1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
23 #define BPF_XADD 0xc0 /* exclusive add - legacy name */
26 #define BPF_MOV 0xb0 /* mov reg to reg */
27 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
29 /* change endianness of a register */
30 #define BPF_END 0xd0 /* flags for endianness conversion: */
31 #define BPF_TO_LE 0x00 /* convert to little-endian */
32 #define BPF_TO_BE 0x08 /* convert to big-endian */
33 #define BPF_FROM_LE BPF_TO_LE
34 #define BPF_FROM_BE BPF_TO_BE
37 #define BPF_JNE 0x50 /* jump != */
38 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
39 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
40 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
41 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
42 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
43 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
44 #define BPF_CALL 0x80 /* function call */
45 #define BPF_EXIT 0x90 /* function return */
47 /* atomic op type fields (stored in immediate) */
48 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
49 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
50 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
52 /* Register numbers */
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG __MAX_BPF_REG
72 __u8 code; /* opcode */
73 __u8 dst_reg:4; /* dest register */
74 __u8 src_reg:4; /* source register */
75 __s16 off; /* signed offset */
76 __s32 imm; /* signed immediate constant */
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
82 __u8 data[0]; /* Arbitrary size */
85 struct bpf_cgroup_storage_key {
86 __u64 cgroup_inode_id; /* cgroup inode id */
87 __u32 attach_type; /* program attach type */
90 union bpf_iter_link_info {
96 /* BPF syscall commands, see bpf(2) man-page for more details. */
98 * DOC: eBPF Syscall Preamble
100 * The operation to be performed by the **bpf**\ () system call is determined
101 * by the *cmd* argument. Each operation takes an accompanying argument,
102 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
103 * below). The size argument is the size of the union pointed to by *attr*.
106 * DOC: eBPF Syscall Commands
110 * Create a map and return a file descriptor that refers to the
111 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
112 * is automatically enabled for the new file descriptor.
114 * Applying **close**\ (2) to the file descriptor returned by
115 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
118 * A new file descriptor (a nonnegative integer), or -1 if an
119 * error occurred (in which case, *errno* is set appropriately).
121 * BPF_MAP_LOOKUP_ELEM
123 * Look up an element with a given *key* in the map referred to
124 * by the file descriptor *map_fd*.
126 * The *flags* argument may be specified as one of the
130 * Look up the value of a spin-locked map without
131 * returning the lock. This must be specified if the
132 * elements contain a spinlock.
135 * Returns zero on success. On error, -1 is returned and *errno*
136 * is set appropriately.
138 * BPF_MAP_UPDATE_ELEM
140 * Create or update an element (key/value pair) in a specified map.
142 * The *flags* argument should be specified as one of the
146 * Create a new element or update an existing element.
148 * Create a new element only if it did not exist.
150 * Update an existing element.
152 * Update a spin_lock-ed map element.
155 * Returns zero on success. On error, -1 is returned and *errno*
156 * is set appropriately.
158 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
159 * **E2BIG**, **EEXIST**, or **ENOENT**.
162 * The number of elements in the map reached the
163 * *max_entries* limit specified at map creation time.
165 * If *flags* specifies **BPF_NOEXIST** and the element
166 * with *key* already exists in the map.
168 * If *flags* specifies **BPF_EXIST** and the element with
169 * *key* does not exist in the map.
171 * BPF_MAP_DELETE_ELEM
173 * Look up and delete an element by key in a specified map.
176 * Returns zero on success. On error, -1 is returned and *errno*
177 * is set appropriately.
179 * BPF_MAP_GET_NEXT_KEY
181 * Look up an element by key in a specified map and return the key
182 * of the next element. Can be used to iterate over all elements
186 * Returns zero on success. On error, -1 is returned and *errno*
187 * is set appropriately.
189 * The following cases can be used to iterate over all elements of
192 * * If *key* is not found, the operation returns zero and sets
193 * the *next_key* pointer to the key of the first element.
194 * * If *key* is found, the operation returns zero and sets the
195 * *next_key* pointer to the key of the next element.
196 * * If *key* is the last element, returns -1 and *errno* is set
199 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
200 * **EINVAL** on error.
204 * Verify and load an eBPF program, returning a new file
205 * descriptor associated with the program.
207 * Applying **close**\ (2) to the file descriptor returned by
208 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
210 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
211 * automatically enabled for the new file descriptor.
214 * A new file descriptor (a nonnegative integer), or -1 if an
215 * error occurred (in which case, *errno* is set appropriately).
219 * Pin an eBPF program or map referred by the specified *bpf_fd*
220 * to the provided *pathname* on the filesystem.
222 * The *pathname* argument must not contain a dot (".").
224 * On success, *pathname* retains a reference to the eBPF object,
225 * preventing deallocation of the object when the original
226 * *bpf_fd* is closed. This allow the eBPF object to live beyond
227 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
230 * Applying **unlink**\ (2) or similar calls to the *pathname*
231 * unpins the object from the filesystem, removing the reference.
232 * If no other file descriptors or filesystem nodes refer to the
233 * same object, it will be deallocated (see NOTES).
235 * The filesystem type for the parent directory of *pathname* must
236 * be **BPF_FS_MAGIC**.
239 * Returns zero on success. On error, -1 is returned and *errno*
240 * is set appropriately.
244 * Open a file descriptor for the eBPF object pinned to the
245 * specified *pathname*.
248 * A new file descriptor (a nonnegative integer), or -1 if an
249 * error occurred (in which case, *errno* is set appropriately).
253 * Attach an eBPF program to a *target_fd* at the specified
254 * *attach_type* hook.
256 * The *attach_type* specifies the eBPF attachment point to
257 * attach the program to, and must be one of *bpf_attach_type*
260 * The *attach_bpf_fd* must be a valid file descriptor for a
261 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
262 * or sock_ops type corresponding to the specified *attach_type*.
264 * The *target_fd* must be a valid file descriptor for a kernel
265 * object which depends on the attach type of *attach_bpf_fd*:
267 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
268 * **BPF_PROG_TYPE_CGROUP_SKB**,
269 * **BPF_PROG_TYPE_CGROUP_SOCK**,
270 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
271 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
272 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
273 * **BPF_PROG_TYPE_SOCK_OPS**
275 * Control Group v2 hierarchy with the eBPF controller
276 * enabled. Requires the kernel to be compiled with
277 * **CONFIG_CGROUP_BPF**.
279 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
281 * Network namespace (eg /proc/self/ns/net).
283 * **BPF_PROG_TYPE_LIRC_MODE2**
285 * LIRC device path (eg /dev/lircN). Requires the kernel
286 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
288 * **BPF_PROG_TYPE_SK_SKB**,
289 * **BPF_PROG_TYPE_SK_MSG**
291 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
294 * Returns zero on success. On error, -1 is returned and *errno*
295 * is set appropriately.
299 * Detach the eBPF program associated with the *target_fd* at the
300 * hook specified by *attach_type*. The program must have been
301 * previously attached using **BPF_PROG_ATTACH**.
304 * Returns zero on success. On error, -1 is returned and *errno*
305 * is set appropriately.
309 * Run the eBPF program associated with the *prog_fd* a *repeat*
310 * number of times against a provided program context *ctx_in* and
311 * data *data_in*, and return the modified program context
312 * *ctx_out*, *data_out* (for example, packet data), result of the
313 * execution *retval*, and *duration* of the test run.
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_XDP**
328 * *ctx_in* and *ctx_out* must be NULL.
330 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
331 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
333 * *ctx_out*, *data_in* and *data_out* must be NULL.
334 * *repeat* must be zero.
337 * Returns zero on success. On error, -1 is returned and *errno*
338 * is set appropriately.
341 * Either *data_size_out* or *ctx_size_out* is too small.
343 * This command is not supported by the program type of
344 * the program referred to by *prog_fd*.
346 * BPF_PROG_GET_NEXT_ID
348 * Fetch the next eBPF program currently loaded into the kernel.
350 * Looks for the eBPF program with an id greater than *start_id*
351 * and updates *next_id* on success. If no other eBPF programs
352 * remain with ids higher than *start_id*, returns -1 and sets
353 * *errno* to **ENOENT**.
356 * Returns zero on success. On error, or when no id remains, -1
357 * is returned and *errno* is set appropriately.
359 * BPF_MAP_GET_NEXT_ID
361 * Fetch the next eBPF map currently loaded into the kernel.
363 * Looks for the eBPF map with an id greater than *start_id*
364 * and updates *next_id* on success. If no other eBPF maps
365 * remain with ids higher than *start_id*, returns -1 and sets
366 * *errno* to **ENOENT**.
369 * Returns zero on success. On error, or when no id remains, -1
370 * is returned and *errno* is set appropriately.
372 * BPF_PROG_GET_FD_BY_ID
374 * Open a file descriptor for the eBPF program corresponding to
378 * A new file descriptor (a nonnegative integer), or -1 if an
379 * error occurred (in which case, *errno* is set appropriately).
381 * BPF_MAP_GET_FD_BY_ID
383 * Open a file descriptor for the eBPF map corresponding to
387 * A new file descriptor (a nonnegative integer), or -1 if an
388 * error occurred (in which case, *errno* is set appropriately).
390 * BPF_OBJ_GET_INFO_BY_FD
392 * Obtain information about the eBPF object corresponding to
395 * Populates up to *info_len* bytes of *info*, which will be in
396 * one of the following formats depending on the eBPF object type
399 * * **struct bpf_prog_info**
400 * * **struct bpf_map_info**
401 * * **struct bpf_btf_info**
402 * * **struct bpf_link_info**
405 * Returns zero on success. On error, -1 is returned and *errno*
406 * is set appropriately.
410 * Obtain information about eBPF programs associated with the
411 * specified *attach_type* hook.
413 * The *target_fd* must be a valid file descriptor for a kernel
414 * object which depends on the attach type of *attach_bpf_fd*:
416 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
417 * **BPF_PROG_TYPE_CGROUP_SKB**,
418 * **BPF_PROG_TYPE_CGROUP_SOCK**,
419 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
420 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
421 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
422 * **BPF_PROG_TYPE_SOCK_OPS**
424 * Control Group v2 hierarchy with the eBPF controller
425 * enabled. Requires the kernel to be compiled with
426 * **CONFIG_CGROUP_BPF**.
428 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
430 * Network namespace (eg /proc/self/ns/net).
432 * **BPF_PROG_TYPE_LIRC_MODE2**
434 * LIRC device path (eg /dev/lircN). Requires the kernel
435 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
437 * **BPF_PROG_QUERY** always fetches the number of programs
438 * attached and the *attach_flags* which were used to attach those
439 * programs. Additionally, if *prog_ids* is nonzero and the number
440 * of attached programs is less than *prog_cnt*, populates
441 * *prog_ids* with the eBPF program ids of the programs attached
444 * The following flags may alter the result:
446 * **BPF_F_QUERY_EFFECTIVE**
447 * Only return information regarding programs which are
448 * currently effective at the specified *target_fd*.
451 * Returns zero on success. On error, -1 is returned and *errno*
452 * is set appropriately.
454 * BPF_RAW_TRACEPOINT_OPEN
456 * Attach an eBPF program to a tracepoint *name* to access kernel
457 * internal arguments of the tracepoint in their raw form.
459 * The *prog_fd* must be a valid file descriptor associated with
460 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
462 * No ABI guarantees are made about the content of tracepoint
463 * arguments exposed to the corresponding eBPF program.
465 * Applying **close**\ (2) to the file descriptor returned by
466 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
469 * A new file descriptor (a nonnegative integer), or -1 if an
470 * error occurred (in which case, *errno* is set appropriately).
474 * Verify and load BPF Type Format (BTF) metadata into the kernel,
475 * returning a new file descriptor associated with the metadata.
476 * BTF is described in more detail at
477 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
479 * The *btf* parameter must point to valid memory providing
480 * *btf_size* bytes of BTF binary metadata.
482 * The returned file descriptor can be passed to other **bpf**\ ()
483 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
484 * associate the BTF with those objects.
486 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
487 * parameters to specify a *btf_log_buf*, *btf_log_size* and
488 * *btf_log_level* which allow the kernel to return freeform log
489 * output regarding the BTF verification process.
492 * A new file descriptor (a nonnegative integer), or -1 if an
493 * error occurred (in which case, *errno* is set appropriately).
495 * BPF_BTF_GET_FD_BY_ID
497 * Open a file descriptor for the BPF Type Format (BTF)
498 * corresponding to *btf_id*.
501 * A new file descriptor (a nonnegative integer), or -1 if an
502 * error occurred (in which case, *errno* is set appropriately).
506 * Obtain information about eBPF programs associated with the
507 * target process identified by *pid* and *fd*.
509 * If the *pid* and *fd* are associated with a tracepoint, kprobe
510 * or uprobe perf event, then the *prog_id* and *fd_type* will
511 * be populated with the eBPF program id and file descriptor type
512 * of type **bpf_task_fd_type**. If associated with a kprobe or
513 * uprobe, the *probe_offset* and *probe_addr* will also be
514 * populated. Optionally, if *buf* is provided, then up to
515 * *buf_len* bytes of *buf* will be populated with the name of
516 * the tracepoint, kprobe or uprobe.
518 * The resulting *prog_id* may be introspected in deeper detail
519 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
522 * Returns zero on success. On error, -1 is returned and *errno*
523 * is set appropriately.
525 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
527 * Look up an element with the given *key* in the map referred to
528 * by the file descriptor *fd*, and if found, delete the element.
530 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
531 * implement this command as a "pop" operation, deleting the top
532 * element rather than one corresponding to *key*.
533 * The *key* and *key_len* parameters should be zeroed when
534 * issuing this operation for these map types.
536 * This command is only valid for the following map types:
537 * * **BPF_MAP_TYPE_QUEUE**
538 * * **BPF_MAP_TYPE_STACK**
541 * Returns zero on success. On error, -1 is returned and *errno*
542 * is set appropriately.
546 * Freeze the permissions of the specified map.
548 * Write permissions may be frozen by passing zero *flags*.
549 * Upon success, no future syscall invocations may alter the
550 * map state of *map_fd*. Write operations from eBPF programs
551 * are still possible for a frozen map.
553 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
556 * Returns zero on success. On error, -1 is returned and *errno*
557 * is set appropriately.
559 * BPF_BTF_GET_NEXT_ID
561 * Fetch the next BPF Type Format (BTF) object currently loaded
564 * Looks for the BTF object with an id greater than *start_id*
565 * and updates *next_id* on success. If no other BTF objects
566 * remain with ids higher than *start_id*, returns -1 and sets
567 * *errno* to **ENOENT**.
570 * Returns zero on success. On error, or when no id remains, -1
571 * is returned and *errno* is set appropriately.
573 * BPF_MAP_LOOKUP_BATCH
575 * Iterate and fetch multiple elements in a map.
577 * Two opaque values are used to manage batch operations,
578 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
579 * to NULL to begin the batched operation. After each subsequent
580 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
581 * *out_batch* as the *in_batch* for the next operation to
582 * continue iteration from the current point.
584 * The *keys* and *values* are output parameters which must point
585 * to memory large enough to hold *count* items based on the key
586 * and value size of the map *map_fd*. The *keys* buffer must be
587 * of *key_size* * *count*. The *values* buffer must be of
588 * *value_size* * *count*.
590 * The *elem_flags* argument may be specified as one of the
594 * Look up the value of a spin-locked map without
595 * returning the lock. This must be specified if the
596 * elements contain a spinlock.
598 * On success, *count* elements from the map are copied into the
599 * user buffer, with the keys copied into *keys* and the values
600 * copied into the corresponding indices in *values*.
602 * If an error is returned and *errno* is not **EFAULT**, *count*
603 * is set to the number of successfully processed elements.
606 * Returns zero on success. On error, -1 is returned and *errno*
607 * is set appropriately.
609 * May set *errno* to **ENOSPC** to indicate that *keys* or
610 * *values* is too small to dump an entire bucket during
611 * iteration of a hash-based map type.
613 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
615 * Iterate and delete all elements in a map.
617 * This operation has the same behavior as
618 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
620 * * Every element that is successfully returned is also deleted
621 * from the map. This is at least *count* elements. Note that
622 * *count* is both an input and an output parameter.
623 * * Upon returning with *errno* set to **EFAULT**, up to
624 * *count* elements may be deleted without returning the keys
625 * and values of the deleted elements.
628 * Returns zero on success. On error, -1 is returned and *errno*
629 * is set appropriately.
631 * BPF_MAP_UPDATE_BATCH
633 * Update multiple elements in a map by *key*.
635 * The *keys* and *values* are input parameters which must point
636 * to memory large enough to hold *count* items based on the key
637 * and value size of the map *map_fd*. The *keys* buffer must be
638 * of *key_size* * *count*. The *values* buffer must be of
639 * *value_size* * *count*.
641 * Each element specified in *keys* is sequentially updated to the
642 * value in the corresponding index in *values*. The *in_batch*
643 * and *out_batch* parameters are ignored and should be zeroed.
645 * The *elem_flags* argument should be specified as one of the
649 * Create new elements or update a existing elements.
651 * Create new elements only if they do not exist.
653 * Update existing elements.
655 * Update spin_lock-ed map elements. This must be
656 * specified if the map value contains a spinlock.
658 * On success, *count* elements from the map are updated.
660 * If an error is returned and *errno* is not **EFAULT**, *count*
661 * is set to the number of successfully processed elements.
664 * Returns zero on success. On error, -1 is returned and *errno*
665 * is set appropriately.
667 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
668 * **E2BIG**. **E2BIG** indicates that the number of elements in
669 * the map reached the *max_entries* limit specified at map
672 * May set *errno* to one of the following error codes under
673 * specific circumstances:
676 * If *flags* specifies **BPF_NOEXIST** and the element
677 * with *key* already exists in the map.
679 * If *flags* specifies **BPF_EXIST** and the element with
680 * *key* does not exist in the map.
682 * BPF_MAP_DELETE_BATCH
684 * Delete multiple elements in a map by *key*.
686 * The *keys* parameter is an input parameter which must point
687 * to memory large enough to hold *count* items based on the key
688 * size of the map *map_fd*, that is, *key_size* * *count*.
690 * Each element specified in *keys* is sequentially deleted. The
691 * *in_batch*, *out_batch*, and *values* parameters are ignored
692 * and should be zeroed.
694 * The *elem_flags* argument may be specified as one of the
698 * Look up the value of a spin-locked map without
699 * returning the lock. This must be specified if the
700 * elements contain a spinlock.
702 * On success, *count* elements from the map are updated.
704 * If an error is returned and *errno* is not **EFAULT**, *count*
705 * is set to the number of successfully processed elements. If
706 * *errno* is **EFAULT**, up to *count* elements may be been
710 * Returns zero on success. On error, -1 is returned and *errno*
711 * is set appropriately.
715 * Attach an eBPF program to a *target_fd* at the specified
716 * *attach_type* hook and return a file descriptor handle for
720 * A new file descriptor (a nonnegative integer), or -1 if an
721 * error occurred (in which case, *errno* is set appropriately).
725 * Update the eBPF program in the specified *link_fd* to
729 * Returns zero on success. On error, -1 is returned and *errno*
730 * is set appropriately.
732 * BPF_LINK_GET_FD_BY_ID
734 * Open a file descriptor for the eBPF Link corresponding to
738 * A new file descriptor (a nonnegative integer), or -1 if an
739 * error occurred (in which case, *errno* is set appropriately).
741 * BPF_LINK_GET_NEXT_ID
743 * Fetch the next eBPF link currently loaded into the kernel.
745 * Looks for the eBPF link with an id greater than *start_id*
746 * and updates *next_id* on success. If no other eBPF links
747 * remain with ids higher than *start_id*, returns -1 and sets
748 * *errno* to **ENOENT**.
751 * Returns zero on success. On error, or when no id remains, -1
752 * is returned and *errno* is set appropriately.
756 * Enable eBPF runtime statistics gathering.
758 * Runtime statistics gathering for the eBPF runtime is disabled
759 * by default to minimize the corresponding performance overhead.
760 * This command enables statistics globally.
762 * Multiple programs may independently enable statistics.
763 * After gathering the desired statistics, eBPF runtime statistics
764 * may be disabled again by calling **close**\ (2) for the file
765 * descriptor returned by this function. Statistics will only be
766 * disabled system-wide when all outstanding file descriptors
767 * returned by prior calls for this subcommand are closed.
770 * A new file descriptor (a nonnegative integer), or -1 if an
771 * error occurred (in which case, *errno* is set appropriately).
775 * Create an iterator on top of the specified *link_fd* (as
776 * previously created using **BPF_LINK_CREATE**) and return a
777 * file descriptor that can be used to trigger the iteration.
779 * If the resulting file descriptor is pinned to the filesystem
780 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
781 * for that path will trigger the iterator to read kernel state
782 * using the eBPF program attached to *link_fd*.
785 * A new file descriptor (a nonnegative integer), or -1 if an
786 * error occurred (in which case, *errno* is set appropriately).
790 * Forcefully detach the specified *link_fd* from its
791 * corresponding attachment point.
794 * Returns zero on success. On error, -1 is returned and *errno*
795 * is set appropriately.
799 * Bind a map to the lifetime of an eBPF program.
801 * The map identified by *map_fd* is bound to the program
802 * identified by *prog_fd* and only released when *prog_fd* is
803 * released. This may be used in cases where metadata should be
804 * associated with a program which otherwise does not contain any
805 * references to the map (for example, embedded in the eBPF
806 * program instructions).
809 * Returns zero on success. On error, -1 is returned and *errno*
810 * is set appropriately.
813 * eBPF objects (maps and programs) can be shared between processes.
815 * * After **fork**\ (2), the child inherits file descriptors
816 * referring to the same eBPF objects.
