brintos / linux-shallow public Read only

1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */2/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com3 *4 * This program is free software; you can redistribute it and/or5 * modify it under the terms of version 2 of the GNU General Public6 * License as published by the Free Software Foundation.7 */8#ifndef _UAPI__LINUX_BPF_H__9#define _UAPI__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_LE35#define BPF_FROM_BE	BPF_TO_BE36 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_JCOND	0xe0	/* conditional pseudo jumps: may_goto, goto_or_nop */46#define BPF_CALL	0x80	/* function call */47#define BPF_EXIT	0x90	/* function return */48 49/* atomic op type fields (stored in immediate) */50#define BPF_FETCH	0x01	/* not an opcode on its own, used to build others */51#define BPF_XCHG	(0xe0 | BPF_FETCH)	/* atomic exchange */52#define BPF_CMPXCHG	(0xf0 | BPF_FETCH)	/* atomic compare-and-write */53 54enum bpf_cond_pseudo_jmp {55	BPF_MAY_GOTO = 0,56};57 58/* Register numbers */59enum {60	BPF_REG_0 = 0,61	BPF_REG_1,62	BPF_REG_2,63	BPF_REG_3,64	BPF_REG_4,65	BPF_REG_5,66	BPF_REG_6,67	BPF_REG_7,68	BPF_REG_8,69	BPF_REG_9,70	BPF_REG_10,71	__MAX_BPF_REG,72};73 74/* BPF has 10 general purpose 64-bit registers and stack frame. */75#define MAX_BPF_REG	__MAX_BPF_REG76 77struct bpf_insn {78	__u8	code;		/* opcode */79	__u8	dst_reg:4;	/* dest register */80	__u8	src_reg:4;	/* source register */81	__s16	off;		/* signed offset */82	__s32	imm;		/* signed immediate constant */83};84 85/* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for86 * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for87 * the trailing flexible array member) instead.88 */89struct bpf_lpm_trie_key {90	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */91	__u8	data[0];	/* Arbitrary size */92};93 94/* Header for bpf_lpm_trie_key structs */95struct bpf_lpm_trie_key_hdr {96	__u32	prefixlen;97};98 99/* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */100struct bpf_lpm_trie_key_u8 {101	union {102		struct bpf_lpm_trie_key_hdr	hdr;103		__u32				prefixlen;104	};105	__u8	data[];		/* Arbitrary size */106};107 108struct bpf_cgroup_storage_key {109	__u64	cgroup_inode_id;	/* cgroup inode id */110	__u32	attach_type;		/* program attach type (enum bpf_attach_type) */111};112 113enum bpf_cgroup_iter_order {114	BPF_CGROUP_ITER_ORDER_UNSPEC = 0,115	BPF_CGROUP_ITER_SELF_ONLY,		/* process only a single object. */116	BPF_CGROUP_ITER_DESCENDANTS_PRE,	/* walk descendants in pre-order. */117	BPF_CGROUP_ITER_DESCENDANTS_POST,	/* walk descendants in post-order. */118	BPF_CGROUP_ITER_ANCESTORS_UP,		/* walk ancestors upward. */119};120 121union bpf_iter_link_info {122	struct {123		__u32	map_fd;124	} map;125	struct {126		enum bpf_cgroup_iter_order order;127 128		/* At most one of cgroup_fd and cgroup_id can be non-zero. If129		 * both are zero, the walk starts from the default cgroup v2130		 * root. For walking v1 hierarchy, one should always explicitly131		 * specify cgroup_fd.132		 */133		__u32	cgroup_fd;134		__u64	cgroup_id;135	} cgroup;136	/* Parameters of task iterators. */137	struct {138		__u32	tid;139		__u32	pid;140		__u32	pid_fd;141	} task;142};143 144/* BPF syscall commands, see bpf(2) man-page for more details. */145/**146 * DOC: eBPF Syscall Preamble147 *148 * The operation to be performed by the **bpf**\ () system call is determined149 * by the *cmd* argument. Each operation takes an accompanying argument,150 * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see151 * below). The size argument is the size of the union pointed to by *attr*.152 */153/**154 * DOC: eBPF Syscall Commands155 *156 * BPF_MAP_CREATE157 *	Description158 *		Create a map and return a file descriptor that refers to the159 *		map. The close-on-exec file descriptor flag (see **fcntl**\ (2))160 *		is automatically enabled for the new file descriptor.161 *162 *		Applying **close**\ (2) to the file descriptor returned by163 *		**BPF_MAP_CREATE** will delete the map (but see NOTES).164 *165 *	Return166 *		A new file descriptor (a nonnegative integer), or -1 if an167 *		error occurred (in which case, *errno* is set appropriately).168 *169 * BPF_MAP_LOOKUP_ELEM170 *	Description171 *		Look up an element with a given *key* in the map referred to172 *		by the file descriptor *map_fd*.173 *174 *		The *flags* argument may be specified as one of the175 *		following:176 *177 *		**BPF_F_LOCK**178 *			Look up the value of a spin-locked map without179 *			returning the lock. This must be specified if the180 *			elements contain a spinlock.181 *182 *	Return183 *		Returns zero on success. On error, -1 is returned and *errno*184 *		is set appropriately.185 *186 * BPF_MAP_UPDATE_ELEM187 *	Description188 *		Create or update an element (key/value pair) in a specified map.189 *190 *		The *flags* argument should be specified as one of the191 *		following:192 *193 *		**BPF_ANY**194 *			Create a new element or update an existing element.195 *		**BPF_NOEXIST**196 *			Create a new element only if it did not exist.197 *		**BPF_EXIST**198 *			Update an existing element.199 *		**BPF_F_LOCK**200 *			Update a spin_lock-ed map element.201 *202 *	Return203 *		Returns zero on success. On error, -1 is returned and *errno*204 *		is set appropriately.205 *206 *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,207 *		**E2BIG**, **EEXIST**, or **ENOENT**.208 *209 *		**E2BIG**210 *			The number of elements in the map reached the211 *			*max_entries* limit specified at map creation time.212 *		**EEXIST**213 *			If *flags* specifies **BPF_NOEXIST** and the element214 *			with *key* already exists in the map.215 *		**ENOENT**216 *			If *flags* specifies **BPF_EXIST** and the element with217 *			*key* does not exist in the map.218 *219 * BPF_MAP_DELETE_ELEM220 *	Description221 *		Look up and delete an element by key in a specified map.222 *223 *	Return224 *		Returns zero on success. On error, -1 is returned and *errno*225 *		is set appropriately.226 *227 * BPF_MAP_GET_NEXT_KEY228 *	Description229 *		Look up an element by key in a specified map and return the key230 *		of the next element. Can be used to iterate over all elements231 *		in the map.232 *233 *	Return234 *		Returns zero on success. On error, -1 is returned and *errno*235 *		is set appropriately.236 *237 *		The following cases can be used to iterate over all elements of238 *		the map:239 *240 *		* If *key* is not found, the operation returns zero and sets241 *		  the *next_key* pointer to the key of the first element.242 *		* If *key* is found, the operation returns zero and sets the243 *		  *next_key* pointer to the key of the next element.244 *		* If *key* is the last element, returns -1 and *errno* is set245 *		  to **ENOENT**.246 *247 *		May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or248 *		**EINVAL** on error.249 *250 * BPF_PROG_LOAD251 *	Description252 *		Verify and load an eBPF program, returning a new file253 *		descriptor associated with the program.254 *255 *		Applying **close**\ (2) to the file descriptor returned by256 *		**BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).257 *258 *		The close-on-exec file descriptor flag (see **fcntl**\ (2)) is259 *		automatically enabled for the new file descriptor.260 *261 *	Return262 *		A new file descriptor (a nonnegative integer), or -1 if an263 *		error occurred (in which case, *errno* is set appropriately).264 *265 * BPF_OBJ_PIN266 *	Description267 *		Pin an eBPF program or map referred by the specified *bpf_fd*268 *		to the provided *pathname* on the filesystem.269 *270 *		The *pathname* argument must not contain a dot (".").271 *272 *		On success, *pathname* retains a reference to the eBPF object,273 *		preventing deallocation of the object when the original274 *		*bpf_fd* is closed. This allow the eBPF object to live beyond275 *		**close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent276 *		process.277 *278 *		Applying **unlink**\ (2) or similar calls to the *pathname*279 *		unpins the object from the filesystem, removing the reference.280 *		If no other file descriptors or filesystem nodes refer to the281 *		same object, it will be deallocated (see NOTES).282 *283 *		The filesystem type for the parent directory of *pathname* must284 *		be **BPF_FS_MAGIC**.285 *286 *	Return287 *		Returns zero on success. On error, -1 is returned and *errno*288 *		is set appropriately.289 *290 * BPF_OBJ_GET291 *	Description292 *		Open a file descriptor for the eBPF object pinned to the293 *		specified *pathname*.294 *295 *	Return296 *		A new file descriptor (a nonnegative integer), or -1 if an297 *		error occurred (in which case, *errno* is set appropriately).298 *299 * BPF_PROG_ATTACH300 *	Description301 *		Attach an eBPF program to a *target_fd* at the specified302 *		*attach_type* hook.303 *304 *		The *attach_type* specifies the eBPF attachment point to305 *		attach the program to, and must be one of *bpf_attach_type*306 *		(see below).307 *308 *		The *attach_bpf_fd* must be a valid file descriptor for a309 *		loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap310 *		or sock_ops type corresponding to the specified *attach_type*.311 *312 *		The *target_fd* must be a valid file descriptor for a kernel313 *		object which depends on the attach type of *attach_bpf_fd*:314 *315 *		**BPF_PROG_TYPE_CGROUP_DEVICE**,316 *		**BPF_PROG_TYPE_CGROUP_SKB**,317 *		**BPF_PROG_TYPE_CGROUP_SOCK**,318 *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,319 *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,320 *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,321 *		**BPF_PROG_TYPE_SOCK_OPS**322 *323 *			Control Group v2 hierarchy with the eBPF controller324 *			enabled. Requires the kernel to be compiled with325 *			**CONFIG_CGROUP_BPF**.326 *327 *		**BPF_PROG_TYPE_FLOW_DISSECTOR**328 *329 *			Network namespace (eg /proc/self/ns/net).330 *331 *		**BPF_PROG_TYPE_LIRC_MODE2**332 *333 *			LIRC device path (eg /dev/lircN). Requires the kernel334 *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.335 *336 *		**BPF_PROG_TYPE_SK_SKB**,337 *		**BPF_PROG_TYPE_SK_MSG**338 *339 *			eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).340 *341 *	Return342 *		Returns zero on success. On error, -1 is returned and *errno*343 *		is set appropriately.344 *345 * BPF_PROG_DETACH346 *	Description347 *		Detach the eBPF program associated with the *target_fd* at the348 *		hook specified by *attach_type*. The program must have been349 *		previously attached using **BPF_PROG_ATTACH**.350 *351 *	Return352 *		Returns zero on success. On error, -1 is returned and *errno*353 *		is set appropriately.354 *355 * BPF_PROG_TEST_RUN356 *	Description357 *		Run the eBPF program associated with the *prog_fd* a *repeat*358 *		number of times against a provided program context *ctx_in* and359 *		data *data_in*, and return the modified program context360 *		*ctx_out*, *data_out* (for example, packet data), result of the361 *		execution *retval*, and *duration* of the test run.362 *363 *		The sizes of the buffers provided as input and output364 *		parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must365 *		be provided in the corresponding variables *ctx_size_in*,366 *		*ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any367 *		of these parameters are not provided (ie set to NULL), the368 *		corresponding size field must be zero.369 *370 *		Some program types have particular requirements:371 *372 *		**BPF_PROG_TYPE_SK_LOOKUP**373 *			*data_in* and *data_out* must be NULL.374 *375 *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,376 *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**377 *378 *			*ctx_out*, *data_in* and *data_out* must be NULL.379 *			*repeat* must be zero.380 *381 *		BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.382 *383 *	Return384 *		Returns zero on success. On error, -1 is returned and *errno*385 *		is set appropriately.386 *387 *		**ENOSPC**388 *			Either *data_size_out* or *ctx_size_out* is too small.389 *		**ENOTSUPP**390 *			This command is not supported by the program type of391 *			the program referred to by *prog_fd*.392 *393 * BPF_PROG_GET_NEXT_ID394 *	Description395 *		Fetch the next eBPF program currently loaded into the kernel.396 *397 *		Looks for the eBPF program with an id greater than *start_id*398 *		and updates *next_id* on success. If no other eBPF programs399 *		remain with ids higher than *start_id*, returns -1 and sets400 *		*errno* to **ENOENT**.401 *402 *	Return403 *		Returns zero on success. On error, or when no id remains, -1404 *		is returned and *errno* is set appropriately.405 *406 * BPF_MAP_GET_NEXT_ID407 *	Description408 *		Fetch the next eBPF map currently loaded into the kernel.409 *410 *		Looks for the eBPF map with an id greater than *start_id*411 *		and updates *next_id* on success. If no other eBPF maps412 *		remain with ids higher than *start_id*, returns -1 and sets413 *		*errno* to **ENOENT**.414 *415 *	Return416 *		Returns zero on success. On error, or when no id remains, -1417 *		is returned and *errno* is set appropriately.418 *419 * BPF_PROG_GET_FD_BY_ID420 *	Description421 *		Open a file descriptor for the eBPF program corresponding to422 *		*prog_id*.423 *424 *	Return425 *		A new file descriptor (a nonnegative integer), or -1 if an426 *		error occurred (in which case, *errno* is set appropriately).427 *428 * BPF_MAP_GET_FD_BY_ID429 *	Description430 *		Open a file descriptor for the eBPF map corresponding to431 *		*map_id*.432 *433 *	Return434 *		A new file descriptor (a nonnegative integer), or -1 if an435 *		error occurred (in which case, *errno* is set appropriately).436 *437 * BPF_OBJ_GET_INFO_BY_FD438 *	Description439 *		Obtain information about the eBPF object corresponding to440 *		*bpf_fd*.441 *442 *		Populates up to *info_len* bytes of *info*, which will be in443 *		one of the following formats depending on the eBPF object type444 *		of *bpf_fd*:445 *446 *		* **struct bpf_prog_info**447 *		* **struct bpf_map_info**448 *		* **struct bpf_btf_info**449 *		* **struct bpf_link_info**450 *451 *	Return452 *		Returns zero on success. On error, -1 is returned and *errno*453 *		is set appropriately.454 *455 * BPF_PROG_QUERY456 *	Description457 *		Obtain information about eBPF programs associated with the458 *		specified *attach_type* hook.459 *460 *		The *target_fd* must be a valid file descriptor for a kernel461 *		object which depends on the attach type of *attach_bpf_fd*:462 *463 *		**BPF_PROG_TYPE_CGROUP_DEVICE**,464 *		**BPF_PROG_TYPE_CGROUP_SKB**,465 *		**BPF_PROG_TYPE_CGROUP_SOCK**,466 *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,467 *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,468 *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,469 *		**BPF_PROG_TYPE_SOCK_OPS**470 *471 *			Control Group v2 hierarchy with the eBPF controller472 *			enabled. Requires the kernel to be compiled with473 *			**CONFIG_CGROUP_BPF**.474 *475 *		**BPF_PROG_TYPE_FLOW_DISSECTOR**476 *477 *			Network namespace (eg /proc/self/ns/net).478 *479 *		**BPF_PROG_TYPE_LIRC_MODE2**480 *481 *			LIRC device path (eg /dev/lircN). Requires the kernel482 *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.483 *484 *		**BPF_PROG_QUERY** always fetches the number of programs485 *		attached and the *attach_flags* which were used to attach those486 *		programs. Additionally, if *prog_ids* is nonzero and the number487 *		of attached programs is less than *prog_cnt*, populates488 *		*prog_ids* with the eBPF program ids of the programs attached489 *		at *target_fd*.490 *491 *		The following flags may alter the result:492 *493 *		**BPF_F_QUERY_EFFECTIVE**494 *			Only return information regarding programs which are495 *			currently effective at the specified *target_fd*.496 *497 *	Return498 *		Returns zero on success. On error, -1 is returned and *errno*499 *		is set appropriately.500 *501 * BPF_RAW_TRACEPOINT_OPEN502 *	Description503 *		Attach an eBPF program to a tracepoint *name* to access kernel504 *		internal arguments of the tracepoint in their raw form.505 *506 *		The *prog_fd* must be a valid file descriptor associated with507 *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.508 *509 *		No ABI guarantees are made about the content of tracepoint510 *		arguments exposed to the corresponding eBPF program.511 *512 *		Applying **close**\ (2) to the file descriptor returned by513 *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).514 *515 *	Return516 *		A new file descriptor (a nonnegative integer), or -1 if an517 *		error occurred (in which case, *errno* is set appropriately).518 *519 * BPF_BTF_LOAD520 *	Description521 *		Verify and load BPF Type Format (BTF) metadata into the kernel,522 *		returning a new file descriptor associated with the metadata.523 *		BTF is described in more detail at524 *		https://www.kernel.org/doc/html/latest/bpf/btf.html.525 *526 *		The *btf* parameter must point to valid memory providing527 *		*btf_size* bytes of BTF binary metadata.528 *529 *		The returned file descriptor can be passed to other **bpf**\ ()530 *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to531 *		associate the BTF with those objects.532 *533 *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional534 *		parameters to specify a *btf_log_buf*, *btf_log_size* and535 *		*btf_log_level* which allow the kernel to return freeform log536 *		output regarding the BTF verification process.537 *538 *	Return539 *		A new file descriptor (a nonnegative integer), or -1 if an540 *		error occurred (in which case, *errno* is set appropriately).541 *542 * BPF_BTF_GET_FD_BY_ID543 *	Description544 *		Open a file descriptor for the BPF Type Format (BTF)545 *		corresponding to *btf_id*.546 *547 *	Return548 *		A new file descriptor (a nonnegative integer), or -1 if an549 *		error occurred (in which case, *errno* is set appropriately).550 *551 * BPF_TASK_FD_QUERY552 *	Description553 *		Obtain information about eBPF programs associated with the554 *		target process identified by *pid* and *fd*.555 *556 *		If the *pid* and *fd* are associated with a tracepoint, kprobe557 *		or uprobe perf event, then the *prog_id* and *fd_type* will558 *		be populated with the eBPF program id and file descriptor type559 *		of type **bpf_task_fd_type**. If associated with a kprobe or560 *		uprobe, the  *probe_offset* and *probe_addr* will also be561 *		populated. Optionally, if *buf* is provided, then up to562 *		*buf_len* bytes of *buf* will be populated with the name of563 *		the tracepoint, kprobe or uprobe.564 *565 *		The resulting *prog_id* may be introspected in deeper detail566 *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.567 *568 *	Return569 *		Returns zero on success. On error, -1 is returned and *errno*570 *		is set appropriately.571 *572 * BPF_MAP_LOOKUP_AND_DELETE_ELEM573 *	Description574 *		Look up an element with the given *key* in the map referred to575 *		by the file descriptor *fd*, and if found, delete the element.576 *577 *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map578 *		types, the *flags* argument needs to be set to 0, but for other579 *		map types, it may be specified as:580 *581 *		**BPF_F_LOCK**582 *			Look up and delete the value of a spin-locked map583 *			without returning the lock. This must be specified if584 *			the elements contain a spinlock.585 *586 *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types587 *		implement this command as a "pop" operation, deleting the top588 *		element rather than one corresponding to *key*.589 *		The *key* and *key_len* parameters should be zeroed when590 *		issuing this operation for these map types.591 *592 *		This command is only valid for the following map types:593 *		* **BPF_MAP_TYPE_QUEUE**594 *		* **BPF_MAP_TYPE_STACK**595 *		* **BPF_MAP_TYPE_HASH**596 *		* **BPF_MAP_TYPE_PERCPU_HASH**597 *		* **BPF_MAP_TYPE_LRU_HASH**598 *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**599 *600 *	Return601 *		Returns zero on success. On error, -1 is returned and *errno*602 *		is set appropriately.603 *604 * BPF_MAP_FREEZE605 *	Description606 *		Freeze the permissions of the specified map.607 *608 *		Write permissions may be frozen by passing zero *flags*.609 *		Upon success, no future syscall invocations may alter the610 *		map state of *map_fd*. Write operations from eBPF programs611 *		are still possible for a frozen map.612 *613 *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.614 *615 *	Return616 *		Returns zero on success. On error, -1 is returned and *errno*617 *		is set appropriately.618 *619 * BPF_BTF_GET_NEXT_ID620 *	Description621 *		Fetch the next BPF Type Format (BTF) object currently loaded622 *		into the kernel.623 *624 *		Looks for the BTF object with an id greater than *start_id*625 *		and updates *next_id* on success. If no other BTF objects626 *		remain with ids higher than *start_id*, returns -1 and sets627 *		*errno* to **ENOENT**.628 *629 *	Return630 *		Returns zero on success. On error, or when no id remains, -1631 *		is returned and *errno* is set appropriately.632 *633 * BPF_MAP_LOOKUP_BATCH634 *	Description635 *		Iterate and fetch multiple elements in a map.636 *637 *		Two opaque values are used to manage batch operations,638 *		*in_batch* and *out_batch*. Initially, *in_batch* must be set639 *		to NULL to begin the batched operation. After each subsequent640 *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant641 *		*out_batch* as the *in_batch* for the next operation to642 *		continue iteration from the current point. Both *in_batch* and643 *		*out_batch* must point to memory large enough to hold a key,644 *		except for maps of type **BPF_MAP_TYPE_{HASH, PERCPU_HASH,645 *		LRU_HASH, LRU_PERCPU_HASH}**, for which batch parameters646 *		must be at least 4 bytes wide regardless of key size.647 *648 *		The *keys* and *values* are output parameters which must point649 *		to memory large enough to hold *count* items based on the key650 *		and value size of the map *map_fd*. The *keys* buffer must be651 *		of *key_size* * *count*. The *values* buffer must be of652 *		*value_size* * *count*.653 *654 *		The *elem_flags* argument may be specified as one of the655 *		following:656 *657 *		**BPF_F_LOCK**658 *			Look up the value of a spin-locked map without659 *			returning the lock. This must be specified if the660 *			elements contain a spinlock.661 *662 *		On success, *count* elements from the map are copied into the663 *		user buffer, with the keys copied into *keys* and the values664 *		copied into the corresponding indices in *values*.665 *666 *		If an error is returned and *errno* is not **EFAULT**, *count*667 *		is set to the number of successfully processed elements.668 *669 *	Return670 *		Returns zero on success. On error, -1 is returned and *errno*671 *		is set appropriately.672 *673 *		May set *errno* to **ENOSPC** to indicate that *keys* or674 *		*values* is too small to dump an entire bucket during675 *		iteration of a hash-based map type.676 *677 * BPF_MAP_LOOKUP_AND_DELETE_BATCH678 *	Description679 *		Iterate and delete all elements in a map.680 *681 *		This operation has the same behavior as682 *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:683 *684 *		* Every element that is successfully returned is also deleted685 *		  from the map. This is at least *count* elements. Note that686 *		  *count* is both an input and an output parameter.687 *		* Upon returning with *errno* set to **EFAULT**, up to688 *		  *count* elements may be deleted without returning the keys689 *		  and values of the deleted elements.690 *691 *	Return692 *		Returns zero on success. On error, -1 is returned and *errno*693 *		is set appropriately.694 *695 * BPF_MAP_UPDATE_BATCH696 *	Description697 *		Update multiple elements in a map by *key*.698 *699 *		The *keys* and *values* are input parameters which must point700 *		to memory large enough to hold *count* items based on the key701 *		and value size of the map *map_fd*. The *keys* buffer must be702 *		of *key_size* * *count*. The *values* buffer must be of703 *		*value_size* * *count*.704 *705 *		Each element specified in *keys* is sequentially updated to the706 *		value in the corresponding index in *values*. The *in_batch*707 *		and *out_batch* parameters are ignored and should be zeroed.708 *709 *		The *elem_flags* argument should be specified as one of the710 *		following:711 *712 *		**BPF_ANY**713 *			Create new elements or update a existing elements.714 *		**BPF_NOEXIST**715 *			Create new elements only if they do not exist.716 *		**BPF_EXIST**717 *			Update existing elements.718 *		**BPF_F_LOCK**719 *			Update spin_lock-ed map elements. This must be720 *			specified if the map value contains a spinlock.721 *722 *		On success, *count* elements from the map are updated.723 *724 *		If an error is returned and *errno* is not **EFAULT**, *count*725 *		is set to the number of successfully processed elements.726 *727 *	Return728 *		Returns zero on success. On error, -1 is returned and *errno*729 *		is set appropriately.730 *731 *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or732 *		**E2BIG**. **E2BIG** indicates that the number of elements in733 *		the map reached the *max_entries* limit specified at map734 *		creation time.735 *736 *		May set *errno* to one of the following error codes under737 *		specific circumstances:738 *739 *		**EEXIST**740 *			If *flags* specifies **BPF_NOEXIST** and the element741 *			with *key* already exists in the map.742 *		**ENOENT**743 *			If *flags* specifies **BPF_EXIST** and the element with744 *			*key* does not exist in the map.745 *746 * BPF_MAP_DELETE_BATCH747 *	Description748 *		Delete multiple elements in a map by *key*.749 *750 *		The *keys* parameter is an input parameter which must point751 *		to memory large enough to hold *count* items based on the key752 *		size of the map *map_fd*, that is, *key_size* * *count*.753 *754 *		Each element specified in *keys* is sequentially deleted. The755 *		*in_batch*, *out_batch*, and *values* parameters are ignored756 *		and should be zeroed.757 *758 *		The *elem_flags* argument may be specified as one of the759 *		following:760 *761 *		**BPF_F_LOCK**762 *			Look up the value of a spin-locked map without763 *			returning the lock. This must be specified if the764 *			elements contain a spinlock.765 *766 *		On success, *count* elements from the map are updated.767 *768 *		If an error is returned and *errno* is not **EFAULT**, *count*769 *		is set to the number of successfully processed elements. If770 *		*errno* is **EFAULT**, up to *count* elements may be been771 *		deleted.772 *773 *	Return774 *		Returns zero on success. On error, -1 is returned and *errno*775 *		is set appropriately.776 *777 * BPF_LINK_CREATE778 *	Description779 *		Attach an eBPF program to a *target_fd* at the specified780 *		*attach_type* hook and return a file descriptor handle for781 *		managing the link.782 *783 *	Return784 *		A new file descriptor (a nonnegative integer), or -1 if an785 *		error occurred (in which case, *errno* is set appropriately).786 *787 * BPF_LINK_UPDATE788 *	Description789 *		Update the eBPF program in the specified *link_fd* to790 *		*new_prog_fd*.791 *792 *	Return793 *		Returns zero on success. On error, -1 is returned and *errno*794 *		is set appropriately.795 *796 * BPF_LINK_GET_FD_BY_ID797 *	Description798 *		Open a file descriptor for the eBPF Link corresponding to799 *		*link_id*.800 *801 *	Return802 *		A new file descriptor (a nonnegative integer), or -1 if an803 *		error occurred (in which case, *errno* is set appropriately).804 *805 * BPF_LINK_GET_NEXT_ID806 *	Description807 *		Fetch the next eBPF link currently loaded into the kernel.808 *809 *		Looks for the eBPF link with an id greater than *start_id*810 *		and updates *next_id* on success. If no other eBPF links811 *		remain with ids higher than *start_id*, returns -1 and sets812 *		*errno* to **ENOENT**.813 *814 *	Return815 *		Returns zero on success. On error, or when no id remains, -1816 *		is returned and *errno* is set appropriately.817 *818 * BPF_ENABLE_STATS819 *	Description820 *		Enable eBPF runtime statistics gathering.821 *822 *		Runtime statistics gathering for the eBPF runtime is disabled823 *		by default to minimize the corresponding performance overhead.824 *		This command enables statistics globally.825 *826 *		Multiple programs may independently enable statistics.827 *		After gathering the desired statistics, eBPF runtime statistics828 *		may be disabled again by calling **close**\ (2) for the file829 *		descriptor returned by this function. Statistics will only be830 *		disabled system-wide when all outstanding file descriptors831 *		returned by prior calls for this subcommand are closed.832 *833 *	Return834 *		A new file descriptor (a nonnegative integer), or -1 if an835 *		error occurred (in which case, *errno* is set appropriately).836 *837 * BPF_ITER_CREATE838 *	Description839 *		Create an iterator on top of the specified *link_fd* (as840 *		previously created using **BPF_LINK_CREATE**) and return a841 *		file descriptor that can be used to trigger the iteration.842 *843 *		If the resulting file descriptor is pinned to the filesystem844 *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls845 *		for that path will trigger the iterator to read