1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_JMP32 0x06 /* jmp mode in word width */
18 #define BPF_ALU64 0x07 /* alu mode in double word width */
21 #define BPF_DW 0x18 /* double word (64-bit) */
22 #define BPF_MEMSX 0x80 /* load with sign extension */
23 #define BPF_ATOMIC 0xc0 /* atomic memory ops - op type in immediate */
24 #define BPF_XADD 0xc0 /* exclusive add - legacy name */
27 #define BPF_MOV 0xb0 /* mov reg to reg */
28 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
30 /* change endianness of a register */
31 #define BPF_END 0xd0 /* flags for endianness conversion: */
32 #define BPF_TO_LE 0x00 /* convert to little-endian */
33 #define BPF_TO_BE 0x08 /* convert to big-endian */
34 #define BPF_FROM_LE BPF_TO_LE
35 #define BPF_FROM_BE BPF_TO_BE
38 #define BPF_JNE 0x50 /* jump != */
39 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
40 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
41 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
42 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
43 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
44 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
45 #define BPF_CALL 0x80 /* function call */
46 #define BPF_EXIT 0x90 /* function return */
48 /* atomic op type fields (stored in immediate) */
49 #define BPF_FETCH 0x01 /* not an opcode on its own, used to build others */
50 #define BPF_XCHG (0xe0 | BPF_FETCH) /* atomic exchange */
51 #define BPF_CMPXCHG (0xf0 | BPF_FETCH) /* atomic compare-and-write */
53 /* Register numbers */
69 /* BPF has 10 general purpose 64-bit registers and stack frame. */
70 #define MAX_BPF_REG __MAX_BPF_REG
73 __u8 code; /* opcode */
74 __u8 dst_reg:4; /* dest register */
75 __u8 src_reg:4; /* source register */
76 __s16 off; /* signed offset */
77 __s32 imm; /* signed immediate constant */
80 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
81 struct bpf_lpm_trie_key {
82 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
83 __u8 data[0]; /* Arbitrary size */
86 struct bpf_cgroup_storage_key {
87 __u64 cgroup_inode_id; /* cgroup inode id */
88 __u32 attach_type; /* program attach type (enum bpf_attach_type) */
91 enum bpf_cgroup_iter_order {
92 BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
93 BPF_CGROUP_ITER_SELF_ONLY, /* process only a single object. */
94 BPF_CGROUP_ITER_DESCENDANTS_PRE, /* walk descendants in pre-order. */
95 BPF_CGROUP_ITER_DESCENDANTS_POST, /* walk descendants in post-order. */
96 BPF_CGROUP_ITER_ANCESTORS_UP, /* walk ancestors upward. */
99 union bpf_iter_link_info {
104 enum bpf_cgroup_iter_order order;
106 /* At most one of cgroup_fd and cgroup_id can be non-zero. If
107 * both are zero, the walk starts from the default cgroup v2
108 * root. For walking v1 hierarchy, one should always explicitly
114 /* Parameters of task iterators. */
122 /* BPF syscall commands, see bpf(2) man-page for more details. */
124 * DOC: eBPF Syscall Preamble
126 * The operation to be performed by the **bpf**\ () system call is determined
127 * by the *cmd* argument. Each operation takes an accompanying argument,
128 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
129 * below). The size argument is the size of the union pointed to by *attr*.
132 * DOC: eBPF Syscall Commands
136 * Create a map and return a file descriptor that refers to the
137 * map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
138 * is automatically enabled for the new file descriptor.
140 * Applying **close**\ (2) to the file descriptor returned by
141 * **BPF_MAP_CREATE** will delete the map (but see NOTES).
144 * A new file descriptor (a nonnegative integer), or -1 if an
145 * error occurred (in which case, *errno* is set appropriately).
147 * BPF_MAP_LOOKUP_ELEM
149 * Look up an element with a given *key* in the map referred to
150 * by the file descriptor *map_fd*.
152 * The *flags* argument may be specified as one of the
156 * Look up the value of a spin-locked map without
157 * returning the lock. This must be specified if the
158 * elements contain a spinlock.
161 * Returns zero on success. On error, -1 is returned and *errno*
162 * is set appropriately.
164 * BPF_MAP_UPDATE_ELEM
166 * Create or update an element (key/value pair) in a specified map.
168 * The *flags* argument should be specified as one of the
172 * Create a new element or update an existing element.
174 * Create a new element only if it did not exist.
176 * Update an existing element.
178 * Update a spin_lock-ed map element.
181 * Returns zero on success. On error, -1 is returned and *errno*
182 * is set appropriately.
184 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
185 * **E2BIG**, **EEXIST**, or **ENOENT**.
188 * The number of elements in the map reached the
189 * *max_entries* limit specified at map creation time.
191 * If *flags* specifies **BPF_NOEXIST** and the element
192 * with *key* already exists in the map.
194 * If *flags* specifies **BPF_EXIST** and the element with
195 * *key* does not exist in the map.
197 * BPF_MAP_DELETE_ELEM
199 * Look up and delete an element by key in a specified map.
202 * Returns zero on success. On error, -1 is returned and *errno*
203 * is set appropriately.
205 * BPF_MAP_GET_NEXT_KEY
207 * Look up an element by key in a specified map and return the key
208 * of the next element. Can be used to iterate over all elements
212 * Returns zero on success. On error, -1 is returned and *errno*
213 * is set appropriately.
215 * The following cases can be used to iterate over all elements of
218 * * If *key* is not found, the operation returns zero and sets
219 * the *next_key* pointer to the key of the first element.
220 * * If *key* is found, the operation returns zero and sets the
221 * *next_key* pointer to the key of the next element.
222 * * If *key* is the last element, returns -1 and *errno* is set
225 * May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
226 * **EINVAL** on error.
230 * Verify and load an eBPF program, returning a new file
231 * descriptor associated with the program.
233 * Applying **close**\ (2) to the file descriptor returned by
234 * **BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
236 * The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
237 * automatically enabled for the new file descriptor.
240 * A new file descriptor (a nonnegative integer), or -1 if an
241 * error occurred (in which case, *errno* is set appropriately).
245 * Pin an eBPF program or map referred by the specified *bpf_fd*
246 * to the provided *pathname* on the filesystem.
248 * The *pathname* argument must not contain a dot (".").
250 * On success, *pathname* retains a reference to the eBPF object,
251 * preventing deallocation of the object when the original
252 * *bpf_fd* is closed. This allow the eBPF object to live beyond
253 * **close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
256 * Applying **unlink**\ (2) or similar calls to the *pathname*
257 * unpins the object from the filesystem, removing the reference.
258 * If no other file descriptors or filesystem nodes refer to the
259 * same object, it will be deallocated (see NOTES).
261 * The filesystem type for the parent directory of *pathname* must
262 * be **BPF_FS_MAGIC**.
265 * Returns zero on success. On error, -1 is returned and *errno*
266 * is set appropriately.
270 * Open a file descriptor for the eBPF object pinned to the
271 * specified *pathname*.
274 * A new file descriptor (a nonnegative integer), or -1 if an
275 * error occurred (in which case, *errno* is set appropriately).
279 * Attach an eBPF program to a *target_fd* at the specified
280 * *attach_type* hook.
282 * The *attach_type* specifies the eBPF attachment point to
283 * attach the program to, and must be one of *bpf_attach_type*
286 * The *attach_bpf_fd* must be a valid file descriptor for a
287 * loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
288 * or sock_ops type corresponding to the specified *attach_type*.
290 * The *target_fd* must be a valid file descriptor for a kernel
291 * object which depends on the attach type of *attach_bpf_fd*:
293 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
294 * **BPF_PROG_TYPE_CGROUP_SKB**,
295 * **BPF_PROG_TYPE_CGROUP_SOCK**,
296 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
297 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
298 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
299 * **BPF_PROG_TYPE_SOCK_OPS**
301 * Control Group v2 hierarchy with the eBPF controller
302 * enabled. Requires the kernel to be compiled with
303 * **CONFIG_CGROUP_BPF**.
305 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
307 * Network namespace (eg /proc/self/ns/net).
309 * **BPF_PROG_TYPE_LIRC_MODE2**
311 * LIRC device path (eg /dev/lircN). Requires the kernel
312 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
314 * **BPF_PROG_TYPE_SK_SKB**,
315 * **BPF_PROG_TYPE_SK_MSG**
317 * eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
320 * Returns zero on success. On error, -1 is returned and *errno*
321 * is set appropriately.
325 * Detach the eBPF program associated with the *target_fd* at the
326 * hook specified by *attach_type*. The program must have been
327 * previously attached using **BPF_PROG_ATTACH**.
330 * Returns zero on success. On error, -1 is returned and *errno*
331 * is set appropriately.
335 * Run the eBPF program associated with the *prog_fd* a *repeat*
336 * number of times against a provided program context *ctx_in* and
337 * data *data_in*, and return the modified program context
338 * *ctx_out*, *data_out* (for example, packet data), result of the
339 * execution *retval*, and *duration* of the test run.
341 * The sizes of the buffers provided as input and output
342 * parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
343 * be provided in the corresponding variables *ctx_size_in*,
344 * *ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
345 * of these parameters are not provided (ie set to NULL), the
346 * corresponding size field must be zero.
348 * Some program types have particular requirements:
350 * **BPF_PROG_TYPE_SK_LOOKUP**
351 * *data_in* and *data_out* must be NULL.
353 * **BPF_PROG_TYPE_RAW_TRACEPOINT**,
354 * **BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
356 * *ctx_out*, *data_in* and *data_out* must be NULL.
357 * *repeat* must be zero.
359 * BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
362 * Returns zero on success. On error, -1 is returned and *errno*
363 * is set appropriately.
366 * Either *data_size_out* or *ctx_size_out* is too small.
368 * This command is not supported by the program type of
369 * the program referred to by *prog_fd*.
371 * BPF_PROG_GET_NEXT_ID
373 * Fetch the next eBPF program currently loaded into the kernel.
375 * Looks for the eBPF program with an id greater than *start_id*
376 * and updates *next_id* on success. If no other eBPF programs
377 * remain with ids higher than *start_id*, returns -1 and sets
378 * *errno* to **ENOENT**.
381 * Returns zero on success. On error, or when no id remains, -1
382 * is returned and *errno* is set appropriately.
384 * BPF_MAP_GET_NEXT_ID
386 * Fetch the next eBPF map currently loaded into the kernel.
388 * Looks for the eBPF map with an id greater than *start_id*
389 * and updates *next_id* on success. If no other eBPF maps
390 * remain with ids higher than *start_id*, returns -1 and sets
391 * *errno* to **ENOENT**.
394 * Returns zero on success. On error, or when no id remains, -1
395 * is returned and *errno* is set appropriately.
397 * BPF_PROG_GET_FD_BY_ID
399 * Open a file descriptor for the eBPF program corresponding to
403 * A new file descriptor (a nonnegative integer), or -1 if an
404 * error occurred (in which case, *errno* is set appropriately).
406 * BPF_MAP_GET_FD_BY_ID
408 * Open a file descriptor for the eBPF map corresponding to
412 * A new file descriptor (a nonnegative integer), or -1 if an
413 * error occurred (in which case, *errno* is set appropriately).
415 * BPF_OBJ_GET_INFO_BY_FD
417 * Obtain information about the eBPF object corresponding to
420 * Populates up to *info_len* bytes of *info*, which will be in
421 * one of the following formats depending on the eBPF object type
424 * * **struct bpf_prog_info**
425 * * **struct bpf_map_info**
426 * * **struct bpf_btf_info**
427 * * **struct bpf_link_info**
430 * Returns zero on success. On error, -1 is returned and *errno*
431 * is set appropriately.
435 * Obtain information about eBPF programs associated with the
436 * specified *attach_type* hook.
438 * The *target_fd* must be a valid file descriptor for a kernel
439 * object which depends on the attach type of *attach_bpf_fd*:
441 * **BPF_PROG_TYPE_CGROUP_DEVICE**,
442 * **BPF_PROG_TYPE_CGROUP_SKB**,
443 * **BPF_PROG_TYPE_CGROUP_SOCK**,
444 * **BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
445 * **BPF_PROG_TYPE_CGROUP_SOCKOPT**,
446 * **BPF_PROG_TYPE_CGROUP_SYSCTL**,
447 * **BPF_PROG_TYPE_SOCK_OPS**
449 * Control Group v2 hierarchy with the eBPF controller
450 * enabled. Requires the kernel to be compiled with
451 * **CONFIG_CGROUP_BPF**.
453 * **BPF_PROG_TYPE_FLOW_DISSECTOR**
455 * Network namespace (eg /proc/self/ns/net).
457 * **BPF_PROG_TYPE_LIRC_MODE2**
459 * LIRC device path (eg /dev/lircN). Requires the kernel
460 * to be compiled with **CONFIG_BPF_LIRC_MODE2**.
462 * **BPF_PROG_QUERY** always fetches the number of programs
463 * attached and the *attach_flags* which were used to attach those
464 * programs. Additionally, if *prog_ids* is nonzero and the number
465 * of attached programs is less than *prog_cnt*, populates
466 * *prog_ids* with the eBPF program ids of the programs attached
469 * The following flags may alter the result:
471 * **BPF_F_QUERY_EFFECTIVE**
472 * Only return information regarding programs which are
473 * currently effective at the specified *target_fd*.
476 * Returns zero on success. On error, -1 is returned and *errno*
477 * is set appropriately.
479 * BPF_RAW_TRACEPOINT_OPEN
481 * Attach an eBPF program to a tracepoint *name* to access kernel
482 * internal arguments of the tracepoint in their raw form.
484 * The *prog_fd* must be a valid file descriptor associated with
485 * a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
487 * No ABI guarantees are made about the content of tracepoint
488 * arguments exposed to the corresponding eBPF program.
490 * Applying **close**\ (2) to the file descriptor returned by
491 * **BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
494 * A new file descriptor (a nonnegative integer), or -1 if an
495 * error occurred (in which case, *errno* is set appropriately).
499 * Verify and load BPF Type Format (BTF) metadata into the kernel,
500 * returning a new file descriptor associated with the metadata.
501 * BTF is described in more detail at
502 * https://www.kernel.org/doc/html/latest/bpf/btf.html.
504 * The *btf* parameter must point to valid memory providing
505 * *btf_size* bytes of BTF binary metadata.
507 * The returned file descriptor can be passed to other **bpf**\ ()
508 * subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
509 * associate the BTF with those objects.
511 * Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
512 * parameters to specify a *btf_log_buf*, *btf_log_size* and
513 * *btf_log_level* which allow the kernel to return freeform log
514 * output regarding the BTF verification process.
517 * A new file descriptor (a nonnegative integer), or -1 if an
518 * error occurred (in which case, *errno* is set appropriately).
520 * BPF_BTF_GET_FD_BY_ID
522 * Open a file descriptor for the BPF Type Format (BTF)
523 * corresponding to *btf_id*.
526 * A new file descriptor (a nonnegative integer), or -1 if an
527 * error occurred (in which case, *errno* is set appropriately).
531 * Obtain information about eBPF programs associated with the
532 * target process identified by *pid* and *fd*.
534 * If the *pid* and *fd* are associated with a tracepoint, kprobe
535 * or uprobe perf event, then the *prog_id* and *fd_type* will
536 * be populated with the eBPF program id and file descriptor type
537 * of type **bpf_task_fd_type**. If associated with a kprobe or
538 * uprobe, the *probe_offset* and *probe_addr* will also be
539 * populated. Optionally, if *buf* is provided, then up to
540 * *buf_len* bytes of *buf* will be populated with the name of
541 * the tracepoint, kprobe or uprobe.
543 * The resulting *prog_id* may be introspected in deeper detail
544 * using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
547 * Returns zero on success. On error, -1 is returned and *errno*
548 * is set appropriately.
550 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
552 * Look up an element with the given *key* in the map referred to
553 * by the file descriptor *fd*, and if found, delete the element.
555 * For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
556 * types, the *flags* argument needs to be set to 0, but for other
557 * map types, it may be specified as:
560 * Look up and delete the value of a spin-locked map
561 * without returning the lock. This must be specified if
562 * the elements contain a spinlock.
564 * The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
565 * implement this command as a "pop" operation, deleting the top
566 * element rather than one corresponding to *key*.
567 * The *key* and *key_len* parameters should be zeroed when
568 * issuing this operation for these map types.
570 * This command is only valid for the following map types:
571 * * **BPF_MAP_TYPE_QUEUE**
572 * * **BPF_MAP_TYPE_STACK**
573 * * **BPF_MAP_TYPE_HASH**
574 * * **BPF_MAP_TYPE_PERCPU_HASH**
575 * * **BPF_MAP_TYPE_LRU_HASH**
576 * * **BPF_MAP_TYPE_LRU_PERCPU_HASH**
579 * Returns zero on success. On error, -1 is returned and *errno*
580 * is set appropriately.
584 * Freeze the permissions of the specified map.
586 * Write permissions may be frozen by passing zero *flags*.
587 * Upon success, no future syscall invocations may alter the
588 * map state of *map_fd*. Write operations from eBPF programs
589 * are still possible for a frozen map.
591 * Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
594 * Returns zero on success. On error, -1 is returned and *errno*
595 * is set appropriately.
597 * BPF_BTF_GET_NEXT_ID
599 * Fetch the next BPF Type Format (BTF) object currently loaded
602 * Looks for the BTF object with an id greater than *start_id*
603 * and updates *next_id* on success. If no other BTF objects
604 * remain with ids higher than *start_id*, returns -1 and sets
605 * *errno* to **ENOENT**.
608 * Returns zero on success. On error, or when no id remains, -1
609 * is returned and *errno* is set appropriately.
611 * BPF_MAP_LOOKUP_BATCH
613 * Iterate and fetch multiple elements in a map.
615 * Two opaque values are used to manage batch operations,
616 * *in_batch* and *out_batch*. Initially, *in_batch* must be set
617 * to NULL to begin the batched operation. After each subsequent
618 * **BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
619 * *out_batch* as the *in_batch* for the next operation to
620 * continue iteration from the current point.
622 * The *keys* and *values* are output parameters which must point
623 * to memory large enough to hold *count* items based on the key
624 * and value size of the map *map_fd*. The *keys* buffer must be
625 * of *key_size* * *count*. The *values* buffer must be of
626 * *value_size* * *count*.
628 * The *elem_flags* argument may be specified as one of the
632 * Look up the value of a spin-locked map without
633 * returning the lock. This must be specified if the
634 * elements contain a spinlock.
636 * On success, *count* elements from the map are copied into the
637 * user buffer, with the keys copied into *keys* and the values
638 * copied into the corresponding indices in *values*.
640 * If an error is returned and *errno* is not **EFAULT**, *count*
641 * is set to the number of successfully processed elements.
644 * Returns zero on success. On error, -1 is returned and *errno*
645 * is set appropriately.
647 * May set *errno* to **ENOSPC** to indicate that *keys* or
648 * *values* is too small to dump an entire bucket during
649 * iteration of a hash-based map type.
651 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
653 * Iterate and delete all elements in a map.
655 * This operation has the same behavior as
656 * **BPF_MAP_LOOKUP_BATCH** with two exceptions:
658 * * Every element that is successfully returned is also deleted
659 * from the map. This is at least *count* elements. Note that
660 * *count* is both an input and an output parameter.
661 * * Upon returning with *errno* set to **EFAULT**, up to
662 * *count* elements may be deleted without returning the keys
663 * and values of the deleted elements.
666 * Returns zero on success. On error, -1 is returned and *errno*
667 * is set appropriately.
669 * BPF_MAP_UPDATE_BATCH
671 * Update multiple elements in a map by *key*.
673 * The *keys* and *values* are input parameters which must point
674 * to memory large enough to hold *count* items based on the key
675 * and value size of the map *map_fd*. The *keys* buffer must be
676 * of *key_size* * *count*. The *values* buffer must be of
677 * *value_size* * *count*.
679 * Each element specified in *keys* is sequentially updated to the
680 * value in the corresponding index in *values*. The *in_batch*
681 * and *out_batch* parameters are ignored and should be zeroed.
683 * The *elem_flags* argument should be specified as one of the
687 * Create new elements or update a existing elements.
689 * Create new elements only if they do not exist.
691 * Update existing elements.
693 * Update spin_lock-ed map elements. This must be
694 * specified if the map value contains a spinlock.
696 * On success, *count* elements from the map are updated.
698 * If an error is returned and *errno* is not **EFAULT**, *count*
699 * is set to the number of successfully processed elements.
702 * Returns zero on success. On error, -1 is returned and *errno*
703 * is set appropriately.
705 * May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
706 * **E2BIG**. **E2BIG** indicates that the number of elements in
707 * the map reached the *max_entries* limit specified at map
710 * May set *errno* to one of the following error codes under
711 * specific circumstances:
714 * If *flags* specifies **BPF_NOEXIST** and the element
715 * with *key* already exists in the map.
717 * If *flags* specifies **BPF_EXIST** and the element with
718 * *key* does not exist in the map.
720 * BPF_MAP_DELETE_BATCH
722 * Delete multiple elements in a map by *key*.
724 * The *keys* parameter is an input parameter which must point
725 * to memory large enough to hold *count* items based on the key
726 * size of the map *map_fd*, that is, *key_size* * *count*.
728 * Each element specified in *keys* is sequentially deleted. The
729 * *in_batch*, *out_batch*, and *values* parameters are ignored
730 * and should be zeroed.
732 * The *elem_flags* argument may be specified as one of the
736 * Look up the value of a spin-locked map without
737 * returning the lock. This must be specified if the
738 * elements contain a spinlock.
740 * On success, *count* elements from the map are updated.
742 * If an error is returned and *errno* is not **EFAULT**, *count*
743 * is set to the number of successfully processed elements. If
744 * *errno* is **EFAULT**, up to *count* elements may be been
748 * Returns zero on success. On error, -1 is returned and *errno*
749 * is set appropriately.
753 * Attach an eBPF program to a *target_fd* at the specified
754 * *attach_type* hook and return a file descriptor handle for
758 * A new file descriptor (a nonnegative integer), or -1 if an
759 * error occurred (in which case, *errno* is set appropriately).
763 * Update the eBPF program in the specified *link_fd* to
767 * Returns zero on success. On error, -1 is returned and *errno*
768 * is set appropriately.
770 * BPF_LINK_GET_FD_BY_ID
772 * Open a file descriptor for the eBPF Link corresponding to
776 * A new file descriptor (a nonnegative integer), or -1 if an
777 * error occurred (in which case, *errno* is set appropriately).
779 * BPF_LINK_GET_NEXT_ID
781 * Fetch the next eBPF link currently loaded into the kernel.
783 * Looks for the eBPF link with an id greater than *start_id*
784 * and updates *next_id* on success. If no other eBPF links
785 * remain with ids higher than *start_id*, returns -1 and sets
786 * *errno* to **ENOENT**.
789 * Returns zero on success. On error, or when no id remains, -1
790 * is returned and *errno* is set appropriately.
794 * Enable eBPF runtime statistics gathering.
796 * Runtime statistics gathering for the eBPF runtime is disabled
797 * by default to minimize the corresponding performance overhead.
798 * This command enables statistics globally.
800 * Multiple programs may independently enable statistics.
801 * After gathering the desired statistics, eBPF runtime statistics
802 * may be disabled again by calling **close**\ (2) for the file
803 * descriptor returned by this function. Statistics will only be
804 * disabled system-wide when all outstanding file descriptors
805 * returned by prior calls for this subcommand are closed.
808 * A new file descriptor (a nonnegative integer), or -1 if an
809 * error occurred (in which case, *errno* is set appropriately).
813 * Create an iterator on top of the specified *link_fd* (as
814 * previously created using **BPF_LINK_CREATE**) and return a
815 * file descriptor that can be used to trigger the iteration.
817 * If the resulting file descriptor is pinned to the filesystem
818 * using **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
819 * for that path will trigger the iterator to read kernel state
820 * using the eBPF program attached to *link_fd*.
823 * A new file descriptor (a nonnegative integer), or -1 if an
824 * error occurred (in which case, *errno* is set appropriately).
828 * Forcefully detach the specified *link_fd* from its
829 * corresponding attachment point.
832 * Returns zero on success. On error, -1 is returned and *errno*
833 * is set appropriately.
837 * Bind a map to the lifetime of an eBPF program.
839 * The map identified by *map_fd* is bound to the program
840 * identified by *prog_fd* and only released when *prog_fd* is
841 * released. This may be used in cases where metadata should be
842 * associated with a program which otherwise does not contain any
843 * references to the map (for example, embedded in the eBPF
844 * program instructions).
847 * Returns zero on success. On error, -1 is returned and *errno*
848 * is set appropriately.
851 * eBPF objects (maps and programs) can be shared between processes.
853 * * After **fork**\ (2), the child inherits file descriptors
854 * referring to the same eBPF objects.
855 * * File descriptors referring to eBPF objects can be transferred over
856 * **unix**\ (7) domain sockets.
857 * * File descriptors referring to eBPF objects can be duplicated in the
858 * usual way, using **dup**\ (2) and similar calls.
859 * * File descriptors referring to eBPF objects can be pinned to the
860 * filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
862 * An eBPF object is deallocated only after all file descriptors referring
863 * to the object have been closed and no references remain pinned to the
864 * filesystem or attached (for example, bound to a program or device).
871 BPF_MAP_GET_NEXT_KEY,
878 BPF_PROG_RUN = BPF_PROG_TEST_RUN,
879 BPF_PROG_GET_NEXT_ID,
881 BPF_PROG_GET_FD_BY_ID,
882 BPF_MAP_GET_FD_BY_ID,
883 BPF_OBJ_GET_INFO_BY_FD,
885 BPF_RAW_TRACEPOINT_OPEN,
887 BPF_BTF_GET_FD_BY_ID,
889 BPF_MAP_LOOKUP_AND_DELETE_ELEM,
892 BPF_MAP_LOOKUP_BATCH,
893 BPF_MAP_LOOKUP_AND_DELETE_BATCH,
894 BPF_MAP_UPDATE_BATCH,
895 BPF_MAP_DELETE_BATCH,
898 BPF_LINK_GET_FD_BY_ID,
899 BPF_LINK_GET_NEXT_ID,
910 BPF_MAP_TYPE_PROG_ARRAY,
911 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
912 BPF_MAP_TYPE_PERCPU_HASH,
913 BPF_MAP_TYPE_PERCPU_ARRAY,
914 BPF_MAP_TYPE_STACK_TRACE,
915 BPF_MAP_TYPE_CGROUP_ARRAY,
916 BPF_MAP_TYPE_LRU_HASH,
917 BPF_MAP_TYPE_LRU_PERCPU_HASH,
918 BPF_MAP_TYPE_LPM_TRIE,
919 BPF_MAP_TYPE_ARRAY_OF_MAPS,
920 BPF_MAP_TYPE_HASH_OF_MAPS,
922 BPF_MAP_TYPE_SOCKMAP,
925 BPF_MAP_TYPE_SOCKHASH,
926 BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
927 /* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
928 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
929 * both cgroup-attached and other progs and supports all functionality
930 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
931 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
933 BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
934 BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
935 BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
938 BPF_MAP_TYPE_SK_STORAGE,
939 BPF_MAP_TYPE_DEVMAP_HASH,
940 BPF_MAP_TYPE_STRUCT_OPS,
941 BPF_MAP_TYPE_RINGBUF,
942 BPF_MAP_TYPE_INODE_STORAGE,
943 BPF_MAP_TYPE_TASK_STORAGE,
944 BPF_MAP_TYPE_BLOOM_FILTER,
945 BPF_MAP_TYPE_USER_RINGBUF,
946 BPF_MAP_TYPE_CGRP_STORAGE,
949 /* Note that tracing related programs such as
950 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
951 * are not subject to a stable API since kernel internal data
952 * structures can change from release to release and may
953 * therefore break existing tracing BPF programs. Tracing BPF
954 * programs correspond to /a/ specific kernel which is to be
955 * analyzed, and not /a/ specific kernel /and/ all future ones.