817 * * File descriptors referring to eBPF objects can be transferred over
818 * **unix**\ (7) domain sockets.
819 * * File descriptors referring to eBPF objects can be duplicated in the
820 * usual way, using **dup**\ (2) and similar calls.
821 * * File descriptors referring to eBPF objects can be pinned to the
822 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
824 * An eBPF object is deallocated only after all file descriptors referring
825 * to the object have been closed and no references remain pinned to the
826 * filesystem or attached (for example, bound to a program or device).
833 BPF_MAP_GET_NEXT_KEY,
840 BPF_PROG_GET_NEXT_ID,
842 BPF_PROG_GET_FD_BY_ID,
843 BPF_MAP_GET_FD_BY_ID,
844 BPF_OBJ_GET_INFO_BY_FD,
846 BPF_RAW_TRACEPOINT_OPEN,
848 BPF_BTF_GET_FD_BY_ID,
850 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
853 BPF_MAP_LOOKUP_BATCH,
854 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
855 BPF_MAP_UPDATE_BATCH,
856 BPF_MAP_DELETE_BATCH,
859 BPF_LINK_GET_FD_BY_ID,
860 BPF_LINK_GET_NEXT_ID,
871 BPF_MAP_TYPE_PROG_ARRAY,
872 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
873 BPF_MAP_TYPE_PERCPU_HASH,
874 BPF_MAP_TYPE_PERCPU_ARRAY,
875 BPF_MAP_TYPE_STACK_TRACE,
876 BPF_MAP_TYPE_CGROUP_ARRAY,
877 BPF_MAP_TYPE_LRU_HASH,
878 BPF_MAP_TYPE_LRU_PERCPU_HASH,
879 BPF_MAP_TYPE_LPM_TRIE,
880 BPF_MAP_TYPE_ARRAY_OF_MAPS,
881 BPF_MAP_TYPE_HASH_OF_MAPS,
883 BPF_MAP_TYPE_SOCKMAP,
886 BPF_MAP_TYPE_SOCKHASH,
887 BPF_MAP_TYPE_CGROUP_STORAGE,
888 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
889 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
892 BPF_MAP_TYPE_SK_STORAGE,
893 BPF_MAP_TYPE_DEVMAP_HASH,
894 BPF_MAP_TYPE_STRUCT_OPS,
895 BPF_MAP_TYPE_RINGBUF,
896 BPF_MAP_TYPE_INODE_STORAGE,
897 BPF_MAP_TYPE_TASK_STORAGE,
900 /* Note that tracing related programs such as
901 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
902 * are not subject to a stable API since kernel internal data
903 * structures can change from release to release and may
904 * therefore break existing tracing BPF programs. Tracing BPF
905 * programs correspond to /a/ specific kernel which is to be
906 * analyzed, and not /a/ specific kernel /and/ all future ones.
909 BPF_PROG_TYPE_UNSPEC,
910 BPF_PROG_TYPE_SOCKET_FILTER,
911 BPF_PROG_TYPE_KPROBE,
912 BPF_PROG_TYPE_SCHED_CLS,
913 BPF_PROG_TYPE_SCHED_ACT,
914 BPF_PROG_TYPE_TRACEPOINT,
916 BPF_PROG_TYPE_PERF_EVENT,
917 BPF_PROG_TYPE_CGROUP_SKB,
918 BPF_PROG_TYPE_CGROUP_SOCK,
919 BPF_PROG_TYPE_LWT_IN,
920 BPF_PROG_TYPE_LWT_OUT,
921 BPF_PROG_TYPE_LWT_XMIT,
922 BPF_PROG_TYPE_SOCK_OPS,
923 BPF_PROG_TYPE_SK_SKB,
924 BPF_PROG_TYPE_CGROUP_DEVICE,
925 BPF_PROG_TYPE_SK_MSG,
926 BPF_PROG_TYPE_RAW_TRACEPOINT,
927 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
928 BPF_PROG_TYPE_LWT_SEG6LOCAL,
929 BPF_PROG_TYPE_LIRC_MODE2,
930 BPF_PROG_TYPE_SK_REUSEPORT,
931 BPF_PROG_TYPE_FLOW_DISSECTOR,
932 BPF_PROG_TYPE_CGROUP_SYSCTL,
933 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
934 BPF_PROG_TYPE_CGROUP_SOCKOPT,
935 BPF_PROG_TYPE_TRACING,
936 BPF_PROG_TYPE_STRUCT_OPS,
939 BPF_PROG_TYPE_SK_LOOKUP,
942 enum bpf_attach_type {
943 BPF_CGROUP_INET_INGRESS,
944 BPF_CGROUP_INET_EGRESS,
945 BPF_CGROUP_INET_SOCK_CREATE,
947 BPF_SK_SKB_STREAM_PARSER,
948 BPF_SK_SKB_STREAM_VERDICT,
951 BPF_CGROUP_INET4_BIND,
952 BPF_CGROUP_INET6_BIND,
953 BPF_CGROUP_INET4_CONNECT,
954 BPF_CGROUP_INET6_CONNECT,
955 BPF_CGROUP_INET4_POST_BIND,
956 BPF_CGROUP_INET6_POST_BIND,
957 BPF_CGROUP_UDP4_SENDMSG,
958 BPF_CGROUP_UDP6_SENDMSG,
962 BPF_CGROUP_UDP4_RECVMSG,
963 BPF_CGROUP_UDP6_RECVMSG,
964 BPF_CGROUP_GETSOCKOPT,
965 BPF_CGROUP_SETSOCKOPT,
972 BPF_CGROUP_INET4_GETPEERNAME,
973 BPF_CGROUP_INET6_GETPEERNAME,
974 BPF_CGROUP_INET4_GETSOCKNAME,
975 BPF_CGROUP_INET6_GETSOCKNAME,
977 BPF_CGROUP_INET_SOCK_RELEASE,
982 __MAX_BPF_ATTACH_TYPE
985 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
988 BPF_LINK_TYPE_UNSPEC = 0,
989 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
990 BPF_LINK_TYPE_TRACING = 2,
991 BPF_LINK_TYPE_CGROUP = 3,
992 BPF_LINK_TYPE_ITER = 4,
993 BPF_LINK_TYPE_NETNS = 5,
994 BPF_LINK_TYPE_XDP = 6,
999 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1001 * NONE(default): No further bpf programs allowed in the subtree.
1003 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1004 * the program in this cgroup yields to sub-cgroup program.
1006 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1007 * that cgroup program gets run in addition to the program in this cgroup.
1009 * Only one program is allowed to be attached to a cgroup with
1010 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1011 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1012 * release old program and attach the new one. Attach flags has to match.
1014 * Multiple programs are allowed to be attached to a cgroup with
1015 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1016 * (those that were attached first, run first)
1017 * The programs of sub-cgroup are executed first, then programs of
1018 * this cgroup and then programs of parent cgroup.
1019 * When children program makes decision (like picking TCP CA or sock bind)
1020 * parent program has a chance to override it.
1022 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1023 * programs for a cgroup. Though it's possible to replace an old program at
1024 * any position by also specifying BPF_F_REPLACE flag and position itself in
1025 * replace_bpf_fd attribute. Old program at this position will be released.
1027 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1028 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1030 * cgrp1 (MULTI progs A, B) ->
1031 * cgrp2 (OVERRIDE prog C) ->
1032 * cgrp3 (MULTI prog D) ->
1033 * cgrp4 (OVERRIDE prog E) ->
1034 * cgrp5 (NONE prog F)
1035 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1036 * if prog F is detached, the execution is E,D,A,B
1037 * if prog F and D are detached, the execution is E,A,B
1038 * if prog F, E and D are detached, the execution is C,A,B
1040 * All eligible programs are executed regardless of return code from
1043 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
1044 #define BPF_F_ALLOW_MULTI (1U << 1)
1045 #define BPF_F_REPLACE (1U << 2)
1047 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1048 * verifier will perform strict alignment checking as if the kernel
1049 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1050 * and NET_IP_ALIGN defined to 2.
1052 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
1054 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1055 * verifier will allow any alignment whatsoever. On platforms
1056 * with strict alignment requirements for loads ands stores (such
1057 * as sparc and mips) the verifier validates that all loads and
1058 * stores provably follow this requirement. This flag turns that
1059 * checking and enforcement off.
1061 * It is mostly used for testing when we want to validate the
1062 * context and memory access aspects of the verifier, but because
1063 * of an unaligned access the alignment check would trigger before
1064 * the one we are interested in.
1066 #define BPF_F_ANY_ALIGNMENT (1U << 1)
1068 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1069 * Verifier does sub-register def/use analysis and identifies instructions whose
1070 * def only matters for low 32-bit, high 32-bit is never referenced later
1071 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1072 * that it is safe to ignore clearing high 32-bit for these instructions. This
1073 * saves some back-ends a lot of code-gen. However such optimization is not
1074 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1075 * hence hasn't used verifier's analysis result. But, we really want to have a
1076 * way to be able to verify the correctness of the described optimization on
1077 * x86_64 on which testsuites are frequently exercised.
1079 * So, this flag is introduced. Once it is set, verifier will randomize high
1080 * 32-bit for those instructions who has been identified as safe to ignore them.
1081 * Then, if verifier is not doing correct analysis, such randomization will
1082 * regress tests to expose bugs.
1084 #define BPF_F_TEST_RND_HI32 (1U << 2)
1086 /* The verifier internal test flag. Behavior is undefined */
1087 #define BPF_F_TEST_STATE_FREQ (1U << 3)
1089 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1090 * restrict map and helper usage for such programs. Sleepable BPF programs can
1091 * only be attached to hooks where kernel execution context allows sleeping.
1092 * Such programs are allowed to use helpers that may sleep like
1093 * bpf_copy_from_user().
1095 #define BPF_F_SLEEPABLE (1U << 4)
1097 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1098 * the following extensions:
1100 * insn[0].src_reg: BPF_PSEUDO_MAP_FD
1101 * insn[0].imm: map fd
1105 * ldimm64 rewrite: address of map
1106 * verifier type: CONST_PTR_TO_MAP
1108 #define BPF_PSEUDO_MAP_FD 1
1109 /* insn[0].src_reg: BPF_PSEUDO_MAP_VALUE
1110 * insn[0].imm: map fd
1111 * insn[1].imm: offset into value
1114 * ldimm64 rewrite: address of map[0]+offset
1115 * verifier type: PTR_TO_MAP_VALUE
1117 #define BPF_PSEUDO_MAP_VALUE 2
1118 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1119 * insn[0].imm: kernel btd id of VAR
1123 * ldimm64 rewrite: address of the kernel variable
1124 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1127 #define BPF_PSEUDO_BTF_ID 3
1128 /* insn[0].src_reg: BPF_PSEUDO_FUNC
1129 * insn[0].imm: insn offset to the func
1133 * ldimm64 rewrite: address of the function
1134 * verifier type: PTR_TO_FUNC.
1136 #define BPF_PSEUDO_FUNC 4
1138 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1139 * offset to another bpf function
1141 #define BPF_PSEUDO_CALL 1
1142 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1143 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1145 #define BPF_PSEUDO_KFUNC_CALL 2
1147 /* flags for BPF_MAP_UPDATE_ELEM command */
1149 BPF_ANY = 0, /* create new element or update existing */
1150 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1151 BPF_EXIST = 2, /* update existing element */
1152 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1155 /* flags for BPF_MAP_CREATE command */
1157 BPF_F_NO_PREALLOC = (1U << 0),
1158 /* Instead of having one common LRU list in the
1159 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1160 * which can scale and perform better.
1161 * Note, the LRU nodes (including free nodes) cannot be moved
1162 * across different LRU lists.
1164 BPF_F_NO_COMMON_LRU = (1U << 1),
1165 /* Specify numa node during map creation */
1166 BPF_F_NUMA_NODE = (1U << 2),
1168 /* Flags for accessing BPF object from syscall side. */
1169 BPF_F_RDONLY = (1U << 3),
1170 BPF_F_WRONLY = (1U << 4),
1172 /* Flag for stack_map, store build_id+offset instead of pointer */
1173 BPF_F_STACK_BUILD_ID = (1U << 5),
1175 /* Zero-initialize hash function seed. This should only be used for testing. */
1176 BPF_F_ZERO_SEED = (1U << 6),
1178 /* Flags for accessing BPF object from program side. */
1179 BPF_F_RDONLY_PROG = (1U << 7),
1180 BPF_F_WRONLY_PROG = (1U << 8),
1182 /* Clone map from listener for newly accepted socket */
1183 BPF_F_CLONE = (1U << 9),
1185 /* Enable memory-mapping BPF map */
1186 BPF_F_MMAPABLE = (1U << 10),
1188 /* Share perf_event among processes */
1189 BPF_F_PRESERVE_ELEMS = (1U << 11),
1191 /* Create a map that is suitable to be an inner map with dynamic max entries */
1192 BPF_F_INNER_MAP = (1U << 12),
1195 /* Flags for BPF_PROG_QUERY. */
1197 /* Query effective (directly attached + inherited from ancestor cgroups)
1198 * programs that will be executed for events within a cgroup.
1199 * attach_flags with this flag are returned only for directly attached programs.
1201 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
1203 /* Flags for BPF_PROG_TEST_RUN */
1205 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1206 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1208 /* type for BPF_ENABLE_STATS */
1209 enum bpf_stats_type {
1210 /* enabled run_time_ns and run_cnt */
1211 BPF_STATS_RUN_TIME = 0,
1214 enum bpf_stack_build_id_status {
1215 /* user space need an empty entry to identify end of a trace */
1216 BPF_STACK_BUILD_ID_EMPTY = 0,
1217 /* with valid build_id and offset */
1218 BPF_STACK_BUILD_ID_VALID = 1,
1219 /* couldn't get build_id, fallback to ip */
1220 BPF_STACK_BUILD_ID_IP = 2,
1223 #define BPF_BUILD_ID_SIZE 20
1224 struct bpf_stack_build_id {
1226 unsigned char build_id[BPF_BUILD_ID_SIZE];
1233 #define BPF_OBJ_NAME_LEN 16U
1236 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1237 __u32 map_type; /* one of enum bpf_map_type */
1238 __u32 key_size; /* size of key in bytes */
1239 __u32 value_size; /* size of value in bytes */
1240 __u32 max_entries; /* max number of entries in a map */
1241 __u32 map_flags; /* BPF_MAP_CREATE related
1242 * flags defined above.
1244 __u32 inner_map_fd; /* fd pointing to the inner map */
1245 __u32 numa_node; /* numa node (effective only if
1246 * BPF_F_NUMA_NODE is set).
1248 char map_name[BPF_OBJ_NAME_LEN];
1249 __u32 map_ifindex; /* ifindex of netdev to create on */
1250 __u32 btf_fd; /* fd pointing to a BTF type data */
1251 __u32 btf_key_type_id; /* BTF type_id of the key */
1252 __u32 btf_value_type_id; /* BTF type_id of the value */
1253 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1254 * struct stored as the
1259 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1263 __aligned_u64 value;
1264 __aligned_u64 next_key;
1269 struct { /* struct used by BPF_MAP_*_BATCH commands */
1270 __aligned_u64 in_batch; /* start batch,
1271 * NULL to start from beginning
1273 __aligned_u64 out_batch; /* output: next start batch */
1275 __aligned_u64 values;
1276 __u32 count; /* input/output:
1277 * input: # of key/value
1279 * output: # of filled elements
1286 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1287 __u32 prog_type; /* one of enum bpf_prog_type */
1289 __aligned_u64 insns;
1290 __aligned_u64 license;
1291 __u32 log_level; /* verbosity level of verifier */
1292 __u32 log_size; /* size of user buffer */
1293 __aligned_u64 log_buf; /* user supplied buffer */
1294 __u32 kern_version; /* not used */
1296 char prog_name[BPF_OBJ_NAME_LEN];
1297 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1298 /* For some prog types expected attach type must be known at
1299 * load time to verify attach type specific parts of prog
1300 * (context accesses, allowed helpers, etc).
1302 __u32 expected_attach_type;
1303 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1304 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1305 __aligned_u64 func_info; /* func info */
1306 __u32 func_info_cnt; /* number of bpf_func_info records */
1307 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1308 __aligned_u64 line_info; /* line info */
1309 __u32 line_info_cnt; /* number of bpf_line_info records */
1310 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1312 /* valid prog_fd to attach to bpf prog */
1313 __u32 attach_prog_fd;
1314 /* or valid module BTF object fd or 0 to attach to vmlinux */
1315 __u32 attach_btf_obj_fd;
1319 struct { /* anonymous struct used by BPF_OBJ_* commands */
1320 __aligned_u64 pathname;
1325 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1326 __u32 target_fd; /* container object to attach to */
1327 __u32 attach_bpf_fd; /* eBPF program to attach */
1330 __u32 replace_bpf_fd; /* previously attached eBPF
1331 * program to replace if
1332 * BPF_F_REPLACE is used
1336 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1339 __u32 data_size_in; /* input: len of data_in */
1340 __u32 data_size_out; /* input/output: len of data_out
1341 * returns ENOSPC if data_out
1344 __aligned_u64 data_in;
1345 __aligned_u64 data_out;
1348 __u32 ctx_size_in; /* input: len of ctx_in */
1349 __u32 ctx_size_out; /* input/output: len of ctx_out
1350 * returns ENOSPC if ctx_out
1353 __aligned_u64 ctx_in;
1354 __aligned_u64 ctx_out;
1359 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1371 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1377 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1378 __u32 target_fd; /* container object to query */
1382 __aligned_u64 prog_ids;
1386 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1391 struct { /* anonymous struct for BPF_BTF_LOAD */
1393 __aligned_u64 btf_log_buf;
1396 __u32 btf_log_level;
1400 __u32 pid; /* input: pid */
1401 __u32 fd; /* input: fd */
1402 __u32 flags; /* input: flags */
1403 __u32 buf_len; /* input/output: buf len */
1404 __aligned_u64 buf; /* input/output:
1405 * tp_name for tracepoint
1407 * filename for uprobe
1409 __u32 prog_id; /* output: prod_id */
1410 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1411 __u64 probe_offset; /* output: probe_offset */
1412 __u64 probe_addr; /* output: probe_addr */
1415 struct { /* struct used by BPF_LINK_CREATE command */
1416 __u32 prog_fd; /* eBPF program to attach */
1418 __u32 target_fd; /* object to attach to */
1419 __u32 target_ifindex; /* target ifindex */
1421 __u32 attach_type; /* attach type */
1422 __u32 flags; /* extra flags */
1424 __u32 target_btf_id; /* btf_id of target to attach to */
1426 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1427 __u32 iter_info_len; /* iter_info length */
1432 struct { /* struct used by BPF_LINK_UPDATE command */
1433 __u32 link_fd; /* link fd */
1434 /* new program fd to update link with */
1436 __u32 flags; /* extra flags */
1437 /* expected link's program fd; is specified only if
1438 * BPF_F_REPLACE flag is set in flags */
1446 struct { /* struct used by BPF_ENABLE_STATS command */
1450 struct { /* struct used by BPF_ITER_CREATE command */
1455 struct { /* struct used by BPF_PROG_BIND_MAP command */
1458 __u32 flags; /* extra flags */
1461 } __attribute__((aligned(8)));
1463 /* The description below is an attempt at providing documentation to eBPF
1464 * developers about the multiple available eBPF helper functions. It can be
1465 * parsed and used to produce a manual page. The workflow is the following,
1466 * and requires the rst2man utility:
1468 * $ ./scripts/bpf_doc.py \
1469 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1470 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1471 * $ man /tmp/bpf-helpers.7
1473 * Note that in order to produce this external documentation, some RST
1474 * formatting is used in the descriptions to get "bold" and "italics" in
1475 * manual pages. Also note that the few trailing white spaces are
1476 * intentional, removing them would break paragraphs for rst2man.
1478 * Start of BPF helper function descriptions:
1480 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1482 * Perform a lookup in *map* for an entry associated to *key*.
1484 * Map value associated to *key*, or **NULL** if no entry was
1487 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1489 * Add or update the value of the entry associated to *key* in
1490 * *map* with *value*. *flags* is one of:
1493 * The entry for *key* must not exist in the map.
1495 * The entry for *key* must already exist in the map.
1497 * No condition on the existence of the entry for *key*.
1499 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1500 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1501 * elements always exist), the helper would return an error.
1503 * 0 on success, or a negative error in case of failure.
1505 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1507 * Delete entry with *key* from *map*.
1509 * 0 on success, or a negative error in case of failure.
1511 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1513 * For tracing programs, safely attempt to read *size* bytes from
1514 * kernel space address *unsafe_ptr* and store the data in *dst*.
1516 * Generally, use **bpf_probe_read_user**\ () or
1517 * **bpf_probe_read_kernel**\ () instead.
1519 * 0 on success, or a negative error in case of failure.
1521 * u64 bpf_ktime_get_ns(void)
1523 * Return the time elapsed since system boot, in nanoseconds.
1524 * Does not include time the system was suspended.
1525 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1529 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1531 * This helper is a "printk()-like" facility for debugging. It
1532 * prints a message defined by format *fmt* (of size *fmt_size*)
1533 * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1534 * available. It can take up to three additional **u64**
1535 * arguments (as an eBPF helpers, the total number of arguments is
1538 * Each time the helper is called, it appends a line to the trace.
1539 * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1540 * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1541 * The format of the trace is customizable, and the exact output
1542 * one will get depends on the options set in
1543 * *\/sys/kernel/debug/tracing/trace_options* (see also the
1544 * *README* file under the same directory). However, it usually
1545 * defaults to something like:
1549 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1553 * * ``telnet`` is the name of the current task.