kernel state846 *		using the eBPF program attached to *link_fd*.847 *848 *	Return849 *		A new file descriptor (a nonnegative integer), or -1 if an850 *		error occurred (in which case, *errno* is set appropriately).851 *852 * BPF_LINK_DETACH853 *	Description854 *		Forcefully detach the specified *link_fd* from its855 *		corresponding attachment point.856 *857 *	Return858 *		Returns zero on success. On error, -1 is returned and *errno*859 *		is set appropriately.860 *861 * BPF_PROG_BIND_MAP862 *	Description863 *		Bind a map to the lifetime of an eBPF program.864 *865 *		The map identified by *map_fd* is bound to the program866 *		identified by *prog_fd* and only released when *prog_fd* is867 *		released. This may be used in cases where metadata should be868 *		associated with a program which otherwise does not contain any869 *		references to the map (for example, embedded in the eBPF870 *		program instructions).871 *872 *	Return873 *		Returns zero on success. On error, -1 is returned and *errno*874 *		is set appropriately.875 *876 * BPF_TOKEN_CREATE877 *	Description878 *		Create BPF token with embedded information about what879 *		BPF-related functionality it allows:880 *		- a set of allowed bpf() syscall commands;881 *		- a set of allowed BPF map types to be created with882 *		BPF_MAP_CREATE command, if BPF_MAP_CREATE itself is allowed;883 *		- a set of allowed BPF program types and BPF program attach884 *		types to be loaded with BPF_PROG_LOAD command, if885 *		BPF_PROG_LOAD itself is allowed.886 *887 *		BPF token is created (derived) from an instance of BPF FS,888 *		assuming it has necessary delegation mount options specified.889 *		This BPF token can be passed as an extra parameter to various890 *		bpf() syscall commands to grant BPF subsystem functionality to891 *		unprivileged processes.892 *893 *		When created, BPF token is "associated" with the owning894 *		user namespace of BPF FS instance (super block) that it was895 *		derived from, and subsequent BPF operations performed with896 *		BPF token would be performing capabilities checks (i.e.,897 *		CAP_BPF, CAP_PERFMON, CAP_NET_ADMIN, CAP_SYS_ADMIN) within898 *		that user namespace. Without BPF token, such capabilities899 *		have to be granted in init user namespace, making bpf()900 *		syscall incompatible with user namespace, for the most part.901 *902 *	Return903 *		A new file descriptor (a nonnegative integer), or -1 if an904 *		error occurred (in which case, *errno* is set appropriately).905 *906 * NOTES907 *	eBPF objects (maps and programs) can be shared between processes.908 *909 *	* After **fork**\ (2), the child inherits file descriptors910 *	  referring to the same eBPF objects.911 *	* File descriptors referring to eBPF objects can be transferred over912 *	  **unix**\ (7) domain sockets.913 *	* File descriptors referring to eBPF objects can be duplicated in the914 *	  usual way, using **dup**\ (2) and similar calls.915 *	* File descriptors referring to eBPF objects can be pinned to the916 *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).917 *918 *	An eBPF object is deallocated only after all file descriptors referring919 *	to the object have been closed and no references remain pinned to the920 *	filesystem or attached (for example, bound to a program or device).921 */922enum bpf_cmd {923	BPF_MAP_CREATE,924	BPF_MAP_LOOKUP_ELEM,925	BPF_MAP_UPDATE_ELEM,926	BPF_MAP_DELETE_ELEM,927	BPF_MAP_GET_NEXT_KEY,928	BPF_PROG_LOAD,929	BPF_OBJ_PIN,930	BPF_OBJ_GET,931	BPF_PROG_ATTACH,932	BPF_PROG_DETACH,933	BPF_PROG_TEST_RUN,934	BPF_PROG_RUN = BPF_PROG_TEST_RUN,935	BPF_PROG_GET_NEXT_ID,936	BPF_MAP_GET_NEXT_ID,937	BPF_PROG_GET_FD_BY_ID,938	BPF_MAP_GET_FD_BY_ID,939	BPF_OBJ_GET_INFO_BY_FD,940	BPF_PROG_QUERY,941	BPF_RAW_TRACEPOINT_OPEN,942	BPF_BTF_LOAD,943	BPF_BTF_GET_FD_BY_ID,944	BPF_TASK_FD_QUERY,945	BPF_MAP_LOOKUP_AND_DELETE_ELEM,946	BPF_MAP_FREEZE,947	BPF_BTF_GET_NEXT_ID,948	BPF_MAP_LOOKUP_BATCH,949	BPF_MAP_LOOKUP_AND_DELETE_BATCH,950	BPF_MAP_UPDATE_BATCH,951	BPF_MAP_DELETE_BATCH,952	BPF_LINK_CREATE,953	BPF_LINK_UPDATE,954	BPF_LINK_GET_FD_BY_ID,955	BPF_LINK_GET_NEXT_ID,956	BPF_ENABLE_STATS,957	BPF_ITER_CREATE,958	BPF_LINK_DETACH,959	BPF_PROG_BIND_MAP,960	BPF_TOKEN_CREATE,961	__MAX_BPF_CMD,962};963 964enum bpf_map_type {965	BPF_MAP_TYPE_UNSPEC,966	BPF_MAP_TYPE_HASH,967	BPF_MAP_TYPE_ARRAY,968	BPF_MAP_TYPE_PROG_ARRAY,969	BPF_MAP_TYPE_PERF_EVENT_ARRAY,970	BPF_MAP_TYPE_PERCPU_HASH,971	BPF_MAP_TYPE_PERCPU_ARRAY,972	BPF_MAP_TYPE_STACK_TRACE,973	BPF_MAP_TYPE_CGROUP_ARRAY,974	BPF_MAP_TYPE_LRU_HASH,975	BPF_MAP_TYPE_LRU_PERCPU_HASH,976	BPF_MAP_TYPE_LPM_TRIE,977	BPF_MAP_TYPE_ARRAY_OF_MAPS,978	BPF_MAP_TYPE_HASH_OF_MAPS,979	BPF_MAP_TYPE_DEVMAP,980	BPF_MAP_TYPE_SOCKMAP,981	BPF_MAP_TYPE_CPUMAP,982	BPF_MAP_TYPE_XSKMAP,983	BPF_MAP_TYPE_SOCKHASH,984	BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,985	/* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching986	 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to987	 * both cgroup-attached and other progs and supports all functionality988	 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark989	 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.990	 */991	BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,992	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,993	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,994	/* BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE is available to bpf programs995	 * attaching to a cgroup. The new mechanism (BPF_MAP_TYPE_CGRP_STORAGE +996	 * local percpu kptr) supports all BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE997	 * functionality and more. So mark * BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE998	 * deprecated.999	 */1000	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE = BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE_DEPRECATED,1001	BPF_MAP_TYPE_QUEUE,1002	BPF_MAP_TYPE_STACK,1003	BPF_MAP_TYPE_SK_STORAGE,1004	BPF_MAP_TYPE_DEVMAP_HASH,1005	BPF_MAP_TYPE_STRUCT_OPS,1006	BPF_MAP_TYPE_RINGBUF,1007	BPF_MAP_TYPE_INODE_STORAGE,1008	BPF_MAP_TYPE_TASK_STORAGE,1009	BPF_MAP_TYPE_BLOOM_FILTER,1010	BPF_MAP_TYPE_USER_RINGBUF,1011	BPF_MAP_TYPE_CGRP_STORAGE,1012	BPF_MAP_TYPE_ARENA,1013	__MAX_BPF_MAP_TYPE1014};1015 1016/* Note that tracing related programs such as1017 * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}1018 * are not subject to a stable API since kernel internal data1019 * structures can change from release to release and may1020 * therefore break existing tracing BPF programs. Tracing BPF1021 * programs correspond to /a/ specific kernel which is to be1022 * analyzed, and not /a/ specific kernel /and/ all future ones.1023 */1024enum bpf_prog_type {1025	BPF_PROG_TYPE_UNSPEC,1026	BPF_PROG_TYPE_SOCKET_FILTER,1027	BPF_PROG_TYPE_KPROBE,1028	BPF_PROG_TYPE_SCHED_CLS,1029	BPF_PROG_TYPE_SCHED_ACT,1030	BPF_PROG_TYPE_TRACEPOINT,1031	BPF_PROG_TYPE_XDP,1032	BPF_PROG_TYPE_PERF_EVENT,1033	BPF_PROG_TYPE_CGROUP_SKB,1034	BPF_PROG_TYPE_CGROUP_SOCK,1035	BPF_PROG_TYPE_LWT_IN,1036	BPF_PROG_TYPE_LWT_OUT,1037	BPF_PROG_TYPE_LWT_XMIT,1038	BPF_PROG_TYPE_SOCK_OPS,1039	BPF_PROG_TYPE_SK_SKB,1040	BPF_PROG_TYPE_CGROUP_DEVICE,1041	BPF_PROG_TYPE_SK_MSG,1042	BPF_PROG_TYPE_RAW_TRACEPOINT,1043	BPF_PROG_TYPE_CGROUP_SOCK_ADDR,1044	BPF_PROG_TYPE_LWT_SEG6LOCAL,1045	BPF_PROG_TYPE_LIRC_MODE2,1046	BPF_PROG_TYPE_SK_REUSEPORT,1047	BPF_PROG_TYPE_FLOW_DISSECTOR,1048	BPF_PROG_TYPE_CGROUP_SYSCTL,1049	BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,1050	BPF_PROG_TYPE_CGROUP_SOCKOPT,1051	BPF_PROG_TYPE_TRACING,1052	BPF_PROG_TYPE_STRUCT_OPS,1053	BPF_PROG_TYPE_EXT,1054	BPF_PROG_TYPE_LSM,1055	BPF_PROG_TYPE_SK_LOOKUP,1056	BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */1057	BPF_PROG_TYPE_NETFILTER,1058	__MAX_BPF_PROG_TYPE1059};1060 1061enum bpf_attach_type {1062	BPF_CGROUP_INET_INGRESS,1063	BPF_CGROUP_INET_EGRESS,1064	BPF_CGROUP_INET_SOCK_CREATE,1065	BPF_CGROUP_SOCK_OPS,1066	BPF_SK_SKB_STREAM_PARSER,1067	BPF_SK_SKB_STREAM_VERDICT,1068	BPF_CGROUP_DEVICE,1069	BPF_SK_MSG_VERDICT,1070	BPF_CGROUP_INET4_BIND,1071	BPF_CGROUP_INET6_BIND,1072	BPF_CGROUP_INET4_CONNECT,1073	BPF_CGROUP_INET6_CONNECT,1074	BPF_CGROUP_INET4_POST_BIND,1075	BPF_CGROUP_INET6_POST_BIND,1076	BPF_CGROUP_UDP4_SENDMSG,1077	BPF_CGROUP_UDP6_SENDMSG,1078	BPF_LIRC_MODE2,1079	BPF_FLOW_DISSECTOR,1080	BPF_CGROUP_SYSCTL,1081	BPF_CGROUP_UDP4_RECVMSG,1082	BPF_CGROUP_UDP6_RECVMSG,1083	BPF_CGROUP_GETSOCKOPT,1084	BPF_CGROUP_SETSOCKOPT,1085	BPF_TRACE_RAW_TP,1086	BPF_TRACE_FENTRY,1087	BPF_TRACE_FEXIT,1088	BPF_MODIFY_RETURN,1089	BPF_LSM_MAC,1090	BPF_TRACE_ITER,1091	BPF_CGROUP_INET4_GETPEERNAME,1092	BPF_CGROUP_INET6_GETPEERNAME,1093	BPF_CGROUP_INET4_GETSOCKNAME,1094	BPF_CGROUP_INET6_GETSOCKNAME,1095	BPF_XDP_DEVMAP,1096	BPF_CGROUP_INET_SOCK_RELEASE,1097	BPF_XDP_CPUMAP,1098	BPF_SK_LOOKUP,1099	BPF_XDP,1100	BPF_SK_SKB_VERDICT,1101	BPF_SK_REUSEPORT_SELECT,1102	BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,1103	BPF_PERF_EVENT,1104	BPF_TRACE_KPROBE_MULTI,1105	BPF_LSM_CGROUP,1106	BPF_STRUCT_OPS,1107	BPF_NETFILTER,1108	BPF_TCX_INGRESS,1109	BPF_TCX_EGRESS,1110	BPF_TRACE_UPROBE_MULTI,1111	BPF_CGROUP_UNIX_CONNECT,1112	BPF_CGROUP_UNIX_SENDMSG,1113	BPF_CGROUP_UNIX_RECVMSG,1114	BPF_CGROUP_UNIX_GETPEERNAME,1115	BPF_CGROUP_UNIX_GETSOCKNAME,1116	BPF_NETKIT_PRIMARY,1117	BPF_NETKIT_PEER,1118	BPF_TRACE_KPROBE_SESSION,1119	__MAX_BPF_ATTACH_TYPE1120};1121 1122#define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE1123 1124/* Add BPF_LINK_TYPE(type, name) in bpf_types.h to keep bpf_link_type_strs[]1125 * in sync with the definitions below.1126 */1127enum bpf_link_type {1128	BPF_LINK_TYPE_UNSPEC = 0,1129	BPF_LINK_TYPE_RAW_TRACEPOINT = 1,1130	BPF_LINK_TYPE_TRACING = 2,1131	BPF_LINK_TYPE_CGROUP = 3,1132	BPF_LINK_TYPE_ITER = 4,1133	BPF_LINK_TYPE_NETNS = 5,1134	BPF_LINK_TYPE_XDP = 6,1135	BPF_LINK_TYPE_PERF_EVENT = 7,1136	BPF_LINK_TYPE_KPROBE_MULTI = 8,1137	BPF_LINK_TYPE_STRUCT_OPS = 9,1138	BPF_LINK_TYPE_NETFILTER = 10,1139	BPF_LINK_TYPE_TCX = 11,1140	BPF_LINK_TYPE_UPROBE_MULTI = 12,1141	BPF_LINK_TYPE_NETKIT = 13,1142	BPF_LINK_TYPE_SOCKMAP = 14,1143	__MAX_BPF_LINK_TYPE,1144};1145 1146#define MAX_BPF_LINK_TYPE __MAX_BPF_LINK_TYPE1147 1148enum bpf_perf_event_type {1149	BPF_PERF_EVENT_UNSPEC = 0,1150	BPF_PERF_EVENT_UPROBE = 1,1151	BPF_PERF_EVENT_URETPROBE = 2,1152	BPF_PERF_EVENT_KPROBE = 3,1153	BPF_PERF_EVENT_KRETPROBE = 4,1154	BPF_PERF_EVENT_TRACEPOINT = 5,1155	BPF_PERF_EVENT_EVENT = 6,1156};1157 1158/* cgroup-bpf attach flags used in BPF_PROG_ATTACH command1159 *1160 * NONE(default): No further bpf programs allowed in the subtree.1161 *1162 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,1163 * the program in this cgroup yields to sub-cgroup program.1164 *1165 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,1166 * that cgroup program gets run in addition to the program in this cgroup.1167 *1168 * Only one program is allowed to be attached to a cgroup with1169 * NONE or BPF_F_ALLOW_OVERRIDE flag.1170 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will1171 * release old program and attach the new one. Attach flags has to match.1172 *1173 * Multiple programs are allowed to be attached to a cgroup with1174 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order1175 * (those that were attached first, run first)1176 * The programs of sub-cgroup are executed first, then programs of1177 * this cgroup and then programs of parent cgroup.1178 * When children program makes decision (like picking TCP CA or sock bind)1179 * parent program has a chance to override it.1180 *1181 * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of1182 * programs for a cgroup. Though it's possible to replace an old program at1183 * any position by also specifying BPF_F_REPLACE flag and position itself in1184 * replace_bpf_fd attribute. Old program at this position will be released.1185 *1186 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.1187 * A cgroup with NONE doesn't allow any programs in sub-cgroups.1188 * Ex1:1189 * cgrp1 (MULTI progs A, B) ->1190 *    cgrp2 (OVERRIDE prog C) ->1191 *      cgrp3 (MULTI prog D) ->1192 *        cgrp4 (OVERRIDE prog E) ->1193 *          cgrp5 (NONE prog F)1194 * the event in cgrp5 triggers execution of F,D,A,B in that order.1195 * if prog F is detached, the execution is E,D,A,B1196 * if prog F and D are detached, the execution is E,A,B1197 * if prog F, E and D are detached, the execution is C,A,B1198 *1199 * All eligible programs are executed regardless of return code from1200 * earlier programs.1201 */1202#define BPF_F_ALLOW_OVERRIDE	(1U << 0)1203#define BPF_F_ALLOW_MULTI	(1U << 1)1204/* Generic attachment flags. */1205#define BPF_F_REPLACE		(1U << 2)1206#define BPF_F_BEFORE		(1U << 3)1207#define BPF_F_AFTER		(1U << 4)1208#define BPF_F_ID		(1U << 5)1209#define BPF_F_LINK		BPF_F_LINK /* 1 << 13 */1210 1211/* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the1212 * verifier will perform strict alignment checking as if the kernel1213 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,1214 * and NET_IP_ALIGN defined to 2.1215 */1216#define BPF_F_STRICT_ALIGNMENT	(1U << 0)1217 1218/* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the1219 * verifier will allow any alignment whatsoever.  On platforms1220 * with strict alignment requirements for loads ands stores (such1221 * as sparc and mips) the verifier validates that all loads and1222 * stores provably follow this requirement.  