Where Online Learning is simpler!

The C and C++ Include Header Files

/usr/include/linux/bpf.h


$ cat -n /usr/include/linux/bpf.h

     1	/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
     2	/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
     3	 *
     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.
     7	 */
     8	#ifndef __LINUX_BPF_H__
     9	#define __LINUX_BPF_H__
    10	
    11	#include <linux/types.h>
    12	#include <linux/bpf_common.h>
    13	
    14	/* Extended instruction set based on top of classic BPF */
    15	
    16	/* instruction classes */
    17	#define BPF_JMP32	0x06	/* jmp mode in word width */
    18	#define BPF_ALU64	0x07	/* alu mode in double word width */
    19	
    20	/* ld/ldx fields */
    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 */
    25	
    26	/* alu/jmp fields */
    27	#define BPF_MOV		0xb0	/* mov reg to reg */
    28	#define BPF_ARSH	0xc0	/* sign extending arithmetic shift right */
    29	
    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
    36	
    37	/* jmp encodings */
    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 */
    47	
    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 */
    52	
    53	/* Register numbers */
    54	enum {
    55		BPF_REG_0 = 0,
    56		BPF_REG_1,
    57		BPF_REG_2,
    58		BPF_REG_3,
    59		BPF_REG_4,
    60		BPF_REG_5,
    61		BPF_REG_6,
    62		BPF_REG_7,
    63		BPF_REG_8,
    64		BPF_REG_9,
    65		BPF_REG_10,
    66		__MAX_BPF_REG,
    67	};
    68	
    69	/* BPF has 10 general purpose 64-bit registers and stack frame. */
    70	#define MAX_BPF_REG	__MAX_BPF_REG
    71	
    72	struct bpf_insn {
    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 */
    78	};
    79	
    80	/* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
    81	 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
    82	 * the trailing flexible array member) instead.
    83	 */
    84	struct bpf_lpm_trie_key {
    85		__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
    86		__u8	data[0];	/* Arbitrary size */
    87	};
    88	
    89	/* Header for bpf_lpm_trie_key structs */
    90	struct bpf_lpm_trie_key_hdr {
    91		__u32	prefixlen;
    92	};
    93	
    94	/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
    95	struct bpf_lpm_trie_key_u8 {
    96		union {
    97			struct bpf_lpm_trie_key_hdr	hdr;
    98			__u32				prefixlen;
    99		};
   100		__u8	data[];		/* Arbitrary size */
   101	};
   102	
   103	struct bpf_cgroup_storage_key {
   104		__u64	cgroup_inode_id;	/* cgroup inode id */
   105		__u32	attach_type;		/* program attach type (enum bpf_attach_type) */
   106	};
   107	
   108	enum bpf_cgroup_iter_order {
   109		BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
   110		BPF_CGROUP_ITER_SELF_ONLY,		/* process only a single object. */
   111		BPF_CGROUP_ITER_DESCENDANTS_PRE,	/* walk descendants in pre-order. */
   112		BPF_CGROUP_ITER_DESCENDANTS_POST,	/* walk descendants in post-order. */
   113		BPF_CGROUP_ITER_ANCESTORS_UP,		/* walk ancestors upward. */
   114	};
   115	
   116	union bpf_iter_link_info {
   117		struct {
   118			__u32	map_fd;
   119		} map;
   120		struct {
   121			enum bpf_cgroup_iter_order order;
   122	
   123			/* At most one of cgroup_fd and cgroup_id can be non-zero. If
   124			 * both are zero, the walk starts from the default cgroup v2
   125			 * root. For walking v1 hierarchy, one should always explicitly
   126			 * specify cgroup_fd.
   127			 */
   128			__u32	cgroup_fd;
   129			__u64	cgroup_id;
   130		} cgroup;
   131		/* Parameters of task iterators. */
   132		struct {
   133			__u32	tid;
   134			__u32	pid;
   135			__u32	pid_fd;
   136		} task;
   137	};
   138	
   139	/* BPF syscall commands, see bpf(2) man-page for more details. */
   140	/**
   141	 * DOC: eBPF Syscall Preamble
   142	 *
   143	 * The operation to be performed by the **bpf**\ () system call is determined
   144	 * by the *cmd* argument. Each operation takes an accompanying argument,
   145	 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
   146	 * below). The size argument is the size of the union pointed to by *attr*.
   147	 */
   148	/**
   149	 * DOC: eBPF Syscall Commands
   150	 *
   151	 * BPF_MAP_CREATE
   152	 *	Description
   153	 *		Create a map and return a file descriptor that refers to the
   154	 *		map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
   155	 *		is automatically enabled for the new file descriptor.
   156	 *
   157	 *		Applying **close**\ (2) to the file descriptor returned by
   158	 *		**BPF_MAP_CREATE** will delete the map (but see NOTES).
   159	 *
   160	 *	Return
   161	 *		A new file descriptor (a nonnegative integer), or -1 if an
   162	 *		error occurred (in which case, *errno* is set appropriately).
   163	 *
   164	 * BPF_MAP_LOOKUP_ELEM
   165	 *	Description
   166	 *		Look up an element with a given *key* in the map referred to
   167	 *		by the file descriptor *map_fd*.
   168	 *
   169	 *		The *flags* argument may be specified as one of the
   170	 *		following:
   171	 *
   172	 *		**BPF_F_LOCK**
   173	 *			Look up the value of a spin-locked map without
   174	 *			returning the lock. This must be specified if the
   175	 *			elements contain a spinlock.
   176	 *
   177	 *	Return
   178	 *		Returns zero on success. On error, -1 is returned and *errno*
   179	 *		is set appropriately.
   180	 *
   181	 * BPF_MAP_UPDATE_ELEM
   182	 *	Description
   183	 *		Create or update an element (key/value pair) in a specified map.
   184	 *
   185	 *		The *flags* argument should be specified as one of the
   186	 *		following:
   187	 *
   188	 *		**BPF_ANY**
   189	 *			Create a new element or update an existing element.
   190	 *		**BPF_NOEXIST**
   191	 *			Create a new element only if it did not exist.
   192	 *		**BPF_EXIST**
   193	 *			Update an existing element.
   194	 *		**BPF_F_LOCK**
   195	 *			Update a spin_lock-ed map element.
   196	 *
   197	 *	Return
   198	 *		Returns zero on success. On error, -1 is returned and *errno*
   199	 *		is set appropriately.
   200	 *
   201	 *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
   202	 *		**E2BIG**, **EEXIST**, or **ENOENT**.
   203	 *
   204	 *		**E2BIG**
   205	 *			The number of elements in the map reached the
   206	 *			*max_entries* limit specified at map creation time.
   207	 *		**EEXIST**
   208	 *			If *flags* specifies **BPF_NOEXIST** and the element
   209	 *			with *key* already exists in the map.
   210	 *		**ENOENT**
   211	 *			If *flags* specifies **BPF_EXIST** and the element with
   212	 *			*key* does not exist in the map.
   213	 *
   214	 * BPF_MAP_DELETE_ELEM
   215	 *	Description
   216	 *		Look up and delete an element by key in a specified map.
   217	 *
   218	 *	Return
   219	 *		Returns zero on success. On error, -1 is returned and *errno*
   220	 *		is set appropriately.
   221	 *
   222	 * BPF_MAP_GET_NEXT_KEY
   223	 *	Description
   224	 *		Look up an element by key in a specified map and return the key
   225	 *		of the next element. Can be used to iterate over all elements
   226	 *		in the map.
   227	 *
   228	 *	Return
   229	 *		Returns zero on success. On error, -1 is returned and *errno*
   230	 *		is set appropriately.
   231	 *
   232	 *		The following cases can be used to iterate over all elements of
   233	 *		the map:
   234	 *
   235	 *		* If *key* is not found, the operation returns zero and sets
   236	 *		  the *next_key* pointer to the key of the first element.
   237	 *		* If *key* is found, the operation returns zero and sets the
   238	 *		  *next_key* pointer to the key of the next element.
   239	 *		* If *key* is the last element, returns -1 and *errno* is set
   240	 *		  to **ENOENT**.
   241	 *
   242	 *		May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
   243	 *		**EINVAL** on error.
   244	 *
   245	 * BPF_PROG_LOAD
   246	 *	Description
   247	 *		Verify and load an eBPF program, returning a new file
   248	 *		descriptor associated with the program.
   249	 *
   250	 *		Applying **close**\ (2) to the file descriptor returned by
   251	 *		**BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
   252	 *
   253	 *		The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
   254	 *		automatically enabled for the new file descriptor.
   255	 *
   256	 *	Return
   257	 *		A new file descriptor (a nonnegative integer), or -1 if an
   258	 *		error occurred (in which case, *errno* is set appropriately).
   259	 *
   260	 * BPF_OBJ_PIN
   261	 *	Description
   262	 *		Pin an eBPF program or map referred by the specified *bpf_fd*
   263	 *		to the provided *pathname* on the filesystem.
   264	 *
   265	 *		The *pathname* argument must not contain a dot (".").
   266	 *
   267	 *		On success, *pathname* retains a reference to the eBPF object,
   268	 *		preventing deallocation of the object when the original
   269	 *		*bpf_fd* is closed. This allow the eBPF object to live beyond
   270	 *		**close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
   271	 *		process.
   272	 *
   273	 *		Applying **unlink**\ (2) or similar calls to the *pathname*
   274	 *		unpins the object from the filesystem, removing the reference.
   275	 *		If no other file descriptors or filesystem nodes refer to the
   276	 *		same object, it will be deallocated (see NOTES).
   277	 *
   278	 *		The filesystem type for the parent directory of *pathname* must
   279	 *		be **BPF_FS_MAGIC**.
   280	 *
   281	 *	Return
   282	 *		Returns zero on success. On error, -1 is returned and *errno*
   283	 *		is set appropriately.
   284	 *
   285	 * BPF_OBJ_GET
   286	 *	Description
   287	 *		Open a file descriptor for the eBPF object pinned to the
   288	 *		specified *pathname*.
   289	 *
   290	 *	Return
   291	 *		A new file descriptor (a nonnegative integer), or -1 if an
   292	 *		error occurred (in which case, *errno* is set appropriately).
   293	 *
   294	 * BPF_PROG_ATTACH
   295	 *	Description
   296	 *		Attach an eBPF program to a *target_fd* at the specified
   297	 *		*attach_type* hook.
   298	 *
   299	 *		The *attach_type* specifies the eBPF attachment point to
   300	 *		attach the program to, and must be one of *bpf_attach_type*
   301	 *		(see below).
   302	 *
   303	 *		The *attach_bpf_fd* must be a valid file descriptor for a
   304	 *		loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
   305	 *		or sock_ops type corresponding to the specified *attach_type*.
   306	 *
   307	 *		The *target_fd* must be a valid file descriptor for a kernel
   308	 *		object which depends on the attach type of *attach_bpf_fd*:
   309	 *
   310	 *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
   311	 *		**BPF_PROG_TYPE_CGROUP_SKB**,
   312	 *		**BPF_PROG_TYPE_CGROUP_SOCK**,
   313	 *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
   314	 *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
   315	 *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
   316	 *		**BPF_PROG_TYPE_SOCK_OPS**
   317	 *
   318	 *			Control Group v2 hierarchy with the eBPF controller
   319	 *			enabled. Requires the kernel to be compiled with
   320	 *			**CONFIG_CGROUP_BPF**.
   321	 *
   322	 *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
   323	 *
   324	 *			Network namespace (eg /proc/self/ns/net).
   325	 *
   326	 *		**BPF_PROG_TYPE_LIRC_MODE2**
   327	 *
   328	 *			LIRC device path (eg /dev/lircN). Requires the kernel
   329	 *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
   330	 *
   331	 *		**BPF_PROG_TYPE_SK_SKB**,
   332	 *		**BPF_PROG_TYPE_SK_MSG**
   333	 *
   334	 *			eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
   335	 *
   336	 *	Return
   337	 *		Returns zero on success. On error, -1 is returned and *errno*
   338	 *		is set appropriately.
   339	 *
   340	 * BPF_PROG_DETACH
   341	 *	Description
   342	 *		Detach the eBPF program associated with the *target_fd* at the
   343	 *		hook specified by *attach_type*. The program must have been
   344	 *		previously attached using **BPF_PROG_ATTACH**.
   345	 *
   346	 *	Return
   347	 *		Returns zero on success. On error, -1 is returned and *errno*
   348	 *		is set appropriately.
   349	 *
   350	 * BPF_PROG_TEST_RUN
   351	 *	Description
   352	 *		Run the eBPF program associated with the *prog_fd* a *repeat*
   353	 *		number of times against a provided program context *ctx_in* and
   354	 *		data *data_in*, and return the modified program context
   355	 *		*ctx_out*, *data_out* (for example, packet data), result of the
   356	 *		execution *retval*, and *duration* of the test run.
   357	 *
   358	 *		The sizes of the buffers provided as input and output
   359	 *		parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
   360	 *		be provided in the corresponding variables *ctx_size_in*,
   361	 *		*ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
   362	 *		of these parameters are not provided (ie set to NULL), the
   363	 *		corresponding size field must be zero.
   364	 *
   365	 *		Some program types have particular requirements:
   366	 *
   367	 *		**BPF_PROG_TYPE_SK_LOOKUP**
   368	 *			*data_in* and *data_out* must be NULL.
   369	 *
   370	 *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,
   371	 *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
   372	 *
   373	 *			*ctx_out*, *data_in* and *data_out* must be NULL.
   374	 *			*repeat* must be zero.
   375	 *
   376	 *		BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
   377	 *
   378	 *	Return
   379	 *		Returns zero on success. On error, -1 is returned and *errno*
   380	 *		is set appropriately.
   381	 *
   382	 *		**ENOSPC**
   383	 *			Either *data_size_out* or *ctx_size_out* is too small.
   384	 *		**ENOTSUPP**
   385	 *			This command is not supported by the program type of
   386	 *			the program referred to by *prog_fd*.
   387	 *
   388	 * BPF_PROG_GET_NEXT_ID
   389	 *	Description
   390	 *		Fetch the next eBPF program currently loaded into the kernel.
   391	 *
   392	 *		Looks for the eBPF program with an id greater than *start_id*
   393	 *		and updates *next_id* on success. If no other eBPF programs
   394	 *		remain with ids higher than *start_id*, returns -1 and sets
   395	 *		*errno* to **ENOENT**.
   396	 *
   397	 *	Return
   398	 *		Returns zero on success. On error, or when no id remains, -1
   399	 *		is returned and *errno* is set appropriately.
   400	 *
   401	 * BPF_MAP_GET_NEXT_ID
   402	 *	Description
   403	 *		Fetch the next eBPF map currently loaded into the kernel.
   404	 *
   405	 *		Looks for the eBPF map with an id greater than *start_id*
   406	 *		and updates *next_id* on success. If no other eBPF maps
   407	 *		remain with ids higher than *start_id*, returns -1 and sets
   408	 *		*errno* to **ENOENT**.
   409	 *
   410	 *	Return
   411	 *		Returns zero on success. On error, or when no id remains, -1
   412	 *		is returned and *errno* is set appropriately.
   413	 *
   414	 * BPF_PROG_GET_FD_BY_ID
   415	 *	Description
   416	 *		Open a file descriptor for the eBPF program corresponding to
   417	 *		*prog_id*.
   418	 *
   419	 *	Return
   420	 *		A new file descriptor (a nonnegative integer), or -1 if an
   421	 *		error occurred (in which case, *errno* is set appropriately).
   422	 *
   423	 * BPF_MAP_GET_FD_BY_ID
   424	 *	Description
   425	 *		Open a file descriptor for the eBPF map corresponding to
   426	 *		*map_id*.
   427	 *
   428	 *	Return
   429	 *		A new file descriptor (a nonnegative integer), or -1 if an
   430	 *		error occurred (in which case, *errno* is set appropriately).
   431	 *
   432	 * BPF_OBJ_GET_INFO_BY_FD
   433	 *	Description
   434	 *		Obtain information about the eBPF object corresponding to
   435	 *		*bpf_fd*.
   436	 *
   437	 *		Populates up to *info_len* bytes of *info*, which will be in
   438	 *		one of the following formats depending on the eBPF object type
   439	 *		of *bpf_fd*:
   440	 *
   441	 *		* **struct bpf_prog_info**
   442	 *		* **struct bpf_map_info**
   443	 *		* **struct bpf_btf_info**
   444	 *		* **struct bpf_link_info**
   445	 *
   446	 *	Return
   447	 *		Returns zero on success. On error, -1 is returned and *errno*
   448	 *		is set appropriately.
   449	 *
   450	 * BPF_PROG_QUERY
   451	 *	Description
   452	 *		Obtain information about eBPF programs associated with the
   453	 *		specified *attach_type* hook.
   454	 *
   455	 *		The *target_fd* must be a valid file descriptor for a kernel
   456	 *		object which depends on the attach type of *attach_bpf_fd*:
   457	 *
   458	 *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
   459	 *		**BPF_PROG_TYPE_CGROUP_SKB**,
   460	 *		**BPF_PROG_TYPE_CGROUP_SOCK**,
   461	 *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
   462	 *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
   463	 *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
   464	 *		**BPF_PROG_TYPE_SOCK_OPS**
   465	 *
   466	 *			Control Group v2 hierarchy with the eBPF controller
   467	 *			enabled. Requires the kernel to be compiled with
   468	 *			**CONFIG_CGROUP_BPF**.
   469	 *
   470	 *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
   471	 *
   472	 *			Network namespace (eg /proc/self/ns/net).
   473	 *
   474	 *		**BPF_PROG_TYPE_LIRC_MODE2**
   475	 *
   476	 *			LIRC device path (eg /dev/lircN). Requires the kernel
   477	 *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
   478	 *
   479	 *		**BPF_PROG_QUERY** always fetches the number of programs
   480	 *		attached and the *attach_flags* which were used to attach those
   481	 *		programs. Additionally, if *prog_ids* is nonzero and the number
   482	 *		of attached programs is less than *prog_cnt*, populates
   483	 *		*prog_ids* with the eBPF program ids of the programs attached
   484	 *		at *target_fd*.
   485	 *
   486	 *		The following flags may alter the result:
   487	 *
   488	 *		**BPF_F_QUERY_EFFECTIVE**
   489	 *			Only return information regarding programs which are
   490	 *			currently effective at the specified *target_fd*.
   491	 *
   492	 *	Return
   493	 *		Returns zero on success. On error, -1 is returned and *errno*
   494	 *		is set appropriately.
   495	 *
   496	 * BPF_RAW_TRACEPOINT_OPEN
   497	 *	Description
   498	 *		Attach an eBPF program to a tracepoint *name* to access kernel
   499	 *		internal arguments of the tracepoint in their raw form.
   500	 *
   501	 *		The *prog_fd* must be a valid file descriptor associated with
   502	 *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
   503	 *
   504	 *		No ABI guarantees are made about the content of tracepoint
   505	 *		arguments exposed to the corresponding eBPF program.
   506	 *
   507	 *		Applying **close**\ (2) to the file descriptor returned by
   508	 *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
   509	 *
   510	 *	Return
   511	 *		A new file descriptor (a nonnegative integer), or -1 if an
   512	 *		error occurred (in which case, *errno* is set appropriately).
   513	 *
   514	 * BPF_BTF_LOAD
   515	 *	Description
   516	 *		Verify and load BPF Type Format (BTF) metadata into the kernel,
   517	 *		returning a new file descriptor associated with the metadata.
   518	 *		BTF is described in more detail at
   519	 *		https://www.kernel.org/doc/html/latest/bpf/btf.html.
   520	 *
   521	 *		The *btf* parameter must point to valid memory providing
   522	 *		*btf_size* bytes of BTF binary metadata.
   523	 *
   524	 *		The returned file descriptor can be passed to other **bpf**\ ()
   525	 *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
   526	 *		associate the BTF with those objects.
   527	 *
   528	 *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
   529	 *		parameters to specify a *btf_log_buf*, *btf_log_size* and
   530	 *		*btf_log_level* which allow the kernel to return freeform log
   531	 *		output regarding the BTF verification process.
   532	 *
   533	 *	Return
   534	 *		A new file descriptor (a nonnegative integer), or -1 if an
   535	 *		error occurred (in which case, *errno* is set appropriately).
   536	 *
   537	 * BPF_BTF_GET_FD_BY_ID
   538	 *	Description
   539	 *		Open a file descriptor for the BPF Type Format (BTF)
   540	 *		corresponding to *btf_id*.
   541	 *
   542	 *	Return
   543	 *		A new file descriptor (a nonnegative integer), or -1 if an
   544	 *		error occurred (in which case, *errno* is set appropriately).
   545	 *
   546	 * BPF_TASK_FD_QUERY
   547	 *	Description
   548	 *		Obtain information about eBPF programs associated with the
   549	 *		target process identified by *pid* and *fd*.
   550	 *
   551	 *		If the *pid* and *fd* are associated with a tracepoint, kprobe
   552	 *		or uprobe perf event, then the *prog_id* and *fd_type* will
   553	 *		be populated with the eBPF program id and file descriptor type
   554	 *		of type **bpf_task_fd_type**. If associated with a kprobe or
   555	 *		uprobe, the  *probe_offset* and *probe_addr* will also be
   556	 *		populated. Optionally, if *buf* is provided, then up to
   557	 *		*buf_len* bytes of *buf* will be populated with the name of
   558	 *		the tracepoint, kprobe or uprobe.
   559	 *
   560	 *		The resulting *prog_id* may be introspected in deeper detail
   561	 *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
   562	 *
   563	 *	Return
   564	 *		Returns zero on success. On error, -1 is returned and *errno*
   565	 *		is set appropriately.
   566	 *
   567	 * BPF_MAP_LOOKUP_AND_DELETE_ELEM
   568	 *	Description
   569	 *		Look up an element with the given *key* in the map referred to
   570	 *		by the file descriptor *fd*, and if found, delete the element.
   571	 *
   572	 *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
   573	 *		types, the *flags* argument needs to be set to 0, but for other
   574	 *		map types, it may be specified as:
   575	 *
   576	 *		**BPF_F_LOCK**
   577	 *			Look up and delete the value of a spin-locked map
   578	 *			without returning the lock. This must be specified if
   579	 *			the elements contain a spinlock.
   580	 *
   581	 *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
   582	 *		implement this command as a "pop" operation, deleting the top
   583	 *		element rather than one corresponding to *key*.
   584	 *		The *key* and *key_len* parameters should be zeroed when
   585	 *		issuing this operation for these map types.
   586	 *
   587	 *		This command is only valid for the following map types:
   588	 *		* **BPF_MAP_TYPE_QUEUE**
   589	 *		* **BPF_MAP_TYPE_STACK**
   590	 *		* **BPF_MAP_TYPE_HASH**
   591	 *		* **BPF_MAP_TYPE_PERCPU_HASH**
   592	 *		* **BPF_MAP_TYPE_LRU_HASH**
   593	 *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**
   594	 *
   595	 *	Return
   596	 *		Returns zero on success. On error, -1 is returned and *errno*
   597	 *		is set appropriately.
   598	 *
   599	 * BPF_MAP_FREEZE
   600	 *	Description
   601	 *		Freeze the permissions of the specified map.
   602	 *
   603	 *		Write permissions may be frozen by passing zero *flags*.
   604	 *		Upon success, no future syscall invocations may alter the
   605	 *		map state of *map_fd*. Write operations from eBPF programs
   606	 *		are still possible for a frozen map.
   607	 *
   608	 *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
   609	 *
   610	 *	Return
   611	 *		Returns zero on success. On error, -1 is returned and *errno*
   612	 *		is set appropriately.
   613	 *
   614	 * BPF_BTF_GET_NEXT_ID
   615	 *	Description
   616	 *		Fetch the next BPF Type Format (BTF) object currently loaded
   617	 *		into the kernel.
   618	 *
   619	 *		Looks for the BTF object with an id greater than *start_id*
   620	 *		and updates *next_id* on success. If no other BTF objects
   621	 *		remain with ids higher than *start_id*, returns -1 and sets
   622	 *		*errno* to **ENOENT**.
   623	 *
   624	 *	Return
   625	 *		Returns zero on success. On error, or when no id remains, -1
   626	 *		is returned and *errno* is set appropriately.
   627	 *
   628	 * BPF_MAP_LOOKUP_BATCH
   629	 *	Description
   630	 *		Iterate and fetch multiple elements in a map.
   631	 *
   632	 *		Two opaque values are used to manage batch operations,
   633	 *		*in_batch* and *out_batch*. Initially, *in_batch* must be set
   634	 *		to NULL to begin the batched operation. After each subsequent
   635	 *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
   636	 *		*out_batch* as the *in_batch* for the next operation to
   637	 *		continue iteration from the current point.
   638	 *
   639	 *		The *keys* and *values* are output parameters which must point
   640	 *		to memory large enough to hold *count* items based on the key
   641	 *		and value size of the map *map_fd*. The *keys* buffer must be
   642	 *		of *key_size* * *count*. The *values* buffer must be of
   643	 *		*value_size* * *count*.
   644	 *
   645	 *		The *elem_flags* argument may be specified as one of the
   646	 *		following:
   647	 *
   648	 *		**BPF_F_LOCK**
   649	 *			Look up the value of a spin-locked map without
   650	 *			returning the lock. This must be specified if the
   651	 *			elements contain a spinlock.
   652	 *
   653	 *		On success, *count* elements from the map are copied into the
   654	 *		user buffer, with the keys copied into *keys* and the values
   655	 *		copied into the corresponding indices in *values*.
   656	 *
   657	 *		If an error is returned and *errno* is not **EFAULT**, *count*
   658	 *		is set to the number of successfully processed elements.
   659	 *
   660	 *	Return
   661	 *		Returns zero on success. On error, -1 is returned and *errno*
   662	 *		is set appropriately.
   663	 *
   664	 *		May set *errno* to **ENOSPC** to indicate that *keys* or
   665	 *		*values* is too small to dump an entire bucket during
   666	 *		iteration of a hash-based map type.
   667	 *
   668	 * BPF_MAP_LOOKUP_AND_DELETE_BATCH
   669	 *	Description
   670	 *		Iterate and delete all elements in a map.
   671	 *
   672	 *		This operation has the same behavior as
   673	 *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:
   674	 *
   675	 *		* Every element that is successfully returned is also deleted
   676	 *		  from the map. This is at least *count* elements. Note that
   677	 *		  *count* is both an input and an output parameter.
   678	 *		* Upon returning with *errno* set to **EFAULT**, up to
   679	 *		  *count* elements may be deleted without returning the keys
   680	 *		  and values of the deleted elements.
   681	 *
   682	 *	Return
   683	 *		Returns zero on success. On error, -1 is returned and *errno*
   684	 *		is set appropriately.
   685	 *
   686	 * BPF_MAP_UPDATE_BATCH
   687	 *	Description
   688	 *		Update multiple elements in a map by *key*.
   689	 *
   690	 *		The *keys* and *values* are input parameters which must point
   691	 *		to memory large enough to hold *count* items based on the key
   692	 *		and value size of the map *map_fd*. The *keys* buffer must be
   693	 *		of *key_size* * *count*. The *values* buffer must be of
   694	 *		*value_size* * *count*.
   695	 *
   696	 *		Each element specified in *keys* is sequentially updated to the
   697	 *		value in the corresponding index in *values*. The *in_batch*
   698	 *		and *out_batch* parameters are ignored and should be zeroed.
   699	 *
   700	 *		The *elem_flags* argument should be specified as one of the
   701	 *		following:
   702	 *
   703	 *		**BPF_ANY**
   704	 *			Create new elements or update a existing elements.
   705	 *		**BPF_NOEXIST**
   706	 *			Create new elements only if they do not exist.
   707	 *		**BPF_EXIST**
   708	 *			Update existing elements.
   709	 *		**BPF_F_LOCK**
   710	 *			Update spin_lock-ed map elements. This must be
   711	 *			specified if the map value contains a spinlock.
   712	 *
   713	 *		On success, *count* elements from the map are updated.
   714	 *
   715	 *		If an error is returned and *errno* is not **EFAULT**, *count*
   716	 *		is set to the number of successfully processed elements.
   717	 *
   718	 *	Return
   719	 *		Returns zero on success. On error, -1 is returned and *errno*
   720	 *		is set appropriately.
   721	 *
   722	 *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
   723	 *		**E2BIG**. **E2BIG** indicates that the number of elements in
   724	 *		the map reached the *max_entries* limit specified at map
   725	 *		creation time.
   726	 *
   727	 *		May set *errno* to one of the following error codes under
   728	 *		specific circumstances:
   729	 *
   730	 *		**EEXIST**
   731	 *			If *flags* specifies **BPF_NOEXIST** and the element
   732	 *			with *key* already exists in the map.
   733	 *		**ENOENT**
   734	 *			If *flags* specifies **BPF_EXIST** and the element with
   735	 *			*key* does not exist in the map.
   736	 *
   737	 * BPF_MAP_DELETE_BATCH
   738	 *	Description
   739	 *		Delete multiple elements in a map by *key*.
   740	 *
   741	 *		The *keys* parameter is an input parameter which must point
   742	 *		to memory large enough to hold *count* items based on the key
   743	 *		size of the map *map_fd*, that is, *key_size* * *count*.
   744	 *
   745	 *		Each element specified in *keys* is sequentially deleted. The
   746	 *		*in_batch*, *out_batch*, and *values* parameters are ignored
   747	 *		and should be zeroed.
   748	 *
   749	 *		The *elem_flags* argument may be specified as one of the
   750	 *		following:
   751	 *
   752	 *		**BPF_F_LOCK**
   753	 *			Look up the value of a spin-locked map without
   754	 *			returning the lock. This must be specified if the
   755	 *			elements contain a spinlock.
   756	 *
   757	 *		On success, *count* elements from the map are updated.
   758	 *
   759	 *		If an error is returned and *errno* is not **EFAULT**, *count*
   760	 *		is set to the number of successfully processed elements. If
   761	 *		*errno* is **EFAULT**, up to *count* elements may be been
   762	 *		deleted.
   763	 *
   764	 *	Return
   765	 *		Returns zero on success. On error, -1 is returned and *errno*
   766	 *		is set appropriately.
   767	 *
   768	 * BPF_LINK_CREATE
   769	 *	Description
   770	 *		Attach an eBPF program to a *target_fd* at the specified
   771	 *		*attach_type* hook and return a file descriptor handle for
   772	 *		managing the link.
   773	 *
   774	 *	Return
   775	 *		A new file descriptor (a nonnegative integer), or -1 if an
   776	 *		error occurred (in which case, *errno* is set appropriately).
   777	 *
   778	 * BPF_LINK_UPDATE
   779	 *	Description
   780	 *		Update the eBPF program in the specified *link_fd* to
   781	 *		*new_prog_fd*.
   782	 *
   783	 *	Return
   784	 *		Returns zero on success. On error, -1 is returned and *errno*
   785	 *		is set appropriately.
   786	 *
   787	 * BPF_LINK_GET_FD_BY_ID
   788	 *	Description
   789	 *		Open a file descriptor for the eBPF Link corresponding to
   790	 *		*link_id*.
   791	 *
   792	 *	Return
   793	 *		A new file descriptor (a nonnegative integer), or -1 if an
   794	 *		error occurred (in which case, *errno* is set appropriately).
   795	 *
   796	 * BPF_LINK_GET_NEXT_ID
   797	 *	Description
   798	 *		Fetch the next eBPF link currently loaded into the kernel.
   799	 *
   800	 *		Looks for the eBPF link with an id greater than *start_id*
   801	 *		and updates *next_id* on success. If no other eBPF links
   802	 *		remain with ids higher than *start_id*, returns -1 and sets
   803	 *		*errno* to **ENOENT**.
   804	 *
   805	 *	Return
   806	 *		Returns zero on success. On error, or when no id remains, -1
   807	 *		is returned and *errno* is set appropriately.
   808	 *
   809	 * BPF_ENABLE_STATS
   810	 *	Description
   811	 *		Enable eBPF runtime statistics gathering.
   812	 *
   813	 *		Runtime statistics gathering for the eBPF runtime is disabled
   814	 *		by default to minimize the corresponding performance overhead.
   815	 *		This command enables statistics globally.
   816	 *
   817	 *		Multiple programs may independently enable statistics.
   818	 *		After gathering the desired statistics, eBPF runtime statistics
   819	 *		may be disabled again by calling **close**\ (2) for the file
   820	 *		descriptor returned by this function. Statistics will only be
   821	 *		disabled system-wide when all outstanding file descriptors
   822	 *		returned by prior calls for this subcommand are closed.
   823	 *
   824	 *	Return
   825	 *		A new file descriptor (a nonnegative integer), or -1 if an
   826	 *		error occurred (in which case, *errno* is set appropriately).
   827	 *
   828	 * BPF_ITER_CREATE
   829	 *	Description
   830	 *		Create an iterator on top of the specified *link_fd* (as
   831	 *		previously created using **BPF_LINK_CREATE**) and return a
   832	 *		file descriptor that can be used to trigger the iteration.
   833	 *
   834	 *		If the resulting file descriptor is pinned to the filesystem
   835	 *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
   836	 *		for that path will trigger the iterator to read kernel state
   837	 *		using the eBPF program attached to *link_fd*.
   838	 *
   839	 *	Return
   840	 *		A new file descriptor (a nonnegative integer), or -1 if an
   841	 *		error occurred (in which case, *errno* is set appropriately).
   842	 *
   843	 * BPF_LINK_DETACH
   844	 *	Description
   845	 *		Forcefully detach the specified *link_fd* from its
   846	 *		corresponding attachment point.
   847	 *
   848	 *	Return
   849	 *		Returns zero on success. On error, -1 is returned and *errno*
   850	 *		is set appropriately.
   851	 *
   852	 * BPF_PROG_BIND_MAP
   853	 *	Description
   854	 *		Bind a map to the lifetime of an eBPF program.
   855	 *
   856	 *		The map identified by *map_fd* is bound to the program
   857	 *		identified by *prog_fd* and only released when *prog_fd* is
   858	 *		released. This may be used in cases where metadata should be
   859	 *		associated with a program which otherwise does not contain any
   860	 *		references to the map (for example, embedded in the eBPF
   861	 *		program instructions).
   862	 *
   863	 *	Return
   864	 *		Returns zero on success. On error, -1 is returned and *errno*
   865	 *		is set appropriately.
   866	 *
   867	 * NOTES
   868	 *	eBPF objects (maps and programs) can be shared between processes.
   869	 *
   870	 *	* After **fork**\ (2), the child inherits file descriptors
   871	 *	  referring to the same eBPF objects.
   872	 *	* File descriptors referring to eBPF objects can be transferred over
   873	 *	  **unix**\ (7) domain sockets.
   874	 *	* File descriptors referring to eBPF objects can be duplicated in the
   875	 *	  usual way, using **dup**\ (2) and similar calls.
   876	 *	* File descriptors referring to eBPF objects can be pinned to the
   877	 *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
   878	 *
   879	 *	An eBPF object is deallocated only after all file descriptors referring
   880	 *	to the object have been closed and no references remain pinned to the
   881	 *	filesystem or attached (for example, bound to a program or device).
   882	 */
   883	enum bpf_cmd {
   884		BPF_MAP_CREATE,
   885		BPF_MAP_LOOKUP_ELEM,
   886		BPF_MAP_UPDATE_ELEM,
   887		BPF_MAP_DELETE_ELEM,
   888		BPF_MAP_GET_NEXT_KEY,
   889		BPF_PROG_LOAD,
   890		BPF_OBJ_PIN,
   891		BPF_OBJ_GET,
   892		BPF_PROG_ATTACH,
   893		BPF_PROG_DETACH,
   894		BPF_PROG_TEST_RUN,
   895		BPF_PROG_RUN = BPF_PROG_TEST_RUN,
   896		BPF_PROG_GET_NEXT_ID,
   897		BPF_MAP_GET_NEXT_ID,
   898		BPF_PROG_GET_FD_BY_ID,
   899		BPF_MAP_GET_FD_BY_ID,
   900		BPF_OBJ_GET_INFO_BY_FD,
   901		BPF_PROG_QUERY,
   902		BPF_RAW_TRACEPOINT_OPEN,
   903		BPF_BTF_LOAD,
   904		BPF_BTF_GET_FD_BY_ID,
   905		BPF_TASK_FD_QUERY,
   906		BPF_MAP_LOOKUP_AND_DELETE_ELEM,
   907		BPF_MAP_FREEZE,
   908		BPF_BTF_GET_NEXT_ID,
   909		BPF_MAP_LOOKUP_BATCH,
   910		BPF_MAP_LOOKUP_AND_DELETE_BATCH,
   911		BPF_MAP_UPDATE_BATCH,
   912		BPF_MAP_DELETE_BATCH,
   913		BPF_LINK_CREATE,
   914		BPF_LINK_UPDATE,
   915		BPF_LINK_GET_FD_BY_ID,
   916		BPF_LINK_GET_NEXT_ID,
   917		BPF_ENABLE_STATS,
   918		BPF_ITER_CREATE,
   919		BPF_LINK_DETACH,
   920		BPF_PROG_BIND_MAP,
   921	};
   922	
   923	enum bpf_map_type {
   924		BPF_MAP_TYPE_UNSPEC,
   925		BPF_MAP_TYPE_HASH,
   926		BPF_MAP_TYPE_ARRAY,
   927		BPF_MAP_TYPE_PROG_ARRAY,
   928		BPF_MAP_TYPE_PERF_EVENT_ARRAY,
   929		BPF_MAP_TYPE_PERCPU_HASH,
   930		BPF_MAP_TYPE_PERCPU_ARRAY,
   931		BPF_MAP_TYPE_STACK_TRACE,
   932		BPF_MAP_TYPE_CGROUP_ARRAY,
   933		BPF_MAP_TYPE_LRU_HASH,
   934		BPF_MAP_TYPE_LRU_PERCPU_HASH,
   935		BPF_MAP_TYPE_LPM_TRIE,
   936		BPF_MAP_TYPE_ARRAY_OF_MAPS,
   937		BPF_MAP_TYPE_HASH_OF_MAPS,
   938		BPF_MAP_TYPE_DEVMAP,
   939		BPF_MAP_TYPE_SOCKMAP,
   940		BPF_MAP_TYPE_CPUMAP,
   941		BPF_MAP_TYPE_XSKMAP,
   942		BPF_MAP_TYPE_SOCKHASH,
   943		BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
   944		/* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
   945		 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
   946		 * both cgroup-attached and other progs and supports all functionality
   947		 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
   948		 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
   949		 */
   950		BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
   951		BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
   952		BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
   953		/* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs
   954		 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +
   955		 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
   956		 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE
   957		 * deprecated.
   958		 */
   959		BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,
   960		BPF_MAP_TYPE_QUEUE,
   961		BPF_MAP_TYPE_STACK,
   962		BPF_MAP_TYPE_SK_STORAGE,
   963		BPF_MAP_TYPE_DEVMAP_HASH,
   964		BPF_MAP_TYPE_STRUCT_OPS,
   965		BPF_MAP_TYPE_RINGBUF,
   966		BPF_MAP_TYPE_INODE_STORAGE,
   967		BPF_MAP_TYPE_TASK_STORAGE,
   968		BPF_MAP_TYPE_BLOOM_FILTER,
   969		BPF_MAP_TYPE_USER_RINGBUF,
   970		BPF_MAP_TYPE_CGRP_STORAGE,
   971	};
   972	
   973	/* Note that tracing related programs such as
   974	 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
   975	 * are not subject to a stable API since kernel internal data
   976	 * structures can change from release to release and may
   977	 * therefore break existing tracing BPF programs. Tracing BPF
   978	 * programs correspond to /a/ specific kernel which is to be
   979	 * analyzed, and not /a/ specific kernel /and/ all future ones.