1554 * * ``470`` is the PID of the current task.
1555 * * ``001`` is the CPU number on which the task is
1557 * * In ``.N..``, each character refers to a set of
1558 * options (whether irqs are enabled, scheduling
1559 * options, whether hard/softirqs are running, level of
1560 * preempt_disabled respectively). **N** means that
1561 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1563 * * ``419421.045894`` is a timestamp.
1564 * * ``0x00000001`` is a fake value used by BPF for the
1565 * instruction pointer register.
1566 * * ``<formatted msg>`` is the message formatted with
1569 * The conversion specifiers supported by *fmt* are similar, but
1570 * more limited than for printk(). They are **%d**, **%i**,
1571 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1572 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1573 * of field, padding with zeroes, etc.) is available, and the
1574 * helper will return **-EINVAL** (but print nothing) if it
1575 * encounters an unknown specifier.
1577 * Also, note that **bpf_trace_printk**\ () is slow, and should
1578 * only be used for debugging purposes. For this reason, a notice
1579 * block (spanning several lines) is printed to kernel logs and
1580 * states that the helper should not be used "for production use"
1581 * the first time this helper is used (or more precisely, when
1582 * **trace_printk**\ () buffers are allocated). For passing values
1583 * to user space, perf events should be preferred.
1585 * The number of bytes written to the buffer, or a negative error
1586 * in case of failure.
1588 * u32 bpf_get_prandom_u32(void)
1590 * Get a pseudo-random number.
1592 * From a security point of view, this helper uses its own
1593 * pseudo-random internal state, and cannot be used to infer the
1594 * seed of other random functions in the kernel. However, it is
1595 * essential to note that the generator used by the helper is not
1596 * cryptographically secure.
1598 * A random 32-bit unsigned value.
1600 * u32 bpf_get_smp_processor_id(void)
1602 * Get the SMP (symmetric multiprocessing) processor id. Note that
1603 * all programs run with preemption disabled, which means that the
1604 * SMP processor id is stable during all the execution of the
1607 * The SMP id of the processor running the program.
1609 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1611 * Store *len* bytes from address *from* into the packet
1612 * associated to *skb*, at *offset*. *flags* are a combination of
1613 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1614 * checksum for the packet after storing the bytes) and
1615 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1616 * **->swhash** and *skb*\ **->l4hash** to 0).
1618 * A call to this helper is susceptible to change the underlying
1619 * packet buffer. Therefore, at load time, all checks on pointers
1620 * previously done by the verifier are invalidated and must be
1621 * performed again, if the helper is used in combination with
1622 * direct packet access.
1624 * 0 on success, or a negative error in case of failure.
1626 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1628 * Recompute the layer 3 (e.g. IP) checksum for the packet
1629 * associated to *skb*. Computation is incremental, so the helper
1630 * must know the former value of the header field that was
1631 * modified (*from*), the new value of this field (*to*), and the
1632 * number of bytes (2 or 4) for this field, stored in *size*.
1633 * Alternatively, it is possible to store the difference between
1634 * the previous and the new values of the header field in *to*, by
1635 * setting *from* and *size* to 0. For both methods, *offset*
1636 * indicates the location of the IP checksum within the packet.
1638 * This helper works in combination with **bpf_csum_diff**\ (),
1639 * which does not update the checksum in-place, but offers more
1640 * flexibility and can handle sizes larger than 2 or 4 for the
1641 * checksum to update.
1643 * A call to this helper is susceptible to change the underlying
1644 * packet buffer. Therefore, at load time, all checks on pointers
1645 * previously done by the verifier are invalidated and must be
1646 * performed again, if the helper is used in combination with
1647 * direct packet access.
1649 * 0 on success, or a negative error in case of failure.
1651 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1653 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1654 * packet associated to *skb*. Computation is incremental, so the
1655 * helper must know the former value of the header field that was
1656 * modified (*from*), the new value of this field (*to*), and the
1657 * number of bytes (2 or 4) for this field, stored on the lowest
1658 * four bits of *flags*. Alternatively, it is possible to store
1659 * the difference between the previous and the new values of the
1660 * header field in *to*, by setting *from* and the four lowest
1661 * bits of *flags* to 0. For both methods, *offset* indicates the
1662 * location of the IP checksum within the packet. In addition to
1663 * the size of the field, *flags* can be added (bitwise OR) actual
1664 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1665 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1666 * for updates resulting in a null checksum the value is set to
1667 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1668 * the checksum is to be computed against a pseudo-header.
1670 * This helper works in combination with **bpf_csum_diff**\ (),
1671 * which does not update the checksum in-place, but offers more
1672 * flexibility and can handle sizes larger than 2 or 4 for the
1673 * checksum to update.
1675 * A call to this helper is susceptible to change the underlying
1676 * packet buffer. Therefore, at load time, all checks on pointers
1677 * previously done by the verifier are invalidated and must be
1678 * performed again, if the helper is used in combination with
1679 * direct packet access.
1681 * 0 on success, or a negative error in case of failure.
1683 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1685 * This special helper is used to trigger a "tail call", or in
1686 * other words, to jump into another eBPF program. The same stack
1687 * frame is used (but values on stack and in registers for the
1688 * caller are not accessible to the callee). This mechanism allows
1689 * for program chaining, either for raising the maximum number of
1690 * available eBPF instructions, or to execute given programs in
1691 * conditional blocks. For security reasons, there is an upper
1692 * limit to the number of successive tail calls that can be
1695 * Upon call of this helper, the program attempts to jump into a
1696 * program referenced at index *index* in *prog_array_map*, a
1697 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1698 * *ctx*, a pointer to the context.
1700 * If the call succeeds, the kernel immediately runs the first
1701 * instruction of the new program. This is not a function call,
1702 * and it never returns to the previous program. If the call
1703 * fails, then the helper has no effect, and the caller continues
1704 * to run its subsequent instructions. A call can fail if the
1705 * destination program for the jump does not exist (i.e. *index*
1706 * is superior to the number of entries in *prog_array_map*), or
1707 * if the maximum number of tail calls has been reached for this
1708 * chain of programs. This limit is defined in the kernel by the
1709 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1710 * which is currently set to 32.
1712 * 0 on success, or a negative error in case of failure.
1714 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1716 * Clone and redirect the packet associated to *skb* to another
1717 * net device of index *ifindex*. Both ingress and egress
1718 * interfaces can be used for redirection. The **BPF_F_INGRESS**
1719 * value in *flags* is used to make the distinction (ingress path
1720 * is selected if the flag is present, egress path otherwise).
1721 * This is the only flag supported for now.
1723 * In comparison with **bpf_redirect**\ () helper,
1724 * **bpf_clone_redirect**\ () has the associated cost of
1725 * duplicating the packet buffer, but this can be executed out of
1726 * the eBPF program. Conversely, **bpf_redirect**\ () is more
1727 * efficient, but it is handled through an action code where the
1728 * redirection happens only after the eBPF program has returned.
1730 * A call to this helper is susceptible to change the underlying
1731 * packet buffer. Therefore, at load time, all checks on pointers
1732 * previously done by the verifier are invalidated and must be
1733 * performed again, if the helper is used in combination with
1734 * direct packet access.
1736 * 0 on success, or a negative error in case of failure.
1738 * u64 bpf_get_current_pid_tgid(void)
1740 * A 64-bit integer containing the current tgid and pid, and
1742 * *current_task*\ **->tgid << 32 \|**
1743 * *current_task*\ **->pid**.
1745 * u64 bpf_get_current_uid_gid(void)
1747 * A 64-bit integer containing the current GID and UID, and
1748 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1750 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1752 * Copy the **comm** attribute of the current task into *buf* of
1753 * *size_of_buf*. The **comm** attribute contains the name of
1754 * the executable (excluding the path) for the current task. The
1755 * *size_of_buf* must be strictly positive. On success, the
1756 * helper makes sure that the *buf* is NUL-terminated. On failure,
1757 * it is filled with zeroes.
1759 * 0 on success, or a negative error in case of failure.
1761 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1763 * Retrieve the classid for the current task, i.e. for the net_cls
1764 * cgroup to which *skb* belongs.
1766 * This helper can be used on TC egress path, but not on ingress.
1768 * The net_cls cgroup provides an interface to tag network packets
1769 * based on a user-provided identifier for all traffic coming from
1770 * the tasks belonging to the related cgroup. See also the related
1771 * kernel documentation, available from the Linux sources in file
1772 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1774 * The Linux kernel has two versions for cgroups: there are
1775 * cgroups v1 and cgroups v2. Both are available to users, who can
1776 * use a mixture of them, but note that the net_cls cgroup is for
1777 * cgroup v1 only. This makes it incompatible with BPF programs
1778 * run on cgroups, which is a cgroup-v2-only feature (a socket can
1779 * only hold data for one version of cgroups at a time).
1781 * This helper is only available is the kernel was compiled with
1782 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1783 * "**y**" or to "**m**".
1785 * The classid, or 0 for the default unconfigured classid.
1787 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1789 * Push a *vlan_tci* (VLAN tag control information) of protocol
1790 * *vlan_proto* to the packet associated to *skb*, then update
1791 * the checksum. Note that if *vlan_proto* is different from
1792 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1793 * be **ETH_P_8021Q**.
1795 * A call to this helper is susceptible to change the underlying
1796 * packet buffer. Therefore, at load time, all checks on pointers
1797 * previously done by the verifier are invalidated and must be
1798 * performed again, if the helper is used in combination with
1799 * direct packet access.
1801 * 0 on success, or a negative error in case of failure.
1803 * long bpf_skb_vlan_pop(struct sk_buff *skb)
1805 * Pop a VLAN header from the packet associated to *skb*.
1807 * A call to this helper is susceptible to change the underlying
1808 * packet buffer. Therefore, at load time, all checks on pointers
1809 * previously done by the verifier are invalidated and must be
1810 * performed again, if the helper is used in combination with
1811 * direct packet access.
1813 * 0 on success, or a negative error in case of failure.
1815 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1817 * Get tunnel metadata. This helper takes a pointer *key* to an
1818 * empty **struct bpf_tunnel_key** of **size**, that will be
1819 * filled with tunnel metadata for the packet associated to *skb*.
1820 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1821 * indicates that the tunnel is based on IPv6 protocol instead of
1824 * The **struct bpf_tunnel_key** is an object that generalizes the
1825 * principal parameters used by various tunneling protocols into a
1826 * single struct. This way, it can be used to easily make a
1827 * decision based on the contents of the encapsulation header,
1828 * "summarized" in this struct. In particular, it holds the IP
1829 * address of the remote end (IPv4 or IPv6, depending on the case)
1830 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1831 * this struct exposes the *key*\ **->tunnel_id**, which is
1832 * generally mapped to a VNI (Virtual Network Identifier), making
1833 * it programmable together with the **bpf_skb_set_tunnel_key**\
1836 * Let's imagine that the following code is part of a program
1837 * attached to the TC ingress interface, on one end of a GRE
1838 * tunnel, and is supposed to filter out all messages coming from
1839 * remote ends with IPv4 address other than 10.0.0.1:
1844 * struct bpf_tunnel_key key = {};
1846 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1848 * return TC_ACT_SHOT; // drop packet
1850 * if (key.remote_ipv4 != 0x0a000001)
1851 * return TC_ACT_SHOT; // drop packet
1853 * return TC_ACT_OK; // accept packet
1855 * This interface can also be used with all encapsulation devices
1856 * that can operate in "collect metadata" mode: instead of having
1857 * one network device per specific configuration, the "collect
1858 * metadata" mode only requires a single device where the
1859 * configuration can be extracted from this helper.
1861 * This can be used together with various tunnels such as VXLan,
1862 * Geneve, GRE or IP in IP (IPIP).
1864 * 0 on success, or a negative error in case of failure.
1866 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1868 * Populate tunnel metadata for packet associated to *skb.* The
1869 * tunnel metadata is set to the contents of *key*, of *size*. The
1870 * *flags* can be set to a combination of the following values:
1872 * **BPF_F_TUNINFO_IPV6**
1873 * Indicate that the tunnel is based on IPv6 protocol
1875 * **BPF_F_ZERO_CSUM_TX**
1876 * For IPv4 packets, add a flag to tunnel metadata
1877 * indicating that checksum computation should be skipped
1878 * and checksum set to zeroes.
1879 * **BPF_F_DONT_FRAGMENT**
1880 * Add a flag to tunnel metadata indicating that the
1881 * packet should not be fragmented.
1882 * **BPF_F_SEQ_NUMBER**
1883 * Add a flag to tunnel metadata indicating that a
1884 * sequence number should be added to tunnel header before
1885 * sending the packet. This flag was added for GRE
1886 * encapsulation, but might be used with other protocols
1887 * as well in the future.
1889 * Here is a typical usage on the transmit path:
1893 * struct bpf_tunnel_key key;
1895 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1896 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1898 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
1899 * helper for additional information.
1901 * 0 on success, or a negative error in case of failure.
1903 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1905 * Read the value of a perf event counter. This helper relies on a
1906 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1907 * the perf event counter is selected when *map* is updated with
1908 * perf event file descriptors. The *map* is an array whose size
1909 * is the number of available CPUs, and each cell contains a value
1910 * relative to one CPU. The value to retrieve is indicated by
1911 * *flags*, that contains the index of the CPU to look up, masked
1912 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1913 * **BPF_F_CURRENT_CPU** to indicate that the value for the
1914 * current CPU should be retrieved.
1916 * Note that before Linux 4.13, only hardware perf event can be
1919 * Also, be aware that the newer helper
1920 * **bpf_perf_event_read_value**\ () is recommended over
1921 * **bpf_perf_event_read**\ () in general. The latter has some ABI
1922 * quirks where error and counter value are used as a return code
1923 * (which is wrong to do since ranges may overlap). This issue is
1924 * fixed with **bpf_perf_event_read_value**\ (), which at the same
1925 * time provides more features over the **bpf_perf_event_read**\
1926 * () interface. Please refer to the description of
1927 * **bpf_perf_event_read_value**\ () for details.
1929 * The value of the perf event counter read from the map, or a
1930 * negative error code in case of failure.
1932 * long bpf_redirect(u32 ifindex, u64 flags)
1934 * Redirect the packet to another net device of index *ifindex*.
1935 * This helper is somewhat similar to **bpf_clone_redirect**\
1936 * (), except that the packet is not cloned, which provides
1937 * increased performance.
1939 * Except for XDP, both ingress and egress interfaces can be used
1940 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
1941 * to make the distinction (ingress path is selected if the flag
1942 * is present, egress path otherwise). Currently, XDP only
1943 * supports redirection to the egress interface, and accepts no
1946 * The same effect can also be attained with the more generic
1947 * **bpf_redirect_map**\ (), which uses a BPF map to store the
1948 * redirect target instead of providing it directly to the helper.
1950 * For XDP, the helper returns **XDP_REDIRECT** on success or
1951 * **XDP_ABORTED** on error. For other program types, the values
1952 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1955 * u32 bpf_get_route_realm(struct sk_buff *skb)
1957 * Retrieve the realm or the route, that is to say the
1958 * **tclassid** field of the destination for the *skb*. The
1959 * identifier retrieved is a user-provided tag, similar to the
1960 * one used with the net_cls cgroup (see description for
1961 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
1962 * held by a route (a destination entry), not by a task.
1964 * Retrieving this identifier works with the clsact TC egress hook
1965 * (see also **tc-bpf(8)**), or alternatively on conventional
1966 * classful egress qdiscs, but not on TC ingress path. In case of
1967 * clsact TC egress hook, this has the advantage that, internally,
1968 * the destination entry has not been dropped yet in the transmit
1969 * path. Therefore, the destination entry does not need to be
1970 * artificially held via **netif_keep_dst**\ () for a classful
1971 * qdisc until the *skb* is freed.
1973 * This helper is available only if the kernel was compiled with
1974 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
1976 * The realm of the route for the packet associated to *skb*, or 0
1977 * if none was found.
1979 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
1981 * Write raw *data* blob into a special BPF perf event held by
1982 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
1983 * event must have the following attributes: **PERF_SAMPLE_RAW**
1984 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
1985 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
1987 * The *flags* are used to indicate the index in *map* for which
1988 * the value must be put, masked with **BPF_F_INDEX_MASK**.
1989 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
1990 * to indicate that the index of the current CPU core should be
1993 * The value to write, of *size*, is passed through eBPF stack and
1994 * pointed by *data*.
1996 * The context of the program *ctx* needs also be passed to the
1999 * On user space, a program willing to read the values needs to
2000 * call **perf_event_open**\ () on the perf event (either for
2001 * one or for all CPUs) and to store the file descriptor into the
2002 * *map*. This must be done before the eBPF program can send data
2003 * into it. An example is available in file
2004 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2005 * tree (the eBPF program counterpart is in
2006 * *samples/bpf/trace_output_kern.c*).
2008 * **bpf_perf_event_output**\ () achieves better performance
2009 * than **bpf_trace_printk**\ () for sharing data with user
2010 * space, and is much better suitable for streaming data from eBPF
2013 * Note that this helper is not restricted to tracing use cases
2014 * and can be used with programs attached to TC or XDP as well,
2015 * where it allows for passing data to user space listeners. Data
2018 * * Only custom structs,
2019 * * Only the packet payload, or
2020 * * A combination of both.
2022 * 0 on success, or a negative error in case of failure.
2024 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2026 * This helper was provided as an easy way to load data from a
2027 * packet. It can be used to load *len* bytes from *offset* from
2028 * the packet associated to *skb*, into the buffer pointed by
2031 * Since Linux 4.7, usage of this helper has mostly been replaced
2032 * by "direct packet access", enabling packet data to be
2033 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2034 * pointing respectively to the first byte of packet data and to
2035 * the byte after the last byte of packet data. However, it
2036 * remains useful if one wishes to read large quantities of data
2037 * at once from a packet into the eBPF stack.
2039 * 0 on success, or a negative error in case of failure.
2041 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2043 * Walk a user or a kernel stack and return its id. To achieve
2044 * this, the helper needs *ctx*, which is a pointer to the context
2045 * on which the tracing program is executed, and a pointer to a
2046 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2048 * The last argument, *flags*, holds the number of stack frames to
2049 * skip (from 0 to 255), masked with
2050 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2051 * a combination of the following flags:
2053 * **BPF_F_USER_STACK**
2054 * Collect a user space stack instead of a kernel stack.
2055 * **BPF_F_FAST_STACK_CMP**
2056 * Compare stacks by hash only.
2057 * **BPF_F_REUSE_STACKID**
2058 * If two different stacks hash into the same *stackid*,
2059 * discard the old one.
2061 * The stack id retrieved is a 32 bit long integer handle which
2062 * can be further combined with other data (including other stack
2063 * ids) and used as a key into maps. This can be useful for
2064 * generating a variety of graphs (such as flame graphs or off-cpu
2067 * For walking a stack, this helper is an improvement over
2068 * **bpf_probe_read**\ (), which can be used with unrolled loops
2069 * but is not efficient and consumes a lot of eBPF instructions.
2070 * Instead, **bpf_get_stackid**\ () can collect up to
2071 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2072 * this limit can be controlled with the **sysctl** program, and
2073 * that it should be manually increased in order to profile long
2074 * user stacks (such as stacks for Java programs). To do so, use:
2078 * # sysctl kernel.perf_event_max_stack=<new value>
2080 * The positive or null stack id on success, or a negative error
2081 * in case of failure.
2083 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2085 * Compute a checksum difference, from the raw buffer pointed by
2086 * *from*, of length *from_size* (that must be a multiple of 4),
2087 * towards the raw buffer pointed by *to*, of size *to_size*
2088 * (same remark). An optional *seed* can be added to the value
2089 * (this can be cascaded, the seed may come from a previous call
2092 * This is flexible enough to be used in several ways:
2094 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2095 * checksum, it can be used when pushing new data.
2096 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2097 * checksum, it can be used when removing data from a packet.
2098 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2099 * can be used to compute a diff. Note that *from_size* and
2100 * *to_size* do not need to be equal.
2102 * This helper can be used in combination with
2103 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2104 * which one can feed in the difference computed with
2105 * **bpf_csum_diff**\ ().
2107 * The checksum result, or a negative error code in case of
2110 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2112 * Retrieve tunnel options metadata for the packet associated to
2113 * *skb*, and store the raw tunnel option data to the buffer *opt*
2116 * This helper can be used with encapsulation devices that can
2117 * operate in "collect metadata" mode (please refer to the related
2118 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2119 * more details). A particular example where this can be used is
2120 * in combination with the Geneve encapsulation protocol, where it
2121 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2122 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2123 * the eBPF program. This allows for full customization of these
2126 * The size of the option data retrieved.
2128 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2130 * Set tunnel options metadata for the packet associated to *skb*
2131 * to the option data contained in the raw buffer *opt* of *size*.
2133 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2134 * helper for additional information.
2136 * 0 on success, or a negative error in case of failure.
2138 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2140 * Change the protocol of the *skb* to *proto*. Currently
2141 * supported are transition from IPv4 to IPv6, and from IPv6 to
2142 * IPv4. The helper takes care of the groundwork for the
2143 * transition, including resizing the socket buffer. The eBPF
2144 * program is expected to fill the new headers, if any, via
2145 * **skb_store_bytes**\ () and to recompute the checksums with
2146 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2147 * (). The main case for this helper is to perform NAT64
2148 * operations out of an eBPF program.
2150 * Internally, the GSO type is marked as dodgy so that headers are
2151 * checked and segments are recalculated by the GSO/GRO engine.
2152 * The size for GSO target is adapted as well.