This flag turns that1223 * checking and enforcement off.1224 *1225 * It is mostly used for testing when we want to validate the1226 * context and memory access aspects of the verifier, but because1227 * of an unaligned access the alignment check would trigger before1228 * the one we are interested in.1229 */1230#define BPF_F_ANY_ALIGNMENT	(1U << 1)1231 1232/* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.1233 * Verifier does sub-register def/use analysis and identifies instructions whose1234 * def only matters for low 32-bit, high 32-bit is never referenced later1235 * through implicit zero extension. Therefore verifier notifies JIT back-ends1236 * that it is safe to ignore clearing high 32-bit for these instructions. This1237 * saves some back-ends a lot of code-gen. However such optimization is not1238 * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends1239 * hence hasn't used verifier's analysis result. But, we really want to have a1240 * way to be able to verify the correctness of the described optimization on1241 * x86_64 on which testsuites are frequently exercised.1242 *1243 * So, this flag is introduced. Once it is set, verifier will randomize high1244 * 32-bit for those instructions who has been identified as safe to ignore them.1245 * Then, if verifier is not doing correct analysis, such randomization will1246 * regress tests to expose bugs.1247 */1248#define BPF_F_TEST_RND_HI32	(1U << 2)1249 1250/* The verifier internal test flag. Behavior is undefined */1251#define BPF_F_TEST_STATE_FREQ	(1U << 3)1252 1253/* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will1254 * restrict map and helper usage for such programs. Sleepable BPF programs can1255 * only be attached to hooks where kernel execution context allows sleeping.1256 * Such programs are allowed to use helpers that may sleep like1257 * bpf_copy_from_user().1258 */1259#define BPF_F_SLEEPABLE		(1U << 4)1260 1261/* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program1262 * fully support xdp frags.1263 */1264#define BPF_F_XDP_HAS_FRAGS	(1U << 5)1265 1266/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded1267 * program becomes device-bound but can access XDP metadata.1268 */1269#define BPF_F_XDP_DEV_BOUND_ONLY	(1U << 6)1270 1271/* The verifier internal test flag. Behavior is undefined */1272#define BPF_F_TEST_REG_INVARIANTS	(1U << 7)1273 1274/* link_create.kprobe_multi.flags used in LINK_CREATE command for1275 * BPF_TRACE_KPROBE_MULTI attach type to create return probe.1276 */1277enum {1278	BPF_F_KPROBE_MULTI_RETURN = (1U << 0)1279};1280 1281/* link_create.uprobe_multi.flags used in LINK_CREATE command for1282 * BPF_TRACE_UPROBE_MULTI attach type to create return probe.1283 */1284enum {1285	BPF_F_UPROBE_MULTI_RETURN = (1U << 0)1286};1287 1288/* link_create.netfilter.flags used in LINK_CREATE command for1289 * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.1290 */1291#define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)1292 1293/* When BPF ldimm64's insn[0].src_reg != 0 then this can have1294 * the following extensions:1295 *1296 * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]1297 * insn[0].imm:      map fd or fd_idx1298 * insn[1].imm:      01299 * insn[0].off:      01300 * insn[1].off:      01301 * ldimm64 rewrite:  address of map1302 * verifier type:    CONST_PTR_TO_MAP1303 */1304#define BPF_PSEUDO_MAP_FD	11305#define BPF_PSEUDO_MAP_IDX	51306 1307/* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE1308 * insn[0].imm:      map fd or fd_idx1309 * insn[1].imm:      offset into value1310 * insn[0].off:      01311 * insn[1].off:      01312 * ldimm64 rewrite:  address of map[0]+offset1313 * verifier type:    PTR_TO_MAP_VALUE1314 */1315#define BPF_PSEUDO_MAP_VALUE		21316#define BPF_PSEUDO_MAP_IDX_VALUE	61317 1318/* insn[0].src_reg:  BPF_PSEUDO_BTF_ID1319 * insn[0].imm:      kernel btd id of VAR1320 * insn[1].imm:      01321 * insn[0].off:      01322 * insn[1].off:      01323 * ldimm64 rewrite:  address of the kernel variable1324 * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var1325 *                   is struct/union.1326 */1327#define BPF_PSEUDO_BTF_ID	31328/* insn[0].src_reg:  BPF_PSEUDO_FUNC1329 * insn[0].imm:      insn offset to the func1330 * insn[1].imm:      01331 * insn[0].off:      01332 * insn[1].off:      01333 * ldimm64 rewrite:  address of the function1334 * verifier type:    PTR_TO_FUNC.1335 */1336#define BPF_PSEUDO_FUNC		41337 1338/* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative1339 * offset to another bpf function1340 */1341#define BPF_PSEUDO_CALL		11342/* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,1343 * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel1344 */1345#define BPF_PSEUDO_KFUNC_CALL	21346 1347enum bpf_addr_space_cast {1348	BPF_ADDR_SPACE_CAST = 1,1349};1350 1351/* flags for BPF_MAP_UPDATE_ELEM command */1352enum {1353	BPF_ANY		= 0, /* create new element or update existing */1354	BPF_NOEXIST	= 1, /* create new element if it didn't exist */1355	BPF_EXIST	= 2, /* update existing element */1356	BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */1357};1358 1359/* flags for BPF_MAP_CREATE command */1360enum {1361	BPF_F_NO_PREALLOC	= (1U << 0),1362/* Instead of having one common LRU list in the1363 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list1364 * which can scale and perform better.1365 * Note, the LRU nodes (including free nodes) cannot be moved1366 * across different LRU lists.1367 */1368	BPF_F_NO_COMMON_LRU	= (1U << 1),1369/* Specify numa node during map creation */1370	BPF_F_NUMA_NODE		= (1U << 2),1371 1372/* Flags for accessing BPF object from syscall side. */1373	BPF_F_RDONLY		= (1U << 3),1374	BPF_F_WRONLY		= (1U << 4),1375 1376/* Flag for stack_map, store build_id+offset instead of pointer */1377	BPF_F_STACK_BUILD_ID	= (1U << 5),1378 1379/* Zero-initialize hash function seed. This should only be used for testing. */1380	BPF_F_ZERO_SEED		= (1U << 6),1381 1382/* Flags for accessing BPF object from program side. */1383	BPF_F_RDONLY_PROG	= (1U << 7),1384	BPF_F_WRONLY_PROG	= (1U << 8),1385 1386/* Clone map from listener for newly accepted socket */1387	BPF_F_CLONE		= (1U << 9),1388 1389/* Enable memory-mapping BPF map */1390	BPF_F_MMAPABLE		= (1U << 10),1391 1392/* Share perf_event among processes */1393	BPF_F_PRESERVE_ELEMS	= (1U << 11),1394 1395/* Create a map that is suitable to be an inner map with dynamic max entries */1396	BPF_F_INNER_MAP		= (1U << 12),1397 1398/* Create a map that will be registered/unregesitered by the backed bpf_link */1399	BPF_F_LINK		= (1U << 13),1400 1401/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */1402	BPF_F_PATH_FD		= (1U << 14),1403 1404/* Flag for value_type_btf_obj_fd, the fd is available */1405	BPF_F_VTYPE_BTF_OBJ_FD	= (1U << 15),1406 1407/* BPF token FD is passed in a corresponding command's token_fd field */1408	BPF_F_TOKEN_FD          = (1U << 16),1409 1410/* When user space page faults in bpf_arena send SIGSEGV instead of inserting new page */1411	BPF_F_SEGV_ON_FAULT	= (1U << 17),1412 1413/* Do not translate kernel bpf_arena pointers to user pointers */1414	BPF_F_NO_USER_CONV	= (1U << 18),1415};1416 1417/* Flags for BPF_PROG_QUERY. */1418 1419/* Query effective (directly attached + inherited from ancestor cgroups)1420 * programs that will be executed for events within a cgroup.1421 * attach_flags with this flag are always returned 0.1422 */1423#define BPF_F_QUERY_EFFECTIVE	(1U << 0)1424 1425/* Flags for BPF_PROG_TEST_RUN */1426 1427/* If set, run the test on the cpu specified by bpf_attr.test.cpu */1428#define BPF_F_TEST_RUN_ON_CPU	(1U << 0)1429/* If set, XDP frames will be transmitted after processing */1430#define BPF_F_TEST_XDP_LIVE_FRAMES	(1U << 1)1431/* If set, apply CHECKSUM_COMPLETE to skb and validate the checksum */1432#define BPF_F_TEST_SKB_CHECKSUM_COMPLETE	(1U << 2)1433 1434/* type for BPF_ENABLE_STATS */1435enum bpf_stats_type {1436	/* enabled run_time_ns and run_cnt */1437	BPF_STATS_RUN_TIME = 0,1438};1439 1440enum bpf_stack_build_id_status {1441	/* user space need an empty entry to identify end of a trace */1442	BPF_STACK_BUILD_ID_EMPTY = 0,1443	/* with valid build_id and offset */1444	BPF_STACK_BUILD_ID_VALID = 1,1445	/* couldn't get build_id, fallback to ip */1446	BPF_STACK_BUILD_ID_IP = 2,1447};1448 1449#define BPF_BUILD_ID_SIZE 201450struct bpf_stack_build_id {1451	__s32		status;1452	unsigned char	build_id[BPF_BUILD_ID_SIZE];1453	union {1454		__u64	offset;1455		__u64	ip;1456	};1457};1458 1459#define BPF_OBJ_NAME_LEN 16U1460 1461union bpf_attr {1462	struct { /* anonymous struct used by BPF_MAP_CREATE command */1463		__u32	map_type;	/* one of enum bpf_map_type */1464		__u32	key_size;	/* size of key in bytes */1465		__u32	value_size;	/* size of value in bytes */1466		__u32	max_entries;	/* max number of entries in a map */1467		__u32	map_flags;	/* BPF_MAP_CREATE related1468					 * flags defined above.1469					 */1470		__u32	inner_map_fd;	/* fd pointing to the inner map */1471		__u32	numa_node;	/* numa node (effective only if1472					 * BPF_F_NUMA_NODE is set).1473					 */1474		char	map_name[BPF_OBJ_NAME_LEN];1475		__u32	map_ifindex;	/* ifindex of netdev to create on */1476		__u32	btf_fd;		/* fd pointing to a BTF type data */1477		__u32	btf_key_type_id;	/* BTF type_id of the key */1478		__u32	btf_value_type_id;	/* BTF type_id of the value */1479		__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-1480						   * struct stored as the1481						   * map value1482						   */1483		/* Any per-map-type extra fields1484		 *1485		 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the1486		 * number of hash functions (if 0, the bloom filter will default1487		 * to using 5 hash functions).1488		 *1489		 * BPF_MAP_TYPE_ARENA - contains the address where user space1490		 * is going to mmap() the arena. It has to be page aligned.1491		 */1492		__u64	map_extra;1493 1494		__s32   value_type_btf_obj_fd;	/* fd pointing to a BTF1495						 * type data for1496						 * btf_vmlinux_value_type_id.1497						 */1498		/* BPF token FD to use with BPF_MAP_CREATE operation.1499		 * If provided, map_flags should have BPF_F_TOKEN_FD flag set.1500		 */1501		__s32	map_token_fd;1502	};1503 1504	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */1505		__u32		map_fd;1506		__aligned_u64	key;1507		union {1508			__aligned_u64 value;1509			__aligned_u64 next_key;1510		};1511		__u64		flags;1512	};1513 1514	struct { /* struct used by BPF_MAP_*_BATCH commands */1515		__aligned_u64	in_batch;	/* start batch,1516						 * NULL to start from beginning1517						 */1518		__aligned_u64	out_batch;	/* output: next start batch */1519		__aligned_u64	keys;1520		__aligned_u64	values;1521		__u32		count;		/* input/output:1522						 * input: # of key/value1523						 * elements1524						 * output: # of filled elements1525						 */1526		__u32		map_fd;1527		__u64		elem_flags;1528		__u64		flags;1529	} batch;1530 1531	struct { /* anonymous struct used by BPF_PROG_LOAD command */1532		__u32		prog_type;	/* one of enum bpf_prog_type */1533		__u32		insn_cnt;1534		__aligned_u64	insns;1535		__aligned_u64	license;1536		__u32		log_level;	/* verbosity level of verifier */1537		__u32		log_size;	/* size of user buffer */1538		__aligned_u64	log_buf;	/* user supplied buffer */1539		__u32		kern_version;	/* not used */1540		__u32		prog_flags;1541		char		prog_name[BPF_OBJ_NAME_LEN];1542		__u32		prog_ifindex;	/* ifindex of netdev to prep for */1543		/* For some prog types expected attach type must be known at1544		 * load time to verify attach type specific parts of prog1545		 * (context accesses, allowed helpers, etc).1546		 */1547		__u32		expected_attach_type;1548		__u32		prog_btf_fd;	/* fd pointing to BTF type data */1549		__u32		func_info_rec_size;	/* userspace bpf_func_info size */1550		__aligned_u64	func_info;	/* func info */1551		__u32		func_info_cnt;	/* number of bpf_func_info records */1552		__u32		line_info_rec_size;	/* userspace bpf_line_info size */1553		__aligned_u64	line_info;	/* line info */1554		__u32		line_info_cnt;	/* number of bpf_line_info records */1555		__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */1556		union {1557			/* valid prog_fd to attach to bpf prog */1558			__u32		attach_prog_fd;1559			/* or valid module BTF object fd or 0 to attach to vmlinux */1560			__u32		attach_btf_obj_fd;1561		};1562		__u32		core_relo_cnt;	/* number of bpf_core_relo */1563		__aligned_u64	fd_array;	/* array of FDs */1564		__aligned_u64	core_relos;1565		__u32		core_relo_rec_size; /* sizeof(struct bpf_core_relo) */1566		/* output: actual total log contents size (including termintaing zero).1567		 * It could be both larger than original log_size (if log was1568		 * truncated), or smaller (if log buffer wasn't filled completely).1569		 */1570		__u32		log_true_size;1571		/* BPF token FD to use with BPF_PROG_LOAD operation.1572		 * If provided, prog_flags should have BPF_F_TOKEN_FD flag set.1573		 */1574		__s32		prog_token_fd;1575	};1576 1577	struct { /* anonymous struct used by BPF_OBJ_* commands */1578		__aligned_u64	pathname;1579		__u32		bpf_fd;1580		__u32		file_flags;1581		/* Same as dirfd in openat() syscall; see openat(2)1582		 * manpage for details of path FD and pathname semantics;1583		 * path_fd should accompanied by BPF_F_PATH_FD flag set in1584		 * file_flags field, otherwise it should be set to zero;1585		 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.1586		 */1587		__s32		path_fd;1588	};1589 1590	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */1591		union {1592			__u32	target_fd;	/* target object to attach to or ... */1593			__u32	target_ifindex;	/* target ifindex */1594		};1595		__u32		attach_bpf_fd;1596		__u32		attach_type;1597		__u32		attach_flags;1598		__u32		replace_bpf_fd;1599		union {1600			__u32	relative_fd;1601			__u32	relative_id;1602		};1603		__u64		expected_revision;1604	};1605 1606	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */1607		__u32		prog_fd;1608		__u32		retval;1609		__u32		data_size_in;	/* input: len of data_in */1610		__u32		data_size_out;	/* input/output: len of data_out1611						 *   returns ENOSPC if data_out1612						 *   is too small.1613						 */1614		__aligned_u64	data_in;1615		__aligned_u64	data_out;1616		__u32		repeat;1617		__u32		duration;1618		__u32		ctx_size_in;	/* input: len of ctx_in */1619		__u32		ctx_size_out;	/* input/output: len of ctx_out1620						 *   returns ENOSPC if ctx_out1621						 *   is too small.1622						 */1623		__aligned_u64	ctx_in;1624		__aligned_u64	ctx_out;1625		__u32		flags;1626		__u32		cpu;1627		__u32		batch_size;1628	} test;1629 1630	struct { /* anonymous struct used by BPF_*_GET_*_ID */1631		union {1632			__u32		start_id;1633			__u32		prog_id;1634			__u32		map_id;1635			__u32		btf_id;1636			__u32		link_id;1637		};1638		__u32		next_id;1639		__u32		open_flags;1640	};1641 1642	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */1643		__u32		bpf_fd;1644		__u32		info_len;1645		__aligned_u64	info;1646	} info;1647 1648	struct { /* anonymous struct used by BPF_PROG_QUERY command */1649		union {1650			__u32	target_fd;	/* target object to query or ... */1651			__u32	target_ifindex;	/* target ifindex */1652		};1653		__u32		attach_type;1654		__u32		query_flags;1655		__u32		attach_flags;1656		__aligned_u64	prog_ids;1657		union {1658			__u32	prog_cnt;1659			__u32	count;1660		};1661		__u32		:32;1662		/* output: per-program attach_flags.1663		 * not allowed to be set during effective query.1664		 */1665		__aligned_u64	prog_attach_flags;1666		__aligned_u64	link_ids;1667		__aligned_u64	link_attach_flags;1668		__u64		revision;1669	} query;1670 1671	struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */1672		__u64		name;1673		__u32		prog_fd;1674		__u32		:32;1675		__aligned_u64	cookie;1676	} raw_tracepoint;1677 1678	struct { /* anonymous struct for BPF_BTF_LOAD */1679		__aligned_u64	btf;1680		__aligned_u64	btf_log_buf;1681		__u32		btf_size;1682		__u32		btf_log_size;1683		__u32		