   980	 */
   981	enum bpf_prog_type {
   982		BPF_PROG_TYPE_UNSPEC,
   983		BPF_PROG_TYPE_SOCKET_FILTER,
   984		BPF_PROG_TYPE_KPROBE,
   985		BPF_PROG_TYPE_SCHED_CLS,
   986		BPF_PROG_TYPE_SCHED_ACT,
   987		BPF_PROG_TYPE_TRACEPOINT,
   988		BPF_PROG_TYPE_XDP,
   989		BPF_PROG_TYPE_PERF_EVENT,
   990		BPF_PROG_TYPE_CGROUP_SKB,
   991		BPF_PROG_TYPE_CGROUP_SOCK,
   992		BPF_PROG_TYPE_LWT_IN,
   993		BPF_PROG_TYPE_LWT_OUT,
   994		BPF_PROG_TYPE_LWT_XMIT,
   995		BPF_PROG_TYPE_SOCK_OPS,
   996		BPF_PROG_TYPE_SK_SKB,
   997		BPF_PROG_TYPE_CGROUP_DEVICE,
   998		BPF_PROG_TYPE_SK_MSG,
   999		BPF_PROG_TYPE_RAW_TRACEPOINT,
  1000		BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
  1001		BPF_PROG_TYPE_LWT_SEG6LOCAL,
  1002		BPF_PROG_TYPE_LIRC_MODE2,
  1003		BPF_PROG_TYPE_SK_REUSEPORT,
  1004		BPF_PROG_TYPE_FLOW_DISSECTOR,
  1005		BPF_PROG_TYPE_CGROUP_SYSCTL,
  1006		BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
  1007		BPF_PROG_TYPE_CGROUP_SOCKOPT,
  1008		BPF_PROG_TYPE_TRACING,
  1009		BPF_PROG_TYPE_STRUCT_OPS,
  1010		BPF_PROG_TYPE_EXT,
  1011		BPF_PROG_TYPE_LSM,
  1012		BPF_PROG_TYPE_SK_LOOKUP,
  1013		BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
  1014		BPF_PROG_TYPE_NETFILTER,
  1015	};
  1016	
  1017	enum bpf_attach_type {
  1018		BPF_CGROUP_INET_INGRESS,
  1019		BPF_CGROUP_INET_EGRESS,
  1020		BPF_CGROUP_INET_SOCK_CREATE,
  1021		BPF_CGROUP_SOCK_OPS,
  1022		BPF_SK_SKB_STREAM_PARSER,
  1023		BPF_SK_SKB_STREAM_VERDICT,
  1024		BPF_CGROUP_DEVICE,
  1025		BPF_SK_MSG_VERDICT,
  1026		BPF_CGROUP_INET4_BIND,
  1027		BPF_CGROUP_INET6_BIND,
  1028		BPF_CGROUP_INET4_CONNECT,
  1029		BPF_CGROUP_INET6_CONNECT,
  1030		BPF_CGROUP_INET4_POST_BIND,
  1031		BPF_CGROUP_INET6_POST_BIND,
  1032		BPF_CGROUP_UDP4_SENDMSG,
  1033		BPF_CGROUP_UDP6_SENDMSG,
  1034		BPF_LIRC_MODE2,
  1035		BPF_FLOW_DISSECTOR,
  1036		BPF_CGROUP_SYSCTL,
  1037		BPF_CGROUP_UDP4_RECVMSG,
  1038		BPF_CGROUP_UDP6_RECVMSG,
  1039		BPF_CGROUP_GETSOCKOPT,
  1040		BPF_CGROUP_SETSOCKOPT,
  1041		BPF_TRACE_RAW_TP,
  1042		BPF_TRACE_FENTRY,
  1043		BPF_TRACE_FEXIT,
  1044		BPF_MODIFY_RETURN,
  1045		BPF_LSM_MAC,
  1046		BPF_TRACE_ITER,
  1047		BPF_CGROUP_INET4_GETPEERNAME,
  1048		BPF_CGROUP_INET6_GETPEERNAME,
  1049		BPF_CGROUP_INET4_GETSOCKNAME,
  1050		BPF_CGROUP_INET6_GETSOCKNAME,
  1051		BPF_XDP_DEVMAP,
  1052		BPF_CGROUP_INET_SOCK_RELEASE,
  1053		BPF_XDP_CPUMAP,
  1054		BPF_SK_LOOKUP,
  1055		BPF_XDP,
  1056		BPF_SK_SKB_VERDICT,
  1057		BPF_SK_REUSEPORT_SELECT,
  1058		BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
  1059		BPF_PERF_EVENT,
  1060		BPF_TRACE_KPROBE_MULTI,
  1061		BPF_LSM_CGROUP,
  1062		BPF_STRUCT_OPS,
  1063		BPF_NETFILTER,
  1064		BPF_TCX_INGRESS,
  1065		BPF_TCX_EGRESS,
  1066		BPF_TRACE_UPROBE_MULTI,
  1067		BPF_CGROUP_UNIX_CONNECT,
  1068		BPF_CGROUP_UNIX_SENDMSG,
  1069		BPF_CGROUP_UNIX_RECVMSG,
  1070		BPF_CGROUP_UNIX_GETPEERNAME,
  1071		BPF_CGROUP_UNIX_GETSOCKNAME,
  1072		BPF_NETKIT_PRIMARY,
  1073		BPF_NETKIT_PEER,
  1074		__MAX_BPF_ATTACH_TYPE
  1075	};
  1076	
  1077	#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  1078	
  1079	enum bpf_link_type {
  1080		BPF_LINK_TYPE_UNSPEC = 0,
  1081		BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
  1082		BPF_LINK_TYPE_TRACING = 2,
  1083		BPF_LINK_TYPE_CGROUP = 3,
  1084		BPF_LINK_TYPE_ITER = 4,
  1085		BPF_LINK_TYPE_NETNS = 5,
  1086		BPF_LINK_TYPE_XDP = 6,
  1087		BPF_LINK_TYPE_PERF_EVENT = 7,
  1088		BPF_LINK_TYPE_KPROBE_MULTI = 8,
  1089		BPF_LINK_TYPE_STRUCT_OPS = 9,
  1090		BPF_LINK_TYPE_NETFILTER = 10,
  1091		BPF_LINK_TYPE_TCX = 11,
  1092		BPF_LINK_TYPE_UPROBE_MULTI = 12,
  1093		BPF_LINK_TYPE_NETKIT = 13,
  1094		__MAX_BPF_LINK_TYPE,
  1095	};
  1096	
  1097	#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE
  1098	
  1099	enum bpf_perf_event_type {
  1100		BPF_PERF_EVENT_UNSPEC = 0,
  1101		BPF_PERF_EVENT_UPROBE = 1,
  1102		BPF_PERF_EVENT_URETPROBE = 2,
  1103		BPF_PERF_EVENT_KPROBE = 3,
  1104		BPF_PERF_EVENT_KRETPROBE = 4,
  1105		BPF_PERF_EVENT_TRACEPOINT = 5,
  1106		BPF_PERF_EVENT_EVENT = 6,
  1107	};
  1108	
  1109	/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
  1110	 *
  1111	 * NONE(default): No further bpf programs allowed in the subtree.
  1112	 *
  1113	 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
  1114	 * the program in this cgroup yields to sub-cgroup program.
  1115	 *
  1116	 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
  1117	 * that cgroup program gets run in addition to the program in this cgroup.
  1118	 *
  1119	 * Only one program is allowed to be attached to a cgroup with
  1120	 * NONE or BPF_F_ALLOW_OVERRIDE flag.
  1121	 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
  1122	 * release old program and attach the new one. Attach flags has to match.
  1123	 *
  1124	 * Multiple programs are allowed to be attached to a cgroup with
  1125	 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
  1126	 * (those that were attached first, run first)
  1127	 * The programs of sub-cgroup are executed first, then programs of
  1128	 * this cgroup and then programs of parent cgroup.
  1129	 * When children program makes decision (like picking TCP CA or sock bind)
  1130	 * parent program has a chance to override it.
  1131	 *
  1132	 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
  1133	 * programs for a cgroup. Though it's possible to replace an old program at
  1134	 * any position by also specifying BPF_F_REPLACE flag and position itself in
  1135	 * replace_bpf_fd attribute. Old program at this position will be released.
  1136	 *
  1137	 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
  1138	 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
  1139	 * Ex1:
  1140	 * cgrp1 (MULTI progs A, B) ->
  1141	 *    cgrp2 (OVERRIDE prog C) ->
  1142	 *      cgrp3 (MULTI prog D) ->
  1143	 *        cgrp4 (OVERRIDE prog E) ->
  1144	 *          cgrp5 (NONE prog F)
  1145	 * the event in cgrp5 triggers execution of F,D,A,B in that order.
  1146	 * if prog F is detached, the execution is E,D,A,B
  1147	 * if prog F and D are detached, the execution is E,A,B
  1148	 * if prog F, E and D are detached, the execution is C,A,B
  1149	 *
  1150	 * All eligible programs are executed regardless of return code from
  1151	 * earlier programs.
  1152	 */
  1153	#define BPF_F_ALLOW_OVERRIDE	(1U << 0)
  1154	#define BPF_F_ALLOW_MULTI	(1U << 1)
  1155	/* Generic attachment flags. */
  1156	#define BPF_F_REPLACE		(1U << 2)
  1157	#define BPF_F_BEFORE		(1U << 3)
  1158	#define BPF_F_AFTER		(1U << 4)
  1159	#define BPF_F_ID		(1U << 5)
  1160	#define BPF_F_PREORDER		(1U << 6)
  1161	#define BPF_F_LINK		BPF_F_LINK /* 1 << 13 */
  1162	
  1163	/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  1164	 * verifier will perform strict alignment checking as if the kernel
  1165	 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  1166	 * and NET_IP_ALIGN defined to 2.
  1167	 */
  1168	#define BPF_F_STRICT_ALIGNMENT	(1U << 0)
  1169	
  1170	/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
  1171	 * verifier will allow any alignment whatsoever.  On platforms
  1172	 * with strict alignment requirements for loads ands stores (such
  1173	 * as sparc and mips) the verifier validates that all loads and
  1174	 * stores provably follow this requirement.  This flag turns that
  1175	 * checking and enforcement off.
  1176	 *
  1177	 * It is mostly used for testing when we want to validate the
  1178	 * context and memory access aspects of the verifier, but because
  1179	 * of an unaligned access the alignment check would trigger before
  1180	 * the one we are interested in.
  1181	 */
  1182	#define BPF_F_ANY_ALIGNMENT	(1U << 1)
  1183	
  1184	/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
  1185	 * Verifier does sub-register def/use analysis and identifies instructions whose
  1186	 * def only matters for low 32-bit, high 32-bit is never referenced later
  1187	 * through implicit zero extension. Therefore verifier notifies JIT back-ends
  1188	 * that it is safe to ignore clearing high 32-bit for these instructions. This
  1189	 * saves some back-ends a lot of code-gen. However such optimization is not
  1190	 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
  1191	 * hence hasn't used verifier's analysis result. But, we really want to have a
  1192	 * way to be able to verify the correctness of the described optimization on
  1193	 * x86_64 on which testsuites are frequently exercised.
  1194	 *
  1195	 * So, this flag is introduced. Once it is set, verifier will randomize high
  1196	 * 32-bit for those instructions who has been identified as safe to ignore them.
  1197	 * Then, if verifier is not doing correct analysis, such randomization will
  1198	 * regress tests to expose bugs.
  1199	 */
  1200	#define BPF_F_TEST_RND_HI32	(1U << 2)
  1201	
  1202	/* The verifier internal test flag. Behavior is undefined */
  1203	#define BPF_F_TEST_STATE_FREQ	(1U << 3)
  1204	
  1205	/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
  1206	 * restrict map and helper usage for such programs. Sleepable BPF programs can
  1207	 * only be attached to hooks where kernel execution context allows sleeping.
  1208	 * Such programs are allowed to use helpers that may sleep like
  1209	 * bpf_copy_from_user().
  1210	 */
  1211	#define BPF_F_SLEEPABLE		(1U << 4)
  1212	
  1213	/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
  1214	 * fully support xdp frags.
  1215	 */
  1216	#define BPF_F_XDP_HAS_FRAGS	(1U << 5)
  1217	
  1218	/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
  1219	 * program becomes device-bound but can access XDP metadata.
  1220	 */
  1221	#define BPF_F_XDP_DEV_BOUND_ONLY	(1U << 6)
  1222	
  1223	/* The verifier internal test flag. Behavior is undefined */
  1224	#define BPF_F_TEST_REG_INVARIANTS	(1U << 7)
  1225	
  1226	/* link_create.kprobe_multi.flags used in LINK_CREATE command for
  1227	 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
  1228	 */
  1229	enum {
  1230		BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
  1231	};
  1232	
  1233	/* link_create.uprobe_multi.flags used in LINK_CREATE command for
  1234	 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
  1235	 */
  1236	enum {
  1237		BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
  1238	};
  1239	
  1240	/* link_create.netfilter.flags used in LINK_CREATE command for
  1241	 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
  1242	 */
  1243	#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
  1244	
  1245	/* When BPF ldimm64's insn[0].src_reg != 0 then this can have
  1246	 * the following extensions:
  1247	 *
  1248	 * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
  1249	 * insn[0].imm:      map fd or fd_idx
  1250	 * insn[1].imm:      0
  1251	 * insn[0].off:      0
  1252	 * insn[1].off:      0
  1253	 * ldimm64 rewrite:  address of map
  1254	 * verifier type:    CONST_PTR_TO_MAP
  1255	 */
  1256	#define BPF_PSEUDO_MAP_FD	1
  1257	#define BPF_PSEUDO_MAP_IDX	5
  1258	
  1259	/* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
  1260	 * insn[0].imm:      map fd or fd_idx
  1261	 * insn[1].imm:      offset into value
  1262	 * insn[0].off:      0
  1263	 * insn[1].off:      0
  1264	 * ldimm64 rewrite:  address of map[0]+offset
  1265	 * verifier type:    PTR_TO_MAP_VALUE
  1266	 */
  1267	#define BPF_PSEUDO_MAP_VALUE		2
  1268	#define BPF_PSEUDO_MAP_IDX_VALUE	6
  1269	
  1270	/* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
  1271	 * insn[0].imm:      kernel btd id of VAR
  1272	 * insn[1].imm:      0
  1273	 * insn[0].off:      0
  1274	 * insn[1].off:      0
  1275	 * ldimm64 rewrite:  address of the kernel variable
  1276	 * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
  1277	 *                   is struct/union.
  1278	 */
  1279	#define BPF_PSEUDO_BTF_ID	3
  1280	/* insn[0].src_reg:  BPF_PSEUDO_FUNC
  1281	 * insn[0].imm:      insn offset to the func
  1282	 * insn[1].imm:      0
  1283	 * insn[0].off:      0
  1284	 * insn[1].off:      0
  1285	 * ldimm64 rewrite:  address of the function
  1286	 * verifier type:    PTR_TO_FUNC.
  1287	 */
  1288	#define BPF_PSEUDO_FUNC		4
  1289	
  1290	/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
  1291	 * offset to another bpf function
  1292	 */
  1293	#define BPF_PSEUDO_CALL		1
  1294	/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
  1295	 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
  1296	 */
  1297	#define BPF_PSEUDO_KFUNC_CALL	2
  1298	
  1299	/* flags for BPF_MAP_UPDATE_ELEM command */
  1300	enum {
  1301		BPF_ANY		= 0, /* create new element or update existing */
  1302		BPF_NOEXIST	= 1, /* create new element if it didn't exist */
  1303		BPF_EXIST	= 2, /* update existing element */
  1304		BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */
  1305	};
  1306	
  1307	/* flags for BPF_MAP_CREATE command */
  1308	enum {
  1309		BPF_F_NO_PREALLOC	= (1U << 0),
  1310	/* Instead of having one common LRU list in the
  1311	 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  1312	 * which can scale and perform better.
  1313	 * Note, the LRU nodes (including free nodes) cannot be moved
  1314	 * across different LRU lists.
  1315	 */
  1316		BPF_F_NO_COMMON_LRU	= (1U << 1),
  1317	/* Specify numa node during map creation */
  1318		BPF_F_NUMA_NODE		= (1U << 2),
  1319	
  1320	/* Flags for accessing BPF object from syscall side. */
  1321		BPF_F_RDONLY		= (1U << 3),
  1322		BPF_F_WRONLY		= (1U << 4),
  1323	
  1324	/* Flag for stack_map, store build_id+offset instead of pointer */
  1325		BPF_F_STACK_BUILD_ID	= (1U << 5),
  1326	
  1327	/* Zero-initialize hash function seed. This should only be used for testing. */
  1328		BPF_F_ZERO_SEED		= (1U << 6),
  1329	
  1330	/* Flags for accessing BPF object from program side. */
  1331		BPF_F_RDONLY_PROG	= (1U << 7),
  1332		BPF_F_WRONLY_PROG	= (1U << 8),
  1333	
  1334	/* Clone map from listener for newly accepted socket */
  1335		BPF_F_CLONE		= (1U << 9),
  1336	
  1337	/* Enable memory-mapping BPF map */
  1338		BPF_F_MMAPABLE		= (1U << 10),
  1339	
  1340	/* Share perf_event among processes */
  1341		BPF_F_PRESERVE_ELEMS	= (1U << 11),
  1342	
  1343	/* Create a map that is suitable to be an inner map with dynamic max entries */
  1344		BPF_F_INNER_MAP		= (1U << 12),
  1345	
  1346	/* Create a map that will be registered/unregesitered by the backed bpf_link */
  1347		BPF_F_LINK		= (1U << 13),
  1348	
  1349	/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
  1350		BPF_F_PATH_FD		= (1U << 14),
  1351	};
  1352	
  1353	/* Flags for BPF_PROG_QUERY. */
  1354	
  1355	/* Query effective (directly attached + inherited from ancestor cgroups)
  1356	 * programs that will be executed for events within a cgroup.
  1357	 * attach_flags with this flag are always returned 0.
  1358	 */
  1359	#define BPF_F_QUERY_EFFECTIVE	(1U << 0)
  1360	
  1361	/* Flags for BPF_PROG_TEST_RUN */
  1362	
  1363	/* If set, run the test on the cpu specified by bpf_attr.test.cpu */
  1364	#define BPF_F_TEST_RUN_ON_CPU	(1U << 0)
  1365	/* If set, XDP frames will be transmitted after processing */
  1366	#define BPF_F_TEST_XDP_LIVE_FRAMES	(1U << 1)
  1367	
  1368	/* type for BPF_ENABLE_STATS */
  1369	enum bpf_stats_type {
  1370		/* enabled run_time_ns and run_cnt */
  1371		BPF_STATS_RUN_TIME = 0,
  1372	};
  1373	
  1374	enum bpf_stack_build_id_status {
  1375		/* user space need an empty entry to identify end of a trace */
  1376		BPF_STACK_BUILD_ID_EMPTY = 0,
  1377		/* with valid build_id and offset */
  1378		BPF_STACK_BUILD_ID_VALID = 1,
  1379		/* couldn't get build_id, fallback to ip */
  1380		BPF_STACK_BUILD_ID_IP = 2,
  1381	};
  1382	
  1383	#define BPF_BUILD_ID_SIZE 20
  1384	struct bpf_stack_build_id {
  1385		__s32		status;
  1386		unsigned char	build_id[BPF_BUILD_ID_SIZE];
  1387		union {
  1388			__u64	offset;
  1389			__u64	ip;
  1390		};
  1391	};
  1392	
  1393	#define BPF_OBJ_NAME_LEN 16U
  1394	
  1395	union bpf_attr {
  1396		struct { /* anonymous struct used by BPF_MAP_CREATE command */
  1397			__u32	map_type;	/* one of enum bpf_map_type */
  1398			__u32	key_size;	/* size of key in bytes */
  1399			__u32	value_size;	/* size of value in bytes */
  1400			__u32	max_entries;	/* max number of entries in a map */
  1401			__u32	map_flags;	/* BPF_MAP_CREATE related
  1402						 * flags defined above.
  1403						 */
  1404			__u32	inner_map_fd;	/* fd pointing to the inner map */
  1405			__u32	numa_node;	/* numa node (effective only if
  1406						 * BPF_F_NUMA_NODE is set).
  1407						 */
  1408			char	map_name[BPF_OBJ_NAME_LEN];
  1409			__u32	map_ifindex;	/* ifindex of netdev to create on */
  1410			__u32	btf_fd;		/* fd pointing to a BTF type data */
  1411			__u32	btf_key_type_id;	/* BTF type_id of the key */
  1412			__u32	btf_value_type_id;	/* BTF type_id of the value */
  1413			__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
  1414							   * struct stored as the
  1415							   * map value
  1416							   */
  1417			/* Any per-map-type extra fields
  1418			 *
  1419			 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
  1420			 * number of hash functions (if 0, the bloom filter will default
  1421			 * to using 5 hash functions).
  1422			 */
  1423			__u64	map_extra;
  1424		};
  1425	
  1426		struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  1427			__u32		map_fd;
  1428			__aligned_u64	key;
  1429			union {
  1430				__aligned_u64 value;
  1431				__aligned_u64 next_key;
  1432			};
  1433			__u64		flags;
  1434		};
  1435	
  1436		struct { /* struct used by BPF_MAP_*_BATCH commands */
  1437			__aligned_u64	in_batch;	/* start batch,
  1438							 * NULL to start from beginning
  1439							 */
  1440			__aligned_u64	out_batch;	/* output: next start batch */
  1441			__aligned_u64	keys;
  1442			__aligned_u64	values;
  1443			__u32		count;		/* input/output:
  1444							 * input: # of key/value
  1445							 * elements
  1446							 * output: # of filled elements
  1447							 */
  1448			__u32		map_fd;
  1449			__u64		elem_flags;
  1450			__u64		flags;
  1451		} batch;
  1452	
  1453		struct { /* anonymous struct used by BPF_PROG_LOAD command */
  1454			__u32		prog_type;	/* one of enum bpf_prog_type */
  1455			__u32		insn_cnt;
  1456			__aligned_u64	insns;
  1457			__aligned_u64	license;
  1458			__u32		log_level;	/* verbosity level of verifier */
  1459			__u32		log_size;	/* size of user buffer */
  1460			__aligned_u64	log_buf;	/* user supplied buffer */
  1461			__u32		kern_version;	/* not used */
  1462			__u32		prog_flags;
  1463			char		prog_name[BPF_OBJ_NAME_LEN];
  1464			__u32		prog_ifindex;	/* ifindex of netdev to prep for */
  1465			/* For some prog types expected attach type must be known at
  1466			 * load time to verify attach type specific parts of prog
  1467			 * (context accesses, allowed helpers, etc).
  1468			 */
  1469			__u32		expected_attach_type;
  1470			__u32		prog_btf_fd;	/* fd pointing to BTF type data */
  1471			__u32		func_info_rec_size;	/* userspace bpf_func_info size */
  1472			__aligned_u64	func_info;	/* func info */
  1473			__u32		func_info_cnt;	/* number of bpf_func_info records */
  1474			__u32		line_info_rec_size;	/* userspace bpf_line_info size */
  1475			__aligned_u64	line_info;	/* line info */
  1476			__u32		line_info_cnt;	/* number of bpf_line_info records */
  1477			__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */
  1478			union {
  1479				/* valid prog_fd to attach to bpf prog */
  1480				__u32		attach_prog_fd;
  1481				/* or valid module BTF object fd or 0 to attach to vmlinux */
  1482				__u32		attach_btf_obj_fd;
  1483			};
  1484			__u32		core_relo_cnt;	/* number of bpf_core_relo */
  1485			__aligned_u64	fd_array;	/* array of FDs */
  1486			__aligned_u64	core_relos;
  1487			__u32		core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
  1488			/* output: actual total log contents size (including termintaing zero).
  1489			 * It could be both larger than original log_size (if log was
  1490			 * truncated), or smaller (if log buffer wasn't filled completely).
  1491			 */
  1492			__u32		log_true_size;
  1493		};
  1494	
  1495		struct { /* anonymous struct used by BPF_OBJ_* commands */
  1496			__aligned_u64	pathname;
  1497			__u32		bpf_fd;
  1498			__u32		file_flags;
  1499			/* Same as dirfd in openat() syscall; see openat(2)
  1500			 * manpage for details of path FD and pathname semantics;
  1501			 * path_fd should accompanied by BPF_F_PATH_FD flag set in
  1502			 * file_flags field, otherwise it should be set to zero;
  1503			 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
  1504			 */
  1505			__s32		path_fd;
  1506		};
  1507	
  1508		struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  1509			union {
  1510				__u32	target_fd;	/* target object to attach to or ... */
  1511				__u32	target_ifindex;	/* target ifindex */
  1512			};
  1513			__u32		attach_bpf_fd;
  1514			__u32		attach_type;
  1515			__u32		attach_flags;
  1516			__u32		replace_bpf_fd;
  1517			union {
  1518				__u32	relative_fd;
  1519				__u32	relative_id;
  1520			};
  1521			__u64		expected_revision;
  1522		};
  1523	
  1524		struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  1525			__u32		prog_fd;
  1526			__u32		retval;
  1527			__u32		data_size_in;	/* input: len of data_in */
  1528			__u32		data_size_out;	/* input/output: len of data_out
  1529							 *   returns ENOSPC if data_out
  1530							 *   is too small.
  1531							 */
  1532			__aligned_u64	data_in;
  1533			__aligned_u64	data_out;
  1534			__u32		repeat;
  1535			__u32		duration;
  1536			__u32		ctx_size_in;	/* input: len of ctx_in */
  1537			__u32		ctx_size_out;	/* input/output: len of ctx_out
  1538							 *   returns ENOSPC if ctx_out
  1539							 *   is too small.
  1540							 */
  1541			__aligned_u64	ctx_in;
  1542			__aligned_u64	ctx_out;
  1543			__u32		flags;
  1544			__u32		cpu;
  1545			__u32		batch_size;
  1546		} test;
  1547	
  1548		struct { /* anonymous struct used by BPF_*_GET_*_ID */
  1549			union {
  1550				__u32		start_id;
  1551				__u32		prog_id;
  1552				__u32		map_id;
  1553				__u32		btf_id;
  1554				__u32		link_id;
  1555			};
  1556			__u32		next_id;
  1557			__u32		open_flags;
  1558		};
  1559	
  1560		struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  1561			__u32		bpf_fd;
  1562			__u32		info_len;
  1563			__aligned_u64	info;
  1564		} info;
  1565	
  1566		struct { /* anonymous struct used by BPF_PROG_QUERY command */
  1567			union {
  1568				__u32	target_fd;	/* target object to query or ... */
  1569				__u32	target_ifindex;	/* target ifindex */
  1570			};
  1571			__u32		attach_type;
  1572			__u32		query_flags;
  1573			__u32		attach_flags;
  1574			__aligned_u64	prog_ids;
  1575			union {
  1576				__u32	prog_cnt;
  1577				__u32	count;
  1578			};
  1579			__u32		:32;
  1580			/* output: per-program attach_flags.
  1581			 * not allowed to be set during effective query.
  1582			 */
  1583			__aligned_u64	prog_attach_flags;
  1584			__aligned_u64	link_ids;
  1585			__aligned_u64	link_attach_flags;
  1586			__u64		revision;
  1587		} query;
  1588	
  1589		struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
  1590			__u64 name;
  1591			__u32 prog_fd;
  1592		} raw_tracepoint;
  1593	
  1594		struct { /* anonymous struct for BPF_BTF_LOAD */
  1595			__aligned_u64	btf;
  1596			__aligned_u64	btf_log_buf;
  1597			__u32		btf_size;
  1598			__u32		btf_log_size;
  1599			__u32		btf_log_level;
  1600			/* output: actual total log contents size (including termintaing zero).
  1601			 * It could be both larger than original log_size (if log was
  1602			 * truncated), or smaller (if log buffer wasn't filled completely).
  1603			 */
  1604			__u32		btf_log_true_size;
  1605		};
  1606	
  1607		struct {
  1608			__u32		pid;		/* input: pid */
  1609			__u32		fd;		/* input: fd */
  1610			__u32		flags;		/* input: flags */
  1611			__u32		buf_len;	/* input/output: buf len */
  1612			__aligned_u64	buf;		/* input/output:
  1613							 *   tp_name for tracepoint
  1614							 *   symbol for kprobe
  1615							 *   filename for uprobe
  1616							 */
  1617			__u32		prog_id;	/* output: prod_id */
  1618			__u32		fd_type;	/* output: BPF_FD_TYPE_* */
  1619			__u64		probe_offset;	/* output: probe_offset */
  1620			__u64		probe_addr;	/* output: probe_addr */
  1621		} task_fd_query;
  1622	
  1623		struct { /* struct used by BPF_LINK_CREATE command */
  1624			union {
  1625				__u32		prog_fd;	/* eBPF program to attach */
  1626				__u32		map_fd;		/* struct_ops to attach */
  1627			};
  1628			union {
  1629				__u32	target_fd;	/* target object to attach to or ... */
  1630				__u32	target_ifindex; /* target ifindex */
  1631			};
  1632			__u32		attach_type;	/* attach type */
  1633			__u32		flags;		/* extra flags */
  1634			union {
  1635				__u32	target_btf_id;	/* btf_id of target to attach to */
  1636				struct {
  1637					__aligned_u64	iter_info;	/* extra bpf_iter_link_info */
  1638					__u32		iter_info_len;	/* iter_info length */
  1639				};
  1640				struct {
  1641					/* black box user-provided value passed through
  1642					 * to BPF program at the execution time and
  1643					 * accessible through bpf_get_attach_cookie() BPF helper
  1644					 */
  1645					__u64		bpf_cookie;
  1646				} perf_event;
  1647				struct {
  1648					__u32		flags;
  1649					__u32		cnt;
  1650					__aligned_u64	syms;
  1651					__aligned_u64	addrs;
  1652					__aligned_u64	cookies;
  1653				} kprobe_multi;
  1654				struct {
  1655					/* this is overlaid with the target_btf_id above. */
  1656					__u32		target_btf_id;
  1657					/* black box user-provided value passed through
  1658					 * to BPF program at the execution time and
  1659					 * accessible through bpf_get_attach_cookie() BPF helper
  1660					 */
  1661					__u64		cookie;
  1662				} tracing;
  1663				struct {
  1664					__u32		pf;
  1665					__u32		hooknum;
  1666					__s32		priority;
  1667					__u32		flags;
  1668				} netfilter;
  1669				struct {
  1670					union {
  1671						__u32	relative_fd;
  1672						__u32	relative_id;
  1673					};
  1674					__u64		expected_revision;
  1675				} tcx;
  1676				struct {
  1677					__aligned_u64	path;
  1678					__aligned_u64	offsets;
  1679					__aligned_u64	ref_ctr_offsets;
  1680					__aligned_u64	cookies;
  1681					__u32		cnt;
  1682					__u32		flags;
  1683					__u32		pid;
  1684				} uprobe_multi;
  1685				struct {
  1686					union {
  1687						__u32	relative_fd;
  1688						__u32	relative_id;
  1689					};
  1690					__u64		expected_revision;
  1691				} netkit;
  1692			};
  1693		} link_create;
  1694	
  1695		struct { /* struct used by BPF_LINK_UPDATE command */
  1696			__u32		link_fd;	/* link fd */
  1697			union {
  1698				/* new program fd to update link with */
  1699				__u32		new_prog_fd;
  1700				/* new struct_ops map fd to update link with */
  1701				__u32           new_map_fd;
  1702			};
  1703			__u32		flags;		/* extra flags */
  1704			union {
  1705				/* expected link's program fd; is specified only if
  1706				 * BPF_F_REPLACE flag is set in flags.
  1707				 */
  1708				__u32		old_prog_fd;
  1709				/* expected link's map fd; is specified only
  1710				 * if BPF_F_REPLACE flag is set.
  1711				 */
  1712				__u32           old_map_fd;
  1713			};
  1714		} link_update;
  1715	
  1716		struct {
  1717			__u32		link_fd;
  1718		} link_detach;
  1719	
  1720		struct { /* struct used by BPF_ENABLE_STATS command */
  1721			__u32		type;
  1722		} enable_stats;
  1723	
  1724		struct { /* struct used by BPF_ITER_CREATE command */
  1725			__u32		link_fd;
  1726			__u32		flags;
  1727		} iter_create;
  1728	
  1729		struct { /* struct used by BPF_PROG_BIND_MAP command */
  1730			__u32		prog_fd;
  1731			__u32		map_fd;
  1732			__u32		flags;		/* extra flags */
  1733		} prog_bind_map;
  1734	
  1735	} __attribute__((aligned(8)));
  1736	
  1737	/* The description below is an attempt at providing documentation to eBPF
  1738	 * developers about the multiple available eBPF helper functions. It can be
  1739	 * parsed and used to produce a manual page. The workflow is the following,
  1740	 * and requires the rst2man utility:
  1741	 *
  1742	 *     $ ./scripts/bpf_doc.py \
  1743	 *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
  1744	 *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
  1745	 *     $ man /tmp/bpf-helpers.7
  1746	 *
  1747	 * Note that in order to produce this external documentation, some RST
  1748	 * formatting is used in the descriptions to get "bold" and "italics" in
  1749	 * manual pages. Also note that the few trailing white spaces are
  1750	 * intentional, removing them would break paragraphs for rst2man.
  1751	 *
  1752	 * Start of BPF helper function descriptions:
  1753	 *
  1754	 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
  1755	 * 	Description
  1756	 * 		Perform a lookup in *map* for an entry associated to *key*.
  1757	 * 	Return
  1758	 * 		Map value associated to *key*, or **NULL** if no entry was
  1759	 * 		found.
  1760	 *
  1761	 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
  1762	 * 	Description
  1763	 * 		Add or update the value of the entry associated to *key* in
  1764	 * 		*map* with *value*. *flags* is one of:
  1765	 *
  1766	 * 		**BPF_NOEXIST**
  1767	 * 			The entry for *key* must not exist in the map.
  1768	 * 		**BPF_EXIST**
  1769	 * 			The entry for *key* must already exist in the map.
  1770	 * 		**BPF_ANY**
  1771	 * 			No condition on the existence of the entry for *key*.
  1772	 *
  1773	 * 		Flag value **BPF_NOEXIST** cannot be used for maps of types
  1774	 * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
  1775	 * 		elements always exist), the helper would return an error.
  1776	 * 	Return
  1777	 * 		0 on success, or a negative error in case of failure.
  1778	 *
  1779	 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
  1780	 * 	Description
  1781	 * 		Delete entry with *key* from *map*.
  1782	 * 	Return
  1783	 * 		0 on success, or a negative error in case of failure.
  1784	 *
  1785	 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
  1786	 * 	Description
  1787	 * 		For tracing programs, safely attempt to read *size* bytes from
  1788	 * 		kernel space address *unsafe_ptr* and store the data in *dst*.
  1789	 *
  1790	 * 		Generally, use **bpf_probe_read_user**\ () or
  1791	 * 		**bpf_probe_read_kernel**\ () instead.
  1792	 * 	Return
  1793	 * 		0 on success, or a negative error in case of failure.
  1794	 *
  1795	 * u64 bpf_ktime_get_ns(void)
  1796	 * 	Description
  1797	 * 		Return the time elapsed since system boot, in nanoseconds.
  1798	 * 		Does not include time the system was suspended.
  1799	 * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
  1800	 * 	Return
  1801	 * 		Current *ktime*.
  1802	 *
  1803	 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
  1804	 * 	Description
  1805	 * 		This helper is a "printk()-like" facility for debugging. It
  1806	 * 		prints a message defined by format *fmt* (of size *fmt_size*)
  1807	 * 		to file *\/sys/kernel/tracing/trace* from TraceFS, if
  1808	 * 		available. It can take up to three additional **u64**
  1809	 * 		arguments (as an eBPF helpers, the total number of arguments is
  1810	 * 		limited to five).
  1811	 *
  1812	 * 		Each time the helper is called, it appends a line to the trace.
  1813	 * 		Lines are discarded while *\/sys/kernel/tracing/trace* is
  1814	 * 		open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
  1815	 * 		The format of the trace is customizable, and the exact output
  1816	 * 		one will get depends on the options set in
  1817	 * 		*\/sys/kernel/tracing/trace_options* (see also the
  1818	 * 		*README* file under the same directory). However, it usually
  1819	 * 		defaults to something like:
  1820	 *
  1821	 * 		::
  1822	 *
  1823	 * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
  1824	 *
  1825	 * 		In the above:
  1826	 *
  1827	 * 			* ``telnet`` is the name of the current task.
  1828	 * 			* ``470`` is the PID of the current task.
  1829	 * 			* ``001`` is the CPU number on which the task is
  1830	 * 			  running.
  1831	 * 			* In ``.N..``, each character refers to a set of
  1832	 * 			  options (whether irqs are enabled, scheduling
  1833	 * 			  options, whether hard/softirqs are running, level of
  1834	 * 			  preempt_disabled respectively). **N** means that
  1835	 * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
  1836	 * 			  are set.
  1837	 * 			* ``419421.045894`` is a timestamp.
  1838	 * 			* ``0x00000001`` is a fake value used by BPF for the
  1839	 * 			  instruction pointer register.
  1840	 * 			* ``<formatted msg>`` is the message formatted with
  1841	 * 			  *fmt*.
  1842	 *
  1843	 * 		The conversion specifiers supported by *fmt* are similar, but
  1844	 * 		more limited than for printk(). They are **%d**, **%i**,
  1845	 * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
  1846	 * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
  1847	 * 		of field, padding with zeroes, etc.) is available, and the
  1848	 * 		helper will return **-EINVAL** (but print nothing) if it
  1849	 * 		encounters an unknown specifier.
  1850	 *
  1851	 * 		Also, note that **bpf_trace_printk**\ () is slow, and should
  1852	 * 		only be used for debugging purposes. For this reason, a notice
  1853	 * 		block (spanning several lines) is printed to kernel logs and
  1854	 * 		states that the helper should not be used "for production use"
  1855	 * 		the first time this helper is used (or more precisely, when
  1856	 * 		**trace_printk**\ () buffers are allocated). For passing values
  1857	 * 		to user space, perf events should be preferred.
  1858	 * 	Return
  1859	 * 		The number of bytes written to the buffer, or a negative error
  1860	 * 		in case of failure.
  1861	 *
  1862	 * u32 bpf_get_prandom_u32(void)
  1863	 * 	Description
  1864	 * 		Get a pseudo-random number.
  1865	 *
  1866	 * 		From a security point of view, this helper uses its own
  1867	 * 		pseudo-random internal state, and cannot be used to infer the
  1868	 * 		seed of other random functions in the kernel. However, it is
  1869	 * 		essential to note that the generator used by the helper is not
  1870	 * 		cryptographically secure.
  1871	 * 	Return
  1872	 * 		A random 32-bit unsigned value.
  1873	 *
  1874	 * u32 bpf_get_smp_processor_id(void)
  1875	 * 	Description
  1876	 * 		Get the SMP (symmetric multiprocessing) processor id. Note that
  1877	 * 		all programs run with migration disabled, which means that the
  1878	 * 		SMP processor id is stable during all the execution of the
  1879	 * 		program.
  1880	 * 	Return
  1881	 * 		The SMP id of the processor running the program.
  1882	 *
  1883	 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
  1884	 * 	Description
  1885	 * 		Store *len* bytes from address *from* into the packet
  1886	 * 		associated to *skb*, at *offset*. *flags* are a combination of
  1887	 * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the
  1888	 * 		checksum for the packet after storing the bytes) and
  1889	 * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
  1890	 * 		**->swhash** and *skb*\ **->l4hash** to 0).