2154 * All values for *flags* are reserved for future usage, and must
2157 * A call to this helper is susceptible to change the underlying
2158 * packet buffer. Therefore, at load time, all checks on pointers
2159 * previously done by the verifier are invalidated and must be
2160 * performed again, if the helper is used in combination with
2161 * direct packet access.
2163 * 0 on success, or a negative error in case of failure.
2165 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2167 * Change the packet type for the packet associated to *skb*. This
2168 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2169 * the eBPF program does not have a write access to *skb*\
2170 * **->pkt_type** beside this helper. Using a helper here allows
2171 * for graceful handling of errors.
2173 * The major use case is to change incoming *skb*s to
2174 * **PACKET_HOST** in a programmatic way instead of having to
2175 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2178 * Note that *type* only allows certain values. At this time, they
2183 * **PACKET_BROADCAST**
2184 * Send packet to all.
2185 * **PACKET_MULTICAST**
2186 * Send packet to group.
2187 * **PACKET_OTHERHOST**
2188 * Send packet to someone else.
2190 * 0 on success, or a negative error in case of failure.
2192 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2194 * Check whether *skb* is a descendant of the cgroup2 held by
2195 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2197 * The return value depends on the result of the test, and can be:
2199 * * 0, if the *skb* failed the cgroup2 descendant test.
2200 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2201 * * A negative error code, if an error occurred.
2203 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2205 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2206 * not set, in particular if the hash was cleared due to mangling,
2207 * recompute this hash. Later accesses to the hash can be done
2208 * directly with *skb*\ **->hash**.
2210 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2211 * prototype with **bpf_skb_change_proto**\ (), or calling
2212 * **bpf_skb_store_bytes**\ () with the
2213 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2214 * the hash and to trigger a new computation for the next call to
2215 * **bpf_get_hash_recalc**\ ().
2219 * u64 bpf_get_current_task(void)
2221 * A pointer to the current task struct.
2223 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2225 * Attempt in a safe way to write *len* bytes from the buffer
2226 * *src* to *dst* in memory. It only works for threads that are in
2227 * user context, and *dst* must be a valid user space address.
2229 * This helper should not be used to implement any kind of
2230 * security mechanism because of TOC-TOU attacks, but rather to
2231 * debug, divert, and manipulate execution of semi-cooperative
2234 * Keep in mind that this feature is meant for experiments, and it
2235 * has a risk of crashing the system and running programs.
2236 * Therefore, when an eBPF program using this helper is attached,
2237 * a warning including PID and process name is printed to kernel
2240 * 0 on success, or a negative error in case of failure.
2242 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2244 * Check whether the probe is being run is the context of a given
2245 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2246 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2248 * The return value depends on the result of the test, and can be:
2250 * * 0, if current task belongs to the cgroup2.
2251 * * 1, if current task does not belong to the cgroup2.
2252 * * A negative error code, if an error occurred.
2254 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2256 * Resize (trim or grow) the packet associated to *skb* to the
2257 * new *len*. The *flags* are reserved for future usage, and must
2260 * The basic idea is that the helper performs the needed work to
2261 * change the size of the packet, then the eBPF program rewrites
2262 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2263 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2264 * and others. This helper is a slow path utility intended for
2265 * replies with control messages. And because it is targeted for
2266 * slow path, the helper itself can afford to be slow: it
2267 * implicitly linearizes, unclones and drops offloads from the
2270 * A call to this helper is susceptible to change the underlying
2271 * packet buffer. Therefore, at load time, all checks on pointers
2272 * previously done by the verifier are invalidated and must be
2273 * performed again, if the helper is used in combination with
2274 * direct packet access.
2276 * 0 on success, or a negative error in case of failure.
2278 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2280 * Pull in non-linear data in case the *skb* is non-linear and not
2281 * all of *len* are part of the linear section. Make *len* bytes
2282 * from *skb* readable and writable. If a zero value is passed for
2283 * *len*, then the whole length of the *skb* is pulled.
2285 * This helper is only needed for reading and writing with direct
2288 * For direct packet access, testing that offsets to access
2289 * are within packet boundaries (test on *skb*\ **->data_end**) is
2290 * susceptible to fail if offsets are invalid, or if the requested
2291 * data is in non-linear parts of the *skb*. On failure the
2292 * program can just bail out, or in the case of a non-linear
2293 * buffer, use a helper to make the data available. The
2294 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2295 * the data. Another one consists in using **bpf_skb_pull_data**
2296 * to pull in once the non-linear parts, then retesting and
2297 * eventually access the data.
2299 * At the same time, this also makes sure the *skb* is uncloned,
2300 * which is a necessary condition for direct write. As this needs
2301 * to be an invariant for the write part only, the verifier
2302 * detects writes and adds a prologue that is calling
2303 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2304 * the very beginning in case it is indeed cloned.
2306 * A call to this helper is susceptible to change the underlying
2307 * packet buffer. Therefore, at load time, all checks on pointers
2308 * previously done by the verifier are invalidated and must be
2309 * performed again, if the helper is used in combination with
2310 * direct packet access.
2312 * 0 on success, or a negative error in case of failure.
2314 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2316 * Add the checksum *csum* into *skb*\ **->csum** in case the
2317 * driver has supplied a checksum for the entire packet into that
2318 * field. Return an error otherwise. This helper is intended to be
2319 * used in combination with **bpf_csum_diff**\ (), in particular
2320 * when the checksum needs to be updated after data has been
2321 * written into the packet through direct packet access.
2323 * The checksum on success, or a negative error code in case of
2326 * void bpf_set_hash_invalid(struct sk_buff *skb)
2328 * Invalidate the current *skb*\ **->hash**. It can be used after
2329 * mangling on headers through direct packet access, in order to
2330 * indicate that the hash is outdated and to trigger a
2331 * recalculation the next time the kernel tries to access this
2332 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2334 * long bpf_get_numa_node_id(void)
2336 * Return the id of the current NUMA node. The primary use case
2337 * for this helper is the selection of sockets for the local NUMA
2338 * node, when the program is attached to sockets using the
2339 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2340 * but the helper is also available to other eBPF program types,
2341 * similarly to **bpf_get_smp_processor_id**\ ().
2343 * The id of current NUMA node.
2345 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2347 * Grows headroom of packet associated to *skb* and adjusts the
2348 * offset of the MAC header accordingly, adding *len* bytes of
2349 * space. It automatically extends and reallocates memory as
2352 * This helper can be used on a layer 3 *skb* to push a MAC header
2353 * for redirection into a layer 2 device.
2355 * All values for *flags* are reserved for future usage, and must
2358 * A call to this helper is susceptible to change the underlying
2359 * packet buffer. Therefore, at load time, all checks on pointers
2360 * previously done by the verifier are invalidated and must be
2361 * performed again, if the helper is used in combination with
2362 * direct packet access.
2364 * 0 on success, or a negative error in case of failure.
2366 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2368 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2369 * it is possible to use a negative value for *delta*. This helper
2370 * can be used to prepare the packet for pushing or popping
2373 * A call to this helper is susceptible to change the underlying
2374 * packet buffer. Therefore, at load time, all checks on pointers
2375 * previously done by the verifier are invalidated and must be
2376 * performed again, if the helper is used in combination with
2377 * direct packet access.
2379 * 0 on success, or a negative error in case of failure.
2381 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2383 * Copy a NUL terminated string from an unsafe kernel address
2384 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2387 * Generally, use **bpf_probe_read_user_str**\ () or
2388 * **bpf_probe_read_kernel_str**\ () instead.
2390 * On success, the strictly positive length of the string,
2391 * including the trailing NUL character. On error, a negative
2394 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2396 * If the **struct sk_buff** pointed by *skb* has a known socket,
2397 * retrieve the cookie (generated by the kernel) of this socket.
2398 * If no cookie has been set yet, generate a new cookie. Once
2399 * generated, the socket cookie remains stable for the life of the
2400 * socket. This helper can be useful for monitoring per socket
2401 * networking traffic statistics as it provides a global socket
2402 * identifier that can be assumed unique.
2404 * A 8-byte long unique number on success, or 0 if the socket
2405 * field is missing inside *skb*.
2407 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2409 * Equivalent to bpf_get_socket_cookie() helper that accepts
2410 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2412 * A 8-byte long unique number.
2414 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2416 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2417 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2419 * A 8-byte long unique number.
2421 * u64 bpf_get_socket_cookie(struct sock *sk)
2423 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2424 * *sk*, but gets socket from a BTF **struct sock**. This helper
2425 * also works for sleepable programs.
2427 * A 8-byte long unique number or 0 if *sk* is NULL.
2429 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2431 * The owner UID of the socket associated to *skb*. If the socket
2432 * is **NULL**, or if it is not a full socket (i.e. if it is a
2433 * time-wait or a request socket instead), **overflowuid** value
2434 * is returned (note that **overflowuid** might also be the actual
2435 * UID value for the socket).
2437 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2439 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2444 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2446 * Emulate a call to **setsockopt()** on the socket associated to
2447 * *bpf_socket*, which must be a full socket. The *level* at
2448 * which the option resides and the name *optname* of the option
2449 * must be specified, see **setsockopt(2)** for more information.
2450 * The option value of length *optlen* is pointed by *optval*.
2452 * *bpf_socket* should be one of the following:
2454 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2455 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2456 * and **BPF_CGROUP_INET6_CONNECT**.
2458 * This helper actually implements a subset of **setsockopt()**.
2459 * It supports the following *level*\ s:
2461 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2462 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2463 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2464 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2465 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2466 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2467 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2468 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2469 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2470 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2471 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2473 * 0 on success, or a negative error in case of failure.
2475 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2477 * Grow or shrink the room for data in the packet associated to
2478 * *skb* by *len_diff*, and according to the selected *mode*.
2480 * By default, the helper will reset any offloaded checksum
2481 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2482 * by the following flag:
2484 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2485 * checksum data of the skb to CHECKSUM_NONE.
2487 * There are two supported modes at this time:
2489 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2490 * (room space is added or removed below the layer 2 header).
2492 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2493 * (room space is added or removed below the layer 3 header).
2495 * The following flags are supported at this time:
2497 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2498 * Adjusting mss in this way is not allowed for datagrams.
2500 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2501 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2502 * Any new space is reserved to hold a tunnel header.
2503 * Configure skb offsets and other fields accordingly.
2505 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2506 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2507 * Use with ENCAP_L3 flags to further specify the tunnel type.
2509 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2510 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2511 * type; *len* is the length of the inner MAC header.
2513 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2514 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2515 * L2 type as Ethernet.
2517 * A call to this helper is susceptible to change the underlying
2518 * packet buffer. Therefore, at load time, all checks on pointers
2519 * previously done by the verifier are invalidated and must be
2520 * performed again, if the helper is used in combination with
2521 * direct packet access.
2523 * 0 on success, or a negative error in case of failure.
2525 * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2527 * Redirect the packet to the endpoint referenced by *map* at
2528 * index *key*. Depending on its type, this *map* can contain
2529 * references to net devices (for forwarding packets through other
2530 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2531 * but this is only implemented for native XDP (with driver
2532 * support) as of this writing).
2534 * The lower two bits of *flags* are used as the return code if
2535 * the map lookup fails. This is so that the return value can be
2536 * one of the XDP program return codes up to **XDP_TX**, as chosen
2537 * by the caller. Any higher bits in the *flags* argument must be
2540 * See also **bpf_redirect**\ (), which only supports redirecting
2541 * to an ifindex, but doesn't require a map to do so.
2543 * **XDP_REDIRECT** on success, or the value of the two lower bits
2544 * of the *flags* argument on error.
2546 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2548 * Redirect the packet to the socket referenced by *map* (of type
2549 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2550 * egress interfaces can be used for redirection. The
2551 * **BPF_F_INGRESS** value in *flags* is used to make the
2552 * distinction (ingress path is selected if the flag is present,
2553 * egress path otherwise). This is the only flag supported for now.
2555 * **SK_PASS** on success, or **SK_DROP** on error.
2557 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2559 * Add an entry to, or update a *map* referencing sockets. The
2560 * *skops* is used as a new value for the entry associated to
2561 * *key*. *flags* is one of:
2564 * The entry for *key* must not exist in the map.
2566 * The entry for *key* must already exist in the map.
2568 * No condition on the existence of the entry for *key*.
2570 * If the *map* has eBPF programs (parser and verdict), those will
2571 * be inherited by the socket being added. If the socket is
2572 * already attached to eBPF programs, this results in an error.
2574 * 0 on success, or a negative error in case of failure.
2576 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2578 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2579 * *delta* (which can be positive or negative). Note that this
2580 * operation modifies the address stored in *xdp_md*\ **->data**,
2581 * so the latter must be loaded only after the helper has been
2584 * The use of *xdp_md*\ **->data_meta** is optional and programs
2585 * are not required to use it. The rationale is that when the
2586 * packet is processed with XDP (e.g. as DoS filter), it is
2587 * possible to push further meta data along with it before passing
2588 * to the stack, and to give the guarantee that an ingress eBPF
2589 * program attached as a TC classifier on the same device can pick
2590 * this up for further post-processing. Since TC works with socket
2591 * buffers, it remains possible to set from XDP the **mark** or
2592 * **priority** pointers, or other pointers for the socket buffer.
2593 * Having this scratch space generic and programmable allows for
2594 * more flexibility as the user is free to store whatever meta
2597 * A call to this helper is susceptible to change the underlying
2598 * packet buffer. Therefore, at load time, all checks on pointers
2599 * previously done by the verifier are invalidated and must be
2600 * performed again, if the helper is used in combination with
2601 * direct packet access.
2603 * 0 on success, or a negative error in case of failure.
2605 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2607 * Read the value of a perf event counter, and store it into *buf*
2608 * of size *buf_size*. This helper relies on a *map* of type
2609 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2610 * counter is selected when *map* is updated with perf event file
2611 * descriptors. The *map* is an array whose size is the number of
2612 * available CPUs, and each cell contains a value relative to one
2613 * CPU. The value to retrieve is indicated by *flags*, that
2614 * contains the index of the CPU to look up, masked with
2615 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2616 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2617 * current CPU should be retrieved.
2619 * This helper behaves in a way close to
2620 * **bpf_perf_event_read**\ () helper, save that instead of
2621 * just returning the value observed, it fills the *buf*
2622 * structure. This allows for additional data to be retrieved: in
2623 * particular, the enabled and running times (in *buf*\
2624 * **->enabled** and *buf*\ **->running**, respectively) are
2625 * copied. In general, **bpf_perf_event_read_value**\ () is
2626 * recommended over **bpf_perf_event_read**\ (), which has some
2627 * ABI issues and provides fewer functionalities.
2629 * These values are interesting, because hardware PMU (Performance
2630 * Monitoring Unit) counters are limited resources. When there are
2631 * more PMU based perf events opened than available counters,
2632 * kernel will multiplex these events so each event gets certain
2633 * percentage (but not all) of the PMU time. In case that
2634 * multiplexing happens, the number of samples or counter value
2635 * will not reflect the case compared to when no multiplexing
2636 * occurs. This makes comparison between different runs difficult.
2637 * Typically, the counter value should be normalized before
2638 * comparing to other experiments. The usual normalization is done
2643 * normalized_counter = counter * t_enabled / t_running
2645 * Where t_enabled is the time enabled for event and t_running is
2646 * the time running for event since last normalization. The
2647 * enabled and running times are accumulated since the perf event
2648 * open. To achieve scaling factor between two invocations of an
2649 * eBPF program, users can use CPU id as the key (which is
2650 * typical for perf array usage model) to remember the previous
2651 * value and do the calculation inside the eBPF program.
2653 * 0 on success, or a negative error in case of failure.
2655 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2657 * For en eBPF program attached to a perf event, retrieve the
2658 * value of the event counter associated to *ctx* and store it in
2659 * the structure pointed by *buf* and of size *buf_size*. Enabled
2660 * and running times are also stored in the structure (see
2661 * description of helper **bpf_perf_event_read_value**\ () for
2664 * 0 on success, or a negative error in case of failure.
2666 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2668 * Emulate a call to **getsockopt()** on the socket associated to
2669 * *bpf_socket*, which must be a full socket. The *level* at
2670 * which the option resides and the name *optname* of the option
2671 * must be specified, see **getsockopt(2)** for more information.
2672 * The retrieved value is stored in the structure pointed by
2673 * *opval* and of length *optlen*.
2675 * *bpf_socket* should be one of the following:
2677 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2678 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2679 * and **BPF_CGROUP_INET6_CONNECT**.
2681 * This helper actually implements a subset of **getsockopt()**.
2682 * It supports the following *level*\ s:
2684 * * **IPPROTO_TCP**, which supports *optname*
2685 * **TCP_CONGESTION**.
2686 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2687 * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2689 * 0 on success, or a negative error in case of failure.
2691 * long bpf_override_return(struct pt_regs *regs, u64 rc)
2693 * Used for error injection, this helper uses kprobes to override
2694 * the return value of the probed function, and to set it to *rc*.
2695 * The first argument is the context *regs* on which the kprobe
2698 * This helper works by setting the PC (program counter)
2699 * to an override function which is run in place of the original
2700 * probed function. This means the probed function is not run at
2701 * all. The replacement function just returns with the required
2704 * This helper has security implications, and thus is subject to
2705 * restrictions. It is only available if the kernel was compiled
2706 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2707 * option, and in this case it only works on functions tagged with
2708 * **ALLOW_ERROR_INJECTION** in the kernel code.
2710 * Also, the helper is only available for the architectures having
2711 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2712 * x86 architecture is the only one to support this feature.
2716 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2718 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2719 * for the full TCP socket associated to *bpf_sock_ops* to
2722 * The primary use of this field is to determine if there should
2723 * be calls to eBPF programs of type
2724 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2725 * code. A program of the same type can change its value, per
2726 * connection and as necessary, when the connection is
2727 * established. This field is directly accessible for reading, but
2728 * this helper must be used for updates in order to return an
2729 * error if an eBPF program tries to set a callback that is not
2730 * supported in the current kernel.
2732 * *argval* is a flag array which can combine these flags:
2734 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2735 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2736 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2737 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2739 * Therefore, this function can be used to clear a callback flag by
2740 * setting the appropriate bit to zero. e.g. to disable the RTO
2743 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
2744 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2746 * Here are some examples of where one could call such eBPF
2750 * * When a packet is retransmitted.
2751 * * When the connection terminates.
2752 * * When a packet is sent.
2753 * * When a packet is received.
2755 * Code **-EINVAL** if the socket is not a full TCP socket;
2756 * otherwise, a positive number containing the bits that could not
2757 * be set is returned (which comes down to 0 if all bits were set
2760 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2762 * This helper is used in programs implementing policies at the
2763 * socket level. If the message *msg* is allowed to pass (i.e. if
2764 * the verdict eBPF program returns **SK_PASS**), redirect it to
2765 * the socket referenced by *map* (of type
2766 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2767 * egress interfaces can be used for redirection. The
2768 * **BPF_F_INGRESS** value in *flags* is used to make the
2769 * distinction (ingress path is selected if the flag is present,
2770 * egress path otherwise). This is the only flag supported for now.
2772 * **SK_PASS** on success, or **SK_DROP** on error.
2774 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2776 * For socket policies, apply the verdict of the eBPF program to
2777 * the next *bytes* (number of bytes) of message *msg*.
2779 * For example, this helper can be used in the following cases:
2781 * * A single **sendmsg**\ () or **sendfile**\ () system call
2782 * contains multiple logical messages that the eBPF program is
2783 * supposed to read and for which it should apply a verdict.
2784 * * An eBPF program only cares to read the first *bytes* of a
2785 * *msg*. If the message has a large payload, then setting up
2786 * and calling the eBPF program repeatedly for all bytes, even
2787 * though the verdict is already known, would create unnecessary
2790 * When called from within an eBPF program, the helper sets a
2791 * counter internal to the BPF infrastructure, that is used to
2792 * apply the last verdict to the next *bytes*. If *bytes* is
2793 * smaller than the current data being processed from a
2794 * **sendmsg**\ () or **sendfile**\ () system call, the first
2795 * *bytes* will be sent and the eBPF program will be re-run with
2796 * the pointer for start of data pointing to byte number *bytes*
2797 * **+ 1**. If *bytes* is larger than the current data being
2798 * processed, then the eBPF verdict will be applied to multiple
2799 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2802 * Note that if a socket closes with the internal counter holding
2803 * a non-zero value, this is not a problem because data is not
2804 * being buffered for *bytes* and is sent as it is received.
2808 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2810 * For socket policies, prevent the execution of the verdict eBPF
2811 * program for message *msg* until *bytes* (byte number) have been
2814 * This can be used when one needs a specific number of bytes
2815 * before a verdict can be assigned, even if the data spans
2816 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2817 * case would be a user calling **sendmsg**\ () repeatedly with
2818 * 1-byte long message segments. Obviously, this is bad for
2819 * performance, but it is still valid. If the eBPF program needs
2820 * *bytes* bytes to validate a header, this helper can be used to
2821 * prevent the eBPF program to be called again until *bytes* have
2826 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2828 * For socket policies, pull in non-linear data from user space
2829 * for *msg* and set pointers *msg*\ **->data** and *msg*\
2830 * **->data_end** to *start* and *end* bytes offsets into *msg*,
2833 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2834 * *msg* it can only parse data that the (**data**, **data_end**)
2835 * pointers have already consumed. For **sendmsg**\ () hooks this
2836 * is likely the first scatterlist element. But for calls relying
2837 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2838 * be the range (**0**, **0**) because the data is shared with
2839 * user space and by default the objective is to avoid allowing
2840 * user space to modify data while (or after) eBPF verdict is
2841 * being decided. This helper can be used to pull in data and to
2842 * set the start and end pointer to given values. Data will be
2843 * copied if necessary (i.e. if data was not linear and if start
2844 * and end pointers do not point to the same chunk).