btf_log_level;1684		/* output: actual total log contents size (including termintaing zero).1685		 * It could be both larger than original log_size (if log was1686		 * truncated), or smaller (if log buffer wasn't filled completely).1687		 */1688		__u32		btf_log_true_size;1689		__u32		btf_flags;1690		/* BPF token FD to use with BPF_BTF_LOAD operation.1691		 * If provided, btf_flags should have BPF_F_TOKEN_FD flag set.1692		 */1693		__s32		btf_token_fd;1694	};1695 1696	struct {1697		__u32		pid;		/* input: pid */1698		__u32		fd;		/* input: fd */1699		__u32		flags;		/* input: flags */1700		__u32		buf_len;	/* input/output: buf len */1701		__aligned_u64	buf;		/* input/output:1702						 *   tp_name for tracepoint1703						 *   symbol for kprobe1704						 *   filename for uprobe1705						 */1706		__u32		prog_id;	/* output: prod_id */1707		__u32		fd_type;	/* output: BPF_FD_TYPE_* */1708		__u64		probe_offset;	/* output: probe_offset */1709		__u64		probe_addr;	/* output: probe_addr */1710	} task_fd_query;1711 1712	struct { /* struct used by BPF_LINK_CREATE command */1713		union {1714			__u32		prog_fd;	/* eBPF program to attach */1715			__u32		map_fd;		/* struct_ops to attach */1716		};1717		union {1718			__u32	target_fd;	/* target object to attach to or ... */1719			__u32	target_ifindex; /* target ifindex */1720		};1721		__u32		attach_type;	/* attach type */1722		__u32		flags;		/* extra flags */1723		union {1724			__u32	target_btf_id;	/* btf_id of target to attach to */1725			struct {1726				__aligned_u64	iter_info;	/* extra bpf_iter_link_info */1727				__u32		iter_info_len;	/* iter_info length */1728			};1729			struct {1730				/* black box user-provided value passed through1731				 * to BPF program at the execution time and1732				 * accessible through bpf_get_attach_cookie() BPF helper1733				 */1734				__u64		bpf_cookie;1735			} perf_event;1736			struct {1737				__u32		flags;1738				__u32		cnt;1739				__aligned_u64	syms;1740				__aligned_u64	addrs;1741				__aligned_u64	cookies;1742			} kprobe_multi;1743			struct {1744				/* this is overlaid with the target_btf_id above. */1745				__u32		target_btf_id;1746				/* black box user-provided value passed through1747				 * to BPF program at the execution time and1748				 * accessible through bpf_get_attach_cookie() BPF helper1749				 */1750				__u64		cookie;1751			} tracing;1752			struct {1753				__u32		pf;1754				__u32		hooknum;1755				__s32		priority;1756				__u32		flags;1757			} netfilter;1758			struct {1759				union {1760					__u32	relative_fd;1761					__u32	relative_id;1762				};1763				__u64		expected_revision;1764			} tcx;1765			struct {1766				__aligned_u64	path;1767				__aligned_u64	offsets;1768				__aligned_u64	ref_ctr_offsets;1769				__aligned_u64	cookies;1770				__u32		cnt;1771				__u32		flags;1772				__u32		pid;1773			} uprobe_multi;1774			struct {1775				union {1776					__u32	relative_fd;1777					__u32	relative_id;1778				};1779				__u64		expected_revision;1780			} netkit;1781		};1782	} link_create;1783 1784	struct { /* struct used by BPF_LINK_UPDATE command */1785		__u32		link_fd;	/* link fd */1786		union {1787			/* new program fd to update link with */1788			__u32		new_prog_fd;1789			/* new struct_ops map fd to update link with */1790			__u32           new_map_fd;1791		};1792		__u32		flags;		/* extra flags */1793		union {1794			/* expected link's program fd; is specified only if1795			 * BPF_F_REPLACE flag is set in flags.1796			 */1797			__u32		old_prog_fd;1798			/* expected link's map fd; is specified only1799			 * if BPF_F_REPLACE flag is set.1800			 */1801			__u32           old_map_fd;1802		};1803	} link_update;1804 1805	struct {1806		__u32		link_fd;1807	} link_detach;1808 1809	struct { /* struct used by BPF_ENABLE_STATS command */1810		__u32		type;1811	} enable_stats;1812 1813	struct { /* struct used by BPF_ITER_CREATE command */1814		__u32		link_fd;1815		__u32		flags;1816	} iter_create;1817 1818	struct { /* struct used by BPF_PROG_BIND_MAP command */1819		__u32		prog_fd;1820		__u32		map_fd;1821		__u32		flags;		/* extra flags */1822	} prog_bind_map;1823 1824	struct { /* struct used by BPF_TOKEN_CREATE command */1825		__u32		flags;1826		__u32		bpffs_fd;1827	} token_create;1828 1829} __attribute__((aligned(8)));1830 1831/* The description below is an attempt at providing documentation to eBPF1832 * developers about the multiple available eBPF helper functions. It can be1833 * parsed and used to produce a manual page. The workflow is the following,1834 * and requires the rst2man utility:1835 *1836 *     $ ./scripts/bpf_doc.py \1837 *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst1838 *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.71839 *     $ man /tmp/bpf-helpers.71840 *1841 * Note that in order to produce this external documentation, some RST1842 * formatting is used in the descriptions to get "bold" and "italics" in1843 * manual pages. Also note that the few trailing white spaces are1844 * intentional, removing them would break paragraphs for rst2man.1845 *1846 * Start of BPF helper function descriptions:1847 *1848 * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)1849 * 	Description1850 * 		Perform a lookup in *map* for an entry associated to *key*.1851 * 	Return1852 * 		Map value associated to *key*, or **NULL** if no entry was1853 * 		found.1854 *1855 * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)1856 * 	Description1857 * 		Add or update the value of the entry associated to *key* in1858 * 		*map* with *value*. *flags* is one of:1859 *1860 * 		**BPF_NOEXIST**1861 * 			The entry for *key* must not exist in the map.1862 * 		**BPF_EXIST**1863 * 			The entry for *key* must already exist in the map.1864 * 		**BPF_ANY**1865 * 			No condition on the existence of the entry for *key*.1866 *1867 * 		Flag value **BPF_NOEXIST** cannot be used for maps of types1868 * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all1869 * 		elements always exist), the helper would return an error.1870 * 	Return1871 * 		0 on success, or a negative error in case of failure.1872 *1873 * long bpf_map_delete_elem(struct bpf_map *map, const void *key)1874 * 	Description1875 * 		Delete entry with *key* from *map*.1876 * 	Return1877 * 		0 on success, or a negative error in case of failure.1878 *1879 * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)1880 * 	Description1881 * 		For tracing programs, safely attempt to read *size* bytes from1882 * 		kernel space address *unsafe_ptr* and store the data in *dst*.1883 *1884 * 		Generally, use **bpf_probe_read_user**\ () or1885 * 		**bpf_probe_read_kernel**\ () instead.1886 * 	Return1887 * 		0 on success, or a negative error in case of failure.1888 *1889 * u64 bpf_ktime_get_ns(void)1890 * 	Description1891 * 		Return the time elapsed since system boot, in nanoseconds.1892 * 		Does not include time the system was suspended.1893 * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)1894 * 	Return1895 * 		Current *ktime*.1896 *1897 * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)1898 * 	Description1899 * 		This helper is a "printk()-like" facility for debugging. It1900 * 		prints a message defined by format *fmt* (of size *fmt_size*)1901 * 		to file *\/sys/kernel/tracing/trace* from TraceFS, if1902 * 		available. It can take up to three additional **u64**1903 * 		arguments (as an eBPF helpers, the total number of arguments is1904 * 		limited to five).1905 *1906 * 		Each time the helper is called, it appends a line to the trace.1907 * 		Lines are discarded while *\/sys/kernel/tracing/trace* is1908 * 		open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.1909 * 		The format of the trace is customizable, and the exact output1910 * 		one will get depends on the options set in1911 * 		*\/sys/kernel/tracing/trace_options* (see also the1912 * 		*README* file under the same directory). However, it usually1913 * 		defaults to something like:1914 *1915 * 		::1916 *1917 * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>1918 *1919 * 		In the above:1920 *1921 * 			* ``telnet`` is the name of the current task.1922 * 			* ``470`` is the PID of the current task.1923 * 			* ``001`` is the CPU number on which the task is1924 * 			  running.1925 * 			* In ``.N..``, each character refers to a set of1926 * 			  options (whether irqs are enabled, scheduling1927 * 			  options, whether hard/softirqs are running, level of1928 * 			  preempt_disabled respectively). **N** means that1929 * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**1930 * 			  are set.1931 * 			* ``419421.045894`` is a timestamp.1932 * 			* ``0x00000001`` is a fake value used by BPF for the1933 * 			  instruction pointer register.1934 * 			* ``<formatted msg>`` is the message formatted with1935 * 			  *fmt*.1936 *1937 * 		The conversion specifiers supported by *fmt* are similar, but1938 * 		more limited than for printk(). They are **%d**, **%i**,1939 * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,1940 * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size1941 * 		of field, padding with zeroes, etc.) is available, and the1942 * 		helper will return **-EINVAL** (but print nothing) if it1943 * 		encounters an unknown specifier.1944 *1945 * 		Also, note that **bpf_trace_printk**\ () is slow, and should1946 * 		only be used for debugging purposes. For this reason, a notice1947 * 		block (spanning several lines) is printed to kernel logs and1948 * 		states that the helper should not be used "for production use"1949 * 		the first time this helper is used (or more precisely, when1950 * 		**trace_printk**\ () buffers are allocated). For passing values1951 * 		to user space, perf events should be preferred.1952 * 	Return1953 * 		The number of bytes written to the buffer, or a negative error1954 * 		in case of failure.1955 *1956 * u32 bpf_get_prandom_u32(void)1957 * 	Description1958 * 		Get a pseudo-random number.1959 *1960 * 		From a security point of view, this helper uses its own1961 * 		pseudo-random internal state, and cannot be used to infer the1962 * 		seed of other random functions in the kernel. However, it is1963 * 		essential to note that the generator used by the helper is not1964 * 		cryptographically secure.1965 * 	Return1966 * 		A random 32-bit unsigned value.1967 *1968 * u32 bpf_get_smp_processor_id(void)1969 * 	Description1970 * 		Get the SMP (symmetric multiprocessing) processor id. Note that1971 * 		all programs run with migration disabled, which means that the1972 * 		SMP processor id is stable during all the execution of the1973 * 		program.1974 * 	Return1975 * 		The SMP id of the processor running the program.1976 *1977 * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)1978 * 	Description1979 * 		Store *len* bytes from address *from* into the packet1980 * 		associated to *skb*, at *offset*. *flags* are a combination of1981 * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the1982 * 		checksum for the packet after storing the bytes) and1983 * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\1984 * 		**->swhash** and *skb*\ **->l4hash** to 0).1985 *1986 * 		A call to this helper is susceptible to change the underlying1987 * 		packet buffer. Therefore, at load time, all checks on pointers1988 * 		previously done by the verifier are invalidated and must be1989 * 		performed again, if the helper is used in combination with1990 * 		direct packet access.1991 * 	Return1992 * 		0 on success, or a negative error in case of failure.1993 *1994 * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)1995 * 	Description1996 * 		Recompute the layer 3 (e.g. IP) checksum for the packet1997 * 		associated to *skb*. Computation is incremental, so the helper1998 * 		must know the former value of the header field that was1999 * 		modified (*from*), the new value of this field (*to*), and the2000 * 		number of bytes (2 or 4) for this field, stored in *size*.2001 * 		Alternatively, it is possible to store the difference between2002 * 		the previous and the new values of the header field in *to*, by2003 * 		setting *from* and *size* to 0. For both methods, *offset*2004 * 		indicates the location of the IP checksum within the packet.2005 *2006 * 		This helper works in combination with **bpf_csum_diff**\ (),2007 * 		which does not update the checksum in-place, but offers more2008 * 		flexibility and can handle sizes larger than 2 or 4 for the2009 * 		checksum to update.2010 *2011 * 		A call to this helper is susceptible to change the underlying2012 * 		packet buffer. Therefore, at load time, all checks on pointers2013 * 		previously done by the verifier are invalidated and must be2014 * 		performed again, if the helper is used in combination with2015 * 		direct packet access.2016 * 	Return2017 * 		0 on success, or a negative error in case of failure.2018 *2019 * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)2020 * 	Description2021 * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the2022 * 		packet associated to *skb*. Computation is incremental, so the2023 * 		helper must know the former value of the header field that was2024 * 		modified (*from*), the new value of this field (*to*), and the2025 * 		number of bytes (2 or 4) for this field, stored on the lowest2026 * 		four bits of *flags*. Alternatively, it is possible to store2027 * 		the difference between the previous and the new values of the2028 * 		header field in *to*, by setting *from* and the four lowest2029 * 		bits of *flags* to 0. For both methods, *offset* indicates the2030 * 		location of the IP checksum within the packet. In addition to2031 * 		the size of the field, *flags* can be added (bitwise OR) actual2032 * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left2033 * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and2034 * 		for updates resulting in a null checksum the value is set to2035 * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates2036 * 		the checksum is to be computed against a pseudo-header.2037 *2038 * 		This helper works in combination with **bpf_csum_diff**\ (),2039 * 		which does not update the checksum in-place, but offers more2040 * 		flexibility and can handle sizes larger than 2 or 4 for the2041 * 		checksum to update.2042 *2043 * 		A call to this helper is susceptible to change the underlying2044 * 		packet buffer. Therefore, at load time, all checks on pointers2045 * 		previously done by the verifier are invalidated and must be2046 * 		performed again, if the helper is used in combination with2047 * 		direct packet access.2048 * 	Return2049 * 		0 on success, or a negative error in case of failure.2050 *2051 * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)2052 * 	Description2053 * 		