  1891	 *
  1892	 * 		A call to this helper is susceptible to change the underlying
  1893	 * 		packet buffer. Therefore, at load time, all checks on pointers
  1894	 * 		previously done by the verifier are invalidated and must be
  1895	 * 		performed again, if the helper is used in combination with
  1896	 * 		direct packet access.
  1897	 * 	Return
  1898	 * 		0 on success, or a negative error in case of failure.
  1899	 *
  1900	 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
  1901	 * 	Description
  1902	 * 		Recompute the layer 3 (e.g. IP) checksum for the packet
  1903	 * 		associated to *skb*. Computation is incremental, so the helper
  1904	 * 		must know the former value of the header field that was
  1905	 * 		modified (*from*), the new value of this field (*to*), and the
  1906	 * 		number of bytes (2 or 4) for this field, stored in *size*.
  1907	 * 		Alternatively, it is possible to store the difference between
  1908	 * 		the previous and the new values of the header field in *to*, by
  1909	 * 		setting *from* and *size* to 0. For both methods, *offset*
  1910	 * 		indicates the location of the IP checksum within the packet.
  1911	 *
  1912	 * 		This helper works in combination with **bpf_csum_diff**\ (),
  1913	 * 		which does not update the checksum in-place, but offers more
  1914	 * 		flexibility and can handle sizes larger than 2 or 4 for the
  1915	 * 		checksum to update.
  1916	 *
  1917	 * 		A call to this helper is susceptible to change the underlying
  1918	 * 		packet buffer. Therefore, at load time, all checks on pointers
  1919	 * 		previously done by the verifier are invalidated and must be
  1920	 * 		performed again, if the helper is used in combination with
  1921	 * 		direct packet access.
  1922	 * 	Return
  1923	 * 		0 on success, or a negative error in case of failure.
  1924	 *
  1925	 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
  1926	 * 	Description
  1927	 * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
  1928	 * 		packet associated to *skb*. Computation is incremental, so the
  1929	 * 		helper must know the former value of the header field that was
  1930	 * 		modified (*from*), the new value of this field (*to*), and the
  1931	 * 		number of bytes (2 or 4) for this field, stored on the lowest
  1932	 * 		four bits of *flags*. Alternatively, it is possible to store
  1933	 * 		the difference between the previous and the new values of the
  1934	 * 		header field in *to*, by setting *from* and the four lowest
  1935	 * 		bits of *flags* to 0. For both methods, *offset* indicates the
  1936	 * 		location of the IP checksum within the packet. In addition to
  1937	 * 		the size of the field, *flags* can be added (bitwise OR) actual
  1938	 * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
  1939	 * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
  1940	 * 		for updates resulting in a null checksum the value is set to
  1941	 * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
  1942	 * 		the checksum is to be computed against a pseudo-header.
  1943	 * 		Flag **BPF_F_IPV6** should be set for IPv6 packets.
  1944	 *
  1945	 * 		This helper works in combination with **bpf_csum_diff**\ (),
  1946	 * 		which does not update the checksum in-place, but offers more
  1947	 * 		flexibility and can handle sizes larger than 2 or 4 for the
  1948	 * 		checksum to update.
  1949	 *
  1950	 * 		A call to this helper is susceptible to change the underlying
  1951	 * 		packet buffer. Therefore, at load time, all checks on pointers
  1952	 * 		previously done by the verifier are invalidated and must be
  1953	 * 		performed again, if the helper is used in combination with
  1954	 * 		direct packet access.
  1955	 * 	Return
  1956	 * 		0 on success, or a negative error in case of failure.
  1957	 *
  1958	 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
  1959	 * 	Description
  1960	 * 		This special helper is used to trigger a "tail call", or in
  1961	 * 		other words, to jump into another eBPF program. The same stack
  1962	 * 		frame is used (but values on stack and in registers for the
  1963	 * 		caller are not accessible to the callee). This mechanism allows
  1964	 * 		for program chaining, either for raising the maximum number of
  1965	 * 		available eBPF instructions, or to execute given programs in
  1966	 * 		conditional blocks. For security reasons, there is an upper
  1967	 * 		limit to the number of successive tail calls that can be
  1968	 * 		performed.
  1969	 *
  1970	 * 		Upon call of this helper, the program attempts to jump into a
  1971	 * 		program referenced at index *index* in *prog_array_map*, a
  1972	 * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
  1973	 * 		*ctx*, a pointer to the context.
  1974	 *
  1975	 * 		If the call succeeds, the kernel immediately runs the first
  1976	 * 		instruction of the new program. This is not a function call,
  1977	 * 		and it never returns to the previous program. If the call
  1978	 * 		fails, then the helper has no effect, and the caller continues
  1979	 * 		to run its subsequent instructions. A call can fail if the
  1980	 * 		destination program for the jump does not exist (i.e. *index*
  1981	 * 		is superior to the number of entries in *prog_array_map*), or
  1982	 * 		if the maximum number of tail calls has been reached for this
  1983	 * 		chain of programs. This limit is defined in the kernel by the
  1984	 * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
  1985	 *		which is currently set to 33.
  1986	 * 	Return
  1987	 * 		0 on success, or a negative error in case of failure.
  1988	 *
  1989	 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
  1990	 * 	Description
  1991	 * 		Clone and redirect the packet associated to *skb* to another
  1992	 * 		net device of index *ifindex*. Both ingress and egress
  1993	 * 		interfaces can be used for redirection. The **BPF_F_INGRESS**
  1994	 * 		value in *flags* is used to make the distinction (ingress path
  1995	 * 		is selected if the flag is present, egress path otherwise).
  1996	 * 		This is the only flag supported for now.
  1997	 *
  1998	 * 		In comparison with **bpf_redirect**\ () helper,
  1999	 * 		**bpf_clone_redirect**\ () has the associated cost of
  2000	 * 		duplicating the packet buffer, but this can be executed out of
  2001	 * 		the eBPF program. Conversely, **bpf_redirect**\ () is more
  2002	 * 		efficient, but it is handled through an action code where the
  2003	 * 		redirection happens only after the eBPF program has returned.
  2004	 *
  2005	 * 		A call to this helper is susceptible to change the underlying
  2006	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2007	 * 		previously done by the verifier are invalidated and must be
  2008	 * 		performed again, if the helper is used in combination with
  2009	 * 		direct packet access.
  2010	 * 	Return
  2011	 * 		0 on success, or a negative error in case of failure. Positive
  2012	 * 		error indicates a potential drop or congestion in the target
  2013	 * 		device. The particular positive error codes are not defined.
  2014	 *
  2015	 * u64 bpf_get_current_pid_tgid(void)
  2016	 * 	Description
  2017	 * 		Get the current pid and tgid.
  2018	 * 	Return
  2019	 * 		A 64-bit integer containing the current tgid and pid, and
  2020	 * 		created as such:
  2021	 * 		*current_task*\ **->tgid << 32 \|**
  2022	 * 		*current_task*\ **->pid**.
  2023	 *
  2024	 * u64 bpf_get_current_uid_gid(void)
  2025	 * 	Description
  2026	 * 		Get the current uid and gid.
  2027	 * 	Return
  2028	 * 		A 64-bit integer containing the current GID and UID, and
  2029	 * 		created as such: *current_gid* **<< 32 \|** *current_uid*.
  2030	 *
  2031	 * long bpf_get_current_comm(void *buf, u32 size_of_buf)
  2032	 * 	Description
  2033	 * 		Copy the **comm** attribute of the current task into *buf* of
  2034	 * 		*size_of_buf*. The **comm** attribute contains the name of
  2035	 * 		the executable (excluding the path) for the current task. The
  2036	 * 		*size_of_buf* must be strictly positive. On success, the
  2037	 * 		helper makes sure that the *buf* is NUL-terminated. On failure,
  2038	 * 		it is filled with zeroes.
  2039	 * 	Return
  2040	 * 		0 on success, or a negative error in case of failure.
  2041	 *
  2042	 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
  2043	 * 	Description
  2044	 * 		Retrieve the classid for the current task, i.e. for the net_cls
  2045	 * 		cgroup to which *skb* belongs.
  2046	 *
  2047	 * 		This helper can be used on TC egress path, but not on ingress.
  2048	 *
  2049	 * 		The net_cls cgroup provides an interface to tag network packets
  2050	 * 		based on a user-provided identifier for all traffic coming from
  2051	 * 		the tasks belonging to the related cgroup. See also the related
  2052	 * 		kernel documentation, available from the Linux sources in file
  2053	 * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.
  2054	 *
  2055	 * 		The Linux kernel has two versions for cgroups: there are
  2056	 * 		cgroups v1 and cgroups v2. Both are available to users, who can
  2057	 * 		use a mixture of them, but note that the net_cls cgroup is for
  2058	 * 		cgroup v1 only. This makes it incompatible with BPF programs
  2059	 * 		run on cgroups, which is a cgroup-v2-only feature (a socket can
  2060	 * 		only hold data for one version of cgroups at a time).
  2061	 *
  2062	 * 		This helper is only available is the kernel was compiled with
  2063	 * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
  2064	 * 		"**y**" or to "**m**".
  2065	 * 	Return
  2066	 * 		The classid, or 0 for the default unconfigured classid.
  2067	 *
  2068	 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
  2069	 * 	Description
  2070	 * 		Push a *vlan_tci* (VLAN tag control information) of protocol
  2071	 * 		*vlan_proto* to the packet associated to *skb*, then update
  2072	 * 		the checksum. Note that if *vlan_proto* is different from
  2073	 * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
  2074	 * 		be **ETH_P_8021Q**.
  2075	 *
  2076	 * 		A call to this helper is susceptible to change the underlying
  2077	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2078	 * 		previously done by the verifier are invalidated and must be
  2079	 * 		performed again, if the helper is used in combination with
  2080	 * 		direct packet access.
  2081	 * 	Return
  2082	 * 		0 on success, or a negative error in case of failure.
  2083	 *
  2084	 * long bpf_skb_vlan_pop(struct sk_buff *skb)
  2085	 * 	Description
  2086	 * 		Pop a VLAN header from the packet associated to *skb*.
  2087	 *
  2088	 * 		A call to this helper is susceptible to change the underlying
  2089	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2090	 * 		previously done by the verifier are invalidated and must be
  2091	 * 		performed again, if the helper is used in combination with
  2092	 * 		direct packet access.
  2093	 * 	Return
  2094	 * 		0 on success, or a negative error in case of failure.
  2095	 *
  2096	 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  2097	 * 	Description
  2098	 * 		Get tunnel metadata. This helper takes a pointer *key* to an
  2099	 * 		empty **struct bpf_tunnel_key** of **size**, that will be
  2100	 * 		filled with tunnel metadata for the packet associated to *skb*.
  2101	 * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
  2102	 * 		indicates that the tunnel is based on IPv6 protocol instead of
  2103	 * 		IPv4.
  2104	 *
  2105	 * 		The **struct bpf_tunnel_key** is an object that generalizes the
  2106	 * 		principal parameters used by various tunneling protocols into a
  2107	 * 		single struct. This way, it can be used to easily make a
  2108	 * 		decision based on the contents of the encapsulation header,
  2109	 * 		"summarized" in this struct. In particular, it holds the IP
  2110	 * 		address of the remote end (IPv4 or IPv6, depending on the case)
  2111	 * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
  2112	 * 		this struct exposes the *key*\ **->tunnel_id**, which is
  2113	 * 		generally mapped to a VNI (Virtual Network Identifier), making
  2114	 * 		it programmable together with the **bpf_skb_set_tunnel_key**\
  2115	 * 		() helper.
  2116	 *
  2117	 * 		Let's imagine that the following code is part of a program
  2118	 * 		attached to the TC ingress interface, on one end of a GRE
  2119	 * 		tunnel, and is supposed to filter out all messages coming from
  2120	 * 		remote ends with IPv4 address other than 10.0.0.1:
  2121	 *
  2122	 * 		::
  2123	 *
  2124	 * 			int ret;
  2125	 * 			struct bpf_tunnel_key key = {};
  2126	 *
  2127	 * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
  2128	 * 			if (ret < 0)
  2129	 * 				return TC_ACT_SHOT;	// drop packet
  2130	 *
  2131	 * 			if (key.remote_ipv4 != 0x0a000001)
  2132	 * 				return TC_ACT_SHOT;	// drop packet
  2133	 *
  2134	 * 			return TC_ACT_OK;		// accept packet
  2135	 *
  2136	 * 		This interface can also be used with all encapsulation devices
  2137	 * 		that can operate in "collect metadata" mode: instead of having
  2138	 * 		one network device per specific configuration, the "collect
  2139	 * 		metadata" mode only requires a single device where the
  2140	 * 		configuration can be extracted from this helper.
  2141	 *
  2142	 * 		This can be used together with various tunnels such as VXLan,
  2143	 * 		Geneve, GRE or IP in IP (IPIP).
  2144	 * 	Return
  2145	 * 		0 on success, or a negative error in case of failure.
  2146	 *
  2147	 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  2148	 * 	Description
  2149	 * 		Populate tunnel metadata for packet associated to *skb.* The
  2150	 * 		tunnel metadata is set to the contents of *key*, of *size*. The
  2151	 * 		*flags* can be set to a combination of the following values:
  2152	 *
  2153	 * 		**BPF_F_TUNINFO_IPV6**
  2154	 * 			Indicate that the tunnel is based on IPv6 protocol
  2155	 * 			instead of IPv4.
  2156	 * 		**BPF_F_ZERO_CSUM_TX**
  2157	 * 			For IPv4 packets, add a flag to tunnel metadata
  2158	 * 			indicating that checksum computation should be skipped
  2159	 * 			and checksum set to zeroes.
  2160	 * 		**BPF_F_DONT_FRAGMENT**
  2161	 * 			Add a flag to tunnel metadata indicating that the
  2162	 * 			packet should not be fragmented.
  2163	 * 		**BPF_F_SEQ_NUMBER**
  2164	 * 			Add a flag to tunnel metadata indicating that a
  2165	 * 			sequence number should be added to tunnel header before
  2166	 * 			sending the packet. This flag was added for GRE
  2167	 * 			encapsulation, but might be used with other protocols
  2168	 * 			as well in the future.
  2169	 * 		**BPF_F_NO_TUNNEL_KEY**
  2170	 * 			Add a flag to tunnel metadata indicating that no tunnel
  2171	 * 			key should be set in the resulting tunnel header.
  2172	 *
  2173	 * 		Here is a typical usage on the transmit path:
  2174	 *
  2175	 * 		::
  2176	 *
  2177	 * 			struct bpf_tunnel_key key;
  2178	 * 			     populate key ...
  2179	 * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
  2180	 * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
  2181	 *
  2182	 * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()
  2183	 * 		helper for additional information.
  2184	 * 	Return
  2185	 * 		0 on success, or a negative error in case of failure.
  2186	 *
  2187	 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
  2188	 * 	Description
  2189	 * 		Read the value of a perf event counter. This helper relies on a
  2190	 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
  2191	 * 		the perf event counter is selected when *map* is updated with
  2192	 * 		perf event file descriptors. The *map* is an array whose size
  2193	 * 		is the number of available CPUs, and each cell contains a value
  2194	 * 		relative to one CPU. The value to retrieve is indicated by
  2195	 * 		*flags*, that contains the index of the CPU to look up, masked
  2196	 * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  2197	 * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
  2198	 * 		current CPU should be retrieved.
  2199	 *
  2200	 * 		Note that before Linux 4.13, only hardware perf event can be
  2201	 * 		retrieved.
  2202	 *
  2203	 * 		Also, be aware that the newer helper
  2204	 * 		**bpf_perf_event_read_value**\ () is recommended over
  2205	 * 		**bpf_perf_event_read**\ () in general. The latter has some ABI
  2206	 * 		quirks where error and counter value are used as a return code
  2207	 * 		(which is wrong to do since ranges may overlap). This issue is
  2208	 * 		fixed with **bpf_perf_event_read_value**\ (), which at the same
  2209	 * 		time provides more features over the **bpf_perf_event_read**\
  2210	 * 		() interface. Please refer to the description of
  2211	 * 		**bpf_perf_event_read_value**\ () for details.
  2212	 * 	Return
  2213	 * 		The value of the perf event counter read from the map, or a
  2214	 * 		negative error code in case of failure.
  2215	 *
  2216	 * long bpf_redirect(u32 ifindex, u64 flags)
  2217	 * 	Description
  2218	 * 		Redirect the packet to another net device of index *ifindex*.
  2219	 * 		This helper is somewhat similar to **bpf_clone_redirect**\
  2220	 * 		(), except that the packet is not cloned, which provides
  2221	 * 		increased performance.
  2222	 *
  2223	 * 		Except for XDP, both ingress and egress interfaces can be used
  2224	 * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used
  2225	 * 		to make the distinction (ingress path is selected if the flag
  2226	 * 		is present, egress path otherwise). Currently, XDP only
  2227	 * 		supports redirection to the egress interface, and accepts no
  2228	 * 		flag at all.
  2229	 *
  2230	 * 		The same effect can also be attained with the more generic
  2231	 * 		**bpf_redirect_map**\ (), which uses a BPF map to store the
  2232	 * 		redirect target instead of providing it directly to the helper.
  2233	 * 	Return
  2234	 * 		For XDP, the helper returns **XDP_REDIRECT** on success or
  2235	 * 		**XDP_ABORTED** on error. For other program types, the values
  2236	 * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
  2237	 * 		error.
  2238	 *
  2239	 * u32 bpf_get_route_realm(struct sk_buff *skb)
  2240	 * 	Description
  2241	 * 		Retrieve the realm or the route, that is to say the
  2242	 * 		**tclassid** field of the destination for the *skb*. The
  2243	 * 		identifier retrieved is a user-provided tag, similar to the
  2244	 * 		one used with the net_cls cgroup (see description for
  2245	 * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is
  2246	 * 		held by a route (a destination entry), not by a task.
  2247	 *
  2248	 * 		Retrieving this identifier works with the clsact TC egress hook
  2249	 * 		(see also **tc-bpf(8)**), or alternatively on conventional
  2250	 * 		classful egress qdiscs, but not on TC ingress path. In case of
  2251	 * 		clsact TC egress hook, this has the advantage that, internally,
  2252	 * 		the destination entry has not been dropped yet in the transmit
  2253	 * 		path. Therefore, the destination entry does not need to be
  2254	 * 		artificially held via **netif_keep_dst**\ () for a classful
  2255	 * 		qdisc until the *skb* is freed.
  2256	 *
  2257	 * 		This helper is available only if the kernel was compiled with
  2258	 * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.
  2259	 * 	Return
  2260	 * 		The realm of the route for the packet associated to *skb*, or 0
  2261	 * 		if none was found.
  2262	 *
  2263	 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  2264	 * 	Description
  2265	 * 		Write raw *data* blob into a special BPF perf event held by
  2266	 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  2267	 * 		event must have the following attributes: **PERF_SAMPLE_RAW**
  2268	 * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  2269	 * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  2270	 *
  2271	 * 		The *flags* are used to indicate the index in *map* for which
  2272	 * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
  2273	 * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  2274	 * 		to indicate that the index of the current CPU core should be
  2275	 * 		used.
  2276	 *
  2277	 * 		The value to write, of *size*, is passed through eBPF stack and
  2278	 * 		pointed by *data*.
  2279	 *
  2280	 * 		The context of the program *ctx* needs also be passed to the
  2281	 * 		helper.
  2282	 *
  2283	 * 		On user space, a program willing to read the values needs to
  2284	 * 		call **perf_event_open**\ () on the perf event (either for
  2285	 * 		one or for all CPUs) and to store the file descriptor into the
  2286	 * 		*map*. This must be done before the eBPF program can send data
  2287	 * 		into it. An example is available in file
  2288	 * 		*samples/bpf/trace_output_user.c* in the Linux kernel source
  2289	 * 		tree (the eBPF program counterpart is in
  2290	 * 		*samples/bpf/trace_output_kern.c*).
  2291	 *
  2292	 * 		**bpf_perf_event_output**\ () achieves better performance
  2293	 * 		than **bpf_trace_printk**\ () for sharing data with user
  2294	 * 		space, and is much better suitable for streaming data from eBPF
  2295	 * 		programs.
  2296	 *
  2297	 * 		Note that this helper is not restricted to tracing use cases
  2298	 * 		and can be used with programs attached to TC or XDP as well,
  2299	 * 		where it allows for passing data to user space listeners. Data
  2300	 * 		can be:
  2301	 *
  2302	 * 		* Only custom structs,
  2303	 * 		* Only the packet payload, or
  2304	 * 		* A combination of both.
  2305	 * 	Return
  2306	 * 		0 on success, or a negative error in case of failure.
  2307	 *
  2308	 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
  2309	 * 	Description
  2310	 * 		This helper was provided as an easy way to load data from a
  2311	 * 		packet. It can be used to load *len* bytes from *offset* from
  2312	 * 		the packet associated to *skb*, into the buffer pointed by
  2313	 * 		*to*.
  2314	 *
  2315	 * 		Since Linux 4.7, usage of this helper has mostly been replaced
  2316	 * 		by "direct packet access", enabling packet data to be
  2317	 * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**
  2318	 * 		pointing respectively to the first byte of packet data and to
  2319	 * 		the byte after the last byte of packet data. However, it
  2320	 * 		remains useful if one wishes to read large quantities of data
  2321	 * 		at once from a packet into the eBPF stack.
  2322	 * 	Return
  2323	 * 		0 on success, or a negative error in case of failure.
  2324	 *
  2325	 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
  2326	 * 	Description
  2327	 * 		Walk a user or a kernel stack and return its id. To achieve
  2328	 * 		this, the helper needs *ctx*, which is a pointer to the context
  2329	 * 		on which the tracing program is executed, and a pointer to a
  2330	 * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.
  2331	 *
  2332	 * 		The last argument, *flags*, holds the number of stack frames to
  2333	 * 		skip (from 0 to 255), masked with
  2334	 * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  2335	 * 		a combination of the following flags:
  2336	 *
  2337	 * 		**BPF_F_USER_STACK**
  2338	 * 			Collect a user space stack instead of a kernel stack.
  2339	 * 		**BPF_F_FAST_STACK_CMP**
  2340	 * 			Compare stacks by hash only.
  2341	 * 		**BPF_F_REUSE_STACKID**
  2342	 * 			If two different stacks hash into the same *stackid*,
  2343	 * 			discard the old one.
  2344	 *
  2345	 * 		The stack id retrieved is a 32 bit long integer handle which
  2346	 * 		can be further combined with other data (including other stack
  2347	 * 		ids) and used as a key into maps. This can be useful for
  2348	 * 		generating a variety of graphs (such as flame graphs or off-cpu
  2349	 * 		graphs).
  2350	 *
  2351	 * 		For walking a stack, this helper is an improvement over
  2352	 * 		**bpf_probe_read**\ (), which can be used with unrolled loops
  2353	 * 		but is not efficient and consumes a lot of eBPF instructions.
  2354	 * 		Instead, **bpf_get_stackid**\ () can collect up to
  2355	 * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
  2356	 * 		this limit can be controlled with the **sysctl** program, and
  2357	 * 		that it should be manually increased in order to profile long
  2358	 * 		user stacks (such as stacks for Java programs). To do so, use:
  2359	 *
  2360	 * 		::
  2361	 *
  2362	 * 			# sysctl kernel.perf_event_max_stack=<new value>
  2363	 * 	Return
  2364	 * 		The positive or null stack id on success, or a negative error
  2365	 * 		in case of failure.
  2366	 *
  2367	 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
  2368	 * 	Description
  2369	 * 		Compute a checksum difference, from the raw buffer pointed by
  2370	 * 		*from*, of length *from_size* (that must be a multiple of 4),
  2371	 * 		towards the raw buffer pointed by *to*, of size *to_size*
  2372	 * 		(same remark). An optional *seed* can be added to the value
  2373	 * 		(this can be cascaded, the seed may come from a previous call
  2374	 * 		to the helper).
  2375	 *
  2376	 * 		This is flexible enough to be used in several ways:
  2377	 *
  2378	 * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to
  2379	 * 		  checksum, it can be used when pushing new data.
  2380	 * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to
  2381	 * 		  checksum, it can be used when removing data from a packet.
  2382	 * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
  2383	 * 		  can be used to compute a diff. Note that *from_size* and
  2384	 * 		  *to_size* do not need to be equal.
  2385	 *
  2386	 * 		This helper can be used in combination with
  2387	 * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
  2388	 * 		which one can feed in the difference computed with
  2389	 * 		**bpf_csum_diff**\ ().
  2390	 * 	Return
  2391	 * 		The checksum result, or a negative error code in case of
  2392	 * 		failure.
  2393	 *
  2394	 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
  2395	 * 	Description
  2396	 * 		Retrieve tunnel options metadata for the packet associated to
  2397	 * 		*skb*, and store the raw tunnel option data to the buffer *opt*
  2398	 * 		of *size*.
  2399	 *
  2400	 * 		This helper can be used with encapsulation devices that can
  2401	 * 		operate in "collect metadata" mode (please refer to the related
  2402	 * 		note in the description of **bpf_skb_get_tunnel_key**\ () for
  2403	 * 		more details). A particular example where this can be used is
  2404	 * 		in combination with the Geneve encapsulation protocol, where it
  2405	 * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
  2406	 * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from
  2407	 * 		the eBPF program. This allows for full customization of these
  2408	 * 		headers.
  2409	 * 	Return
  2410	 * 		The size of the option data retrieved.
  2411	 *
  2412	 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
  2413	 * 	Description
  2414	 * 		Set tunnel options metadata for the packet associated to *skb*
  2415	 * 		to the option data contained in the raw buffer *opt* of *size*.
  2416	 *
  2417	 * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()
  2418	 * 		helper for additional information.
  2419	 * 	Return
  2420	 * 		0 on success, or a negative error in case of failure.
  2421	 *
  2422	 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
  2423	 * 	Description
  2424	 * 		Change the protocol of the *skb* to *proto*. Currently
  2425	 * 		supported are transition from IPv4 to IPv6, and from IPv6 to
  2426	 * 		IPv4. The helper takes care of the groundwork for the
  2427	 * 		transition, including resizing the socket buffer. The eBPF
  2428	 * 		program is expected to fill the new headers, if any, via
  2429	 * 		**skb_store_bytes**\ () and to recompute the checksums with
  2430	 * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
  2431	 * 		(). The main case for this helper is to perform NAT64
  2432	 * 		operations out of an eBPF program.
  2433	 *
  2434	 * 		Internally, the GSO type is marked as dodgy so that headers are
  2435	 * 		checked and segments are recalculated by the GSO/GRO engine.
  2436	 * 		The size for GSO target is adapted as well.
  2437	 *
  2438	 * 		All values for *flags* are reserved for future usage, and must
  2439	 * 		be left at zero.
  2440	 *
  2441	 * 		A call to this helper is susceptible to change the underlying
  2442	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2443	 * 		previously done by the verifier are invalidated and must be
  2444	 * 		performed again, if the helper is used in combination with
  2445	 * 		direct packet access.
  2446	 * 	Return
  2447	 * 		0 on success, or a negative error in case of failure.
  2448	 *
  2449	 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
  2450	 * 	Description
  2451	 * 		Change the packet type for the packet associated to *skb*. This
  2452	 * 		comes down to setting *skb*\ **->pkt_type** to *type*, except
  2453	 * 		the eBPF program does not have a write access to *skb*\
  2454	 * 		**->pkt_type** beside this helper. Using a helper here allows
  2455	 * 		for graceful handling of errors.
  2456	 *
  2457	 * 		The major use case is to change incoming *skb*s to
  2458	 * 		**PACKET_HOST** in a programmatic way instead of having to
  2459	 * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
  2460	 * 		example.
  2461	 *
  2462	 * 		Note that *type* only allows certain values. At this time, they
  2463	 * 		are:
  2464	 *
  2465	 * 		**PACKET_HOST**
  2466	 * 			Packet is for us.
  2467	 * 		**PACKET_BROADCAST**
  2468	 * 			Send packet to all.
  2469	 * 		**PACKET_MULTICAST**
  2470	 * 			Send packet to group.
  2471	 * 		**PACKET_OTHERHOST**
  2472	 * 			Send packet to someone else.
  2473	 * 	Return
  2474	 * 		0 on success, or a negative error in case of failure.
  2475	 *
  2476	 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
  2477	 * 	Description
  2478	 * 		Check whether *skb* is a descendant of the cgroup2 held by
  2479	 * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  2480	 * 	Return
  2481	 * 		The return value depends on the result of the test, and can be:
  2482	 *
  2483	 * 		* 0, if the *skb* failed the cgroup2 descendant test.
  2484	 * 		* 1, if the *skb* succeeded the cgroup2 descendant test.
  2485	 * 		* A negative error code, if an error occurred.
  2486	 *
  2487	 * u32 bpf_get_hash_recalc(struct sk_buff *skb)
  2488	 * 	Description
  2489	 * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is
  2490	 * 		not set, in particular if the hash was cleared due to mangling,
  2491	 * 		recompute this hash. Later accesses to the hash can be done
  2492	 * 		directly with *skb*\ **->hash**.
  2493	 *
  2494	 * 		Calling **bpf_set_hash_invalid**\ (), changing a packet
  2495	 * 		prototype with **bpf_skb_change_proto**\ (), or calling
  2496	 * 		**bpf_skb_store_bytes**\ () with the
  2497	 * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear
  2498	 * 		the hash and to trigger a new computation for the next call to
  2499	 * 		**bpf_get_hash_recalc**\ ().
  2500	 * 	Return
  2501	 * 		The 32-bit hash.
  2502	 *
  2503	 * u64 bpf_get_current_task(void)
  2504	 * 	Description
  2505	 * 		Get the current task.
  2506	 * 	Return
  2507	 * 		A pointer to the current task struct.
  2508	 *
  2509	 * long bpf_probe_write_user(void *dst, const void *src, u32 len)
  2510	 * 	Description
  2511	 * 		Attempt in a safe way to write *len* bytes from the buffer
  2512	 * 		*src* to *dst* in memory. It only works for threads that are in
  2513	 * 		user context, and *dst* must be a valid user space address.
  2514	 *
  2515	 * 		This helper should not be used to implement any kind of
  2516	 * 		security mechanism because of TOC-TOU attacks, but rather to
  2517	 * 		debug, divert, and manipulate execution of semi-cooperative
  2518	 * 		processes.
  2519	 *
  2520	 * 		Keep in mind that this feature is meant for experiments, and it
  2521	 * 		has a risk of crashing the system and running programs.
  2522	 * 		Therefore, when an eBPF program using this helper is attached,
  2523	 * 		a warning including PID and process name is printed to kernel
  2524	 * 		logs.
  2525	 * 	Return
  2526	 * 		0 on success, or a negative error in case of failure.
  2527	 *
  2528	 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
  2529	 * 	Description
  2530	 * 		Check whether the probe is being run is the context of a given
  2531	 * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by
  2532	 * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  2533	 * 	Return
  2534	 * 		The return value depends on the result of the test, and can be:
  2535	 *
  2536	 *		* 1, if current task belongs to the cgroup2.
  2537	 *		* 0, if current task does not belong to the cgroup2.
  2538	 * 		* A negative error code, if an error occurred.
  2539	 *
  2540	 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
  2541	 * 	Description
  2542	 * 		Resize (trim or grow) the packet associated to *skb* to the
  2543	 * 		new *len*. The *flags* are reserved for future usage, and must
  2544	 * 		be left at zero.
  2545	 *
  2546	 * 		The basic idea is that the helper performs the needed work to
  2547	 * 		change the size of the packet, then the eBPF program rewrites
  2548	 * 		the rest via helpers like **bpf_skb_store_bytes**\ (),
  2549	 * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
  2550	 * 		and others. This helper is a slow path utility intended for
  2551	 * 		replies with control messages. And because it is targeted for
  2552	 * 		slow path, the helper itself can afford to be slow: it
  2553	 * 		implicitly linearizes, unclones and drops offloads from the
  2554	 * 		*skb*.
  2555	 *
  2556	 * 		A call to this helper is susceptible to change the underlying
  2557	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2558	 * 		previously done by the verifier are invalidated and must be
  2559	 * 		performed again, if the helper is used in combination with
  2560	 * 		direct packet access.
  2561	 * 	Return
  2562	 * 		0 on success, or a negative error in case of failure.
  2563	 *
  2564	 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
  2565	 * 	Description
  2566	 * 		Pull in non-linear data in case the *skb* is non-linear and not
  2567	 * 		all of *len* are part of the linear section. Make *len* bytes
  2568	 * 		from *skb* readable and writable. If a zero value is passed for
  2569	 *		*len*, then all bytes in the linear part of *skb* will be made
  2570	 *		readable and writable.
  2571	 *
  2572	 * 		This helper is only needed for reading and writing with direct
  2573	 * 		packet access.
  2574	 *
  2575	 * 		For direct packet access, testing that offsets to access
  2576	 * 		are within packet boundaries (test on *skb*\ **->data_end**) is
  2577	 * 		susceptible to fail if offsets are invalid, or if the requested
  2578	 * 		data is in non-linear parts of the *skb*. On failure the
  2579	 * 		program can just bail out, or in the case of a non-linear
  2580	 * 		buffer, use a helper to make the data available. The
  2581	 * 		**bpf_skb_load_bytes**\ () helper is a first solution to access
  2582	 * 		the data. Another one consists in using **bpf_skb_pull_data**
  2583	 * 		to pull in once the non-linear parts, then retesting and
  2584	 * 		eventually access the data.
  2585	 *
  2586	 * 		At the same time, this also makes sure the *skb* is uncloned,
  2587	 * 		which is a necessary condition for direct write. As this needs
  2588	 * 		to be an invariant for the write part only, the verifier
  2589	 * 		detects writes and adds a prologue that is calling
  2590	 * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from
  2591	 * 		the very beginning in case it is indeed cloned.
  2592	 *
  2593	 * 		A call to this helper is susceptible to change the underlying
  2594	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2595	 * 		previously done by the verifier are invalidated and must be
  2596	 * 		performed again, if the helper is used in combination with
  2597	 * 		direct packet access.
  2598	 * 	Return
  2599	 * 		0 on success, or a negative error in case of failure.
  2600	 *
  2601	 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
  2602	 * 	Description
  2603	 * 		Add the checksum *csum* into *skb*\ **->csum** in case the
  2604	 * 		driver has supplied a checksum for the entire packet into that
  2605	 * 		field. Return an error otherwise. This helper is intended to be
  2606	 * 		used in combination with **bpf_csum_diff**\ (), in particular
  2607	 * 		when the checksum needs to be updated after data has been
  2608	 * 		written into the packet through direct packet access.
  2609	 * 	Return
  2610	 * 		The checksum on success, or a negative error code in case of
  2611	 * 		failure.
  2612	 *
  2613	 * void bpf_set_hash_invalid(struct sk_buff *skb)
  2614	 * 	Description
  2615	 * 		Invalidate the current *skb*\ **->hash**. It can be used after
  2616	 * 		mangling on headers through direct packet access, in order to
  2617	 * 		indicate that the hash is outdated and to trigger a
  2618	 * 		recalculation the next time the kernel tries to access this
  2619	 * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.
  2620	 * 	Return
  2621	 * 		void.
  2622	 *
  2623	 * long bpf_get_numa_node_id(void)
  2624	 * 	Description
  2625	 * 		Return the id of the current NUMA node. The primary use case
  2626	 * 		for this helper is the selection of sockets for the local NUMA
  2627	 * 		node, when the program is attached to sockets using the
  2628	 * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
  2629	 * 		but the helper is also available to other eBPF program types,
  2630	 * 		similarly to **bpf_get_smp_processor_id**\ ().
  2631	 * 	Return
  2632	 * 		The id of current NUMA node.
  2633	 *
  2634	 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
  2635	 * 	Description
  2636	 * 		Grows headroom of packet associated to *skb* and adjusts the
  2637	 * 		offset of the MAC header accordingly, adding *len* bytes of
  2638	 * 		space. It automatically extends and reallocates memory as
  2639	 * 		required.
  2640	 *
  2641	 * 		This helper can be used on a layer 3 *skb* to push a MAC header
  2642	 * 		for redirection into a layer 2 device.
  2643	 *
  2644	 * 		All values for *flags* are reserved for future usage, and must
  2645	 * 		be left at zero.
  2646	 *
  2647	 * 		A call to this helper is susceptible to change the underlying
  2648	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2649	 * 		previously done by the verifier are invalidated and must be
  2650	 * 		performed again, if the helper is used in combination with
  2651	 * 		direct packet access.
  2652	 * 	Return
  2653	 * 		0 on success, or a negative error in case of failure.
  2654	 *
  2655	 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
  2656	 * 	Description
  2657	 * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
  2658	 * 		it is possible to use a negative value for *delta*. This helper
  2659	 * 		can be used to prepare the packet for pushing or popping
  2660	 * 		headers.
  2661	 *
  2662	 * 		A call to this helper is susceptible to change the underlying
  2663	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2664	 * 		previously done by the verifier are invalidated and must be
  2665	 * 		performed again, if the helper is used in combination with
  2666	 * 		direct packet access.
  2667	 * 	Return
  2668	 * 		0 on success, or a negative error in case of failure.
  2669	 *
  2670	 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
  2671	 * 	Description
  2672	 * 		Copy a NUL terminated string from an unsafe kernel address
  2673	 * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
  2674	 * 		more details.
  2675	 *
  2676	 * 		Generally, use **bpf_probe_read_user_str**\ () or
  2677	 * 		**bpf_probe_read_kernel_str**\ () instead.
  2678	 * 	Return
  2679	 * 		On success, the strictly positive length of the string,
  2680	 * 		including the trailing NUL character. On error, a negative
  2681	 * 		value.
  2682	 *
  2683	 * u64 bpf_get_socket_cookie(struct sk_buff *skb)
  2684	 * 	Description
  2685	 * 		If the **struct sk_buff** pointed by *skb* has a known socket,
  2686	 * 		retrieve the cookie (generated by the kernel) of this socket.
  2687	 * 		If no cookie has been set yet, generate a new cookie. Once
  2688	 * 		generated, the socket cookie remains stable for the life of the
  2689	 * 		socket. This helper can be useful for monitoring per socket
  2690	 * 		networking traffic statistics as it provides a global socket
  2691	 * 		identifier that can be assumed unique.
  2692	 * 	Return
  2693	 * 		A 8-byte long unique number on success, or 0 if the socket
  2694	 * 		field is missing inside *skb*.
  2695	 *
  2696	 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
  2697	 * 	Description
  2698	 * 		Equivalent to bpf_get_socket_cookie() helper that accepts
  2699	 * 		*skb*, but gets socket from **struct bpf_sock_addr** context.
  2700	 * 	Return
  2701	 * 		A 8-byte long unique number.
  2702	 *
  2703	 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
  2704	 * 	Description
  2705	 * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
  2706	 * 		*skb*, but gets socket from **struct bpf_sock_ops** context.
  2707	 * 	Return
  2708	 * 		A 8-byte long unique number.
  2709	 *
  2710	 * u64 bpf_get_socket_cookie(struct sock *sk)
  2711	 * 	Description
  2712	 * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
  2713	 * 		*sk*, but gets socket from a BTF **struct sock**. This helper
  2714	 * 		also works for sleepable programs.
  2715	 * 	Return
  2716	 * 		A 8-byte long unique number or 0 if *sk* is NULL.
  2717	 *
  2718	 * u32 bpf_get_socket_uid(struct sk_buff *skb)
  2719	 * 	Description
  2720	 * 		Get the owner UID of the socked associated to *skb*.