958 BPF_PROG_TYPE_UNSPEC,
959 BPF_PROG_TYPE_SOCKET_FILTER,
960 BPF_PROG_TYPE_KPROBE,
961 BPF_PROG_TYPE_SCHED_CLS,
962 BPF_PROG_TYPE_SCHED_ACT,
963 BPF_PROG_TYPE_TRACEPOINT,
965 BPF_PROG_TYPE_PERF_EVENT,
966 BPF_PROG_TYPE_CGROUP_SKB,
967 BPF_PROG_TYPE_CGROUP_SOCK,
968 BPF_PROG_TYPE_LWT_IN,
969 BPF_PROG_TYPE_LWT_OUT,
970 BPF_PROG_TYPE_LWT_XMIT,
971 BPF_PROG_TYPE_SOCK_OPS,
972 BPF_PROG_TYPE_SK_SKB,
973 BPF_PROG_TYPE_CGROUP_DEVICE,
974 BPF_PROG_TYPE_SK_MSG,
975 BPF_PROG_TYPE_RAW_TRACEPOINT,
976 BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
977 BPF_PROG_TYPE_LWT_SEG6LOCAL,
978 BPF_PROG_TYPE_LIRC_MODE2,
979 BPF_PROG_TYPE_SK_REUSEPORT,
980 BPF_PROG_TYPE_FLOW_DISSECTOR,
981 BPF_PROG_TYPE_CGROUP_SYSCTL,
982 BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
983 BPF_PROG_TYPE_CGROUP_SOCKOPT,
984 BPF_PROG_TYPE_TRACING,
985 BPF_PROG_TYPE_STRUCT_OPS,
988 BPF_PROG_TYPE_SK_LOOKUP,
989 BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
990 BPF_PROG_TYPE_NETFILTER,
993 enum bpf_attach_type {
994 BPF_CGROUP_INET_INGRESS,
995 BPF_CGROUP_INET_EGRESS,
996 BPF_CGROUP_INET_SOCK_CREATE,
998 BPF_SK_SKB_STREAM_PARSER,
999 BPF_SK_SKB_STREAM_VERDICT,
1002 BPF_CGROUP_INET4_BIND,
1003 BPF_CGROUP_INET6_BIND,
1004 BPF_CGROUP_INET4_CONNECT,
1005 BPF_CGROUP_INET6_CONNECT,
1006 BPF_CGROUP_INET4_POST_BIND,
1007 BPF_CGROUP_INET6_POST_BIND,
1008 BPF_CGROUP_UDP4_SENDMSG,
1009 BPF_CGROUP_UDP6_SENDMSG,
1013 BPF_CGROUP_UDP4_RECVMSG,
1014 BPF_CGROUP_UDP6_RECVMSG,
1015 BPF_CGROUP_GETSOCKOPT,
1016 BPF_CGROUP_SETSOCKOPT,
1023 BPF_CGROUP_INET4_GETPEERNAME,
1024 BPF_CGROUP_INET6_GETPEERNAME,
1025 BPF_CGROUP_INET4_GETSOCKNAME,
1026 BPF_CGROUP_INET6_GETSOCKNAME,
1028 BPF_CGROUP_INET_SOCK_RELEASE,
1033 BPF_SK_REUSEPORT_SELECT,
1034 BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1036 BPF_TRACE_KPROBE_MULTI,
1042 BPF_TRACE_UPROBE_MULTI,
1043 __MAX_BPF_ATTACH_TYPE
1046 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1048 enum bpf_link_type {
1049 BPF_LINK_TYPE_UNSPEC = 0,
1050 BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1051 BPF_LINK_TYPE_TRACING = 2,
1052 BPF_LINK_TYPE_CGROUP = 3,
1053 BPF_LINK_TYPE_ITER = 4,
1054 BPF_LINK_TYPE_NETNS = 5,
1055 BPF_LINK_TYPE_XDP = 6,
1056 BPF_LINK_TYPE_PERF_EVENT = 7,
1057 BPF_LINK_TYPE_KPROBE_MULTI = 8,
1058 BPF_LINK_TYPE_STRUCT_OPS = 9,
1059 BPF_LINK_TYPE_NETFILTER = 10,
1060 BPF_LINK_TYPE_TCX = 11,
1061 BPF_LINK_TYPE_UPROBE_MULTI = 12,
1065 enum bpf_perf_event_type {
1066 BPF_PERF_EVENT_UNSPEC = 0,
1067 BPF_PERF_EVENT_UPROBE = 1,
1068 BPF_PERF_EVENT_URETPROBE = 2,
1069 BPF_PERF_EVENT_KPROBE = 3,
1070 BPF_PERF_EVENT_KRETPROBE = 4,
1071 BPF_PERF_EVENT_TRACEPOINT = 5,
1072 BPF_PERF_EVENT_EVENT = 6,
1075 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1077 * NONE(default): No further bpf programs allowed in the subtree.
1079 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1080 * the program in this cgroup yields to sub-cgroup program.
1082 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1083 * that cgroup program gets run in addition to the program in this cgroup.
1085 * Only one program is allowed to be attached to a cgroup with
1086 * NONE or BPF_F_ALLOW_OVERRIDE flag.
1087 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1088 * release old program and attach the new one. Attach flags has to match.
1090 * Multiple programs are allowed to be attached to a cgroup with
1091 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1092 * (those that were attached first, run first)
1093 * The programs of sub-cgroup are executed first, then programs of
1094 * this cgroup and then programs of parent cgroup.
1095 * When children program makes decision (like picking TCP CA or sock bind)
1096 * parent program has a chance to override it.
1098 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1099 * programs for a cgroup. Though it's possible to replace an old program at
1100 * any position by also specifying BPF_F_REPLACE flag and position itself in
1101 * replace_bpf_fd attribute. Old program at this position will be released.
1103 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1104 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1106 * cgrp1 (MULTI progs A, B) ->
1107 * cgrp2 (OVERRIDE prog C) ->
1108 * cgrp3 (MULTI prog D) ->
1109 * cgrp4 (OVERRIDE prog E) ->
1110 * cgrp5 (NONE prog F)
1111 * the event in cgrp5 triggers execution of F,D,A,B in that order.
1112 * if prog F is detached, the execution is E,D,A,B
1113 * if prog F and D are detached, the execution is E,A,B
1114 * if prog F, E and D are detached, the execution is C,A,B
1116 * All eligible programs are executed regardless of return code from
1119 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
1120 #define BPF_F_ALLOW_MULTI (1U << 1)
1121 /* Generic attachment flags. */
1122 #define BPF_F_REPLACE (1U << 2)
1123 #define BPF_F_BEFORE (1U << 3)
1124 #define BPF_F_AFTER (1U << 4)
1125 #define BPF_F_ID (1U << 5)
1126 #define BPF_F_LINK BPF_F_LINK /* 1 << 13 */
1128 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1129 * verifier will perform strict alignment checking as if the kernel
1130 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1131 * and NET_IP_ALIGN defined to 2.
1133 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
1135 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1136 * verifier will allow any alignment whatsoever. On platforms
1137 * with strict alignment requirements for loads ands stores (such
1138 * as sparc and mips) the verifier validates that all loads and
1139 * stores provably follow this requirement. This flag turns that
1140 * checking and enforcement off.
1142 * It is mostly used for testing when we want to validate the
1143 * context and memory access aspects of the verifier, but because
1144 * of an unaligned access the alignment check would trigger before
1145 * the one we are interested in.
1147 #define BPF_F_ANY_ALIGNMENT (1U << 1)
1149 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1150 * Verifier does sub-register def/use analysis and identifies instructions whose
1151 * def only matters for low 32-bit, high 32-bit is never referenced later
1152 * through implicit zero extension. Therefore verifier notifies JIT back-ends
1153 * that it is safe to ignore clearing high 32-bit for these instructions. This
1154 * saves some back-ends a lot of code-gen. However such optimization is not
1155 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1156 * hence hasn't used verifier's analysis result. But, we really want to have a
1157 * way to be able to verify the correctness of the described optimization on
1158 * x86_64 on which testsuites are frequently exercised.
1160 * So, this flag is introduced. Once it is set, verifier will randomize high
1161 * 32-bit for those instructions who has been identified as safe to ignore them.
1162 * Then, if verifier is not doing correct analysis, such randomization will
1163 * regress tests to expose bugs.
1165 #define BPF_F_TEST_RND_HI32 (1U << 2)
1167 /* The verifier internal test flag. Behavior is undefined */
1168 #define BPF_F_TEST_STATE_FREQ (1U << 3)
1170 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1171 * restrict map and helper usage for such programs. Sleepable BPF programs can
1172 * only be attached to hooks where kernel execution context allows sleeping.
1173 * Such programs are allowed to use helpers that may sleep like
1174 * bpf_copy_from_user().
1176 #define BPF_F_SLEEPABLE (1U << 4)
1178 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1179 * fully support xdp frags.
1181 #define BPF_F_XDP_HAS_FRAGS (1U << 5)
1183 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1184 * program becomes device-bound but can access XDP metadata.
1186 #define BPF_F_XDP_DEV_BOUND_ONLY (1U << 6)
1188 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1189 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1192 BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
1195 /* link_create.uprobe_multi.flags used in LINK_CREATE command for
1196 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1199 BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
1202 /* link_create.netfilter.flags used in LINK_CREATE command for
1203 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1205 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1207 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1208 * the following extensions:
1210 * insn[0].src_reg: BPF_PSEUDO_MAP_[FD|IDX]
1211 * insn[0].imm: map fd or fd_idx
1215 * ldimm64 rewrite: address of map
1216 * verifier type: CONST_PTR_TO_MAP
1218 #define BPF_PSEUDO_MAP_FD 1
1219 #define BPF_PSEUDO_MAP_IDX 5
1221 /* insn[0].src_reg: BPF_PSEUDO_MAP_[IDX_]VALUE
1222 * insn[0].imm: map fd or fd_idx
1223 * insn[1].imm: offset into value
1226 * ldimm64 rewrite: address of map[0]+offset
1227 * verifier type: PTR_TO_MAP_VALUE
1229 #define BPF_PSEUDO_MAP_VALUE 2
1230 #define BPF_PSEUDO_MAP_IDX_VALUE 6
1232 /* insn[0].src_reg: BPF_PSEUDO_BTF_ID
1233 * insn[0].imm: kernel btd id of VAR
1237 * ldimm64 rewrite: address of the kernel variable
1238 * verifier type: PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1241 #define BPF_PSEUDO_BTF_ID 3
1242 /* insn[0].src_reg: BPF_PSEUDO_FUNC
1243 * insn[0].imm: insn offset to the func
1247 * ldimm64 rewrite: address of the function
1248 * verifier type: PTR_TO_FUNC.
1250 #define BPF_PSEUDO_FUNC 4
1252 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1253 * offset to another bpf function
1255 #define BPF_PSEUDO_CALL 1
1256 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1257 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1259 #define BPF_PSEUDO_KFUNC_CALL 2
1261 /* flags for BPF_MAP_UPDATE_ELEM command */
1263 BPF_ANY = 0, /* create new element or update existing */
1264 BPF_NOEXIST = 1, /* create new element if it didn't exist */
1265 BPF_EXIST = 2, /* update existing element */
1266 BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
1269 /* flags for BPF_MAP_CREATE command */
1271 BPF_F_NO_PREALLOC = (1U << 0),
1272 /* Instead of having one common LRU list in the
1273 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1274 * which can scale and perform better.
1275 * Note, the LRU nodes (including free nodes) cannot be moved
1276 * across different LRU lists.
1278 BPF_F_NO_COMMON_LRU = (1U << 1),
1279 /* Specify numa node during map creation */
1280 BPF_F_NUMA_NODE = (1U << 2),
1282 /* Flags for accessing BPF object from syscall side. */
1283 BPF_F_RDONLY = (1U << 3),
1284 BPF_F_WRONLY = (1U << 4),
1286 /* Flag for stack_map, store build_id+offset instead of pointer */
1287 BPF_F_STACK_BUILD_ID = (1U << 5),
1289 /* Zero-initialize hash function seed. This should only be used for testing. */
1290 BPF_F_ZERO_SEED = (1U << 6),
1292 /* Flags for accessing BPF object from program side. */
1293 BPF_F_RDONLY_PROG = (1U << 7),
1294 BPF_F_WRONLY_PROG = (1U << 8),
1296 /* Clone map from listener for newly accepted socket */
1297 BPF_F_CLONE = (1U << 9),
1299 /* Enable memory-mapping BPF map */
1300 BPF_F_MMAPABLE = (1U << 10),
1302 /* Share perf_event among processes */
1303 BPF_F_PRESERVE_ELEMS = (1U << 11),
1305 /* Create a map that is suitable to be an inner map with dynamic max entries */
1306 BPF_F_INNER_MAP = (1U << 12),
1308 /* Create a map that will be registered/unregesitered by the backed bpf_link */
1309 BPF_F_LINK = (1U << 13),
1311 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1312 BPF_F_PATH_FD = (1U << 14),
1315 /* Flags for BPF_PROG_QUERY. */
1317 /* Query effective (directly attached + inherited from ancestor cgroups)
1318 * programs that will be executed for events within a cgroup.
1319 * attach_flags with this flag are always returned 0.
1321 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
1323 /* Flags for BPF_PROG_TEST_RUN */
1325 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1326 #define BPF_F_TEST_RUN_ON_CPU (1U << 0)
1327 /* If set, XDP frames will be transmitted after processing */
1328 #define BPF_F_TEST_XDP_LIVE_FRAMES (1U << 1)
1330 /* type for BPF_ENABLE_STATS */
1331 enum bpf_stats_type {
1332 /* enabled run_time_ns and run_cnt */
1333 BPF_STATS_RUN_TIME = 0,
1336 enum bpf_stack_build_id_status {
1337 /* user space need an empty entry to identify end of a trace */
1338 BPF_STACK_BUILD_ID_EMPTY = 0,
1339 /* with valid build_id and offset */
1340 BPF_STACK_BUILD_ID_VALID = 1,
1341 /* couldn't get build_id, fallback to ip */
1342 BPF_STACK_BUILD_ID_IP = 2,
1345 #define BPF_BUILD_ID_SIZE 20
1346 struct bpf_stack_build_id {
1348 unsigned char build_id[BPF_BUILD_ID_SIZE];
1355 #define BPF_OBJ_NAME_LEN 16U
1358 struct { /* anonymous struct used by BPF_MAP_CREATE command */
1359 __u32 map_type; /* one of enum bpf_map_type */
1360 __u32 key_size; /* size of key in bytes */
1361 __u32 value_size; /* size of value in bytes */
1362 __u32 max_entries; /* max number of entries in a map */
1363 __u32 map_flags; /* BPF_MAP_CREATE related
1364 * flags defined above.
1366 __u32 inner_map_fd; /* fd pointing to the inner map */
1367 __u32 numa_node; /* numa node (effective only if
1368 * BPF_F_NUMA_NODE is set).
1370 char map_name[BPF_OBJ_NAME_LEN];
1371 __u32 map_ifindex; /* ifindex of netdev to create on */
1372 __u32 btf_fd; /* fd pointing to a BTF type data */
1373 __u32 btf_key_type_id; /* BTF type_id of the key */
1374 __u32 btf_value_type_id; /* BTF type_id of the value */
1375 __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1376 * struct stored as the
1379 /* Any per-map-type extra fields
1381 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1382 * number of hash functions (if 0, the bloom filter will default
1383 * to using 5 hash functions).
1388 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1392 __aligned_u64 value;
1393 __aligned_u64 next_key;
1398 struct { /* struct used by BPF_MAP_*_BATCH commands */
1399 __aligned_u64 in_batch; /* start batch,
1400 * NULL to start from beginning
1402 __aligned_u64 out_batch; /* output: next start batch */
1404 __aligned_u64 values;
1405 __u32 count; /* input/output:
1406 * input: # of key/value
1408 * output: # of filled elements
1415 struct { /* anonymous struct used by BPF_PROG_LOAD command */
1416 __u32 prog_type; /* one of enum bpf_prog_type */
1418 __aligned_u64 insns;
1419 __aligned_u64 license;
1420 __u32 log_level; /* verbosity level of verifier */
1421 __u32 log_size; /* size of user buffer */
1422 __aligned_u64 log_buf; /* user supplied buffer */
1423 __u32 kern_version; /* not used */
1425 char prog_name[BPF_OBJ_NAME_LEN];
1426 __u32 prog_ifindex; /* ifindex of netdev to prep for */
1427 /* For some prog types expected attach type must be known at
1428 * load time to verify attach type specific parts of prog
1429 * (context accesses, allowed helpers, etc).
1431 __u32 expected_attach_type;
1432 __u32 prog_btf_fd; /* fd pointing to BTF type data */
1433 __u32 func_info_rec_size; /* userspace bpf_func_info size */
1434 __aligned_u64 func_info; /* func info */
1435 __u32 func_info_cnt; /* number of bpf_func_info records */
1436 __u32 line_info_rec_size; /* userspace bpf_line_info size */
1437 __aligned_u64 line_info; /* line info */
1438 __u32 line_info_cnt; /* number of bpf_line_info records */
1439 __u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1441 /* valid prog_fd to attach to bpf prog */
1442 __u32 attach_prog_fd;
1443 /* or valid module BTF object fd or 0 to attach to vmlinux */
1444 __u32 attach_btf_obj_fd;
1446 __u32 core_relo_cnt; /* number of bpf_core_relo */
1447 __aligned_u64 fd_array; /* array of FDs */
1448 __aligned_u64 core_relos;
1449 __u32 core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1450 /* output: actual total log contents size (including termintaing zero).
1451 * It could be both larger than original log_size (if log was
1452 * truncated), or smaller (if log buffer wasn't filled completely).
1454 __u32 log_true_size;
1457 struct { /* anonymous struct used by BPF_OBJ_* commands */
1458 __aligned_u64 pathname;
1461 /* Same as dirfd in openat() syscall; see openat(2)
1462 * manpage for details of path FD and pathname semantics;
1463 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1464 * file_flags field, otherwise it should be set to zero;
1465 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1470 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1472 __u32 target_fd; /* target object to attach to or ... */
1473 __u32 target_ifindex; /* target ifindex */
1475 __u32 attach_bpf_fd;
1478 __u32 replace_bpf_fd;
1483 __u64 expected_revision;
1486 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1489 __u32 data_size_in; /* input: len of data_in */
1490 __u32 data_size_out; /* input/output: len of data_out
1491 * returns ENOSPC if data_out
1494 __aligned_u64 data_in;
1495 __aligned_u64 data_out;
1498 __u32 ctx_size_in; /* input: len of ctx_in */
1499 __u32 ctx_size_out; /* input/output: len of ctx_out
1500 * returns ENOSPC if ctx_out
1503 __aligned_u64 ctx_in;
1504 __aligned_u64 ctx_out;
1510 struct { /* anonymous struct used by BPF_*_GET_*_ID */
1522 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1528 struct { /* anonymous struct used by BPF_PROG_QUERY command */
1530 __u32 target_fd; /* target object to query or ... */
1531 __u32 target_ifindex; /* target ifindex */
1536 __aligned_u64 prog_ids;
1542 /* output: per-program attach_flags.
1543 * not allowed to be set during effective query.
1545 __aligned_u64 prog_attach_flags;
1546 __aligned_u64 link_ids;
1547 __aligned_u64 link_attach_flags;
1551 struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1556 struct { /* anonymous struct for BPF_BTF_LOAD */
1558 __aligned_u64 btf_log_buf;
1561 __u32 btf_log_level;
1562 /* output: actual total log contents size (including termintaing zero).
1563 * It could be both larger than original log_size (if log was
1564 * truncated), or smaller (if log buffer wasn't filled completely).
1566 __u32 btf_log_true_size;
1570 __u32 pid; /* input: pid */
1571 __u32 fd; /* input: fd */
1572 __u32 flags; /* input: flags */
1573 __u32 buf_len; /* input/output: buf len */
1574 __aligned_u64 buf; /* input/output:
1575 * tp_name for tracepoint
1577 * filename for uprobe
1579 __u32 prog_id; /* output: prod_id */
1580 __u32 fd_type; /* output: BPF_FD_TYPE_* */
1581 __u64 probe_offset; /* output: probe_offset */
1582 __u64 probe_addr; /* output: probe_addr */
1585 struct { /* struct used by BPF_LINK_CREATE command */
1587 __u32 prog_fd; /* eBPF program to attach */
1588 __u32 map_fd; /* struct_ops to attach */
1591 __u32 target_fd; /* target object to attach to or ... */
1592 __u32 target_ifindex; /* target ifindex */
1594 __u32 attach_type; /* attach type */
1595 __u32 flags; /* extra flags */
1597 __u32 target_btf_id; /* btf_id of target to attach to */
1599 __aligned_u64 iter_info; /* extra bpf_iter_link_info */
1600 __u32 iter_info_len; /* iter_info length */
1603 /* black box user-provided value passed through
1604 * to BPF program at the execution time and
1605 * accessible through bpf_get_attach_cookie() BPF helper
1613 __aligned_u64 addrs;
1614 __aligned_u64 cookies;
1617 /* this is overlaid with the target_btf_id above. */
1618 __u32 target_btf_id;
1619 /* black box user-provided value passed through
1620 * to BPF program at the execution time and
1621 * accessible through bpf_get_attach_cookie() BPF helper
1636 __u64 expected_revision;
1640 __aligned_u64 offsets;
1641 __aligned_u64 ref_ctr_offsets;
1642 __aligned_u64 cookies;
1650 struct { /* struct used by BPF_LINK_UPDATE command */
1651 __u32 link_fd; /* link fd */
1653 /* new program fd to update link with */
1655 /* new struct_ops map fd to update link with */
1658 __u32 flags; /* extra flags */
1660 /* expected link's program fd; is specified only if
1661 * BPF_F_REPLACE flag is set in flags.
1664 /* expected link's map fd; is specified only
1665 * if BPF_F_REPLACE flag is set.
1675 struct { /* struct used by BPF_ENABLE_STATS command */
1679 struct { /* struct used by BPF_ITER_CREATE command */
1684 struct { /* struct used by BPF_PROG_BIND_MAP command */
1687 __u32 flags; /* extra flags */
1690 } __attribute__((aligned(8)));
1692 /* The description below is an attempt at providing documentation to eBPF
1693 * developers about the multiple available eBPF helper functions. It can be
1694 * parsed and used to produce a manual page. The workflow is the following,
1695 * and requires the rst2man utility:
1697 * $ ./scripts/bpf_doc.py \
1698 * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1699 * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1700 * $ man /tmp/bpf-helpers.7
1702 * Note that in order to produce this external documentation, some RST
1703 * formatting is used in the descriptions to get "bold" and "italics" in
1704 * manual pages. Also note that the few trailing white spaces are
1705 * intentional, removing them would break paragraphs for rst2man.
1707 * Start of BPF helper function descriptions:
1709 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1711 * Perform a lookup in *map* for an entry associated to *key*.
1713 * Map value associated to *key*, or **NULL** if no entry was
1716 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1718 * Add or update the value of the entry associated to *key* in
1719 * *map* with *value*. *flags* is one of:
1722 * The entry for *key* must not exist in the map.
1724 * The entry for *key* must already exist in the map.
1726 * No condition on the existence of the entry for *key*.
1728 * Flag value **BPF_NOEXIST** cannot be used for maps of types
1729 * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
1730 * elements always exist), the helper would return an error.
1732 * 0 on success, or a negative error in case of failure.
1734 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1736 * Delete entry with *key* from *map*.
1738 * 0 on success, or a negative error in case of failure.
1740 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1742 * For tracing programs, safely attempt to read *size* bytes from
1743 * kernel space address *unsafe_ptr* and store the data in *dst*.
1745 * Generally, use **bpf_probe_read_user**\ () or
1746 * **bpf_probe_read_kernel**\ () instead.
1748 * 0 on success, or a negative error in case of failure.
1750 * u64 bpf_ktime_get_ns(void)
1752 * Return the time elapsed since system boot, in nanoseconds.
1753 * Does not include time the system was suspended.
1754 * See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1758 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1760 * This helper is a "printk()-like" facility for debugging. It
1761 * prints a message defined by format *fmt* (of size *fmt_size*)
1762 * to file *\/sys/kernel/tracing/trace* from TraceFS, if
1763 * available. It can take up to three additional **u64**
1764 * arguments (as an eBPF helpers, the total number of arguments is
1767 * Each time the helper is called, it appends a line to the trace.
1768 * Lines are discarded while *\/sys/kernel/tracing/trace* is
1769 * open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
1770 * The format of the trace is customizable, and the exact output
1771 * one will get depends on the options set in
1772 * *\/sys/kernel/tracing/trace_options* (see also the
1773 * *README* file under the same directory). However, it usually
1774 * defaults to something like:
1778 * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1782 * * ``telnet`` is the name of the current task.
1783 * * ``470`` is the PID of the current task.
1784 * * ``001`` is the CPU number on which the task is
1786 * * In ``.N..``, each character refers to a set of
1787 * options (whether irqs are enabled, scheduling
1788 * options, whether hard/softirqs are running, level of
1789 * preempt_disabled respectively). **N** means that
1790 * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1792 * * ``419421.045894`` is a timestamp.
1793 * * ``0x00000001`` is a fake value used by BPF for the
1794 * instruction pointer register.
1795 * * ``<formatted msg>`` is the message formatted with
1798 * The conversion specifiers supported by *fmt* are similar, but
1799 * more limited than for printk(). They are **%d**, **%i**,
1800 * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1801 * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1802 * of field, padding with zeroes, etc.) is available, and the
1803 * helper will return **-EINVAL** (but print nothing) if it
1804 * encounters an unknown specifier.
1806 * Also, note that **bpf_trace_printk**\ () is slow, and should
1807 * only be used for debugging purposes. For this reason, a notice
1808 * block (spanning several lines) is printed to kernel logs and
1809 * states that the helper should not be used "for production use"
1810 * the first time this helper is used (or more precisely, when
1811 * **trace_printk**\ () buffers are allocated). For passing values
1812 * to user space, perf events should be preferred.
1814 * The number of bytes written to the buffer, or a negative error
1815 * in case of failure.
1817 * u32 bpf_get_prandom_u32(void)
1819 * Get a pseudo-random number.
1821 * From a security point of view, this helper uses its own
1822 * pseudo-random internal state, and cannot be used to infer the
1823 * seed of other random functions in the kernel. However, it is
1824 * essential to note that the generator used by the helper is not
1825 * cryptographically secure.
1827 * A random 32-bit unsigned value.
1829 * u32 bpf_get_smp_processor_id(void)
1831 * Get the SMP (symmetric multiprocessing) processor id. Note that
1832 * all programs run with migration disabled, which means that the
1833 * SMP processor id is stable during all the execution of the
1836 * The SMP id of the processor running the program.
1838 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1840 * Store *len* bytes from address *from* into the packet
1841 * associated to *skb*, at *offset*. *flags* are a combination of
1842 * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1843 * checksum for the packet after storing the bytes) and
1844 * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1845 * **->swhash** and *skb*\ **->l4hash** to 0).
1847 * A call to this helper is susceptible to change the underlying
1848 * packet buffer. Therefore, at load time, all checks on pointers
1849 * previously done by the verifier are invalidated and must be
1850 * performed again, if the helper is used in combination with
1851 * direct packet access.
1853 * 0 on success, or a negative error in case of failure.
1855 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1857 * Recompute the layer 3 (e.g. IP) checksum for the packet
1858 * associated to *skb*. Computation is incremental, so the helper
1859 * must know the former value of the header field that was
1860 * modified (*from*), the new value of this field (*to*), and the
1861 * number of bytes (2 or 4) for this field, stored in *size*.
1862 * Alternatively, it is possible to store the difference between
1863 * the previous and the new values of the header field in *to*, by
1864 * setting *from* and *size* to 0. For both methods, *offset*
1865 * indicates the location of the IP checksum within the packet.
1867 * This helper works in combination with **bpf_csum_diff**\ (),
1868 * which does not update the checksum in-place, but offers more
1869 * flexibility and can handle sizes larger than 2 or 4 for the
1870 * checksum to update.
1872 * A call to this helper is susceptible to change the underlying
1873 * packet buffer. Therefore, at load time, all checks on pointers
1874 * previously done by the verifier are invalidated and must be
1875 * performed again, if the helper is used in combination with
1876 * direct packet access.
1878 * 0 on success, or a negative error in case of failure.