This special helper is used to trigger a "tail call", or in2054 * 		other words, to jump into another eBPF program. The same stack2055 * 		frame is used (but values on stack and in registers for the2056 * 		caller are not accessible to the callee). This mechanism allows2057 * 		for program chaining, either for raising the maximum number of2058 * 		available eBPF instructions, or to execute given programs in2059 * 		conditional blocks. For security reasons, there is an upper2060 * 		limit to the number of successive tail calls that can be2061 * 		performed.2062 *2063 * 		Upon call of this helper, the program attempts to jump into a2064 * 		program referenced at index *index* in *prog_array_map*, a2065 * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes2066 * 		*ctx*, a pointer to the context.2067 *2068 * 		If the call succeeds, the kernel immediately runs the first2069 * 		instruction of the new program. This is not a function call,2070 * 		and it never returns to the previous program. If the call2071 * 		fails, then the helper has no effect, and the caller continues2072 * 		to run its subsequent instructions. A call can fail if the2073 * 		destination program for the jump does not exist (i.e. *index*2074 * 		is superior to the number of entries in *prog_array_map*), or2075 * 		if the maximum number of tail calls has been reached for this2076 * 		chain of programs. This limit is defined in the kernel by the2077 * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),2078 *		which is currently set to 33.2079 * 	Return2080 * 		0 on success, or a negative error in case of failure.2081 *2082 * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)2083 * 	Description2084 * 		Clone and redirect the packet associated to *skb* to another2085 * 		net device of index *ifindex*. Both ingress and egress2086 * 		interfaces can be used for redirection. The **BPF_F_INGRESS**2087 * 		value in *flags* is used to make the distinction (ingress path2088 * 		is selected if the flag is present, egress path otherwise).2089 * 		This is the only flag supported for now.2090 *2091 * 		In comparison with **bpf_redirect**\ () helper,2092 * 		**bpf_clone_redirect**\ () has the associated cost of2093 * 		duplicating the packet buffer, but this can be executed out of2094 * 		the eBPF program. Conversely, **bpf_redirect**\ () is more2095 * 		efficient, but it is handled through an action code where the2096 * 		redirection happens only after the eBPF program has returned.2097 *2098 * 		A call to this helper is susceptible to change the underlying2099 * 		packet buffer. Therefore, at load time, all checks on pointers2100 * 		previously done by the verifier are invalidated and must be2101 * 		performed again, if the helper is used in combination with2102 * 		direct packet access.2103 * 	Return2104 * 		0 on success, or a negative error in case of failure. Positive2105 * 		error indicates a potential drop or congestion in the target2106 * 		device. The particular positive error codes are not defined.2107 *2108 * u64 bpf_get_current_pid_tgid(void)2109 * 	Description2110 * 		Get the current pid and tgid.2111 * 	Return2112 * 		A 64-bit integer containing the current tgid and pid, and2113 * 		created as such:2114 * 		*current_task*\ **->tgid << 32 \|**2115 * 		*current_task*\ **->pid**.2116 *2117 * u64 bpf_get_current_uid_gid(void)2118 * 	Description2119 * 		Get the current uid and gid.2120 * 	Return2121 * 		A 64-bit integer containing the current GID and UID, and2122 * 		created as such: *current_gid* **<< 32 \|** *current_uid*.2123 *2124 * long bpf_get_current_comm(void *buf, u32 size_of_buf)2125 * 	Description2126 * 		Copy the **comm** attribute of the current task into *buf* of2127 * 		*size_of_buf*. The **comm** attribute contains the name of2128 * 		the executable (excluding the path) for the current task. The2129 * 		*size_of_buf* must be strictly positive. On success, the2130 * 		helper makes sure that the *buf* is NUL-terminated. On failure,2131 * 		it is filled with zeroes.2132 * 	Return2133 * 		0 on success, or a negative error in case of failure.2134 *2135 * u32 bpf_get_cgroup_classid(struct sk_buff *skb)2136 * 	Description2137 * 		Retrieve the classid for the current task, i.e. for the net_cls2138 * 		cgroup to which *skb* belongs.2139 *2140 * 		This helper can be used on TC egress path, but not on ingress.2141 *2142 * 		The net_cls cgroup provides an interface to tag network packets2143 * 		based on a user-provided identifier for all traffic coming from2144 * 		the tasks belonging to the related cgroup. See also the related2145 * 		kernel documentation, available from the Linux sources in file2146 * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.2147 *2148 * 		The Linux kernel has two versions for cgroups: there are2149 * 		cgroups v1 and cgroups v2. Both are available to users, who can2150 * 		use a mixture of them, but note that the net_cls cgroup is for2151 * 		cgroup v1 only. This makes it incompatible with BPF programs2152 * 		run on cgroups, which is a cgroup-v2-only feature (a socket can2153 * 		only hold data for one version of cgroups at a time).2154 *2155 * 		This helper is only available is the kernel was compiled with2156 * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to2157 * 		"**y**" or to "**m**".2158 * 	Return2159 * 		The classid, or 0 for the default unconfigured classid.2160 *2161 * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)2162 * 	Description2163 * 		Push a *vlan_tci* (VLAN tag control information) of protocol2164 * 		*vlan_proto* to the packet associated to *skb*, then update2165 * 		the checksum. Note that if *vlan_proto* is different from2166 * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to2167 * 		be **ETH_P_8021Q**.2168 *2169 * 		A call to this helper is susceptible to change the underlying2170 * 		packet buffer. Therefore, at load time, all checks on pointers2171 * 		previously done by the verifier are invalidated and must be2172 * 		performed again, if the helper is used in combination with2173 * 		direct packet access.2174 * 	Return2175 * 		0 on success, or a negative error in case of failure.2176 *2177 * long bpf_skb_vlan_pop(struct sk_buff *skb)2178 * 	Description2179 * 		Pop a VLAN header from the packet associated to *skb*.2180 *2181 * 		A call to this helper is susceptible to change the underlying2182 * 		packet buffer. Therefore, at load time, all checks on pointers2183 * 		previously done by the verifier are invalidated and must be2184 * 		performed again, if the helper is used in combination with2185 * 		direct packet access.2186 * 	Return2187 * 		0 on success, or a negative error in case of failure.2188 *2189 * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)2190 * 	Description2191 * 		Get tunnel metadata. This helper takes a pointer *key* to an2192 * 		empty **struct bpf_tunnel_key** of **size**, that will be2193 * 		filled with tunnel metadata for the packet associated to *skb*.2194 * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which2195 * 		indicates that the tunnel is based on IPv6 protocol instead of2196 * 		IPv4.2197 *2198 * 		The **struct bpf_tunnel_key** is an object that generalizes the2199 * 		principal parameters used by various tunneling protocols into a2200 * 		single struct. This way, it can be used to easily make a2201 * 		decision based on the contents of the encapsulation header,2202 * 		"summarized" in this struct. In particular, it holds the IP2203 * 		address of the remote end (IPv4 or IPv6, depending on the case)2204 * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,2205 * 		this struct exposes the *key*\ **->tunnel_id**, which is2206 * 		generally mapped to a VNI (Virtual Network Identifier), making2207 * 		it programmable together with the **bpf_skb_set_tunnel_key**\2208 * 		() helper.2209 *2210 * 		Let's imagine that the following code is part of a program2211 * 		attached to the TC ingress interface, on one end of a GRE2212 * 		tunnel, and is supposed to filter out all messages coming from2213 * 		remote ends with IPv4 address other than 10.0.0.1:2214 *2215 * 		::2216 *2217 * 			int ret;2218 * 			struct bpf_tunnel_key key = {};2219 *2220 * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);2221 * 			if (ret < 0)2222 * 				return TC_ACT_SHOT;	// drop packet2223 *2224 * 			if (key.remote_ipv4 != 0x0a000001)2225 * 				return TC_ACT_SHOT;	// drop packet2226 *2227 * 			return TC_ACT_OK;		// accept packet2228 *2229 * 		This interface can also be used with all encapsulation devices2230 * 		that can operate in "collect metadata" mode: instead of having2231 * 		one network device per specific configuration, the "collect2232 * 		metadata" mode only requires a single device where the2233 * 		configuration can be extracted from this helper.2234 *2235 * 		This can be used together with various tunnels such as VXLan,2236 * 		Geneve, GRE or IP in IP (IPIP).2237 * 	Return2238 * 		0 on success, or a negative error in case of failure.2239 *2240 * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)2241 * 	Description2242 * 		Populate tunnel metadata for packet associated to *skb.* The2243 * 		tunnel metadata is set to the contents of *key*, of *size*. The2244 * 		*flags* can be set to a combination of the following values:2245 *2246 * 		**BPF_F_TUNINFO_IPV6**2247 * 			Indicate that the tunnel is based on IPv6 protocol2248 * 			instead of IPv4.2249 * 		**BPF_F_ZERO_CSUM_TX**2250 * 			For IPv4 packets, add a flag to tunnel metadata2251 * 			indicating that checksum computation should be skipped2252 * 			and checksum set to zeroes.2253 * 		**BPF_F_DONT_FRAGMENT**2254 * 			Add a flag to tunnel metadata indicating that the2255 * 			packet should not be fragmented.2256 * 		**BPF_F_SEQ_NUMBER**2257 * 			Add a flag to tunnel metadata indicating that a2258 * 			sequence number should be added to tunnel header before2259 * 			sending the packet. This flag was added for GRE2260 * 			encapsulation, but might be used with other protocols2261 * 			as well in the future.2262 * 		**BPF_F_NO_TUNNEL_KEY**2263 * 			Add a flag to tunnel metadata indicating that no tunnel2264 * 			key should be set in the resulting tunnel header.2265 *2266 * 		Here is a typical usage on the transmit path:2267 *2268 * 		::2269 *2270 * 			struct bpf_tunnel_key key;2271 * 			     populate key ...2272 * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);2273 * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);2274 *2275 * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()2276 * 		helper for additional information.2277 * 	Return2278 * 		0 on success, or a negative error in case of failure.2279 *2280 * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)2281 * 	Description2282 * 		Read the value of a perf event counter. This helper relies on a2283 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of2284 * 		the perf event counter is selected when *map* is updated with2285 * 		perf event file descriptors. The *map* is an array whose size2286 * 		is the number of available CPUs, and each cell contains a value2287 * 		relative to one CPU. The value to retrieve is indicated by2288 * 		*flags*, that contains the index of the CPU to look up, masked2289 * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to2290 * 		**BPF_F_CURRENT_CPU** to indicate that the value for the2291 * 		current CPU should be retrieved.2292 *2293 * 		Note that before Linux 4.13, only hardware perf event can be2294 * 		retrieved.2295 *2296 * 		Also, be aware that the newer helper2297 * 		**bpf_perf_event_read_value**\ () is recommended over2298 * 		**bpf_perf_event_read**\ () in general. The latter has some ABI2299 * 		quirks where error and counter value are used as a return code2300 * 		(which is wrong to do since ranges may overlap). This issue is2301 * 		fixed with **bpf_perf_event_read_value**\ (), which at the same2302 * 		time provides more features over the **bpf_perf_event_read**\2303 * 		() interface. Please refer to the description of2304 * 		**bpf_perf_event_read_value**\ () for details.2305 * 	Return2306 * 		The value of the perf event counter read from the map, or a2307 * 		negative error code in case of failure.2308 *2309 * long bpf_redirect(u32 ifindex, u64 flags)2310 * 	Description2311 * 		Redirect the packet to another net device of index *ifindex*.2312 * 		This helper is somewhat similar to **bpf_clone_redirect**\2313 * 		(), except that the packet is not cloned, which provides2314 * 		increased performance.2315 *2316 * 		Except for XDP, both ingress and egress interfaces can be used2317 * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used2318 * 		to make the distinction (ingress path is selected if the flag2319 * 		is present, egress path otherwise). Currently, XDP only2320 * 		supports redirection to the egress interface, and accepts no2321 * 		flag at all.2322 *2323 * 		The same effect can also be attained with the more generic2324 * 		**bpf_redirect_map**\ (), which uses a BPF map to store the2325 * 		redirect target instead of providing it directly to the helper.2326 * 	Return2327 * 		For XDP, the helper returns **XDP_REDIRECT** on success or2328 * 		**XDP_ABORTED** on error. For other program types, the values2329 * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on2330 * 		error.2331 *2332 * u32 bpf_get_route_realm(struct sk_buff *skb)2333 * 	Description2334 * 		Retrieve the realm or the route, that is to say the2335 * 		**tclassid** field of the destination for the *skb*. The2336 * 		identifier retrieved is a user-provided tag, similar to the2337 * 		one used with the net_cls cgroup (see description for2338 * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is2339 * 		held by a route (a destination entry), not by a task.2340 *2341 * 		Retrieving this identifier works with the clsact TC egress hook2342 * 		(see also **tc-bpf(8)**), or alternatively on conventional2343 * 		classful egress qdiscs, but not on TC ingress path. In case of2344 * 		clsact TC egress hook, this has the advantage that, internally,2345 * 		the destination entry has not been dropped yet in the transmit2346 * 		path. Therefore, the destination entry does not need to be2347 * 		artificially held via **netif_keep_dst**\ () for a classful2348 * 		qdisc until the *skb* is freed.2349 *2350 * 		This helper is available only if the kernel was compiled with2351 * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.2352 * 	Return2353 * 		The realm of the route for the packet associated to *skb*, or 02354 * 		if none was found.2355 *2356 * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)2357 * 	Description2358 * 		Write raw *data* blob into a special BPF perf event