  2721	 * 	Return
  2722	 * 		The owner UID of the socket associated to *skb*. If the socket
  2723	 * 		is **NULL**, or if it is not a full socket (i.e. if it is a
  2724	 * 		time-wait or a request socket instead), **overflowuid** value
  2725	 * 		is returned (note that **overflowuid** might also be the actual
  2726	 * 		UID value for the socket).
  2727	 *
  2728	 * long bpf_set_hash(struct sk_buff *skb, u32 hash)
  2729	 * 	Description
  2730	 * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)
  2731	 * 		to value *hash*.
  2732	 * 	Return
  2733	 * 		0
  2734	 *
  2735	 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
  2736	 * 	Description
  2737	 * 		Emulate a call to **setsockopt()** on the socket associated to
  2738	 * 		*bpf_socket*, which must be a full socket. The *level* at
  2739	 * 		which the option resides and the name *optname* of the option
  2740	 * 		must be specified, see **setsockopt(2)** for more information.
  2741	 * 		The option value of length *optlen* is pointed by *optval*.
  2742	 *
  2743	 * 		*bpf_socket* should be one of the following:
  2744	 *
  2745	 * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
  2746	 *		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
  2747	 *		  **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
  2748	 *
  2749	 * 		This helper actually implements a subset of **setsockopt()**.
  2750	 * 		It supports the following *level*\ s:
  2751	 *
  2752	 * 		* **SOL_SOCKET**, which supports the following *optname*\ s:
  2753	 * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
  2754	 * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
  2755	 * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
  2756	 * 		  **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
  2757	 * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:
  2758	 * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,
  2759	 * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
  2760	 * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
  2761	 * 		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
  2762	 * 		  **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
  2763	 * 		  **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
  2764	 * 		  **TCP_BPF_RTO_MIN**.
  2765	 * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  2766	 * 		* **IPPROTO_IPV6**, which supports the following *optname*\ s:
  2767	 * 		  **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
  2768	 * 	Return
  2769	 * 		0 on success, or a negative error in case of failure.
  2770	 *
  2771	 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
  2772	 * 	Description
  2773	 * 		Grow or shrink the room for data in the packet associated to
  2774	 * 		*skb* by *len_diff*, and according to the selected *mode*.
  2775	 *
  2776	 * 		By default, the helper will reset any offloaded checksum
  2777	 * 		indicator of the skb to CHECKSUM_NONE. This can be avoided
  2778	 * 		by the following flag:
  2779	 *
  2780	 * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
  2781	 * 		  checksum data of the skb to CHECKSUM_NONE.
  2782	 *
  2783	 *		There are two supported modes at this time:
  2784	 *
  2785	 *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
  2786	 * 		  (room space is added or removed between the layer 2 and
  2787	 * 		  layer 3 headers).
  2788	 *
  2789	 * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  2790	 * 		  (room space is added or removed between the layer 3 and
  2791	 * 		  layer 4 headers).
  2792	 *
  2793	 *		The following flags are supported at this time:
  2794	 *
  2795	 *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
  2796	 *		  Adjusting mss in this way is not allowed for datagrams.
  2797	 *
  2798	 *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
  2799	 *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
  2800	 *		  Any new space is reserved to hold a tunnel header.
  2801	 *		  Configure skb offsets and other fields accordingly.
  2802	 *
  2803	 *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
  2804	 *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
  2805	 *		  Use with ENCAP_L3 flags to further specify the tunnel type.
  2806	 *
  2807	 *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
  2808	 *		  Use with ENCAP_L3/L4 flags to further specify the tunnel
  2809	 *		  type; *len* is the length of the inner MAC header.
  2810	 *
  2811	 *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
  2812	 *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
  2813	 *		  L2 type as Ethernet.
  2814	 *
  2815	 *		* **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
  2816	 *		  **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
  2817	 *		  Indicate the new IP header version after decapsulating the outer
  2818	 *		  IP header. Used when the inner and outer IP versions are different.
  2819	 *
  2820	 * 		A call to this helper is susceptible to change the underlying
  2821	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2822	 * 		previously done by the verifier are invalidated and must be
  2823	 * 		performed again, if the helper is used in combination with
  2824	 * 		direct packet access.
  2825	 * 	Return
  2826	 * 		0 on success, or a negative error in case of failure.
  2827	 *
  2828	 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
  2829	 * 	Description
  2830	 * 		Redirect the packet to the endpoint referenced by *map* at
  2831	 * 		index *key*. Depending on its type, this *map* can contain
  2832	 * 		references to net devices (for forwarding packets through other
  2833	 * 		ports), or to CPUs (for redirecting XDP frames to another CPU;
  2834	 * 		but this is only implemented for native XDP (with driver
  2835	 * 		support) as of this writing).
  2836	 *
  2837	 * 		The lower two bits of *flags* are used as the return code if
  2838	 * 		the map lookup fails. This is so that the return value can be
  2839	 * 		one of the XDP program return codes up to **XDP_TX**, as chosen
  2840	 * 		by the caller. The higher bits of *flags* can be set to
  2841	 * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
  2842	 *
  2843	 * 		With BPF_F_BROADCAST the packet will be broadcasted to all the
  2844	 * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
  2845	 * 		interface will be excluded when do broadcasting.
  2846	 *
  2847	 * 		See also **bpf_redirect**\ (), which only supports redirecting
  2848	 * 		to an ifindex, but doesn't require a map to do so.
  2849	 * 	Return
  2850	 * 		**XDP_REDIRECT** on success, or the value of the two lower bits
  2851	 * 		of the *flags* argument on error.
  2852	 *
  2853	 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
  2854	 * 	Description
  2855	 * 		Redirect the packet to the socket referenced by *map* (of type
  2856	 * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  2857	 * 		egress interfaces can be used for redirection. The
  2858	 * 		**BPF_F_INGRESS** value in *flags* is used to make the
  2859	 * 		distinction (ingress path is selected if the flag is present,
  2860	 * 		egress path otherwise). This is the only flag supported for now.
  2861	 * 	Return
  2862	 * 		**SK_PASS** on success, or **SK_DROP** on error.
  2863	 *
  2864	 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  2865	 * 	Description
  2866	 * 		Add an entry to, or update a *map* referencing sockets. The
  2867	 * 		*skops* is used as a new value for the entry associated to
  2868	 * 		*key*. *flags* is one of:
  2869	 *
  2870	 * 		**BPF_NOEXIST**
  2871	 * 			The entry for *key* must not exist in the map.
  2872	 * 		**BPF_EXIST**
  2873	 * 			The entry for *key* must already exist in the map.
  2874	 * 		**BPF_ANY**
  2875	 * 			No condition on the existence of the entry for *key*.
  2876	 *
  2877	 * 		If the *map* has eBPF programs (parser and verdict), those will
  2878	 * 		be inherited by the socket being added. If the socket is
  2879	 * 		already attached to eBPF programs, this results in an error.
  2880	 * 	Return
  2881	 * 		0 on success, or a negative error in case of failure.
  2882	 *
  2883	 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
  2884	 * 	Description
  2885	 * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by
  2886	 * 		*delta* (which can be positive or negative). Note that this
  2887	 * 		operation modifies the address stored in *xdp_md*\ **->data**,
  2888	 * 		so the latter must be loaded only after the helper has been
  2889	 * 		called.
  2890	 *
  2891	 * 		The use of *xdp_md*\ **->data_meta** is optional and programs
  2892	 * 		are not required to use it. The rationale is that when the
  2893	 * 		packet is processed with XDP (e.g. as DoS filter), it is
  2894	 * 		possible to push further meta data along with it before passing
  2895	 * 		to the stack, and to give the guarantee that an ingress eBPF
  2896	 * 		program attached as a TC classifier on the same device can pick
  2897	 * 		this up for further post-processing. Since TC works with socket
  2898	 * 		buffers, it remains possible to set from XDP the **mark** or
  2899	 * 		**priority** pointers, or other pointers for the socket buffer.
  2900	 * 		Having this scratch space generic and programmable allows for
  2901	 * 		more flexibility as the user is free to store whatever meta
  2902	 * 		data they need.
  2903	 *
  2904	 * 		A call to this helper is susceptible to change the underlying
  2905	 * 		packet buffer. Therefore, at load time, all checks on pointers
  2906	 * 		previously done by the verifier are invalidated and must be
  2907	 * 		performed again, if the helper is used in combination with
  2908	 * 		direct packet access.
  2909	 * 	Return
  2910	 * 		0 on success, or a negative error in case of failure.
  2911	 *
  2912	 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
  2913	 * 	Description
  2914	 * 		Read the value of a perf event counter, and store it into *buf*
  2915	 * 		of size *buf_size*. This helper relies on a *map* of type
  2916	 * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
  2917	 * 		counter is selected when *map* is updated with perf event file
  2918	 * 		descriptors. The *map* is an array whose size is the number of
  2919	 * 		available CPUs, and each cell contains a value relative to one
  2920	 * 		CPU. The value to retrieve is indicated by *flags*, that
  2921	 * 		contains the index of the CPU to look up, masked with
  2922	 * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  2923	 * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
  2924	 * 		current CPU should be retrieved.
  2925	 *
  2926	 * 		This helper behaves in a way close to
  2927	 * 		**bpf_perf_event_read**\ () helper, save that instead of
  2928	 * 		just returning the value observed, it fills the *buf*
  2929	 * 		structure. This allows for additional data to be retrieved: in
  2930	 * 		particular, the enabled and running times (in *buf*\
  2931	 * 		**->enabled** and *buf*\ **->running**, respectively) are
  2932	 * 		copied. In general, **bpf_perf_event_read_value**\ () is
  2933	 * 		recommended over **bpf_perf_event_read**\ (), which has some
  2934	 * 		ABI issues and provides fewer functionalities.
  2935	 *
  2936	 * 		These values are interesting, because hardware PMU (Performance
  2937	 * 		Monitoring Unit) counters are limited resources. When there are
  2938	 * 		more PMU based perf events opened than available counters,
  2939	 * 		kernel will multiplex these events so each event gets certain
  2940	 * 		percentage (but not all) of the PMU time. In case that
  2941	 * 		multiplexing happens, the number of samples or counter value
  2942	 * 		will not reflect the case compared to when no multiplexing
  2943	 * 		occurs. This makes comparison between different runs difficult.
  2944	 * 		Typically, the counter value should be normalized before
  2945	 * 		comparing to other experiments. The usual normalization is done
  2946	 * 		as follows.
  2947	 *
  2948	 * 		::
  2949	 *
  2950	 * 			normalized_counter = counter * t_enabled / t_running
  2951	 *
  2952	 * 		Where t_enabled is the time enabled for event and t_running is
  2953	 * 		the time running for event since last normalization. The
  2954	 * 		enabled and running times are accumulated since the perf event
  2955	 * 		open. To achieve scaling factor between two invocations of an
  2956	 * 		eBPF program, users can use CPU id as the key (which is
  2957	 * 		typical for perf array usage model) to remember the previous
  2958	 * 		value and do the calculation inside the eBPF program.
  2959	 * 	Return
  2960	 * 		0 on success, or a negative error in case of failure.
  2961	 *
  2962	 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
  2963	 * 	Description
  2964	 * 		For an eBPF program attached to a perf event, retrieve the
  2965	 * 		value of the event counter associated to *ctx* and store it in
  2966	 * 		the structure pointed by *buf* and of size *buf_size*. Enabled
  2967	 * 		and running times are also stored in the structure (see
  2968	 * 		description of helper **bpf_perf_event_read_value**\ () for
  2969	 * 		more details).
  2970	 * 	Return
  2971	 * 		0 on success, or a negative error in case of failure.
  2972	 *
  2973	 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
  2974	 * 	Description
  2975	 * 		Emulate a call to **getsockopt()** on the socket associated to
  2976	 * 		*bpf_socket*, which must be a full socket. The *level* at
  2977	 * 		which the option resides and the name *optname* of the option
  2978	 * 		must be specified, see **getsockopt(2)** for more information.
  2979	 * 		The retrieved value is stored in the structure pointed by
  2980	 * 		*opval* and of length *optlen*.
  2981	 *
  2982	 * 		*bpf_socket* should be one of the following:
  2983	 *
  2984	 * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
  2985	 *		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,
  2986	 *		  **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.
  2987	 *
  2988	 * 		This helper actually implements a subset of **getsockopt()**.
  2989	 * 		It supports the same set of *optname*\ s that is supported by
  2990	 * 		the **bpf_setsockopt**\ () helper.  The exceptions are
  2991	 * 		**TCP_BPF_*** is **bpf_setsockopt**\ () only and
  2992	 * 		**TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
  2993	 * 	Return
  2994	 * 		0 on success, or a negative error in case of failure.
  2995	 *
  2996	 * long bpf_override_return(struct pt_regs *regs, u64 rc)
  2997	 * 	Description
  2998	 * 		Used for error injection, this helper uses kprobes to override
  2999	 * 		the return value of the probed function, and to set it to *rc*.
  3000	 * 		The first argument is the context *regs* on which the kprobe
  3001	 * 		works.
  3002	 *
  3003	 * 		This helper works by setting the PC (program counter)
  3004	 * 		to an override function which is run in place of the original
  3005	 * 		probed function. This means the probed function is not run at
  3006	 * 		all. The replacement function just returns with the required
  3007	 * 		value.
  3008	 *
  3009	 * 		This helper has security implications, and thus is subject to
  3010	 * 		restrictions. It is only available if the kernel was compiled
  3011	 * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
  3012	 * 		option, and in this case it only works on functions tagged with
  3013	 * 		**ALLOW_ERROR_INJECTION** in the kernel code.
  3014	 *
  3015	 * 		Also, the helper is only available for the architectures having
  3016	 * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
  3017	 * 		x86 architecture is the only one to support this feature.
  3018	 * 	Return
  3019	 * 		0
  3020	 *
  3021	 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
  3022	 * 	Description
  3023	 * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field
  3024	 * 		for the full TCP socket associated to *bpf_sock_ops* to
  3025	 * 		*argval*.
  3026	 *
  3027	 * 		The primary use of this field is to determine if there should
  3028	 * 		be calls to eBPF programs of type
  3029	 * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
  3030	 * 		code. A program of the same type can change its value, per
  3031	 * 		connection and as necessary, when the connection is
  3032	 * 		established. This field is directly accessible for reading, but
  3033	 * 		this helper must be used for updates in order to return an
  3034	 * 		error if an eBPF program tries to set a callback that is not
  3035	 * 		supported in the current kernel.
  3036	 *
  3037	 * 		*argval* is a flag array which can combine these flags:
  3038	 *
  3039	 * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
  3040	 * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
  3041	 * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
  3042	 * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
  3043	 *
  3044	 * 		Therefore, this function can be used to clear a callback flag by
  3045	 * 		setting the appropriate bit to zero. e.g. to disable the RTO
  3046	 * 		callback:
  3047	 *
  3048	 * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**
  3049	 * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
  3050	 *
  3051	 * 		Here are some examples of where one could call such eBPF
  3052	 * 		program:
  3053	 *
  3054	 * 		* When RTO fires.
  3055	 * 		* When a packet is retransmitted.
  3056	 * 		* When the connection terminates.
  3057	 * 		* When a packet is sent.
  3058	 * 		* When a packet is received.
  3059	 * 	Return
  3060	 * 		Code **-EINVAL** if the socket is not a full TCP socket;
  3061	 * 		otherwise, a positive number containing the bits that could not
  3062	 * 		be set is returned (which comes down to 0 if all bits were set
  3063	 * 		as required).
  3064	 *
  3065	 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
  3066	 * 	Description
  3067	 * 		This helper is used in programs implementing policies at the
  3068	 * 		socket level. If the message *msg* is allowed to pass (i.e. if
  3069	 * 		the verdict eBPF program returns **SK_PASS**), redirect it to
  3070	 * 		the socket referenced by *map* (of type
  3071	 * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  3072	 * 		egress interfaces can be used for redirection. The
  3073	 * 		**BPF_F_INGRESS** value in *flags* is used to make the
  3074	 * 		distinction (ingress path is selected if the flag is present,
  3075	 * 		egress path otherwise). This is the only flag supported for now.
  3076	 * 	Return
  3077	 * 		**SK_PASS** on success, or **SK_DROP** on error.
  3078	 *
  3079	 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
  3080	 * 	Description
  3081	 * 		For socket policies, apply the verdict of the eBPF program to
  3082	 * 		the next *bytes* (number of bytes) of message *msg*.
  3083	 *
  3084	 * 		For example, this helper can be used in the following cases:
  3085	 *
  3086	 * 		* A single **sendmsg**\ () or **sendfile**\ () system call
  3087	 * 		  contains multiple logical messages that the eBPF program is
  3088	 * 		  supposed to read and for which it should apply a verdict.
  3089	 * 		* An eBPF program only cares to read the first *bytes* of a
  3090	 * 		  *msg*. If the message has a large payload, then setting up
  3091	 * 		  and calling the eBPF program repeatedly for all bytes, even
  3092	 * 		  though the verdict is already known, would create unnecessary
  3093	 * 		  overhead.
  3094	 *
  3095	 * 		When called from within an eBPF program, the helper sets a
  3096	 * 		counter internal to the BPF infrastructure, that is used to
  3097	 * 		apply the last verdict to the next *bytes*. If *bytes* is
  3098	 * 		smaller than the current data being processed from a
  3099	 * 		**sendmsg**\ () or **sendfile**\ () system call, the first
  3100	 * 		*bytes* will be sent and the eBPF program will be re-run with
  3101	 * 		the pointer for start of data pointing to byte number *bytes*
  3102	 * 		**+ 1**. If *bytes* is larger than the current data being
  3103	 * 		processed, then the eBPF verdict will be applied to multiple
  3104	 * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are
  3105	 * 		consumed.
  3106	 *
  3107	 * 		Note that if a socket closes with the internal counter holding
  3108	 * 		a non-zero value, this is not a problem because data is not
  3109	 * 		being buffered for *bytes* and is sent as it is received.
  3110	 * 	Return
  3111	 * 		0
  3112	 *
  3113	 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
  3114	 * 	Description
  3115	 * 		For socket policies, prevent the execution of the verdict eBPF
  3116	 * 		program for message *msg* until *bytes* (byte number) have been
  3117	 * 		accumulated.
  3118	 *
  3119	 * 		This can be used when one needs a specific number of bytes
  3120	 * 		before a verdict can be assigned, even if the data spans
  3121	 * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
  3122	 * 		case would be a user calling **sendmsg**\ () repeatedly with
  3123	 * 		1-byte long message segments. Obviously, this is bad for
  3124	 * 		performance, but it is still valid. If the eBPF program needs
  3125	 * 		*bytes* bytes to validate a header, this helper can be used to
  3126	 * 		prevent the eBPF program to be called again until *bytes* have
  3127	 * 		been accumulated.
  3128	 * 	Return
  3129	 * 		0
  3130	 *
  3131	 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
  3132	 * 	Description
  3133	 * 		For socket policies, pull in non-linear data from user space
  3134	 * 		for *msg* and set pointers *msg*\ **->data** and *msg*\
  3135	 * 		**->data_end** to *start* and *end* bytes offsets into *msg*,
  3136	 * 		respectively.
  3137	 *
  3138	 * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  3139	 * 		*msg* it can only parse data that the (**data**, **data_end**)
  3140	 * 		pointers have already consumed. For **sendmsg**\ () hooks this
  3141	 * 		is likely the first scatterlist element. But for calls relying
  3142	 * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will
  3143	 * 		be the range (**0**, **0**) because the data is shared with
  3144	 * 		user space and by default the objective is to avoid allowing
  3145	 * 		user space to modify data while (or after) eBPF verdict is
  3146	 * 		being decided. This helper can be used to pull in data and to
  3147	 * 		set the start and end pointer to given values. Data will be
  3148	 * 		copied if necessary (i.e. if data was not linear and if start
  3149	 * 		and end pointers do not point to the same chunk).
  3150	 *
  3151	 * 		A call to this helper is susceptible to change the underlying
  3152	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3153	 * 		previously done by the verifier are invalidated and must be
  3154	 * 		performed again, if the helper is used in combination with
  3155	 * 		direct packet access.
  3156	 *
  3157	 * 		All values for *flags* are reserved for future usage, and must
  3158	 * 		be left at zero.
  3159	 * 	Return
  3160	 * 		0 on success, or a negative error in case of failure.
  3161	 *
  3162	 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
  3163	 * 	Description
  3164	 * 		Bind the socket associated to *ctx* to the address pointed by
  3165	 * 		*addr*, of length *addr_len*. This allows for making outgoing
  3166	 * 		connection from the desired IP address, which can be useful for
  3167	 * 		example when all processes inside a cgroup should use one
  3168	 * 		single IP address on a host that has multiple IP configured.
  3169	 *
  3170	 * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The
  3171	 * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or
  3172	 * 		**AF_INET6**). It's advised to pass zero port (**sin_port**
  3173	 * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
  3174	 * 		behavior and lets the kernel efficiently pick up an unused
  3175	 * 		port as long as 4-tuple is unique. Passing non-zero port might
  3176	 * 		lead to degraded performance.
  3177	 * 	Return
  3178	 * 		0 on success, or a negative error in case of failure.
  3179	 *
  3180	 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
  3181	 * 	Description
  3182	 * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
  3183	 * 		possible to both shrink and grow the packet tail.
  3184	 * 		Shrink done via *delta* being a negative integer.
  3185	 *
  3186	 * 		A call to this helper is susceptible to change the underlying
  3187	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3188	 * 		previously done by the verifier are invalidated and must be
  3189	 * 		performed again, if the helper is used in combination with
  3190	 * 		direct packet access.
  3191	 * 	Return
  3192	 * 		0 on success, or a negative error in case of failure.
  3193	 *
  3194	 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
  3195	 * 	Description
  3196	 * 		Retrieve the XFRM state (IP transform framework, see also
  3197	 * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
  3198	 *
  3199	 * 		The retrieved value is stored in the **struct bpf_xfrm_state**
  3200	 * 		pointed by *xfrm_state* and of length *size*.
  3201	 *
  3202	 * 		All values for *flags* are reserved for future usage, and must
  3203	 * 		be left at zero.
  3204	 *
  3205	 * 		This helper is available only if the kernel was compiled with
  3206	 * 		**CONFIG_XFRM** configuration option.
  3207	 * 	Return
  3208	 * 		0 on success, or a negative error in case of failure.
  3209	 *
  3210	 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
  3211	 * 	Description
  3212	 * 		Return a user or a kernel stack in bpf program provided buffer.
  3213	 * 		To achieve this, the helper needs *ctx*, which is a pointer
  3214	 * 		to the context on which the tracing program is executed.
  3215	 * 		To store the stacktrace, the bpf program provides *buf* with
  3216	 * 		a nonnegative *size*.
  3217	 *
  3218	 * 		The last argument, *flags*, holds the number of stack frames to
  3219	 * 		skip (from 0 to 255), masked with
  3220	 * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  3221	 * 		the following flags:
  3222	 *
  3223	 * 		**BPF_F_USER_STACK**
  3224	 * 			Collect a user space stack instead of a kernel stack.
  3225	 * 		**BPF_F_USER_BUILD_ID**
  3226	 * 			Collect (build_id, file_offset) instead of ips for user
  3227	 * 			stack, only valid if **BPF_F_USER_STACK** is also
  3228	 * 			specified.
  3229	 *
  3230	 * 			*file_offset* is an offset relative to the beginning
  3231	 * 			of the executable or shared object file backing the vma
  3232	 * 			which the *ip* falls in. It is *not* an offset relative
  3233	 * 			to that object's base address. Accordingly, it must be
  3234	 * 			adjusted by adding (sh_addr - sh_offset), where
  3235	 * 			sh_{addr,offset} correspond to the executable section
  3236	 * 			containing *file_offset* in the object, for comparisons
  3237	 * 			to symbols' st_value to be valid.
  3238	 *
  3239	 * 		**bpf_get_stack**\ () can collect up to
  3240	 * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  3241	 * 		to sufficient large buffer size. Note that
  3242	 * 		this limit can be controlled with the **sysctl** program, and
  3243	 * 		that it should be manually increased in order to profile long
  3244	 * 		user stacks (such as stacks for Java programs). To do so, use:
  3245	 *
  3246	 * 		::
  3247	 *
  3248	 * 			# sysctl kernel.perf_event_max_stack=<new value>
  3249	 * 	Return
  3250	 * 		The non-negative copied *buf* length equal to or less than
  3251	 * 		*size* on success, or a negative error in case of failure.
  3252	 *
  3253	 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
  3254	 * 	Description
  3255	 * 		This helper is similar to **bpf_skb_load_bytes**\ () in that
  3256	 * 		it provides an easy way to load *len* bytes from *offset*
  3257	 * 		from the packet associated to *skb*, into the buffer pointed
  3258	 * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that
  3259	 * 		a fifth argument *start_header* exists in order to select a
  3260	 * 		base offset to start from. *start_header* can be one of:
  3261	 *
  3262	 * 		**BPF_HDR_START_MAC**
  3263	 * 			Base offset to load data from is *skb*'s mac header.
  3264	 * 		**BPF_HDR_START_NET**
  3265	 * 			Base offset to load data from is *skb*'s network header.
  3266	 *
  3267	 * 		In general, "direct packet access" is the preferred method to
  3268	 * 		access packet data, however, this helper is in particular useful
  3269	 * 		in socket filters where *skb*\ **->data** does not always point
  3270	 * 		to the start of the mac header and where "direct packet access"
  3271	 * 		is not available.
  3272	 * 	Return
  3273	 * 		0 on success, or a negative error in case of failure.
  3274	 *
  3275	 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
  3276	 *	Description
  3277	 *		Do FIB lookup in kernel tables using parameters in *params*.
  3278	 *		If lookup is successful and result shows packet is to be
  3279	 *		forwarded, the neighbor tables are searched for the nexthop.
  3280	 *		If successful (ie., FIB lookup shows forwarding and nexthop
  3281	 *		is resolved), the nexthop address is returned in ipv4_dst
  3282	 *		or ipv6_dst based on family, smac is set to mac address of
  3283	 *		egress device, dmac is set to nexthop mac address, rt_metric
  3284	 *		is set to metric from route (IPv4/IPv6 only), and ifindex
  3285	 *		is set to the device index of the nexthop from the FIB lookup.
  3286	 *
  3287	 *		*plen* argument is the size of the passed in struct.
  3288	 *		*flags* argument can be a combination of one or more of the
  3289	 *		following values:
  3290	 *
  3291	 *		**BPF_FIB_LOOKUP_DIRECT**
  3292	 *			Do a direct table lookup vs full lookup using FIB
  3293	 *			rules.
  3294	 *		**BPF_FIB_LOOKUP_TBID**
  3295	 *			Used with BPF_FIB_LOOKUP_DIRECT.
  3296	 *			Use the routing table ID present in *params*->tbid
  3297	 *			for the fib lookup.
  3298	 *		**BPF_FIB_LOOKUP_OUTPUT**
  3299	 *			Perform lookup from an egress perspective (default is
  3300	 *			ingress).
  3301	 *		**BPF_FIB_LOOKUP_SKIP_NEIGH**
  3302	 *			Skip the neighbour table lookup. *params*->dmac
  3303	 *			and *params*->smac will not be set as output. A common
  3304	 *			use case is to call **bpf_redirect_neigh**\ () after
  3305	 *			doing **bpf_fib_lookup**\ ().
  3306	 *		**BPF_FIB_LOOKUP_SRC**
  3307	 *			Derive and set source IP addr in *params*->ipv{4,6}_src
  3308	 *			for the nexthop. If the src addr cannot be derived,
  3309	 *			**BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
  3310	 *			case, *params*->dmac and *params*->smac are not set either.
  3311	 *
  3312	 *		*ctx* is either **struct xdp_md** for XDP programs or
  3313	 *		**struct sk_buff** tc cls_act programs.
  3314	 *	Return
  3315	 *		* < 0 if any input argument is invalid
  3316	 *		*   0 on success (packet is forwarded, nexthop neighbor exists)
  3317	 *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
  3318	 *		  packet is not forwarded or needs assist from full stack
  3319	 *
  3320	 *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
  3321	 *		was exceeded and output params->mtu_result contains the MTU.
  3322	 *
  3323	 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  3324	 *	Description
  3325	 *		Add an entry to, or update a sockhash *map* referencing sockets.
  3326	 *		The *skops* is used as a new value for the entry associated to
  3327	 *		*key*. *flags* is one of:
  3328	 *
  3329	 *		**BPF_NOEXIST**
  3330	 *			The entry for *key* must not exist in the map.
  3331	 *		**BPF_EXIST**
  3332	 *			The entry for *key* must already exist in the map.
  3333	 *		**BPF_ANY**
  3334	 *			No condition on the existence of the entry for *key*.
  3335	 *
  3336	 *		If the *map* has eBPF programs (parser and verdict), those will
  3337	 *		be inherited by the socket being added. If the socket is
  3338	 *		already attached to eBPF programs, this results in an error.
  3339	 *	Return
  3340	 *		0 on success, or a negative error in case of failure.
  3341	 *
  3342	 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
  3343	 *	Description
  3344	 *		This helper is used in programs implementing policies at the
  3345	 *		socket level. If the message *msg* is allowed to pass (i.e. if
  3346	 *		the verdict eBPF program returns **SK_PASS**), redirect it to
  3347	 *		the socket referenced by *map* (of type
  3348	 *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  3349	 *		egress interfaces can be used for redirection. The
  3350	 *		**BPF_F_INGRESS** value in *flags* is used to make the
  3351	 *		distinction (ingress path is selected if the flag is present,
  3352	 *		egress path otherwise). This is the only flag supported for now.
  3353	 *	Return
  3354	 *		**SK_PASS** on success, or **SK_DROP** on error.
  3355	 *
  3356	 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
  3357	 *	Description
  3358	 *		This helper is used in programs implementing policies at the
  3359	 *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
  3360	 *		if the verdict eBPF program returns **SK_PASS**), redirect it
  3361	 *		to the socket referenced by *map* (of type
  3362	 *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  3363	 *		egress interfaces can be used for redirection. The
  3364	 *		**BPF_F_INGRESS** value in *flags* is used to make the
  3365	 *		distinction (ingress path is selected if the flag is present,
  3366	 *		egress otherwise). This is the only flag supported for now.
  3367	 *	Return
  3368	 *		**SK_PASS** on success, or **SK_DROP** on error.
  3369	 *
  3370	 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
  3371	 *	Description
  3372	 *		Encapsulate the packet associated to *skb* within a Layer 3
  3373	 *		protocol header. This header is provided in the buffer at
  3374	 *		address *hdr*, with *len* its size in bytes. *type* indicates
  3375	 *		the protocol of the header and can be one of:
  3376	 *
  3377	 *		**BPF_LWT_ENCAP_SEG6**
  3378	 *			IPv6 encapsulation with Segment Routing Header
  3379	 *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
  3380	 *			the IPv6 header is computed by the kernel.
  3381	 *		**BPF_LWT_ENCAP_SEG6_INLINE**
  3382	 *			Only works if *skb* contains an IPv6 packet. Insert a
  3383	 *			Segment Routing Header (**struct ipv6_sr_hdr**) inside
  3384	 *			the IPv6 header.
  3385	 *		**BPF_LWT_ENCAP_IP**
  3386	 *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header
  3387	 *			must be IPv4 or IPv6, followed by zero or more
  3388	 *			additional headers, up to **LWT_BPF_MAX_HEADROOM**
  3389	 *			total bytes in all prepended headers. Please note that
  3390	 *			if **skb_is_gso**\ (*skb*) is true, no more than two
  3391	 *			headers can be prepended, and the inner header, if
  3392	 *			present, should be either GRE or UDP/GUE.
  3393	 *
  3394	 *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
  3395	 *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
  3396	 *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
  3397	 *		**BPF_PROG_TYPE_LWT_XMIT**.
  3398	 *
  3399	 * 		A call to this helper is susceptible to change the underlying
  3400	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3401	 * 		previously done by the verifier are invalidated and must be
  3402	 * 		performed again, if the helper is used in combination with
  3403	 * 		direct packet access.
  3404	 *	Return
  3405	 * 		0 on success, or a negative error in case of failure.
  3406	 *
  3407	 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
  3408	 *	Description
  3409	 *		Store *len* bytes from address *from* into the packet
  3410	 *		associated to *skb*, at *offset*. Only the flags, tag and TLVs
  3411	 *		inside the outermost IPv6 Segment Routing Header can be
  3412	 *		modified through this helper.
  3413	 *
  3414	 * 		A call to this helper is susceptible to change the underlying
  3415	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3416	 * 		previously done by the verifier are invalidated and must be
  3417	 * 		performed again, if the helper is used in combination with
  3418	 * 		direct packet access.
  3419	 *	Return
  3420	 * 		0 on success, or a negative error in case of failure.
  3421	 *
  3422	 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
  3423	 *	Description
  3424	 *		Adjust the size allocated to TLVs in the outermost IPv6
  3425	 *		Segment Routing Header contained in the packet associated to
  3426	 *		*skb*, at position *offset* by *delta* bytes. Only offsets
  3427	 *		after the segments are accepted. *delta* can be as well
  3428	 *		positive (growing) as negative (shrinking).
  3429	 *
  3430	 * 		A call to this helper is susceptible to change the underlying
  3431	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3432	 * 		previously done by the verifier are invalidated and must be
  3433	 * 		performed again, if the helper is used in combination with
  3434	 * 		direct packet access.
  3435	 *	Return
  3436	 * 		0 on success, or a negative error in case of failure.
  3437	 *
  3438	 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
  3439	 *	Description
  3440	 *		Apply an IPv6 Segment Routing action of type *action* to the
  3441	 *		packet associated to *skb*. Each action takes a parameter
  3442	 *		contained at address *param*, and of length *param_len* bytes.
  3443	 *		*action* can be one of:
  3444	 *
  3445	 *		**SEG6_LOCAL_ACTION_END_X**
  3446	 *			End.X action: Endpoint with Layer-3 cross-connect.
  3447	 *			Type of *param*: **struct in6_addr**.
  3448	 *		**SEG6_LOCAL_ACTION_END_T**
  3449	 *			End.T action: Endpoint with specific IPv6 table lookup.
  3450	 *			Type of *param*: **int**.
  3451	 *		**SEG6_LOCAL_ACTION_END_B6**
  3452	 *			End.B6 action: Endpoint bound to an SRv6 policy.
  3453	 *			Type of *param*: **struct ipv6_sr_hdr**.
  3454	 *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**
  3455	 *			End.B6.Encap action: Endpoint bound to an SRv6
  3456	 *			encapsulation policy.
  3457	 *			Type of *param*: **struct ipv6_sr_hdr**.
  3458	 *
  3459	 * 		A call to this helper is susceptible to change the underlying
  3460	 * 		packet buffer. Therefore, at load time, all checks on pointers
  3461	 * 		previously done by the verifier are invalidated and must be
  3462	 * 		performed again, if the helper is used in combination with
  3463	 * 		direct packet access.
  3464	 *	Return
  3465	 * 		0 on success, or a negative error in case of failure.
  3466	 *
  3467	 * long bpf_rc_repeat(void *ctx)
  3468	 *	Description
  3469	 *		This helper is used in programs implementing IR decoding, to
  3470	 *		report a successfully decoded repeat key message. This delays
  3471	 *		the generation of a key up event for previously generated
  3472	 *		key down event.
  3473	 *
  3474	 *		Some IR protocols like NEC have a special IR message for
  3475	 *		repeating last button, for when a button is held down.
  3476	 *
  3477	 *		The *ctx* should point to the lirc sample as passed into
  3478	 *		the program.
  3479	 *
  3480	 *		This helper is only available is the kernel was compiled with
  3481	 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3482	 *		"**y**".
  3483	 *	Return
  3484	 *		0
  3485	 *
  3486	 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
  3487	 *	Description
  3488	 *		This helper is used in programs implementing IR decoding, to
  3489	 *		report a successfully decoded key press with *scancode*,
  3490	 *		*toggle* value in the given *protocol*. The scancode will be
  3491	 *		translated to a keycode using the rc keymap, and reported as
  3492	 *		an input key down event. After a period a key up event is
  3493	 *		generated. This period can be extended by calling either
  3494	 *		**bpf_rc_keydown**\ () again with the same values, or calling
  3495	 *		**bpf_rc_repeat**\ ().
  3496	 *
  3497	 *		Some protocols include a toggle bit, in case the button was
  3498	 *		released and pressed again between consecutive scancodes.
  3499	 *
  3500	 *		The *ctx* should point to the lirc sample as passed into
  3501	 *		the program.
  3502	 *
  3503	 *		The *protocol* is the decoded protocol number (see
  3504	 *		**enum rc_proto** for some predefined values).
  3505	 *
  3506	 *		This helper is only available is the kernel was compiled with
  3507	 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3508	 *		"**y**".
  3509	 *	Return
  3510	 *		0
  3511	 *
  3512	 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
  3513	 * 	Description
  3514	 * 		Return the cgroup v2 id of the socket associated with the *skb*.
  3515	 * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()
  3516	 * 		helper for cgroup v1 by providing a tag resp. identifier that
  3517	 * 		can be matched on or used for map lookups e.g. to implement
  3518	 * 		policy. The cgroup v2 id of a given path in the hierarchy is
  3519	 * 		exposed in user space through the f_handle API in order to get
  3520	 * 		to the same 64-bit id.
  3521	 *
  3522	 * 		This helper can be used on TC egress path, but not on ingress,
  3523	 * 		and is available only if the kernel was compiled with the
  3524	 * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.
  3525	 * 	Return
  3526	 * 		The id is returned or 0 in case the id could not be retrieved.
  3527	 *
  3528	 * u64 bpf_get_current_cgroup_id(void)
  3529	 * 	Description
  3530	 * 		Get the current cgroup id based on the cgroup within which
  3531	 * 		the current task is running.
  3532	 * 	Return
  3533	 * 		A 64-bit integer containing the current cgroup id based
  3534	 * 		on the cgroup within which the current task is running.
  3535	 *
  3536	 * void *bpf_get_local_storage(void *map, u64 flags)
  3537	 *	Description
  3538	 *		Get the pointer to the local storage area.
  3539	 *		The type and the size of the local storage is defined
  3540	 *		by the *map* argument.
  3541	 *		The *flags* meaning is specific for each map type,
  3542	 *		and has to be 0 for cgroup local storage.
  3543	 *
  3544	 *		Depending on the BPF program type, a local storage area
  3545	 *		can be shared between multiple instances of the BPF program,
  3546	 *		running simultaneously.
  3547	 *
  3548	 *		A user should care about the synchronization by himself.
  3549	 *		For example, by using the **BPF_ATOMIC** instructions to alter
  3550	 *		the shared data.
  3551	 *	Return
  3552	 *		A pointer to the local storage area.
  3553	 *
  3554	 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
  3555	 *	Description
  3556	 *		Select a **SO_REUSEPORT** socket from a
  3557	 *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
  3558	 *		It checks the selected socket is matching the incoming
  3559	 *		request in the socket buffer.
  3560	 *	Return
  3561	 *		0 on success, or a negative error in case of failure.
  3562	 *
  3563	 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
  3564	 *	Description
  3565	 *		Return id of cgroup v2 that is ancestor of cgroup associated
  3566	 *		with the *skb* at the *ancestor_level*.  The root cgroup is at
  3567	 *		*ancestor_level* zero and each step down the hierarchy
  3568	 *		increments the level. If *ancestor_level* == level of cgroup
  3569	 *		associated with *skb*, then return value will be same as that
  3570	 *		of **bpf_skb_cgroup_id**\ ().