1880 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1882 * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1883 * packet associated to *skb*. Computation is incremental, so the
1884 * helper must know the former value of the header field that was
1885 * modified (*from*), the new value of this field (*to*), and the
1886 * number of bytes (2 or 4) for this field, stored on the lowest
1887 * four bits of *flags*. Alternatively, it is possible to store
1888 * the difference between the previous and the new values of the
1889 * header field in *to*, by setting *from* and the four lowest
1890 * bits of *flags* to 0. For both methods, *offset* indicates the
1891 * location of the IP checksum within the packet. In addition to
1892 * the size of the field, *flags* can be added (bitwise OR) actual
1893 * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1894 * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1895 * for updates resulting in a null checksum the value is set to
1896 * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1897 * the checksum is to be computed against a pseudo-header.
1899 * This helper works in combination with **bpf_csum_diff**\ (),
1900 * which does not update the checksum in-place, but offers more
1901 * flexibility and can handle sizes larger than 2 or 4 for the
1902 * checksum to update.
1904 * A call to this helper is susceptible to change the underlying
1905 * packet buffer. Therefore, at load time, all checks on pointers
1906 * previously done by the verifier are invalidated and must be
1907 * performed again, if the helper is used in combination with
1908 * direct packet access.
1910 * 0 on success, or a negative error in case of failure.
1912 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1914 * This special helper is used to trigger a "tail call", or in
1915 * other words, to jump into another eBPF program. The same stack
1916 * frame is used (but values on stack and in registers for the
1917 * caller are not accessible to the callee). This mechanism allows
1918 * for program chaining, either for raising the maximum number of
1919 * available eBPF instructions, or to execute given programs in
1920 * conditional blocks. For security reasons, there is an upper
1921 * limit to the number of successive tail calls that can be
1924 * Upon call of this helper, the program attempts to jump into a
1925 * program referenced at index *index* in *prog_array_map*, a
1926 * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1927 * *ctx*, a pointer to the context.
1929 * If the call succeeds, the kernel immediately runs the first
1930 * instruction of the new program. This is not a function call,
1931 * and it never returns to the previous program. If the call
1932 * fails, then the helper has no effect, and the caller continues
1933 * to run its subsequent instructions. A call can fail if the
1934 * destination program for the jump does not exist (i.e. *index*
1935 * is superior to the number of entries in *prog_array_map*), or
1936 * if the maximum number of tail calls has been reached for this
1937 * chain of programs. This limit is defined in the kernel by the
1938 * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1939 * which is currently set to 33.
1941 * 0 on success, or a negative error in case of failure.
1943 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1945 * Clone and redirect the packet associated to *skb* to another
1946 * net device of index *ifindex*. Both ingress and egress
1947 * interfaces can be used for redirection. The **BPF_F_INGRESS**
1948 * value in *flags* is used to make the distinction (ingress path
1949 * is selected if the flag is present, egress path otherwise).
1950 * This is the only flag supported for now.
1952 * In comparison with **bpf_redirect**\ () helper,
1953 * **bpf_clone_redirect**\ () has the associated cost of
1954 * duplicating the packet buffer, but this can be executed out of
1955 * the eBPF program. Conversely, **bpf_redirect**\ () is more
1956 * efficient, but it is handled through an action code where the
1957 * redirection happens only after the eBPF program has returned.
1959 * A call to this helper is susceptible to change the underlying
1960 * packet buffer. Therefore, at load time, all checks on pointers
1961 * previously done by the verifier are invalidated and must be
1962 * performed again, if the helper is used in combination with
1963 * direct packet access.
1965 * 0 on success, or a negative error in case of failure. Positive
1966 * error indicates a potential drop or congestion in the target
1967 * device. The particular positive error codes are not defined.
1969 * u64 bpf_get_current_pid_tgid(void)
1971 * Get the current pid and tgid.
1973 * A 64-bit integer containing the current tgid and pid, and
1975 * *current_task*\ **->tgid << 32 \|**
1976 * *current_task*\ **->pid**.
1978 * u64 bpf_get_current_uid_gid(void)
1980 * Get the current uid and gid.
1982 * A 64-bit integer containing the current GID and UID, and
1983 * created as such: *current_gid* **<< 32 \|** *current_uid*.
1985 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1987 * Copy the **comm** attribute of the current task into *buf* of
1988 * *size_of_buf*. The **comm** attribute contains the name of
1989 * the executable (excluding the path) for the current task. The
1990 * *size_of_buf* must be strictly positive. On success, the
1991 * helper makes sure that the *buf* is NUL-terminated. On failure,
1992 * it is filled with zeroes.
1994 * 0 on success, or a negative error in case of failure.
1996 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1998 * Retrieve the classid for the current task, i.e. for the net_cls
1999 * cgroup to which *skb* belongs.
2001 * This helper can be used on TC egress path, but not on ingress.
2003 * The net_cls cgroup provides an interface to tag network packets
2004 * based on a user-provided identifier for all traffic coming from
2005 * the tasks belonging to the related cgroup. See also the related
2006 * kernel documentation, available from the Linux sources in file
2007 * *Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2009 * The Linux kernel has two versions for cgroups: there are
2010 * cgroups v1 and cgroups v2. Both are available to users, who can
2011 * use a mixture of them, but note that the net_cls cgroup is for
2012 * cgroup v1 only. This makes it incompatible with BPF programs
2013 * run on cgroups, which is a cgroup-v2-only feature (a socket can
2014 * only hold data for one version of cgroups at a time).
2016 * This helper is only available is the kernel was compiled with
2017 * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2018 * "**y**" or to "**m**".
2020 * The classid, or 0 for the default unconfigured classid.
2022 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2024 * Push a *vlan_tci* (VLAN tag control information) of protocol
2025 * *vlan_proto* to the packet associated to *skb*, then update
2026 * the checksum. Note that if *vlan_proto* is different from
2027 * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2028 * be **ETH_P_8021Q**.
2030 * A call to this helper is susceptible to change the underlying
2031 * packet buffer. Therefore, at load time, all checks on pointers
2032 * previously done by the verifier are invalidated and must be
2033 * performed again, if the helper is used in combination with
2034 * direct packet access.
2036 * 0 on success, or a negative error in case of failure.
2038 * long bpf_skb_vlan_pop(struct sk_buff *skb)
2040 * Pop a VLAN header from the packet associated to *skb*.
2042 * A call to this helper is susceptible to change the underlying
2043 * packet buffer. Therefore, at load time, all checks on pointers
2044 * previously done by the verifier are invalidated and must be
2045 * performed again, if the helper is used in combination with
2046 * direct packet access.
2048 * 0 on success, or a negative error in case of failure.
2050 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2052 * Get tunnel metadata. This helper takes a pointer *key* to an
2053 * empty **struct bpf_tunnel_key** of **size**, that will be
2054 * filled with tunnel metadata for the packet associated to *skb*.
2055 * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2056 * indicates that the tunnel is based on IPv6 protocol instead of
2059 * The **struct bpf_tunnel_key** is an object that generalizes the
2060 * principal parameters used by various tunneling protocols into a
2061 * single struct. This way, it can be used to easily make a
2062 * decision based on the contents of the encapsulation header,
2063 * "summarized" in this struct. In particular, it holds the IP
2064 * address of the remote end (IPv4 or IPv6, depending on the case)
2065 * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2066 * this struct exposes the *key*\ **->tunnel_id**, which is
2067 * generally mapped to a VNI (Virtual Network Identifier), making
2068 * it programmable together with the **bpf_skb_set_tunnel_key**\
2071 * Let's imagine that the following code is part of a program
2072 * attached to the TC ingress interface, on one end of a GRE
2073 * tunnel, and is supposed to filter out all messages coming from
2074 * remote ends with IPv4 address other than 10.0.0.1:
2079 * struct bpf_tunnel_key key = {};
2081 * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2083 * return TC_ACT_SHOT; // drop packet
2085 * if (key.remote_ipv4 != 0x0a000001)
2086 * return TC_ACT_SHOT; // drop packet
2088 * return TC_ACT_OK; // accept packet
2090 * This interface can also be used with all encapsulation devices
2091 * that can operate in "collect metadata" mode: instead of having
2092 * one network device per specific configuration, the "collect
2093 * metadata" mode only requires a single device where the
2094 * configuration can be extracted from this helper.
2096 * This can be used together with various tunnels such as VXLan,
2097 * Geneve, GRE or IP in IP (IPIP).
2099 * 0 on success, or a negative error in case of failure.
2101 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2103 * Populate tunnel metadata for packet associated to *skb.* The
2104 * tunnel metadata is set to the contents of *key*, of *size*. The
2105 * *flags* can be set to a combination of the following values:
2107 * **BPF_F_TUNINFO_IPV6**
2108 * Indicate that the tunnel is based on IPv6 protocol
2110 * **BPF_F_ZERO_CSUM_TX**
2111 * For IPv4 packets, add a flag to tunnel metadata
2112 * indicating that checksum computation should be skipped
2113 * and checksum set to zeroes.
2114 * **BPF_F_DONT_FRAGMENT**
2115 * Add a flag to tunnel metadata indicating that the
2116 * packet should not be fragmented.
2117 * **BPF_F_SEQ_NUMBER**
2118 * Add a flag to tunnel metadata indicating that a
2119 * sequence number should be added to tunnel header before
2120 * sending the packet. This flag was added for GRE
2121 * encapsulation, but might be used with other protocols
2122 * as well in the future.
2123 * **BPF_F_NO_TUNNEL_KEY**
2124 * Add a flag to tunnel metadata indicating that no tunnel
2125 * key should be set in the resulting tunnel header.
2127 * Here is a typical usage on the transmit path:
2131 * struct bpf_tunnel_key key;
2133 * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2134 * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2136 * See also the description of the **bpf_skb_get_tunnel_key**\ ()
2137 * helper for additional information.
2139 * 0 on success, or a negative error in case of failure.
2141 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2143 * Read the value of a perf event counter. This helper relies on a
2144 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2145 * the perf event counter is selected when *map* is updated with
2146 * perf event file descriptors. The *map* is an array whose size
2147 * is the number of available CPUs, and each cell contains a value
2148 * relative to one CPU. The value to retrieve is indicated by
2149 * *flags*, that contains the index of the CPU to look up, masked
2150 * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2151 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2152 * current CPU should be retrieved.
2154 * Note that before Linux 4.13, only hardware perf event can be
2157 * Also, be aware that the newer helper
2158 * **bpf_perf_event_read_value**\ () is recommended over
2159 * **bpf_perf_event_read**\ () in general. The latter has some ABI
2160 * quirks where error and counter value are used as a return code
2161 * (which is wrong to do since ranges may overlap). This issue is
2162 * fixed with **bpf_perf_event_read_value**\ (), which at the same
2163 * time provides more features over the **bpf_perf_event_read**\
2164 * () interface. Please refer to the description of
2165 * **bpf_perf_event_read_value**\ () for details.
2167 * The value of the perf event counter read from the map, or a
2168 * negative error code in case of failure.
2170 * long bpf_redirect(u32 ifindex, u64 flags)
2172 * Redirect the packet to another net device of index *ifindex*.
2173 * This helper is somewhat similar to **bpf_clone_redirect**\
2174 * (), except that the packet is not cloned, which provides
2175 * increased performance.
2177 * Except for XDP, both ingress and egress interfaces can be used
2178 * for redirection. The **BPF_F_INGRESS** value in *flags* is used
2179 * to make the distinction (ingress path is selected if the flag
2180 * is present, egress path otherwise). Currently, XDP only
2181 * supports redirection to the egress interface, and accepts no
2184 * The same effect can also be attained with the more generic
2185 * **bpf_redirect_map**\ (), which uses a BPF map to store the
2186 * redirect target instead of providing it directly to the helper.
2188 * For XDP, the helper returns **XDP_REDIRECT** on success or
2189 * **XDP_ABORTED** on error. For other program types, the values
2190 * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2193 * u32 bpf_get_route_realm(struct sk_buff *skb)
2195 * Retrieve the realm or the route, that is to say the
2196 * **tclassid** field of the destination for the *skb*. The
2197 * identifier retrieved is a user-provided tag, similar to the
2198 * one used with the net_cls cgroup (see description for
2199 * **bpf_get_cgroup_classid**\ () helper), but here this tag is
2200 * held by a route (a destination entry), not by a task.
2202 * Retrieving this identifier works with the clsact TC egress hook
2203 * (see also **tc-bpf(8)**), or alternatively on conventional
2204 * classful egress qdiscs, but not on TC ingress path. In case of
2205 * clsact TC egress hook, this has the advantage that, internally,
2206 * the destination entry has not been dropped yet in the transmit
2207 * path. Therefore, the destination entry does not need to be
2208 * artificially held via **netif_keep_dst**\ () for a classful
2209 * qdisc until the *skb* is freed.
2211 * This helper is available only if the kernel was compiled with
2212 * **CONFIG_IP_ROUTE_CLASSID** configuration option.
2214 * The realm of the route for the packet associated to *skb*, or 0
2215 * if none was found.
2217 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2219 * Write raw *data* blob into a special BPF perf event held by
2220 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2221 * event must have the following attributes: **PERF_SAMPLE_RAW**
2222 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2223 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2225 * The *flags* are used to indicate the index in *map* for which
2226 * the value must be put, masked with **BPF_F_INDEX_MASK**.
2227 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2228 * to indicate that the index of the current CPU core should be
2231 * The value to write, of *size*, is passed through eBPF stack and
2232 * pointed by *data*.
2234 * The context of the program *ctx* needs also be passed to the
2237 * On user space, a program willing to read the values needs to
2238 * call **perf_event_open**\ () on the perf event (either for
2239 * one or for all CPUs) and to store the file descriptor into the
2240 * *map*. This must be done before the eBPF program can send data
2241 * into it. An example is available in file
2242 * *samples/bpf/trace_output_user.c* in the Linux kernel source
2243 * tree (the eBPF program counterpart is in
2244 * *samples/bpf/trace_output_kern.c*).
2246 * **bpf_perf_event_output**\ () achieves better performance
2247 * than **bpf_trace_printk**\ () for sharing data with user
2248 * space, and is much better suitable for streaming data from eBPF
2251 * Note that this helper is not restricted to tracing use cases
2252 * and can be used with programs attached to TC or XDP as well,
2253 * where it allows for passing data to user space listeners. Data
2256 * * Only custom structs,
2257 * * Only the packet payload, or
2258 * * A combination of both.
2260 * 0 on success, or a negative error in case of failure.
2262 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2264 * This helper was provided as an easy way to load data from a
2265 * packet. It can be used to load *len* bytes from *offset* from
2266 * the packet associated to *skb*, into the buffer pointed by
2269 * Since Linux 4.7, usage of this helper has mostly been replaced
2270 * by "direct packet access", enabling packet data to be
2271 * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2272 * pointing respectively to the first byte of packet data and to
2273 * the byte after the last byte of packet data. However, it
2274 * remains useful if one wishes to read large quantities of data
2275 * at once from a packet into the eBPF stack.
2277 * 0 on success, or a negative error in case of failure.
2279 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2281 * Walk a user or a kernel stack and return its id. To achieve
2282 * this, the helper needs *ctx*, which is a pointer to the context
2283 * on which the tracing program is executed, and a pointer to a
2284 * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2286 * The last argument, *flags*, holds the number of stack frames to
2287 * skip (from 0 to 255), masked with
2288 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2289 * a combination of the following flags:
2291 * **BPF_F_USER_STACK**
2292 * Collect a user space stack instead of a kernel stack.
2293 * **BPF_F_FAST_STACK_CMP**
2294 * Compare stacks by hash only.
2295 * **BPF_F_REUSE_STACKID**
2296 * If two different stacks hash into the same *stackid*,
2297 * discard the old one.
2299 * The stack id retrieved is a 32 bit long integer handle which
2300 * can be further combined with other data (including other stack
2301 * ids) and used as a key into maps. This can be useful for
2302 * generating a variety of graphs (such as flame graphs or off-cpu
2305 * For walking a stack, this helper is an improvement over
2306 * **bpf_probe_read**\ (), which can be used with unrolled loops
2307 * but is not efficient and consumes a lot of eBPF instructions.
2308 * Instead, **bpf_get_stackid**\ () can collect up to
2309 * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2310 * this limit can be controlled with the **sysctl** program, and
2311 * that it should be manually increased in order to profile long
2312 * user stacks (such as stacks for Java programs). To do so, use:
2316 * # sysctl kernel.perf_event_max_stack=<new value>
2318 * The positive or null stack id on success, or a negative error
2319 * in case of failure.
2321 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2323 * Compute a checksum difference, from the raw buffer pointed by
2324 * *from*, of length *from_size* (that must be a multiple of 4),
2325 * towards the raw buffer pointed by *to*, of size *to_size*
2326 * (same remark). An optional *seed* can be added to the value
2327 * (this can be cascaded, the seed may come from a previous call
2330 * This is flexible enough to be used in several ways:
2332 * * With *from_size* == 0, *to_size* > 0 and *seed* set to
2333 * checksum, it can be used when pushing new data.
2334 * * With *from_size* > 0, *to_size* == 0 and *seed* set to
2335 * checksum, it can be used when removing data from a packet.
2336 * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2337 * can be used to compute a diff. Note that *from_size* and
2338 * *to_size* do not need to be equal.
2340 * This helper can be used in combination with
2341 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2342 * which one can feed in the difference computed with
2343 * **bpf_csum_diff**\ ().
2345 * The checksum result, or a negative error code in case of
2348 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2350 * Retrieve tunnel options metadata for the packet associated to
2351 * *skb*, and store the raw tunnel option data to the buffer *opt*
2354 * This helper can be used with encapsulation devices that can
2355 * operate in "collect metadata" mode (please refer to the related
2356 * note in the description of **bpf_skb_get_tunnel_key**\ () for
2357 * more details). A particular example where this can be used is
2358 * in combination with the Geneve encapsulation protocol, where it
2359 * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2360 * and retrieving arbitrary TLVs (Type-Length-Value headers) from
2361 * the eBPF program. This allows for full customization of these
2364 * The size of the option data retrieved.
2366 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2368 * Set tunnel options metadata for the packet associated to *skb*
2369 * to the option data contained in the raw buffer *opt* of *size*.
2371 * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2372 * helper for additional information.
2374 * 0 on success, or a negative error in case of failure.
2376 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2378 * Change the protocol of the *skb* to *proto*. Currently
2379 * supported are transition from IPv4 to IPv6, and from IPv6 to
2380 * IPv4. The helper takes care of the groundwork for the
2381 * transition, including resizing the socket buffer. The eBPF
2382 * program is expected to fill the new headers, if any, via
2383 * **skb_store_bytes**\ () and to recompute the checksums with
2384 * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2385 * (). The main case for this helper is to perform NAT64
2386 * operations out of an eBPF program.
2388 * Internally, the GSO type is marked as dodgy so that headers are
2389 * checked and segments are recalculated by the GSO/GRO engine.
2390 * The size for GSO target is adapted as well.
2392 * All values for *flags* are reserved for future usage, and must
2395 * A call to this helper is susceptible to change the underlying
2396 * packet buffer. Therefore, at load time, all checks on pointers
2397 * previously done by the verifier are invalidated and must be
2398 * performed again, if the helper is used in combination with
2399 * direct packet access.
2401 * 0 on success, or a negative error in case of failure.
2403 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2405 * Change the packet type for the packet associated to *skb*. This
2406 * comes down to setting *skb*\ **->pkt_type** to *type*, except
2407 * the eBPF program does not have a write access to *skb*\
2408 * **->pkt_type** beside this helper. Using a helper here allows
2409 * for graceful handling of errors.
2411 * The major use case is to change incoming *skb*s to
2412 * **PACKET_HOST** in a programmatic way instead of having to
2413 * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2416 * Note that *type* only allows certain values. At this time, they
2421 * **PACKET_BROADCAST**
2422 * Send packet to all.
2423 * **PACKET_MULTICAST**
2424 * Send packet to group.
2425 * **PACKET_OTHERHOST**
2426 * Send packet to someone else.
2428 * 0 on success, or a negative error in case of failure.
2430 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2432 * Check whether *skb* is a descendant of the cgroup2 held by
2433 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2435 * The return value depends on the result of the test, and can be:
2437 * * 0, if the *skb* failed the cgroup2 descendant test.
2438 * * 1, if the *skb* succeeded the cgroup2 descendant test.
2439 * * A negative error code, if an error occurred.
2441 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2443 * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2444 * not set, in particular if the hash was cleared due to mangling,
2445 * recompute this hash. Later accesses to the hash can be done
2446 * directly with *skb*\ **->hash**.
2448 * Calling **bpf_set_hash_invalid**\ (), changing a packet
2449 * prototype with **bpf_skb_change_proto**\ (), or calling
2450 * **bpf_skb_store_bytes**\ () with the
2451 * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2452 * the hash and to trigger a new computation for the next call to
2453 * **bpf_get_hash_recalc**\ ().
2457 * u64 bpf_get_current_task(void)
2459 * Get the current task.
2461 * A pointer to the current task struct.
2463 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2465 * Attempt in a safe way to write *len* bytes from the buffer
2466 * *src* to *dst* in memory. It only works for threads that are in
2467 * user context, and *dst* must be a valid user space address.
2469 * This helper should not be used to implement any kind of
2470 * security mechanism because of TOC-TOU attacks, but rather to
2471 * debug, divert, and manipulate execution of semi-cooperative
2474 * Keep in mind that this feature is meant for experiments, and it
2475 * has a risk of crashing the system and running programs.
2476 * Therefore, when an eBPF program using this helper is attached,
2477 * a warning including PID and process name is printed to kernel
2480 * 0 on success, or a negative error in case of failure.
2482 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2484 * Check whether the probe is being run is the context of a given
2485 * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2486 * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2488 * The return value depends on the result of the test, and can be:
2490 * * 1, if current task belongs to the cgroup2.
2491 * * 0, if current task does not belong to the cgroup2.
2492 * * A negative error code, if an error occurred.
2494 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2496 * Resize (trim or grow) the packet associated to *skb* to the
2497 * new *len*. The *flags* are reserved for future usage, and must
2500 * The basic idea is that the helper performs the needed work to
2501 * change the size of the packet, then the eBPF program rewrites
2502 * the rest via helpers like **bpf_skb_store_bytes**\ (),
2503 * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2504 * and others. This helper is a slow path utility intended for
2505 * replies with control messages. And because it is targeted for
2506 * slow path, the helper itself can afford to be slow: it
2507 * implicitly linearizes, unclones and drops offloads from the
2510 * A call to this helper is susceptible to change the underlying
2511 * packet buffer. Therefore, at load time, all checks on pointers
2512 * previously done by the verifier are invalidated and must be
2513 * performed again, if the helper is used in combination with
2514 * direct packet access.
2516 * 0 on success, or a negative error in case of failure.
2518 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2520 * Pull in non-linear data in case the *skb* is non-linear and not
2521 * all of *len* are part of the linear section. Make *len* bytes
2522 * from *skb* readable and writable. If a zero value is passed for
2523 * *len*, then all bytes in the linear part of *skb* will be made
2524 * readable and writable.
2526 * This helper is only needed for reading and writing with direct
2529 * For direct packet access, testing that offsets to access
2530 * are within packet boundaries (test on *skb*\ **->data_end**) is
2531 * susceptible to fail if offsets are invalid, or if the requested
2532 * data is in non-linear parts of the *skb*. On failure the
2533 * program can just bail out, or in the case of a non-linear
2534 * buffer, use a helper to make the data available. The
2535 * **bpf_skb_load_bytes**\ () helper is a first solution to access
2536 * the data. Another one consists in using **bpf_skb_pull_data**
2537 * to pull in once the non-linear parts, then retesting and
2538 * eventually access the data.
2540 * At the same time, this also makes sure the *skb* is uncloned,
2541 * which is a necessary condition for direct write. As this needs
2542 * to be an invariant for the write part only, the verifier
2543 * detects writes and adds a prologue that is calling
2544 * **bpf_skb_pull_data()** to effectively unclone the *skb* from
2545 * the very beginning in case it is indeed cloned.
2547 * A call to this helper is susceptible to change the underlying
2548 * packet buffer. Therefore, at load time, all checks on pointers
2549 * previously done by the verifier are invalidated and must be
2550 * performed again, if the helper is used in combination with
2551 * direct packet access.
2553 * 0 on success, or a negative error in case of failure.
2555 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2557 * Add the checksum *csum* into *skb*\ **->csum** in case the
2558 * driver has supplied a checksum for the entire packet into that
2559 * field. Return an error otherwise. This helper is intended to be
2560 * used in combination with **bpf_csum_diff**\ (), in particular
2561 * when the checksum needs to be updated after data has been
2562 * written into the packet through direct packet access.
2564 * The checksum on success, or a negative error code in case of
2567 * void bpf_set_hash_invalid(struct sk_buff *skb)
2569 * Invalidate the current *skb*\ **->hash**. It can be used after
2570 * mangling on headers through direct packet access, in order to
2571 * indicate that the hash is outdated and to trigger a
2572 * recalculation the next time the kernel tries to access this
2573 * hash or when the **bpf_get_hash_recalc**\ () helper is called.
2577 * long bpf_get_numa_node_id(void)
2579 * Return the id of the current NUMA node. The primary use case
2580 * for this helper is the selection of sockets for the local NUMA
2581 * node, when the program is attached to sockets using the
2582 * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2583 * but the helper is also available to other eBPF program types,
2584 * similarly to **bpf_get_smp_processor_id**\ ().
2586 * The id of current NUMA node.
2588 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2590 * Grows headroom of packet associated to *skb* and adjusts the
2591 * offset of the MAC header accordingly, adding *len* bytes of
2592 * space. It automatically extends and reallocates memory as
2595 * This helper can be used on a layer 3 *skb* to push a MAC header
2596 * for redirection into a layer 2 device.
2598 * All values for *flags* are reserved for future usage, and must
2601 * A call to this helper is susceptible to change the underlying
2602 * packet buffer. Therefore, at load time, all checks on pointers
2603 * previously done by the verifier are invalidated and must be
2604 * performed again, if the helper is used in combination with
2605 * direct packet access.
2607 * 0 on success, or a negative error in case of failure.
2609 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2611 * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2612 * it is possible to use a negative value for *delta*. This helper
2613 * can be used to prepare the packet for pushing or popping
2616 * A call to this helper is susceptible to change the underlying
2617 * packet buffer. Therefore, at load time, all checks on pointers
2618 * previously done by the verifier are invalidated and must be
2619 * performed again, if the helper is used in combination with
2620 * direct packet access.
2622 * 0 on success, or a negative error in case of failure.
2624 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2626 * Copy a NUL terminated string from an unsafe kernel address
2627 * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2630 * Generally, use **bpf_probe_read_user_str**\ () or
2631 * **bpf_probe_read_kernel_str**\ () instead.
2633 * On success, the strictly positive length of the string,
2634 * including the trailing NUL character. On error, a negative
2637 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2639 * If the **struct sk_buff** pointed by *skb* has a known socket,
2640 * retrieve the cookie (generated by the kernel) of this socket.
2641 * If no cookie has been set yet, generate a new cookie. Once
2642 * generated, the socket cookie remains stable for the life of the
2643 * socket. This helper can be useful for monitoring per socket
2644 * networking traffic statistics as it provides a global socket
2645 * identifier that can be assumed unique.
2647 * A 8-byte long unique number on success, or 0 if the socket
2648 * field is missing inside *skb*.
2650 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2652 * Equivalent to bpf_get_socket_cookie() helper that accepts
2653 * *skb*, but gets socket from **struct bpf_sock_addr** context.
2655 * A 8-byte long unique number.
2657 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2659 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2660 * *skb*, but gets socket from **struct bpf_sock_ops** context.
2662 * A 8-byte long unique number.
2664 * u64 bpf_get_socket_cookie(struct sock *sk)
2666 * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2667 * *sk*, but gets socket from a BTF **struct sock**. This helper
2668 * also works for sleepable programs.
2670 * A 8-byte long unique number or 0 if *sk* is NULL.
2672 * u32 bpf_get_socket_uid(struct sk_buff *skb)
2674 * Get the owner UID of the socked associated to *skb*.
2676 * The owner UID of the socket associated to *skb*. If the socket
2677 * is **NULL**, or if it is not a full socket (i.e. if it is a
2678 * time-wait or a request socket instead), **overflowuid** value
2679 * is returned (note that **overflowuid** might also be the actual
2680 * UID value for the socket).