2846 * A call to this helper is susceptible to change the underlying
2847 * packet buffer. Therefore, at load time, all checks on pointers
2848 * previously done by the verifier are invalidated and must be
2849 * performed again, if the helper is used in combination with
2850 * direct packet access.
2852 * All values for *flags* are reserved for future usage, and must
2855 * 0 on success, or a negative error in case of failure.
2857 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2859 * Bind the socket associated to *ctx* to the address pointed by
2860 * *addr*, of length *addr_len*. This allows for making outgoing
2861 * connection from the desired IP address, which can be useful for
2862 * example when all processes inside a cgroup should use one
2863 * single IP address on a host that has multiple IP configured.
2865 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2866 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2867 * **AF_INET6**). It's advised to pass zero port (**sin_port**
2868 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2869 * behavior and lets the kernel efficiently pick up an unused
2870 * port as long as 4-tuple is unique. Passing non-zero port might
2871 * lead to degraded performance.
2873 * 0 on success, or a negative error in case of failure.
2875 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2877 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2878 * possible to both shrink and grow the packet tail.
2879 * Shrink done via *delta* being a negative integer.
2881 * A call to this helper is susceptible to change the underlying
2882 * packet buffer. Therefore, at load time, all checks on pointers
2883 * previously done by the verifier are invalidated and must be
2884 * performed again, if the helper is used in combination with
2885 * direct packet access.
2887 * 0 on success, or a negative error in case of failure.
2889 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2891 * Retrieve the XFRM state (IP transform framework, see also
2892 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2894 * The retrieved value is stored in the **struct bpf_xfrm_state**
2895 * pointed by *xfrm_state* and of length *size*.
2897 * All values for *flags* are reserved for future usage, and must
2900 * This helper is available only if the kernel was compiled with
2901 * **CONFIG_XFRM** configuration option.
2903 * 0 on success, or a negative error in case of failure.
2905 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2907 * Return a user or a kernel stack in bpf program provided buffer.
2908 * To achieve this, the helper needs *ctx*, which is a pointer
2909 * to the context on which the tracing program is executed.
2910 * To store the stacktrace, the bpf program provides *buf* with
2911 * a nonnegative *size*.
2913 * The last argument, *flags*, holds the number of stack frames to
2914 * skip (from 0 to 255), masked with
2915 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2916 * the following flags:
2918 * **BPF_F_USER_STACK**
2919 * Collect a user space stack instead of a kernel stack.
2920 * **BPF_F_USER_BUILD_ID**
2921 * Collect buildid+offset instead of ips for user stack,
2922 * only valid if **BPF_F_USER_STACK** is also specified.
2924 * **bpf_get_stack**\ () can collect up to
2925 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2926 * to sufficient large buffer size. Note that
2927 * this limit can be controlled with the **sysctl** program, and
2928 * that it should be manually increased in order to profile long
2929 * user stacks (such as stacks for Java programs). To do so, use:
2933 * # sysctl kernel.perf_event_max_stack=<new value>
2935 * A non-negative value equal to or less than *size* on success,
2936 * or a negative error in case of failure.
2938 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2940 * This helper is similar to **bpf_skb_load_bytes**\ () in that
2941 * it provides an easy way to load *len* bytes from *offset*
2942 * from the packet associated to *skb*, into the buffer pointed
2943 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2944 * a fifth argument *start_header* exists in order to select a
2945 * base offset to start from. *start_header* can be one of:
2947 * **BPF_HDR_START_MAC**
2948 * Base offset to load data from is *skb*'s mac header.
2949 * **BPF_HDR_START_NET**
2950 * Base offset to load data from is *skb*'s network header.
2952 * In general, "direct packet access" is the preferred method to
2953 * access packet data, however, this helper is in particular useful
2954 * in socket filters where *skb*\ **->data** does not always point
2955 * to the start of the mac header and where "direct packet access"
2958 * 0 on success, or a negative error in case of failure.
2960 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
2962 * Do FIB lookup in kernel tables using parameters in *params*.
2963 * If lookup is successful and result shows packet is to be
2964 * forwarded, the neighbor tables are searched for the nexthop.
2965 * If successful (ie., FIB lookup shows forwarding and nexthop
2966 * is resolved), the nexthop address is returned in ipv4_dst
2967 * or ipv6_dst based on family, smac is set to mac address of
2968 * egress device, dmac is set to nexthop mac address, rt_metric
2969 * is set to metric from route (IPv4/IPv6 only), and ifindex
2970 * is set to the device index of the nexthop from the FIB lookup.
2972 * *plen* argument is the size of the passed in struct.
2973 * *flags* argument can be a combination of one or more of the
2976 * **BPF_FIB_LOOKUP_DIRECT**
2977 * Do a direct table lookup vs full lookup using FIB
2979 * **BPF_FIB_LOOKUP_OUTPUT**
2980 * Perform lookup from an egress perspective (default is
2983 * *ctx* is either **struct xdp_md** for XDP programs or
2984 * **struct sk_buff** tc cls_act programs.
2986 * * < 0 if any input argument is invalid
2987 * * 0 on success (packet is forwarded, nexthop neighbor exists)
2988 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
2989 * packet is not forwarded or needs assist from full stack
2991 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
2992 * was exceeded and output params->mtu_result contains the MTU.
2994 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2996 * Add an entry to, or update a sockhash *map* referencing sockets.
2997 * The *skops* is used as a new value for the entry associated to
2998 * *key*. *flags* is one of:
3001 * The entry for *key* must not exist in the map.
3003 * The entry for *key* must already exist in the map.
3005 * No condition on the existence of the entry for *key*.
3007 * If the *map* has eBPF programs (parser and verdict), those will
3008 * be inherited by the socket being added. If the socket is
3009 * already attached to eBPF programs, this results in an error.
3011 * 0 on success, or a negative error in case of failure.
3013 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3015 * This helper is used in programs implementing policies at the
3016 * socket level. If the message *msg* is allowed to pass (i.e. if
3017 * the verdict eBPF program returns **SK_PASS**), redirect it to
3018 * the socket referenced by *map* (of type
3019 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3020 * egress interfaces can be used for redirection. The
3021 * **BPF_F_INGRESS** value in *flags* is used to make the
3022 * distinction (ingress path is selected if the flag is present,
3023 * egress path otherwise). This is the only flag supported for now.
3025 * **SK_PASS** on success, or **SK_DROP** on error.
3027 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3029 * This helper is used in programs implementing policies at the
3030 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3031 * if the verdict eBPF program returns **SK_PASS**), redirect it
3032 * to the socket referenced by *map* (of type
3033 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3034 * egress interfaces can be used for redirection. The
3035 * **BPF_F_INGRESS** value in *flags* is used to make the
3036 * distinction (ingress path is selected if the flag is present,
3037 * egress otherwise). This is the only flag supported for now.
3039 * **SK_PASS** on success, or **SK_DROP** on error.
3041 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3043 * Encapsulate the packet associated to *skb* within a Layer 3
3044 * protocol header. This header is provided in the buffer at
3045 * address *hdr*, with *len* its size in bytes. *type* indicates
3046 * the protocol of the header and can be one of:
3048 * **BPF_LWT_ENCAP_SEG6**
3049 * IPv6 encapsulation with Segment Routing Header
3050 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3051 * the IPv6 header is computed by the kernel.
3052 * **BPF_LWT_ENCAP_SEG6_INLINE**
3053 * Only works if *skb* contains an IPv6 packet. Insert a
3054 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3056 * **BPF_LWT_ENCAP_IP**
3057 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3058 * must be IPv4 or IPv6, followed by zero or more
3059 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3060 * total bytes in all prepended headers. Please note that
3061 * if **skb_is_gso**\ (*skb*) is true, no more than two
3062 * headers can be prepended, and the inner header, if
3063 * present, should be either GRE or UDP/GUE.
3065 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3066 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3067 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3068 * **BPF_PROG_TYPE_LWT_XMIT**.
3070 * A call to this helper is susceptible to change the underlying
3071 * packet buffer. Therefore, at load time, all checks on pointers
3072 * previously done by the verifier are invalidated and must be
3073 * performed again, if the helper is used in combination with
3074 * direct packet access.
3076 * 0 on success, or a negative error in case of failure.
3078 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3080 * Store *len* bytes from address *from* into the packet
3081 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3082 * inside the outermost IPv6 Segment Routing Header can be
3083 * modified through this helper.
3085 * A call to this helper is susceptible to change the underlying
3086 * packet buffer. Therefore, at load time, all checks on pointers
3087 * previously done by the verifier are invalidated and must be
3088 * performed again, if the helper is used in combination with
3089 * direct packet access.
3091 * 0 on success, or a negative error in case of failure.
3093 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3095 * Adjust the size allocated to TLVs in the outermost IPv6
3096 * Segment Routing Header contained in the packet associated to
3097 * *skb*, at position *offset* by *delta* bytes. Only offsets
3098 * after the segments are accepted. *delta* can be as well
3099 * positive (growing) as negative (shrinking).
3101 * A call to this helper is susceptible to change the underlying
3102 * packet buffer. Therefore, at load time, all checks on pointers
3103 * previously done by the verifier are invalidated and must be
3104 * performed again, if the helper is used in combination with
3105 * direct packet access.
3107 * 0 on success, or a negative error in case of failure.
3109 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3111 * Apply an IPv6 Segment Routing action of type *action* to the
3112 * packet associated to *skb*. Each action takes a parameter
3113 * contained at address *param*, and of length *param_len* bytes.
3114 * *action* can be one of:
3116 * **SEG6_LOCAL_ACTION_END_X**
3117 * End.X action: Endpoint with Layer-3 cross-connect.
3118 * Type of *param*: **struct in6_addr**.
3119 * **SEG6_LOCAL_ACTION_END_T**
3120 * End.T action: Endpoint with specific IPv6 table lookup.
3121 * Type of *param*: **int**.
3122 * **SEG6_LOCAL_ACTION_END_B6**
3123 * End.B6 action: Endpoint bound to an SRv6 policy.
3124 * Type of *param*: **struct ipv6_sr_hdr**.
3125 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3126 * End.B6.Encap action: Endpoint bound to an SRv6
3127 * encapsulation policy.
3128 * Type of *param*: **struct ipv6_sr_hdr**.
3130 * A call to this helper is susceptible to change the underlying
3131 * packet buffer. Therefore, at load time, all checks on pointers
3132 * previously done by the verifier are invalidated and must be
3133 * performed again, if the helper is used in combination with
3134 * direct packet access.
3136 * 0 on success, or a negative error in case of failure.
3138 * long bpf_rc_repeat(void *ctx)
3140 * This helper is used in programs implementing IR decoding, to
3141 * report a successfully decoded repeat key message. This delays
3142 * the generation of a key up event for previously generated
3145 * Some IR protocols like NEC have a special IR message for
3146 * repeating last button, for when a button is held down.
3148 * The *ctx* should point to the lirc sample as passed into
3151 * This helper is only available is the kernel was compiled with
3152 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3157 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3159 * This helper is used in programs implementing IR decoding, to
3160 * report a successfully decoded key press with *scancode*,
3161 * *toggle* value in the given *protocol*. The scancode will be
3162 * translated to a keycode using the rc keymap, and reported as
3163 * an input key down event. After a period a key up event is
3164 * generated. This period can be extended by calling either
3165 * **bpf_rc_keydown**\ () again with the same values, or calling
3166 * **bpf_rc_repeat**\ ().
3168 * Some protocols include a toggle bit, in case the button was
3169 * released and pressed again between consecutive scancodes.
3171 * The *ctx* should point to the lirc sample as passed into
3174 * The *protocol* is the decoded protocol number (see
3175 * **enum rc_proto** for some predefined values).
3177 * This helper is only available is the kernel was compiled with
3178 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3183 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3185 * Return the cgroup v2 id of the socket associated with the *skb*.
3186 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3187 * helper for cgroup v1 by providing a tag resp. identifier that
3188 * can be matched on or used for map lookups e.g. to implement
3189 * policy. The cgroup v2 id of a given path in the hierarchy is
3190 * exposed in user space through the f_handle API in order to get
3191 * to the same 64-bit id.
3193 * This helper can be used on TC egress path, but not on ingress,
3194 * and is available only if the kernel was compiled with the
3195 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3197 * The id is returned or 0 in case the id could not be retrieved.
3199 * u64 bpf_get_current_cgroup_id(void)
3201 * A 64-bit integer containing the current cgroup id based
3202 * on the cgroup within which the current task is running.
3204 * void *bpf_get_local_storage(void *map, u64 flags)
3206 * Get the pointer to the local storage area.
3207 * The type and the size of the local storage is defined
3208 * by the *map* argument.
3209 * The *flags* meaning is specific for each map type,
3210 * and has to be 0 for cgroup local storage.
3212 * Depending on the BPF program type, a local storage area
3213 * can be shared between multiple instances of the BPF program,
3214 * running simultaneously.
3216 * A user should care about the synchronization by himself.
3217 * For example, by using the **BPF_ATOMIC** instructions to alter
3220 * A pointer to the local storage area.
3222 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3224 * Select a **SO_REUSEPORT** socket from a
3225 * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*.
3226 * It checks the selected socket is matching the incoming
3227 * request in the socket buffer.
3229 * 0 on success, or a negative error in case of failure.
3231 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3233 * Return id of cgroup v2 that is ancestor of cgroup associated
3234 * with the *skb* at the *ancestor_level*. The root cgroup is at
3235 * *ancestor_level* zero and each step down the hierarchy
3236 * increments the level. If *ancestor_level* == level of cgroup
3237 * associated with *skb*, then return value will be same as that
3238 * of **bpf_skb_cgroup_id**\ ().
3240 * The helper is useful to implement policies based on cgroups
3241 * that are upper in hierarchy than immediate cgroup associated
3244 * The format of returned id and helper limitations are same as in
3245 * **bpf_skb_cgroup_id**\ ().
3247 * The id is returned or 0 in case the id could not be retrieved.
3249 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3251 * Look for TCP socket matching *tuple*, optionally in a child
3252 * network namespace *netns*. The return value must be checked,
3253 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3255 * The *ctx* should point to the context of the program, such as
3256 * the skb or socket (depending on the hook in use). This is used
3257 * to determine the base network namespace for the lookup.
3259 * *tuple_size* must be one of:
3261 * **sizeof**\ (*tuple*\ **->ipv4**)
3262 * Look for an IPv4 socket.
3263 * **sizeof**\ (*tuple*\ **->ipv6**)
3264 * Look for an IPv6 socket.
3266 * If the *netns* is a negative signed 32-bit integer, then the
3267 * socket lookup table in the netns associated with the *ctx*
3268 * will be used. For the TC hooks, this is the netns of the device
3269 * in the skb. For socket hooks, this is the netns of the socket.
3270 * If *netns* is any other signed 32-bit value greater than or
3271 * equal to zero then it specifies the ID of the netns relative to
3272 * the netns associated with the *ctx*. *netns* values beyond the
3273 * range of 32-bit integers are reserved for future use.
3275 * All values for *flags* are reserved for future usage, and must
3278 * This helper is available only if the kernel was compiled with
3279 * **CONFIG_NET** configuration option.
3281 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3282 * For sockets with reuseport option, the **struct bpf_sock**
3283 * result is from *reuse*\ **->socks**\ [] using the hash of the
3286 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3288 * Look for UDP socket matching *tuple*, optionally in a child
3289 * network namespace *netns*. The return value must be checked,
3290 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3292 * The *ctx* should point to the context of the program, such as
3293 * the skb or socket (depending on the hook in use). This is used
3294 * to determine the base network namespace for the lookup.
3296 * *tuple_size* must be one of:
3298 * **sizeof**\ (*tuple*\ **->ipv4**)
3299 * Look for an IPv4 socket.
3300 * **sizeof**\ (*tuple*\ **->ipv6**)
3301 * Look for an IPv6 socket.
3303 * If the *netns* is a negative signed 32-bit integer, then the
3304 * socket lookup table in the netns associated with the *ctx*
3305 * will be used. For the TC hooks, this is the netns of the device
3306 * in the skb. For socket hooks, this is the netns of the socket.
3307 * If *netns* is any other signed 32-bit value greater than or
3308 * equal to zero then it specifies the ID of the netns relative to
3309 * the netns associated with the *ctx*. *netns* values beyond the
3310 * range of 32-bit integers are reserved for future use.
3312 * All values for *flags* are reserved for future usage, and must
3315 * This helper is available only if the kernel was compiled with
3316 * **CONFIG_NET** configuration option.
3318 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3319 * For sockets with reuseport option, the **struct bpf_sock**
3320 * result is from *reuse*\ **->socks**\ [] using the hash of the
3323 * long bpf_sk_release(void *sock)
3325 * Release the reference held by *sock*. *sock* must be a
3326 * non-**NULL** pointer that was returned from
3327 * **bpf_sk_lookup_xxx**\ ().
3329 * 0 on success, or a negative error in case of failure.
3331 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3333 * Push an element *value* in *map*. *flags* is one of:
3336 * If the queue/stack is full, the oldest element is
3337 * removed to make room for this.
3339 * 0 on success, or a negative error in case of failure.
3341 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3343 * Pop an element from *map*.
3345 * 0 on success, or a negative error in case of failure.
3347 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3349 * Get an element from *map* without removing it.
3351 * 0 on success, or a negative error in case of failure.
3353 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3355 * For socket policies, insert *len* bytes into *msg* at offset
3358 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3359 * *msg* it may want to insert metadata or options into the *msg*.
3360 * This can later be read and used by any of the lower layer BPF
3363 * This helper may fail if under memory pressure (a malloc
3364 * fails) in these cases BPF programs will get an appropriate
3365 * error and BPF programs will need to handle them.
3367 * 0 on success, or a negative error in case of failure.
3369 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3371 * Will remove *len* bytes from a *msg* starting at byte *start*.
3372 * This may result in **ENOMEM** errors under certain situations if
3373 * an allocation and copy are required due to a full ring buffer.
3374 * However, the helper will try to avoid doing the allocation
3375 * if possible. Other errors can occur if input parameters are
3376 * invalid either due to *start* byte not being valid part of *msg*
3377 * payload and/or *pop* value being to large.
3379 * 0 on success, or a negative error in case of failure.
3381 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3383 * This helper is used in programs implementing IR decoding, to
3384 * report a successfully decoded pointer movement.
3386 * The *ctx* should point to the lirc sample as passed into
3389 * This helper is only available is the kernel was compiled with
3390 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3395 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3397 * Acquire a spinlock represented by the pointer *lock*, which is
3398 * stored as part of a value of a map. Taking the lock allows to
3399 * safely update the rest of the fields in that value. The
3400 * spinlock can (and must) later be released with a call to
3401 * **bpf_spin_unlock**\ (\ *lock*\ ).
3403 * Spinlocks in BPF programs come with a number of restrictions
3406 * * **bpf_spin_lock** objects are only allowed inside maps of
3407 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3408 * list could be extended in the future).
3409 * * BTF description of the map is mandatory.
3410 * * The BPF program can take ONE lock at a time, since taking two
3411 * or more could cause dead locks.
3412 * * Only one **struct bpf_spin_lock** is allowed per map element.
3413 * * When the lock is taken, calls (either BPF to BPF or helpers)
3415 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3416 * allowed inside a spinlock-ed region.
3417 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3418 * the lock, on all execution paths, before it returns.
3419 * * The BPF program can access **struct bpf_spin_lock** only via
3420 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3421 * helpers. Loading or storing data into the **struct
3422 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3423 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3424 * of the map value must be a struct and have **struct
3425 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3426 * Nested lock inside another struct is not allowed.
3427 * * The **struct bpf_spin_lock** *lock* field in a map value must
3428 * be aligned on a multiple of 4 bytes in that value.
3429 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3430 * the **bpf_spin_lock** field to user space.
3431 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3432 * a BPF program, do not update the **bpf_spin_lock** field.
3433 * * **bpf_spin_lock** cannot be on the stack or inside a
3434 * networking packet (it can only be inside of a map values).
3435 * * **bpf_spin_lock** is available to root only.
3436 * * Tracing programs and socket filter programs cannot use
3437 * **bpf_spin_lock**\ () due to insufficient preemption checks
3438 * (but this may change in the future).
3439 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3443 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3445 * Release the *lock* previously locked by a call to
3446 * **bpf_spin_lock**\ (\ *lock*\ ).
3450 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3452 * This helper gets a **struct bpf_sock** pointer such
3453 * that all the fields in this **bpf_sock** can be accessed.
3455 * A **struct bpf_sock** pointer on success, or **NULL** in
3458 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3460 * This helper gets a **struct bpf_tcp_sock** pointer from a
3461 * **struct bpf_sock** pointer.
3463 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3466 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3468 * Set ECN (Explicit Congestion Notification) field of IP header
3469 * to **CE** (Congestion Encountered) if current value is **ECT**
3470 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3473 * 1 if the **CE** flag is set (either by the current helper call
3474 * or because it was already present), 0 if it is not set.
3476 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3478 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3479 * **bpf_sk_release**\ () is unnecessary and not allowed.
3481 * A **struct bpf_sock** pointer on success, or **NULL** in
3484 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3486 * Look for TCP socket matching *tuple*, optionally in a child
3487 * network namespace *netns*. The return value must be checked,
3488 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3490 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3491 * that it also returns timewait or request sockets. Use
3492 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3495 * This helper is available only if the kernel was compiled with
3496 * **CONFIG_NET** configuration option.
3498 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3499 * For sockets with reuseport option, the **struct bpf_sock**
3500 * result is from *reuse*\ **->socks**\ [] using the hash of the
3503 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3505 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3506 * the listening socket in *sk*.