  3571	 *
  3572	 *		The helper is useful to implement policies based on cgroups
  3573	 *		that are upper in hierarchy than immediate cgroup associated
  3574	 *		with *skb*.
  3575	 *
  3576	 *		The format of returned id and helper limitations are same as in
  3577	 *		**bpf_skb_cgroup_id**\ ().
  3578	 *	Return
  3579	 *		The id is returned or 0 in case the id could not be retrieved.
  3580	 *
  3581	 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3582	 *	Description
  3583	 *		Look for TCP socket matching *tuple*, optionally in a child
  3584	 *		network namespace *netns*. The return value must be checked,
  3585	 *		and if non-**NULL**, released via **bpf_sk_release**\ ().
  3586	 *
  3587	 *		The *ctx* should point to the context of the program, such as
  3588	 *		the skb or socket (depending on the hook in use). This is used
  3589	 *		to determine the base network namespace for the lookup.
  3590	 *
  3591	 *		*tuple_size* must be one of:
  3592	 *
  3593	 *		**sizeof**\ (*tuple*\ **->ipv4**)
  3594	 *			Look for an IPv4 socket.
  3595	 *		**sizeof**\ (*tuple*\ **->ipv6**)
  3596	 *			Look for an IPv6 socket.
  3597	 *
  3598	 *		If the *netns* is a negative signed 32-bit integer, then the
  3599	 *		socket lookup table in the netns associated with the *ctx*
  3600	 *		will be used. For the TC hooks, this is the netns of the device
  3601	 *		in the skb. For socket hooks, this is the netns of the socket.
  3602	 *		If *netns* is any other signed 32-bit value greater than or
  3603	 *		equal to zero then it specifies the ID of the netns relative to
  3604	 *		the netns associated with the *ctx*. *netns* values beyond the
  3605	 *		range of 32-bit integers are reserved for future use.
  3606	 *
  3607	 *		All values for *flags* are reserved for future usage, and must
  3608	 *		be left at zero.
  3609	 *
  3610	 *		This helper is available only if the kernel was compiled with
  3611	 *		**CONFIG_NET** configuration option.
  3612	 *	Return
  3613	 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3614	 *		For sockets with reuseport option, the **struct bpf_sock**
  3615	 *		result is from *reuse*\ **->socks**\ [] using the hash of the
  3616	 *		tuple.
  3617	 *
  3618	 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3619	 *	Description
  3620	 *		Look for UDP socket matching *tuple*, optionally in a child
  3621	 *		network namespace *netns*. The return value must be checked,
  3622	 *		and if non-**NULL**, released via **bpf_sk_release**\ ().
  3623	 *
  3624	 *		The *ctx* should point to the context of the program, such as
  3625	 *		the skb or socket (depending on the hook in use). This is used
  3626	 *		to determine the base network namespace for the lookup.
  3627	 *
  3628	 *		*tuple_size* must be one of:
  3629	 *
  3630	 *		**sizeof**\ (*tuple*\ **->ipv4**)
  3631	 *			Look for an IPv4 socket.
  3632	 *		**sizeof**\ (*tuple*\ **->ipv6**)
  3633	 *			Look for an IPv6 socket.
  3634	 *
  3635	 *		If the *netns* is a negative signed 32-bit integer, then the
  3636	 *		socket lookup table in the netns associated with the *ctx*
  3637	 *		will be used. For the TC hooks, this is the netns of the device
  3638	 *		in the skb. For socket hooks, this is the netns of the socket.
  3639	 *		If *netns* is any other signed 32-bit value greater than or
  3640	 *		equal to zero then it specifies the ID of the netns relative to
  3641	 *		the netns associated with the *ctx*. *netns* values beyond the
  3642	 *		range of 32-bit integers are reserved for future use.
  3643	 *
  3644	 *		All values for *flags* are reserved for future usage, and must
  3645	 *		be left at zero.
  3646	 *
  3647	 *		This helper is available only if the kernel was compiled with
  3648	 *		**CONFIG_NET** configuration option.
  3649	 *	Return
  3650	 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3651	 *		For sockets with reuseport option, the **struct bpf_sock**
  3652	 *		result is from *reuse*\ **->socks**\ [] using the hash of the
  3653	 *		tuple.
  3654	 *
  3655	 * long bpf_sk_release(void *sock)
  3656	 *	Description
  3657	 *		Release the reference held by *sock*. *sock* must be a
  3658	 *		non-**NULL** pointer that was returned from
  3659	 *		**bpf_sk_lookup_xxx**\ ().
  3660	 *	Return
  3661	 *		0 on success, or a negative error in case of failure.
  3662	 *
  3663	 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
  3664	 * 	Description
  3665	 * 		Push an element *value* in *map*. *flags* is one of:
  3666	 *
  3667	 * 		**BPF_EXIST**
  3668	 * 			If the queue/stack is full, the oldest element is
  3669	 * 			removed to make room for this.
  3670	 * 	Return
  3671	 * 		0 on success, or a negative error in case of failure.
  3672	 *
  3673	 * long bpf_map_pop_elem(struct bpf_map *map, void *value)
  3674	 * 	Description
  3675	 * 		Pop an element from *map*.
  3676	 * 	Return
  3677	 * 		0 on success, or a negative error in case of failure.
  3678	 *
  3679	 * long bpf_map_peek_elem(struct bpf_map *map, void *value)
  3680	 * 	Description
  3681	 * 		Get an element from *map* without removing it.
  3682	 * 	Return
  3683	 * 		0 on success, or a negative error in case of failure.
  3684	 *
  3685	 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
  3686	 *	Description
  3687	 *		For socket policies, insert *len* bytes into *msg* at offset
  3688	 *		*start*.
  3689	 *
  3690	 *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  3691	 *		*msg* it may want to insert metadata or options into the *msg*.
  3692	 *		This can later be read and used by any of the lower layer BPF
  3693	 *		hooks.
  3694	 *
  3695	 *		This helper may fail if under memory pressure (a malloc
  3696	 *		fails) in these cases BPF programs will get an appropriate
  3697	 *		error and BPF programs will need to handle them.
  3698	 *	Return
  3699	 *		0 on success, or a negative error in case of failure.
  3700	 *
  3701	 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
  3702	 *	Description
  3703	 *		Will remove *len* bytes from a *msg* starting at byte *start*.
  3704	 *		This may result in **ENOMEM** errors under certain situations if
  3705	 *		an allocation and copy are required due to a full ring buffer.
  3706	 *		However, the helper will try to avoid doing the allocation
  3707	 *		if possible. Other errors can occur if input parameters are
  3708	 *		invalid either due to *start* byte not being valid part of *msg*
  3709	 *		payload and/or *pop* value being to large.
  3710	 *	Return
  3711	 *		0 on success, or a negative error in case of failure.
  3712	 *
  3713	 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
  3714	 *	Description
  3715	 *		This helper is used in programs implementing IR decoding, to
  3716	 *		report a successfully decoded pointer movement.
  3717	 *
  3718	 *		The *ctx* should point to the lirc sample as passed into
  3719	 *		the program.
  3720	 *
  3721	 *		This helper is only available is the kernel was compiled with
  3722	 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  3723	 *		"**y**".
  3724	 *	Return
  3725	 *		0
  3726	 *
  3727	 * long bpf_spin_lock(struct bpf_spin_lock *lock)
  3728	 *	Description
  3729	 *		Acquire a spinlock represented by the pointer *lock*, which is
  3730	 *		stored as part of a value of a map. Taking the lock allows to
  3731	 *		safely update the rest of the fields in that value. The
  3732	 *		spinlock can (and must) later be released with a call to
  3733	 *		**bpf_spin_unlock**\ (\ *lock*\ ).
  3734	 *
  3735	 *		Spinlocks in BPF programs come with a number of restrictions
  3736	 *		and constraints:
  3737	 *
  3738	 *		* **bpf_spin_lock** objects are only allowed inside maps of
  3739	 *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
  3740	 *		  list could be extended in the future).
  3741	 *		* BTF description of the map is mandatory.
  3742	 *		* The BPF program can take ONE lock at a time, since taking two
  3743	 *		  or more could cause dead locks.
  3744	 *		* Only one **struct bpf_spin_lock** is allowed per map element.
  3745	 *		* When the lock is taken, calls (either BPF to BPF or helpers)
  3746	 *		  are not allowed.
  3747	 *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
  3748	 *		  allowed inside a spinlock-ed region.
  3749	 *		* The BPF program MUST call **bpf_spin_unlock**\ () to release
  3750	 *		  the lock, on all execution paths, before it returns.
  3751	 *		* The BPF program can access **struct bpf_spin_lock** only via
  3752	 *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
  3753	 *		  helpers. Loading or storing data into the **struct
  3754	 *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
  3755	 *		* To use the **bpf_spin_lock**\ () helper, the BTF description
  3756	 *		  of the map value must be a struct and have **struct
  3757	 *		  bpf_spin_lock** *anyname*\ **;** field at the top level.
  3758	 *		  Nested lock inside another struct is not allowed.
  3759	 *		* The **struct bpf_spin_lock** *lock* field in a map value must
  3760	 *		  be aligned on a multiple of 4 bytes in that value.
  3761	 *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
  3762	 *		  the **bpf_spin_lock** field to user space.
  3763	 *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
  3764	 *		  a BPF program, do not update the **bpf_spin_lock** field.
  3765	 *		* **bpf_spin_lock** cannot be on the stack or inside a
  3766	 *		  networking packet (it can only be inside of a map values).
  3767	 *		* **bpf_spin_lock** is available to root only.
  3768	 *		* Tracing programs and socket filter programs cannot use
  3769	 *		  **bpf_spin_lock**\ () due to insufficient preemption checks
  3770	 *		  (but this may change in the future).
  3771	 *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.
  3772	 *	Return
  3773	 *		0
  3774	 *
  3775	 * long bpf_spin_unlock(struct bpf_spin_lock *lock)
  3776	 *	Description
  3777	 *		Release the *lock* previously locked by a call to
  3778	 *		**bpf_spin_lock**\ (\ *lock*\ ).
  3779	 *	Return
  3780	 *		0
  3781	 *
  3782	 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
  3783	 *	Description
  3784	 *		This helper gets a **struct bpf_sock** pointer such
  3785	 *		that all the fields in this **bpf_sock** can be accessed.
  3786	 *	Return
  3787	 *		A **struct bpf_sock** pointer on success, or **NULL** in
  3788	 *		case of failure.
  3789	 *
  3790	 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
  3791	 *	Description
  3792	 *		This helper gets a **struct bpf_tcp_sock** pointer from a
  3793	 *		**struct bpf_sock** pointer.
  3794	 *	Return
  3795	 *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in
  3796	 *		case of failure.
  3797	 *
  3798	 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
  3799	 *	Description
  3800	 *		Set ECN (Explicit Congestion Notification) field of IP header
  3801	 *		to **CE** (Congestion Encountered) if current value is **ECT**
  3802	 *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv6
  3803	 *		and IPv4.
  3804	 *	Return
  3805	 *		1 if the **CE** flag is set (either by the current helper call
  3806	 *		or because it was already present), 0 if it is not set.
  3807	 *
  3808	 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
  3809	 *	Description
  3810	 *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
  3811	 *		**bpf_sk_release**\ () is unnecessary and not allowed.
  3812	 *	Return
  3813	 *		A **struct bpf_sock** pointer on success, or **NULL** in
  3814	 *		case of failure.
  3815	 *
  3816	 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  3817	 *	Description
  3818	 *		Look for TCP socket matching *tuple*, optionally in a child
  3819	 *		network namespace *netns*. The return value must be checked,
  3820	 *		and if non-**NULL**, released via **bpf_sk_release**\ ().
  3821	 *
  3822	 *		This function is identical to **bpf_sk_lookup_tcp**\ (), except
  3823	 *		that it also returns timewait or request sockets. Use
  3824	 *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
  3825	 *		full structure.
  3826	 *
  3827	 *		This helper is available only if the kernel was compiled with
  3828	 *		**CONFIG_NET** configuration option.
  3829	 *	Return
  3830	 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
  3831	 *		For sockets with reuseport option, the **struct bpf_sock**
  3832	 *		result is from *reuse*\ **->socks**\ [] using the hash of the
  3833	 *		tuple.
  3834	 *
  3835	 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
  3836	 * 	Description
  3837	 * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for
  3838	 * 		the listening socket in *sk*.
  3839	 *
  3840	 * 		*iph* points to the start of the IPv4 or IPv6 header, while
  3841	 * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or
  3842	 * 		**sizeof**\ (**struct ipv6hdr**).
  3843	 *
  3844	 * 		*th* points to the start of the TCP header, while *th_len*
  3845	 *		contains the length of the TCP header (at least
  3846	 *		**sizeof**\ (**struct tcphdr**)).
  3847	 * 	Return
  3848	 * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
  3849	 * 		error otherwise.
  3850	 *
  3851	 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
  3852	 *	Description
  3853	 *		Get name of sysctl in /proc/sys/ and copy it into provided by
  3854	 *		program buffer *buf* of size *buf_len*.
  3855	 *
  3856	 *		The buffer is always NUL terminated, unless it's zero-sized.
  3857	 *
  3858	 *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
  3859	 *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
  3860	 *		only (e.g. "tcp_mem").
  3861	 *	Return
  3862	 *		Number of character copied (not including the trailing NUL).
  3863	 *
  3864	 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3865	 *		truncated name in this case).
  3866	 *
  3867	 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
  3868	 *	Description
  3869	 *		Get current value of sysctl as it is presented in /proc/sys
  3870	 *		(incl. newline, etc), and copy it as a string into provided
  3871	 *		by program buffer *buf* of size *buf_len*.
  3872	 *
  3873	 *		The whole value is copied, no matter what file position user
  3874	 *		space issued e.g. sys_read at.
  3875	 *
  3876	 *		The buffer is always NUL terminated, unless it's zero-sized.
  3877	 *	Return
  3878	 *		Number of character copied (not including the trailing NUL).
  3879	 *
  3880	 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3881	 *		truncated name in this case).
  3882	 *
  3883	 *		**-EINVAL** if current value was unavailable, e.g. because
  3884	 *		sysctl is uninitialized and read returns -EIO for it.
  3885	 *
  3886	 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
  3887	 *	Description
  3888	 *		Get new value being written by user space to sysctl (before
  3889	 *		the actual write happens) and copy it as a string into
  3890	 *		provided by program buffer *buf* of size *buf_len*.
  3891	 *
  3892	 *		User space may write new value at file position > 0.
  3893	 *
  3894	 *		The buffer is always NUL terminated, unless it's zero-sized.
  3895	 *	Return
  3896	 *		Number of character copied (not including the trailing NUL).
  3897	 *
  3898	 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
  3899	 *		truncated name in this case).
  3900	 *
  3901	 *		**-EINVAL** if sysctl is being read.
  3902	 *
  3903	 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
  3904	 *	Description
  3905	 *		Override new value being written by user space to sysctl with
  3906	 *		value provided by program in buffer *buf* of size *buf_len*.
  3907	 *
  3908	 *		*buf* should contain a string in same form as provided by user
  3909	 *		space on sysctl write.
  3910	 *
  3911	 *		User space may write new value at file position > 0. To override
  3912	 *		the whole sysctl value file position should be set to zero.
  3913	 *	Return
  3914	 *		0 on success.
  3915	 *
  3916	 *		**-E2BIG** if the *buf_len* is too big.
  3917	 *
  3918	 *		**-EINVAL** if sysctl is being read.
  3919	 *
  3920	 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
  3921	 *	Description
  3922	 *		Convert the initial part of the string from buffer *buf* of
  3923	 *		size *buf_len* to a long integer according to the given base
  3924	 *		and save the result in *res*.
  3925	 *
  3926	 *		The string may begin with an arbitrary amount of white space
  3927	 *		(as determined by **isspace**\ (3)) followed by a single
  3928	 *		optional '**-**' sign.
  3929	 *
  3930	 *		Five least significant bits of *flags* encode base, other bits
  3931	 *		are currently unused.
  3932	 *
  3933	 *		Base must be either 8, 10, 16 or 0 to detect it automatically
  3934	 *		similar to user space **strtol**\ (3).
  3935	 *	Return
  3936	 *		Number of characters consumed on success. Must be positive but
  3937	 *		no more than *buf_len*.
  3938	 *
  3939	 *		**-EINVAL** if no valid digits were found or unsupported base
  3940	 *		was provided.
  3941	 *
  3942	 *		**-ERANGE** if resulting value was out of range.
  3943	 *
  3944	 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
  3945	 *	Description
  3946	 *		Convert the initial part of the string from buffer *buf* of
  3947	 *		size *buf_len* to an unsigned long integer according to the
  3948	 *		given base and save the result in *res*.
  3949	 *
  3950	 *		The string may begin with an arbitrary amount of white space
  3951	 *		(as determined by **isspace**\ (3)).
  3952	 *
  3953	 *		Five least significant bits of *flags* encode base, other bits
  3954	 *		are currently unused.
  3955	 *
  3956	 *		Base must be either 8, 10, 16 or 0 to detect it automatically
  3957	 *		similar to user space **strtoul**\ (3).
  3958	 *	Return
  3959	 *		Number of characters consumed on success. Must be positive but
  3960	 *		no more than *buf_len*.
  3961	 *
  3962	 *		**-EINVAL** if no valid digits were found or unsupported base
  3963	 *		was provided.
  3964	 *
  3965	 *		**-ERANGE** if resulting value was out of range.
  3966	 *
  3967	 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
  3968	 *	Description
  3969	 *		Get a bpf-local-storage from a *sk*.
  3970	 *
  3971	 *		Logically, it could be thought of getting the value from
  3972	 *		a *map* with *sk* as the **key**.  From this
  3973	 *		perspective,  the usage is not much different from
  3974	 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
  3975	 *		helper enforces the key must be a full socket and the map must
  3976	 *		be a **BPF_MAP_TYPE_SK_STORAGE** also.
  3977	 *
  3978	 *		Underneath, the value is stored locally at *sk* instead of
  3979	 *		the *map*.  The *map* is used as the bpf-local-storage
  3980	 *		"type". The bpf-local-storage "type" (i.e. the *map*) is
  3981	 *		searched against all bpf-local-storages residing at *sk*.
  3982	 *
  3983	 *		*sk* is a kernel **struct sock** pointer for LSM program.
  3984	 *		*sk* is a **struct bpf_sock** pointer for other program types.
  3985	 *
  3986	 *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
  3987	 *		used such that a new bpf-local-storage will be
  3988	 *		created if one does not exist.  *value* can be used
  3989	 *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
  3990	 *		the initial value of a bpf-local-storage.  If *value* is
  3991	 *		**NULL**, the new bpf-local-storage will be zero initialized.
  3992	 *	Return
  3993	 *		A bpf-local-storage pointer is returned on success.
  3994	 *
  3995	 *		**NULL** if not found or there was an error in adding
  3996	 *		a new bpf-local-storage.
  3997	 *
  3998	 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
  3999	 *	Description
  4000	 *		Delete a bpf-local-storage from a *sk*.
  4001	 *	Return
  4002	 *		0 on success.
  4003	 *
  4004	 *		**-ENOENT** if the bpf-local-storage cannot be found.
  4005	 *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).
  4006	 *
  4007	 * long bpf_send_signal(u32 sig)
  4008	 *	Description
  4009	 *		Send signal *sig* to the process of the current task.
  4010	 *		The signal may be delivered to any of this process's threads.
  4011	 *	Return
  4012	 *		0 on success or successfully queued.
  4013	 *
  4014	 *		**-EBUSY** if work queue under nmi is full.
  4015	 *
  4016	 *		**-EINVAL** if *sig* is invalid.
  4017	 *
  4018	 *		**-EPERM** if no permission to send the *sig*.
  4019	 *
  4020	 *		**-EAGAIN** if bpf program can try again.
  4021	 *
  4022	 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
  4023	 *	Description
  4024	 *		Try to issue a SYN cookie for the packet with corresponding
  4025	 *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
  4026	 *
  4027	 *		*iph* points to the start of the IPv4 or IPv6 header, while
  4028	 *		*iph_len* contains **sizeof**\ (**struct iphdr**) or
  4029	 *		**sizeof**\ (**struct ipv6hdr**).
  4030	 *
  4031	 *		*th* points to the start of the TCP header, while *th_len*
  4032	 *		contains the length of the TCP header with options (at least
  4033	 *		**sizeof**\ (**struct tcphdr**)).
  4034	 *	Return
  4035	 *		On success, lower 32 bits hold the generated SYN cookie in
  4036	 *		followed by 16 bits which hold the MSS value for that cookie,
  4037	 *		and the top 16 bits are unused.
  4038	 *
  4039	 *		On failure, the returned value is one of the following:
  4040	 *
  4041	 *		**-EINVAL** SYN cookie cannot be issued due to error
  4042	 *
  4043	 *		**-ENOENT** SYN cookie should not be issued (no SYN flood)
  4044	 *
  4045	 *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies
  4046	 *
  4047	 *		**-EPROTONOSUPPORT** IP packet version is not 4 or 6
  4048	 *
  4049	 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  4050	 * 	Description
  4051	 * 		Write raw *data* blob into a special BPF perf event held by
  4052	 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  4053	 * 		event must have the following attributes: **PERF_SAMPLE_RAW**
  4054	 * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  4055	 * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  4056	 *
  4057	 * 		The *flags* are used to indicate the index in *map* for which
  4058	 * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
  4059	 * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  4060	 * 		to indicate that the index of the current CPU core should be
  4061	 * 		used.
  4062	 *
  4063	 * 		The value to write, of *size*, is passed through eBPF stack and
  4064	 * 		pointed by *data*.
  4065	 *
  4066	 * 		*ctx* is a pointer to in-kernel struct sk_buff.
  4067	 *
  4068	 * 		This helper is similar to **bpf_perf_event_output**\ () but
  4069	 * 		restricted to raw_tracepoint bpf programs.
  4070	 * 	Return
  4071	 * 		0 on success, or a negative error in case of failure.
  4072	 *
  4073	 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
  4074	 * 	Description
  4075	 * 		Safely attempt to read *size* bytes from user space address
  4076	 * 		*unsafe_ptr* and store the data in *dst*.
  4077	 * 	Return
  4078	 * 		0 on success, or a negative error in case of failure.
  4079	 *
  4080	 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
  4081	 * 	Description
  4082	 * 		Safely attempt to read *size* bytes from kernel space address
  4083	 * 		*unsafe_ptr* and store the data in *dst*.
  4084	 * 	Return
  4085	 * 		0 on success, or a negative error in case of failure.
  4086	 *
  4087	 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
  4088	 * 	Description
  4089	 * 		Copy a NUL terminated string from an unsafe user address
  4090	 * 		*unsafe_ptr* to *dst*. The *size* should include the
  4091	 * 		terminating NUL byte. In case the string length is smaller than
  4092	 * 		*size*, the target is not padded with further NUL bytes. If the
  4093	 * 		string length is larger than *size*, just *size*-1 bytes are
  4094	 * 		copied and the last byte is set to NUL.
  4095	 *
  4096	 * 		On success, returns the number of bytes that were written,
  4097	 * 		including the terminal NUL. This makes this helper useful in
  4098	 * 		tracing programs for reading strings, and more importantly to
  4099	 * 		get its length at runtime. See the following snippet:
  4100	 *
  4101	 * 		::
  4102	 *
  4103	 * 			SEC("kprobe/sys_open")
  4104	 * 			void bpf_sys_open(struct pt_regs *ctx)
  4105	 * 			{
  4106	 * 			        char buf[PATHLEN]; // PATHLEN is defined to 256
  4107	 * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),
  4108	 * 				                                  ctx->di);
  4109	 *
  4110	 * 				// Consume buf, for example push it to
  4111	 * 				// userspace via bpf_perf_event_output(); we
  4112	 * 				// can use res (the string length) as event
  4113	 * 				// size, after checking its boundaries.
  4114	 * 			}
  4115	 *
  4116	 * 		In comparison, using **bpf_probe_read_user**\ () helper here
  4117	 * 		instead to read the string would require to estimate the length
  4118	 * 		at compile time, and would often result in copying more memory
  4119	 * 		than necessary.
  4120	 *
  4121	 * 		Another useful use case is when parsing individual process
  4122	 * 		arguments or individual environment variables navigating
  4123	 * 		*current*\ **->mm->arg_start** and *current*\
  4124	 * 		**->mm->env_start**: using this helper and the return value,
  4125	 * 		one can quickly iterate at the right offset of the memory area.
  4126	 * 	Return
  4127	 * 		On success, the strictly positive length of the output string,
  4128	 * 		including the trailing NUL character. On error, a negative
  4129	 * 		value.
  4130	 *
  4131	 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
  4132	 * 	Description
  4133	 * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
  4134	 * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
  4135	 * 	Return
  4136	 * 		On success, the strictly positive length of the string, including
  4137	 * 		the trailing NUL character. On error, a negative value.
  4138	 *
  4139	 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
  4140	 *	Description
  4141	 *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
  4142	 *		*rcv_nxt* is the ack_seq to be sent out.
  4143	 *	Return
  4144	 *		0 on success, or a negative error in case of failure.
  4145	 *
  4146	 * long bpf_send_signal_thread(u32 sig)
  4147	 *	Description
  4148	 *		Send signal *sig* to the thread corresponding to the current task.
  4149	 *	Return
  4150	 *		0 on success or successfully queued.
  4151	 *
  4152	 *		**-EBUSY** if work queue under nmi is full.
  4153	 *
  4154	 *		**-EINVAL** if *sig* is invalid.
  4155	 *
  4156	 *		**-EPERM** if no permission to send the *sig*.
  4157	 *
  4158	 *		**-EAGAIN** if bpf program can try again.
  4159	 *
  4160	 * u64 bpf_jiffies64(void)
  4161	 *	Description
  4162	 *		Obtain the 64bit jiffies
  4163	 *	Return
  4164	 *		The 64 bit jiffies
  4165	 *
  4166	 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
  4167	 *	Description
  4168	 *		For an eBPF program attached to a perf event, retrieve the
  4169	 *		branch records (**struct perf_branch_entry**) associated to *ctx*
  4170	 *		and store it in the buffer pointed by *buf* up to size
  4171	 *		*size* bytes.
  4172	 *	Return
  4173	 *		On success, number of bytes written to *buf*. On error, a
  4174	 *		negative value.
  4175	 *
  4176	 *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
  4177	 *		instead return the number of bytes required to store all the
  4178	 *		branch entries. If this flag is set, *buf* may be NULL.
  4179	 *
  4180	 *		**-EINVAL** if arguments invalid or **size** not a multiple
  4181	 *		of **sizeof**\ (**struct perf_branch_entry**\ ).
  4182	 *
  4183	 *		**-ENOENT** if architecture does not support branch records.
  4184	 *
  4185	 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
  4186	 *	Description
  4187	 *		Returns 0 on success, values for *pid* and *tgid* as seen from the current
  4188	 *		*namespace* will be returned in *nsdata*.
  4189	 *	Return
  4190	 *		0 on success, or one of the following in case of failure:
  4191	 *
  4192	 *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number
  4193	 *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
  4194	 *
  4195	 *		**-ENOENT** if pidns does not exists for the current task.
  4196	 *
  4197	 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  4198	 *	Description
  4199	 *		Write raw *data* blob into a special BPF perf event held by
  4200	 *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  4201	 *		event must have the following attributes: **PERF_SAMPLE_RAW**
  4202	 *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  4203	 *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  4204	 *
  4205	 *		The *flags* are used to indicate the index in *map* for which
  4206	 *		the value must be put, masked with **BPF_F_INDEX_MASK**.
  4207	 *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  4208	 *		to indicate that the index of the current CPU core should be
  4209	 *		used.
  4210	 *
  4211	 *		The value to write, of *size*, is passed through eBPF stack and
  4212	 *		pointed by *data*.
  4213	 *
  4214	 *		*ctx* is a pointer to in-kernel struct xdp_buff.
  4215	 *
  4216	 *		This helper is similar to **bpf_perf_eventoutput**\ () but
  4217	 *		restricted to raw_tracepoint bpf programs.
  4218	 *	Return
  4219	 *		0 on success, or a negative error in case of failure.
  4220	 *
  4221	 * u64 bpf_get_netns_cookie(void *ctx)
  4222	 * 	Description
  4223	 * 		Retrieve the cookie (generated by the kernel) of the network
  4224	 * 		namespace the input *ctx* is associated with. The network
  4225	 * 		namespace cookie remains stable for its lifetime and provides
  4226	 * 		a global identifier that can be assumed unique. If *ctx* is
  4227	 * 		NULL, then the helper returns the cookie for the initial
  4228	 * 		network namespace. The cookie itself is very similar to that
  4229	 * 		of **bpf_get_socket_cookie**\ () helper, but for network
  4230	 * 		namespaces instead of sockets.
  4231	 * 	Return
  4232	 * 		A 8-byte long opaque number.
  4233	 *
  4234	 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
  4235	 * 	Description
  4236	 * 		Return id of cgroup v2 that is ancestor of the cgroup associated
  4237	 * 		with the current task at the *ancestor_level*. The root cgroup
  4238	 * 		is at *ancestor_level* zero and each step down the hierarchy
  4239	 * 		increments the level. If *ancestor_level* == level of cgroup
  4240	 * 		associated with the current task, then return value will be the
  4241	 * 		same as that of **bpf_get_current_cgroup_id**\ ().
  4242	 *
  4243	 * 		The helper is useful to implement policies based on cgroups
  4244	 * 		that are upper in hierarchy than immediate cgroup associated
  4245	 * 		with the current task.
  4246	 *
  4247	 * 		The format of returned id and helper limitations are same as in
  4248	 * 		**bpf_get_current_cgroup_id**\ ().
  4249	 * 	Return
  4250	 * 		The id is returned or 0 in case the id could not be retrieved.
  4251	 *
  4252	 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
  4253	 *	Description
  4254	 *		Helper is overloaded depending on BPF program type. This
  4255	 *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and
  4256	 *		**BPF_PROG_TYPE_SCHED_ACT** programs.
  4257	 *
  4258	 *		Assign the *sk* to the *skb*. When combined with appropriate
  4259	 *		routing configuration to receive the packet towards the socket,
  4260	 *		will cause *skb* to be delivered to the specified socket.
  4261	 *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),
  4262	 *		**bpf_clone_redirect**\ () or other methods outside of BPF may
  4263	 *		interfere with successful delivery to the socket.
  4264	 *
  4265	 *		This operation is only valid from TC ingress path.
  4266	 *
  4267	 *		The *flags* argument must be zero.
  4268	 *	Return
  4269	 *		0 on success, or a negative error in case of failure:
  4270	 *
  4271	 *		**-EINVAL** if specified *flags* are not supported.
  4272	 *
  4273	 *		**-ENOENT** if the socket is unavailable for assignment.
  4274	 *
  4275	 *		**-ENETUNREACH** if the socket is unreachable (wrong netns).
  4276	 *
  4277	 *		**-EOPNOTSUPP** if the operation is not supported, for example
  4278	 *		a call from outside of TC ingress.
  4279	 *
  4280	 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
  4281	 *	Description
  4282	 *		Helper is overloaded depending on BPF program type. This
  4283	 *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
  4284	 *
  4285	 *		Select the *sk* as a result of a socket lookup.
  4286	 *
  4287	 *		For the operation to succeed passed socket must be compatible
  4288	 *		with the packet description provided by the *ctx* object.
  4289	 *
  4290	 *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
  4291	 *		be an exact match. While IP family (**AF_INET** or
  4292	 *		**AF_INET6**) must be compatible, that is IPv6 sockets
  4293	 *		that are not v6-only can be selected for IPv4 packets.
  4294	 *
  4295	 *		Only TCP listeners and UDP unconnected sockets can be
  4296	 *		selected. *sk* can also be NULL to reset any previous
  4297	 *		selection.
  4298	 *
  4299	 *		*flags* argument can combination of following values:
  4300	 *
  4301	 *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous
  4302	 *		  socket selection, potentially done by a BPF program
  4303	 *		  that ran before us.
  4304	 *
  4305	 *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
  4306	 *		  load-balancing within reuseport group for the socket
  4307	 *		  being selected.
  4308	 *
  4309	 *		On success *ctx->sk* will point to the selected socket.
  4310	 *
  4311	 *	Return
  4312	 *		0 on success, or a negative errno in case of failure.
  4313	 *
  4314	 *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is
  4315	 *		  not compatible with packet family (*ctx->family*).
  4316	 *
  4317	 *		* **-EEXIST** if socket has been already selected,
  4318	 *		  potentially by another program, and
  4319	 *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
  4320	 *
  4321	 *		* **-EINVAL** if unsupported flags were specified.
  4322	 *
  4323	 *		* **-EPROTOTYPE** if socket L4 protocol
  4324	 *		  (*sk->protocol*) doesn't match packet protocol
  4325	 *		  (*ctx->protocol*).
  4326	 *
  4327	 *		* **-ESOCKTNOSUPPORT** if socket is not in allowed
  4328	 *		  state (TCP listening or UDP unconnected).
  4329	 *
  4330	 * u64 bpf_ktime_get_boot_ns(void)
  4331	 * 	Description
  4332	 * 		Return the time elapsed since system boot, in nanoseconds.
  4333	 * 		Does include the time the system was suspended.
  4334	 * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
  4335	 * 	Return
  4336	 * 		Current *ktime*.
  4337	 *
  4338	 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
  4339	 * 	Description
  4340	 * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
  4341	 * 		out the format string.
  4342	 * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for
  4343	 * 		the format string itself. The *data* and *data_len* are format string
  4344	 * 		arguments. The *data* are a **u64** array and corresponding format string
  4345	 * 		values are stored in the array. For strings and pointers where pointees
  4346	 * 		are accessed, only the pointer values are stored in the *data* array.
  4347	 * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.
  4348	 *
  4349	 *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
  4350	 *		Reading kernel memory may fail due to either invalid address or
  4351	 *		valid address but requiring a major memory fault. If reading kernel memory
  4352	 *		fails, the string for **%s** will be an empty string, and the ip
  4353	 *		address for **%p{i,I}{4,6}** will be 0. Not returning error to
  4354	 *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.
  4355	 * 	Return
  4356	 * 		0 on success, or a negative error in case of failure:
  4357	 *
  4358	 *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again
  4359	 *		by returning 1 from bpf program.
  4360	 *
  4361	 *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
  4362	 *
  4363	 *		**-E2BIG** if *fmt* contains too many format specifiers.
  4364	 *
  4365	 *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
  4366	 *
  4367	 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
  4368	 * 	Description
  4369	 * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
  4370	 * 		The *m* represents the seq_file. The *data* and *len* represent the
  4371	 * 		data to write in bytes.
  4372	 * 	Return
  4373	 * 		0 on success, or a negative error in case of failure:
  4374	 *
  4375	 *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
  4376	 *
  4377	 * u64 bpf_sk_cgroup_id(void *sk)
  4378	 *	Description
  4379	 *		Return the cgroup v2 id of the socket *sk*.
  4380	 *
  4381	 *		*sk* must be a non-**NULL** pointer to a socket, e.g. one
  4382	 *		returned from **bpf_sk_lookup_xxx**\ (),
  4383	 *		**bpf_sk_fullsock**\ (), etc. The format of returned id is
  4384	 *		same as in **bpf_skb_cgroup_id**\ ().
  4385	 *
  4386	 *		This helper is available only if the kernel was compiled with
  4387	 *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.
  4388	 *	Return
  4389	 *		The id is returned or 0 in case the id could not be retrieved.
  4390	 *
  4391	 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
  4392	 *	Description
  4393	 *		Return id of cgroup v2 that is ancestor of cgroup associated
  4394	 *		with the *sk* at the *ancestor_level*.  The root cgroup is at
  4395	 *		*ancestor_level* zero and each step down the hierarchy
  4396	 *		increments the level. If *ancestor_level* == level of cgroup
  4397	 *		associated with *sk*, then return value will be same as that
  4398	 *		of **bpf_sk_cgroup_id**\ ().
  4399	 *
  4400	 *		The helper is useful to implement policies based on cgroups
  4401	 *		that are upper in hierarchy than immediate cgroup associated
  4402	 *		with *sk*.
  4403	 *
  4404	 *		The format of returned id and helper limitations are same as in
  4405	 *		**bpf_sk_cgroup_id**\ ().
  4406	 *	Return
  4407	 *		The id is returned or 0 in case the id could not be retrieved.
  4408	 *
  4409	 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
  4410	 * 	Description
  4411	 * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.
  4412	 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4413	 * 		of new data availability is sent.
  4414	 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4415	 * 		of new data availability is sent unconditionally.
  4416	 * 		If **0** is specified in *flags*, an adaptive notification
  4417	 * 		of new data availability is sent.
  4418	 *
  4419	 * 		An adaptive notification is a notification sent whenever the user-space
  4420	 * 		process has caught up and consumed all available payloads. In case the user-space
  4421	 * 		process is still processing a previous payload, then no notification is needed
  4422	 * 		as it will process the newly added payload automatically.
  4423	 * 	Return
  4424	 * 		0 on success, or a negative error in case of failure.
  4425	 *
  4426	 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
  4427	 * 	Description
  4428	 * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.
  4429	 * 		*flags* must be 0.
  4430	 * 	Return
  4431	 * 		Valid pointer with *size* bytes of memory available; NULL,
  4432	 * 		otherwise.
  4433	 *
  4434	 * void bpf_ringbuf_submit(void *data, u64 flags)
  4435	 * 	Description
  4436	 * 		Submit reserved ring buffer sample, pointed to by *data*.
  4437	 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4438	 * 		of new data availability is sent.
  4439	 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4440	 * 		of new data availability is sent unconditionally.
  4441	 * 		If **0** is specified in *flags*, an adaptive notification
  4442	 * 		of new data availability is sent.
  4443	 *
  4444	 * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
  4445	 * 	Return
  4446	 * 		Nothing. Always succeeds.
  4447	 *
  4448	 * void bpf_ringbuf_discard(void *data, u64 flags)
  4449	 * 	Description
  4450	 * 		Discard reserved ring buffer sample, pointed to by *data*.
  4451	 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
  4452	 * 		of new data availability is sent.
  4453	 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
  4454	 * 		of new data availability is sent unconditionally.
  4455	 * 		If **0** is specified in *flags*, an adaptive notification
  4456	 * 		of new data availability is sent.
  4457	 *
  4458	 * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
  4459	 * 	Return
  4460	 * 		Nothing. Always succeeds.
  4461	 *
  4462	 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
  4463	 *	Description
  4464	 *		Query various characteristics of provided ring buffer. What
  4465	 *		exactly is queries is determined by *flags*:
  4466	 *
  4467	 *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
  4468	 *		* **BPF_RB_RING_SIZE**: The size of ring buffer.
  4469	 *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).
  4470	 *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
  4471	 *
  4472	 *		Data returned is just a momentary snapshot of actual values
  4473	 *		and could be inaccurate, so this facility should be used to
  4474	 *		power heuristics and for reporting, not to make 100% correct
  4475	 *		calculation.
  4476	 *	Return
  4477	 *		Requested value, or 0, if *flags* are not recognized.
  4478	 *
  4479	 * long bpf_csum_level(struct sk_buff *skb, u64 level)
  4480	 * 	Description
  4481	 * 		Change the skbs checksum level by one layer up or down, or
  4482	 * 		reset it entirely to none in order to have the stack perform
  4483	 * 		checksum validation. The level is applicable to the following
  4484	 * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
  4485	 * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
  4486	 * 		through **bpf_skb_adjust_room**\ () helper with passing in
  4487	 * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call
  4488	 * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
  4489	 * 		the UDP header is removed. Similarly, an encap of the latter
  4490	 * 		into the former could be accompanied by a helper call to
  4491	 * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
  4492	 * 		skb is still intended to be processed in higher layers of the
  4493	 * 		stack instead of just egressing at tc.