2682 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2684 * Set the full hash for *skb* (set the field *skb*\ **->hash**)
2689 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2691 * Emulate a call to **setsockopt()** on the socket associated to
2692 * *bpf_socket*, which must be a full socket. The *level* at
2693 * which the option resides and the name *optname* of the option
2694 * must be specified, see **setsockopt(2)** for more information.
2695 * The option value of length *optlen* is pointed by *optval*.
2697 * *bpf_socket* should be one of the following:
2699 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2700 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2701 * and **BPF_CGROUP_INET6_CONNECT**.
2703 * This helper actually implements a subset of **setsockopt()**.
2704 * It supports the following *level*\ s:
2706 * * **SOL_SOCKET**, which supports the following *optname*\ s:
2707 * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2708 * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2709 * **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2710 * **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2711 * * **IPPROTO_TCP**, which supports the following *optname*\ s:
2712 * **TCP_CONGESTION**, **TCP_BPF_IW**,
2713 * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2714 * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2715 * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2716 * **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2717 * **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2718 * **TCP_BPF_RTO_MIN**.
2719 * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2720 * * **IPPROTO_IPV6**, which supports the following *optname*\ s:
2721 * **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2723 * 0 on success, or a negative error in case of failure.
2725 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2727 * Grow or shrink the room for data in the packet associated to
2728 * *skb* by *len_diff*, and according to the selected *mode*.
2730 * By default, the helper will reset any offloaded checksum
2731 * indicator of the skb to CHECKSUM_NONE. This can be avoided
2732 * by the following flag:
2734 * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2735 * checksum data of the skb to CHECKSUM_NONE.
2737 * There are two supported modes at this time:
2739 * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2740 * (room space is added or removed between the layer 2 and
2743 * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2744 * (room space is added or removed between the layer 3 and
2747 * The following flags are supported at this time:
2749 * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2750 * Adjusting mss in this way is not allowed for datagrams.
2752 * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2753 * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2754 * Any new space is reserved to hold a tunnel header.
2755 * Configure skb offsets and other fields accordingly.
2757 * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2758 * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2759 * Use with ENCAP_L3 flags to further specify the tunnel type.
2761 * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2762 * Use with ENCAP_L3/L4 flags to further specify the tunnel
2763 * type; *len* is the length of the inner MAC header.
2765 * * **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2766 * Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2767 * L2 type as Ethernet.
2769 * * **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
2770 * **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
2771 * Indicate the new IP header version after decapsulating the outer
2772 * IP header. Used when the inner and outer IP versions are different.
2774 * A call to this helper is susceptible to change the underlying
2775 * packet buffer. Therefore, at load time, all checks on pointers
2776 * previously done by the verifier are invalidated and must be
2777 * performed again, if the helper is used in combination with
2778 * direct packet access.
2780 * 0 on success, or a negative error in case of failure.
2782 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2784 * Redirect the packet to the endpoint referenced by *map* at
2785 * index *key*. Depending on its type, this *map* can contain
2786 * references to net devices (for forwarding packets through other
2787 * ports), or to CPUs (for redirecting XDP frames to another CPU;
2788 * but this is only implemented for native XDP (with driver
2789 * support) as of this writing).
2791 * The lower two bits of *flags* are used as the return code if
2792 * the map lookup fails. This is so that the return value can be
2793 * one of the XDP program return codes up to **XDP_TX**, as chosen
2794 * by the caller. The higher bits of *flags* can be set to
2795 * BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2797 * With BPF_F_BROADCAST the packet will be broadcasted to all the
2798 * interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2799 * interface will be excluded when do broadcasting.
2801 * See also **bpf_redirect**\ (), which only supports redirecting
2802 * to an ifindex, but doesn't require a map to do so.
2804 * **XDP_REDIRECT** on success, or the value of the two lower bits
2805 * of the *flags* argument on error.
2807 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2809 * Redirect the packet to the socket referenced by *map* (of type
2810 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2811 * egress interfaces can be used for redirection. The
2812 * **BPF_F_INGRESS** value in *flags* is used to make the
2813 * distinction (ingress path is selected if the flag is present,
2814 * egress path otherwise). This is the only flag supported for now.
2816 * **SK_PASS** on success, or **SK_DROP** on error.
2818 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2820 * Add an entry to, or update a *map* referencing sockets. The
2821 * *skops* is used as a new value for the entry associated to
2822 * *key*. *flags* is one of:
2825 * The entry for *key* must not exist in the map.
2827 * The entry for *key* must already exist in the map.
2829 * No condition on the existence of the entry for *key*.
2831 * If the *map* has eBPF programs (parser and verdict), those will
2832 * be inherited by the socket being added. If the socket is
2833 * already attached to eBPF programs, this results in an error.
2835 * 0 on success, or a negative error in case of failure.
2837 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2839 * Adjust the address pointed by *xdp_md*\ **->data_meta** by
2840 * *delta* (which can be positive or negative). Note that this
2841 * operation modifies the address stored in *xdp_md*\ **->data**,
2842 * so the latter must be loaded only after the helper has been
2845 * The use of *xdp_md*\ **->data_meta** is optional and programs
2846 * are not required to use it. The rationale is that when the
2847 * packet is processed with XDP (e.g. as DoS filter), it is
2848 * possible to push further meta data along with it before passing
2849 * to the stack, and to give the guarantee that an ingress eBPF
2850 * program attached as a TC classifier on the same device can pick
2851 * this up for further post-processing. Since TC works with socket
2852 * buffers, it remains possible to set from XDP the **mark** or
2853 * **priority** pointers, or other pointers for the socket buffer.
2854 * Having this scratch space generic and programmable allows for
2855 * more flexibility as the user is free to store whatever meta
2858 * A call to this helper is susceptible to change the underlying
2859 * packet buffer. Therefore, at load time, all checks on pointers
2860 * previously done by the verifier are invalidated and must be
2861 * performed again, if the helper is used in combination with
2862 * direct packet access.
2864 * 0 on success, or a negative error in case of failure.
2866 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2868 * Read the value of a perf event counter, and store it into *buf*
2869 * of size *buf_size*. This helper relies on a *map* of type
2870 * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2871 * counter is selected when *map* is updated with perf event file
2872 * descriptors. The *map* is an array whose size is the number of
2873 * available CPUs, and each cell contains a value relative to one
2874 * CPU. The value to retrieve is indicated by *flags*, that
2875 * contains the index of the CPU to look up, masked with
2876 * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2877 * **BPF_F_CURRENT_CPU** to indicate that the value for the
2878 * current CPU should be retrieved.
2880 * This helper behaves in a way close to
2881 * **bpf_perf_event_read**\ () helper, save that instead of
2882 * just returning the value observed, it fills the *buf*
2883 * structure. This allows for additional data to be retrieved: in
2884 * particular, the enabled and running times (in *buf*\
2885 * **->enabled** and *buf*\ **->running**, respectively) are
2886 * copied. In general, **bpf_perf_event_read_value**\ () is
2887 * recommended over **bpf_perf_event_read**\ (), which has some
2888 * ABI issues and provides fewer functionalities.
2890 * These values are interesting, because hardware PMU (Performance
2891 * Monitoring Unit) counters are limited resources. When there are
2892 * more PMU based perf events opened than available counters,
2893 * kernel will multiplex these events so each event gets certain
2894 * percentage (but not all) of the PMU time. In case that
2895 * multiplexing happens, the number of samples or counter value
2896 * will not reflect the case compared to when no multiplexing
2897 * occurs. This makes comparison between different runs difficult.
2898 * Typically, the counter value should be normalized before
2899 * comparing to other experiments. The usual normalization is done
2904 * normalized_counter = counter * t_enabled / t_running
2906 * Where t_enabled is the time enabled for event and t_running is
2907 * the time running for event since last normalization. The
2908 * enabled and running times are accumulated since the perf event
2909 * open. To achieve scaling factor between two invocations of an
2910 * eBPF program, users can use CPU id as the key (which is
2911 * typical for perf array usage model) to remember the previous
2912 * value and do the calculation inside the eBPF program.
2914 * 0 on success, or a negative error in case of failure.
2916 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2918 * For an eBPF program attached to a perf event, retrieve the
2919 * value of the event counter associated to *ctx* and store it in
2920 * the structure pointed by *buf* and of size *buf_size*. Enabled
2921 * and running times are also stored in the structure (see
2922 * description of helper **bpf_perf_event_read_value**\ () for
2925 * 0 on success, or a negative error in case of failure.
2927 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2929 * Emulate a call to **getsockopt()** on the socket associated to
2930 * *bpf_socket*, which must be a full socket. The *level* at
2931 * which the option resides and the name *optname* of the option
2932 * must be specified, see **getsockopt(2)** for more information.
2933 * The retrieved value is stored in the structure pointed by
2934 * *opval* and of length *optlen*.
2936 * *bpf_socket* should be one of the following:
2938 * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2939 * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2940 * and **BPF_CGROUP_INET6_CONNECT**.
2942 * This helper actually implements a subset of **getsockopt()**.
2943 * It supports the same set of *optname*\ s that is supported by
2944 * the **bpf_setsockopt**\ () helper. The exceptions are
2945 * **TCP_BPF_*** is **bpf_setsockopt**\ () only and
2946 * **TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
2948 * 0 on success, or a negative error in case of failure.
2950 * long bpf_override_return(struct pt_regs *regs, u64 rc)
2952 * Used for error injection, this helper uses kprobes to override
2953 * the return value of the probed function, and to set it to *rc*.
2954 * The first argument is the context *regs* on which the kprobe
2957 * This helper works by setting the PC (program counter)
2958 * to an override function which is run in place of the original
2959 * probed function. This means the probed function is not run at
2960 * all. The replacement function just returns with the required
2963 * This helper has security implications, and thus is subject to
2964 * restrictions. It is only available if the kernel was compiled
2965 * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2966 * option, and in this case it only works on functions tagged with
2967 * **ALLOW_ERROR_INJECTION** in the kernel code.
2969 * Also, the helper is only available for the architectures having
2970 * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2971 * x86 architecture is the only one to support this feature.
2975 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2977 * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2978 * for the full TCP socket associated to *bpf_sock_ops* to
2981 * The primary use of this field is to determine if there should
2982 * be calls to eBPF programs of type
2983 * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2984 * code. A program of the same type can change its value, per
2985 * connection and as necessary, when the connection is
2986 * established. This field is directly accessible for reading, but
2987 * this helper must be used for updates in order to return an
2988 * error if an eBPF program tries to set a callback that is not
2989 * supported in the current kernel.
2991 * *argval* is a flag array which can combine these flags:
2993 * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2994 * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2995 * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2996 * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2998 * Therefore, this function can be used to clear a callback flag by
2999 * setting the appropriate bit to zero. e.g. to disable the RTO
3002 * **bpf_sock_ops_cb_flags_set(bpf_sock,**
3003 * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3005 * Here are some examples of where one could call such eBPF
3009 * * When a packet is retransmitted.
3010 * * When the connection terminates.
3011 * * When a packet is sent.
3012 * * When a packet is received.
3014 * Code **-EINVAL** if the socket is not a full TCP socket;
3015 * otherwise, a positive number containing the bits that could not
3016 * be set is returned (which comes down to 0 if all bits were set
3019 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3021 * This helper is used in programs implementing policies at the
3022 * socket level. If the message *msg* is allowed to pass (i.e. if
3023 * the verdict eBPF program returns **SK_PASS**), redirect it to
3024 * the socket referenced by *map* (of type
3025 * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3026 * egress interfaces can be used for redirection. The
3027 * **BPF_F_INGRESS** value in *flags* is used to make the
3028 * distinction (ingress path is selected if the flag is present,
3029 * egress path otherwise). This is the only flag supported for now.
3031 * **SK_PASS** on success, or **SK_DROP** on error.
3033 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3035 * For socket policies, apply the verdict of the eBPF program to
3036 * the next *bytes* (number of bytes) of message *msg*.
3038 * For example, this helper can be used in the following cases:
3040 * * A single **sendmsg**\ () or **sendfile**\ () system call
3041 * contains multiple logical messages that the eBPF program is
3042 * supposed to read and for which it should apply a verdict.
3043 * * An eBPF program only cares to read the first *bytes* of a
3044 * *msg*. If the message has a large payload, then setting up
3045 * and calling the eBPF program repeatedly for all bytes, even
3046 * though the verdict is already known, would create unnecessary
3049 * When called from within an eBPF program, the helper sets a
3050 * counter internal to the BPF infrastructure, that is used to
3051 * apply the last verdict to the next *bytes*. If *bytes* is
3052 * smaller than the current data being processed from a
3053 * **sendmsg**\ () or **sendfile**\ () system call, the first
3054 * *bytes* will be sent and the eBPF program will be re-run with
3055 * the pointer for start of data pointing to byte number *bytes*
3056 * **+ 1**. If *bytes* is larger than the current data being
3057 * processed, then the eBPF verdict will be applied to multiple
3058 * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3061 * Note that if a socket closes with the internal counter holding
3062 * a non-zero value, this is not a problem because data is not
3063 * being buffered for *bytes* and is sent as it is received.
3067 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3069 * For socket policies, prevent the execution of the verdict eBPF
3070 * program for message *msg* until *bytes* (byte number) have been
3073 * This can be used when one needs a specific number of bytes
3074 * before a verdict can be assigned, even if the data spans
3075 * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3076 * case would be a user calling **sendmsg**\ () repeatedly with
3077 * 1-byte long message segments. Obviously, this is bad for
3078 * performance, but it is still valid. If the eBPF program needs
3079 * *bytes* bytes to validate a header, this helper can be used to
3080 * prevent the eBPF program to be called again until *bytes* have
3085 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3087 * For socket policies, pull in non-linear data from user space
3088 * for *msg* and set pointers *msg*\ **->data** and *msg*\
3089 * **->data_end** to *start* and *end* bytes offsets into *msg*,
3092 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3093 * *msg* it can only parse data that the (**data**, **data_end**)
3094 * pointers have already consumed. For **sendmsg**\ () hooks this
3095 * is likely the first scatterlist element. But for calls relying
3096 * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3097 * be the range (**0**, **0**) because the data is shared with
3098 * user space and by default the objective is to avoid allowing
3099 * user space to modify data while (or after) eBPF verdict is
3100 * being decided. This helper can be used to pull in data and to
3101 * set the start and end pointer to given values. Data will be
3102 * copied if necessary (i.e. if data was not linear and if start
3103 * and end pointers do not point to the same chunk).
3105 * A call to this helper is susceptible to change the underlying
3106 * packet buffer. Therefore, at load time, all checks on pointers
3107 * previously done by the verifier are invalidated and must be
3108 * performed again, if the helper is used in combination with
3109 * direct packet access.
3111 * All values for *flags* are reserved for future usage, and must
3114 * 0 on success, or a negative error in case of failure.
3116 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3118 * Bind the socket associated to *ctx* to the address pointed by
3119 * *addr*, of length *addr_len*. This allows for making outgoing
3120 * connection from the desired IP address, which can be useful for
3121 * example when all processes inside a cgroup should use one
3122 * single IP address on a host that has multiple IP configured.
3124 * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3125 * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3126 * **AF_INET6**). It's advised to pass zero port (**sin_port**
3127 * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3128 * behavior and lets the kernel efficiently pick up an unused
3129 * port as long as 4-tuple is unique. Passing non-zero port might
3130 * lead to degraded performance.
3132 * 0 on success, or a negative error in case of failure.
3134 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3136 * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3137 * possible to both shrink and grow the packet tail.
3138 * Shrink done via *delta* being a negative integer.
3140 * A call to this helper is susceptible to change the underlying
3141 * packet buffer. Therefore, at load time, all checks on pointers
3142 * previously done by the verifier are invalidated and must be
3143 * performed again, if the helper is used in combination with
3144 * direct packet access.
3146 * 0 on success, or a negative error in case of failure.
3148 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3150 * Retrieve the XFRM state (IP transform framework, see also
3151 * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3153 * The retrieved value is stored in the **struct bpf_xfrm_state**
3154 * pointed by *xfrm_state* and of length *size*.
3156 * All values for *flags* are reserved for future usage, and must
3159 * This helper is available only if the kernel was compiled with
3160 * **CONFIG_XFRM** configuration option.
3162 * 0 on success, or a negative error in case of failure.
3164 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3166 * Return a user or a kernel stack in bpf program provided buffer.
3167 * To achieve this, the helper needs *ctx*, which is a pointer
3168 * to the context on which the tracing program is executed.
3169 * To store the stacktrace, the bpf program provides *buf* with
3170 * a nonnegative *size*.
3172 * The last argument, *flags*, holds the number of stack frames to
3173 * skip (from 0 to 255), masked with
3174 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3175 * the following flags:
3177 * **BPF_F_USER_STACK**
3178 * Collect a user space stack instead of a kernel stack.
3179 * **BPF_F_USER_BUILD_ID**
3180 * Collect (build_id, file_offset) instead of ips for user
3181 * stack, only valid if **BPF_F_USER_STACK** is also
3184 * *file_offset* is an offset relative to the beginning
3185 * of the executable or shared object file backing the vma
3186 * which the *ip* falls in. It is *not* an offset relative
3187 * to that object's base address. Accordingly, it must be
3188 * adjusted by adding (sh_addr - sh_offset), where
3189 * sh_{addr,offset} correspond to the executable section
3190 * containing *file_offset* in the object, for comparisons
3191 * to symbols' st_value to be valid.
3193 * **bpf_get_stack**\ () can collect up to
3194 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3195 * to sufficient large buffer size. Note that
3196 * this limit can be controlled with the **sysctl** program, and
3197 * that it should be manually increased in order to profile long
3198 * user stacks (such as stacks for Java programs). To do so, use:
3202 * # sysctl kernel.perf_event_max_stack=<new value>
3204 * The non-negative copied *buf* length equal to or less than
3205 * *size* on success, or a negative error in case of failure.
3207 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3209 * This helper is similar to **bpf_skb_load_bytes**\ () in that
3210 * it provides an easy way to load *len* bytes from *offset*
3211 * from the packet associated to *skb*, into the buffer pointed
3212 * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3213 * a fifth argument *start_header* exists in order to select a
3214 * base offset to start from. *start_header* can be one of:
3216 * **BPF_HDR_START_MAC**
3217 * Base offset to load data from is *skb*'s mac header.
3218 * **BPF_HDR_START_NET**
3219 * Base offset to load data from is *skb*'s network header.
3221 * In general, "direct packet access" is the preferred method to
3222 * access packet data, however, this helper is in particular useful
3223 * in socket filters where *skb*\ **->data** does not always point
3224 * to the start of the mac header and where "direct packet access"
3227 * 0 on success, or a negative error in case of failure.
3229 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3231 * Do FIB lookup in kernel tables using parameters in *params*.
3232 * If lookup is successful and result shows packet is to be
3233 * forwarded, the neighbor tables are searched for the nexthop.
3234 * If successful (ie., FIB lookup shows forwarding and nexthop
3235 * is resolved), the nexthop address is returned in ipv4_dst
3236 * or ipv6_dst based on family, smac is set to mac address of
3237 * egress device, dmac is set to nexthop mac address, rt_metric
3238 * is set to metric from route (IPv4/IPv6 only), and ifindex
3239 * is set to the device index of the nexthop from the FIB lookup.
3241 * *plen* argument is the size of the passed in struct.
3242 * *flags* argument can be a combination of one or more of the
3245 * **BPF_FIB_LOOKUP_DIRECT**
3246 * Do a direct table lookup vs full lookup using FIB
3248 * **BPF_FIB_LOOKUP_TBID**
3249 * Used with BPF_FIB_LOOKUP_DIRECT.
3250 * Use the routing table ID present in *params*->tbid
3251 * for the fib lookup.
3252 * **BPF_FIB_LOOKUP_OUTPUT**
3253 * Perform lookup from an egress perspective (default is
3255 * **BPF_FIB_LOOKUP_SKIP_NEIGH**
3256 * Skip the neighbour table lookup. *params*->dmac
3257 * and *params*->smac will not be set as output. A common
3258 * use case is to call **bpf_redirect_neigh**\ () after
3259 * doing **bpf_fib_lookup**\ ().
3261 * *ctx* is either **struct xdp_md** for XDP programs or
3262 * **struct sk_buff** tc cls_act programs.
3264 * * < 0 if any input argument is invalid
3265 * * 0 on success (packet is forwarded, nexthop neighbor exists)
3266 * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3267 * packet is not forwarded or needs assist from full stack
3269 * If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3270 * was exceeded and output params->mtu_result contains the MTU.
3272 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3274 * Add an entry to, or update a sockhash *map* referencing sockets.
3275 * The *skops* is used as a new value for the entry associated to
3276 * *key*. *flags* is one of:
3279 * The entry for *key* must not exist in the map.
3281 * The entry for *key* must already exist in the map.
3283 * No condition on the existence of the entry for *key*.
3285 * If the *map* has eBPF programs (parser and verdict), those will
3286 * be inherited by the socket being added. If the socket is
3287 * already attached to eBPF programs, this results in an error.
3289 * 0 on success, or a negative error in case of failure.
3291 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3293 * This helper is used in programs implementing policies at the
3294 * socket level. If the message *msg* is allowed to pass (i.e. if
3295 * the verdict eBPF program returns **SK_PASS**), redirect it to
3296 * the socket referenced by *map* (of type
3297 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3298 * egress interfaces can be used for redirection. The
3299 * **BPF_F_INGRESS** value in *flags* is used to make the
3300 * distinction (ingress path is selected if the flag is present,
3301 * egress path otherwise). This is the only flag supported for now.
3303 * **SK_PASS** on success, or **SK_DROP** on error.
3305 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3307 * This helper is used in programs implementing policies at the
3308 * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3309 * if the verdict eBPF program returns **SK_PASS**), redirect it
3310 * to the socket referenced by *map* (of type
3311 * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3312 * egress interfaces can be used for redirection. The
3313 * **BPF_F_INGRESS** value in *flags* is used to make the
3314 * distinction (ingress path is selected if the flag is present,
3315 * egress otherwise). This is the only flag supported for now.
3317 * **SK_PASS** on success, or **SK_DROP** on error.
3319 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3321 * Encapsulate the packet associated to *skb* within a Layer 3
3322 * protocol header. This header is provided in the buffer at
3323 * address *hdr*, with *len* its size in bytes. *type* indicates
3324 * the protocol of the header and can be one of:
3326 * **BPF_LWT_ENCAP_SEG6**
3327 * IPv6 encapsulation with Segment Routing Header
3328 * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3329 * the IPv6 header is computed by the kernel.
3330 * **BPF_LWT_ENCAP_SEG6_INLINE**
3331 * Only works if *skb* contains an IPv6 packet. Insert a
3332 * Segment Routing Header (**struct ipv6_sr_hdr**) inside
3334 * **BPF_LWT_ENCAP_IP**
3335 * IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3336 * must be IPv4 or IPv6, followed by zero or more
3337 * additional headers, up to **LWT_BPF_MAX_HEADROOM**
3338 * total bytes in all prepended headers. Please note that
3339 * if **skb_is_gso**\ (*skb*) is true, no more than two
3340 * headers can be prepended, and the inner header, if
3341 * present, should be either GRE or UDP/GUE.
3343 * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3344 * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3345 * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3346 * **BPF_PROG_TYPE_LWT_XMIT**.
3348 * A call to this helper is susceptible to change the underlying
3349 * packet buffer. Therefore, at load time, all checks on pointers
3350 * previously done by the verifier are invalidated and must be
3351 * performed again, if the helper is used in combination with
3352 * direct packet access.
3354 * 0 on success, or a negative error in case of failure.
3356 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3358 * Store *len* bytes from address *from* into the packet
3359 * associated to *skb*, at *offset*. Only the flags, tag and TLVs
3360 * inside the outermost IPv6 Segment Routing Header can be
3361 * modified through this helper.
3363 * A call to this helper is susceptible to change the underlying
3364 * packet buffer. Therefore, at load time, all checks on pointers
3365 * previously done by the verifier are invalidated and must be
3366 * performed again, if the helper is used in combination with
3367 * direct packet access.
3369 * 0 on success, or a negative error in case of failure.
3371 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3373 * Adjust the size allocated to TLVs in the outermost IPv6
3374 * Segment Routing Header contained in the packet associated to
3375 * *skb*, at position *offset* by *delta* bytes. Only offsets
3376 * after the segments are accepted. *delta* can be as well
3377 * positive (growing) as negative (shrinking).
3379 * A call to this helper is susceptible to change the underlying
3380 * packet buffer. Therefore, at load time, all checks on pointers
3381 * previously done by the verifier are invalidated and must be
3382 * performed again, if the helper is used in combination with
3383 * direct packet access.
3385 * 0 on success, or a negative error in case of failure.
3387 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3389 * Apply an IPv6 Segment Routing action of type *action* to the
3390 * packet associated to *skb*. Each action takes a parameter
3391 * contained at address *param*, and of length *param_len* bytes.
3392 * *action* can be one of:
3394 * **SEG6_LOCAL_ACTION_END_X**
3395 * End.X action: Endpoint with Layer-3 cross-connect.
3396 * Type of *param*: **struct in6_addr**.
3397 * **SEG6_LOCAL_ACTION_END_T**
3398 * End.T action: Endpoint with specific IPv6 table lookup.
3399 * Type of *param*: **int**.
3400 * **SEG6_LOCAL_ACTION_END_B6**
3401 * End.B6 action: Endpoint bound to an SRv6 policy.
3402 * Type of *param*: **struct ipv6_sr_hdr**.
3403 * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
3404 * End.B6.Encap action: Endpoint bound to an SRv6
3405 * encapsulation policy.
3406 * Type of *param*: **struct ipv6_sr_hdr**.
3408 * A call to this helper is susceptible to change the underlying
3409 * packet buffer. Therefore, at load time, all checks on pointers
3410 * previously done by the verifier are invalidated and must be
3411 * performed again, if the helper is used in combination with
3412 * direct packet access.
3414 * 0 on success, or a negative error in case of failure.
3416 * long bpf_rc_repeat(void *ctx)
3418 * This helper is used in programs implementing IR decoding, to
3419 * report a successfully decoded repeat key message. This delays
3420 * the generation of a key up event for previously generated
3423 * Some IR protocols like NEC have a special IR message for
3424 * repeating last button, for when a button is held down.
3426 * The *ctx* should point to the lirc sample as passed into
3429 * This helper is only available is the kernel was compiled with
3430 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3435 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3437 * This helper is used in programs implementing IR decoding, to
3438 * report a successfully decoded key press with *scancode*,
3439 * *toggle* value in the given *protocol*. The scancode will be
3440 * translated to a keycode using the rc keymap, and reported as
3441 * an input key down event. After a period a key up event is
3442 * generated. This period can be extended by calling either
3443 * **bpf_rc_keydown**\ () again with the same values, or calling
3444 * **bpf_rc_repeat**\ ().
3446 * Some protocols include a toggle bit, in case the button was
3447 * released and pressed again between consecutive scancodes.
3449 * The *ctx* should point to the lirc sample as passed into
3452 * The *protocol* is the decoded protocol number (see
3453 * **enum rc_proto** for some predefined values).
3455 * This helper is only available is the kernel was compiled with
3456 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3461 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3463 * Return the cgroup v2 id of the socket associated with the *skb*.
3464 * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3465 * helper for cgroup v1 by providing a tag resp. identifier that
3466 * can be matched on or used for map lookups e.g. to implement
3467 * policy. The cgroup v2 id of a given path in the hierarchy is
3468 * exposed in user space through the f_handle API in order to get
3469 * to the same 64-bit id.
3471 * This helper can be used on TC egress path, but not on ingress,
3472 * and is available only if the kernel was compiled with the
3473 * **CONFIG_SOCK_CGROUP_DATA** configuration option.
3475 * The id is returned or 0 in case the id could not be retrieved.
3477 * u64 bpf_get_current_cgroup_id(void)
3479 * Get the current cgroup id based on the cgroup within which
3480 * the current task is running.
3482 * A 64-bit integer containing the current cgroup id based
3483 * on the cgroup within which the current task is running.
3485 * void *bpf_get_local_storage(void *map, u64 flags)
3487 * Get the pointer to the local storage area.
