32 file_pos; /* Sysctl file position to read from, write to.
6707 * Allows 1,2,4-byte read an 4-byte write.
6711 struct bpf_sockopt {
6712 __bpf_md_ptr(struct bpf_sock *, sk);
6713 __bpf_md_ptr(void *, optval);
6714 __bpf_md_ptr(void *, optval_end);
6722 struct bpf_pidns_info {
6727 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6728 struct bpf_sk_lookup {
6730 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6731 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6734 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
6735 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6736 __u32 remote_ip4; /* Network byte order */
6737 __u32 remote_ip6[4]; /* Network byte order */
6738 __be16 remote_port; /* Network byte order */
6739 __u16 :16; /* Zero padding */
6740 __u32 local_ip4; /* Network byte order */
6741 __u32 local_ip6[4]; /* Network byte order */
6742 __u32 local_port; /* Host byte order */
6743 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */
6747 * struct btf_ptr is used for typed pointer representation; the
6748 * type id is used to render the pointer data as the appropriate type
6749 * via the bpf_snprintf_btf() helper described above. A flags field -
6750 * potentially to specify additional details about the BTF pointer
6751 * (rather than its mode of display) - is included for future use.
6752 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6757 __u32 flags; /* BTF ptr flags; unused at present. */
6761 * Flags to control bpf_snprintf_btf() behaviour.
6762 * - BTF_F_COMPACT: no formatting around type information
6763 * - BTF_F_NONAME: no struct/union member names/types
6764 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6765 * equivalent to %px.
6766 * - BTF_F_ZERO: show zero-valued struct/union members; they
6767 * are not displayed by default
6770 BTF_F_COMPACT = (1ULL << 0),
6771 BTF_F_NONAME = (1ULL << 1),
6772 BTF_F_PTR_RAW = (1ULL << 2),
6773 BTF_F_ZERO = (1ULL << 3),
6776 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
6777 * has to be adjusted by relocations. It is emitted by llvm and passed to
6778 * libbpf and later to the kernel.
6780 enum bpf_core_relo_kind {
6781 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */
6782 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */
6783 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */
6784 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */
6785 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */
6786 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */
6787 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */
6788 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */
6789 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */
6790 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */
6791 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */
6792 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */
6793 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */
6797 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
6798 * and from libbpf to the kernel.
6800 * CO-RE relocation captures the following data:
6801 * - insn_off - instruction offset (in bytes) within a BPF program that needs
6802 * its insn->imm field to be relocated with actual field info;
6803 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
6805 * - access_str_off - offset into corresponding .BTF string section. String
6806 * interpretation depends on specific relocation kind:
6807 * - for field-based relocations, string encodes an accessed field using
6808 * a sequence of field and array indices, separated by colon (:). It's
6809 * conceptually very close to LLVM's getelementptr ([0]) instruction's
6810 * arguments for identifying offset to a field.
6811 * - for type-based relocations, strings is expected to be just "0";
6812 * - for enum value-based relocations, string contains an index of enum
6813 * value within its enum type;
6814 * - kind - one of enum bpf_core_relo_kind;
6824 * struct sample *s = ...;
6825 * int *x = &s->a; // encoded as "0:0" (a is field #0)
6826 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
6827 * // b is field #0 inside anon struct, accessing elem #5)
6828 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
6830 * type_id for all relocs in this example will capture BTF type id of
6833 * Such relocation is emitted when using __builtin_preserve_access_index()
6834 * Clang built-in, passing expression that captures field address, e.g.:
6836 * bpf_probe_read(&dst, sizeof(dst),
6837 * __builtin_preserve_access_index(&src->a.b.c));
6839 * In this case Clang will emit field relocation recording necessary data to
6840 * be able to find offset of embedded `a.b.c` field within `src` struct.
6842 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
6844 struct bpf_core_relo {
6847 __u32 access_str_off;
6848 enum bpf_core_relo_kind kind;
6851 #endif /* _UAPI__LINUX_BPF_H__ */