  4494	 *
  4495	 * 		There are three supported level settings at this time:
  4496	 *
  4497	 * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
  4498	 * 		  with CHECKSUM_UNNECESSARY.
  4499	 * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
  4500	 * 		  with CHECKSUM_UNNECESSARY.
  4501	 * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
  4502	 * 		  sets CHECKSUM_NONE to force checksum validation by the stack.
  4503	 * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
  4504	 * 		  skb->csum_level.
  4505	 * 	Return
  4506	 * 		0 on success, or a negative error in case of failure. In the
  4507	 * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
  4508	 * 		is returned or the error code -EACCES in case the skb is not
  4509	 * 		subject to CHECKSUM_UNNECESSARY.
  4510	 *
  4511	 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
  4512	 *	Description
  4513	 *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
  4514	 *	Return
  4515	 *		*sk* if casting is valid, or **NULL** otherwise.
  4516	 *
  4517	 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
  4518	 *	Description
  4519	 *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
  4520	 *	Return
  4521	 *		*sk* if casting is valid, or **NULL** otherwise.
  4522	 *
  4523	 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
  4524	 * 	Description
  4525	 *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
  4526	 *	Return
  4527	 *		*sk* if casting is valid, or **NULL** otherwise.
  4528	 *
  4529	 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
  4530	 * 	Description
  4531	 *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
  4532	 *	Return
  4533	 *		*sk* if casting is valid, or **NULL** otherwise.
  4534	 *
  4535	 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
  4536	 * 	Description
  4537	 *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
  4538	 *	Return
  4539	 *		*sk* if casting is valid, or **NULL** otherwise.
  4540	 *
  4541	 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
  4542	 *	Description
  4543	 *		Return a user or a kernel stack in bpf program provided buffer.
  4544	 *		Note: the user stack will only be populated if the *task* is
  4545	 *		the current task; all other tasks will return -EOPNOTSUPP.
  4546	 *		To achieve this, the helper needs *task*, which is a valid
  4547	 *		pointer to **struct task_struct**. To store the stacktrace, the
  4548	 *		bpf program provides *buf* with a nonnegative *size*.
  4549	 *
  4550	 *		The last argument, *flags*, holds the number of stack frames to
  4551	 *		skip (from 0 to 255), masked with
  4552	 *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  4553	 *		the following flags:
  4554	 *
  4555	 *		**BPF_F_USER_STACK**
  4556	 *			Collect a user space stack instead of a kernel stack.
  4557	 *			The *task* must be the current task.
  4558	 *		**BPF_F_USER_BUILD_ID**
  4559	 *			Collect buildid+offset instead of ips for user stack,
  4560	 *			only valid if **BPF_F_USER_STACK** is also specified.
  4561	 *
  4562	 *		**bpf_get_task_stack**\ () can collect up to
  4563	 *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  4564	 *		to sufficient large buffer size. Note that
  4565	 *		this limit can be controlled with the **sysctl** program, and
  4566	 *		that it should be manually increased in order to profile long
  4567	 *		user stacks (such as stacks for Java programs). To do so, use:
  4568	 *
  4569	 *		::
  4570	 *
  4571	 *			# sysctl kernel.perf_event_max_stack=<new value>
  4572	 *	Return
  4573	 * 		The non-negative copied *buf* length equal to or less than
  4574	 * 		*size* on success, or a negative error in case of failure.
  4575	 *
  4576	 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
  4577	 *	Description
  4578	 *		Load header option.  Support reading a particular TCP header
  4579	 *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
  4580	 *
  4581	 *		If *flags* is 0, it will search the option from the
  4582	 *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
  4583	 *		has details on what skb_data contains under different
  4584	 *		*skops*\ **->op**.
  4585	 *
  4586	 *		The first byte of the *searchby_res* specifies the
  4587	 *		kind that it wants to search.
  4588	 *
  4589	 *		If the searching kind is an experimental kind
  4590	 *		(i.e. 253 or 254 according to RFC6994).  It also
  4591	 *		needs to specify the "magic" which is either
  4592	 *		2 bytes or 4 bytes.  It then also needs to
  4593	 *		specify the size of the magic by using
  4594	 *		the 2nd byte which is "kind-length" of a TCP
  4595	 *		header option and the "kind-length" also
  4596	 *		includes the first 2 bytes "kind" and "kind-length"
  4597	 *		itself as a normal TCP header option also does.
  4598	 *
  4599	 *		For example, to search experimental kind 254 with
  4600	 *		2 byte magic 0xeB9F, the searchby_res should be
  4601	 *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
  4602	 *
  4603	 *		To search for the standard window scale option (3),
  4604	 *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].
  4605	 *		Note, kind-length must be 0 for regular option.
  4606	 *
  4607	 *		Searching for No-Op (0) and End-of-Option-List (1) are
  4608	 *		not supported.
  4609	 *
  4610	 *		*len* must be at least 2 bytes which is the minimal size
  4611	 *		of a header option.
  4612	 *
  4613	 *		Supported flags:
  4614	 *
  4615	 *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
  4616	 *		  saved_syn packet or the just-received syn packet.
  4617	 *
  4618	 *	Return
  4619	 *		> 0 when found, the header option is copied to *searchby_res*.
  4620	 *		The return value is the total length copied. On failure, a
  4621	 *		negative error code is returned:
  4622	 *
  4623	 *		**-EINVAL** if a parameter is invalid.
  4624	 *
  4625	 *		**-ENOMSG** if the option is not found.
  4626	 *
  4627	 *		**-ENOENT** if no syn packet is available when
  4628	 *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.
  4629	 *
  4630	 *		**-ENOSPC** if there is not enough space.  Only *len* number of
  4631	 *		bytes are copied.
  4632	 *
  4633	 *		**-EFAULT** on failure to parse the header options in the
  4634	 *		packet.
  4635	 *
  4636	 *		**-EPERM** if the helper cannot be used under the current
  4637	 *		*skops*\ **->op**.
  4638	 *
  4639	 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
  4640	 *	Description
  4641	 *		Store header option.  The data will be copied
  4642	 *		from buffer *from* with length *len* to the TCP header.
  4643	 *
  4644	 *		The buffer *from* should have the whole option that
  4645	 *		includes the kind, kind-length, and the actual
  4646	 *		option data.  The *len* must be at least kind-length
  4647	 *		long.  The kind-length does not have to be 4 byte
  4648	 *		aligned.  The kernel will take care of the padding
  4649	 *		and setting the 4 bytes aligned value to th->doff.
  4650	 *
  4651	 *		This helper will check for duplicated option
  4652	 *		by searching the same option in the outgoing skb.
  4653	 *
  4654	 *		This helper can only be called during
  4655	 *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
  4656	 *
  4657	 *	Return
  4658	 *		0 on success, or negative error in case of failure:
  4659	 *
  4660	 *		**-EINVAL** If param is invalid.
  4661	 *
  4662	 *		**-ENOSPC** if there is not enough space in the header.
  4663	 *		Nothing has been written
  4664	 *
  4665	 *		**-EEXIST** if the option already exists.
  4666	 *
  4667	 *		**-EFAULT** on failure to parse the existing header options.
  4668	 *
  4669	 *		**-EPERM** if the helper cannot be used under the current
  4670	 *		*skops*\ **->op**.
  4671	 *
  4672	 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
  4673	 *	Description
  4674	 *		Reserve *len* bytes for the bpf header option.  The
  4675	 *		space will be used by **bpf_store_hdr_opt**\ () later in
  4676	 *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
  4677	 *
  4678	 *		If **bpf_reserve_hdr_opt**\ () is called multiple times,
  4679	 *		the total number of bytes will be reserved.
  4680	 *
  4681	 *		This helper can only be called during
  4682	 *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
  4683	 *
  4684	 *	Return
  4685	 *		0 on success, or negative error in case of failure:
  4686	 *
  4687	 *		**-EINVAL** if a parameter is invalid.
  4688	 *
  4689	 *		**-ENOSPC** if there is not enough space in the header.
  4690	 *
  4691	 *		**-EPERM** if the helper cannot be used under the current
  4692	 *		*skops*\ **->op**.
  4693	 *
  4694	 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
  4695	 *	Description
  4696	 *		Get a bpf_local_storage from an *inode*.
  4697	 *
  4698	 *		Logically, it could be thought of as getting the value from
  4699	 *		a *map* with *inode* as the **key**.  From this
  4700	 *		perspective,  the usage is not much different from
  4701	 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
  4702	 *		helper enforces the key must be an inode and the map must also
  4703	 *		be a **BPF_MAP_TYPE_INODE_STORAGE**.
  4704	 *
  4705	 *		Underneath, the value is stored locally at *inode* instead of
  4706	 *		the *map*.  The *map* is used as the bpf-local-storage
  4707	 *		"type". The bpf-local-storage "type" (i.e. the *map*) is
  4708	 *		searched against all bpf_local_storage residing at *inode*.
  4709	 *
  4710	 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  4711	 *		used such that a new bpf_local_storage will be
  4712	 *		created if one does not exist.  *value* can be used
  4713	 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  4714	 *		the initial value of a bpf_local_storage.  If *value* is
  4715	 *		**NULL**, the new bpf_local_storage will be zero initialized.
  4716	 *	Return
  4717	 *		A bpf_local_storage pointer is returned on success.
  4718	 *
  4719	 *		**NULL** if not found or there was an error in adding
  4720	 *		a new bpf_local_storage.
  4721	 *
  4722	 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
  4723	 *	Description
  4724	 *		Delete a bpf_local_storage from an *inode*.
  4725	 *	Return
  4726	 *		0 on success.
  4727	 *
  4728	 *		**-ENOENT** if the bpf_local_storage cannot be found.
  4729	 *
  4730	 * long bpf_d_path(struct path *path, char *buf, u32 sz)
  4731	 *	Description
  4732	 *		Return full path for given **struct path** object, which
  4733	 *		needs to be the kernel BTF *path* object. The path is
  4734	 *		returned in the provided buffer *buf* of size *sz* and
  4735	 *		is zero terminated.
  4736	 *
  4737	 *	Return
  4738	 *		On success, the strictly positive length of the string,
  4739	 *		including the trailing NUL character. On error, a negative
  4740	 *		value.
  4741	 *
  4742	 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
  4743	 * 	Description
  4744	 * 		Read *size* bytes from user space address *user_ptr* and store
  4745	 * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().
  4746	 * 	Return
  4747	 * 		0 on success, or a negative error in case of failure.
  4748	 *
  4749	 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
  4750	 *	Description
  4751	 *		Use BTF to store a string representation of *ptr*->ptr in *str*,
  4752	 *		using *ptr*->type_id.  This value should specify the type
  4753	 *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
  4754	 *		can be used to look up vmlinux BTF type ids. Traversing the
  4755	 *		data structure using BTF, the type information and values are
  4756	 *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
  4757	 *		the pointer data is carried out to avoid kernel crashes during
  4758	 *		operation.  Smaller types can use string space on the stack;
  4759	 *		larger programs can use map data to store the string
  4760	 *		representation.
  4761	 *
  4762	 *		The string can be subsequently shared with userspace via
  4763	 *		bpf_perf_event_output() or ring buffer interfaces.
  4764	 *		bpf_trace_printk() is to be avoided as it places too small
  4765	 *		a limit on string size to be useful.
  4766	 *
  4767	 *		*flags* is a combination of
  4768	 *
  4769	 *		**BTF_F_COMPACT**
  4770	 *			no formatting around type information
  4771	 *		**BTF_F_NONAME**
  4772	 *			no struct/union member names/types
  4773	 *		**BTF_F_PTR_RAW**
  4774	 *			show raw (unobfuscated) pointer values;
  4775	 *			equivalent to printk specifier %px.
  4776	 *		**BTF_F_ZERO**
  4777	 *			show zero-valued struct/union members; they
  4778	 *			are not displayed by default
  4779	 *
  4780	 *	Return
  4781	 *		The number of bytes that were written (or would have been
  4782	 *		written if output had to be truncated due to string size),
  4783	 *		or a negative error in cases of failure.
  4784	 *
  4785	 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
  4786	 *	Description
  4787	 *		Use BTF to write to seq_write a string representation of
  4788	 *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
  4789	 *		*flags* are identical to those used for bpf_snprintf_btf.
  4790	 *	Return
  4791	 *		0 on success or a negative error in case of failure.
  4792	 *
  4793	 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
  4794	 * 	Description
  4795	 * 		See **bpf_get_cgroup_classid**\ () for the main description.
  4796	 * 		This helper differs from **bpf_get_cgroup_classid**\ () in that
  4797	 * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s
  4798	 * 		associated socket instead of the current process.
  4799	 * 	Return
  4800	 * 		The id is returned or 0 in case the id could not be retrieved.
  4801	 *
  4802	 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
  4803	 * 	Description
  4804	 * 		Redirect the packet to another net device of index *ifindex*
  4805	 * 		and fill in L2 addresses from neighboring subsystem. This helper
  4806	 * 		is somewhat similar to **bpf_redirect**\ (), except that it
  4807	 * 		populates L2 addresses as well, meaning, internally, the helper
  4808	 * 		relies on the neighbor lookup for the L2 address of the nexthop.
  4809	 *
  4810	 * 		The helper will perform a FIB lookup based on the skb's
  4811	 * 		networking header to get the address of the next hop, unless
  4812	 * 		this is supplied by the caller in the *params* argument. The
  4813	 * 		*plen* argument indicates the len of *params* and should be set
  4814	 * 		to 0 if *params* is NULL.
  4815	 *
  4816	 * 		The *flags* argument is reserved and must be 0. The helper is
  4817	 * 		currently only supported for tc BPF program types, and enabled
  4818	 * 		for IPv4 and IPv6 protocols.
  4819	 * 	Return
  4820	 * 		The helper returns **TC_ACT_REDIRECT** on success or
  4821	 * 		**TC_ACT_SHOT** on error.
  4822	 *
  4823	 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
  4824	 *     Description
  4825	 *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
  4826	 *             pointer to the percpu kernel variable on *cpu*. A ksym is an
  4827	 *             extern variable decorated with '__ksym'. For ksym, there is a
  4828	 *             global var (either static or global) defined of the same name
  4829	 *             in the kernel. The ksym is percpu if the global var is percpu.
  4830	 *             The returned pointer points to the global percpu var on *cpu*.
  4831	 *
  4832	 *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
  4833	 *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
  4834	 *             happens if *cpu* is larger than nr_cpu_ids. The caller of
  4835	 *             bpf_per_cpu_ptr() must check the returned value.
  4836	 *     Return
  4837	 *             A pointer pointing to the kernel percpu variable on *cpu*, or
  4838	 *             NULL, if *cpu* is invalid.
  4839	 *
  4840	 * void *bpf_this_cpu_ptr(const void *percpu_ptr)
  4841	 *	Description
  4842	 *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a
  4843	 *		pointer to the percpu kernel variable on this cpu. See the
  4844	 *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().
  4845	 *
  4846	 *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
  4847	 *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
  4848	 *		never return NULL.
  4849	 *	Return
  4850	 *		A pointer pointing to the kernel percpu variable on this cpu.
  4851	 *
  4852	 * long bpf_redirect_peer(u32 ifindex, u64 flags)
  4853	 * 	Description
  4854	 * 		Redirect the packet to another net device of index *ifindex*.
  4855	 * 		This helper is somewhat similar to **bpf_redirect**\ (), except
  4856	 * 		that the redirection happens to the *ifindex*' peer device and
  4857	 * 		the netns switch takes place from ingress to ingress without
  4858	 * 		going through the CPU's backlog queue.
  4859	 *
  4860	 * 		The *flags* argument is reserved and must be 0. The helper is
  4861	 * 		currently only supported for tc BPF program types at the ingress
  4862	 * 		hook and for veth device types. The peer device must reside in a
  4863	 * 		different network namespace.
  4864	 * 	Return
  4865	 * 		The helper returns **TC_ACT_REDIRECT** on success or
  4866	 * 		**TC_ACT_SHOT** on error.
  4867	 *
  4868	 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
  4869	 *	Description
  4870	 *		Get a bpf_local_storage from the *task*.
  4871	 *
  4872	 *		Logically, it could be thought of as getting the value from
  4873	 *		a *map* with *task* as the **key**.  From this
  4874	 *		perspective,  the usage is not much different from
  4875	 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
  4876	 *		helper enforces the key must be a task_struct and the map must also
  4877	 *		be a **BPF_MAP_TYPE_TASK_STORAGE**.
  4878	 *
  4879	 *		Underneath, the value is stored locally at *task* instead of
  4880	 *		the *map*.  The *map* is used as the bpf-local-storage
  4881	 *		"type". The bpf-local-storage "type" (i.e. the *map*) is
  4882	 *		searched against all bpf_local_storage residing at *task*.
  4883	 *
  4884	 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  4885	 *		used such that a new bpf_local_storage will be
  4886	 *		created if one does not exist.  *value* can be used
  4887	 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  4888	 *		the initial value of a bpf_local_storage.  If *value* is
  4889	 *		**NULL**, the new bpf_local_storage will be zero initialized.
  4890	 *	Return
  4891	 *		A bpf_local_storage pointer is returned on success.
  4892	 *
  4893	 *		**NULL** if not found or there was an error in adding
  4894	 *		a new bpf_local_storage.
  4895	 *
  4896	 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
  4897	 *	Description
  4898	 *		Delete a bpf_local_storage from a *task*.
  4899	 *	Return
  4900	 *		0 on success.
  4901	 *
  4902	 *		**-ENOENT** if the bpf_local_storage cannot be found.
  4903	 *
  4904	 * struct task_struct *bpf_get_current_task_btf(void)
  4905	 *	Description
  4906	 *		Return a BTF pointer to the "current" task.
  4907	 *		This pointer can also be used in helpers that accept an
  4908	 *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.
  4909	 *	Return
  4910	 *		Pointer to the current task.
  4911	 *
  4912	 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
  4913	 *	Description
  4914	 *		Set or clear certain options on *bprm*:
  4915	 *
  4916	 *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
  4917	 *		which sets the **AT_SECURE** auxv for glibc. The bit
  4918	 *		is cleared if the flag is not specified.
  4919	 *	Return
  4920	 *		**-EINVAL** if invalid *flags* are passed, zero otherwise.
  4921	 *
  4922	 * u64 bpf_ktime_get_coarse_ns(void)
  4923	 * 	Description
  4924	 * 		Return a coarse-grained version of the time elapsed since
  4925	 * 		system boot, in nanoseconds. Does not include time the system
  4926	 * 		was suspended.
  4927	 *
  4928	 * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
  4929	 * 	Return
  4930	 * 		Current *ktime*.
  4931	 *
  4932	 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
  4933	 *	Description
  4934	 *		Returns the stored IMA hash of the *inode* (if it's available).
  4935	 *		If the hash is larger than *size*, then only *size*
  4936	 *		bytes will be copied to *dst*
  4937	 *	Return
  4938	 *		The **hash_algo** is returned on success,
  4939	 *		**-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
  4940	 *		invalid arguments are passed.
  4941	 *
  4942	 * struct socket *bpf_sock_from_file(struct file *file)
  4943	 *	Description
  4944	 *		If the given file represents a socket, returns the associated
  4945	 *		socket.
  4946	 *	Return
  4947	 *		A pointer to a struct socket on success or NULL if the file is
  4948	 *		not a socket.
  4949	 *
  4950	 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
  4951	 *	Description
  4952	 *		Check packet size against exceeding MTU of net device (based
  4953	 *		on *ifindex*).  This helper will likely be used in combination
  4954	 *		with helpers that adjust/change the packet size.
  4955	 *
  4956	 *		The argument *len_diff* can be used for querying with a planned
  4957	 *		size change. This allows to check MTU prior to changing packet
  4958	 *		ctx. Providing a *len_diff* adjustment that is larger than the
  4959	 *		actual packet size (resulting in negative packet size) will in
  4960	 *		principle not exceed the MTU, which is why it is not considered
  4961	 *		a failure.  Other BPF helpers are needed for performing the
  4962	 *		planned size change; therefore the responsibility for catching
  4963	 *		a negative packet size belongs in those helpers.
  4964	 *
  4965	 *		Specifying *ifindex* zero means the MTU check is performed
  4966	 *		against the current net device.  This is practical if this isn't
  4967	 *		used prior to redirect.
  4968	 *
  4969	 *		On input *mtu_len* must be a valid pointer, else verifier will
  4970	 *		reject BPF program.  If the value *mtu_len* is initialized to
  4971	 *		zero then the ctx packet size is use.  When value *mtu_len* is
  4972	 *		provided as input this specify the L3 length that the MTU check
  4973	 *		is done against. Remember XDP and TC length operate at L2, but
  4974	 *		this value is L3 as this correlate to MTU and IP-header tot_len
  4975	 *		values which are L3 (similar behavior as bpf_fib_lookup).
  4976	 *
  4977	 *		The Linux kernel route table can configure MTUs on a more
  4978	 *		specific per route level, which is not provided by this helper.
  4979	 *		For route level MTU checks use the **bpf_fib_lookup**\ ()
  4980	 *		helper.
  4981	 *
  4982	 *		*ctx* is either **struct xdp_md** for XDP programs or
  4983	 *		**struct sk_buff** for tc cls_act programs.
  4984	 *
  4985	 *		The *flags* argument can be a combination of one or more of the
  4986	 *		following values:
  4987	 *
  4988	 *		**BPF_MTU_CHK_SEGS**
  4989	 *			This flag will only works for *ctx* **struct sk_buff**.
  4990	 *			If packet context contains extra packet segment buffers
  4991	 *			(often knows as GSO skb), then MTU check is harder to
  4992	 *			check at this point, because in transmit path it is
  4993	 *			possible for the skb packet to get re-segmented
  4994	 *			(depending on net device features).  This could still be
  4995	 *			a MTU violation, so this flag enables performing MTU
  4996	 *			check against segments, with a different violation
  4997	 *			return code to tell it apart. Check cannot use len_diff.
  4998	 *
  4999	 *		On return *mtu_len* pointer contains the MTU value of the net
  5000	 *		device.  Remember the net device configured MTU is the L3 size,
  5001	 *		which is returned here and XDP and TC length operate at L2.
  5002	 *		Helper take this into account for you, but remember when using
  5003	 *		MTU value in your BPF-code.
  5004	 *
  5005	 *	Return
  5006	 *		* 0 on success, and populate MTU value in *mtu_len* pointer.
  5007	 *
  5008	 *		* < 0 if any input argument is invalid (*mtu_len* not updated)
  5009	 *
  5010	 *		MTU violations return positive values, but also populate MTU
  5011	 *		value in *mtu_len* pointer, as this can be needed for
  5012	 *		implementing PMTU handing:
  5013	 *
  5014	 *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**
  5015	 *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**
  5016	 *
  5017	 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
  5018	 *	Description
  5019	 *		For each element in **map**, call **callback_fn** function with
  5020	 *		**map**, **callback_ctx** and other map-specific parameters.
  5021	 *		The **callback_fn** should be a static function and
  5022	 *		the **callback_ctx** should be a pointer to the stack.
  5023	 *		The **flags** is used to control certain aspects of the helper.
  5024	 *		Currently, the **flags** must be 0.
  5025	 *
  5026	 *		The following are a list of supported map types and their
  5027	 *		respective expected callback signatures:
  5028	 *
  5029	 *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
  5030	 *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
  5031	 *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
  5032	 *
  5033	 *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
  5034	 *
  5035	 *		For per_cpu maps, the map_value is the value on the cpu where the
  5036	 *		bpf_prog is running.
  5037	 *
  5038	 *		If **callback_fn** return 0, the helper will continue to the next
  5039	 *		element. If return value is 1, the helper will skip the rest of
  5040	 *		elements and return. Other return values are not used now.
  5041	 *
  5042	 *	Return
  5043	 *		The number of traversed map elements for success, **-EINVAL** for
  5044	 *		invalid **flags**.
  5045	 *
  5046	 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
  5047	 *	Description
  5048	 *		Outputs a string into the **str** buffer of size **str_size**
  5049	 *		based on a format string stored in a read-only map pointed by
  5050	 *		**fmt**.
  5051	 *
  5052	 *		Each format specifier in **fmt** corresponds to one u64 element
  5053	 *		in the **data** array. For strings and pointers where pointees
  5054	 *		are accessed, only the pointer values are stored in the *data*
  5055	 *		array. The *data_len* is the size of *data* in bytes - must be
  5056	 *		a multiple of 8.
  5057	 *
  5058	 *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel
  5059	 *		memory. Reading kernel memory may fail due to either invalid
  5060	 *		address or valid address but requiring a major memory fault. If
  5061	 *		reading kernel memory fails, the string for **%s** will be an
  5062	 *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
  5063	 *		Not returning error to bpf program is consistent with what
  5064	 *		**bpf_trace_printk**\ () does for now.
  5065	 *
  5066	 *	Return
  5067	 *		The strictly positive length of the formatted string, including
  5068	 *		the trailing zero character. If the return value is greater than
  5069	 *		**str_size**, **str** contains a truncated string, guaranteed to
  5070	 *		be zero-terminated except when **str_size** is 0.
  5071	 *
  5072	 *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.
  5073	 *
  5074	 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
  5075	 * 	Description
  5076	 * 		Execute bpf syscall with given arguments.
  5077	 * 	Return
  5078	 * 		A syscall result.
  5079	 *
  5080	 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
  5081	 * 	Description
  5082	 * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
  5083	 * 	Return
  5084	 * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.
  5085	 *
  5086	 * long bpf_sys_close(u32 fd)
  5087	 * 	Description
  5088	 * 		Execute close syscall for given FD.
  5089	 * 	Return
  5090	 * 		A syscall result.
  5091	 *
  5092	 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
  5093	 *	Description
  5094	 *		Initialize the timer.
  5095	 *		First 4 bits of *flags* specify clockid.
  5096	 *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
  5097	 *		All other bits of *flags* are reserved.
  5098	 *		The verifier will reject the program if *timer* is not from
  5099	 *		the same *map*.
  5100	 *	Return
  5101	 *		0 on success.
  5102	 *		**-EBUSY** if *timer* is already initialized.
  5103	 *		**-EINVAL** if invalid *flags* are passed.
  5104	 *		**-EPERM** if *timer* is in a map that doesn't have any user references.
  5105	 *		The user space should either hold a file descriptor to a map with timers
  5106	 *		or pin such map in bpffs. When map is unpinned or file descriptor is
  5107	 *		closed all timers in the map will be cancelled and freed.
  5108	 *
  5109	 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
  5110	 *	Description
  5111	 *		Configure the timer to call *callback_fn* static function.
  5112	 *	Return
  5113	 *		0 on success.
  5114	 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
  5115	 *		**-EPERM** if *timer* is in a map that doesn't have any user references.
  5116	 *		The user space should either hold a file descriptor to a map with timers
  5117	 *		or pin such map in bpffs. When map is unpinned or file descriptor is
  5118	 *		closed all timers in the map will be cancelled and freed.
  5119	 *
  5120	 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
  5121	 *	Description
  5122	 *		Set timer expiration N nanoseconds from the current time. The
  5123	 *		configured callback will be invoked in soft irq context on some cpu
  5124	 *		and will not repeat unless another bpf_timer_start() is made.
  5125	 *		In such case the next invocation can migrate to a different cpu.
  5126	 *		Since struct bpf_timer is a field inside map element the map
  5127	 *		owns the timer. The bpf_timer_set_callback() will increment refcnt
  5128	 *		of BPF program to make sure that callback_fn code stays valid.
  5129	 *		When user space reference to a map reaches zero all timers
  5130	 *		in a map are cancelled and corresponding program's refcnts are
  5131	 *		decremented. This is done to make sure that Ctrl-C of a user
  5132	 *		process doesn't leave any timers running. If map is pinned in
  5133	 *		bpffs the callback_fn can re-arm itself indefinitely.
  5134	 *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands
  5135	 *		cancel and free the timer in the given map element.
  5136	 *		The map can contain timers that invoke callback_fn-s from different
  5137	 *		programs. The same callback_fn can serve different timers from
  5138	 *		different maps if key/value layout matches across maps.
  5139	 *		Every bpf_timer_set_callback() can have different callback_fn.
  5140	 *
  5141	 *		*flags* can be one of:
  5142	 *
  5143	 *		**BPF_F_TIMER_ABS**
  5144	 *			Start the timer in absolute expire value instead of the
  5145	 *			default relative one.
  5146	 *		**BPF_F_TIMER_CPU_PIN**
  5147	 *			Timer will be pinned to the CPU of the caller.
  5148	 *
  5149	 *	Return
  5150	 *		0 on success.
  5151	 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
  5152	 *		or invalid *flags* are passed.
  5153	 *
  5154	 * long bpf_timer_cancel(struct bpf_timer *timer)
  5155	 *	Description
  5156	 *		Cancel the timer and wait for callback_fn to finish if it was running.
  5157	 *	Return
  5158	 *		0 if the timer was not active.
  5159	 *		1 if the timer was active.
  5160	 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
  5161	 *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
  5162	 *		own timer which would have led to a deadlock otherwise.
  5163	 *
  5164	 * u64 bpf_get_func_ip(void *ctx)
  5165	 * 	Description
  5166	 * 		Get address of the traced function (for tracing and kprobe programs).
  5167	 *
  5168	 * 		When called for kprobe program attached as uprobe it returns
  5169	 * 		probe address for both entry and return uprobe.
  5170	 *
  5171	 * 	Return
  5172	 * 		Address of the traced function for kprobe.
  5173	 * 		0 for kprobes placed within the function (not at the entry).
  5174	 * 		Address of the probe for uprobe and return uprobe.
  5175	 *
  5176	 * u64 bpf_get_attach_cookie(void *ctx)
  5177	 * 	Description
  5178	 * 		Get bpf_cookie value provided (optionally) during the program
  5179	 * 		attachment. It might be different for each individual
  5180	 * 		attachment, even if BPF program itself is the same.
  5181	 * 		Expects BPF program context *ctx* as a first argument.
  5182	 *
  5183	 * 		Supported for the following program types:
  5184	 *			- kprobe/uprobe;
  5185	 *			- tracepoint;
  5186	 *			- perf_event.
  5187	 * 	Return
  5188	 *		Value specified by user at BPF link creation/attachment time
  5189	 *		or 0, if it was not specified.
  5190	 *
  5191	 * long bpf_task_pt_regs(struct task_struct *task)
  5192	 *	Description
  5193	 *		Get the struct pt_regs associated with **task**.
  5194	 *	Return
  5195	 *		A pointer to struct pt_regs.
  5196	 *
  5197	 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
  5198	 *	Description
  5199	 *		Get branch trace from hardware engines like Intel LBR. The
  5200	 *		hardware engine is stopped shortly after the helper is
  5201	 *		called. Therefore, the user need to filter branch entries
  5202	 *		based on the actual use case. To capture branch trace
  5203	 *		before the trigger point of the BPF program, the helper
  5204	 *		should be called at the beginning of the BPF program.
  5205	 *
  5206	 *		The data is stored as struct perf_branch_entry into output
  5207	 *		buffer *entries*. *size* is the size of *entries* in bytes.
  5208	 *		*flags* is reserved for now and must be zero.
  5209	 *
  5210	 *	Return
  5211	 *		On success, number of bytes written to *buf*. On error, a
  5212	 *		negative value.
  5213	 *
  5214	 *		**-EINVAL** if *flags* is not zero.
  5215	 *
  5216	 *		**-ENOENT** if architecture does not support branch records.
  5217	 *
  5218	 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
  5219	 *	Description
  5220	 *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
  5221	 *		to format and can handle more format args as a result.
  5222	 *
  5223	 *		Arguments are to be used as in **bpf_seq_printf**\ () helper.
  5224	 *	Return
  5225	 *		The number of bytes written to the buffer, or a negative error
  5226	 *		in case of failure.
  5227	 *
  5228	 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
  5229	 * 	Description
  5230	 *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
  5231	 *	Return
  5232	 *		*sk* if casting is valid, or **NULL** otherwise.
  5233	 *
  5234	 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
  5235	 *	Description
  5236	 *		Get the address of a kernel symbol, returned in *res*. *res* is
  5237	 *		set to 0 if the symbol is not found.
  5238	 *	Return
  5239	 *		On success, zero. On error, a negative value.
  5240	 *
  5241	 *		**-EINVAL** if *flags* is not zero.
  5242	 *
  5243	 *		**-EINVAL** if string *name* is not the same size as *name_sz*.
  5244	 *
  5245	 *		**-ENOENT** if symbol is not found.
  5246	 *
  5247	 *		**-EPERM** if caller does not have permission to obtain kernel address.
  5248	 *
  5249	 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
  5250	 *	Description
  5251	 *		Find vma of *task* that contains *addr*, call *callback_fn*
  5252	 *		function with *task*, *vma*, and *callback_ctx*.
  5253	 *		The *callback_fn* should be a static function and
  5254	 *		the *callback_ctx* should be a pointer to the stack.
  5255	 *		The *flags* is used to control certain aspects of the helper.
  5256	 *		Currently, the *flags* must be 0.
  5257	 *
  5258	 *		The expected callback signature is
  5259	 *
  5260	 *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
  5261	 *
  5262	 *	Return
  5263	 *		0 on success.
  5264	 *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
  5265	 *		**-EBUSY** if failed to try lock mmap_lock.
  5266	 *		**-EINVAL** for invalid **flags**.
  5267	 *
  5268	 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
  5269	 *	Description
  5270	 *		For **nr_loops**, call **callback_fn** function
  5271	 *		with **callback_ctx** as the context parameter.
  5272	 *		The **callback_fn** should be a static function and
  5273	 *		the **callback_ctx** should be a pointer to the stack.
  5274	 *		The **flags** is used to control certain aspects of the helper.
  5275	 *		Currently, the **flags** must be 0. Currently, nr_loops is
  5276	 *		limited to 1 << 23 (~8 million) loops.
  5277	 *
  5278	 *		long (\*callback_fn)(u32 index, void \*ctx);
  5279	 *
  5280	 *		where **index** is the current index in the loop. The index
  5281	 *		is zero-indexed.
  5282	 *
  5283	 *		If **callback_fn** returns 0, the helper will continue to the next
  5284	 *		loop. If return value is 1, the helper will skip the rest of
  5285	 *		the loops and return. Other return values are not used now,
  5286	 *		and will be rejected by the verifier.
  5287	 *
  5288	 *	Return
  5289	 *		The number of loops performed, **-EINVAL** for invalid **flags**,
  5290	 *		**-E2BIG** if **nr_loops** exceeds the maximum number of loops.
  5291	 *
  5292	 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
  5293	 *	Description
  5294	 *		Do strncmp() between **s1** and **s2**. **s1** doesn't need
  5295	 *		to be null-terminated and **s1_sz** is the maximum storage
  5296	 *		size of **s1**. **s2** must be a read-only string.
  5297	 *	Return
  5298	 *		An integer less than, equal to, or greater than zero
  5299	 *		if the first **s1_sz** bytes of **s1** is found to be
  5300	 *		less than, to match, or be greater than **s2**.
  5301	 *
  5302	 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
  5303	 *	Description
  5304	 *		Get **n**-th argument register (zero based) of the traced function (for tracing programs)
  5305	 *		returned in **value**.
  5306	 *
  5307	 *	Return
  5308	 *		0 on success.
  5309	 *		**-EINVAL** if n >= argument register count of traced function.
  5310	 *
  5311	 * long bpf_get_func_ret(void *ctx, u64 *value)
  5312	 *	Description
  5313	 *		Get return value of the traced function (for tracing programs)
  5314	 *		in **value**.
  5315	 *
  5316	 *	Return
  5317	 *		0 on success.
  5318	 *		**-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
  5319	 *
  5320	 * long bpf_get_func_arg_cnt(void *ctx)
  5321	 *	Description
  5322	 *		Get number of registers of the traced function (for tracing programs) where
  5323	 *		function arguments are stored in these registers.
  5324	 *
  5325	 *	Return
  5326	 *		The number of argument registers of the traced function.
  5327	 *
  5328	 * int bpf_get_retval(void)
  5329	 *	Description
  5330	 *		Get the BPF program's return value that will be returned to the upper layers.
  5331	 *
  5332	 *		This helper is currently supported by cgroup programs and only by the hooks
  5333	 *		where BPF program's return value is returned to the userspace via errno.
  5334	 *	Return
  5335	 *		The BPF program's return value.
  5336	 *
  5337	 * int bpf_set_retval(int retval)
  5338	 *	Description
  5339	 *		Set the BPF program's return value that will be returned to the upper layers.
  5340	 *
  5341	 *		This helper is currently supported by cgroup programs and only by the hooks
  5342	 *		where BPF program's return value is returned to the userspace via errno.
  5343	 *
  5344	 *		Note that there is the following corner case where the program exports an error
  5345	 *		via bpf_set_retval but signals success via 'return 1':
  5346	 *
  5347	 *			bpf_set_retval(-EPERM);
  5348	 *			return 1;
  5349	 *
  5350	 *		In this case, the BPF program's return value will use helper's -EPERM. This
  5351	 *		still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
  5352	 *
  5353	 *	Return
  5354	 *		0 on success, or a negative error in case of failure.
  5355	 *
  5356	 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
  5357	 *	Description
  5358	 *		Get the total size of a given xdp buff (linear and paged area)
  5359	 *	Return
  5360	 *		The total size of a given xdp buffer.
  5361	 *
  5362	 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
  5363	 *	Description
  5364	 *		This helper is provided as an easy way to load data from a
  5365	 *		xdp buffer. It can be used to load *len* bytes from *offset* from
  5366	 *		the frame associated to *xdp_md*, into the buffer pointed by
  5367	 *		*buf*.
  5368	 *	Return
  5369	 *		0 on success, or a negative error in case of failure.
  5370	 *
  5371	 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
  5372	 *	Description
  5373	 *		Store *len* bytes from buffer *buf* into the frame
  5374	 *		associated to *xdp_md*, at *offset*.
  5375	 *	Return
  5376	 *		0 on success, or a negative error in case of failure.
  5377	 *
  5378	 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
  5379	 *	Description
  5380	 *		Read *size* bytes from user space address *user_ptr* in *tsk*'s
  5381	 *		address space, and stores the data in *dst*. *flags* is not
  5382	 *		used yet and is provided for future extensibility. This helper
  5383	 *		can only be used by sleepable programs.
  5384	 *	Return
  5385	 *		0 on success, or a negative error in case of failure. On error
  5386	 *		*dst* buffer is zeroed out.
  5387	 *
  5388	 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
  5389	 *	Description
  5390	 *		Change the __sk_buff->tstamp_type to *tstamp_type*
  5391	 *		and set *tstamp* to the __sk_buff->tstamp together.
  5392	 *
  5393	 *		If there is no need to change the __sk_buff->tstamp_type,
  5394	 *		the tstamp value can be directly written to __sk_buff->tstamp
  5395	 *		instead.
  5396	 *
  5397	 *		BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
  5398	 *		will be kept during bpf_redirect_*().  A non zero
  5399	 *		*tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
  5400	 *		*tstamp_type*.
  5401	 *
  5402	 *		A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
  5403	 *		with a zero *tstamp*.
  5404	 *
  5405	 *		Only IPv4 and IPv6 skb->protocol are supported.
  5406	 *
  5407	 *		This function is most useful when it needs to set a
  5408	 *		mono delivery time to __sk_buff->tstamp and then
  5409	 *		bpf_redirect_*() to the egress of an iface.  For example,
  5410	 *		changing the (rcv) timestamp in __sk_buff->tstamp at
  5411	 *		ingress to a mono delivery time and then bpf_redirect_*()
  5412	 *		to sch_fq@phy-dev.