held by2359 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf2360 * 		event must have the following attributes: **PERF_SAMPLE_RAW**2361 * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and2362 * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.2363 *2364 * 		The *flags* are used to indicate the index in *map* for which2365 * 		the value must be put, masked with **BPF_F_INDEX_MASK**.2366 * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**2367 * 		to indicate that the index of the current CPU core should be2368 * 		used.2369 *2370 * 		The value to write, of *size*, is passed through eBPF stack and2371 * 		pointed by *data*.2372 *2373 * 		The context of the program *ctx* needs also be passed to the2374 * 		helper.2375 *2376 * 		On user space, a program willing to read the values needs to2377 * 		call **perf_event_open**\ () on the perf event (either for2378 * 		one or for all CPUs) and to store the file descriptor into the2379 * 		*map*. This must be done before the eBPF program can send data2380 * 		into it. An example is available in file2381 * 		*samples/bpf/trace_output_user.c* in the Linux kernel source2382 * 		tree (the eBPF program counterpart is in2383 * 		*samples/bpf/trace_output_kern.c*).2384 *2385 * 		**bpf_perf_event_output**\ () achieves better performance2386 * 		than **bpf_trace_printk**\ () for sharing data with user2387 * 		space, and is much better suitable for streaming data from eBPF2388 * 		programs.2389 *2390 * 		Note that this helper is not restricted to tracing use cases2391 * 		and can be used with programs attached to TC or XDP as well,2392 * 		where it allows for passing data to user space listeners. Data2393 * 		can be:2394 *2395 * 		* Only custom structs,2396 * 		* Only the packet payload, or2397 * 		* A combination of both.2398 * 	Return2399 * 		0 on success, or a negative error in case of failure.2400 *2401 * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)2402 * 	Description2403 * 		This helper was provided as an easy way to load data from a2404 * 		packet. It can be used to load *len* bytes from *offset* from2405 * 		the packet associated to *skb*, into the buffer pointed by2406 * 		*to*.2407 *2408 * 		Since Linux 4.7, usage of this helper has mostly been replaced2409 * 		by "direct packet access", enabling packet data to be2410 * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**2411 * 		pointing respectively to the first byte of packet data and to2412 * 		the byte after the last byte of packet data. However, it2413 * 		remains useful if one wishes to read large quantities of data2414 * 		at once from a packet into the eBPF stack.2415 * 	Return2416 * 		0 on success, or a negative error in case of failure.2417 *2418 * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)2419 * 	Description2420 * 		Walk a user or a kernel stack and return its id. To achieve2421 * 		this, the helper needs *ctx*, which is a pointer to the context2422 * 		on which the tracing program is executed, and a pointer to a2423 * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.2424 *2425 * 		The last argument, *flags*, holds the number of stack frames to2426 * 		skip (from 0 to 255), masked with2427 * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set2428 * 		a combination of the following flags:2429 *2430 * 		**BPF_F_USER_STACK**2431 * 			Collect a user space stack instead of a kernel stack.2432 * 		**BPF_F_FAST_STACK_CMP**2433 * 			Compare stacks by hash only.2434 * 		**BPF_F_REUSE_STACKID**2435 * 			If two different stacks hash into the same *stackid*,2436 * 			discard the old one.2437 *2438 * 		The stack id retrieved is a 32 bit long integer handle which2439 * 		can be further combined with other data (including other stack2440 * 		ids) and used as a key into maps. This can be useful for2441 * 		generating a variety of graphs (such as flame graphs or off-cpu2442 * 		graphs).2443 *2444 * 		For walking a stack, this helper is an improvement over2445 * 		**bpf_probe_read**\ (), which can be used with unrolled loops2446 * 		but is not efficient and consumes a lot of eBPF instructions.2447 * 		Instead, **bpf_get_stackid**\ () can collect up to2448 * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that2449 * 		this limit can be controlled with the **sysctl** program, and2450 * 		that it should be manually increased in order to profile long2451 * 		user stacks (such as stacks for Java programs). To do so, use:2452 *2453 * 		::2454 *2455 * 			# sysctl kernel.perf_event_max_stack=<new value>2456 * 	Return2457 * 		The positive or null stack id on success, or a negative error2458 * 		in case of failure.2459 *2460 * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)2461 * 	Description2462 * 		Compute a checksum difference, from the raw buffer pointed by2463 * 		*from*, of length *from_size* (that must be a multiple of 4),2464 * 		towards the raw buffer pointed by *to*, of size *to_size*2465 * 		(same remark). An optional *seed* can be added to the value2466 * 		(this can be cascaded, the seed may come from a previous call2467 * 		to the helper).2468 *2469 * 		This is flexible enough to be used in several ways:2470 *2471 * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to2472 * 		  checksum, it can be used when pushing new data.2473 * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to2474 * 		  checksum, it can be used when removing data from a packet.2475 * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it2476 * 		  can be used to compute a diff. Note that *from_size* and2477 * 		  *to_size* do not need to be equal.2478 *2479 * 		This helper can be used in combination with2480 * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to2481 * 		which one can feed in the difference computed with2482 * 		**bpf_csum_diff**\ ().2483 * 	Return2484 * 		The checksum result, or a negative error code in case of2485 * 		failure.2486 *2487 * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)2488 * 	Description2489 * 		Retrieve tunnel options metadata for the packet associated to2490 * 		*skb*, and store the raw tunnel option data to the buffer *opt*2491 * 		of *size*.2492 *2493 * 		This helper can be used with encapsulation devices that can2494 * 		operate in "collect metadata" mode (please refer to the related2495 * 		note in the description of **bpf_skb_get_tunnel_key**\ () for2496 * 		more details). A particular example where this can be used is2497 * 		in combination with the Geneve encapsulation protocol, where it2498 * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)2499 * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from2500 * 		the eBPF program. This allows for full customization of these2501 * 		headers.2502 * 	Return2503 * 		The size of the option data retrieved.2504 *2505 * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)2506 * 	Description2507 * 		Set tunnel options metadata for the packet associated to *skb*2508 * 		to the option data contained in the raw buffer *opt* of *size*.2509 *2510 * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()2511 * 		helper for additional information.2512 * 	Return2513 * 		0 on success, or a negative error in case of failure.2514 *2515 * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)2516 * 	Description2517 * 		Change the protocol of the *skb* to *proto*. Currently2518 * 		supported are transition from IPv4 to IPv6, and from IPv6 to2519 * 		IPv4. The helper takes care of the groundwork for the2520 * 		transition, including resizing the socket buffer. The eBPF2521 * 		program is expected to fill the new headers, if any, via2522 * 		**skb_store_bytes**\ () and to recompute the checksums with2523 * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\2524 * 		(). The main case for this helper is to perform NAT642525 * 		operations out of an eBPF program.2526 *2527 * 		Internally, the GSO type is marked as dodgy so that headers are2528 * 		checked and segments are recalculated by the GSO/GRO engine.2529 * 		The size for GSO target is adapted as well.2530 *2531 * 		All values for *flags* are reserved for future usage, and must2532 * 		be left at zero.2533 *2534 * 		A call to this helper is susceptible to change the underlying2535 * 		packet buffer. Therefore, at load time, all checks on pointers2536 * 		previously done by the verifier are invalidated and must be2537 * 		performed again, if the helper is used in combination with2538 * 		direct packet access.2539 * 	Return2540 * 		0 on success, or a negative error in case of failure.2541 *2542 * long bpf_skb_change_type(struct sk_buff *skb, u32 type)2543 * 	Description2544 * 		Change the packet type for the packet associated to *skb*. This2545 * 		comes down to setting *skb*\ **->pkt_type** to *type*, except2546 * 		the eBPF program does not have a write access to *skb*\2547 * 		**->pkt_type** beside this helper. Using a helper here allows2548 * 		for graceful handling of errors.2549 *2550 * 		The major use case is to change incoming *skb*s to2551 * 		**PACKET_HOST** in a programmatic way instead of having to2552 * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for2553 * 		example.2554 *2555 * 		Note that *type* only allows certain values. At this time, they2556 * 		are:2557 *2558 * 		**PACKET_HOST**2559 * 			Packet is for us.2560 * 		**PACKET_BROADCAST**2561 * 			Send packet to all.2562 * 		**PACKET_MULTICAST**2563 * 			Send packet to group.2564 * 		**PACKET_OTHERHOST**2565 * 			Send packet to someone else.2566 * 	Return2567 * 		0 on success, or a negative error in case of failure.2568 *2569 * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)2570 * 	Description2571 * 		Check whether *skb* is a descendant of the cgroup2 held by2572 * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.2573 * 	Return2574 * 		The return value depends on the result of the test, and can be:2575 *2576 * 		* 0, if the *skb* failed the cgroup2 descendant test.2577 * 		* 1, if the *skb* succeeded the cgroup2 descendant test.2578 * 		* A negative error code, if an error occurred.2579 *2580 * u32 bpf_get_hash_recalc(struct sk_buff *skb)2581 * 	Description2582 * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is2583 * 		not set, in particular if the hash was cleared due to mangling,2584 * 		recompute this hash. Later accesses to the hash can be done2585 * 		directly with *skb*\ **->hash**.2586 *2587 * 		Calling **bpf_set_hash_invalid**\ (), changing a packet2588 * 		prototype with **bpf_skb_change_proto**\ (), or calling2589 * 		**bpf_skb_store_bytes**\ () with the2590 * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear2591 * 		the hash and to trigger a new computation for the next call to2592 * 		**bpf_get_hash_recalc**\ ().2593 * 	Return2594 * 		The 32-bit hash.2595 *2596 * u64 bpf_get_current_task(void)2597 * 	Description2598 * 		Get the current task.2599 * 	Return2600 * 		A pointer to the current task struct.2601 *2602 * long bpf_probe_write_user(void *dst, const void *src, u32 len)2603 * 	Description2604 * 		Attempt in a safe way to write *len* bytes from the buffer2605 * 		*src* to *dst* in memory. It only works for threads that are in2606 * 		user context, and *dst* must be a valid user space address.2607 *2608 * 		This helper should not be used to implement any kind of2609 * 		security mechanism because of TOC-TOU attacks, but rather to2610 * 		debug, divert, and manipulate execution of semi-cooperative2611 * 		processes.2612 *2613 * 		Keep in mind that this feature is meant for experiments, and it2614 * 		has a risk of crashing the system and running programs.2615 * 		Therefore, when an eBPF program using this helper is attached,2616 * 		a warning including PID and process name is printed to kernel2617 * 		logs.2618 * 	Return2619 * 		0 on success, or a negative error in case of failure.2620 *2621 * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)2622 * 	Description2623 * 		Check whether the probe is being run is the context of a given2624 * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by2625 * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.2626 * 	Return2627 * 		The return value depends on the result of the test, and can be:2628 *2629 *		* 1, if current task belongs to the cgroup2.2630 *		* 0, if current task does not belong to the cgroup2.2631 * 		* A negative error code, if an error occurred.2632 *2633 * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)2634 * 	Description2635 * 		Resize (trim or grow) the packet associated to *skb* to the2636 * 		new *len*. The *flags* are reserved for future usage, and must2637 * 		be left at zero.2638 *2639 * 		The basic idea is that the helper performs the needed work to2640 * 		change the size of the packet, then the eBPF program rewrites2641 * 		the rest via helpers like **bpf_skb_store_bytes**\ (),2642 * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()2643 * 		and others. This helper is a slow path utility intended for2644 * 		replies with control messages. And because it is targeted for2645 * 		slow path, the helper itself can afford to be slow: it2646 * 		implicitly linearizes, unclones and drops offloads from the2647 * 		*skb*.2648 *2649 * 		A call to this helper is susceptible to change the underlying2650 * 		packet buffer. Therefore, at load time, all checks on pointers2651 * 		previously done by the verifier are invalidated and must be2652 * 		performed again, if the helper is used in combination with2653 * 		direct packet access.2654 * 	Return2655 * 		0 on success, or a negative error in case of failure.2656 *2657 * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)2658 * 	Description2659 * 		Pull in non-linear data in case the *skb* is non-linear and not2660 * 		all of *len* are part of the linear section. Make *len* bytes2661 * 		from *skb* readable and writable. If a zero value is passed for2662 *		*len*, then all bytes in the linear part of *skb* will be made2663 *		readable and writable.2664 *2665 * 		This helper is only needed for reading and writing with direct2666 * 		packet access.2667 *2668 * 		For direct packet access, testing that offsets to access2669 * 		are within packet boundaries (test on *skb*\ **->data_end**) is2670 * 		susceptible to fail if offsets are invalid, or if the requested2671 * 		data is in non-linear parts of the *skb*. On failure the2672 * 		program can just bail out, or in the case of a non-linear2673 * 		buffer, use a helper to make the data available. The2674 * 		**bpf_skb_load_bytes**\ () helper is a first solution to access2675 * 		the data. Another one consists in using **bpf_skb_pull_data**2676 * 		to pull in once the non-linear parts, then retesting and2677 * 		eventually access the data.2678 *2679 * 		At