3488 * The type and the size of the local storage is defined
3489 * by the *map* argument.
3490 * The *flags* meaning is specific for each map type,
3491 * and has to be 0 for cgroup local storage.
3493 * Depending on the BPF program type, a local storage area
3494 * can be shared between multiple instances of the BPF program,
3495 * running simultaneously.
3497 * A user should care about the synchronization by himself.
3498 * For example, by using the **BPF_ATOMIC** instructions to alter
3501 * A pointer to the local storage area.
3503 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3505 * Select a **SO_REUSEPORT** socket from a
3506 * **BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3507 * It checks the selected socket is matching the incoming
3508 * request in the socket buffer.
3510 * 0 on success, or a negative error in case of failure.
3512 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3514 * Return id of cgroup v2 that is ancestor of cgroup associated
3515 * with the *skb* at the *ancestor_level*. The root cgroup is at
3516 * *ancestor_level* zero and each step down the hierarchy
3517 * increments the level. If *ancestor_level* == level of cgroup
3518 * associated with *skb*, then return value will be same as that
3519 * of **bpf_skb_cgroup_id**\ ().
3521 * The helper is useful to implement policies based on cgroups
3522 * that are upper in hierarchy than immediate cgroup associated
3525 * The format of returned id and helper limitations are same as in
3526 * **bpf_skb_cgroup_id**\ ().
3528 * The id is returned or 0 in case the id could not be retrieved.
3530 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3532 * Look for TCP socket matching *tuple*, optionally in a child
3533 * network namespace *netns*. The return value must be checked,
3534 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3536 * The *ctx* should point to the context of the program, such as
3537 * the skb or socket (depending on the hook in use). This is used
3538 * to determine the base network namespace for the lookup.
3540 * *tuple_size* must be one of:
3542 * **sizeof**\ (*tuple*\ **->ipv4**)
3543 * Look for an IPv4 socket.
3544 * **sizeof**\ (*tuple*\ **->ipv6**)
3545 * Look for an IPv6 socket.
3547 * If the *netns* is a negative signed 32-bit integer, then the
3548 * socket lookup table in the netns associated with the *ctx*
3549 * will be used. For the TC hooks, this is the netns of the device
3550 * in the skb. For socket hooks, this is the netns of the socket.
3551 * If *netns* is any other signed 32-bit value greater than or
3552 * equal to zero then it specifies the ID of the netns relative to
3553 * the netns associated with the *ctx*. *netns* values beyond the
3554 * range of 32-bit integers are reserved for future use.
3556 * All values for *flags* are reserved for future usage, and must
3559 * This helper is available only if the kernel was compiled with
3560 * **CONFIG_NET** configuration option.
3562 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3563 * For sockets with reuseport option, the **struct bpf_sock**
3564 * result is from *reuse*\ **->socks**\ [] using the hash of the
3567 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3569 * Look for UDP socket matching *tuple*, optionally in a child
3570 * network namespace *netns*. The return value must be checked,
3571 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3573 * The *ctx* should point to the context of the program, such as
3574 * the skb or socket (depending on the hook in use). This is used
3575 * to determine the base network namespace for the lookup.
3577 * *tuple_size* must be one of:
3579 * **sizeof**\ (*tuple*\ **->ipv4**)
3580 * Look for an IPv4 socket.
3581 * **sizeof**\ (*tuple*\ **->ipv6**)
3582 * Look for an IPv6 socket.
3584 * If the *netns* is a negative signed 32-bit integer, then the
3585 * socket lookup table in the netns associated with the *ctx*
3586 * will be used. For the TC hooks, this is the netns of the device
3587 * in the skb. For socket hooks, this is the netns of the socket.
3588 * If *netns* is any other signed 32-bit value greater than or
3589 * equal to zero then it specifies the ID of the netns relative to
3590 * the netns associated with the *ctx*. *netns* values beyond the
3591 * range of 32-bit integers are reserved for future use.
3593 * All values for *flags* are reserved for future usage, and must
3596 * This helper is available only if the kernel was compiled with
3597 * **CONFIG_NET** configuration option.
3599 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3600 * For sockets with reuseport option, the **struct bpf_sock**
3601 * result is from *reuse*\ **->socks**\ [] using the hash of the
3604 * long bpf_sk_release(void *sock)
3606 * Release the reference held by *sock*. *sock* must be a
3607 * non-**NULL** pointer that was returned from
3608 * **bpf_sk_lookup_xxx**\ ().
3610 * 0 on success, or a negative error in case of failure.
3612 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3614 * Push an element *value* in *map*. *flags* is one of:
3617 * If the queue/stack is full, the oldest element is
3618 * removed to make room for this.
3620 * 0 on success, or a negative error in case of failure.
3622 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3624 * Pop an element from *map*.
3626 * 0 on success, or a negative error in case of failure.
3628 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3630 * Get an element from *map* without removing it.
3632 * 0 on success, or a negative error in case of failure.
3634 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3636 * For socket policies, insert *len* bytes into *msg* at offset
3639 * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3640 * *msg* it may want to insert metadata or options into the *msg*.
3641 * This can later be read and used by any of the lower layer BPF
3644 * This helper may fail if under memory pressure (a malloc
3645 * fails) in these cases BPF programs will get an appropriate
3646 * error and BPF programs will need to handle them.
3648 * 0 on success, or a negative error in case of failure.
3650 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3652 * Will remove *len* bytes from a *msg* starting at byte *start*.
3653 * This may result in **ENOMEM** errors under certain situations if
3654 * an allocation and copy are required due to a full ring buffer.
3655 * However, the helper will try to avoid doing the allocation
3656 * if possible. Other errors can occur if input parameters are
3657 * invalid either due to *start* byte not being valid part of *msg*
3658 * payload and/or *pop* value being to large.
3660 * 0 on success, or a negative error in case of failure.
3662 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3664 * This helper is used in programs implementing IR decoding, to
3665 * report a successfully decoded pointer movement.
3667 * The *ctx* should point to the lirc sample as passed into
3670 * This helper is only available is the kernel was compiled with
3671 * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3676 * long bpf_spin_lock(struct bpf_spin_lock *lock)
3678 * Acquire a spinlock represented by the pointer *lock*, which is
3679 * stored as part of a value of a map. Taking the lock allows to
3680 * safely update the rest of the fields in that value. The
3681 * spinlock can (and must) later be released with a call to
3682 * **bpf_spin_unlock**\ (\ *lock*\ ).
3684 * Spinlocks in BPF programs come with a number of restrictions
3687 * * **bpf_spin_lock** objects are only allowed inside maps of
3688 * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3689 * list could be extended in the future).
3690 * * BTF description of the map is mandatory.
3691 * * The BPF program can take ONE lock at a time, since taking two
3692 * or more could cause dead locks.
3693 * * Only one **struct bpf_spin_lock** is allowed per map element.
3694 * * When the lock is taken, calls (either BPF to BPF or helpers)
3696 * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3697 * allowed inside a spinlock-ed region.
3698 * * The BPF program MUST call **bpf_spin_unlock**\ () to release
3699 * the lock, on all execution paths, before it returns.
3700 * * The BPF program can access **struct bpf_spin_lock** only via
3701 * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3702 * helpers. Loading or storing data into the **struct
3703 * bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3704 * * To use the **bpf_spin_lock**\ () helper, the BTF description
3705 * of the map value must be a struct and have **struct
3706 * bpf_spin_lock** *anyname*\ **;** field at the top level.
3707 * Nested lock inside another struct is not allowed.
3708 * * The **struct bpf_spin_lock** *lock* field in a map value must
3709 * be aligned on a multiple of 4 bytes in that value.
3710 * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3711 * the **bpf_spin_lock** field to user space.
3712 * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3713 * a BPF program, do not update the **bpf_spin_lock** field.
3714 * * **bpf_spin_lock** cannot be on the stack or inside a
3715 * networking packet (it can only be inside of a map values).
3716 * * **bpf_spin_lock** is available to root only.
3717 * * Tracing programs and socket filter programs cannot use
3718 * **bpf_spin_lock**\ () due to insufficient preemption checks
3719 * (but this may change in the future).
3720 * * **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3724 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3726 * Release the *lock* previously locked by a call to
3727 * **bpf_spin_lock**\ (\ *lock*\ ).
3731 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3733 * This helper gets a **struct bpf_sock** pointer such
3734 * that all the fields in this **bpf_sock** can be accessed.
3736 * A **struct bpf_sock** pointer on success, or **NULL** in
3739 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3741 * This helper gets a **struct bpf_tcp_sock** pointer from a
3742 * **struct bpf_sock** pointer.
3744 * A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3747 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3749 * Set ECN (Explicit Congestion Notification) field of IP header
3750 * to **CE** (Congestion Encountered) if current value is **ECT**
3751 * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3754 * 1 if the **CE** flag is set (either by the current helper call
3755 * or because it was already present), 0 if it is not set.
3757 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3759 * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3760 * **bpf_sk_release**\ () is unnecessary and not allowed.
3762 * A **struct bpf_sock** pointer on success, or **NULL** in
3765 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3767 * Look for TCP socket matching *tuple*, optionally in a child
3768 * network namespace *netns*. The return value must be checked,
3769 * and if non-**NULL**, released via **bpf_sk_release**\ ().
3771 * This function is identical to **bpf_sk_lookup_tcp**\ (), except
3772 * that it also returns timewait or request sockets. Use
3773 * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3776 * This helper is available only if the kernel was compiled with
3777 * **CONFIG_NET** configuration option.
3779 * Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3780 * For sockets with reuseport option, the **struct bpf_sock**
3781 * result is from *reuse*\ **->socks**\ [] using the hash of the
3784 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3786 * Check whether *iph* and *th* contain a valid SYN cookie ACK for
3787 * the listening socket in *sk*.
3789 * *iph* points to the start of the IPv4 or IPv6 header, while
3790 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3791 * **sizeof**\ (**struct ipv6hdr**).
3793 * *th* points to the start of the TCP header, while *th_len*
3794 * contains the length of the TCP header (at least
3795 * **sizeof**\ (**struct tcphdr**)).
3797 * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3800 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3802 * Get name of sysctl in /proc/sys/ and copy it into provided by
3803 * program buffer *buf* of size *buf_len*.
3805 * The buffer is always NUL terminated, unless it's zero-sized.
3807 * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3808 * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3809 * only (e.g. "tcp_mem").
3811 * Number of character copied (not including the trailing NUL).
3813 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3814 * truncated name in this case).
3816 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3818 * Get current value of sysctl as it is presented in /proc/sys
3819 * (incl. newline, etc), and copy it as a string into provided
3820 * by program buffer *buf* of size *buf_len*.
3822 * The whole value is copied, no matter what file position user
3823 * space issued e.g. sys_read at.
3825 * The buffer is always NUL terminated, unless it's zero-sized.
3827 * Number of character copied (not including the trailing NUL).
3829 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3830 * truncated name in this case).
3832 * **-EINVAL** if current value was unavailable, e.g. because
3833 * sysctl is uninitialized and read returns -EIO for it.
3835 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3837 * Get new value being written by user space to sysctl (before
3838 * the actual write happens) and copy it as a string into
3839 * provided by program buffer *buf* of size *buf_len*.
3841 * User space may write new value at file position > 0.
3843 * The buffer is always NUL terminated, unless it's zero-sized.
3845 * Number of character copied (not including the trailing NUL).
3847 * **-E2BIG** if the buffer wasn't big enough (*buf* will contain
3848 * truncated name in this case).
3850 * **-EINVAL** if sysctl is being read.
3852 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3854 * Override new value being written by user space to sysctl with
3855 * value provided by program in buffer *buf* of size *buf_len*.
3857 * *buf* should contain a string in same form as provided by user
3858 * space on sysctl write.
3860 * User space may write new value at file position > 0. To override
3861 * the whole sysctl value file position should be set to zero.
3865 * **-E2BIG** if the *buf_len* is too big.
3867 * **-EINVAL** if sysctl is being read.
3869 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3871 * Convert the initial part of the string from buffer *buf* of
3872 * size *buf_len* to a long integer according to the given base
3873 * and save the result in *res*.
3875 * The string may begin with an arbitrary amount of white space
3876 * (as determined by **isspace**\ (3)) followed by a single
3877 * optional '**-**' sign.
3879 * Five least significant bits of *flags* encode base, other bits
3880 * are currently unused.
3882 * Base must be either 8, 10, 16 or 0 to detect it automatically
3883 * similar to user space **strtol**\ (3).
3885 * Number of characters consumed on success. Must be positive but
3886 * no more than *buf_len*.
3888 * **-EINVAL** if no valid digits were found or unsupported base
3891 * **-ERANGE** if resulting value was out of range.
3893 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3895 * Convert the initial part of the string from buffer *buf* of
3896 * size *buf_len* to an unsigned long integer according to the
3897 * given base and save the result in *res*.
3899 * The string may begin with an arbitrary amount of white space
3900 * (as determined by **isspace**\ (3)).
3902 * Five least significant bits of *flags* encode base, other bits
3903 * are currently unused.
3905 * Base must be either 8, 10, 16 or 0 to detect it automatically
3906 * similar to user space **strtoul**\ (3).
3908 * Number of characters consumed on success. Must be positive but
3909 * no more than *buf_len*.
3911 * **-EINVAL** if no valid digits were found or unsupported base
3914 * **-ERANGE** if resulting value was out of range.
3916 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3918 * Get a bpf-local-storage from a *sk*.
3920 * Logically, it could be thought of getting the value from
3921 * a *map* with *sk* as the **key**. From this
3922 * perspective, the usage is not much different from
3923 * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3924 * helper enforces the key must be a full socket and the map must
3925 * be a **BPF_MAP_TYPE_SK_STORAGE** also.
3927 * Underneath, the value is stored locally at *sk* instead of
3928 * the *map*. The *map* is used as the bpf-local-storage
3929 * "type". The bpf-local-storage "type" (i.e. the *map*) is
3930 * searched against all bpf-local-storages residing at *sk*.
3932 * *sk* is a kernel **struct sock** pointer for LSM program.
3933 * *sk* is a **struct bpf_sock** pointer for other program types.
3935 * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3936 * used such that a new bpf-local-storage will be
3937 * created if one does not exist. *value* can be used
3938 * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3939 * the initial value of a bpf-local-storage. If *value* is
3940 * **NULL**, the new bpf-local-storage will be zero initialized.
3942 * A bpf-local-storage pointer is returned on success.
3944 * **NULL** if not found or there was an error in adding
3945 * a new bpf-local-storage.
3947 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3949 * Delete a bpf-local-storage from a *sk*.
3953 * **-ENOENT** if the bpf-local-storage cannot be found.
3954 * **-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3956 * long bpf_send_signal(u32 sig)
3958 * Send signal *sig* to the process of the current task.
3959 * The signal may be delivered to any of this process's threads.
3961 * 0 on success or successfully queued.
3963 * **-EBUSY** if work queue under nmi is full.
3965 * **-EINVAL** if *sig* is invalid.
3967 * **-EPERM** if no permission to send the *sig*.
3969 * **-EAGAIN** if bpf program can try again.
3971 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3973 * Try to issue a SYN cookie for the packet with corresponding
3974 * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3976 * *iph* points to the start of the IPv4 or IPv6 header, while
3977 * *iph_len* contains **sizeof**\ (**struct iphdr**) or
3978 * **sizeof**\ (**struct ipv6hdr**).
3980 * *th* points to the start of the TCP header, while *th_len*
3981 * contains the length of the TCP header with options (at least
3982 * **sizeof**\ (**struct tcphdr**)).
3984 * On success, lower 32 bits hold the generated SYN cookie in
3985 * followed by 16 bits which hold the MSS value for that cookie,
3986 * and the top 16 bits are unused.
3988 * On failure, the returned value is one of the following:
3990 * **-EINVAL** SYN cookie cannot be issued due to error
3992 * **-ENOENT** SYN cookie should not be issued (no SYN flood)
3994 * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3996 * **-EPROTONOSUPPORT** IP packet version is not 4 or 6
3998 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4000 * Write raw *data* blob into a special BPF perf event held by
4001 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4002 * event must have the following attributes: **PERF_SAMPLE_RAW**
4003 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4004 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4006 * The *flags* are used to indicate the index in *map* for which
4007 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4008 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4009 * to indicate that the index of the current CPU core should be
4012 * The value to write, of *size*, is passed through eBPF stack and
4013 * pointed by *data*.
4015 * *ctx* is a pointer to in-kernel struct sk_buff.
4017 * This helper is similar to **bpf_perf_event_output**\ () but
4018 * restricted to raw_tracepoint bpf programs.
4020 * 0 on success, or a negative error in case of failure.
4022 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4024 * Safely attempt to read *size* bytes from user space address
4025 * *unsafe_ptr* and store the data in *dst*.
4027 * 0 on success, or a negative error in case of failure.
4029 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4031 * Safely attempt to read *size* bytes from kernel space address
4032 * *unsafe_ptr* and store the data in *dst*.
4034 * 0 on success, or a negative error in case of failure.
4036 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4038 * Copy a NUL terminated string from an unsafe user address
4039 * *unsafe_ptr* to *dst*. The *size* should include the
4040 * terminating NUL byte. In case the string length is smaller than
4041 * *size*, the target is not padded with further NUL bytes. If the
4042 * string length is larger than *size*, just *size*-1 bytes are
4043 * copied and the last byte is set to NUL.
4045 * On success, returns the number of bytes that were written,
4046 * including the terminal NUL. This makes this helper useful in
4047 * tracing programs for reading strings, and more importantly to
4048 * get its length at runtime. See the following snippet:
4052 * SEC("kprobe/sys_open")
4053 * void bpf_sys_open(struct pt_regs *ctx)
4055 * char buf[PATHLEN]; // PATHLEN is defined to 256
4056 * int res = bpf_probe_read_user_str(buf, sizeof(buf),
4059 * // Consume buf, for example push it to
4060 * // userspace via bpf_perf_event_output(); we
4061 * // can use res (the string length) as event
4062 * // size, after checking its boundaries.
4065 * In comparison, using **bpf_probe_read_user**\ () helper here
4066 * instead to read the string would require to estimate the length
4067 * at compile time, and would often result in copying more memory
4070 * Another useful use case is when parsing individual process
4071 * arguments or individual environment variables navigating
4072 * *current*\ **->mm->arg_start** and *current*\
4073 * **->mm->env_start**: using this helper and the return value,
4074 * one can quickly iterate at the right offset of the memory area.
4076 * On success, the strictly positive length of the output string,
4077 * including the trailing NUL character. On error, a negative
4080 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4082 * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4083 * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4085 * On success, the strictly positive length of the string, including
4086 * the trailing NUL character. On error, a negative value.
4088 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4090 * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4091 * *rcv_nxt* is the ack_seq to be sent out.
4093 * 0 on success, or a negative error in case of failure.
4095 * long bpf_send_signal_thread(u32 sig)
4097 * Send signal *sig* to the thread corresponding to the current task.
4099 * 0 on success or successfully queued.
4101 * **-EBUSY** if work queue under nmi is full.
4103 * **-EINVAL** if *sig* is invalid.
4105 * **-EPERM** if no permission to send the *sig*.
4107 * **-EAGAIN** if bpf program can try again.
4109 * u64 bpf_jiffies64(void)
4111 * Obtain the 64bit jiffies
4113 * The 64 bit jiffies
4115 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4117 * For an eBPF program attached to a perf event, retrieve the
4118 * branch records (**struct perf_branch_entry**) associated to *ctx*
4119 * and store it in the buffer pointed by *buf* up to size
4122 * On success, number of bytes written to *buf*. On error, a
4125 * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4126 * instead return the number of bytes required to store all the
4127 * branch entries. If this flag is set, *buf* may be NULL.
4129 * **-EINVAL** if arguments invalid or **size** not a multiple
4130 * of **sizeof**\ (**struct perf_branch_entry**\ ).
4132 * **-ENOENT** if architecture does not support branch records.
4134 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4136 * Returns 0 on success, values for *pid* and *tgid* as seen from the current
4137 * *namespace* will be returned in *nsdata*.
4139 * 0 on success, or one of the following in case of failure:
4141 * **-EINVAL** if dev and inum supplied don't match dev_t and inode number
4142 * with nsfs of current task, or if dev conversion to dev_t lost high bits.
4144 * **-ENOENT** if pidns does not exists for the current task.
4146 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4148 * Write raw *data* blob into a special BPF perf event held by
4149 * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4150 * event must have the following attributes: **PERF_SAMPLE_RAW**
4151 * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4152 * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4154 * The *flags* are used to indicate the index in *map* for which
4155 * the value must be put, masked with **BPF_F_INDEX_MASK**.
4156 * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4157 * to indicate that the index of the current CPU core should be
4160 * The value to write, of *size*, is passed through eBPF stack and
4161 * pointed by *data*.
4163 * *ctx* is a pointer to in-kernel struct xdp_buff.
4165 * This helper is similar to **bpf_perf_eventoutput**\ () but
4166 * restricted to raw_tracepoint bpf programs.
4168 * 0 on success, or a negative error in case of failure.
4170 * u64 bpf_get_netns_cookie(void *ctx)
4172 * Retrieve the cookie (generated by the kernel) of the network
4173 * namespace the input *ctx* is associated with. The network
4174 * namespace cookie remains stable for its lifetime and provides
4175 * a global identifier that can be assumed unique. If *ctx* is
4176 * NULL, then the helper returns the cookie for the initial
4177 * network namespace. The cookie itself is very similar to that
4178 * of **bpf_get_socket_cookie**\ () helper, but for network
4179 * namespaces instead of sockets.
4181 * A 8-byte long opaque number.
4183 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4185 * Return id of cgroup v2 that is ancestor of the cgroup associated
4186 * with the current task at the *ancestor_level*. The root cgroup
4187 * is at *ancestor_level* zero and each step down the hierarchy
4188 * increments the level. If *ancestor_level* == level of cgroup
4189 * associated with the current task, then return value will be the
4190 * same as that of **bpf_get_current_cgroup_id**\ ().
4192 * The helper is useful to implement policies based on cgroups
4193 * that are upper in hierarchy than immediate cgroup associated
4194 * with the current task.
4196 * The format of returned id and helper limitations are same as in
4197 * **bpf_get_current_cgroup_id**\ ().
4199 * The id is returned or 0 in case the id could not be retrieved.
4201 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4203 * Helper is overloaded depending on BPF program type. This
4204 * description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4205 * **BPF_PROG_TYPE_SCHED_ACT** programs.
4207 * Assign the *sk* to the *skb*. When combined with appropriate
4208 * routing configuration to receive the packet towards the socket,
4209 * will cause *skb* to be delivered to the specified socket.
4210 * Subsequent redirection of *skb* via **bpf_redirect**\ (),
4211 * **bpf_clone_redirect**\ () or other methods outside of BPF may
4212 * interfere with successful delivery to the socket.
4214 * This operation is only valid from TC ingress path.
4216 * The *flags* argument must be zero.
4218 * 0 on success, or a negative error in case of failure:
4220 * **-EINVAL** if specified *flags* are not supported.
4222 * **-ENOENT** if the socket is unavailable for assignment.
4224 * **-ENETUNREACH** if the socket is unreachable (wrong netns).
4226 * **-EOPNOTSUPP** if the operation is not supported, for example
4227 * a call from outside of TC ingress.
4229 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4231 * Helper is overloaded depending on BPF program type. This
4232 * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4234 * Select the *sk* as a result of a socket lookup.
4236 * For the operation to succeed passed socket must be compatible
4237 * with the packet description provided by the *ctx* object.
4239 * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4240 * be an exact match. While IP family (**AF_INET** or
4241 * **AF_INET6**) must be compatible, that is IPv6 sockets
4242 * that are not v6-only can be selected for IPv4 packets.
4244 * Only TCP listeners and UDP unconnected sockets can be
4245 * selected. *sk* can also be NULL to reset any previous
4248 * *flags* argument can combination of following values:
4250 * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4251 * socket selection, potentially done by a BPF program
4252 * that ran before us.
4254 * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4255 * load-balancing within reuseport group for the socket
4258 * On success *ctx->sk* will point to the selected socket.
4261 * 0 on success, or a negative errno in case of failure.
4263 * * **-EAFNOSUPPORT** if socket family (*sk->family*) is
4264 * not compatible with packet family (*ctx->family*).
4266 * * **-EEXIST** if socket has been already selected,
4267 * potentially by another program, and
4268 * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4270 * * **-EINVAL** if unsupported flags were specified.
4272 * * **-EPROTOTYPE** if socket L4 protocol
4273 * (*sk->protocol*) doesn't match packet protocol
4274 * (*ctx->protocol*).
4276 * * **-ESOCKTNOSUPPORT** if socket is not in allowed
4277 * state (TCP listening or UDP unconnected).
4279 * u64 bpf_ktime_get_boot_ns(void)
4281 * Return the time elapsed since system boot, in nanoseconds.
4282 * Does include the time the system was suspended.
4283 * See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4287 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4289 * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4290 * out the format string.
4291 * The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4292 * the format string itself. The *data* and *data_len* are format string
4293 * arguments. The *data* are a **u64** array and corresponding format string
4294 * values are stored in the array. For strings and pointers where pointees
4295 * are accessed, only the pointer values are stored in the *data* array.
4296 * The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4298 * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4299 * Reading kernel memory may fail due to either invalid address or
4300 * valid address but requiring a major memory fault. If reading kernel memory
4301 * fails, the string for **%s** will be an empty string, and the ip
4302 * address for **%p{i,I}{4,6}** will be 0. Not returning error to
4303 * bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4305 * 0 on success, or a negative error in case of failure:
4307 * **-EBUSY** if per-CPU memory copy buffer is busy, can try again
4308 * by returning 1 from bpf program.
4310 * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4312 * **-E2BIG** if *fmt* contains too many format specifiers.
4314 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4316 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4318 * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4319 * The *m* represents the seq_file. The *data* and *len* represent the
4320 * data to write in bytes.
4322 * 0 on success, or a negative error in case of failure:
4324 * **-EOVERFLOW** if an overflow happened: The same object will be tried again.
4326 * u64 bpf_sk_cgroup_id(void *sk)
4328 * Return the cgroup v2 id of the socket *sk*.
4330 * *sk* must be a non-**NULL** pointer to a socket, e.g. one
4331 * returned from **bpf_sk_lookup_xxx**\ (),
4332 * **bpf_sk_fullsock**\ (), etc. The format of returned id is
4333 * same as in **bpf_skb_cgroup_id**\ ().
4335 * This helper is available only if the kernel was compiled with
4336 * the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4338 * The id is returned or 0 in case the id could not be retrieved.
4340 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4342 * Return id of cgroup v2 that is ancestor of cgroup associated
4343 * with the *sk* at the *ancestor_level*. The root cgroup is at
4344 * *ancestor_level* zero and each step down the hierarchy
4345 * increments the level. If *ancestor_level* == level of cgroup
4346 * associated with *sk*, then return value will be same as that
4347 * of **bpf_sk_cgroup_id**\ ().
4349 * The helper is useful to implement policies based on cgroups
4350 * that are upper in hierarchy than immediate cgroup associated
4353 * The format of returned id and helper limitations are same as in
4354 * **bpf_sk_cgroup_id**\ ().
4356 * The id is returned or 0 in case the id could not be retrieved.
4358 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4360 * Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4361 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4362 * of new data availability is sent.
4363 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4364 * of new data availability is sent unconditionally.
4365 * If **0** is specified in *flags*, an adaptive notification
4366 * of new data availability is sent.
4368 * An adaptive notification is a notification sent whenever the user-space
4369 * process has caught up and consumed all available payloads. In case the user-space
4370 * process is still processing a previous payload, then no notification is needed
4371 * as it will process the newly added payload automatically.
4373 * 0 on success, or a negative error in case of failure.
4375 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4377 * Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4378 * *flags* must be 0.
4380 * Valid pointer with *size* bytes of memory available; NULL,
4383 * void bpf_ringbuf_submit(void *data, u64 flags)
4385 * Submit reserved ring buffer sample, pointed to by *data*.
4386 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4387 * of new data availability is sent.
4388 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4389 * of new data availability is sent unconditionally.
4390 * If **0** is specified in *flags*, an adaptive notification
4391 * of new data availability is sent.