  5413	 *	Return
  5414	 *		0 on success.
  5415	 *		**-EINVAL** for invalid input
  5416	 *		**-EOPNOTSUPP** for unsupported protocol
  5417	 *
  5418	 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
  5419	 *	Description
  5420	 *		Returns a calculated IMA hash of the *file*.
  5421	 *		If the hash is larger than *size*, then only *size*
  5422	 *		bytes will be copied to *dst*
  5423	 *	Return
  5424	 *		The **hash_algo** is returned on success,
  5425	 *		**-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
  5426	 *		invalid arguments are passed.
  5427	 *
  5428	 * void *bpf_kptr_xchg(void *map_value, void *ptr)
  5429	 *	Description
  5430	 *		Exchange kptr at pointer *map_value* with *ptr*, and return the
  5431	 *		old value. *ptr* can be NULL, otherwise it must be a referenced
  5432	 *		pointer which will be released when this helper is called.
  5433	 *	Return
  5434	 *		The old value of kptr (which can be NULL). The returned pointer
  5435	 *		if not NULL, is a reference which must be released using its
  5436	 *		corresponding release function, or moved into a BPF map before
  5437	 *		program exit.
  5438	 *
  5439	 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
  5440	 * 	Description
  5441	 * 		Perform a lookup in *percpu map* for an entry associated to
  5442	 * 		*key* on *cpu*.
  5443	 * 	Return
  5444	 * 		Map value associated to *key* on *cpu*, or **NULL** if no entry
  5445	 * 		was found or *cpu* is invalid.
  5446	 *
  5447	 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
  5448	 *	Description
  5449	 *		Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
  5450	 *	Return
  5451	 *		*sk* if casting is valid, or **NULL** otherwise.
  5452	 *
  5453	 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
  5454	 *	Description
  5455	 *		Get a dynptr to local memory *data*.
  5456	 *
  5457	 *		*data* must be a ptr to a map value.
  5458	 *		The maximum *size* supported is DYNPTR_MAX_SIZE.
  5459	 *		*flags* is currently unused.
  5460	 *	Return
  5461	 *		0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
  5462	 *		-EINVAL if flags is not 0.
  5463	 *
  5464	 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
  5465	 *	Description
  5466	 *		Reserve *size* bytes of payload in a ring buffer *ringbuf*
  5467	 *		through the dynptr interface. *flags* must be 0.
  5468	 *
  5469	 *		Please note that a corresponding bpf_ringbuf_submit_dynptr or
  5470	 *		bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
  5471	 *		reservation fails. This is enforced by the verifier.
  5472	 *	Return
  5473	 *		0 on success, or a negative error in case of failure.
  5474	 *
  5475	 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
  5476	 *	Description
  5477	 *		Submit reserved ring buffer sample, pointed to by *data*,
  5478	 *		through the dynptr interface. This is a no-op if the dynptr is
  5479	 *		invalid/null.
  5480	 *
  5481	 *		For more information on *flags*, please see
  5482	 *		'bpf_ringbuf_submit'.
  5483	 *	Return
  5484	 *		Nothing. Always succeeds.
  5485	 *
  5486	 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
  5487	 *	Description
  5488	 *		Discard reserved ring buffer sample through the dynptr
  5489	 *		interface. This is a no-op if the dynptr is invalid/null.
  5490	 *
  5491	 *		For more information on *flags*, please see
  5492	 *		'bpf_ringbuf_discard'.
  5493	 *	Return
  5494	 *		Nothing. Always succeeds.
  5495	 *
  5496	 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
  5497	 *	Description
  5498	 *		Read *len* bytes from *src* into *dst*, starting from *offset*
  5499	 *		into *src*.
  5500	 *		*flags* is currently unused.
  5501	 *	Return
  5502	 *		0 on success, -E2BIG if *offset* + *len* exceeds the length
  5503	 *		of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
  5504	 *		*flags* is not 0.
  5505	 *
  5506	 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
  5507	 *	Description
  5508	 *		Write *len* bytes from *src* into *dst*, starting from *offset*
  5509	 *		into *dst*.
  5510	 *
  5511	 *		*flags* must be 0 except for skb-type dynptrs.
  5512	 *
  5513	 *		For skb-type dynptrs:
  5514	 *		    *  All data slices of the dynptr are automatically
  5515	 *		       invalidated after **bpf_dynptr_write**\ (). This is
  5516	 *		       because writing may pull the skb and change the
  5517	 *		       underlying packet buffer.
  5518	 *
  5519	 *		    *  For *flags*, please see the flags accepted by
  5520	 *		       **bpf_skb_store_bytes**\ ().
  5521	 *	Return
  5522	 *		0 on success, -E2BIG if *offset* + *len* exceeds the length
  5523	 *		of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
  5524	 *		is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
  5525	 *		other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
  5526	 *
  5527	 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
  5528	 *	Description
  5529	 *		Get a pointer to the underlying dynptr data.
  5530	 *
  5531	 *		*len* must be a statically known value. The returned data slice
  5532	 *		is invalidated whenever the dynptr is invalidated.
  5533	 *
  5534	 *		skb and xdp type dynptrs may not use bpf_dynptr_data. They should
  5535	 *		instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
  5536	 *	Return
  5537	 *		Pointer to the underlying dynptr data, NULL if the dynptr is
  5538	 *		read-only, if the dynptr is invalid, or if the offset and length
  5539	 *		is out of bounds.
  5540	 *
  5541	 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
  5542	 *	Description
  5543	 *		Try to issue a SYN cookie for the packet with corresponding
  5544	 *		IPv4/TCP headers, *iph* and *th*, without depending on a
  5545	 *		listening socket.
  5546	 *
  5547	 *		*iph* points to the IPv4 header.
  5548	 *
  5549	 *		*th* points to the start of the TCP header, while *th_len*
  5550	 *		contains the length of the TCP header (at least
  5551	 *		**sizeof**\ (**struct tcphdr**)).
  5552	 *	Return
  5553	 *		On success, lower 32 bits hold the generated SYN cookie in
  5554	 *		followed by 16 bits which hold the MSS value for that cookie,
  5555	 *		and the top 16 bits are unused.
  5556	 *
  5557	 *		On failure, the returned value is one of the following:
  5558	 *
  5559	 *		**-EINVAL** if *th_len* is invalid.
  5560	 *
  5561	 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
  5562	 *	Description
  5563	 *		Try to issue a SYN cookie for the packet with corresponding
  5564	 *		IPv6/TCP headers, *iph* and *th*, without depending on a
  5565	 *		listening socket.
  5566	 *
  5567	 *		*iph* points to the IPv6 header.
  5568	 *
  5569	 *		*th* points to the start of the TCP header, while *th_len*
  5570	 *		contains the length of the TCP header (at least
  5571	 *		**sizeof**\ (**struct tcphdr**)).
  5572	 *	Return
  5573	 *		On success, lower 32 bits hold the generated SYN cookie in
  5574	 *		followed by 16 bits which hold the MSS value for that cookie,
  5575	 *		and the top 16 bits are unused.
  5576	 *
  5577	 *		On failure, the returned value is one of the following:
  5578	 *
  5579	 *		**-EINVAL** if *th_len* is invalid.
  5580	 *
  5581	 *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
  5582	 *
  5583	 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
  5584	 *	Description
  5585	 *		Check whether *iph* and *th* contain a valid SYN cookie ACK
  5586	 *		without depending on a listening socket.
  5587	 *
  5588	 *		*iph* points to the IPv4 header.
  5589	 *
  5590	 *		*th* points to the TCP header.
  5591	 *	Return
  5592	 *		0 if *iph* and *th* are a valid SYN cookie ACK.
  5593	 *
  5594	 *		On failure, the returned value is one of the following:
  5595	 *
  5596	 *		**-EACCES** if the SYN cookie is not valid.
  5597	 *
  5598	 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
  5599	 *	Description
  5600	 *		Check whether *iph* and *th* contain a valid SYN cookie ACK
  5601	 *		without depending on a listening socket.
  5602	 *
  5603	 *		*iph* points to the IPv6 header.
  5604	 *
  5605	 *		*th* points to the TCP header.
  5606	 *	Return
  5607	 *		0 if *iph* and *th* are a valid SYN cookie ACK.
  5608	 *
  5609	 *		On failure, the returned value is one of the following:
  5610	 *
  5611	 *		**-EACCES** if the SYN cookie is not valid.
  5612	 *
  5613	 *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
  5614	 *
  5615	 * u64 bpf_ktime_get_tai_ns(void)
  5616	 *	Description
  5617	 *		A nonsettable system-wide clock derived from wall-clock time but
  5618	 *		ignoring leap seconds.  This clock does not experience
  5619	 *		discontinuities and backwards jumps caused by NTP inserting leap
  5620	 *		seconds as CLOCK_REALTIME does.
  5621	 *
  5622	 *		See: **clock_gettime**\ (**CLOCK_TAI**)
  5623	 *	Return
  5624	 *		Current *ktime*.
  5625	 *
  5626	 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
  5627	 *	Description
  5628	 *		Drain samples from the specified user ring buffer, and invoke
  5629	 *		the provided callback for each such sample:
  5630	 *
  5631	 *		long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
  5632	 *
  5633	 *		If **callback_fn** returns 0, the helper will continue to try
  5634	 *		and drain the next sample, up to a maximum of
  5635	 *		BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
  5636	 *		the helper will skip the rest of the samples and return. Other
  5637	 *		return values are not used now, and will be rejected by the
  5638	 *		verifier.
  5639	 *	Return
  5640	 *		The number of drained samples if no error was encountered while
  5641	 *		draining samples, or 0 if no samples were present in the ring
  5642	 *		buffer. If a user-space producer was epoll-waiting on this map,
  5643	 *		and at least one sample was drained, they will receive an event
  5644	 *		notification notifying them of available space in the ring
  5645	 *		buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
  5646	 *		function, no wakeup notification will be sent. If the
  5647	 *		BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
  5648	 *		be sent even if no sample was drained.
  5649	 *
  5650	 *		On failure, the returned value is one of the following:
  5651	 *
  5652	 *		**-EBUSY** if the ring buffer is contended, and another calling
  5653	 *		context was concurrently draining the ring buffer.
  5654	 *
  5655	 *		**-EINVAL** if user-space is not properly tracking the ring
  5656	 *		buffer due to the producer position not being aligned to 8
  5657	 *		bytes, a sample not being aligned to 8 bytes, or the producer
  5658	 *		position not matching the advertised length of a sample.
  5659	 *
  5660	 *		**-E2BIG** if user-space has tried to publish a sample which is
  5661	 *		larger than the size of the ring buffer, or which cannot fit
  5662	 *		within a struct bpf_dynptr.
  5663	 *
  5664	 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
  5665	 *	Description
  5666	 *		Get a bpf_local_storage from the *cgroup*.
  5667	 *
  5668	 *		Logically, it could be thought of as getting the value from
  5669	 *		a *map* with *cgroup* as the **key**.  From this
  5670	 *		perspective,  the usage is not much different from
  5671	 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
  5672	 *		helper enforces the key must be a cgroup struct and the map must also
  5673	 *		be a **BPF_MAP_TYPE_CGRP_STORAGE**.
  5674	 *
  5675	 *		In reality, the local-storage value is embedded directly inside of the
  5676	 *		*cgroup* object itself, rather than being located in the
  5677	 *		**BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
  5678	 *		queried for some *map* on a *cgroup* object, the kernel will perform an
  5679	 *		O(n) iteration over all of the live local-storage values for that
  5680	 *		*cgroup* object until the local-storage value for the *map* is found.
  5681	 *
  5682	 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
  5683	 *		used such that a new bpf_local_storage will be
  5684	 *		created if one does not exist.  *value* can be used
  5685	 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
  5686	 *		the initial value of a bpf_local_storage.  If *value* is
  5687	 *		**NULL**, the new bpf_local_storage will be zero initialized.
  5688	 *	Return
  5689	 *		A bpf_local_storage pointer is returned on success.
  5690	 *
  5691	 *		**NULL** if not found or there was an error in adding
  5692	 *		a new bpf_local_storage.
  5693	 *
  5694	 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
  5695	 *	Description
  5696	 *		Delete a bpf_local_storage from a *cgroup*.
  5697	 *	Return
  5698	 *		0 on success.
  5699	 *
  5700	 *		**-ENOENT** if the bpf_local_storage cannot be found.
  5701	 */
  5702	#define ___BPF_FUNC_MAPPER(FN, ctx...)			\
  5703		FN(unspec, 0, ##ctx)				\
  5704		FN(map_lookup_elem, 1, ##ctx)			\
  5705		FN(map_update_elem, 2, ##ctx)			\
  5706		FN(map_delete_elem, 3, ##ctx)			\
  5707		FN(probe_read, 4, ##ctx)			\
  5708		FN(ktime_get_ns, 5, ##ctx)			\
  5709		FN(trace_printk, 6, ##ctx)			\
  5710		FN(get_prandom_u32, 7, ##ctx)			\
  5711		FN(get_smp_processor_id, 8, ##ctx)		\
  5712		FN(skb_store_bytes, 9, ##ctx)			\
  5713		FN(l3_csum_replace, 10, ##ctx)			\
  5714		FN(l4_csum_replace, 11, ##ctx)			\
  5715		FN(tail_call, 12, ##ctx)			\
  5716		FN(clone_redirect, 13, ##ctx)			\
  5717		FN(get_current_pid_tgid, 14, ##ctx)		\
  5718		FN(get_current_uid_gid, 15, ##ctx)		\
  5719		FN(get_current_comm, 16, ##ctx)			\
  5720		FN(get_cgroup_classid, 17, ##ctx)		\
  5721		FN(skb_vlan_push, 18, ##ctx)			\
  5722		FN(skb_vlan_pop, 19, ##ctx)			\
  5723		FN(skb_get_tunnel_key, 20, ##ctx)		\
  5724		FN(skb_set_tunnel_key, 21, ##ctx)		\
  5725		FN(perf_event_read, 22, ##ctx)			\
  5726		FN(redirect, 23, ##ctx)				\
  5727		FN(get_route_realm, 24, ##ctx)			\
  5728		FN(perf_event_output, 25, ##ctx)		\
  5729		FN(skb_load_bytes, 26, ##ctx)			\
  5730		FN(get_stackid, 27, ##ctx)			\
  5731		FN(csum_diff, 28, ##ctx)			\
  5732		FN(skb_get_tunnel_opt, 29, ##ctx)		\
  5733		FN(skb_set_tunnel_opt, 30, ##ctx)		\
  5734		FN(skb_change_proto, 31, ##ctx)			\
  5735		FN(skb_change_type, 32, ##ctx)			\
  5736		FN(skb_under_cgroup, 33, ##ctx)			\
  5737		FN(get_hash_recalc, 34, ##ctx)			\
  5738		FN(get_current_task, 35, ##ctx)			\
  5739		FN(probe_write_user, 36, ##ctx)			\
  5740		FN(current_task_under_cgroup, 37, ##ctx)	\
  5741		FN(skb_change_tail, 38, ##ctx)			\
  5742		FN(skb_pull_data, 39, ##ctx)			\
  5743		FN(csum_update, 40, ##ctx)			\
  5744		FN(set_hash_invalid, 41, ##ctx)			\
  5745		FN(get_numa_node_id, 42, ##ctx)			\
  5746		FN(skb_change_head, 43, ##ctx)			\
  5747		FN(xdp_adjust_head, 44, ##ctx)			\
  5748		FN(probe_read_str, 45, ##ctx)			\
  5749		FN(get_socket_cookie, 46, ##ctx)		\
  5750		FN(get_socket_uid, 47, ##ctx)			\
  5751		FN(set_hash, 48, ##ctx)				\
  5752		FN(setsockopt, 49, ##ctx)			\
  5753		FN(skb_adjust_room, 50, ##ctx)			\
  5754		FN(redirect_map, 51, ##ctx)			\
  5755		FN(sk_redirect_map, 52, ##ctx)			\
  5756		FN(sock_map_update, 53, ##ctx)			\
  5757		FN(xdp_adjust_meta, 54, ##ctx)			\
  5758		FN(perf_event_read_value, 55, ##ctx)		\
  5759		FN(perf_prog_read_value, 56, ##ctx)		\
  5760		FN(getsockopt, 57, ##ctx)			\
  5761		FN(override_return, 58, ##ctx)			\
  5762		FN(sock_ops_cb_flags_set, 59, ##ctx)		\
  5763		FN(msg_redirect_map, 60, ##ctx)			\
  5764		FN(msg_apply_bytes, 61, ##ctx)			\
  5765		FN(msg_cork_bytes, 62, ##ctx)			\
  5766		FN(msg_pull_data, 63, ##ctx)			\
  5767		FN(bind, 64, ##ctx)				\
  5768		FN(xdp_adjust_tail, 65, ##ctx)			\
  5769		FN(skb_get_xfrm_state, 66, ##ctx)		\
  5770		FN(get_stack, 67, ##ctx)			\
  5771		FN(skb_load_bytes_relative, 68, ##ctx)		\
  5772		FN(fib_lookup, 69, ##ctx)			\
  5773		FN(sock_hash_update, 70, ##ctx)			\
  5774		FN(msg_redirect_hash, 71, ##ctx)		\
  5775		FN(sk_redirect_hash, 72, ##ctx)			\
  5776		FN(lwt_push_encap, 73, ##ctx)			\
  5777		FN(lwt_seg6_store_bytes, 74, ##ctx)		\
  5778		FN(lwt_seg6_adjust_srh, 75, ##ctx)		\
  5779		FN(lwt_seg6_action, 76, ##ctx)			\
  5780		FN(rc_repeat, 77, ##ctx)			\
  5781		FN(rc_keydown, 78, ##ctx)			\
  5782		FN(skb_cgroup_id, 79, ##ctx)			\
  5783		FN(get_current_cgroup_id, 80, ##ctx)		\
  5784		FN(get_local_storage, 81, ##ctx)		\
  5785		FN(sk_select_reuseport, 82, ##ctx)		\
  5786		FN(skb_ancestor_cgroup_id, 83, ##ctx)		\
  5787		FN(sk_lookup_tcp, 84, ##ctx)			\
  5788		FN(sk_lookup_udp, 85, ##ctx)			\
  5789		FN(sk_release, 86, ##ctx)			\
  5790		FN(map_push_elem, 87, ##ctx)			\
  5791		FN(map_pop_elem, 88, ##ctx)			\
  5792		FN(map_peek_elem, 89, ##ctx)			\
  5793		FN(msg_push_data, 90, ##ctx)			\
  5794		FN(msg_pop_data, 91, ##ctx)			\
  5795		FN(rc_pointer_rel, 92, ##ctx)			\
  5796		FN(spin_lock, 93, ##ctx)			\
  5797		FN(spin_unlock, 94, ##ctx)			\
  5798		FN(sk_fullsock, 95, ##ctx)			\
  5799		FN(tcp_sock, 96, ##ctx)				\
  5800		FN(skb_ecn_set_ce, 97, ##ctx)			\
  5801		FN(get_listener_sock, 98, ##ctx)		\
  5802		FN(skc_lookup_tcp, 99, ##ctx)			\
  5803		FN(tcp_check_syncookie, 100, ##ctx)		\
  5804		FN(sysctl_get_name, 101, ##ctx)			\
  5805		FN(sysctl_get_current_value, 102, ##ctx)	\
  5806		FN(sysctl_get_new_value, 103, ##ctx)		\
  5807		FN(sysctl_set_new_value, 104, ##ctx)		\
  5808		FN(strtol, 105, ##ctx)				\
  5809		FN(strtoul, 106, ##ctx)				\
  5810		FN(sk_storage_get, 107, ##ctx)			\
  5811		FN(sk_storage_delete, 108, ##ctx)		\
  5812		FN(send_signal, 109, ##ctx)			\
  5813		FN(tcp_gen_syncookie, 110, ##ctx)		\
  5814		FN(skb_output, 111, ##ctx)			\
  5815		FN(probe_read_user, 112, ##ctx)			\
  5816		FN(probe_read_kernel, 113, ##ctx)		\
  5817		FN(probe_read_user_str, 114, ##ctx)		\
  5818		FN(probe_read_kernel_str, 115, ##ctx)		\
  5819		FN(tcp_send_ack, 116, ##ctx)			\
  5820		FN(send_signal_thread, 117, ##ctx)		\
  5821		FN(jiffies64, 118, ##ctx)			\
  5822		FN(read_branch_records, 119, ##ctx)		\
  5823		FN(get_ns_current_pid_tgid, 120, ##ctx)		\
  5824		FN(xdp_output, 121, ##ctx)			\
  5825		FN(get_netns_cookie, 122, ##ctx)		\
  5826		FN(get_current_ancestor_cgroup_id, 123, ##ctx)	\
  5827		FN(sk_assign, 124, ##ctx)			\
  5828		FN(ktime_get_boot_ns, 125, ##ctx)		\
  5829		FN(seq_printf, 126, ##ctx)			\
  5830		FN(seq_write, 127, ##ctx)			\
  5831		FN(sk_cgroup_id, 128, ##ctx)			\
  5832		FN(sk_ancestor_cgroup_id, 129, ##ctx)		\
  5833		FN(ringbuf_output, 130, ##ctx)			\
  5834		FN(ringbuf_reserve, 131, ##ctx)			\
  5835		FN(ringbuf_submit, 132, ##ctx)			\
  5836		FN(ringbuf_discard, 133, ##ctx)			\
  5837		FN(ringbuf_query, 134, ##ctx)			\
  5838		FN(csum_level, 135, ##ctx)			\
  5839		FN(skc_to_tcp6_sock, 136, ##ctx)		\
  5840		FN(skc_to_tcp_sock, 137, ##ctx)			\
  5841		FN(skc_to_tcp_timewait_sock, 138, ##ctx)	\
  5842		FN(skc_to_tcp_request_sock, 139, ##ctx)		\
  5843		FN(skc_to_udp6_sock, 140, ##ctx)		\
  5844		FN(get_task_stack, 141, ##ctx)			\
  5845		FN(load_hdr_opt, 142, ##ctx)			\
  5846		FN(store_hdr_opt, 143, ##ctx)			\
  5847		FN(reserve_hdr_opt, 144, ##ctx)			\
  5848		FN(inode_storage_get, 145, ##ctx)		\
  5849		FN(inode_storage_delete, 146, ##ctx)		\
  5850		FN(d_path, 147, ##ctx)				\
  5851		FN(copy_from_user, 148, ##ctx)			\
  5852		FN(snprintf_btf, 149, ##ctx)			\
  5853		FN(seq_printf_btf, 150, ##ctx)			\
  5854		FN(skb_cgroup_classid, 151, ##ctx)		\
  5855		FN(redirect_neigh, 152, ##ctx)			\
  5856		FN(per_cpu_ptr, 153, ##ctx)			\
  5857		FN(this_cpu_ptr, 154, ##ctx)			\
  5858		FN(redirect_peer, 155, ##ctx)			\
  5859		FN(task_storage_get, 156, ##ctx)		\
  5860		FN(task_storage_delete, 157, ##ctx)		\
  5861		FN(get_current_task_btf, 158, ##ctx)		\
  5862		FN(bprm_opts_set, 159, ##ctx)			\
  5863		FN(ktime_get_coarse_ns, 160, ##ctx)		\
  5864		FN(ima_inode_hash, 161, ##ctx)			\
  5865		FN(sock_from_file, 162, ##ctx)			\
  5866		FN(check_mtu, 163, ##ctx)			\
  5867		FN(for_each_map_elem, 164, ##ctx)		\
  5868		FN(snprintf, 165, ##ctx)			\
  5869		FN(sys_bpf, 166, ##ctx)				\
  5870		FN(btf_find_by_name_kind, 167, ##ctx)		\
  5871		FN(sys_close, 168, ##ctx)			\
  5872		FN(timer_init, 169, ##ctx)			\
  5873		FN(timer_set_callback, 170, ##ctx)		\
  5874		FN(timer_start, 171, ##ctx)			\
  5875		FN(timer_cancel, 172, ##ctx)			\
  5876		FN(get_func_ip, 173, ##ctx)			\
  5877		FN(get_attach_cookie, 174, ##ctx)		\
  5878		FN(task_pt_regs, 175, ##ctx)			\
  5879		FN(get_branch_snapshot, 176, ##ctx)		\
  5880		FN(trace_vprintk, 177, ##ctx)			\
  5881		FN(skc_to_unix_sock, 178, ##ctx)		\
  5882		FN(kallsyms_lookup_name, 179, ##ctx)		\
  5883		FN(find_vma, 180, ##ctx)			\
  5884		FN(loop, 181, ##ctx)				\
  5885		FN(strncmp, 182, ##ctx)				\
  5886		FN(get_func_arg, 183, ##ctx)			\
  5887		FN(get_func_ret, 184, ##ctx)			\
  5888		FN(get_func_arg_cnt, 185, ##ctx)		\
  5889		FN(get_retval, 186, ##ctx)			\
  5890		FN(set_retval, 187, ##ctx)			\
  5891		FN(xdp_get_buff_len, 188, ##ctx)		\
  5892		FN(xdp_load_bytes, 189, ##ctx)			\
  5893		FN(xdp_store_bytes, 190, ##ctx)			\
  5894		FN(copy_from_user_task, 191, ##ctx)		\
  5895		FN(skb_set_tstamp, 192, ##ctx)			\
  5896		FN(ima_file_hash, 193, ##ctx)			\
  5897		FN(kptr_xchg, 194, ##ctx)			\
  5898		FN(map_lookup_percpu_elem, 195, ##ctx)		\
  5899		FN(skc_to_mptcp_sock, 196, ##ctx)		\
  5900		FN(dynptr_from_mem, 197, ##ctx)			\
  5901		FN(ringbuf_reserve_dynptr, 198, ##ctx)		\
  5902		FN(ringbuf_submit_dynptr, 199, ##ctx)		\
  5903		FN(ringbuf_discard_dynptr, 200, ##ctx)		\
  5904		FN(dynptr_read, 201, ##ctx)			\
  5905		FN(dynptr_write, 202, ##ctx)			\
  5906		FN(dynptr_data, 203, ##ctx)			\
  5907		FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx)	\
  5908		FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx)	\
  5909		FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx)	\
  5910		FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)	\
  5911		FN(ktime_get_tai_ns, 208, ##ctx)		\
  5912		FN(user_ringbuf_drain, 209, ##ctx)		\
  5913		FN(cgrp_storage_get, 210, ##ctx)		\
  5914		FN(cgrp_storage_delete, 211, ##ctx)		\
  5915		/* */
  5916	
  5917	/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
  5918	 * know or care about integer value that is now passed as second argument
  5919	 */
  5920	#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
  5921	#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
  5922	
  5923	/* integer value in 'imm' field of BPF_CALL instruction selects which helper
  5924	 * function eBPF program intends to call
  5925	 */
  5926	#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
  5927	enum bpf_func_id {
  5928		___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  5929		__BPF_FUNC_MAX_ID,
  5930	};
  5931	#undef __BPF_ENUM_FN
  5932	
  5933	/* All flags used by eBPF helper functions, placed here. */
  5934	
  5935	/* BPF_FUNC_skb_store_bytes flags. */
  5936	enum {
  5937		BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),
  5938		BPF_F_INVALIDATE_HASH		= (1ULL << 1),
  5939	};
  5940	
  5941	/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  5942	 * First 4 bits are for passing the header field size.
  5943	 */
  5944	enum {
  5945		BPF_F_HDR_FIELD_MASK		= 0xfULL,
  5946	};
  5947	
  5948	/* BPF_FUNC_l4_csum_replace flags. */
  5949	enum {
  5950		BPF_F_PSEUDO_HDR		= (1ULL << 4),
  5951		BPF_F_MARK_MANGLED_0		= (1ULL << 5),
  5952		BPF_F_MARK_ENFORCE		= (1ULL << 6),
  5953		BPF_F_IPV6			= (1ULL << 7),
  5954	};
  5955	
  5956	/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  5957	enum {
  5958		BPF_F_TUNINFO_IPV6		= (1ULL << 0),
  5959	};
  5960	
  5961	/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
  5962	enum {
  5963		BPF_F_SKIP_FIELD_MASK		= 0xffULL,
  5964		BPF_F_USER_STACK		= (1ULL << 8),
  5965	/* flags used by BPF_FUNC_get_stackid only. */
  5966		BPF_F_FAST_STACK_CMP		= (1ULL << 9),
  5967		BPF_F_REUSE_STACKID		= (1ULL << 10),
  5968	/* flags used by BPF_FUNC_get_stack only. */
  5969		BPF_F_USER_BUILD_ID		= (1ULL << 11),
  5970	};
  5971	
  5972	/* BPF_FUNC_skb_set_tunnel_key flags. */
  5973	enum {
  5974		BPF_F_ZERO_CSUM_TX		= (1ULL << 1),
  5975		BPF_F_DONT_FRAGMENT		= (1ULL << 2),
  5976		BPF_F_SEQ_NUMBER		= (1ULL << 3),
  5977		BPF_F_NO_TUNNEL_KEY		= (1ULL << 4),
  5978	};
  5979	
  5980	/* BPF_FUNC_skb_get_tunnel_key flags. */
  5981	enum {
  5982		BPF_F_TUNINFO_FLAGS		= (1ULL << 4),
  5983	};
  5984	
  5985	/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
  5986	 * BPF_FUNC_perf_event_read_value flags.
  5987	 */
  5988	enum {
  5989		BPF_F_INDEX_MASK		= 0xffffffffULL,
  5990		BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,
  5991	/* BPF_FUNC_perf_event_output for sk_buff input context. */
  5992		BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),
  5993	};
  5994	
  5995	/* Current network namespace */
  5996	enum {
  5997		BPF_F_CURRENT_NETNS		= (-1L),
  5998	};
  5999	
  6000	/* BPF_FUNC_csum_level level values. */
  6001	enum {
  6002		BPF_CSUM_LEVEL_QUERY,
  6003		BPF_CSUM_LEVEL_INC,
  6004		BPF_CSUM_LEVEL_DEC,
  6005		BPF_CSUM_LEVEL_RESET,
  6006	};
  6007	
  6008	/* BPF_FUNC_skb_adjust_room flags. */
  6009	enum {
  6010		BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),
  6011		BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),
  6012		BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),
  6013		BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),
  6014		BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),
  6015		BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),
  6016		BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),
  6017		BPF_F_ADJ_ROOM_DECAP_L3_IPV4	= (1ULL << 7),
  6018		BPF_F_ADJ_ROOM_DECAP_L3_IPV6	= (1ULL << 8),
  6019	};
  6020	
  6021	enum {
  6022		BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,
  6023		BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,
  6024	};
  6025	
  6026	#define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \
  6027						  BPF_ADJ_ROOM_ENCAP_L2_MASK) \
  6028						 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
  6029	
  6030	/* BPF_FUNC_sysctl_get_name flags. */
  6031	enum {
  6032		BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),
  6033	};
  6034	
  6035	/* BPF_FUNC_<kernel_obj>_storage_get flags */
  6036	enum {
  6037		BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),
  6038		/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
  6039		 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
  6040		 */
  6041		BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
  6042	};
  6043	
  6044	/* BPF_FUNC_read_branch_records flags. */
  6045	enum {
  6046		BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),
  6047	};
  6048	
  6049	/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
  6050	 * BPF_FUNC_bpf_ringbuf_output flags.
  6051	 */
  6052	enum {
  6053		BPF_RB_NO_WAKEUP		= (1ULL << 0),
  6054		BPF_RB_FORCE_WAKEUP		= (1ULL << 1),
  6055	};
  6056	
  6057	/* BPF_FUNC_bpf_ringbuf_query flags */
  6058	enum {
  6059		BPF_RB_AVAIL_DATA = 0,
  6060		BPF_RB_RING_SIZE = 1,
  6061		BPF_RB_CONS_POS = 2,
  6062		BPF_RB_PROD_POS = 3,
  6063	};
  6064	
  6065	/* BPF ring buffer constants */
  6066	enum {
  6067		BPF_RINGBUF_BUSY_BIT		= (1U << 31),
  6068		BPF_RINGBUF_DISCARD_BIT		= (1U << 30),
  6069		BPF_RINGBUF_HDR_SZ		= 8,
  6070	};
  6071	
  6072	/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
  6073	enum {
  6074		BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),
  6075		BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),
  6076	};
  6077	
  6078	/* Mode for BPF_FUNC_skb_adjust_room helper. */
  6079	enum bpf_adj_room_mode {
  6080		BPF_ADJ_ROOM_NET,
  6081		BPF_ADJ_ROOM_MAC,
  6082	};
  6083	
  6084	/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
  6085	enum bpf_hdr_start_off {
  6086		BPF_HDR_START_MAC,
  6087		BPF_HDR_START_NET,
  6088	};
  6089	
  6090	/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
  6091	enum bpf_lwt_encap_mode {
  6092		BPF_LWT_ENCAP_SEG6,
  6093		BPF_LWT_ENCAP_SEG6_INLINE,
  6094		BPF_LWT_ENCAP_IP,
  6095	};
  6096	
  6097	/* Flags for bpf_bprm_opts_set helper */
  6098	enum {
  6099		BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
  6100	};
  6101	
  6102	/* Flags for bpf_redirect and bpf_redirect_map helpers */
  6103	enum {
  6104		BPF_F_INGRESS		= (1ULL << 0), /* used for skb path */
  6105		BPF_F_BROADCAST		= (1ULL << 3), /* used for XDP path */
  6106		BPF_F_EXCLUDE_INGRESS	= (1ULL << 4), /* used for XDP path */
  6107	#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)
  6108	};
  6109	
  6110	#define __bpf_md_ptr(type, name)	\
  6111	union {					\
  6112		type name;			\
  6113		__u64 :64;			\
  6114	} __attribute__((aligned(8)))
  6115	
  6116	enum {
  6117		BPF_SKB_TSTAMP_UNSPEC,
  6118		BPF_SKB_TSTAMP_DELIVERY_MONO,	/* tstamp has mono delivery time */
  6119		/* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
  6120		 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
  6121		 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
  6122		 */
  6123	};
  6124	
  6125	/* user accessible mirror of in-kernel sk_buff.
  6126	 * new fields can only be added to the end of this structure
  6127	 */
  6128	struct __sk_buff {
  6129		__u32 len;
  6130		__u32 pkt_type;
  6131		__u32 mark;
  6132		__u32 queue_mapping;
  6133		__u32 protocol;
  6134		__u32 vlan_present;
  6135		__u32 vlan_tci;
  6136		__u32 vlan_proto;
  6137		__u32 priority;
  6138		__u32 ingress_ifindex;
  6139		__u32 ifindex;
  6140		__u32 tc_index;
  6141		__u32 cb[5];
  6142		__u32 hash;
  6143		__u32 tc_classid;
  6144		__u32 data;
  6145		__u32 data_end;
  6146		__u32 napi_id;
  6147	
  6148		/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
  6149		__u32 family;
  6150		__u32 remote_ip4;	/* Stored in network byte order */
  6151		__u32 local_ip4;	/* Stored in network byte order */
  6152		__u32 remote_ip6[4];	/* Stored in network byte order */
  6153		__u32 local_ip6[4];	/* Stored in network byte order */
  6154		__u32 remote_port;	/* Stored in network byte order */
  6155		__u32 local_port;	/* stored in host byte order */
  6156		/* ... here. */
  6157	
  6158		__u32 data_meta;
  6159		__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
  6160		__u64 tstamp;
  6161		__u32 wire_len;
  6162		__u32 gso_segs;
  6163		__bpf_md_ptr(struct bpf_sock *, sk);
  6164		__u32 gso_size;
  6165		__u8  tstamp_type;
  6166		__u32 :24;		/* Padding, future use. */
  6167		__u64 hwtstamp;
  6168	};
  6169	
  6170	struct bpf_tunnel_key {
  6171		__u32 tunnel_id;
  6172		union {
  6173			__u32 remote_ipv4;
  6174			__u32 remote_ipv6[4];
  6175		};
  6176		__u8 tunnel_tos;
  6177		__u8 tunnel_ttl;
  6178		union {
  6179			__u16 tunnel_ext;	/* compat */
  6180			__be16 tunnel_flags;
  6181		};
  6182		__u32 tunnel_label;
  6183		union {
  6184			__u32 local_ipv4;
  6185			__u32 local_ipv6[4];
  6186		};
  6187	};
  6188	
  6189	/* user accessible mirror of in-kernel xfrm_state.
  6190	 * new fields can only be added to the end of this structure
  6191	 */
  6192	struct bpf_xfrm_state {
  6193		__u32 reqid;
  6194		__u32 spi;	/* Stored in network byte order */
  6195		__u16 family;
  6196		__u16 ext;	/* Padding, future use. */
  6197		union {
  6198			__u32 remote_ipv4;	/* Stored in network byte order */
  6199			__u32 remote_ipv6[4];	/* Stored in network byte order */
  6200		};
  6201	};
  6202	
  6203	/* Generic BPF return codes which all BPF program types may support.
  6204	 * The values are binary compatible with their TC_ACT_* counter-part to
  6205	 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  6206	 * programs.
  6207	 *
  6208	 * XDP is handled seprately, see XDP_*.
  6209	 */
  6210	enum bpf_ret_code {
  6211		BPF_OK = 0,
  6212		/* 1 reserved */
  6213		BPF_DROP = 2,
  6214		/* 3-6 reserved */
  6215		BPF_REDIRECT = 7,
  6216		/* >127 are reserved for prog type specific return codes.
  6217		 *
  6218		 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
  6219		 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
  6220		 *    changed and should be routed based on its new L3 header.
  6221		 *    (This is an L3 redirect, as opposed to L2 redirect
  6222		 *    represented by BPF_REDIRECT above).
  6223		 */
  6224		BPF_LWT_REROUTE = 128,
  6225		/* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
  6226		 *   to indicate that no custom dissection was performed, and
  6227		 *   fallback to standard dissector is requested.
  6228		 */
  6229		BPF_FLOW_DISSECTOR_CONTINUE = 129,
  6230	};
  6231	
  6232	struct bpf_sock {
  6233		__u32 bound_dev_if;
  6234		__u32 family;
  6235		__u32 type;
  6236		__u32 protocol;
  6237		__u32 mark;
  6238		__u32 priority;
  6239		/* IP address also allows 1 and 2 bytes access */
  6240		__u32 src_ip4;
  6241		__u32 src_ip6[4];
  6242		__u32 src_port;		/* host byte order */
  6243		__be16 dst_port;	/* network byte order */
  6244		__u16 :16;		/* zero padding */
  6245		__u32 dst_ip4;
  6246		__u32 dst_ip6[4];
  6247		__u32 state;
  6248		__s32 rx_queue_mapping;
  6249	};
  6250	
  6251	struct bpf_tcp_sock {
  6252		__u32 snd_cwnd;		/* Sending congestion window		*/
  6253		__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
  6254		__u32 rtt_min;
  6255		__u32 snd_ssthresh;	/* Slow start size threshold		*/
  6256		__u32 rcv_nxt;		/* What we want to receive next		*/
  6257		__u32 snd_nxt;		/* Next sequence we send		*/
  6258		__u32 snd_una;		/* First byte we want an ack for	*/
  6259		__u32 mss_cache;	/* Cached effective mss, not including SACKS */
  6260		__u32 ecn_flags;	/* ECN status bits.			*/
  6261		__u32 rate_delivered;	/* saved rate sample: packets delivered */
  6262		__u32 rate_interval_us;	/* saved rate sample: time elapsed */
  6263		__u32 packets_out;	/* Packets which are "in flight"	*/
  6264		__u32 retrans_out;	/* Retransmitted packets out		*/
  6265		__u32 total_retrans;	/* Total retransmits for entire connection */
  6266		__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
  6267					 * total number of segments in.