3508 * *iph* points to the start of the IPv4 or IPv6 header, while
3509 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3510 * **sizeof**\ (**struct ip6hdr**).
3512 * *th* points to the start of the TCP header, while *th_len*
3513 * contains **sizeof**\ (**struct tcphdr**).
3515 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3518 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3520 * Get name of sysctl in /proc/sys/ and copy it into provided by
3521 * program buffer *buf* of size *buf_len*.
3523 * The buffer is always NUL terminated, unless it's zero-sized.
3525 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3526 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3527 * only (e.g. "tcp_mem").
3529 * Number of character copied (not including the trailing NUL).
3531 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3532 * truncated name in this case).
3534 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3536 * Get current value of sysctl as it is presented in /proc/sys
3537 * (incl. newline, etc), and copy it as a string into provided
3538 * by program buffer *buf* of size *buf_len*.
3540 * The whole value is copied, no matter what file position user
3541 * space issued e.g. sys_read at.
3543 * The buffer is always NUL terminated, unless it's zero-sized.
3545 * Number of character copied (not including the trailing NUL).
3547 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3548 * truncated name in this case).
3550 * **-EINVAL** if current value was unavailable, e.g. because
3551 * sysctl is uninitialized and read returns -EIO for it.
3553 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3555 * Get new value being written by user space to sysctl (before
3556 * the actual write happens) and copy it as a string into
3557 * provided by program buffer *buf* of size *buf_len*.
3559 * User space may write new value at file position > 0.
3561 * The buffer is always NUL terminated, unless it's zero-sized.
3563 * Number of character copied (not including the trailing NUL).
3565 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3566 * truncated name in this case).
3568 * **-EINVAL** if sysctl is being read.
3570 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3572 * Override new value being written by user space to sysctl with
3573 * value provided by program in buffer *buf* of size *buf_len*.
3575 * *buf* should contain a string in same form as provided by user
3576 * space on sysctl write.
3578 * User space may write new value at file position > 0. To override
3579 * the whole sysctl value file position should be set to zero.
3583 * **-E2BIG** if the *buf_len* is too big.
3585 * **-EINVAL** if sysctl is being read.
3587 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3589 * Convert the initial part of the string from buffer *buf* of
3590 * size *buf_len* to a long integer according to the given base
3591 * and save the result in *res*.
3593 * The string may begin with an arbitrary amount of white space
3594 * (as determined by **isspace**\ (3)) followed by a single
3595 * optional '**-**' sign.
3597 * Five least significant bits of *flags* encode base, other bits
3598 * are currently unused.
3600 * Base must be either 8, 10, 16 or 0 to detect it automatically
3601 * similar to user space **strtol**\ (3).
3603 * Number of characters consumed on success. Must be positive but
3604 * no more than *buf_len*.
3606 * **-EINVAL** if no valid digits were found or unsupported base
3609 * **-ERANGE** if resulting value was out of range.
3611 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3613 * Convert the initial part of the string from buffer *buf* of
3614 * size *buf_len* to an unsigned long integer according to the
3615 * given base and save the result in *res*.
3617 * The string may begin with an arbitrary amount of white space
3618 * (as determined by **isspace**\ (3)).
3620 * Five least significant bits of *flags* encode base, other bits
3621 * are currently unused.
3623 * Base must be either 8, 10, 16 or 0 to detect it automatically
3624 * similar to user space **strtoul**\ (3).
3626 * Number of characters consumed on success. Must be positive but
3627 * no more than *buf_len*.
3629 * **-EINVAL** if no valid digits were found or unsupported base
3632 * **-ERANGE** if resulting value was out of range.
3634 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3636 * Get a bpf-local-storage from a *sk*.
3638 * Logically, it could be thought of getting the value from
3639 * a *map* with *sk* as the **key**. From this
3640 * perspective, the usage is not much different from
3641 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3642 * helper enforces the key must be a full socket and the map must
3643 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
3645 * Underneath, the value is stored locally at *sk* instead of
3646 * the *map*. The *map* is used as the bpf-local-storage
3647 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3648 * searched against all bpf-local-storages residing at *sk*.
3650 * *sk* is a kernel **struct sock** pointer for LSM program.
3651 * *sk* is a **struct bpf_sock** pointer for other program types.
3653 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3654 * used such that a new bpf-local-storage will be
3655 * created if one does not exist. *value* can be used
3656 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3657 * the initial value of a bpf-local-storage. If *value* is
3658 * **NULL**, the new bpf-local-storage will be zero initialized.
3660 * A bpf-local-storage pointer is returned on success.
3662 * **NULL** if not found or there was an error in adding
3663 * a new bpf-local-storage.
3665 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3667 * Delete a bpf-local-storage from a *sk*.
3671 * **-ENOENT** if the bpf-local-storage cannot be found.
3672 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3674 * long bpf_send_signal(u32 sig)
3676 * Send signal *sig* to the process of the current task.
3677 * The signal may be delivered to any of this process's threads.
3679 * 0 on success or successfully queued.
3681 * **-EBUSY** if work queue under nmi is full.
3683 * **-EINVAL** if *sig* is invalid.
3685 * **-EPERM** if no permission to send the *sig*.
3687 * **-EAGAIN** if bpf program can try again.
3689 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3691 * Try to issue a SYN cookie for the packet with corresponding
3692 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3694 * *iph* points to the start of the IPv4 or IPv6 header, while
3695 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3696 * **sizeof**\ (**struct ip6hdr**).
3698 * *th* points to the start of the TCP header, while *th_len*
3699 * contains the length of the TCP header.
3701 * On success, lower 32 bits hold the generated SYN cookie in
3702 * followed by 16 bits which hold the MSS value for that cookie,
3703 * and the top 16 bits are unused.
3705 * On failure, the returned value is one of the following:
3707 * **-EINVAL** SYN cookie cannot be issued due to error
3709 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
3711 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3713 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
3715 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3717 * Write raw *data* blob into a special BPF perf event held by
3718 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3719 * event must have the following attributes: **PERF_SAMPLE_RAW**
3720 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3721 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3723 * The *flags* are used to indicate the index in *map* for which
3724 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3725 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3726 * to indicate that the index of the current CPU core should be
3729 * The value to write, of *size*, is passed through eBPF stack and
3730 * pointed by *data*.
3732 * *ctx* is a pointer to in-kernel struct sk_buff.
3734 * This helper is similar to **bpf_perf_event_output**\ () but
3735 * restricted to raw_tracepoint bpf programs.
3737 * 0 on success, or a negative error in case of failure.
3739 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3741 * Safely attempt to read *size* bytes from user space address
3742 * *unsafe_ptr* and store the data in *dst*.
3744 * 0 on success, or a negative error in case of failure.
3746 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3748 * Safely attempt to read *size* bytes from kernel space address
3749 * *unsafe_ptr* and store the data in *dst*.
3751 * 0 on success, or a negative error in case of failure.
3753 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3755 * Copy a NUL terminated string from an unsafe user address
3756 * *unsafe_ptr* to *dst*. The *size* should include the
3757 * terminating NUL byte. In case the string length is smaller than
3758 * *size*, the target is not padded with further NUL bytes. If the
3759 * string length is larger than *size*, just *size*-1 bytes are
3760 * copied and the last byte is set to NUL.
3762 * On success, returns the number of bytes that were written,
3763 * including the terminal NUL. This makes this helper useful in
3764 * tracing programs for reading strings, and more importantly to
3765 * get its length at runtime. See the following snippet:
3769 * SEC("kprobe/sys_open")
3770 * void bpf_sys_open(struct pt_regs *ctx)
3772 * char buf[PATHLEN]; // PATHLEN is defined to 256
3773 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
3776 * // Consume buf, for example push it to
3777 * // userspace via bpf_perf_event_output(); we
3778 * // can use res (the string length) as event
3779 * // size, after checking its boundaries.
3782 * In comparison, using **bpf_probe_read_user**\ () helper here
3783 * instead to read the string would require to estimate the length
3784 * at compile time, and would often result in copying more memory
3787 * Another useful use case is when parsing individual process
3788 * arguments or individual environment variables navigating
3789 * *current*\ **->mm->arg_start** and *current*\
3790 * **->mm->env_start**: using this helper and the return value,
3791 * one can quickly iterate at the right offset of the memory area.
3793 * On success, the strictly positive length of the output string,
3794 * including the trailing NUL character. On error, a negative
3797 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3799 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3800 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3802 * On success, the strictly positive length of the string, including
3803 * the trailing NUL character. On error, a negative value.
3805 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3807 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3808 * *rcv_nxt* is the ack_seq to be sent out.
3810 * 0 on success, or a negative error in case of failure.
3812 * long bpf_send_signal_thread(u32 sig)
3814 * Send signal *sig* to the thread corresponding to the current task.
3816 * 0 on success or successfully queued.
3818 * **-EBUSY** if work queue under nmi is full.
3820 * **-EINVAL** if *sig* is invalid.
3822 * **-EPERM** if no permission to send the *sig*.
3824 * **-EAGAIN** if bpf program can try again.
3826 * u64 bpf_jiffies64(void)
3828 * Obtain the 64bit jiffies
3830 * The 64 bit jiffies
3832 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3834 * For an eBPF program attached to a perf event, retrieve the
3835 * branch records (**struct perf_branch_entry**) associated to *ctx*
3836 * and store it in the buffer pointed by *buf* up to size
3839 * On success, number of bytes written to *buf*. On error, a
3842 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3843 * instead return the number of bytes required to store all the
3844 * branch entries. If this flag is set, *buf* may be NULL.
3846 * **-EINVAL** if arguments invalid or **size** not a multiple
3847 * of **sizeof**\ (**struct perf_branch_entry**\ ).
3849 * **-ENOENT** if architecture does not support branch records.
3851 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3853 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
3854 * *namespace* will be returned in *nsdata*.
3856 * 0 on success, or one of the following in case of failure:
3858 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
3859 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
3861 * **-ENOENT** if pidns does not exists for the current task.
3863 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3865 * Write raw *data* blob into a special BPF perf event held by
3866 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3867 * event must have the following attributes: **PERF_SAMPLE_RAW**
3868 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3869 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3871 * The *flags* are used to indicate the index in *map* for which
3872 * the value must be put, masked with **BPF_F_INDEX_MASK**.
3873 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3874 * to indicate that the index of the current CPU core should be
3877 * The value to write, of *size*, is passed through eBPF stack and
3878 * pointed by *data*.
3880 * *ctx* is a pointer to in-kernel struct xdp_buff.
3882 * This helper is similar to **bpf_perf_eventoutput**\ () but
3883 * restricted to raw_tracepoint bpf programs.
3885 * 0 on success, or a negative error in case of failure.
3887 * u64 bpf_get_netns_cookie(void *ctx)
3889 * Retrieve the cookie (generated by the kernel) of the network
3890 * namespace the input *ctx* is associated with. The network
3891 * namespace cookie remains stable for its lifetime and provides
3892 * a global identifier that can be assumed unique. If *ctx* is
3893 * NULL, then the helper returns the cookie for the initial
3894 * network namespace. The cookie itself is very similar to that
3895 * of **bpf_get_socket_cookie**\ () helper, but for network
3896 * namespaces instead of sockets.
3898 * A 8-byte long opaque number.
3900 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3902 * Return id of cgroup v2 that is ancestor of the cgroup associated
3903 * with the current task at the *ancestor_level*. The root cgroup
3904 * is at *ancestor_level* zero and each step down the hierarchy
3905 * increments the level. If *ancestor_level* == level of cgroup
3906 * associated with the current task, then return value will be the
3907 * same as that of **bpf_get_current_cgroup_id**\ ().
3909 * The helper is useful to implement policies based on cgroups
3910 * that are upper in hierarchy than immediate cgroup associated
3911 * with the current task.
3913 * The format of returned id and helper limitations are same as in
3914 * **bpf_get_current_cgroup_id**\ ().
3916 * The id is returned or 0 in case the id could not be retrieved.
3918 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3920 * Helper is overloaded depending on BPF program type. This
3921 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3922 * **BPF_PROG_TYPE_SCHED_ACT** programs.
3924 * Assign the *sk* to the *skb*. When combined with appropriate
3925 * routing configuration to receive the packet towards the socket,
3926 * will cause *skb* to be delivered to the specified socket.
3927 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
3928 * **bpf_clone_redirect**\ () or other methods outside of BPF may
3929 * interfere with successful delivery to the socket.
3931 * This operation is only valid from TC ingress path.
3933 * The *flags* argument must be zero.
3935 * 0 on success, or a negative error in case of failure:
3937 * **-EINVAL** if specified *flags* are not supported.
3939 * **-ENOENT** if the socket is unavailable for assignment.
3941 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
3943 * **-EOPNOTSUPP** if the operation is not supported, for example
3944 * a call from outside of TC ingress.
3946 * **-ESOCKTNOSUPPORT** if the socket type is not supported
3949 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3951 * Helper is overloaded depending on BPF program type. This
3952 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3954 * Select the *sk* as a result of a socket lookup.
3956 * For the operation to succeed passed socket must be compatible
3957 * with the packet description provided by the *ctx* object.
3959 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
3960 * be an exact match. While IP family (**AF_INET** or
3961 * **AF_INET6**) must be compatible, that is IPv6 sockets
3962 * that are not v6-only can be selected for IPv4 packets.
3964 * Only TCP listeners and UDP unconnected sockets can be
3965 * selected. *sk* can also be NULL to reset any previous
3968 * *flags* argument can combination of following values:
3970 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
3971 * socket selection, potentially done by a BPF program
3972 * that ran before us.
3974 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
3975 * load-balancing within reuseport group for the socket
3978 * On success *ctx->sk* will point to the selected socket.
3981 * 0 on success, or a negative errno in case of failure.
3983 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
3984 * not compatible with packet family (*ctx->family*).
3986 * * **-EEXIST** if socket has been already selected,
3987 * potentially by another program, and
3988 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
3990 * * **-EINVAL** if unsupported flags were specified.
3992 * * **-EPROTOTYPE** if socket L4 protocol
3993 * (*sk->protocol*) doesn't match packet protocol
3994 * (*ctx->protocol*).
3996 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
3997 * state (TCP listening or UDP unconnected).
3999 * u64 bpf_ktime_get_boot_ns(void)
4001 * Return the time elapsed since system boot, in nanoseconds.
4002 * Does include the time the system was suspended.
4003 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4007 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4009 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4010 * out the format string.
4011 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4012 * the format string itself. The *data* and *data_len* are format string
4013 * arguments. The *data* are a **u64** array and corresponding format string
4014 * values are stored in the array. For strings and pointers where pointees
4015 * are accessed, only the pointer values are stored in the *data* array.
4016 * The *data_len* is the size of *data* in bytes.
4018 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4019 * Reading kernel memory may fail due to either invalid address or
4020 * valid address but requiring a major memory fault. If reading kernel memory
4021 * fails, the string for **%s** will be an empty string, and the ip
4022 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4023 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4025 * 0 on success, or a negative error in case of failure:
4027 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4028 * by returning 1 from bpf program.
4030 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4032 * **-E2BIG** if *fmt* contains too many format specifiers.
4034 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4036 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4038 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4039 * The *m* represents the seq_file. The *data* and *len* represent the
4040 * data to write in bytes.
4042 * 0 on success, or a negative error in case of failure:
4044 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4046 * u64 bpf_sk_cgroup_id(void *sk)
4048 * Return the cgroup v2 id of the socket *sk*.
4050 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4051 * returned from **bpf_sk_lookup_xxx**\ (),
4052 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4053 * same as in **bpf_skb_cgroup_id**\ ().
4055 * This helper is available only if the kernel was compiled with
4056 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4058 * The id is returned or 0 in case the id could not be retrieved.
4060 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4062 * Return id of cgroup v2 that is ancestor of cgroup associated
4063 * with the *sk* at the *ancestor_level*. The root cgroup is at
4064 * *ancestor_level* zero and each step down the hierarchy
4065 * increments the level. If *ancestor_level* == level of cgroup
4066 * associated with *sk*, then return value will be same as that
4067 * of **bpf_sk_cgroup_id**\ ().
4069 * The helper is useful to implement policies based on cgroups
4070 * that are upper in hierarchy than immediate cgroup associated
4073 * The format of returned id and helper limitations are same as in
4074 * **bpf_sk_cgroup_id**\ ().
4076 * The id is returned or 0 in case the id could not be retrieved.
4078 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4080 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4081 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4082 * of new data availability is sent.
4083 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4084 * of new data availability is sent unconditionally.
4085 * If **0** is specified in *flags*, an adaptive notification
4086 * of new data availability is sent.
4088 * An adaptive notification is a notification sent whenever the user-space
4089 * process has caught up and consumed all available payloads. In case the user-space
4090 * process is still processing a previous payload, then no notification is needed
4091 * as it will process the newly added payload automatically.
4093 * 0 on success, or a negative error in case of failure.
4095 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4097 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4098 * *flags* must be 0.
4100 * Valid pointer with *size* bytes of memory available; NULL,
4103 * void bpf_ringbuf_submit(void *data, u64 flags)
4105 * Submit reserved ring buffer sample, pointed to by *data*.
4106 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4107 * of new data availability is sent.
4108 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4109 * of new data availability is sent unconditionally.
4110 * If **0** is specified in *flags*, an adaptive notification
4111 * of new data availability is sent.
4113 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4115 * Nothing. Always succeeds.
4117 * void bpf_ringbuf_discard(void *data, u64 flags)
4119 * Discard reserved ring buffer sample, pointed to by *data*.
4120 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4121 * of new data availability is sent.
4122 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4123 * of new data availability is sent unconditionally.
4124 * If **0** is specified in *flags*, an adaptive notification
4125 * of new data availability is sent.
4127 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4129 * Nothing. Always succeeds.
4131 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4133 * Query various characteristics of provided ring buffer. What
4134 * exactly is queries is determined by *flags*:
4136 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4137 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4138 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4139 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4141 * Data returned is just a momentary snapshot of actual values
4142 * and could be inaccurate, so this facility should be used to
4143 * power heuristics and for reporting, not to make 100% correct
4146 * Requested value, or 0, if *flags* are not recognized.
4148 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4150 * Change the skbs checksum level by one layer up or down, or
4151 * reset it entirely to none in order to have the stack perform
4152 * checksum validation. The level is applicable to the following
4153 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4154 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4155 * through **bpf_skb_adjust_room**\ () helper with passing in
4156 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4157 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4158 * the UDP header is removed. Similarly, an encap of the latter
4159 * into the former could be accompanied by a helper call to
4160 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4161 * skb is still intended to be processed in higher layers of the
4162 * stack instead of just egressing at tc.
4164 * There are three supported level settings at this time:
4166 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4167 * with CHECKSUM_UNNECESSARY.
4168 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4169 * with CHECKSUM_UNNECESSARY.
4170 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4171 * sets CHECKSUM_NONE to force checksum validation by the stack.
4172 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4175 * 0 on success, or a negative error in case of failure. In the
4176 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4177 * is returned or the error code -EACCES in case the skb is not
4178 * subject to CHECKSUM_UNNECESSARY.
4180 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4182 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4184 * *sk* if casting is valid, or **NULL** otherwise.
4186 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4188 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4190 * *sk* if casting is valid, or **NULL** otherwise.
4192 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4194 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4196 * *sk* if casting is valid, or **NULL** otherwise.
4198 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4200 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4202 * *sk* if casting is valid, or **NULL** otherwise.
4204 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4206 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4208 * *sk* if casting is valid, or **NULL** otherwise.
4210 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4212 * Return a user or a kernel stack in bpf program provided buffer.
4213 * To achieve this, the helper needs *task*, which is a valid
4214 * pointer to **struct task_struct**. To store the stacktrace, the
4215 * bpf program provides *buf* with a nonnegative *size*.
4217 * The last argument, *flags*, holds the number of stack frames to
4218 * skip (from 0 to 255), masked with
4219 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4220 * the following flags:
4222 * **BPF_F_USER_STACK**
4223 * Collect a user space stack instead of a kernel stack.
4224 * **BPF_F_USER_BUILD_ID**
4225 * Collect buildid+offset instead of ips for user stack,
4226 * only valid if **BPF_F_USER_STACK** is also specified.
4228 * **bpf_get_task_stack**\ () can collect up to
4229 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4230 * to sufficient large buffer size. Note that
4231 * this limit can be controlled with the **sysctl** program, and
4232 * that it should be manually increased in order to profile long
4233 * user stacks (such as stacks for Java programs). To do so, use:
4237 * # sysctl kernel.perf_event_max_stack=<new value>
4239 * A non-negative value equal to or less than *size* on success,
4240 * or a negative error in case of failure.
4242 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4244 * Load header option. Support reading a particular TCP header
4245 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4247 * If *flags* is 0, it will search the option from the
4248 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4249 * has details on what skb_data contains under different
4250 * *skops*\ **->op**.
4252 * The first byte of the *searchby_res* specifies the
4253 * kind that it wants to search.
4255 * If the searching kind is an experimental kind
4256 * (i.e. 253 or 254 according to RFC6994). It also
4257 * needs to specify the "magic" which is either
4258 * 2 bytes or 4 bytes. It then also needs to
4259 * specify the size of the magic by using
4260 * the 2nd byte which is "kind-length" of a TCP
4261 * header option and the "kind-length" also
4262 * includes the first 2 bytes "kind" and "kind-length"
4263 * itself as a normal TCP header option also does.
4265 * For example, to search experimental kind 254 with
4266 * 2 byte magic 0xeB9F, the searchby_res should be
4267 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4269 * To search for the standard window scale option (3),
4270 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4271 * Note, kind-length must be 0 for regular option.
4273 * Searching for No-Op (0) and End-of-Option-List (1) are
4276 * *len* must be at least 2 bytes which is the minimal size
4277 * of a header option.