4393 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4395 * Nothing. Always succeeds.
4397 * void bpf_ringbuf_discard(void *data, u64 flags)
4399 * Discard reserved ring buffer sample, pointed to by *data*.
4400 * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4401 * of new data availability is sent.
4402 * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4403 * of new data availability is sent unconditionally.
4404 * If **0** is specified in *flags*, an adaptive notification
4405 * of new data availability is sent.
4407 * See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4409 * Nothing. Always succeeds.
4411 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4413 * Query various characteristics of provided ring buffer. What
4414 * exactly is queries is determined by *flags*:
4416 * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4417 * * **BPF_RB_RING_SIZE**: The size of ring buffer.
4418 * * **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4419 * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4421 * Data returned is just a momentary snapshot of actual values
4422 * and could be inaccurate, so this facility should be used to
4423 * power heuristics and for reporting, not to make 100% correct
4426 * Requested value, or 0, if *flags* are not recognized.
4428 * long bpf_csum_level(struct sk_buff *skb, u64 level)
4430 * Change the skbs checksum level by one layer up or down, or
4431 * reset it entirely to none in order to have the stack perform
4432 * checksum validation. The level is applicable to the following
4433 * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4434 * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4435 * through **bpf_skb_adjust_room**\ () helper with passing in
4436 * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call
4437 * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4438 * the UDP header is removed. Similarly, an encap of the latter
4439 * into the former could be accompanied by a helper call to
4440 * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4441 * skb is still intended to be processed in higher layers of the
4442 * stack instead of just egressing at tc.
4444 * There are three supported level settings at this time:
4446 * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4447 * with CHECKSUM_UNNECESSARY.
4448 * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4449 * with CHECKSUM_UNNECESSARY.
4450 * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4451 * sets CHECKSUM_NONE to force checksum validation by the stack.
4452 * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4455 * 0 on success, or a negative error in case of failure. In the
4456 * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4457 * is returned or the error code -EACCES in case the skb is not
4458 * subject to CHECKSUM_UNNECESSARY.
4460 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4462 * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4464 * *sk* if casting is valid, or **NULL** otherwise.
4466 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4468 * Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4470 * *sk* if casting is valid, or **NULL** otherwise.
4472 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4474 * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4476 * *sk* if casting is valid, or **NULL** otherwise.
4478 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4480 * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4482 * *sk* if casting is valid, or **NULL** otherwise.
4484 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4486 * Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4488 * *sk* if casting is valid, or **NULL** otherwise.
4490 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4492 * Return a user or a kernel stack in bpf program provided buffer.
4493 * To achieve this, the helper needs *task*, which is a valid
4494 * pointer to **struct task_struct**. To store the stacktrace, the
4495 * bpf program provides *buf* with a nonnegative *size*.
4497 * The last argument, *flags*, holds the number of stack frames to
4498 * skip (from 0 to 255), masked with
4499 * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4500 * the following flags:
4502 * **BPF_F_USER_STACK**
4503 * Collect a user space stack instead of a kernel stack.
4504 * **BPF_F_USER_BUILD_ID**
4505 * Collect buildid+offset instead of ips for user stack,
4506 * only valid if **BPF_F_USER_STACK** is also specified.
4508 * **bpf_get_task_stack**\ () can collect up to
4509 * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4510 * to sufficient large buffer size. Note that
4511 * this limit can be controlled with the **sysctl** program, and
4512 * that it should be manually increased in order to profile long
4513 * user stacks (such as stacks for Java programs). To do so, use:
4517 * # sysctl kernel.perf_event_max_stack=<new value>
4519 * The non-negative copied *buf* length equal to or less than
4520 * *size* on success, or a negative error in case of failure.
4522 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4524 * Load header option. Support reading a particular TCP header
4525 * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4527 * If *flags* is 0, it will search the option from the
4528 * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops**
4529 * has details on what skb_data contains under different
4530 * *skops*\ **->op**.
4532 * The first byte of the *searchby_res* specifies the
4533 * kind that it wants to search.
4535 * If the searching kind is an experimental kind
4536 * (i.e. 253 or 254 according to RFC6994). It also
4537 * needs to specify the "magic" which is either
4538 * 2 bytes or 4 bytes. It then also needs to
4539 * specify the size of the magic by using
4540 * the 2nd byte which is "kind-length" of a TCP
4541 * header option and the "kind-length" also
4542 * includes the first 2 bytes "kind" and "kind-length"
4543 * itself as a normal TCP header option also does.
4545 * For example, to search experimental kind 254 with
4546 * 2 byte magic 0xeB9F, the searchby_res should be
4547 * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4549 * To search for the standard window scale option (3),
4550 * the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4551 * Note, kind-length must be 0 for regular option.
4553 * Searching for No-Op (0) and End-of-Option-List (1) are
4556 * *len* must be at least 2 bytes which is the minimal size
4557 * of a header option.
4561 * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4562 * saved_syn packet or the just-received syn packet.
4565 * > 0 when found, the header option is copied to *searchby_res*.
4566 * The return value is the total length copied. On failure, a
4567 * negative error code is returned:
4569 * **-EINVAL** if a parameter is invalid.
4571 * **-ENOMSG** if the option is not found.
4573 * **-ENOENT** if no syn packet is available when
4574 * **BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4576 * **-ENOSPC** if there is not enough space. Only *len* number of
4579 * **-EFAULT** on failure to parse the header options in the
4582 * **-EPERM** if the helper cannot be used under the current
4583 * *skops*\ **->op**.
4585 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4587 * Store header option. The data will be copied
4588 * from buffer *from* with length *len* to the TCP header.
4590 * The buffer *from* should have the whole option that
4591 * includes the kind, kind-length, and the actual
4592 * option data. The *len* must be at least kind-length
4593 * long. The kind-length does not have to be 4 byte
4594 * aligned. The kernel will take care of the padding
4595 * and setting the 4 bytes aligned value to th->doff.
4597 * This helper will check for duplicated option
4598 * by searching the same option in the outgoing skb.
4600 * This helper can only be called during
4601 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4604 * 0 on success, or negative error in case of failure:
4606 * **-EINVAL** If param is invalid.
4608 * **-ENOSPC** if there is not enough space in the header.
4609 * Nothing has been written
4611 * **-EEXIST** if the option already exists.
4613 * **-EFAULT** on failure to parse the existing header options.
4615 * **-EPERM** if the helper cannot be used under the current
4616 * *skops*\ **->op**.
4618 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4620 * Reserve *len* bytes for the bpf header option. The
4621 * space will be used by **bpf_store_hdr_opt**\ () later in
4622 * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4624 * If **bpf_reserve_hdr_opt**\ () is called multiple times,
4625 * the total number of bytes will be reserved.
4627 * This helper can only be called during
4628 * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4631 * 0 on success, or negative error in case of failure:
4633 * **-EINVAL** if a parameter is invalid.
4635 * **-ENOSPC** if there is not enough space in the header.
4637 * **-EPERM** if the helper cannot be used under the current
4638 * *skops*\ **->op**.
4640 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4642 * Get a bpf_local_storage from an *inode*.
4644 * Logically, it could be thought of as getting the value from
4645 * a *map* with *inode* as the **key**. From this
4646 * perspective, the usage is not much different from
4647 * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4648 * helper enforces the key must be an inode and the map must also
4649 * be a **BPF_MAP_TYPE_INODE_STORAGE**.
4651 * Underneath, the value is stored locally at *inode* instead of
4652 * the *map*. The *map* is used as the bpf-local-storage
4653 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4654 * searched against all bpf_local_storage residing at *inode*.
4656 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4657 * used such that a new bpf_local_storage will be
4658 * created if one does not exist. *value* can be used
4659 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4660 * the initial value of a bpf_local_storage. If *value* is
4661 * **NULL**, the new bpf_local_storage will be zero initialized.
4663 * A bpf_local_storage pointer is returned on success.
4665 * **NULL** if not found or there was an error in adding
4666 * a new bpf_local_storage.
4668 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4670 * Delete a bpf_local_storage from an *inode*.
4674 * **-ENOENT** if the bpf_local_storage cannot be found.
4676 * long bpf_d_path(struct path *path, char *buf, u32 sz)
4678 * Return full path for given **struct path** object, which
4679 * needs to be the kernel BTF *path* object. The path is
4680 * returned in the provided buffer *buf* of size *sz* and
4681 * is zero terminated.
4684 * On success, the strictly positive length of the string,
4685 * including the trailing NUL character. On error, a negative
4688 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4690 * Read *size* bytes from user space address *user_ptr* and store
4691 * the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4693 * 0 on success, or a negative error in case of failure.
4695 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4697 * Use BTF to store a string representation of *ptr*->ptr in *str*,
4698 * using *ptr*->type_id. This value should specify the type
4699 * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4700 * can be used to look up vmlinux BTF type ids. Traversing the
4701 * data structure using BTF, the type information and values are
4702 * stored in the first *str_size* - 1 bytes of *str*. Safe copy of
4703 * the pointer data is carried out to avoid kernel crashes during
4704 * operation. Smaller types can use string space on the stack;
4705 * larger programs can use map data to store the string
4708 * The string can be subsequently shared with userspace via
4709 * bpf_perf_event_output() or ring buffer interfaces.
4710 * bpf_trace_printk() is to be avoided as it places too small
4711 * a limit on string size to be useful.
4713 * *flags* is a combination of
4716 * no formatting around type information
4718 * no struct/union member names/types
4720 * show raw (unobfuscated) pointer values;
4721 * equivalent to printk specifier %px.
4723 * show zero-valued struct/union members; they
4724 * are not displayed by default
4727 * The number of bytes that were written (or would have been
4728 * written if output had to be truncated due to string size),
4729 * or a negative error in cases of failure.
4731 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4733 * Use BTF to write to seq_write a string representation of
4734 * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4735 * *flags* are identical to those used for bpf_snprintf_btf.
4737 * 0 on success or a negative error in case of failure.
4739 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4741 * See **bpf_get_cgroup_classid**\ () for the main description.
4742 * This helper differs from **bpf_get_cgroup_classid**\ () in that
4743 * the cgroup v1 net_cls class is retrieved only from the *skb*'s
4744 * associated socket instead of the current process.
4746 * The id is returned or 0 in case the id could not be retrieved.
4748 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4750 * Redirect the packet to another net device of index *ifindex*
4751 * and fill in L2 addresses from neighboring subsystem. This helper
4752 * is somewhat similar to **bpf_redirect**\ (), except that it
4753 * populates L2 addresses as well, meaning, internally, the helper
4754 * relies on the neighbor lookup for the L2 address of the nexthop.
4756 * The helper will perform a FIB lookup based on the skb's
4757 * networking header to get the address of the next hop, unless
4758 * this is supplied by the caller in the *params* argument. The
4759 * *plen* argument indicates the len of *params* and should be set
4760 * to 0 if *params* is NULL.
4762 * The *flags* argument is reserved and must be 0. The helper is
4763 * currently only supported for tc BPF program types, and enabled
4764 * for IPv4 and IPv6 protocols.
4766 * The helper returns **TC_ACT_REDIRECT** on success or
4767 * **TC_ACT_SHOT** on error.
4769 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4771 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4772 * pointer to the percpu kernel variable on *cpu*. A ksym is an
4773 * extern variable decorated with '__ksym'. For ksym, there is a
4774 * global var (either static or global) defined of the same name
4775 * in the kernel. The ksym is percpu if the global var is percpu.
4776 * The returned pointer points to the global percpu var on *cpu*.
4778 * bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4779 * kernel, except that bpf_per_cpu_ptr() may return NULL. This
4780 * happens if *cpu* is larger than nr_cpu_ids. The caller of
4781 * bpf_per_cpu_ptr() must check the returned value.
4783 * A pointer pointing to the kernel percpu variable on *cpu*, or
4784 * NULL, if *cpu* is invalid.
4786 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4788 * Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4789 * pointer to the percpu kernel variable on this cpu. See the
4790 * description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4792 * bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4793 * the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4794 * never return NULL.
4796 * A pointer pointing to the kernel percpu variable on this cpu.
4798 * long bpf_redirect_peer(u32 ifindex, u64 flags)
4800 * Redirect the packet to another net device of index *ifindex*.
4801 * This helper is somewhat similar to **bpf_redirect**\ (), except
4802 * that the redirection happens to the *ifindex*' peer device and
4803 * the netns switch takes place from ingress to ingress without
4804 * going through the CPU's backlog queue.
4806 * The *flags* argument is reserved and must be 0. The helper is
4807 * currently only supported for tc BPF program types at the ingress
4808 * hook and for veth device types. The peer device must reside in a
4809 * different network namespace.
4811 * The helper returns **TC_ACT_REDIRECT** on success or
4812 * **TC_ACT_SHOT** on error.
4814 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4816 * Get a bpf_local_storage from the *task*.
4818 * Logically, it could be thought of as getting the value from
4819 * a *map* with *task* as the **key**. From this
4820 * perspective, the usage is not much different from
4821 * **bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4822 * helper enforces the key must be a task_struct and the map must also
4823 * be a **BPF_MAP_TYPE_TASK_STORAGE**.
4825 * Underneath, the value is stored locally at *task* instead of
4826 * the *map*. The *map* is used as the bpf-local-storage
4827 * "type". The bpf-local-storage "type" (i.e. the *map*) is
4828 * searched against all bpf_local_storage residing at *task*.
4830 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4831 * used such that a new bpf_local_storage will be
4832 * created if one does not exist. *value* can be used
4833 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4834 * the initial value of a bpf_local_storage. If *value* is
4835 * **NULL**, the new bpf_local_storage will be zero initialized.
4837 * A bpf_local_storage pointer is returned on success.
4839 * **NULL** if not found or there was an error in adding
4840 * a new bpf_local_storage.
4842 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4844 * Delete a bpf_local_storage from a *task*.
4848 * **-ENOENT** if the bpf_local_storage cannot be found.
4850 * struct task_struct *bpf_get_current_task_btf(void)
4852 * Return a BTF pointer to the "current" task.
4853 * This pointer can also be used in helpers that accept an
4854 * *ARG_PTR_TO_BTF_ID* of type *task_struct*.
4856 * Pointer to the current task.
4858 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4860 * Set or clear certain options on *bprm*:
4862 * **BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4863 * which sets the **AT_SECURE** auxv for glibc. The bit
4864 * is cleared if the flag is not specified.
4866 * **-EINVAL** if invalid *flags* are passed, zero otherwise.
4868 * u64 bpf_ktime_get_coarse_ns(void)
4870 * Return a coarse-grained version of the time elapsed since
4871 * system boot, in nanoseconds. Does not include time the system
4874 * See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4878 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4880 * Returns the stored IMA hash of the *inode* (if it's available).
4881 * If the hash is larger than *size*, then only *size*
4882 * bytes will be copied to *dst*
4884 * The **hash_algo** is returned on success,
4885 * **-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4886 * invalid arguments are passed.
4888 * struct socket *bpf_sock_from_file(struct file *file)
4890 * If the given file represents a socket, returns the associated
4893 * A pointer to a struct socket on success or NULL if the file is
4896 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4898 * Check packet size against exceeding MTU of net device (based
4899 * on *ifindex*). This helper will likely be used in combination
4900 * with helpers that adjust/change the packet size.
4902 * The argument *len_diff* can be used for querying with a planned
4903 * size change. This allows to check MTU prior to changing packet
4904 * ctx. Providing a *len_diff* adjustment that is larger than the
4905 * actual packet size (resulting in negative packet size) will in
4906 * principle not exceed the MTU, which is why it is not considered
4907 * a failure. Other BPF helpers are needed for performing the
4908 * planned size change; therefore the responsibility for catching
4909 * a negative packet size belongs in those helpers.
4911 * Specifying *ifindex* zero means the MTU check is performed
4912 * against the current net device. This is practical if this isn't
4913 * used prior to redirect.
4915 * On input *mtu_len* must be a valid pointer, else verifier will
4916 * reject BPF program. If the value *mtu_len* is initialized to
4917 * zero then the ctx packet size is use. When value *mtu_len* is
4918 * provided as input this specify the L3 length that the MTU check
4919 * is done against. Remember XDP and TC length operate at L2, but
4920 * this value is L3 as this correlate to MTU and IP-header tot_len
4921 * values which are L3 (similar behavior as bpf_fib_lookup).
4923 * The Linux kernel route table can configure MTUs on a more
4924 * specific per route level, which is not provided by this helper.
4925 * For route level MTU checks use the **bpf_fib_lookup**\ ()
4928 * *ctx* is either **struct xdp_md** for XDP programs or
4929 * **struct sk_buff** for tc cls_act programs.
4931 * The *flags* argument can be a combination of one or more of the
4934 * **BPF_MTU_CHK_SEGS**
4935 * This flag will only works for *ctx* **struct sk_buff**.
4936 * If packet context contains extra packet segment buffers
4937 * (often knows as GSO skb), then MTU check is harder to
4938 * check at this point, because in transmit path it is
4939 * possible for the skb packet to get re-segmented
4940 * (depending on net device features). This could still be
4941 * a MTU violation, so this flag enables performing MTU
4942 * check against segments, with a different violation
4943 * return code to tell it apart. Check cannot use len_diff.
4945 * On return *mtu_len* pointer contains the MTU value of the net
4946 * device. Remember the net device configured MTU is the L3 size,
4947 * which is returned here and XDP and TC length operate at L2.
4948 * Helper take this into account for you, but remember when using
4949 * MTU value in your BPF-code.
4952 * * 0 on success, and populate MTU value in *mtu_len* pointer.
4954 * * < 0 if any input argument is invalid (*mtu_len* not updated)
4956 * MTU violations return positive values, but also populate MTU
4957 * value in *mtu_len* pointer, as this can be needed for
4958 * implementing PMTU handing:
4960 * * **BPF_MTU_CHK_RET_FRAG_NEEDED**
4961 * * **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4963 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4965 * For each element in **map**, call **callback_fn** function with
4966 * **map**, **callback_ctx** and other map-specific parameters.
4967 * The **callback_fn** should be a static function and
4968 * the **callback_ctx** should be a pointer to the stack.
4969 * The **flags** is used to control certain aspects of the helper.
4970 * Currently, the **flags** must be 0.
4972 * The following are a list of supported map types and their
4973 * respective expected callback signatures:
4975 * BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4976 * BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4977 * BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4979 * long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4981 * For per_cpu maps, the map_value is the value on the cpu where the
4982 * bpf_prog is running.
4984 * If **callback_fn** return 0, the helper will continue to the next
4985 * element. If return value is 1, the helper will skip the rest of
4986 * elements and return. Other return values are not used now.
4989 * The number of traversed map elements for success, **-EINVAL** for
4990 * invalid **flags**.
4992 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4994 * Outputs a string into the **str** buffer of size **str_size**
4995 * based on a format string stored in a read-only map pointed by
4998 * Each format specifier in **fmt** corresponds to one u64 element
4999 * in the **data** array. For strings and pointers where pointees
5000 * are accessed, only the pointer values are stored in the *data*
5001 * array. The *data_len* is the size of *data* in bytes - must be
5004 * Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5005 * memory. Reading kernel memory may fail due to either invalid
5006 * address or valid address but requiring a major memory fault. If
5007 * reading kernel memory fails, the string for **%s** will be an
5008 * empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5009 * Not returning error to bpf program is consistent with what
5010 * **bpf_trace_printk**\ () does for now.
5013 * The strictly positive length of the formatted string, including
5014 * the trailing zero character. If the return value is greater than
5015 * **str_size**, **str** contains a truncated string, guaranteed to
5016 * be zero-terminated except when **str_size** is 0.
5018 * Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5020 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5022 * Execute bpf syscall with given arguments.
5026 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5028 * Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5030 * Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5032 * long bpf_sys_close(u32 fd)
5034 * Execute close syscall for given FD.
5038 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5040 * Initialize the timer.
5041 * First 4 bits of *flags* specify clockid.
5042 * Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5043 * All other bits of *flags* are reserved.
5044 * The verifier will reject the program if *timer* is not from
5048 * **-EBUSY** if *timer* is already initialized.
5049 * **-EINVAL** if invalid *flags* are passed.
5050 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5051 * The user space should either hold a file descriptor to a map with timers
5052 * or pin such map in bpffs. When map is unpinned or file descriptor is
5053 * closed all timers in the map will be cancelled and freed.
5055 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5057 * Configure the timer to call *callback_fn* static function.
5060 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5061 * **-EPERM** if *timer* is in a map that doesn't have any user references.
5062 * The user space should either hold a file descriptor to a map with timers
5063 * or pin such map in bpffs. When map is unpinned or file descriptor is
5064 * closed all timers in the map will be cancelled and freed.
5066 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5068 * Set timer expiration N nanoseconds from the current time. The
5069 * configured callback will be invoked in soft irq context on some cpu
5070 * and will not repeat unless another bpf_timer_start() is made.
5071 * In such case the next invocation can migrate to a different cpu.
5072 * Since struct bpf_timer is a field inside map element the map
5073 * owns the timer. The bpf_timer_set_callback() will increment refcnt
5074 * of BPF program to make sure that callback_fn code stays valid.
5075 * When user space reference to a map reaches zero all timers
5076 * in a map are cancelled and corresponding program's refcnts are
5077 * decremented. This is done to make sure that Ctrl-C of a user
5078 * process doesn't leave any timers running. If map is pinned in
5079 * bpffs the callback_fn can re-arm itself indefinitely.
5080 * bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5081 * cancel and free the timer in the given map element.
5082 * The map can contain timers that invoke callback_fn-s from different
5083 * programs. The same callback_fn can serve different timers from
5084 * different maps if key/value layout matches across maps.
5085 * Every bpf_timer_set_callback() can have different callback_fn.
5087 * *flags* can be one of:
5089 * **BPF_F_TIMER_ABS**
5090 * Start the timer in absolute expire value instead of the
5091 * default relative one.
5095 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5096 * or invalid *flags* are passed.
5098 * long bpf_timer_cancel(struct bpf_timer *timer)
5100 * Cancel the timer and wait for callback_fn to finish if it was running.
5102 * 0 if the timer was not active.
5103 * 1 if the timer was active.
5104 * **-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5105 * **-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5106 * own timer which would have led to a deadlock otherwise.
5108 * u64 bpf_get_func_ip(void *ctx)
5110 * Get address of the traced function (for tracing and kprobe programs).
5112 * When called for kprobe program attached as uprobe it returns
5113 * probe address for both entry and return uprobe.
5116 * Address of the traced function for kprobe.
5117 * 0 for kprobes placed within the function (not at the entry).
5118 * Address of the probe for uprobe and return uprobe.
5120 * u64 bpf_get_attach_cookie(void *ctx)
5122 * Get bpf_cookie value provided (optionally) during the program
5123 * attachment. It might be different for each individual
5124 * attachment, even if BPF program itself is the same.
5125 * Expects BPF program context *ctx* as a first argument.
5127 * Supported for the following program types:
5132 * Value specified by user at BPF link creation/attachment time
5133 * or 0, if it was not specified.
5135 * long bpf_task_pt_regs(struct task_struct *task)
5137 * Get the struct pt_regs associated with **task**.
5139 * A pointer to struct pt_regs.
5141 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5143 * Get branch trace from hardware engines like Intel LBR. The
5144 * hardware engine is stopped shortly after the helper is
5145 * called. Therefore, the user need to filter branch entries
5146 * based on the actual use case. To capture branch trace
5147 * before the trigger point of the BPF program, the helper
5148 * should be called at the beginning of the BPF program.
5150 * The data is stored as struct perf_branch_entry into output
5151 * buffer *entries*. *size* is the size of *entries* in bytes.
5152 * *flags* is reserved for now and must be zero.
5155 * On success, number of bytes written to *buf*. On error, a
5158 * **-EINVAL** if *flags* is not zero.
5160 * **-ENOENT** if architecture does not support branch records.
5162 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5164 * Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5165 * to format and can handle more format args as a result.
5167 * Arguments are to be used as in **bpf_seq_printf**\ () helper.
5169 * The number of bytes written to the buffer, or a negative error
5170 * in case of failure.
5172 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5174 * Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5176 * *sk* if casting is valid, or **NULL** otherwise.
5178 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5180 * Get the address of a kernel symbol, returned in *res*. *res* is
5181 * set to 0 if the symbol is not found.
5183 * On success, zero. On error, a negative value.
5185 * **-EINVAL** if *flags* is not zero.
5187 * **-EINVAL** if string *name* is not the same size as *name_sz*.
5189 * **-ENOENT** if symbol is not found.
5191 * **-EPERM** if caller does not have permission to obtain kernel address.
5193 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5195 * Find vma of *task* that contains *addr*, call *callback_fn*
5196 * function with *task*, *vma*, and *callback_ctx*.
5197 * The *callback_fn* should be a static function and
5198 * the *callback_ctx* should be a pointer to the stack.
5199 * The *flags* is used to control certain aspects of the helper.
5200 * Currently, the *flags* must be 0.
5202 * The expected callback signature is
5204 * long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5208 * **-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5209 * **-EBUSY** if failed to try lock mmap_lock.
5210 * **-EINVAL** for invalid **flags**.
5212 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5214 * For **nr_loops**, call **callback_fn** function
5215 * with **callback_ctx** as the context parameter.
5216 * The **callback_fn** should be a static function and
5217 * the **callback_ctx** should be a pointer to the stack.
5218 * The **flags** is used to control certain aspects of the helper.
5219 * Currently, the **flags** must be 0. Currently, nr_loops is
5220 * limited to 1 << 23 (~8 million) loops.
5222 * long (\*callback_fn)(u32 index, void \*ctx);
5224 * where **index** is the current index in the loop. The index
5227 * If **callback_fn** returns 0, the helper will continue to the next
5228 * loop. If return value is 1, the helper will skip the rest of
5229 * the loops and return. Other return values are not used now,
5230 * and will be rejected by the verifier.
5233 * The number of loops performed, **-EINVAL** for invalid **flags**,
5234 * **-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5236 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5238 * Do strncmp() between **s1** and **s2**. **s1** doesn't need
5239 * to be null-terminated and **s1_sz** is the maximum storage
5240 * size of **s1**. **s2** must be a read-only string.
5242 * An integer less than, equal to, or greater than zero
5243 * if the first **s1_sz** bytes of **s1** is found to be
5244 * less than, to match, or be greater than **s2**.
5246 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5248 * Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5249 * returned in **value**.
5253 * **-EINVAL** if n >= argument register count of traced function.
5255 * long bpf_get_func_ret(void *ctx, u64 *value)
5257 * Get return value of the traced function (for tracing programs)
5262 * **-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5264 * long bpf_get_func_arg_cnt(void *ctx)
5266 * Get number of registers of the traced function (for tracing programs) where
5267 * function arguments are stored in these registers.
5270 * The number of argument registers of the traced function.
5272 * int bpf_get_retval(void)
5274 * Get the BPF program's return value that will be returned to the upper layers.
5276 * This helper is currently supported by cgroup programs and only by the hooks
5277 * where BPF program's return value is returned to the userspace via errno.
5279 * The BPF program's return value.
5281 * int bpf_set_retval(int retval)
5283 * Set the BPF program's return value that will be returned to the upper layers.
5285 * This helper is currently supported by cgroup programs and only by the hooks
5286 * where BPF program's return value is returned to the userspace via errno.
5288 * Note that there is the following corner case where the program exports an error
5289 * via bpf_set_retval but signals success via 'return 1':
5291 * bpf_set_retval(-EPERM);
5294 * In this case, the BPF program's return value will use helper's -EPERM. This
5295 * still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5298 * 0 on success, or a negative error in case of failure.
5300 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5302 * Get the total size of a given xdp buff (linear and paged area)
5304 * The total size of a given xdp buffer.
5306 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5308 * This helper is provided as an easy way to load data from a
5309 * xdp buffer. It can be used to load *len* bytes from *offset* from
5310 * the frame associated to *xdp_md*, into the buffer pointed by
5313 * 0 on success, or a negative error in case of failure.
5315 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5317 * Store *len* bytes from buffer *buf* into the frame
5318 * associated to *xdp_md*, at *offset*.
5320 * 0 on success, or a negative error in case of failure.
5322 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5324 * Read *size* bytes from user space address *user_ptr* in *tsk*'s
5325 * address space, and stores the data in *dst*. *flags* is not
5326 * used yet and is provided for future extensibility. This helper
5327 * can only be used by sleepable programs.