  6268					 */
  6269		__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
  6270					 * total number of data segments in.
  6271					 */
  6272		__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
  6273					 * The total number of segments sent.
  6274					 */
  6275		__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
  6276					 * total number of data segments sent.
  6277					 */
  6278		__u32 lost_out;		/* Lost packets			*/
  6279		__u32 sacked_out;	/* SACK'd packets			*/
  6280		__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
  6281					 * sum(delta(rcv_nxt)), or how many bytes
  6282					 * were acked.
  6283					 */
  6284		__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
  6285					 * sum(delta(snd_una)), or how many bytes
  6286					 * were acked.
  6287					 */
  6288		__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups
  6289					 * total number of DSACK blocks received
  6290					 */
  6291		__u32 delivered;	/* Total data packets delivered incl. rexmits */
  6292		__u32 delivered_ce;	/* Like the above but only ECE marked packets */
  6293		__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */
  6294	};
  6295	
  6296	struct bpf_sock_tuple {
  6297		union {
  6298			struct {
  6299				__be32 saddr;
  6300				__be32 daddr;
  6301				__be16 sport;
  6302				__be16 dport;
  6303			} ipv4;
  6304			struct {
  6305				__be32 saddr[4];
  6306				__be32 daddr[4];
  6307				__be16 sport;
  6308				__be16 dport;
  6309			} ipv6;
  6310		};
  6311	};
  6312	
  6313	/* (Simplified) user return codes for tcx prog type.
  6314	 * A valid tcx program must return one of these defined values. All other
  6315	 * return codes are reserved for future use. Must remain compatible with
  6316	 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
  6317	 * return codes are mapped to TCX_NEXT.
  6318	 */
  6319	enum tcx_action_base {
  6320		TCX_NEXT	= -1,
  6321		TCX_PASS	= 0,
  6322		TCX_DROP	= 2,
  6323		TCX_REDIRECT	= 7,
  6324	};
  6325	
  6326	struct bpf_xdp_sock {
  6327		__u32 queue_id;
  6328	};
  6329	
  6330	#define XDP_PACKET_HEADROOM 256
  6331	
  6332	/* User return codes for XDP prog type.
  6333	 * A valid XDP program must return one of these defined values. All other
  6334	 * return codes are reserved for future use. Unknown return codes will
  6335	 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
  6336	 */
  6337	enum xdp_action {
  6338		XDP_ABORTED = 0,
  6339		XDP_DROP,
  6340		XDP_PASS,
  6341		XDP_TX,
  6342		XDP_REDIRECT,
  6343	};
  6344	
  6345	/* user accessible metadata for XDP packet hook
  6346	 * new fields must be added to the end of this structure
  6347	 */
  6348	struct xdp_md {
  6349		__u32 data;
  6350		__u32 data_end;
  6351		__u32 data_meta;
  6352		/* Below access go through struct xdp_rxq_info */
  6353		__u32 ingress_ifindex; /* rxq->dev->ifindex */
  6354		__u32 rx_queue_index;  /* rxq->queue_index  */
  6355	
  6356		__u32 egress_ifindex;  /* txq->dev->ifindex */
  6357	};
  6358	
  6359	/* DEVMAP map-value layout
  6360	 *
  6361	 * The struct data-layout of map-value is a configuration interface.
  6362	 * New members can only be added to the end of this structure.
  6363	 */
  6364	struct bpf_devmap_val {
  6365		__u32 ifindex;   /* device index */
  6366		union {
  6367			int   fd;  /* prog fd on map write */
  6368			__u32 id;  /* prog id on map read */
  6369		} bpf_prog;
  6370	};
  6371	
  6372	/* CPUMAP map-value layout
  6373	 *
  6374	 * The struct data-layout of map-value is a configuration interface.
  6375	 * New members can only be added to the end of this structure.
  6376	 */
  6377	struct bpf_cpumap_val {
  6378		__u32 qsize;	/* queue size to remote target CPU */
  6379		union {
  6380			int   fd;	/* prog fd on map write */
  6381			__u32 id;	/* prog id on map read */
  6382		} bpf_prog;
  6383	};
  6384	
  6385	enum sk_action {
  6386		SK_DROP = 0,
  6387		SK_PASS,
  6388	};
  6389	
  6390	/* user accessible metadata for SK_MSG packet hook, new fields must
  6391	 * be added to the end of this structure
  6392	 */
  6393	struct sk_msg_md {
  6394		__bpf_md_ptr(void *, data);
  6395		__bpf_md_ptr(void *, data_end);
  6396	
  6397		__u32 family;
  6398		__u32 remote_ip4;	/* Stored in network byte order */
  6399		__u32 local_ip4;	/* Stored in network byte order */
  6400		__u32 remote_ip6[4];	/* Stored in network byte order */
  6401		__u32 local_ip6[4];	/* Stored in network byte order */
  6402		__u32 remote_port;	/* Stored in network byte order */
  6403		__u32 local_port;	/* stored in host byte order */
  6404		__u32 size;		/* Total size of sk_msg */
  6405	
  6406		__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
  6407	};
  6408	
  6409	struct sk_reuseport_md {
  6410		/*
  6411		 * Start of directly accessible data. It begins from
  6412		 * the tcp/udp header.
  6413		 */
  6414		__bpf_md_ptr(void *, data);
  6415		/* End of directly accessible data */
  6416		__bpf_md_ptr(void *, data_end);
  6417		/*
  6418		 * Total length of packet (starting from the tcp/udp header).
  6419		 * Note that the directly accessible bytes (data_end - data)
  6420		 * could be less than this "len".  Those bytes could be
  6421		 * indirectly read by a helper "bpf_skb_load_bytes()".
  6422		 */
  6423		__u32 len;
  6424		/*
  6425		 * Eth protocol in the mac header (network byte order). e.g.
  6426		 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
  6427		 */
  6428		__u32 eth_protocol;
  6429		__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
  6430		__u32 bind_inany;	/* Is sock bound to an INANY address? */
  6431		__u32 hash;		/* A hash of the packet 4 tuples */
  6432		/* When reuse->migrating_sk is NULL, it is selecting a sk for the
  6433		 * new incoming connection request (e.g. selecting a listen sk for
  6434		 * the received SYN in the TCP case).  reuse->sk is one of the sk
  6435		 * in the reuseport group. The bpf prog can use reuse->sk to learn
  6436		 * the local listening ip/port without looking into the skb.
  6437		 *
  6438		 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
  6439		 * reuse->migrating_sk is the socket that needs to be migrated
  6440		 * to another listening socket.  migrating_sk could be a fullsock
  6441		 * sk that is fully established or a reqsk that is in-the-middle
  6442		 * of 3-way handshake.
  6443		 */
  6444		__bpf_md_ptr(struct bpf_sock *, sk);
  6445		__bpf_md_ptr(struct bpf_sock *, migrating_sk);
  6446	};
  6447	
  6448	#define BPF_TAG_SIZE	8
  6449	
  6450	struct bpf_prog_info {
  6451		__u32 type;
  6452		__u32 id;
  6453		__u8  tag[BPF_TAG_SIZE];
  6454		__u32 jited_prog_len;
  6455		__u32 xlated_prog_len;
  6456		__aligned_u64 jited_prog_insns;
  6457		__aligned_u64 xlated_prog_insns;
  6458		__u64 load_time;	/* ns since boottime */
  6459		__u32 created_by_uid;
  6460		__u32 nr_map_ids;
  6461		__aligned_u64 map_ids;
  6462		char name[BPF_OBJ_NAME_LEN];
  6463		__u32 ifindex;
  6464		__u32 gpl_compatible:1;
  6465		__u32 :31; /* alignment pad */
  6466		__u64 netns_dev;
  6467		__u64 netns_ino;
  6468		__u32 nr_jited_ksyms;
  6469		__u32 nr_jited_func_lens;
  6470		__aligned_u64 jited_ksyms;
  6471		__aligned_u64 jited_func_lens;
  6472		__u32 btf_id;
  6473		__u32 func_info_rec_size;
  6474		__aligned_u64 func_info;
  6475		__u32 nr_func_info;
  6476		__u32 nr_line_info;
  6477		__aligned_u64 line_info;
  6478		__aligned_u64 jited_line_info;
  6479		__u32 nr_jited_line_info;
  6480		__u32 line_info_rec_size;
  6481		__u32 jited_line_info_rec_size;
  6482		__u32 nr_prog_tags;
  6483		__aligned_u64 prog_tags;
  6484		__u64 run_time_ns;
  6485		__u64 run_cnt;
  6486		__u64 recursion_misses;
  6487		__u32 verified_insns;
  6488		__u32 attach_btf_obj_id;
  6489		__u32 attach_btf_id;
  6490	} __attribute__((aligned(8)));
  6491	
  6492	struct bpf_map_info {
  6493		__u32 type;
  6494		__u32 id;
  6495		__u32 key_size;
  6496		__u32 value_size;
  6497		__u32 max_entries;
  6498		__u32 map_flags;
  6499		char  name[BPF_OBJ_NAME_LEN];
  6500		__u32 ifindex;
  6501		__u32 btf_vmlinux_value_type_id;
  6502		__u64 netns_dev;
  6503		__u64 netns_ino;
  6504		__u32 btf_id;
  6505		__u32 btf_key_type_id;
  6506		__u32 btf_value_type_id;
  6507		__u32 :32;	/* alignment pad */
  6508		__u64 map_extra;
  6509	} __attribute__((aligned(8)));
  6510	
  6511	struct bpf_btf_info {
  6512		__aligned_u64 btf;
  6513		__u32 btf_size;
  6514		__u32 id;
  6515		__aligned_u64 name;
  6516		__u32 name_len;
  6517		__u32 kernel_btf;
  6518	} __attribute__((aligned(8)));
  6519	
  6520	struct bpf_link_info {
  6521		__u32 type;
  6522		__u32 id;
  6523		__u32 prog_id;
  6524		union {
  6525			struct {
  6526				__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
  6527				__u32 tp_name_len;     /* in/out: tp_name buffer len */
  6528			} raw_tracepoint;
  6529			struct {
  6530				__u32 attach_type;
  6531				__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
  6532				__u32 target_btf_id; /* BTF type id inside the object */
  6533			} tracing;
  6534			struct {
  6535				__u64 cgroup_id;
  6536				__u32 attach_type;
  6537			} cgroup;
  6538			struct {
  6539				__aligned_u64 target_name; /* in/out: target_name buffer ptr */
  6540				__u32 target_name_len;	   /* in/out: target_name buffer len */
  6541	
  6542				/* If the iter specific field is 32 bits, it can be put
  6543				 * in the first or second union. Otherwise it should be
  6544				 * put in the second union.
  6545				 */
  6546				union {
  6547					struct {
  6548						__u32 map_id;
  6549					} map;
  6550				};
  6551				union {
  6552					struct {
  6553						__u64 cgroup_id;
  6554						__u32 order;
  6555					} cgroup;
  6556					struct {
  6557						__u32 tid;
  6558						__u32 pid;
  6559					} task;
  6560				};
  6561			} iter;
  6562			struct  {
  6563				__u32 netns_ino;
  6564				__u32 attach_type;
  6565			} netns;
  6566			struct {
  6567				__u32 ifindex;
  6568			} xdp;
  6569			struct {
  6570				__u32 map_id;
  6571			} struct_ops;
  6572			struct {
  6573				__u32 pf;
  6574				__u32 hooknum;
  6575				__s32 priority;
  6576				__u32 flags;
  6577			} netfilter;
  6578			struct {
  6579				__aligned_u64 addrs;
  6580				__u32 count; /* in/out: kprobe_multi function count */
  6581				__u32 flags;
  6582				__u64 missed;
  6583			} kprobe_multi;
  6584			struct {
  6585				__aligned_u64 path;
  6586				__aligned_u64 offsets;
  6587				__aligned_u64 ref_ctr_offsets;
  6588				__aligned_u64 cookies;
  6589				__u32 path_size; /* in/out: real path size on success, including zero byte */
  6590				__u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */
  6591				__u32 flags;
  6592				__u32 pid;
  6593			} uprobe_multi;
  6594			struct {
  6595				__u32 type; /* enum bpf_perf_event_type */
  6596				__u32 :32;
  6597				union {
  6598					struct {
  6599						__aligned_u64 file_name; /* in/out */
  6600						__u32 name_len;
  6601						__u32 offset; /* offset from file_name */
  6602						__u64 cookie;
  6603					} uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
  6604					struct {
  6605						__aligned_u64 func_name; /* in/out */
  6606						__u32 name_len;
  6607						__u32 offset; /* offset from func_name */
  6608						__u64 addr;
  6609						__u64 missed;
  6610						__u64 cookie;
  6611					} kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
  6612					struct {
  6613						__aligned_u64 tp_name;   /* in/out */
  6614						__u32 name_len;
  6615						__u32 :32;
  6616						__u64 cookie;
  6617					} tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
  6618					struct {
  6619						__u64 config;
  6620						__u32 type;
  6621						__u32 :32;
  6622						__u64 cookie;
  6623					} event; /* BPF_PERF_EVENT_EVENT */
  6624				};
  6625			} perf_event;
  6626			struct {
  6627				__u32 ifindex;
  6628				__u32 attach_type;
  6629			} tcx;
  6630			struct {
  6631				__u32 ifindex;
  6632				__u32 attach_type;
  6633			} netkit;
  6634		};
  6635	} __attribute__((aligned(8)));
  6636	
  6637	/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
  6638	 * by user and intended to be used by socket (e.g. to bind to, depends on
  6639	 * attach type).
  6640	 */
  6641	struct bpf_sock_addr {
  6642		__u32 user_family;	/* Allows 4-byte read, but no write. */
  6643		__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.
  6644					 * Stored in network byte order.
  6645					 */
  6646		__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
  6647					 * Stored in network byte order.
  6648					 */
  6649		__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.
  6650					 * Stored in network byte order
  6651					 */
  6652		__u32 family;		/* Allows 4-byte read, but no write */
  6653		__u32 type;		/* Allows 4-byte read, but no write */
  6654		__u32 protocol;		/* Allows 4-byte read, but no write */
  6655		__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.
  6656					 * Stored in network byte order.
  6657					 */
  6658		__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
  6659					 * Stored in network byte order.
  6660					 */
  6661		__bpf_md_ptr(struct bpf_sock *, sk);
  6662	};
  6663	
  6664	/* User bpf_sock_ops struct to access socket values and specify request ops
  6665	 * and their replies.
  6666	 * Some of this fields are in network (bigendian) byte order and may need
  6667	 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
  6668	 * New fields can only be added at the end of this structure
  6669	 */
  6670	struct bpf_sock_ops {
  6671		__u32 op;
  6672		union {
  6673			__u32 args[4];		/* Optionally passed to bpf program */
  6674			__u32 reply;		/* Returned by bpf program	    */
  6675			__u32 replylong[4];	/* Optionally returned by bpf prog  */
  6676		};
  6677		__u32 family;
  6678		__u32 remote_ip4;	/* Stored in network byte order */
  6679		__u32 local_ip4;	/* Stored in network byte order */
  6680		__u32 remote_ip6[4];	/* Stored in network byte order */
  6681		__u32 local_ip6[4];	/* Stored in network byte order */
  6682		__u32 remote_port;	/* Stored in network byte order */
  6683		__u32 local_port;	/* stored in host byte order */
  6684		__u32 is_fullsock;	/* Some TCP fields are only valid if
  6685					 * there is a full socket. If not, the
  6686					 * fields read as zero.
  6687					 */
  6688		__u32 snd_cwnd;
  6689		__u32 srtt_us;		/* Averaged RTT << 3 in usecs */
  6690		__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
  6691		__u32 state;
  6692		__u32 rtt_min;
  6693		__u32 snd_ssthresh;
  6694		__u32 rcv_nxt;
  6695		__u32 snd_nxt;
  6696		__u32 snd_una;
  6697		__u32 mss_cache;
  6698		__u32 ecn_flags;
  6699		__u32 rate_delivered;
  6700		__u32 rate_interval_us;
  6701		__u32 packets_out;
  6702		__u32 retrans_out;
  6703		__u32 total_retrans;
  6704		__u32 segs_in;
  6705		__u32 data_segs_in;
  6706		__u32 segs_out;
  6707		__u32 data_segs_out;
  6708		__u32 lost_out;
  6709		__u32 sacked_out;
  6710		__u32 sk_txhash;
  6711		__u64 bytes_received;
  6712		__u64 bytes_acked;
  6713		__bpf_md_ptr(struct bpf_sock *, sk);
  6714		/* [skb_data, skb_data_end) covers the whole TCP header.
  6715		 *
  6716		 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
  6717		 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
  6718		 *                                header has not been written.
  6719		 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
  6720		 *				  been written so far.
  6721		 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
  6722		 *					the 3WHS.
  6723		 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
  6724		 *					the 3WHS.
  6725		 *
  6726		 * bpf_load_hdr_opt() can also be used to read a particular option.
  6727		 */
  6728		__bpf_md_ptr(void *, skb_data);
  6729		__bpf_md_ptr(void *, skb_data_end);
  6730		__u32 skb_len;		/* The total length of a packet.
  6731					 * It includes the header, options,
  6732					 * and payload.
  6733					 */
  6734		__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides
  6735					 * an easy way to check for tcp_flags
  6736					 * without parsing skb_data.
  6737					 *
  6738					 * In particular, the skb_tcp_flags
  6739					 * will still be available in
  6740					 * BPF_SOCK_OPS_HDR_OPT_LEN even though
  6741					 * the outgoing header has not
  6742					 * been written yet.
  6743					 */
  6744		__u64 skb_hwtstamp;
  6745	};
  6746	
  6747	/* Definitions for bpf_sock_ops_cb_flags */
  6748	enum {
  6749		BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),
  6750		BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),
  6751		BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),
  6752		BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),
  6753		/* Call bpf for all received TCP headers.  The bpf prog will be
  6754		 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6755		 *
  6756		 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6757		 * for the header option related helpers that will be useful
  6758		 * to the bpf programs.
  6759		 *
  6760		 * It could be used at the client/active side (i.e. connect() side)
  6761		 * when the server told it that the server was in syncookie
  6762		 * mode and required the active side to resend the bpf-written
  6763		 * options.  The active side can keep writing the bpf-options until
  6764		 * it received a valid packet from the server side to confirm
  6765		 * the earlier packet (and options) has been received.  The later
  6766		 * example patch is using it like this at the active side when the
  6767		 * server is in syncookie mode.
  6768		 *
  6769		 * The bpf prog will usually turn this off in the common cases.
  6770		 */
  6771		BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),
  6772		/* Call bpf when kernel has received a header option that
  6773		 * the kernel cannot handle.  The bpf prog will be called under
  6774		 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
  6775		 *
  6776		 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
  6777		 * for the header option related helpers that will be useful
  6778		 * to the bpf programs.
  6779		 */
  6780		BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
  6781		/* Call bpf when the kernel is writing header options for the
  6782		 * outgoing packet.  The bpf prog will first be called
  6783		 * to reserve space in a skb under
  6784		 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
  6785		 * the bpf prog will be called to write the header option(s)
  6786		 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6787		 *
  6788		 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
  6789		 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
  6790		 * related helpers that will be useful to the bpf programs.
  6791		 *
  6792		 * The kernel gets its chance to reserve space and write
  6793		 * options first before the BPF program does.
  6794		 */
  6795		BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
  6796	/* Mask of all currently supported cb flags */
  6797		BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
  6798	};
  6799	
  6800	/* List of known BPF sock_ops operators.
  6801	 * New entries can only be added at the end
  6802	 */
  6803	enum {
  6804		BPF_SOCK_OPS_VOID,
  6805		BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
  6806						 * -1 if default value should be used
  6807						 */
  6808		BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
  6809						 * window (in packets) or -1 if default
  6810						 * value should be used
  6811						 */
  6812		BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
  6813						 * active connection is initialized
  6814						 */
  6815		BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
  6816							 * active connection is
  6817							 * established
  6818							 */
  6819		BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
  6820							 * passive connection is
  6821							 * established
  6822							 */
  6823		BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
  6824						 * needs ECN
  6825						 */
  6826		BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is
  6827						 * based on the path and may be
  6828						 * dependent on the congestion control
  6829						 * algorithm. In general it indicates
  6830						 * a congestion threshold. RTTs above
  6831						 * this indicate congestion
  6832						 */
  6833		BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.
  6834						 * Arg1: value of icsk_retransmits
  6835						 * Arg2: value of icsk_rto
  6836						 * Arg3: whether RTO has expired
  6837						 */
  6838		BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.
  6839						 * Arg1: sequence number of 1st byte
  6840						 * Arg2: # segments
  6841						 * Arg3: return value of
  6842						 *       tcp_transmit_skb (0 => success)
  6843						 */
  6844		BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.
  6845						 * Arg1: old_state
  6846						 * Arg2: new_state
  6847						 */
  6848		BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
  6849						 * socket transition to LISTEN state.
  6850						 */
  6851		BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.
  6852						 */
  6853		BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.
  6854						 * It will be called to handle
  6855						 * the packets received at
  6856						 * an already established
  6857						 * connection.
  6858						 *
  6859						 * sock_ops->skb_data:
  6860						 * Referring to the received skb.
  6861						 * It covers the TCP header only.
  6862						 *
  6863						 * bpf_load_hdr_opt() can also
  6864						 * be used to search for a
  6865						 * particular option.
  6866						 */
  6867		BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the
  6868						 * header option later in
  6869						 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6870						 * Arg1: bool want_cookie. (in
  6871						 *       writing SYNACK only)
  6872						 *
  6873						 * sock_ops->skb_data:
  6874						 * Not available because no header has
  6875						 * been	written yet.
  6876						 *
  6877						 * sock_ops->skb_tcp_flags:
  6878						 * The tcp_flags of the
  6879						 * outgoing skb. (e.g. SYN, ACK, FIN).
  6880						 *
  6881						 * bpf_reserve_hdr_opt() should
  6882						 * be used to reserve space.
  6883						 */
  6884		BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options
  6885						 * Arg1: bool want_cookie. (in
  6886						 *       writing SYNACK only)
  6887						 *
  6888						 * sock_ops->skb_data:
  6889						 * Referring to the outgoing skb.
  6890						 * It covers the TCP header
  6891						 * that has already been written
  6892						 * by the kernel and the
  6893						 * earlier bpf-progs.
  6894						 *
  6895						 * sock_ops->skb_tcp_flags:
  6896						 * The tcp_flags of the outgoing
  6897						 * skb. (e.g. SYN, ACK, FIN).
  6898						 *
  6899						 * bpf_store_hdr_opt() should
  6900						 * be used to write the
  6901						 * option.
  6902						 *
  6903						 * bpf_load_hdr_opt() can also
  6904						 * be used to search for a
  6905						 * particular option that
  6906						 * has already been written
  6907						 * by the kernel or the
  6908						 * earlier bpf-progs.
  6909						 */
  6910	};
  6911	
  6912	/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
  6913	 * changes between the TCP and BPF versions. Ideally this should never happen.
  6914	 * If it does, we need to add code to convert them before calling
  6915	 * the BPF sock_ops function.
  6916	 */
  6917	enum {
  6918		BPF_TCP_ESTABLISHED = 1,
  6919		BPF_TCP_SYN_SENT,
  6920		BPF_TCP_SYN_RECV,
  6921		BPF_TCP_FIN_WAIT1,
  6922		BPF_TCP_FIN_WAIT2,
  6923		BPF_TCP_TIME_WAIT,
  6924		BPF_TCP_CLOSE,
  6925		BPF_TCP_CLOSE_WAIT,
  6926		BPF_TCP_LAST_ACK,
  6927		BPF_TCP_LISTEN,
  6928		BPF_TCP_CLOSING,	/* Now a valid state */
  6929		BPF_TCP_NEW_SYN_RECV,
  6930		BPF_TCP_BOUND_INACTIVE,
  6931	
  6932		BPF_TCP_MAX_STATES	/* Leave at the end! */
  6933	};
  6934	
  6935	enum {
  6936		TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */
  6937		TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */
  6938		TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */
  6939		TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */
  6940		/* Copy the SYN pkt to optval
  6941		 *
  6942		 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
  6943		 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
  6944		 * to only getting from the saved_syn.  It can either get the
  6945		 * syn packet from:
  6946		 *
  6947		 * 1. the just-received SYN packet (only available when writing the
  6948		 *    SYNACK).  It will be useful when it is not necessary to
  6949		 *    save the SYN packet for latter use.  It is also the only way
  6950		 *    to get the SYN during syncookie mode because the syn
  6951		 *    packet cannot be saved during syncookie.
  6952		 *
  6953		 * OR
  6954		 *
  6955		 * 2. the earlier saved syn which was done by
  6956		 *    bpf_setsockopt(TCP_SAVE_SYN).
  6957		 *
  6958		 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
  6959		 * SYN packet is obtained.
  6960		 *
  6961		 * If the bpf-prog does not need the IP[46] header,  the
  6962		 * bpf-prog can avoid parsing the IP header by using
  6963		 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
  6964		 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
  6965		 *
  6966		 *      >0: Total number of bytes copied
  6967		 * -ENOSPC: Not enough space in optval. Only optlen number of
  6968		 *          bytes is copied.
  6969		 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
  6970		 *	    is not saved by setsockopt(TCP_SAVE_SYN).
  6971		 */
  6972		TCP_BPF_SYN		= 1005, /* Copy the TCP header */
  6973		TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */
  6974		TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
  6975	};
  6976	
  6977	enum {
  6978		BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
  6979	};
  6980	
  6981	/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
  6982	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
  6983	 */
  6984	enum {
  6985		BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the
  6986							 * total option spaces
  6987							 * required for an established
  6988							 * sk in order to calculate the
  6989							 * MSS.  No skb is actually
  6990							 * sent.
  6991							 */
  6992		BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode
  6993							 * when sending a SYN.
  6994							 */
  6995	};
  6996	
  6997	struct bpf_perf_event_value {
  6998		__u64 counter;
  6999		__u64 enabled;
  7000		__u64 running;
  7001	};
  7002	
  7003	enum {
  7004		BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),
  7005		BPF_DEVCG_ACC_READ	= (1ULL << 1),
  7006		BPF_DEVCG_ACC_WRITE	= (1ULL << 2),
  7007	};
  7008	
  7009	enum {
  7010		BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),
  7011		BPF_DEVCG_DEV_CHAR	= (1ULL << 1),
  7012	};
  7013	
  7014	struct bpf_cgroup_dev_ctx {
  7015		/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
  7016		__u32 access_type;
  7017		__u32 major;
  7018		__u32 minor;
  7019	};
  7020	
  7021	struct bpf_raw_tracepoint_args {
  7022		__u64 args[0];
  7023	};
  7024	
  7025	/* DIRECT:  Skip the FIB rules and go to FIB table associated with device
  7026	 * OUTPUT:  Do lookup from egress perspective; default is ingress
  7027	 */
  7028	enum {
  7029		BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
  7030		BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
  7031		BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
  7032		BPF_FIB_LOOKUP_TBID    = (1U << 3),
  7033		BPF_FIB_LOOKUP_SRC     = (1U << 4),
  7034	};
  7035	
  7036	enum {
  7037		BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
  7038		BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
  7039		BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
  7040		BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
  7041		BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
  7042		BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
  7043		BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
  7044		BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
  7045		BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
  7046		BPF_FIB_LKUP_RET_NO_SRC_ADDR,  /* failed to derive IP src addr */
  7047	};
  7048	
  7049	struct bpf_fib_lookup {
  7050		/* input:  network family for lookup (AF_INET, AF_INET6)
  7051		 * output: network family of egress nexthop
  7052		 */
  7053		__u8	family;
  7054	
  7055		/* set if lookup is to consider L4 data - e.g., FIB rules */
  7056		__u8	l4_protocol;
  7057		__be16	sport;
  7058		__be16	dport;
  7059	
  7060		union {	/* used for MTU check */
  7061			/* input to lookup */
  7062			__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */
  7063	
  7064			/* output: MTU value */
  7065			__u16	mtu_result;
  7066		} __attribute__((packed, aligned(2)));
  7067		/* input: L3 device index for lookup
  7068		 * output: device index from FIB lookup
  7069		 */
  7070		__u32	ifindex;
  7071	
  7072		union {
  7073			/* inputs to lookup */
  7074			__u8	tos;		/* AF_INET  */
  7075			__be32	flowinfo;	/* AF_INET6, flow_label + priority */
  7076	
  7077			/* output: metric of fib result (IPv4/IPv6 only) */
  7078			__u32	rt_metric;
  7079		};
  7080	
  7081		/* input: source address to consider for lookup
  7082		 * output: source address result from lookup
  7083		 */
  7084		union {
  7085			__be32		ipv4_src;
  7086			__u32		ipv6_src[4];  /* in6_addr; network order */
  7087		};
  7088	
  7089		/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
  7090		 * network header. output: bpf_fib_lookup sets to gateway address
  7091		 * if FIB lookup returns gateway route
  7092		 */
  7093		union {
  7094			__be32		ipv4_dst;
  7095			__u32		ipv6_dst[4];  /* in6_addr; network order */
  7096		};
  7097	
  7098		union {
  7099			struct {
  7100				/* output */
  7101				__be16	h_vlan_proto;
  7102				__be16	h_vlan_TCI;
  7103			};
  7104			/* input: when accompanied with the
  7105			 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
  7106			 * specific routing table to use for the fib lookup.
  7107			 */
  7108			__u32	tbid;
  7109		};
  7110	
  7111		__u8	smac[6];     /* ETH_ALEN */
  7112		__u8	dmac[6];     /* ETH_ALEN */
  7113	};
  7114	
  7115	struct bpf_redir_neigh {
  7116		/* network family for lookup (AF_INET, AF_INET6) */
  7117		__u32 nh_family;
  7118		/* network address of nexthop; skips fib lookup to find gateway */
  7119		union {
  7120			__be32		ipv4_nh;
  7121			__u32		ipv6_nh[4];  /* in6_addr; network order */
  7122		};
  7123	};
  7124	
  7125	/* bpf_check_mtu flags*/
  7126	enum  bpf_check_mtu_flags {
  7127		BPF_MTU_CHK_SEGS  = (1U << 0),
  7128	};
  7129	
  7130	enum bpf_check_mtu_ret {
  7131		BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
  7132		BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
  7133		BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
  7134	};
  7135	
  7136	enum bpf_task_fd_type {
  7137		BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */
  7138		BPF_FD_TYPE_TRACEPOINT,		/* tp name */
  7139		BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */
  7140		BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */
  7141		BPF_FD_TYPE_UPROBE,		/* filename + offset */
  7142		BPF_FD_TYPE_URETPROBE,		/* filename + offset */
  7143	};
  7144	
  7145	enum {
  7146		BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),
  7147		BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),
  7148		BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),
  7149	};
  7150	
  7151	struct bpf_flow_keys {
  7152		__u16	nhoff;
  7153		__u16	thoff;
  7154		__u16	addr_proto;			/* ETH_P_* of valid addrs */
  7155		__u8	is_frag;
  7156		__u8	is_first_frag;
  7157		__u8	is_encap;
  7158		__u8	ip_proto;
  7159		__be16	n_proto;
  7160		__be16	sport;
  7161		__be16	dport;
  7162		union {
  7163			struct {
  7164				__be32	ipv4_src;
  7165				__be32	ipv4_dst;
  7166			};
  7167			struct {
  7168				__u32	ipv6_src[4];	/* in6_addr; network order */
  7169				__u32	ipv6_dst[4];	/* in6_addr; network order */
  7170			};
  7171		};
  7172		__u32	flags;
  7173		__be32	flow_label;
  7174	};
  7175	
  7176	struct bpf_func_info {
  7177		__u32	insn_off;
  7178		__u32	type_id;
  7179	};
  7180	
  7181	#define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)
  7182	#define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)
  7183	
  7184	struct bpf_line_info {
  7185		__u32	insn_off;
  7186		__u32	file_name_off;
  7187		__u32	line_off;
  7188		__u32	line_col;
  7189	};
  7190	
  7191	struct bpf_spin_lock {
  7192		__u32	val;
  7193	};
  7194	
  7195	struct bpf_timer {
  7196		__u64 __opaque[2];
  7197	} __attribute__((aligned(8)));
  7198	
  7199	struct bpf_dynptr {
  7200		__u64 __opaque[2];
  7201	} __attribute__((aligned(8)));
  7202	
  7203	struct bpf_list_head {
  7204		__u64 __opaque[2];
  7205	} __attribute__((aligned(8)));
  7206	
  7207	struct bpf_list_node {
  7208		__u64 __opaque[3];
  7209	} __attribute__((aligned(8)));
  7210	
  7211	struct bpf_rb_root {
  7212		__u64 __opaque[2];
  7213	} __attribute__((aligned(8)));
  7214	
  7215	struct bpf_rb_node {
  7216		__u64 __opaque[4];
  7217	} __attribute__((aligned(8)));
  7218	
  7219	struct bpf_refcount {
  7220		__u32 __opaque[1];
  7221	} __attribute__((aligned(4)));
  7222	
  7223	struct bpf_sysctl {
  7224		__u32	write;		/* Sysctl is being read (= 0) or written (= 1).
  7225					 * Allows 1,2,4-byte read, but no write.
  7226					 */
  7227		__u32	file_pos;	/* Sysctl file position to read from, write to.
  7228					 * Allows 1,2,4-byte read an 4-byte write.
  7229					 */
  7230	};
  7231	
  7232	struct bpf_sockopt {
  7233		__bpf_md_ptr(struct bpf_sock *, sk);
  7234		__bpf_md_ptr(void *, optval);
  7235		__bpf_md_ptr(void *, optval_end);
  7236	
  7237		__s32	level;
  7238		__s32	optname;
  7239		__s32	optlen;
  7240		__s32	retval;
  7241	};
  7242	
  7243	struct bpf_pidns_info {
  7244		__u32 pid;
  7245		__u32 tgid;
  7246	};
  7247	
  7248	/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
  7249	struct bpf_sk_lookup {
  7250		union {
  7251			__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
  7252			__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
  7253		};
  7254	
  7255		__u32 family;		/* Protocol family (AF_INET, AF_INET6) */
  7256		__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
  7257		__u32 remote_ip4;	/* Network byte order */
  7258		__u32 remote_ip6[4];	/* Network byte order */
  7259		__be16 remote_port;	/* Network byte order */
  7260		__u16 :16;		/* Zero padding */
  7261		__u32 local_ip4;	/* Network byte order */
  7262		__u32 local_ip6[4];	/* Network byte order */
  7263		__u32 local_port;	/* Host byte order */
  7264		__u32 ingress_ifindex;		/* The arriving interface. Determined by inet_iif. */
  7265	};
  7266	
  7267	/*
  7268	 * struct btf_ptr is used for typed pointer representation; the
  7269	 * type id is used to render the pointer data as the appropriate type
  7270	 * via the bpf_snprintf_btf() helper described above.  A flags field -
  7271	 * potentially to specify additional details about the BTF pointer
  7272	 * (rather than its mode of display) - is included for future use.
  7273	 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
  7274	 */
  7275	struct btf_ptr {
  7276		void *ptr;
  7277		__u32 type_id;
  7278		__u32 flags;		/* BTF ptr flags; unused at present. */
  7279	};
  7280	
  7281	/*
  7282	 * Flags to control bpf_snprintf_btf() behaviour.
  7283	 *     - BTF_F_COMPACT: no formatting around type information
  7284	 *     - BTF_F_NONAME: no struct/union member names/types
  7285	 *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
  7286	 *       equivalent to %px.
  7287	 *     - BTF_F_ZERO: show zero-valued struct/union members; they
  7288	 *       are not displayed by default
  7289	 */
  7290	enum {
  7291		BTF_F_COMPACT	=	(1ULL << 0),
  7292		BTF_F_NONAME	=	(1ULL << 1),
  7293		BTF_F_PTR_RAW	=	(1ULL << 2),
  7294		BTF_F_ZERO	=	(1ULL << 3),
  7295	};
  7296	
  7297	/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
  7298	 * has to be adjusted by relocations. It is emitted by llvm and passed to
  7299	 * libbpf and later to the kernel.
  7300	 */
  7301	enum bpf_core_relo_kind {
  7302		BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */
  7303		BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */
  7304		BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */
  7305		BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */
  7306		BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */
  7307		BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */
  7308		BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */
  7309		BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */
  7310		BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */
  7311		BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */
  7312		BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */
  7313		BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */
  7314		BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */
  7315	};
  7316	
  7317	/*
  7318	 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
  7319	 * and from libbpf to the kernel.
  7320	 *
  7321	 * CO-RE relocation captures the following data:
  7322	 * - insn_off - instruction offset (in bytes) within a BPF program that needs
  7323	 *   its insn->imm field to be relocated with actual field info;
  7324	 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
  7325	 *   type or field;
  7326	 * - access_str_off - offset into corresponding .BTF string section. String
  7327	 *   interpretation depends on specific relocation kind:
  7328	 *     - for field-based relocations, string encodes an accessed field using
  7329	 *       a sequence of field and array indices, separated by colon (:). It's
  7330	 *       conceptually very close to LLVM's getelementptr ([0]) instruction's
  7331	 *       arguments for identifying offset to a field.
  7332	 *     - for type-based relocations, strings is expected to be just "0";
  7333	 *     - for enum value-based relocations, string contains an index of enum
  7334	 *       value within its enum type;
  7335	 * - kind - one of enum bpf_core_relo_kind;
  7336	 *
  7337	 * Example:
  7338	 *   struct sample {
  7339	 *       int a;
  7340	 *       struct {
  7341	 *           int b[10];
  7342	 *       };
  7343	 *   };
  7344	 *
  7345	 *   struct sample *s = ...;
  7346	 *   int *x = &s->a;     // encoded as "0:0" (a is field #0)
  7347	 *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,
  7348	 *                       // b is field #0 inside anon struct, accessing elem #5)
  7349	 *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
  7350	 *
  7351	 * type_id for all relocs in this example will capture BTF type id of
  7352	 * `struct sample`.
  7353	 *
  7354	 * Such relocation is emitted when using __builtin_preserve_access_index()
  7355	 * Clang built-in, passing expression that captures field address, e.g.:
  7356	 *
  7357	 * bpf_probe_read(&dst, sizeof(dst),
  7358	 *		  __builtin_preserve_access_index(&src->a.b.c));
  7359	 *
  7360	 * In this case Clang will emit field relocation recording necessary data to
  7361	 * be able to find offset of embedded `a.b.c` field within `src` struct.
  7362	 *
  7363	 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
  7364	 */
  7365	struct bpf_core_relo {
  7366		__u32 insn_off;
  7367		__u32 type_id;
  7368		__u32 access_str_off;
  7369		enum bpf_core_relo_kind kind;
  7370	};
  7371	
  7372	/*
  7373	 * Flags to control bpf_timer_start() behaviour.
  7374	 *     - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
  7375	 *       relative to current time.
  7376	 *     - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.
  7377	 */
  7378	enum {
  7379		BPF_F_TIMER_ABS = (1ULL << 0),
  7380		BPF_F_TIMER_CPU_PIN = (1ULL << 1),
  7381	};
  7382	
  7383	/* BPF numbers iterator state */
  7384	struct bpf_iter_num {
  7385		/* opaque iterator state; having __u64 here allows to preserve correct
  7386		 * alignment requirements in vmlinux.h, generated from BTF
  7387		 */
  7388		__u64 __opaque[1];
  7389	} __attribute__((aligned(8)));
  7390	
  7391	#endif /* __LINUX_BPF_H__ */