4281 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4282 * saved_syn packet or the just-received syn packet.
4285 * > 0 when found, the header option is copied to *searchby_res*.
4286 * The return value is the total length copied. On failure, a
4287 * negative error code is returned:
4289 * **-EINVAL** if a parameter is invalid.
4291 * **-ENOMSG** if the option is not found.
4293 * **-ENOENT** if no syn packet is available when
4294 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4296 * **-ENOSPC** if there is not enough space. Only *len* number of
4299 * **-EFAULT** on failure to parse the header options in the
4302 * **-EPERM** if the helper cannot be used under the current
4303 * *skops*\ **->op**.
4305 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4307 * Store header option. The data will be copied
4308 * from buffer *from* with length *len* to the TCP header.
4310 * The buffer *from* should have the whole option that
4311 * includes the kind, kind-length, and the actual
4312 * option data. The *len* must be at least kind-length
4313 * long. The kind-length does not have to be 4 byte
4314 * aligned. The kernel will take care of the padding
4315 * and setting the 4 bytes aligned value to th->doff.
4317 * This helper will check for duplicated option
4318 * by searching the same option in the outgoing skb.
4320 * This helper can only be called during
4321 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4324 * 0 on success, or negative error in case of failure:
4326 * **-EINVAL** If param is invalid.
4328 * **-ENOSPC** if there is not enough space in the header.
4329 * Nothing has been written
4331 * **-EEXIST** if the option already exists.
4333 * **-EFAULT** on failrue to parse the existing header options.
4335 * **-EPERM** if the helper cannot be used under the current
4336 * *skops*\ **->op**.
4338 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4340 * Reserve *len* bytes for the bpf header option. The
4341 * space will be used by **bpf_store_hdr_opt**\ () later in
4342 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4344 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4345 * the total number of bytes will be reserved.
4347 * This helper can only be called during
4348 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4351 * 0 on success, or negative error in case of failure:
4353 * **-EINVAL** if a parameter is invalid.
4355 * **-ENOSPC** if there is not enough space in the header.
4357 * **-EPERM** if the helper cannot be used under the current
4358 * *skops*\ **->op**.
4360 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4362 * Get a bpf_local_storage from an *inode*.
4364 * Logically, it could be thought of as getting the value from
4365 * a *map* with *inode* as the **key**. From this
4366 * perspective, the usage is not much different from
4367 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4368 * helper enforces the key must be an inode and the map must also
4369 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4371 * Underneath, the value is stored locally at *inode* instead of
4372 * the *map*. The *map* is used as the bpf-local-storage
4373 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4374 * searched against all bpf_local_storage residing at *inode*.
4376 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4377 * used such that a new bpf_local_storage will be
4378 * created if one does not exist. *value* can be used
4379 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4380 * the initial value of a bpf_local_storage. If *value* is
4381 * **NULL**, the new bpf_local_storage will be zero initialized.
4383 * A bpf_local_storage pointer is returned on success.
4385 * **NULL** if not found or there was an error in adding
4386 * a new bpf_local_storage.
4388 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4390 * Delete a bpf_local_storage from an *inode*.
4394 * **-ENOENT** if the bpf_local_storage cannot be found.
4396 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4398 * Return full path for given **struct path** object, which
4399 * needs to be the kernel BTF *path* object. The path is
4400 * returned in the provided buffer *buf* of size *sz* and
4401 * is zero terminated.
4404 * On success, the strictly positive length of the string,
4405 * including the trailing NUL character. On error, a negative
4408 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4410 * Read *size* bytes from user space address *user_ptr* and store
4411 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4413 * 0 on success, or a negative error in case of failure.
4415 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4417 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4418 * using *ptr*->type_id. This value should specify the type
4419 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4420 * can be used to look up vmlinux BTF type ids. Traversing the
4421 * data structure using BTF, the type information and values are
4422 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4423 * the pointer data is carried out to avoid kernel crashes during
4424 * operation. Smaller types can use string space on the stack;
4425 * larger programs can use map data to store the string
4428 * The string can be subsequently shared with userspace via
4429 * bpf_perf_event_output() or ring buffer interfaces.
4430 * bpf_trace_printk() is to be avoided as it places too small
4431 * a limit on string size to be useful.
4433 * *flags* is a combination of
4436 * no formatting around type information
4438 * no struct/union member names/types
4440 * show raw (unobfuscated) pointer values;
4441 * equivalent to printk specifier %px.
4443 * show zero-valued struct/union members; they
4444 * are not displayed by default
4447 * The number of bytes that were written (or would have been
4448 * written if output had to be truncated due to string size),
4449 * or a negative error in cases of failure.
4451 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4453 * Use BTF to write to seq_write a string representation of
4454 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4455 * *flags* are identical to those used for bpf_snprintf_btf.
4457 * 0 on success or a negative error in case of failure.
4459 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4461 * See **bpf_get_cgroup_classid**\ () for the main description.
4462 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4463 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4464 * associated socket instead of the current process.
4466 * The id is returned or 0 in case the id could not be retrieved.
4468 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4470 * Redirect the packet to another net device of index *ifindex*
4471 * and fill in L2 addresses from neighboring subsystem. This helper
4472 * is somewhat similar to **bpf_redirect**\ (), except that it
4473 * populates L2 addresses as well, meaning, internally, the helper
4474 * relies on the neighbor lookup for the L2 address of the nexthop.
4476 * The helper will perform a FIB lookup based on the skb's
4477 * networking header to get the address of the next hop, unless
4478 * this is supplied by the caller in the *params* argument. The
4479 * *plen* argument indicates the len of *params* and should be set
4480 * to 0 if *params* is NULL.
4482 * The *flags* argument is reserved and must be 0. The helper is
4483 * currently only supported for tc BPF program types, and enabled
4484 * for IPv4 and IPv6 protocols.
4486 * The helper returns **TC_ACT_REDIRECT** on success or
4487 * **TC_ACT_SHOT** on error.
4489 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4491 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4492 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4493 * extern variable decorated with '__ksym'. For ksym, there is a
4494 * global var (either static or global) defined of the same name
4495 * in the kernel. The ksym is percpu if the global var is percpu.
4496 * The returned pointer points to the global percpu var on *cpu*.
4498 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4499 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4500 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4501 * bpf_per_cpu_ptr() must check the returned value.
4503 * A pointer pointing to the kernel percpu variable on *cpu*, or
4504 * NULL, if *cpu* is invalid.
4506 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4508 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4509 * pointer to the percpu kernel variable on this cpu. See the
4510 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4512 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4513 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4514 * never return NULL.
4516 * A pointer pointing to the kernel percpu variable on this cpu.
4518 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4520 * Redirect the packet to another net device of index *ifindex*.
4521 * This helper is somewhat similar to **bpf_redirect**\ (), except
4522 * that the redirection happens to the *ifindex*' peer device and
4523 * the netns switch takes place from ingress to ingress without
4524 * going through the CPU's backlog queue.
4526 * The *flags* argument is reserved and must be 0. The helper is
4527 * currently only supported for tc BPF program types at the ingress
4528 * hook and for veth device types. The peer device must reside in a
4529 * different network namespace.
4531 * The helper returns **TC_ACT_REDIRECT** on success or
4532 * **TC_ACT_SHOT** on error.
4534 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4536 * Get a bpf_local_storage from the *task*.
4538 * Logically, it could be thought of as getting the value from
4539 * a *map* with *task* as the **key**. From this
4540 * perspective, the usage is not much different from
4541 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4542 * helper enforces the key must be an task_struct and the map must also
4543 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4545 * Underneath, the value is stored locally at *task* instead of
4546 * the *map*. The *map* is used as the bpf-local-storage
4547 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4548 * searched against all bpf_local_storage residing at *task*.
4550 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4551 * used such that a new bpf_local_storage will be
4552 * created if one does not exist. *value* can be used
4553 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4554 * the initial value of a bpf_local_storage. If *value* is
4555 * **NULL**, the new bpf_local_storage will be zero initialized.
4557 * A bpf_local_storage pointer is returned on success.
4559 * **NULL** if not found or there was an error in adding
4560 * a new bpf_local_storage.
4562 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4564 * Delete a bpf_local_storage from a *task*.
4568 * **-ENOENT** if the bpf_local_storage cannot be found.
4570 * struct task_struct *bpf_get_current_task_btf(void)
4572 * Return a BTF pointer to the "current" task.
4573 * This pointer can also be used in helpers that accept an
4574 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
4576 * Pointer to the current task.
4578 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4580 * Set or clear certain options on *bprm*:
4582 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4583 * which sets the **AT_SECURE** auxv for glibc. The bit
4584 * is cleared if the flag is not specified.
4586 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
4588 * u64 bpf_ktime_get_coarse_ns(void)
4590 * Return a coarse-grained version of the time elapsed since
4591 * system boot, in nanoseconds. Does not include time the system
4594 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4598 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4600 * Returns the stored IMA hash of the *inode* (if it's avaialable).
4601 * If the hash is larger than *size*, then only *size*
4602 * bytes will be copied to *dst*
4604 * The **hash_algo** is returned on success,
4605 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4606 * invalid arguments are passed.
4608 * struct socket *bpf_sock_from_file(struct file *file)
4610 * If the given file represents a socket, returns the associated
4613 * A pointer to a struct socket on success or NULL if the file is
4616 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4618 * Check packet size against exceeding MTU of net device (based
4619 * on *ifindex*). This helper will likely be used in combination
4620 * with helpers that adjust/change the packet size.
4622 * The argument *len_diff* can be used for querying with a planned
4623 * size change. This allows to check MTU prior to changing packet
4624 * ctx. Providing an *len_diff* adjustment that is larger than the
4625 * actual packet size (resulting in negative packet size) will in
4626 * principle not exceed the MTU, why it is not considered a
4627 * failure. Other BPF-helpers are needed for performing the
4628 * planned size change, why the responsability for catch a negative
4629 * packet size belong in those helpers.
4631 * Specifying *ifindex* zero means the MTU check is performed
4632 * against the current net device. This is practical if this isn't
4633 * used prior to redirect.
4635 * On input *mtu_len* must be a valid pointer, else verifier will
4636 * reject BPF program. If the value *mtu_len* is initialized to
4637 * zero then the ctx packet size is use. When value *mtu_len* is
4638 * provided as input this specify the L3 length that the MTU check
4639 * is done against. Remember XDP and TC length operate at L2, but
4640 * this value is L3 as this correlate to MTU and IP-header tot_len
4641 * values which are L3 (similar behavior as bpf_fib_lookup).
4643 * The Linux kernel route table can configure MTUs on a more
4644 * specific per route level, which is not provided by this helper.
4645 * For route level MTU checks use the **bpf_fib_lookup**\ ()
4648 * *ctx* is either **struct xdp_md** for XDP programs or
4649 * **struct sk_buff** for tc cls_act programs.
4651 * The *flags* argument can be a combination of one or more of the
4654 * **BPF_MTU_CHK_SEGS**
4655 * This flag will only works for *ctx* **struct sk_buff**.
4656 * If packet context contains extra packet segment buffers
4657 * (often knows as GSO skb), then MTU check is harder to
4658 * check at this point, because in transmit path it is
4659 * possible for the skb packet to get re-segmented
4660 * (depending on net device features). This could still be
4661 * a MTU violation, so this flag enables performing MTU
4662 * check against segments, with a different violation
4663 * return code to tell it apart. Check cannot use len_diff.
4665 * On return *mtu_len* pointer contains the MTU value of the net
4666 * device. Remember the net device configured MTU is the L3 size,
4667 * which is returned here and XDP and TC length operate at L2.
4668 * Helper take this into account for you, but remember when using
4669 * MTU value in your BPF-code.
4672 * * 0 on success, and populate MTU value in *mtu_len* pointer.
4674 * * < 0 if any input argument is invalid (*mtu_len* not updated)
4676 * MTU violations return positive values, but also populate MTU
4677 * value in *mtu_len* pointer, as this can be needed for
4678 * implementing PMTU handing:
4680 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
4681 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4683 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4685 * For each element in **map**, call **callback_fn** function with
4686 * **map**, **callback_ctx** and other map-specific parameters.
4687 * The **callback_fn** should be a static function and
4688 * the **callback_ctx** should be a pointer to the stack.
4689 * The **flags** is used to control certain aspects of the helper.
4690 * Currently, the **flags** must be 0.
4692 * The following are a list of supported map types and their
4693 * respective expected callback signatures:
4695 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4696 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4697 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4699 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4701 * For per_cpu maps, the map_value is the value on the cpu where the
4702 * bpf_prog is running.
4704 * If **callback_fn** return 0, the helper will continue to the next
4705 * element. If return value is 1, the helper will skip the rest of
4706 * elements and return. Other return values are not used now.
4709 * The number of traversed map elements for success, **-EINVAL** for
4710 * invalid **flags**.
4712 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4714 * Outputs a string into the **str** buffer of size **str_size**
4715 * based on a format string stored in a read-only map pointed by
4718 * Each format specifier in **fmt** corresponds to one u64 element
4719 * in the **data** array. For strings and pointers where pointees
4720 * are accessed, only the pointer values are stored in the *data*
4721 * array. The *data_len* is the size of *data* in bytes.
4723 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4724 * memory. Reading kernel memory may fail due to either invalid
4725 * address or valid address but requiring a major memory fault. If
4726 * reading kernel memory fails, the string for **%s** will be an
4727 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4728 * Not returning error to bpf program is consistent with what
4729 * **bpf_trace_printk**\ () does for now.
4732 * The strictly positive length of the formatted string, including
4733 * the trailing zero character. If the return value is greater than
4734 * **str_size**, **str** contains a truncated string, guaranteed to
4735 * be zero-terminated except when **str_size** is 0.
4737 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4739 #define __BPF_FUNC_MAPPER(FN) \
4741 FN(map_lookup_elem), \
4742 FN(map_update_elem), \
4743 FN(map_delete_elem), \
4747 FN(get_prandom_u32), \
4748 FN(get_smp_processor_id), \
4749 FN(skb_store_bytes), \
4750 FN(l3_csum_replace), \
4751 FN(l4_csum_replace), \
4753 FN(clone_redirect), \
4754 FN(get_current_pid_tgid), \
4755 FN(get_current_uid_gid), \
4756 FN(get_current_comm), \
4757 FN(get_cgroup_classid), \
4758 FN(skb_vlan_push), \
4760 FN(skb_get_tunnel_key), \
4761 FN(skb_set_tunnel_key), \
4762 FN(perf_event_read), \
4764 FN(get_route_realm), \
4765 FN(perf_event_output), \
4766 FN(skb_load_bytes), \
4769 FN(skb_get_tunnel_opt), \
4770 FN(skb_set_tunnel_opt), \
4771 FN(skb_change_proto), \
4772 FN(skb_change_type), \
4773 FN(skb_under_cgroup), \
4774 FN(get_hash_recalc), \
4775 FN(get_current_task), \
4776 FN(probe_write_user), \
4777 FN(current_task_under_cgroup), \
4778 FN(skb_change_tail), \
4779 FN(skb_pull_data), \
4781 FN(set_hash_invalid), \
4782 FN(get_numa_node_id), \
4783 FN(skb_change_head), \
4784 FN(xdp_adjust_head), \
4785 FN(probe_read_str), \
4786 FN(get_socket_cookie), \
4787 FN(get_socket_uid), \
4790 FN(skb_adjust_room), \
4792 FN(sk_redirect_map), \
4793 FN(sock_map_update), \
4794 FN(xdp_adjust_meta), \
4795 FN(perf_event_read_value), \
4796 FN(perf_prog_read_value), \
4798 FN(override_return), \
4799 FN(sock_ops_cb_flags_set), \
4800 FN(msg_redirect_map), \
4801 FN(msg_apply_bytes), \
4802 FN(msg_cork_bytes), \
4803 FN(msg_pull_data), \
4805 FN(xdp_adjust_tail), \
4806 FN(skb_get_xfrm_state), \
4808 FN(skb_load_bytes_relative), \
4810 FN(sock_hash_update), \
4811 FN(msg_redirect_hash), \
4812 FN(sk_redirect_hash), \
4813 FN(lwt_push_encap), \
4814 FN(lwt_seg6_store_bytes), \
4815 FN(lwt_seg6_adjust_srh), \
4816 FN(lwt_seg6_action), \
4819 FN(skb_cgroup_id), \
4820 FN(get_current_cgroup_id), \
4821 FN(get_local_storage), \
4822 FN(sk_select_reuseport), \
4823 FN(skb_ancestor_cgroup_id), \
4824 FN(sk_lookup_tcp), \
4825 FN(sk_lookup_udp), \
4827 FN(map_push_elem), \
4829 FN(map_peek_elem), \
4830 FN(msg_push_data), \
4832 FN(rc_pointer_rel), \
4837 FN(skb_ecn_set_ce), \
4838 FN(get_listener_sock), \
4839 FN(skc_lookup_tcp), \
4840 FN(tcp_check_syncookie), \
4841 FN(sysctl_get_name), \
4842 FN(sysctl_get_current_value), \
4843 FN(sysctl_get_new_value), \
4844 FN(sysctl_set_new_value), \
4847 FN(sk_storage_get), \
4848 FN(sk_storage_delete), \
4850 FN(tcp_gen_syncookie), \
4852 FN(probe_read_user), \
4853 FN(probe_read_kernel), \
4854 FN(probe_read_user_str), \
4855 FN(probe_read_kernel_str), \
4857 FN(send_signal_thread), \
4859 FN(read_branch_records), \
4860 FN(get_ns_current_pid_tgid), \
4862 FN(get_netns_cookie), \
4863 FN(get_current_ancestor_cgroup_id), \
4865 FN(ktime_get_boot_ns), \
4869 FN(sk_ancestor_cgroup_id), \
4870 FN(ringbuf_output), \
4871 FN(ringbuf_reserve), \
4872 FN(ringbuf_submit), \
4873 FN(ringbuf_discard), \
4874 FN(ringbuf_query), \
4876 FN(skc_to_tcp6_sock), \
4877 FN(skc_to_tcp_sock), \
4878 FN(skc_to_tcp_timewait_sock), \
4879 FN(skc_to_tcp_request_sock), \
4880 FN(skc_to_udp6_sock), \
4881 FN(get_task_stack), \
4883 FN(store_hdr_opt), \
4884 FN(reserve_hdr_opt), \
4885 FN(inode_storage_get), \
4886 FN(inode_storage_delete), \
4888 FN(copy_from_user), \
4890 FN(seq_printf_btf), \
4891 FN(skb_cgroup_classid), \
4892 FN(redirect_neigh), \
4895 FN(redirect_peer), \
4896 FN(task_storage_get), \
4897 FN(task_storage_delete), \
4898 FN(get_current_task_btf), \
4899 FN(bprm_opts_set), \
4900 FN(ktime_get_coarse_ns), \
4901 FN(ima_inode_hash), \
4902 FN(sock_from_file), \
4904 FN(for_each_map_elem), \
4908 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
4909 * function eBPF program intends to call
4911 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
4913 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
4916 #undef __BPF_ENUM_FN
4918 /* All flags used by eBPF helper functions, placed here. */
4920 /* BPF_FUNC_skb_store_bytes flags. */
4922 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
4923 BPF_F_INVALIDATE_HASH = (1ULL << 1),
4926 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
4927 * First 4 bits are for passing the header field size.
4930 BPF_F_HDR_FIELD_MASK = 0xfULL,
4933 /* BPF_FUNC_l4_csum_replace flags. */
4935 BPF_F_PSEUDO_HDR = (1ULL << 4),
4936 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
4937 BPF_F_MARK_ENFORCE = (1ULL << 6),
4940 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
4942 BPF_F_INGRESS = (1ULL << 0),
4945 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
4947 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
4950 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
4952 BPF_F_SKIP_FIELD_MASK = 0xffULL,
4953 BPF_F_USER_STACK = (1ULL << 8),
4954 /* flags used by BPF_FUNC_get_stackid only. */
4955 BPF_F_FAST_STACK_CMP = (1ULL << 9),
4956 BPF_F_REUSE_STACKID = (1ULL << 10),
4957 /* flags used by BPF_FUNC_get_stack only. */
4958 BPF_F_USER_BUILD_ID = (1ULL << 11),
4961 /* BPF_FUNC_skb_set_tunnel_key flags. */
4963 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
4964 BPF_F_DONT_FRAGMENT = (1ULL << 2),
4965 BPF_F_SEQ_NUMBER = (1ULL << 3),
4968 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
4969 * BPF_FUNC_perf_event_read_value flags.
4972 BPF_F_INDEX_MASK = 0xffffffffULL,
4973 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
4974 /* BPF_FUNC_perf_event_output for sk_buff input context. */
4975 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
4978 /* Current network namespace */
4980 BPF_F_CURRENT_NETNS = (-1L),
4983 /* BPF_FUNC_csum_level level values. */
4985 BPF_CSUM_LEVEL_QUERY,
4988 BPF_CSUM_LEVEL_RESET,
4991 /* BPF_FUNC_skb_adjust_room flags. */
4993 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
4994 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
4995 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
4996 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
4997 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
4998 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
4999 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
5003 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
5004 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
5007 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
5008 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5009 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5011 /* BPF_FUNC_sysctl_get_name flags. */
5013 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
5016 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5018 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
5019 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5020 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5022 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
5025 /* BPF_FUNC_read_branch_records flags. */
5027 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
5030 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5031 * BPF_FUNC_bpf_ringbuf_output flags.