5329 * 0 on success, or a negative error in case of failure. On error
5330 * *dst* buffer is zeroed out.
5332 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5334 * Change the __sk_buff->tstamp_type to *tstamp_type*
5335 * and set *tstamp* to the __sk_buff->tstamp together.
5337 * If there is no need to change the __sk_buff->tstamp_type,
5338 * the tstamp value can be directly written to __sk_buff->tstamp
5341 * BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5342 * will be kept during bpf_redirect_*(). A non zero
5343 * *tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5346 * A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5347 * with a zero *tstamp*.
5349 * Only IPv4 and IPv6 skb->protocol are supported.
5351 * This function is most useful when it needs to set a
5352 * mono delivery time to __sk_buff->tstamp and then
5353 * bpf_redirect_*() to the egress of an iface. For example,
5354 * changing the (rcv) timestamp in __sk_buff->tstamp at
5355 * ingress to a mono delivery time and then bpf_redirect_*()
5356 * to sch_fq@phy-dev.
5359 * **-EINVAL** for invalid input
5360 * **-EOPNOTSUPP** for unsupported protocol
5362 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5364 * Returns a calculated IMA hash of the *file*.
5365 * If the hash is larger than *size*, then only *size*
5366 * bytes will be copied to *dst*
5368 * The **hash_algo** is returned on success,
5369 * **-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
5370 * invalid arguments are passed.
5372 * void *bpf_kptr_xchg(void *map_value, void *ptr)
5374 * Exchange kptr at pointer *map_value* with *ptr*, and return the
5375 * old value. *ptr* can be NULL, otherwise it must be a referenced
5376 * pointer which will be released when this helper is called.
5378 * The old value of kptr (which can be NULL). The returned pointer
5379 * if not NULL, is a reference which must be released using its
5380 * corresponding release function, or moved into a BPF map before
5383 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5385 * Perform a lookup in *percpu map* for an entry associated to
5388 * Map value associated to *key* on *cpu*, or **NULL** if no entry
5389 * was found or *cpu* is invalid.
5391 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5393 * Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5395 * *sk* if casting is valid, or **NULL** otherwise.
5397 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5399 * Get a dynptr to local memory *data*.
5401 * *data* must be a ptr to a map value.
5402 * The maximum *size* supported is DYNPTR_MAX_SIZE.
5403 * *flags* is currently unused.
5405 * 0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5406 * -EINVAL if flags is not 0.
5408 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5410 * Reserve *size* bytes of payload in a ring buffer *ringbuf*
5411 * through the dynptr interface. *flags* must be 0.
5413 * Please note that a corresponding bpf_ringbuf_submit_dynptr or
5414 * bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5415 * reservation fails. This is enforced by the verifier.
5417 * 0 on success, or a negative error in case of failure.
5419 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5421 * Submit reserved ring buffer sample, pointed to by *data*,
5422 * through the dynptr interface. This is a no-op if the dynptr is
5425 * For more information on *flags*, please see
5426 * 'bpf_ringbuf_submit'.
5428 * Nothing. Always succeeds.
5430 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5432 * Discard reserved ring buffer sample through the dynptr
5433 * interface. This is a no-op if the dynptr is invalid/null.
5435 * For more information on *flags*, please see
5436 * 'bpf_ringbuf_discard'.
5438 * Nothing. Always succeeds.
5440 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
5442 * Read *len* bytes from *src* into *dst*, starting from *offset*
5444 * *flags* is currently unused.
5446 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5447 * of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5450 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5452 * Write *len* bytes from *src* into *dst*, starting from *offset*
5455 * *flags* must be 0 except for skb-type dynptrs.
5457 * For skb-type dynptrs:
5458 * * All data slices of the dynptr are automatically
5459 * invalidated after **bpf_dynptr_write**\ (). This is
5460 * because writing may pull the skb and change the
5461 * underlying packet buffer.
5463 * * For *flags*, please see the flags accepted by
5464 * **bpf_skb_store_bytes**\ ().
5466 * 0 on success, -E2BIG if *offset* + *len* exceeds the length
5467 * of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5468 * is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5469 * other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5471 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
5473 * Get a pointer to the underlying dynptr data.
5475 * *len* must be a statically known value. The returned data slice
5476 * is invalidated whenever the dynptr is invalidated.
5478 * skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5479 * instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5481 * Pointer to the underlying dynptr data, NULL if the dynptr is
5482 * read-only, if the dynptr is invalid, or if the offset and length
5485 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5487 * Try to issue a SYN cookie for the packet with corresponding
5488 * IPv4/TCP headers, *iph* and *th*, without depending on a
5491 * *iph* points to the IPv4 header.
5493 * *th* points to the start of the TCP header, while *th_len*
5494 * contains the length of the TCP header (at least
5495 * **sizeof**\ (**struct tcphdr**)).
5497 * On success, lower 32 bits hold the generated SYN cookie in
5498 * followed by 16 bits which hold the MSS value for that cookie,
5499 * and the top 16 bits are unused.
5501 * On failure, the returned value is one of the following:
5503 * **-EINVAL** if *th_len* is invalid.
5505 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5507 * Try to issue a SYN cookie for the packet with corresponding
5508 * IPv6/TCP headers, *iph* and *th*, without depending on a
5511 * *iph* points to the IPv6 header.
5513 * *th* points to the start of the TCP header, while *th_len*
5514 * contains the length of the TCP header (at least
5515 * **sizeof**\ (**struct tcphdr**)).
5517 * On success, lower 32 bits hold the generated SYN cookie in
5518 * followed by 16 bits which hold the MSS value for that cookie,
5519 * and the top 16 bits are unused.
5521 * On failure, the returned value is one of the following:
5523 * **-EINVAL** if *th_len* is invalid.
5525 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5527 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5529 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5530 * without depending on a listening socket.
5532 * *iph* points to the IPv4 header.
5534 * *th* points to the TCP header.
5536 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5538 * On failure, the returned value is one of the following:
5540 * **-EACCES** if the SYN cookie is not valid.
5542 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5544 * Check whether *iph* and *th* contain a valid SYN cookie ACK
5545 * without depending on a listening socket.
5547 * *iph* points to the IPv6 header.
5549 * *th* points to the TCP header.
5551 * 0 if *iph* and *th* are a valid SYN cookie ACK.
5553 * On failure, the returned value is one of the following:
5555 * **-EACCES** if the SYN cookie is not valid.
5557 * **-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5559 * u64 bpf_ktime_get_tai_ns(void)
5561 * A nonsettable system-wide clock derived from wall-clock time but
5562 * ignoring leap seconds. This clock does not experience
5563 * discontinuities and backwards jumps caused by NTP inserting leap
5564 * seconds as CLOCK_REALTIME does.
5566 * See: **clock_gettime**\ (**CLOCK_TAI**)
5570 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5572 * Drain samples from the specified user ring buffer, and invoke
5573 * the provided callback for each such sample:
5575 * long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5577 * If **callback_fn** returns 0, the helper will continue to try
5578 * and drain the next sample, up to a maximum of
5579 * BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5580 * the helper will skip the rest of the samples and return. Other
5581 * return values are not used now, and will be rejected by the
5584 * The number of drained samples if no error was encountered while
5585 * draining samples, or 0 if no samples were present in the ring
5586 * buffer. If a user-space producer was epoll-waiting on this map,
5587 * and at least one sample was drained, they will receive an event
5588 * notification notifying them of available space in the ring
5589 * buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5590 * function, no wakeup notification will be sent. If the
5591 * BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5592 * be sent even if no sample was drained.
5594 * On failure, the returned value is one of the following:
5596 * **-EBUSY** if the ring buffer is contended, and another calling
5597 * context was concurrently draining the ring buffer.
5599 * **-EINVAL** if user-space is not properly tracking the ring
5600 * buffer due to the producer position not being aligned to 8
5601 * bytes, a sample not being aligned to 8 bytes, or the producer
5602 * position not matching the advertised length of a sample.
5604 * **-E2BIG** if user-space has tried to publish a sample which is
5605 * larger than the size of the ring buffer, or which cannot fit
5606 * within a struct bpf_dynptr.
5608 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5610 * Get a bpf_local_storage from the *cgroup*.
5612 * Logically, it could be thought of as getting the value from
5613 * a *map* with *cgroup* as the **key**. From this
5614 * perspective, the usage is not much different from
5615 * **bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5616 * helper enforces the key must be a cgroup struct and the map must also
5617 * be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5619 * In reality, the local-storage value is embedded directly inside of the
5620 * *cgroup* object itself, rather than being located in the
5621 * **BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5622 * queried for some *map* on a *cgroup* object, the kernel will perform an
5623 * O(n) iteration over all of the live local-storage values for that
5624 * *cgroup* object until the local-storage value for the *map* is found.
5626 * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5627 * used such that a new bpf_local_storage will be
5628 * created if one does not exist. *value* can be used
5629 * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5630 * the initial value of a bpf_local_storage. If *value* is
5631 * **NULL**, the new bpf_local_storage will be zero initialized.
5633 * A bpf_local_storage pointer is returned on success.
5635 * **NULL** if not found or there was an error in adding
5636 * a new bpf_local_storage.
5638 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5640 * Delete a bpf_local_storage from a *cgroup*.
5644 * **-ENOENT** if the bpf_local_storage cannot be found.
5646 #define ___BPF_FUNC_MAPPER(FN, ctx...) \
5647 FN(unspec, 0, ##ctx) \
5648 FN(map_lookup_elem, 1, ##ctx) \
5649 FN(map_update_elem, 2, ##ctx) \
5650 FN(map_delete_elem, 3, ##ctx) \
5651 FN(probe_read, 4, ##ctx) \
5652 FN(ktime_get_ns, 5, ##ctx) \
5653 FN(trace_printk, 6, ##ctx) \
5654 FN(get_prandom_u32, 7, ##ctx) \
5655 FN(get_smp_processor_id, 8, ##ctx) \
5656 FN(skb_store_bytes, 9, ##ctx) \
5657 FN(l3_csum_replace, 10, ##ctx) \
5658 FN(l4_csum_replace, 11, ##ctx) \
5659 FN(tail_call, 12, ##ctx) \
5660 FN(clone_redirect, 13, ##ctx) \
5661 FN(get_current_pid_tgid, 14, ##ctx) \
5662 FN(get_current_uid_gid, 15, ##ctx) \
5663 FN(get_current_comm, 16, ##ctx) \
5664 FN(get_cgroup_classid, 17, ##ctx) \
5665 FN(skb_vlan_push, 18, ##ctx) \
5666 FN(skb_vlan_pop, 19, ##ctx) \
5667 FN(skb_get_tunnel_key, 20, ##ctx) \
5668 FN(skb_set_tunnel_key, 21, ##ctx) \
5669 FN(perf_event_read, 22, ##ctx) \
5670 FN(redirect, 23, ##ctx) \
5671 FN(get_route_realm, 24, ##ctx) \
5672 FN(perf_event_output, 25, ##ctx) \
5673 FN(skb_load_bytes, 26, ##ctx) \
5674 FN(get_stackid, 27, ##ctx) \
5675 FN(csum_diff, 28, ##ctx) \
5676 FN(skb_get_tunnel_opt, 29, ##ctx) \
5677 FN(skb_set_tunnel_opt, 30, ##ctx) \
5678 FN(skb_change_proto, 31, ##ctx) \
5679 FN(skb_change_type, 32, ##ctx) \
5680 FN(skb_under_cgroup, 33, ##ctx) \
5681 FN(get_hash_recalc, 34, ##ctx) \
5682 FN(get_current_task, 35, ##ctx) \
5683 FN(probe_write_user, 36, ##ctx) \
5684 FN(current_task_under_cgroup, 37, ##ctx) \
5685 FN(skb_change_tail, 38, ##ctx) \
5686 FN(skb_pull_data, 39, ##ctx) \
5687 FN(csum_update, 40, ##ctx) \
5688 FN(set_hash_invalid, 41, ##ctx) \
5689 FN(get_numa_node_id, 42, ##ctx) \
5690 FN(skb_change_head, 43, ##ctx) \
5691 FN(xdp_adjust_head, 44, ##ctx) \
5692 FN(probe_read_str, 45, ##ctx) \
5693 FN(get_socket_cookie, 46, ##ctx) \
5694 FN(get_socket_uid, 47, ##ctx) \
5695 FN(set_hash, 48, ##ctx) \
5696 FN(setsockopt, 49, ##ctx) \
5697 FN(skb_adjust_room, 50, ##ctx) \
5698 FN(redirect_map, 51, ##ctx) \
5699 FN(sk_redirect_map, 52, ##ctx) \
5700 FN(sock_map_update, 53, ##ctx) \
5701 FN(xdp_adjust_meta, 54, ##ctx) \
5702 FN(perf_event_read_value, 55, ##ctx) \
5703 FN(perf_prog_read_value, 56, ##ctx) \
5704 FN(getsockopt, 57, ##ctx) \
5705 FN(override_return, 58, ##ctx) \
5706 FN(sock_ops_cb_flags_set, 59, ##ctx) \
5707 FN(msg_redirect_map, 60, ##ctx) \
5708 FN(msg_apply_bytes, 61, ##ctx) \
5709 FN(msg_cork_bytes, 62, ##ctx) \
5710 FN(msg_pull_data, 63, ##ctx) \
5711 FN(bind, 64, ##ctx) \
5712 FN(xdp_adjust_tail, 65, ##ctx) \
5713 FN(skb_get_xfrm_state, 66, ##ctx) \
5714 FN(get_stack, 67, ##ctx) \
5715 FN(skb_load_bytes_relative, 68, ##ctx) \
5716 FN(fib_lookup, 69, ##ctx) \
5717 FN(sock_hash_update, 70, ##ctx) \
5718 FN(msg_redirect_hash, 71, ##ctx) \
5719 FN(sk_redirect_hash, 72, ##ctx) \
5720 FN(lwt_push_encap, 73, ##ctx) \
5721 FN(lwt_seg6_store_bytes, 74, ##ctx) \
5722 FN(lwt_seg6_adjust_srh, 75, ##ctx) \
5723 FN(lwt_seg6_action, 76, ##ctx) \
5724 FN(rc_repeat, 77, ##ctx) \
5725 FN(rc_keydown, 78, ##ctx) \
5726 FN(skb_cgroup_id, 79, ##ctx) \
5727 FN(get_current_cgroup_id, 80, ##ctx) \
5728 FN(get_local_storage, 81, ##ctx) \
5729 FN(sk_select_reuseport, 82, ##ctx) \
5730 FN(skb_ancestor_cgroup_id, 83, ##ctx) \
5731 FN(sk_lookup_tcp, 84, ##ctx) \
5732 FN(sk_lookup_udp, 85, ##ctx) \
5733 FN(sk_release, 86, ##ctx) \
5734 FN(map_push_elem, 87, ##ctx) \
5735 FN(map_pop_elem, 88, ##ctx) \
5736 FN(map_peek_elem, 89, ##ctx) \
5737 FN(msg_push_data, 90, ##ctx) \
5738 FN(msg_pop_data, 91, ##ctx) \
5739 FN(rc_pointer_rel, 92, ##ctx) \
5740 FN(spin_lock, 93, ##ctx) \
5741 FN(spin_unlock, 94, ##ctx) \
5742 FN(sk_fullsock, 95, ##ctx) \
5743 FN(tcp_sock, 96, ##ctx) \
5744 FN(skb_ecn_set_ce, 97, ##ctx) \
5745 FN(get_listener_sock, 98, ##ctx) \
5746 FN(skc_lookup_tcp, 99, ##ctx) \
5747 FN(tcp_check_syncookie, 100, ##ctx) \
5748 FN(sysctl_get_name, 101, ##ctx) \
5749 FN(sysctl_get_current_value, 102, ##ctx) \
5750 FN(sysctl_get_new_value, 103, ##ctx) \
5751 FN(sysctl_set_new_value, 104, ##ctx) \
5752 FN(strtol, 105, ##ctx) \
5753 FN(strtoul, 106, ##ctx) \
5754 FN(sk_storage_get, 107, ##ctx) \
5755 FN(sk_storage_delete, 108, ##ctx) \
5756 FN(send_signal, 109, ##ctx) \
5757 FN(tcp_gen_syncookie, 110, ##ctx) \
5758 FN(skb_output, 111, ##ctx) \
5759 FN(probe_read_user, 112, ##ctx) \
5760 FN(probe_read_kernel, 113, ##ctx) \
5761 FN(probe_read_user_str, 114, ##ctx) \
5762 FN(probe_read_kernel_str, 115, ##ctx) \
5763 FN(tcp_send_ack, 116, ##ctx) \
5764 FN(send_signal_thread, 117, ##ctx) \
5765 FN(jiffies64, 118, ##ctx) \
5766 FN(read_branch_records, 119, ##ctx) \
5767 FN(get_ns_current_pid_tgid, 120, ##ctx) \
5768 FN(xdp_output, 121, ##ctx) \
5769 FN(get_netns_cookie, 122, ##ctx) \
5770 FN(get_current_ancestor_cgroup_id, 123, ##ctx) \
5771 FN(sk_assign, 124, ##ctx) \
5772 FN(ktime_get_boot_ns, 125, ##ctx) \
5773 FN(seq_printf, 126, ##ctx) \
5774 FN(seq_write, 127, ##ctx) \
5775 FN(sk_cgroup_id, 128, ##ctx) \
5776 FN(sk_ancestor_cgroup_id, 129, ##ctx) \
5777 FN(ringbuf_output, 130, ##ctx) \
5778 FN(ringbuf_reserve, 131, ##ctx) \
5779 FN(ringbuf_submit, 132, ##ctx) \
5780 FN(ringbuf_discard, 133, ##ctx) \
5781 FN(ringbuf_query, 134, ##ctx) \
5782 FN(csum_level, 135, ##ctx) \
5783 FN(skc_to_tcp6_sock, 136, ##ctx) \
5784 FN(skc_to_tcp_sock, 137, ##ctx) \
5785 FN(skc_to_tcp_timewait_sock, 138, ##ctx) \
5786 FN(skc_to_tcp_request_sock, 139, ##ctx) \
5787 FN(skc_to_udp6_sock, 140, ##ctx) \
5788 FN(get_task_stack, 141, ##ctx) \
5789 FN(load_hdr_opt, 142, ##ctx) \
5790 FN(store_hdr_opt, 143, ##ctx) \
5791 FN(reserve_hdr_opt, 144, ##ctx) \
5792 FN(inode_storage_get, 145, ##ctx) \
5793 FN(inode_storage_delete, 146, ##ctx) \
5794 FN(d_path, 147, ##ctx) \
5795 FN(copy_from_user, 148, ##ctx) \
5796 FN(snprintf_btf, 149, ##ctx) \
5797 FN(seq_printf_btf, 150, ##ctx) \
5798 FN(skb_cgroup_classid, 151, ##ctx) \
5799 FN(redirect_neigh, 152, ##ctx) \
5800 FN(per_cpu_ptr, 153, ##ctx) \
5801 FN(this_cpu_ptr, 154, ##ctx) \
5802 FN(redirect_peer, 155, ##ctx) \
5803 FN(task_storage_get, 156, ##ctx) \
5804 FN(task_storage_delete, 157, ##ctx) \
5805 FN(get_current_task_btf, 158, ##ctx) \
5806 FN(bprm_opts_set, 159, ##ctx) \
5807 FN(ktime_get_coarse_ns, 160, ##ctx) \
5808 FN(ima_inode_hash, 161, ##ctx) \
5809 FN(sock_from_file, 162, ##ctx) \
5810 FN(check_mtu, 163, ##ctx) \
5811 FN(for_each_map_elem, 164, ##ctx) \
5812 FN(snprintf, 165, ##ctx) \
5813 FN(sys_bpf, 166, ##ctx) \
5814 FN(btf_find_by_name_kind, 167, ##ctx) \
5815 FN(sys_close, 168, ##ctx) \
5816 FN(timer_init, 169, ##ctx) \
5817 FN(timer_set_callback, 170, ##ctx) \
5818 FN(timer_start, 171, ##ctx) \
5819 FN(timer_cancel, 172, ##ctx) \
5820 FN(get_func_ip, 173, ##ctx) \
5821 FN(get_attach_cookie, 174, ##ctx) \
5822 FN(task_pt_regs, 175, ##ctx) \
5823 FN(get_branch_snapshot, 176, ##ctx) \
5824 FN(trace_vprintk, 177, ##ctx) \
5825 FN(skc_to_unix_sock, 178, ##ctx) \
5826 FN(kallsyms_lookup_name, 179, ##ctx) \
5827 FN(find_vma, 180, ##ctx) \
5828 FN(loop, 181, ##ctx) \
5829 FN(strncmp, 182, ##ctx) \
5830 FN(get_func_arg, 183, ##ctx) \
5831 FN(get_func_ret, 184, ##ctx) \
5832 FN(get_func_arg_cnt, 185, ##ctx) \
5833 FN(get_retval, 186, ##ctx) \
5834 FN(set_retval, 187, ##ctx) \
5835 FN(xdp_get_buff_len, 188, ##ctx) \
5836 FN(xdp_load_bytes, 189, ##ctx) \
5837 FN(xdp_store_bytes, 190, ##ctx) \
5838 FN(copy_from_user_task, 191, ##ctx) \
5839 FN(skb_set_tstamp, 192, ##ctx) \
5840 FN(ima_file_hash, 193, ##ctx) \
5841 FN(kptr_xchg, 194, ##ctx) \
5842 FN(map_lookup_percpu_elem, 195, ##ctx) \
5843 FN(skc_to_mptcp_sock, 196, ##ctx) \
5844 FN(dynptr_from_mem, 197, ##ctx) \
5845 FN(ringbuf_reserve_dynptr, 198, ##ctx) \
5846 FN(ringbuf_submit_dynptr, 199, ##ctx) \
5847 FN(ringbuf_discard_dynptr, 200, ##ctx) \
5848 FN(dynptr_read, 201, ##ctx) \
5849 FN(dynptr_write, 202, ##ctx) \
5850 FN(dynptr_data, 203, ##ctx) \
5851 FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx) \
5852 FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx) \
5853 FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx) \
5854 FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx) \
5855 FN(ktime_get_tai_ns, 208, ##ctx) \
5856 FN(user_ringbuf_drain, 209, ##ctx) \
5857 FN(cgrp_storage_get, 210, ##ctx) \
5858 FN(cgrp_storage_delete, 211, ##ctx) \
5861 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
5862 * know or care about integer value that is now passed as second argument
5864 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
5865 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
5867 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5868 * function eBPF program intends to call
5870 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
5872 ___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5875 #undef __BPF_ENUM_FN
5877 /* All flags used by eBPF helper functions, placed here. */
5879 /* BPF_FUNC_skb_store_bytes flags. */
5881 BPF_F_RECOMPUTE_CSUM = (1ULL << 0),
5882 BPF_F_INVALIDATE_HASH = (1ULL << 1),
5885 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5886 * First 4 bits are for passing the header field size.
5889 BPF_F_HDR_FIELD_MASK = 0xfULL,
5892 /* BPF_FUNC_l4_csum_replace flags. */
5894 BPF_F_PSEUDO_HDR = (1ULL << 4),
5895 BPF_F_MARK_MANGLED_0 = (1ULL << 5),
5896 BPF_F_MARK_ENFORCE = (1ULL << 6),
5899 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5901 BPF_F_INGRESS = (1ULL << 0),
5904 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5906 BPF_F_TUNINFO_IPV6 = (1ULL << 0),
5909 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5911 BPF_F_SKIP_FIELD_MASK = 0xffULL,
5912 BPF_F_USER_STACK = (1ULL << 8),
5913 /* flags used by BPF_FUNC_get_stackid only. */
5914 BPF_F_FAST_STACK_CMP = (1ULL << 9),
5915 BPF_F_REUSE_STACKID = (1ULL << 10),
5916 /* flags used by BPF_FUNC_get_stack only. */
5917 BPF_F_USER_BUILD_ID = (1ULL << 11),
5920 /* BPF_FUNC_skb_set_tunnel_key flags. */
5922 BPF_F_ZERO_CSUM_TX = (1ULL << 1),
5923 BPF_F_DONT_FRAGMENT = (1ULL << 2),
5924 BPF_F_SEQ_NUMBER = (1ULL << 3),
5925 BPF_F_NO_TUNNEL_KEY = (1ULL << 4),
5928 /* BPF_FUNC_skb_get_tunnel_key flags. */
5930 BPF_F_TUNINFO_FLAGS = (1ULL << 4),
5933 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5934 * BPF_FUNC_perf_event_read_value flags.
5937 BPF_F_INDEX_MASK = 0xffffffffULL,
5938 BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK,
5939 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5940 BPF_F_CTXLEN_MASK = (0xfffffULL << 32),
5943 /* Current network namespace */
5945 BPF_F_CURRENT_NETNS = (-1L),
5948 /* BPF_FUNC_csum_level level values. */
5950 BPF_CSUM_LEVEL_QUERY,
5953 BPF_CSUM_LEVEL_RESET,
5956 /* BPF_FUNC_skb_adjust_room flags. */
5958 BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0),
5959 BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1),
5960 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2),
5961 BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3),
5962 BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4),
5963 BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5),
5964 BPF_F_ADJ_ROOM_ENCAP_L2_ETH = (1ULL << 6),
5965 BPF_F_ADJ_ROOM_DECAP_L3_IPV4 = (1ULL << 7),
5966 BPF_F_ADJ_ROOM_DECAP_L3_IPV6 = (1ULL << 8),
5970 BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff,
5971 BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56,
5974 #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
5975 BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5976 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5978 /* BPF_FUNC_sysctl_get_name flags. */
5980 BPF_F_SYSCTL_BASE_NAME = (1ULL << 0),
5983 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5985 BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0),
5986 /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5987 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5989 BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE,
5992 /* BPF_FUNC_read_branch_records flags. */
5994 BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0),
5997 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5998 * BPF_FUNC_bpf_ringbuf_output flags.
6001 BPF_RB_NO_WAKEUP = (1ULL << 0),
6002 BPF_RB_FORCE_WAKEUP = (1ULL << 1),
6005 /* BPF_FUNC_bpf_ringbuf_query flags */
6007 BPF_RB_AVAIL_DATA = 0,
6008 BPF_RB_RING_SIZE = 1,
6009 BPF_RB_CONS_POS = 2,
6010 BPF_RB_PROD_POS = 3,
6013 /* BPF ring buffer constants */
6015 BPF_RINGBUF_BUSY_BIT = (1U << 31),
6016 BPF_RINGBUF_DISCARD_BIT = (1U << 30),
6017 BPF_RINGBUF_HDR_SZ = 8,
6020 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6022 BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0),
6023 BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1),
6026 /* Mode for BPF_FUNC_skb_adjust_room helper. */
6027 enum bpf_adj_room_mode {
6032 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6033 enum bpf_hdr_start_off {
6038 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6039 enum bpf_lwt_encap_mode {
6041 BPF_LWT_ENCAP_SEG6_INLINE,
6045 /* Flags for bpf_bprm_opts_set helper */
6047 BPF_F_BPRM_SECUREEXEC = (1ULL << 0),
6050 /* Flags for bpf_redirect_map helper */
6052 BPF_F_BROADCAST = (1ULL << 3),
6053 BPF_F_EXCLUDE_INGRESS = (1ULL << 4),
6056 #define __bpf_md_ptr(type, name) \
6060 } __attribute__((aligned(8)))
6063 BPF_SKB_TSTAMP_UNSPEC,
6064 BPF_SKB_TSTAMP_DELIVERY_MONO, /* tstamp has mono delivery time */
6065 /* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
6066 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
6067 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
6071 /* user accessible mirror of in-kernel sk_buff.
6072 * new fields can only be added to the end of this structure
6078 __u32 queue_mapping;
6084 __u32 ingress_ifindex;
6094 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6096 __u32 remote_ip4; /* Stored in network byte order */
6097 __u32 local_ip4; /* Stored in network byte order */
6098 __u32 remote_ip6[4]; /* Stored in network byte order */
6099 __u32 local_ip6[4]; /* Stored in network byte order */
6100 __u32 remote_port; /* Stored in network byte order */
6101 __u32 local_port; /* stored in host byte order */
6105 __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
6109 __bpf_md_ptr(struct bpf_sock *, sk);
6112 __u32 :24; /* Padding, future use. */
6116 struct bpf_tunnel_key {
6120 __u32 remote_ipv6[4];
6125 __u16 tunnel_ext; /* compat */
6126 __be16 tunnel_flags;
6131 __u32 local_ipv6[4];
6135 /* user accessible mirror of in-kernel xfrm_state.