5034 BPF_RB_NO_WAKEUP = (1ULL << 0),
5035 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
5038 /* BPF_FUNC_bpf_ringbuf_query flags */
5040 BPF_RB_AVAIL_DATA = 0,
5041 BPF_RB_RING_SIZE = 1,
5042 BPF_RB_CONS_POS = 2,
5043 BPF_RB_PROD_POS = 3,
5046 /* BPF ring buffer constants */
5048 BPF_RINGBUF_BUSY_BIT = (1U << 31),
5049 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
5050 BPF_RINGBUF_HDR_SZ = 8,
5053 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5055 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
5056 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
5059 /* Mode for BPF_FUNC_skb_adjust_room helper. */
5060 enum bpf_adj_room_mode {
5065 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5066 enum bpf_hdr_start_off {
5071 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5072 enum bpf_lwt_encap_mode {
5074 BPF_LWT_ENCAP_SEG6_INLINE,
5078 /* Flags for bpf_bprm_opts_set helper */
5080 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
5083 #define __bpf_md_ptr(type, name) \
5087 } __attribute__((aligned(8)))
5089 /* user accessible mirror of in-kernel sk_buff.
5090 * new fields can only be added to the end of this structure
5096 __u32 queue_mapping;
5102 __u32 ingress_ifindex;
5112 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5114 __u32 remote_ip4; /* Stored in network byte order */
5115 __u32 local_ip4; /* Stored in network byte order */
5116 __u32 remote_ip6[4]; /* Stored in network byte order */
5117 __u32 local_ip6[4]; /* Stored in network byte order */
5118 __u32 remote_port; /* Stored in network byte order */
5119 __u32 local_port; /* stored in host byte order */
5123 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5127 __bpf_md_ptr(struct bpf_sock *, sk);
5131 struct bpf_tunnel_key {
5135 __u32 remote_ipv6[4];
5139 __u16 tunnel_ext; /* Padding, future use. */
5143 /* user accessible mirror of in-kernel xfrm_state.
5144 * new fields can only be added to the end of this structure
5146 struct bpf_xfrm_state {
5148 __u32 spi; /* Stored in network byte order */
5150 __u16 ext; /* Padding, future use. */
5152 __u32 remote_ipv4; /* Stored in network byte order */
5153 __u32 remote_ipv6[4]; /* Stored in network byte order */
5157 /* Generic BPF return codes which all BPF program types may support.
5158 * The values are binary compatible with their TC_ACT_* counter-part to
5159 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5162 * XDP is handled seprately, see XDP_*.
5170 /* >127 are reserved for prog type specific return codes.
5172 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5173 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5174 * changed and should be routed based on its new L3 header.
5175 * (This is an L3 redirect, as opposed to L2 redirect
5176 * represented by BPF_REDIRECT above).
5178 BPF_LWT_REROUTE = 128,
5188 /* IP address also allows 1 and 2 bytes access */
5191 __u32 src_port; /* host byte order */
5192 __u32 dst_port; /* network byte order */
5196 __s32 rx_queue_mapping;
5199 struct bpf_tcp_sock {
5200 __u32 snd_cwnd; /* Sending congestion window */
5201 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
5203 __u32 snd_ssthresh; /* Slow start size threshold */
5204 __u32 rcv_nxt; /* What we want to receive next */
5205 __u32 snd_nxt; /* Next sequence we send */
5206 __u32 snd_una; /* First byte we want an ack for */
5207 __u32 mss_cache; /* Cached effective mss, not including SACKS */
5208 __u32 ecn_flags; /* ECN status bits. */
5209 __u32 rate_delivered; /* saved rate sample: packets delivered */
5210 __u32 rate_interval_us; /* saved rate sample: time elapsed */
5211 __u32 packets_out; /* Packets which are "in flight" */
5212 __u32 retrans_out; /* Retransmitted packets out */
5213 __u32 total_retrans; /* Total retransmits for entire connection */
5214 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
5215 * total number of segments in.
5217 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
5218 * total number of data segments in.
5220 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
5221 * The total number of segments sent.
5223 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
5224 * total number of data segments sent.
5226 __u32 lost_out; /* Lost packets */
5227 __u32 sacked_out; /* SACK'd packets */
5228 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
5229 * sum(delta(rcv_nxt)), or how many bytes
5232 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
5233 * sum(delta(snd_una)), or how many bytes
5236 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
5237 * total number of DSACK blocks received
5239 __u32 delivered; /* Total data packets delivered incl. rexmits */
5240 __u32 delivered_ce; /* Like the above but only ECE marked packets */
5241 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
5244 struct bpf_sock_tuple {
5261 struct bpf_xdp_sock {
5265 #define XDP_PACKET_HEADROOM 256
5267 /* User return codes for XDP prog type.
5268 * A valid XDP program must return one of these defined values. All other
5269 * return codes are reserved for future use. Unknown return codes will
5270 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5280 /* user accessible metadata for XDP packet hook
5281 * new fields must be added to the end of this structure
5287 /* Below access go through struct xdp_rxq_info */
5288 __u32 ingress_ifindex; /* rxq->dev->ifindex */
5289 __u32 rx_queue_index; /* rxq->queue_index */
5291 __u32 egress_ifindex; /* txq->dev->ifindex */
5294 /* DEVMAP map-value layout
5296 * The struct data-layout of map-value is a configuration interface.
5297 * New members can only be added to the end of this structure.
5299 struct bpf_devmap_val {
5300 __u32 ifindex; /* device index */
5302 int fd; /* prog fd on map write */
5303 __u32 id; /* prog id on map read */
5307 /* CPUMAP map-value layout
5309 * The struct data-layout of map-value is a configuration interface.
5310 * New members can only be added to the end of this structure.
5312 struct bpf_cpumap_val {
5313 __u32 qsize; /* queue size to remote target CPU */
5315 int fd; /* prog fd on map write */
5316 __u32 id; /* prog id on map read */
5325 /* user accessible metadata for SK_MSG packet hook, new fields must
5326 * be added to the end of this structure
5329 __bpf_md_ptr(void *, data);
5330 __bpf_md_ptr(void *, data_end);
5333 __u32 remote_ip4; /* Stored in network byte order */
5334 __u32 local_ip4; /* Stored in network byte order */
5335 __u32 remote_ip6[4]; /* Stored in network byte order */
5336 __u32 local_ip6[4]; /* Stored in network byte order */
5337 __u32 remote_port; /* Stored in network byte order */
5338 __u32 local_port; /* stored in host byte order */
5339 __u32 size; /* Total size of sk_msg */
5341 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
5344 struct sk_reuseport_md {
5346 * Start of directly accessible data. It begins from
5347 * the tcp/udp header.
5349 __bpf_md_ptr(void *, data);
5350 /* End of directly accessible data */
5351 __bpf_md_ptr(void *, data_end);
5353 * Total length of packet (starting from the tcp/udp header).
5354 * Note that the directly accessible bytes (data_end - data)
5355 * could be less than this "len". Those bytes could be
5356 * indirectly read by a helper "bpf_skb_load_bytes()".
5360 * Eth protocol in the mac header (network byte order). e.g.
5361 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
5364 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
5365 __u32 bind_inany; /* Is sock bound to an INANY address? */
5366 __u32 hash; /* A hash of the packet 4 tuples */
5369 #define BPF_TAG_SIZE 8
5371 struct bpf_prog_info {
5374 __u8 tag[BPF_TAG_SIZE];
5375 __u32 jited_prog_len;
5376 __u32 xlated_prog_len;
5377 __aligned_u64 jited_prog_insns;
5378 __aligned_u64 xlated_prog_insns;
5379 __u64 load_time; /* ns since boottime */
5380 __u32 created_by_uid;
5382 __aligned_u64 map_ids;
5383 char name[BPF_OBJ_NAME_LEN];
5385 __u32 gpl_compatible:1;
5386 __u32 :31; /* alignment pad */
5389 __u32 nr_jited_ksyms;
5390 __u32 nr_jited_func_lens;
5391 __aligned_u64 jited_ksyms;
5392 __aligned_u64 jited_func_lens;
5394 __u32 func_info_rec_size;
5395 __aligned_u64 func_info;
5398 __aligned_u64 line_info;
5399 __aligned_u64 jited_line_info;
5400 __u32 nr_jited_line_info;
5401 __u32 line_info_rec_size;
5402 __u32 jited_line_info_rec_size;
5404 __aligned_u64 prog_tags;
5407 __u64 recursion_misses;
5408 } __attribute__((aligned(8)));
5410 struct bpf_map_info {
5417 char name[BPF_OBJ_NAME_LEN];
5419 __u32 btf_vmlinux_value_type_id;
5423 __u32 btf_key_type_id;
5424 __u32 btf_value_type_id;
5425 } __attribute__((aligned(8)));
5427 struct bpf_btf_info {
5434 } __attribute__((aligned(8)));
5436 struct bpf_link_info {
5442 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
5443 __u32 tp_name_len; /* in/out: tp_name buffer len */
5447 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
5448 __u32 target_btf_id; /* BTF type id inside the object */
5455 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
5456 __u32 target_name_len; /* in/out: target_name buffer len */
5471 } __attribute__((aligned(8)));
5473 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
5474 * by user and intended to be used by socket (e.g. to bind to, depends on
5477 struct bpf_sock_addr {
5478 __u32 user_family; /* Allows 4-byte read, but no write. */
5479 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5480 * Stored in network byte order.
5482 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5483 * Stored in network byte order.
5485 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
5486 * Stored in network byte order
5488 __u32 family; /* Allows 4-byte read, but no write */
5489 __u32 type; /* Allows 4-byte read, but no write */
5490 __u32 protocol; /* Allows 4-byte read, but no write */
5491 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
5492 * Stored in network byte order.
5494 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
5495 * Stored in network byte order.
5497 __bpf_md_ptr(struct bpf_sock *, sk);
5500 /* User bpf_sock_ops struct to access socket values and specify request ops
5501 * and their replies.
5502 * Some of this fields are in network (bigendian) byte order and may need
5503 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
5504 * New fields can only be added at the end of this structure
5506 struct bpf_sock_ops {
5509 __u32 args[4]; /* Optionally passed to bpf program */
5510 __u32 reply; /* Returned by bpf program */
5511 __u32 replylong[4]; /* Optionally returned by bpf prog */
5514 __u32 remote_ip4; /* Stored in network byte order */
5515 __u32 local_ip4; /* Stored in network byte order */
5516 __u32 remote_ip6[4]; /* Stored in network byte order */
5517 __u32 local_ip6[4]; /* Stored in network byte order */
5518 __u32 remote_port; /* Stored in network byte order */
5519 __u32 local_port; /* stored in host byte order */
5520 __u32 is_fullsock; /* Some TCP fields are only valid if
5521 * there is a full socket. If not, the
5522 * fields read as zero.
5525 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
5526 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
5535 __u32 rate_delivered;
5536 __u32 rate_interval_us;
5539 __u32 total_retrans;
5543 __u32 data_segs_out;
5547 __u64 bytes_received;
5549 __bpf_md_ptr(struct bpf_sock *, sk);
5550 /* [skb_data, skb_data_end) covers the whole TCP header.
5552 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
5553 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
5554 * header has not been written.
5555 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
5556 * been written so far.
5557 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
5559 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
5562 * bpf_load_hdr_opt() can also be used to read a particular option.
5564 __bpf_md_ptr(void *, skb_data);
5565 __bpf_md_ptr(void *, skb_data_end);
5566 __u32 skb_len; /* The total length of a packet.
5567 * It includes the header, options,
5570 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
5571 * an easy way to check for tcp_flags
5572 * without parsing skb_data.
5574 * In particular, the skb_tcp_flags
5575 * will still be available in
5576 * BPF_SOCK_OPS_HDR_OPT_LEN even though
5577 * the outgoing header has not
5582 /* Definitions for bpf_sock_ops_cb_flags */
5584 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
5585 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
5586 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
5587 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
5588 /* Call bpf for all received TCP headers. The bpf prog will be
5589 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5591 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5592 * for the header option related helpers that will be useful
5593 * to the bpf programs.
5595 * It could be used at the client/active side (i.e. connect() side)
5596 * when the server told it that the server was in syncookie
5597 * mode and required the active side to resend the bpf-written
5598 * options. The active side can keep writing the bpf-options until
5599 * it received a valid packet from the server side to confirm
5600 * the earlier packet (and options) has been received. The later
5601 * example patch is using it like this at the active side when the
5602 * server is in syncookie mode.
5604 * The bpf prog will usually turn this off in the common cases.
5606 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
5607 /* Call bpf when kernel has received a header option that
5608 * the kernel cannot handle. The bpf prog will be called under
5609 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
5611 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5612 * for the header option related helpers that will be useful
5613 * to the bpf programs.
5615 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
5616 /* Call bpf when the kernel is writing header options for the
5617 * outgoing packet. The bpf prog will first be called
5618 * to reserve space in a skb under
5619 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
5620 * the bpf prog will be called to write the header option(s)
5621 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5623 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
5624 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
5625 * related helpers that will be useful to the bpf programs.
5627 * The kernel gets its chance to reserve space and write
5628 * options first before the BPF program does.
5630 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
5631 /* Mask of all currently supported cb flags */
5632 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
5635 /* List of known BPF sock_ops operators.
5636 * New entries can only be added at the end
5640 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
5641 * -1 if default value should be used
5643 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
5644 * window (in packets) or -1 if default
5645 * value should be used
5647 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
5648 * active connection is initialized
5650 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
5651 * active connection is
5654 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
5655 * passive connection is
5658 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
5661 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
5662 * based on the path and may be
5663 * dependent on the congestion control
5664 * algorithm. In general it indicates
5665 * a congestion threshold. RTTs above
5666 * this indicate congestion
5668 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
5669 * Arg1: value of icsk_retransmits
5670 * Arg2: value of icsk_rto
5671 * Arg3: whether RTO has expired
5673 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
5674 * Arg1: sequence number of 1st byte
5676 * Arg3: return value of
5677 * tcp_transmit_skb (0 => success)
5679 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
5683 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
5684 * socket transition to LISTEN state.
5686 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
5688 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
5689 * It will be called to handle
5690 * the packets received at
5691 * an already established
5694 * sock_ops->skb_data:
5695 * Referring to the received skb.
5696 * It covers the TCP header only.
5698 * bpf_load_hdr_opt() can also
5699 * be used to search for a
5700 * particular option.
5702 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
5703 * header option later in
5704 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5705 * Arg1: bool want_cookie. (in
5706 * writing SYNACK only)
5708 * sock_ops->skb_data:
5709 * Not available because no header has
5712 * sock_ops->skb_tcp_flags:
5713 * The tcp_flags of the
5714 * outgoing skb. (e.g. SYN, ACK, FIN).
5716 * bpf_reserve_hdr_opt() should
5717 * be used to reserve space.
5719 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
5720 * Arg1: bool want_cookie. (in
5721 * writing SYNACK only)
5723 * sock_ops->skb_data:
5724 * Referring to the outgoing skb.
5725 * It covers the TCP header
5726 * that has already been written
5727 * by the kernel and the
5728 * earlier bpf-progs.
5730 * sock_ops->skb_tcp_flags:
5731 * The tcp_flags of the outgoing
5732 * skb. (e.g. SYN, ACK, FIN).
5734 * bpf_store_hdr_opt() should
5735 * be used to write the
5738 * bpf_load_hdr_opt() can also
5739 * be used to search for a
5740 * particular option that
5741 * has already been written
5742 * by the kernel or the
5743 * earlier bpf-progs.
5747 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
5748 * changes between the TCP and BPF versions. Ideally this should never happen.
5749 * If it does, we need to add code to convert them before calling
5750 * the BPF sock_ops function.
5753 BPF_TCP_ESTABLISHED = 1,
5763 BPF_TCP_CLOSING, /* Now a valid state */
5764 BPF_TCP_NEW_SYN_RECV,
5766 BPF_TCP_MAX_STATES /* Leave at the end! */
5770 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
5771 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
5772 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
5773 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
5774 /* Copy the SYN pkt to optval
5776 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
5777 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
5778 * to only getting from the saved_syn. It can either get the
5781 * 1. the just-received SYN packet (only available when writing the
5782 * SYNACK). It will be useful when it is not necessary to
5783 * save the SYN packet for latter use. It is also the only way
5784 * to get the SYN during syncookie mode because the syn
5785 * packet cannot be saved during syncookie.
5789 * 2. the earlier saved syn which was done by
5790 * bpf_setsockopt(TCP_SAVE_SYN).
5792 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
5793 * SYN packet is obtained.
5795 * If the bpf-prog does not need the IP[46] header, the
5796 * bpf-prog can avoid parsing the IP header by using
5797 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
5798 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
5800 * >0: Total number of bytes copied
5801 * -ENOSPC: Not enough space in optval. Only optlen number of
5803 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
5804 * is not saved by setsockopt(TCP_SAVE_SYN).
5806 TCP_BPF_SYN = 1005, /* Copy the TCP header */
5807 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
5808 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
5812 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
5815 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
5816 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5819 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
5820 * total option spaces
5821 * required for an established
5822 * sk in order to calculate the
5823 * MSS. No skb is actually
5826 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
5827 * when sending a SYN.
5831 struct bpf_perf_event_value {
5838 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
5839 BPF_DEVCG_ACC_READ = (1ULL << 1),
5840 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
5844 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
5845 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
5848 struct bpf_cgroup_dev_ctx {
5849 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
5855 struct bpf_raw_tracepoint_args {
5859 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
5860 * OUTPUT: Do lookup from egress perspective; default is ingress
5863 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
5864 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
5868 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
5869 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
5870 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
5871 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
5872 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
5873 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
5874 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
5875 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
5876 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5879 struct bpf_fib_lookup {
5880 /* input: network family for lookup (AF_INET, AF_INET6)
5881 * output: network family of egress nexthop
5885 /* set if lookup is to consider L4 data - e.g., FIB rules */
5890 union { /* used for MTU check */
5891 /* input to lookup */
5892 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
5894 /* output: MTU value */
5897 /* input: L3 device index for lookup
5898 * output: device index from FIB lookup
5903 /* inputs to lookup */
5904 __u8 tos; /* AF_INET */
5905 __be32 flowinfo; /* AF_INET6, flow_label + priority */
5907 /* output: metric of fib result (IPv4/IPv6 only) */
5913 __u32 ipv6_src[4]; /* in6_addr; network order */
5916 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
5917 * network header. output: bpf_fib_lookup sets to gateway address
5918 * if FIB lookup returns gateway route
5922 __u32 ipv6_dst[4]; /* in6_addr; network order */
5926 __be16 h_vlan_proto;
5928 __u8 smac[6]; /* ETH_ALEN */
5929 __u8 dmac[6]; /* ETH_ALEN */
5932 struct bpf_redir_neigh {
5933 /* network family for lookup (AF_INET, AF_INET6) */
5935 /* network address of nexthop; skips fib lookup to find gateway */
5938 __u32 ipv6_nh[4]; /* in6_addr; network order */
5942 /* bpf_check_mtu flags*/
5943 enum bpf_check_mtu_flags {
5944 BPF_MTU_CHK_SEGS = (1U << 0),
5947 enum bpf_check_mtu_ret {
5948 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
5949 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
5950 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
5953 enum bpf_task_fd_type {
5954 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
5955 BPF_FD_TYPE_TRACEPOINT, /* tp name */
5956 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
5957 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
5958 BPF_FD_TYPE_UPROBE, /* filename + offset */
5959 BPF_FD_TYPE_URETPROBE, /* filename + offset */
5963 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
5964 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
5965 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
5968 struct bpf_flow_keys {
5971 __u16 addr_proto; /* ETH_P_* of valid addrs */
5985 __u32 ipv6_src[4]; /* in6_addr; network order */
5986 __u32 ipv6_dst[4]; /* in6_addr; network order */
5993 struct bpf_func_info {
5998 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
5999 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
6001 struct bpf_line_info {
6003 __u32 file_name_off;
6008 struct bpf_spin_lock {
6013 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
6014 * Allows 1,2,4-byte read, but no write.
6016 __u32 file_pos; /* Sysctl file position to read from, write to.
6017 * Allows 1,2,4-byte read an 4-byte write.
6021 struct bpf_sockopt {
6022 __bpf_md_ptr(struct bpf_sock *, sk);
6023 __bpf_md_ptr(void *, optval);
6024 __bpf_md_ptr(void *, optval_end);
6032 struct bpf_pidns_info {
6037 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6038 struct bpf_sk_lookup {
6040 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6041 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6044 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
6045 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6046 __u32 remote_ip4; /* Network byte order */
6047 __u32 remote_ip6[4]; /* Network byte order */
6048 __u32 remote_port; /* Network byte order */
6049 __u32 local_ip4; /* Network byte order */
6050 __u32 local_ip6[4]; /* Network byte order */
6051 __u32 local_port; /* Host byte order */
6055 * struct btf_ptr is used for typed pointer representation; the
6056 * type id is used to render the pointer data as the appropriate type
6057 * via the bpf_snprintf_btf() helper described above. A flags field -
6058 * potentially to specify additional details about the BTF pointer
6059 * (rather than its mode of display) - is included for future use.
6060 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6065 __u32 flags; /* BTF ptr flags; unused at present. */
6069 * Flags to control bpf_snprintf_btf() behaviour.
6070 * - BTF_F_COMPACT: no formatting around type information
6071 * - BTF_F_NONAME: no struct/union member names/types
6072 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6073 * equivalent to %px.
6074 * - BTF_F_ZERO: show zero-valued struct/union members; they
6075 * are not displayed by default
6078 BTF_F_COMPACT = (1ULL << 0),
6079 BTF_F_NONAME = (1ULL << 1),
6080 BTF_F_PTR_RAW = (1ULL << 2),
6081 BTF_F_ZERO = (1ULL << 3),
6084 #endif /* _UAPI__LINUX_BPF_H__ */