6136 * new fields can only be added to the end of this structure
6138 struct bpf_xfrm_state {
6140 __u32 spi; /* Stored in network byte order */
6142 __u16 ext; /* Padding, future use. */
6144 __u32 remote_ipv4; /* Stored in network byte order */
6145 __u32 remote_ipv6[4]; /* Stored in network byte order */
6149 /* Generic BPF return codes which all BPF program types may support.
6150 * The values are binary compatible with their TC_ACT_* counter-part to
6151 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6154 * XDP is handled seprately, see XDP_*.
6162 /* >127 are reserved for prog type specific return codes.
6164 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6165 * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6166 * changed and should be routed based on its new L3 header.
6167 * (This is an L3 redirect, as opposed to L2 redirect
6168 * represented by BPF_REDIRECT above).
6170 BPF_LWT_REROUTE = 128,
6171 /* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6172 * to indicate that no custom dissection was performed, and
6173 * fallback to standard dissector is requested.
6175 BPF_FLOW_DISSECTOR_CONTINUE = 129,
6185 /* IP address also allows 1 and 2 bytes access */
6188 __u32 src_port; /* host byte order */
6189 __be16 dst_port; /* network byte order */
6190 __u16 :16; /* zero padding */
6194 __s32 rx_queue_mapping;
6197 struct bpf_tcp_sock {
6198 __u32 snd_cwnd; /* Sending congestion window */
6199 __u32 srtt_us; /* smoothed round trip time << 3 in usecs */
6201 __u32 snd_ssthresh; /* Slow start size threshold */
6202 __u32 rcv_nxt; /* What we want to receive next */
6203 __u32 snd_nxt; /* Next sequence we send */
6204 __u32 snd_una; /* First byte we want an ack for */
6205 __u32 mss_cache; /* Cached effective mss, not including SACKS */
6206 __u32 ecn_flags; /* ECN status bits. */
6207 __u32 rate_delivered; /* saved rate sample: packets delivered */
6208 __u32 rate_interval_us; /* saved rate sample: time elapsed */
6209 __u32 packets_out; /* Packets which are "in flight" */
6210 __u32 retrans_out; /* Retransmitted packets out */
6211 __u32 total_retrans; /* Total retransmits for entire connection */
6212 __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn
6213 * total number of segments in.
6215 __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn
6216 * total number of data segments in.
6218 __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut
6219 * The total number of segments sent.
6221 __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut
6222 * total number of data segments sent.
6224 __u32 lost_out; /* Lost packets */
6225 __u32 sacked_out; /* SACK'd packets */
6226 __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived
6227 * sum(delta(rcv_nxt)), or how many bytes
6230 __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked
6231 * sum(delta(snd_una)), or how many bytes
6234 __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups
6235 * total number of DSACK blocks received
6237 __u32 delivered; /* Total data packets delivered incl. rexmits */
6238 __u32 delivered_ce; /* Like the above but only ECE marked packets */
6239 __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */
6242 struct bpf_sock_tuple {
6259 /* (Simplified) user return codes for tcx prog type.
6260 * A valid tcx program must return one of these defined values. All other
6261 * return codes are reserved for future use. Must remain compatible with
6262 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6263 * return codes are mapped to TCX_NEXT.
6265 enum tcx_action_base {
6272 struct bpf_xdp_sock {
6276 #define XDP_PACKET_HEADROOM 256
6278 /* User return codes for XDP prog type.
6279 * A valid XDP program must return one of these defined values. All other
6280 * return codes are reserved for future use. Unknown return codes will
6281 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6291 /* user accessible metadata for XDP packet hook
6292 * new fields must be added to the end of this structure
6298 /* Below access go through struct xdp_rxq_info */
6299 __u32 ingress_ifindex; /* rxq->dev->ifindex */
6300 __u32 rx_queue_index; /* rxq->queue_index */
6302 __u32 egress_ifindex; /* txq->dev->ifindex */
6305 /* DEVMAP map-value layout
6307 * The struct data-layout of map-value is a configuration interface.
6308 * New members can only be added to the end of this structure.
6310 struct bpf_devmap_val {
6311 __u32 ifindex; /* device index */
6313 int fd; /* prog fd on map write */
6314 __u32 id; /* prog id on map read */
6318 /* CPUMAP map-value layout
6320 * The struct data-layout of map-value is a configuration interface.
6321 * New members can only be added to the end of this structure.
6323 struct bpf_cpumap_val {
6324 __u32 qsize; /* queue size to remote target CPU */
6326 int fd; /* prog fd on map write */
6327 __u32 id; /* prog id on map read */
6336 /* user accessible metadata for SK_MSG packet hook, new fields must
6337 * be added to the end of this structure
6340 __bpf_md_ptr(void *, data);
6341 __bpf_md_ptr(void *, data_end);
6344 __u32 remote_ip4; /* Stored in network byte order */
6345 __u32 local_ip4; /* Stored in network byte order */
6346 __u32 remote_ip6[4]; /* Stored in network byte order */
6347 __u32 local_ip6[4]; /* Stored in network byte order */
6348 __u32 remote_port; /* Stored in network byte order */
6349 __u32 local_port; /* stored in host byte order */
6350 __u32 size; /* Total size of sk_msg */
6352 __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6355 struct sk_reuseport_md {
6357 * Start of directly accessible data. It begins from
6358 * the tcp/udp header.
6360 __bpf_md_ptr(void *, data);
6361 /* End of directly accessible data */
6362 __bpf_md_ptr(void *, data_end);
6364 * Total length of packet (starting from the tcp/udp header).
6365 * Note that the directly accessible bytes (data_end - data)
6366 * could be less than this "len". Those bytes could be
6367 * indirectly read by a helper "bpf_skb_load_bytes()".
6371 * Eth protocol in the mac header (network byte order). e.g.
6372 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6375 __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6376 __u32 bind_inany; /* Is sock bound to an INANY address? */
6377 __u32 hash; /* A hash of the packet 4 tuples */
6378 /* When reuse->migrating_sk is NULL, it is selecting a sk for the
6379 * new incoming connection request (e.g. selecting a listen sk for
6380 * the received SYN in the TCP case). reuse->sk is one of the sk
6381 * in the reuseport group. The bpf prog can use reuse->sk to learn
6382 * the local listening ip/port without looking into the skb.
6384 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6385 * reuse->migrating_sk is the socket that needs to be migrated
6386 * to another listening socket. migrating_sk could be a fullsock
6387 * sk that is fully established or a reqsk that is in-the-middle
6388 * of 3-way handshake.
6390 __bpf_md_ptr(struct bpf_sock *, sk);
6391 __bpf_md_ptr(struct bpf_sock *, migrating_sk);
6394 #define BPF_TAG_SIZE 8
6396 struct bpf_prog_info {
6399 __u8 tag[BPF_TAG_SIZE];
6400 __u32 jited_prog_len;
6401 __u32 xlated_prog_len;
6402 __aligned_u64 jited_prog_insns;
6403 __aligned_u64 xlated_prog_insns;
6404 __u64 load_time; /* ns since boottime */
6405 __u32 created_by_uid;
6407 __aligned_u64 map_ids;
6408 char name[BPF_OBJ_NAME_LEN];
6410 __u32 gpl_compatible:1;
6411 __u32 :31; /* alignment pad */
6414 __u32 nr_jited_ksyms;
6415 __u32 nr_jited_func_lens;
6416 __aligned_u64 jited_ksyms;
6417 __aligned_u64 jited_func_lens;
6419 __u32 func_info_rec_size;
6420 __aligned_u64 func_info;
6423 __aligned_u64 line_info;
6424 __aligned_u64 jited_line_info;
6425 __u32 nr_jited_line_info;
6426 __u32 line_info_rec_size;
6427 __u32 jited_line_info_rec_size;
6429 __aligned_u64 prog_tags;
6432 __u64 recursion_misses;
6433 __u32 verified_insns;
6434 __u32 attach_btf_obj_id;
6435 __u32 attach_btf_id;
6436 } __attribute__((aligned(8)));
6438 struct bpf_map_info {
6445 char name[BPF_OBJ_NAME_LEN];
6447 __u32 btf_vmlinux_value_type_id;
6451 __u32 btf_key_type_id;
6452 __u32 btf_value_type_id;
6453 __u32 :32; /* alignment pad */
6455 } __attribute__((aligned(8)));
6457 struct bpf_btf_info {
6464 } __attribute__((aligned(8)));
6466 struct bpf_link_info {
6472 __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6473 __u32 tp_name_len; /* in/out: tp_name buffer len */
6477 __u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6478 __u32 target_btf_id; /* BTF type id inside the object */
6485 __aligned_u64 target_name; /* in/out: target_name buffer ptr */
6486 __u32 target_name_len; /* in/out: target_name buffer len */
6488 /* If the iter specific field is 32 bits, it can be put
6489 * in the first or second union. Otherwise it should be
6490 * put in the second union.
6525 __aligned_u64 addrs;
6526 __u32 count; /* in/out: kprobe_multi function count */
6530 __u32 type; /* enum bpf_perf_event_type */
6534 __aligned_u64 file_name; /* in/out */
6536 __u32 offset; /* offset from file_name */
6537 } uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6539 __aligned_u64 func_name; /* in/out */
6541 __u32 offset; /* offset from func_name */
6543 } kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6545 __aligned_u64 tp_name; /* in/out */
6547 } tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
6551 } event; /* BPF_PERF_EVENT_EVENT */
6559 } __attribute__((aligned(8)));
6561 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6562 * by user and intended to be used by socket (e.g. to bind to, depends on
6565 struct bpf_sock_addr {
6566 __u32 user_family; /* Allows 4-byte read, but no write. */
6567 __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6568 * Stored in network byte order.
6570 __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6571 * Stored in network byte order.
6573 __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write.
6574 * Stored in network byte order
6576 __u32 family; /* Allows 4-byte read, but no write */
6577 __u32 type; /* Allows 4-byte read, but no write */
6578 __u32 protocol; /* Allows 4-byte read, but no write */
6579 __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write.
6580 * Stored in network byte order.
6582 __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write.
6583 * Stored in network byte order.
6585 __bpf_md_ptr(struct bpf_sock *, sk);
6588 /* User bpf_sock_ops struct to access socket values and specify request ops
6589 * and their replies.
6590 * Some of this fields are in network (bigendian) byte order and may need
6591 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6592 * New fields can only be added at the end of this structure
6594 struct bpf_sock_ops {
6597 __u32 args[4]; /* Optionally passed to bpf program */
6598 __u32 reply; /* Returned by bpf program */
6599 __u32 replylong[4]; /* Optionally returned by bpf prog */
6602 __u32 remote_ip4; /* Stored in network byte order */
6603 __u32 local_ip4; /* Stored in network byte order */
6604 __u32 remote_ip6[4]; /* Stored in network byte order */
6605 __u32 local_ip6[4]; /* Stored in network byte order */
6606 __u32 remote_port; /* Stored in network byte order */
6607 __u32 local_port; /* stored in host byte order */
6608 __u32 is_fullsock; /* Some TCP fields are only valid if
6609 * there is a full socket. If not, the
6610 * fields read as zero.
6613 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
6614 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6623 __u32 rate_delivered;
6624 __u32 rate_interval_us;
6627 __u32 total_retrans;
6631 __u32 data_segs_out;
6635 __u64 bytes_received;
6637 __bpf_md_ptr(struct bpf_sock *, sk);
6638 /* [skb_data, skb_data_end) covers the whole TCP header.
6640 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6641 * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the
6642 * header has not been written.
6643 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6644 * been written so far.
6645 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes
6647 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6650 * bpf_load_hdr_opt() can also be used to read a particular option.
6652 __bpf_md_ptr(void *, skb_data);
6653 __bpf_md_ptr(void *, skb_data_end);
6654 __u32 skb_len; /* The total length of a packet.
6655 * It includes the header, options,
6658 __u32 skb_tcp_flags; /* tcp_flags of the header. It provides
6659 * an easy way to check for tcp_flags
6660 * without parsing skb_data.
6662 * In particular, the skb_tcp_flags
6663 * will still be available in
6664 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6665 * the outgoing header has not
6671 /* Definitions for bpf_sock_ops_cb_flags */
6673 BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0),
6674 BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1),
6675 BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2),
6676 BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3),
6677 /* Call bpf for all received TCP headers. The bpf prog will be
6678 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6680 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6681 * for the header option related helpers that will be useful
6682 * to the bpf programs.
6684 * It could be used at the client/active side (i.e. connect() side)
6685 * when the server told it that the server was in syncookie
6686 * mode and required the active side to resend the bpf-written
6687 * options. The active side can keep writing the bpf-options until
6688 * it received a valid packet from the server side to confirm
6689 * the earlier packet (and options) has been received. The later
6690 * example patch is using it like this at the active side when the
6691 * server is in syncookie mode.
6693 * The bpf prog will usually turn this off in the common cases.
6695 BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4),
6696 /* Call bpf when kernel has received a header option that
6697 * the kernel cannot handle. The bpf prog will be called under
6698 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6700 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6701 * for the header option related helpers that will be useful
6702 * to the bpf programs.
6704 BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6705 /* Call bpf when the kernel is writing header options for the
6706 * outgoing packet. The bpf prog will first be called
6707 * to reserve space in a skb under
6708 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then
6709 * the bpf prog will be called to write the header option(s)
6710 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6712 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6713 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6714 * related helpers that will be useful to the bpf programs.
6716 * The kernel gets its chance to reserve space and write
6717 * options first before the BPF program does.
6719 BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6720 /* Mask of all currently supported cb flags */
6721 BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F,
6724 /* List of known BPF sock_ops operators.
6725 * New entries can only be added at the end
6729 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
6730 * -1 if default value should be used
6732 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
6733 * window (in packets) or -1 if default
6734 * value should be used
6736 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
6737 * active connection is initialized
6739 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
6740 * active connection is
6743 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
6744 * passive connection is
6747 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
6750 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
6751 * based on the path and may be
6752 * dependent on the congestion control
6753 * algorithm. In general it indicates
6754 * a congestion threshold. RTTs above
6755 * this indicate congestion
6757 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
6758 * Arg1: value of icsk_retransmits
6759 * Arg2: value of icsk_rto
6760 * Arg3: whether RTO has expired
6762 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
6763 * Arg1: sequence number of 1st byte
6765 * Arg3: return value of
6766 * tcp_transmit_skb (0 => success)
6768 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
6772 BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
6773 * socket transition to LISTEN state.
6775 BPF_SOCK_OPS_RTT_CB, /* Called on every RTT.
6777 BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option.
6778 * It will be called to handle
6779 * the packets received at
6780 * an already established
6783 * sock_ops->skb_data:
6784 * Referring to the received skb.
6785 * It covers the TCP header only.
6787 * bpf_load_hdr_opt() can also
6788 * be used to search for a
6789 * particular option.
6791 BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the
6792 * header option later in
6793 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6794 * Arg1: bool want_cookie. (in
6795 * writing SYNACK only)
6797 * sock_ops->skb_data:
6798 * Not available because no header has
6801 * sock_ops->skb_tcp_flags:
6802 * The tcp_flags of the
6803 * outgoing skb. (e.g. SYN, ACK, FIN).
6805 * bpf_reserve_hdr_opt() should
6806 * be used to reserve space.
6808 BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options
6809 * Arg1: bool want_cookie. (in
6810 * writing SYNACK only)
6812 * sock_ops->skb_data:
6813 * Referring to the outgoing skb.
6814 * It covers the TCP header
6815 * that has already been written
6816 * by the kernel and the
6817 * earlier bpf-progs.
6819 * sock_ops->skb_tcp_flags:
6820 * The tcp_flags of the outgoing
6821 * skb. (e.g. SYN, ACK, FIN).
6823 * bpf_store_hdr_opt() should
6824 * be used to write the
6827 * bpf_load_hdr_opt() can also
6828 * be used to search for a
6829 * particular option that
6830 * has already been written
6831 * by the kernel or the
6832 * earlier bpf-progs.
6836 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
6837 * changes between the TCP and BPF versions. Ideally this should never happen.
6838 * If it does, we need to add code to convert them before calling
6839 * the BPF sock_ops function.
6842 BPF_TCP_ESTABLISHED = 1,
6852 BPF_TCP_CLOSING, /* Now a valid state */
6853 BPF_TCP_NEW_SYN_RECV,
6855 BPF_TCP_MAX_STATES /* Leave at the end! */
6859 TCP_BPF_IW = 1001, /* Set TCP initial congestion window */
6860 TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */
6861 TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */
6862 TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */
6863 /* Copy the SYN pkt to optval
6865 * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the
6866 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6867 * to only getting from the saved_syn. It can either get the
6870 * 1. the just-received SYN packet (only available when writing the
6871 * SYNACK). It will be useful when it is not necessary to
6872 * save the SYN packet for latter use. It is also the only way
6873 * to get the SYN during syncookie mode because the syn
6874 * packet cannot be saved during syncookie.
6878 * 2. the earlier saved syn which was done by
6879 * bpf_setsockopt(TCP_SAVE_SYN).
6881 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6882 * SYN packet is obtained.
6884 * If the bpf-prog does not need the IP[46] header, the
6885 * bpf-prog can avoid parsing the IP header by using
6886 * TCP_BPF_SYN. Otherwise, the bpf-prog can get both
6887 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6889 * >0: Total number of bytes copied
6890 * -ENOSPC: Not enough space in optval. Only optlen number of
6892 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6893 * is not saved by setsockopt(TCP_SAVE_SYN).
6895 TCP_BPF_SYN = 1005, /* Copy the TCP header */
6896 TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */
6897 TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */
6901 BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6904 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6905 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6908 BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the
6909 * total option spaces
6910 * required for an established
6911 * sk in order to calculate the
6912 * MSS. No skb is actually
6915 BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode
6916 * when sending a SYN.
6920 struct bpf_perf_event_value {
6927 BPF_DEVCG_ACC_MKNOD = (1ULL << 0),
6928 BPF_DEVCG_ACC_READ = (1ULL << 1),
6929 BPF_DEVCG_ACC_WRITE = (1ULL << 2),
6933 BPF_DEVCG_DEV_BLOCK = (1ULL << 0),
6934 BPF_DEVCG_DEV_CHAR = (1ULL << 1),
6937 struct bpf_cgroup_dev_ctx {
6938 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6944 struct bpf_raw_tracepoint_args {
6948 /* DIRECT: Skip the FIB rules and go to FIB table associated with device
6949 * OUTPUT: Do lookup from egress perspective; default is ingress
6952 BPF_FIB_LOOKUP_DIRECT = (1U << 0),
6953 BPF_FIB_LOOKUP_OUTPUT = (1U << 1),
6954 BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
6955 BPF_FIB_LOOKUP_TBID = (1U << 3),
6959 BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
6960 BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
6961 BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
6962 BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
6963 BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
6964 BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6965 BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
6966 BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
6967 BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
6970 struct bpf_fib_lookup {
6971 /* input: network family for lookup (AF_INET, AF_INET6)
6972 * output: network family of egress nexthop
6976 /* set if lookup is to consider L4 data - e.g., FIB rules */
6981 union { /* used for MTU check */
6982 /* input to lookup */
6983 __u16 tot_len; /* L3 length from network hdr (iph->tot_len) */
6985 /* output: MTU value */
6988 /* input: L3 device index for lookup
6989 * output: device index from FIB lookup
6994 /* inputs to lookup */
6995 __u8 tos; /* AF_INET */
6996 __be32 flowinfo; /* AF_INET6, flow_label + priority */
6998 /* output: metric of fib result (IPv4/IPv6 only) */
7004 __u32 ipv6_src[4]; /* in6_addr; network order */
7007 /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7008 * network header. output: bpf_fib_lookup sets to gateway address
7009 * if FIB lookup returns gateway route
7013 __u32 ipv6_dst[4]; /* in6_addr; network order */
7019 __be16 h_vlan_proto;
7022 /* input: when accompanied with the
7023 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7024 * specific routing table to use for the fib lookup.
7029 __u8 smac[6]; /* ETH_ALEN */
7030 __u8 dmac[6]; /* ETH_ALEN */
7033 struct bpf_redir_neigh {
7034 /* network family for lookup (AF_INET, AF_INET6) */
7036 /* network address of nexthop; skips fib lookup to find gateway */
7039 __u32 ipv6_nh[4]; /* in6_addr; network order */
7043 /* bpf_check_mtu flags*/
7044 enum bpf_check_mtu_flags {
7045 BPF_MTU_CHK_SEGS = (1U << 0),
7048 enum bpf_check_mtu_ret {
7049 BPF_MTU_CHK_RET_SUCCESS, /* check and lookup successful */
7050 BPF_MTU_CHK_RET_FRAG_NEEDED, /* fragmentation required to fwd */
7051 BPF_MTU_CHK_RET_SEGS_TOOBIG, /* GSO re-segmentation needed to fwd */
7054 enum bpf_task_fd_type {
7055 BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
7056 BPF_FD_TYPE_TRACEPOINT, /* tp name */
7057 BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
7058 BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
7059 BPF_FD_TYPE_UPROBE, /* filename + offset */
7060 BPF_FD_TYPE_URETPROBE, /* filename + offset */
7064 BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0),
7065 BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1),
7066 BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2),
7069 struct bpf_flow_keys {
7072 __u16 addr_proto; /* ETH_P_* of valid addrs */
7086 __u32 ipv6_src[4]; /* in6_addr; network order */
7087 __u32 ipv6_dst[4]; /* in6_addr; network order */
7094 struct bpf_func_info {
7099 #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10)
7100 #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff)
7102 struct bpf_line_info {
7104 __u32 file_name_off;
7109 struct bpf_spin_lock {
7116 } __attribute__((aligned(8)));
7121 } __attribute__((aligned(8)));
7123 struct bpf_list_head {
7126 } __attribute__((aligned(8)));
7128 struct bpf_list_node {
7132 } __attribute__((aligned(8)));
7134 struct bpf_rb_root {
7137 } __attribute__((aligned(8)));
7139 struct bpf_rb_node {
7144 } __attribute__((aligned(8)));
7146 struct bpf_refcount {
7148 } __attribute__((aligned(4)));
7151 __u32 write; /* Sysctl is being read (= 0) or written (= 1).
7152 * Allows 1,2,4-byte read, but no write.
7154 __u32 file_pos; /* Sysctl file position to read from, write to.
7155 * Allows 1,2,4-byte read an 4-byte write.
7159 struct bpf_sockopt {
7160 __bpf_md_ptr(struct bpf_sock *, sk);
7161 __bpf_md_ptr(void *, optval);
7162 __bpf_md_ptr(void *, optval_end);
7170 struct bpf_pidns_info {
7175 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7176 struct bpf_sk_lookup {
7178 __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
7179 __u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
7182 __u32 family; /* Protocol family (AF_INET, AF_INET6) */
7183 __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7184 __u32 remote_ip4; /* Network byte order */
7185 __u32 remote_ip6[4]; /* Network byte order */
7186 __be16 remote_port; /* Network byte order */
7187 __u16 :16; /* Zero padding */
7188 __u32 local_ip4; /* Network byte order */
7189 __u32 local_ip6[4]; /* Network byte order */
7190 __u32 local_port; /* Host byte order */
7191 __u32 ingress_ifindex; /* The arriving interface. Determined by inet_iif. */
7195 * struct btf_ptr is used for typed pointer representation; the
7196 * type id is used to render the pointer data as the appropriate type
7197 * via the bpf_snprintf_btf() helper described above. A flags field -
7198 * potentially to specify additional details about the BTF pointer
7199 * (rather than its mode of display) - is included for future use.
7200 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7205 __u32 flags; /* BTF ptr flags; unused at present. */
7209 * Flags to control bpf_snprintf_btf() behaviour.
7210 * - BTF_F_COMPACT: no formatting around type information
7211 * - BTF_F_NONAME: no struct/union member names/types
7212 * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7213 * equivalent to %px.
7214 * - BTF_F_ZERO: show zero-valued struct/union members; they
7215 * are not displayed by default
7218 BTF_F_COMPACT = (1ULL << 0),
7219 BTF_F_NONAME = (1ULL << 1),
7220 BTF_F_PTR_RAW = (1ULL << 2),
7221 BTF_F_ZERO = (1ULL << 3),
7224 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7225 * has to be adjusted by relocations. It is emitted by llvm and passed to
7226 * libbpf and later to the kernel.
7228 enum bpf_core_relo_kind {
7229 BPF_CORE_FIELD_BYTE_OFFSET = 0, /* field byte offset */
7230 BPF_CORE_FIELD_BYTE_SIZE = 1, /* field size in bytes */
7231 BPF_CORE_FIELD_EXISTS = 2, /* field existence in target kernel */
7232 BPF_CORE_FIELD_SIGNED = 3, /* field signedness (0 - unsigned, 1 - signed) */
7233 BPF_CORE_FIELD_LSHIFT_U64 = 4, /* bitfield-specific left bitshift */
7234 BPF_CORE_FIELD_RSHIFT_U64 = 5, /* bitfield-specific right bitshift */
7235 BPF_CORE_TYPE_ID_LOCAL = 6, /* type ID in local BPF object */
7236 BPF_CORE_TYPE_ID_TARGET = 7, /* type ID in target kernel */
7237 BPF_CORE_TYPE_EXISTS = 8, /* type existence in target kernel */
7238 BPF_CORE_TYPE_SIZE = 9, /* type size in bytes */
7239 BPF_CORE_ENUMVAL_EXISTS = 10, /* enum value existence in target kernel */
7240 BPF_CORE_ENUMVAL_VALUE = 11, /* enum value integer value */
7241 BPF_CORE_TYPE_MATCHES = 12, /* type match in target kernel */
7245 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7246 * and from libbpf to the kernel.
7248 * CO-RE relocation captures the following data:
7249 * - insn_off - instruction offset (in bytes) within a BPF program that needs
7250 * its insn->imm field to be relocated with actual field info;
7251 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7253 * - access_str_off - offset into corresponding .BTF string section. String
7254 * interpretation depends on specific relocation kind:
7255 * - for field-based relocations, string encodes an accessed field using
7256 * a sequence of field and array indices, separated by colon (:). It's
7257 * conceptually very close to LLVM's getelementptr ([0]) instruction's
7258 * arguments for identifying offset to a field.
7259 * - for type-based relocations, strings is expected to be just "0";
7260 * - for enum value-based relocations, string contains an index of enum
7261 * value within its enum type;
7262 * - kind - one of enum bpf_core_relo_kind;
7272 * struct sample *s = ...;
7273 * int *x = &s->a; // encoded as "0:0" (a is field #0)
7274 * int *y = &s->b[5]; // encoded as "0:1:0:5" (anon struct is field #1,
7275 * // b is field #0 inside anon struct, accessing elem #5)
7276 * int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7278 * type_id for all relocs in this example will capture BTF type id of
7281 * Such relocation is emitted when using __builtin_preserve_access_index()
7282 * Clang built-in, passing expression that captures field address, e.g.:
7284 * bpf_probe_read(&dst, sizeof(dst),
7285 * __builtin_preserve_access_index(&src->a.b.c));
7287 * In this case Clang will emit field relocation recording necessary data to
7288 * be able to find offset of embedded `a.b.c` field within `src` struct.
7290 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7292 struct bpf_core_relo {
7295 __u32 access_str_off;
7296 enum bpf_core_relo_kind kind;
7300 * Flags to control bpf_timer_start() behaviour.
7301 * - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7302 * relative to current time.
7305 BPF_F_TIMER_ABS = (1ULL << 0),
7308 /* BPF numbers iterator state */
7309 struct bpf_iter_num {
7310 /* opaque iterator state; having __u64 here allows to preserve correct
7311 * alignment requirements in vmlinux.h, generated from BTF
7314 } __attribute__((aligned(8)));
7316 #endif /* _UAPI__LINUX_BPF_H__ */