the same time, this also makes sure the *skb* is uncloned,2680 * 		which is a necessary condition for direct write. As this needs2681 * 		to be an invariant for the write part only, the verifier2682 * 		detects writes and adds a prologue that is calling2683 * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from2684 * 		the very beginning in case it is indeed cloned.2685 *2686 * 		A call to this helper is susceptible to change the underlying2687 * 		packet buffer. Therefore, at load time, all checks on pointers2688 * 		previously done by the verifier are invalidated and must be2689 * 		performed again, if the helper is used in combination with2690 * 		direct packet access.2691 * 	Return2692 * 		0 on success, or a negative error in case of failure.2693 *2694 * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)2695 * 	Description2696 * 		Add the checksum *csum* into *skb*\ **->csum** in case the2697 * 		driver has supplied a checksum for the entire packet into that2698 * 		field. Return an error otherwise. This helper is intended to be2699 * 		used in combination with **bpf_csum_diff**\ (), in particular2700 * 		when the checksum needs to be updated after data has been2701 * 		written into the packet through direct packet access.2702 * 	Return2703 * 		The checksum on success, or a negative error code in case of2704 * 		failure.2705 *2706 * void bpf_set_hash_invalid(struct sk_buff *skb)2707 * 	Description2708 * 		Invalidate the current *skb*\ **->hash**. It can be used after2709 * 		mangling on headers through direct packet access, in order to2710 * 		indicate that the hash is outdated and to trigger a2711 * 		recalculation the next time the kernel tries to access this2712 * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.2713 * 	Return2714 * 		void.2715 *2716 * long bpf_get_numa_node_id(void)2717 * 	Description2718 * 		Return the id of the current NUMA node. The primary use case2719 * 		for this helper is the selection of sockets for the local NUMA2720 * 		node, when the program is attached to sockets using the2721 * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),2722 * 		but the helper is also available to other eBPF program types,2723 * 		similarly to **bpf_get_smp_processor_id**\ ().2724 * 	Return2725 * 		The id of current NUMA node.2726 *2727 * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)2728 * 	Description2729 * 		Grows headroom of packet associated to *skb* and adjusts the2730 * 		offset of the MAC header accordingly, adding *len* bytes of2731 * 		space. It automatically extends and reallocates memory as2732 * 		required.2733 *2734 * 		This helper can be used on a layer 3 *skb* to push a MAC header2735 * 		for redirection into a layer 2 device.2736 *2737 * 		All values for *flags* are reserved for future usage, and must2738 * 		be left at zero.2739 *2740 * 		A call to this helper is susceptible to change the underlying2741 * 		packet buffer. Therefore, at load time, all checks on pointers2742 * 		previously done by the verifier are invalidated and must be2743 * 		performed again, if the helper is used in combination with2744 * 		direct packet access.2745 * 	Return2746 * 		0 on success, or a negative error in case of failure.2747 *2748 * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)2749 * 	Description2750 * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that2751 * 		it is possible to use a negative value for *delta*. This helper2752 * 		can be used to prepare the packet for pushing or popping2753 * 		headers.2754 *2755 * 		A call to this helper is susceptible to change the underlying2756 * 		packet buffer. Therefore, at load time, all checks on pointers2757 * 		previously done by the verifier are invalidated and must be2758 * 		performed again, if the helper is used in combination with2759 * 		direct packet access.2760 * 	Return2761 * 		0 on success, or a negative error in case of failure.2762 *2763 * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)2764 * 	Description2765 * 		Copy a NUL terminated string from an unsafe kernel address2766 * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for2767 * 		more details.2768 *2769 * 		Generally, use **bpf_probe_read_user_str**\ () or2770 * 		**bpf_probe_read_kernel_str**\ () instead.2771 * 	Return2772 * 		On success, the strictly positive length of the string,2773 * 		including the trailing NUL character. On error, a negative2774 * 		value.2775 *2776 * u64 bpf_get_socket_cookie(struct sk_buff *skb)2777 * 	Description2778 * 		If the **struct sk_buff** pointed by *skb* has a known socket,2779 * 		retrieve the cookie (generated by the kernel) of this socket.2780 * 		If no cookie has been set yet, generate a new cookie. Once2781 * 		generated, the socket cookie remains stable for the life of the2782 * 		socket. This helper can be useful for monitoring per socket2783 * 		networking traffic statistics as it provides a global socket2784 * 		identifier that can be assumed unique.2785 * 	Return2786 * 		A 8-byte long unique number on success, or 0 if the socket2787 * 		field is missing inside *skb*.2788 *2789 * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)2790 * 	Description2791 * 		Equivalent to bpf_get_socket_cookie() helper that accepts2792 * 		*skb*, but gets socket from **struct bpf_sock_addr** context.2793 * 	Return2794 * 		A 8-byte long unique number.2795 *2796 * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)2797 * 	Description2798 * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts2799 * 		*skb*, but gets socket from **struct bpf_sock_ops** context.2800 * 	Return2801 * 		A 8-byte long unique number.2802 *2803 * u64 bpf_get_socket_cookie(struct sock *sk)2804 * 	Description2805 * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts2806 * 		*sk*, but gets socket from a BTF **struct sock**. This helper2807 * 		also works for sleepable programs.2808 * 	Return2809 * 		A 8-byte long unique number or 0 if *sk* is NULL.2810 *2811 * u32 bpf_get_socket_uid(struct sk_buff *skb)2812 * 	Description2813 * 		Get the owner UID of the socked associated to *skb*.2814 * 	Return2815 * 		The owner UID of the socket associated to *skb*. If the socket2816 * 		is **NULL**, or if it is not a full socket (i.e. if it is a2817 * 		time-wait or a request socket instead), **overflowuid** value2818 * 		is returned (note that **overflowuid** might also be the actual2819 * 		UID value for the socket).2820 *2821 * long bpf_set_hash(struct sk_buff *skb, u32 hash)2822 * 	Description2823 * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)2824 * 		to value *hash*.2825 * 	Return2826 * 		02827 *2828 * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)2829 * 	Description2830 * 		Emulate a call to **setsockopt()** on the socket associated to2831 * 		*bpf_socket*, which must be a full socket. The *level* at2832 * 		which the option resides and the name *optname* of the option2833 * 		must be specified, see **setsockopt(2)** for more information.2834 * 		The option value of length *optlen* is pointed by *optval*.2835 *2836 * 		*bpf_socket* should be one of the following:2837 *2838 * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.2839 *		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,2840 *		  **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.2841 *2842 * 		This helper actually implements a subset of **setsockopt()**.2843 * 		It supports the following *level*\ s:2844 *2845 * 		* **SOL_SOCKET**, which supports the following *optname*\ s:2846 * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,2847 * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,2848 * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,2849 * 		  **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.2850 * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:2851 * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,2852 * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,2853 * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,2854 * 		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,2855 * 		  **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,2856 * 		  **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,2857 *		  **TCP_BPF_RTO_MIN**, **TCP_BPF_SOCK_OPS_CB_FLAGS**.2858 * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.2859 * 		* **IPPROTO_IPV6**, which supports the following *optname*\ s:2860 * 		  **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.2861 * 	Return2862 * 		0 on success, or a negative error in case of failure.2863 *2864 * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)2865 * 	Description2866 * 		Grow or shrink the room for data in the packet associated to2867 * 		*skb* by *len_diff*, and according to the selected *mode*.2868 *2869 * 		By default, the helper will reset any offloaded checksum2870 * 		indicator of the skb to CHECKSUM_NONE. This can be avoided2871 * 		by the following flag:2872 *2873 * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded2874 * 		  checksum data of the skb to CHECKSUM_NONE.2875 *2876 *		There are two supported modes at this time:2877 *2878 *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer2879 * 		  (room space is added or removed between the layer 2 and2880 * 		  layer 3 headers).2881 *2882 * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer2883 * 		  (room space is added or removed between the layer 3 and2884 * 		  layer 4 headers).2885 *2886 *		The following flags are supported at this time:2887 *2888 *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.2889 *		  Adjusting mss in this way is not allowed for datagrams.2890 *2891 *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,2892 *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:2893 *		  Any new space is reserved to hold a tunnel header.2894 *		  Configure skb offsets and other fields accordingly.2895 *2896 *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,2897 *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:2898 *		  Use with ENCAP_L3 flags to further specify the tunnel type.2899 *2900 *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):2901 *		  Use with ENCAP_L3/L4 flags to further specify the tunnel2902 *		  type; *len* is the length of the inner MAC header.2903 *2904 *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:2905 *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the2906 *		  L2 type as Ethernet.2907 *2908 *		* **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,2909 *		  **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:2910 *		  Indicate the new IP header version after decapsulating the outer2911 *		  IP header. Used when the inner and outer IP versions are different.2912 *2913 * 		A call to this helper is susceptible to change the underlying2914 * 		packet buffer. Therefore, at load time, all checks on pointers2915 * 		previously done by the verifier are invalidated and must be2916 * 		performed again, if the helper is used in combination with2917 * 		direct packet access.2918 * 	Return2919 * 		0 on success, or a negative error in case of failure.2920 *2921 * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)2922 * 	Description2923 * 		Redirect the packet to the endpoint referenced by *map* at2924 * 		index *key*. Depending on its type, this *map* can contain2925 * 		references to net devices (for forwarding packets through other2926 * 		ports), or to CPUs (for redirecting XDP frames to another CPU;2927 * 		but this is only implemented for native XDP (with driver2928 * 		support) as of this writing).2929 *2930 * 		The lower two bits of *flags* are used as the return code if2931 * 		the map lookup fails. This is so that the return value can be2932 * 		one of the XDP program return codes up to **XDP_TX**, as chosen2933 * 		by the caller. The higher bits of *flags* can be set to2934 * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.2935 *2936 * 		With BPF_F_BROADCAST the packet will be broadcasted to all the2937 * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress2938 * 		interface will be excluded when do broadcasting.2939 *2940 * 		See also **bpf_redirect**\ (), which only supports redirecting2941 * 		to an ifindex, but doesn't require a map to do so.2942 * 	Return2943 * 		**XDP_REDIRECT** on success, or the value of the two lower bits2944 * 		of the *flags* argument on error.2945 *2946 * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)2947 * 	Description2948 * 		Redirect the packet to the socket referenced by *map* (of type2949 * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and2950 * 		egress interfaces can be used for redirection. The2951 * 		**BPF_F_INGRESS** value in *flags* is used to make the2952 * 		distinction (ingress path is selected if the flag is present,2953 * 		egress path otherwise). This is the only flag supported for now.2954 * 	Return2955 * 		**SK_PASS** on success, or **SK_DROP** on error.2956 *2957 * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)2958 * 	Description2959 * 		Add an entry to, or update a *map* referencing sockets. The2960 * 		*skops* is used as a new value for the entry associated to2961 * 		*key*. *flags* is one of:2962 *2963 * 		**BPF_NOEXIST**2964 * 			The entry for *key* must not exist in the map.2965 * 		**BPF_EXIST**2966 * 			The entry for *key* must already exist in the map.2967 * 		**BPF_ANY**2968 * 			No condition on the existence of the entry for *key*.2969 *2970 * 		If the *map* has eBPF programs (parser and verdict), those will2971 * 		be inherited by the socket being added. If the socket is2972 * 		already attached to eBPF programs, this results in an error.2973 * 	Return2974 * 		0 on success, or a negative error in case of failure.2975 *2976 * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)2977 * 	Description2978 * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by2979 * 		*delta* (which can be positive or negative). Note that this2980 * 		operation modifies the address stored in *xdp_md*\ **->data**,2981 * 		so the latter must be loaded only after the helper has been2982 * 		called.2983 *2984 * 		The use of *xdp_md*\ **->data_meta** is optional and programs2985 * 		are not required to use it. The rationale is that when the2986 * 		packet is processed with XDP (e.g. as DoS filter), it is2987 * 		possible to push further meta data along with it before passing2988 * 		to the stack, and to give the guarantee that an ingress eBPF2989 * 		program attached as a TC classifier on the same device can pick2990 * 		this up for further post-processing. Since TC works with socket2991 * 		buffers, it remains possible to set from XDP the **mark** or2992 * 		**priority** pointers, or other pointers for the socket buffer.2993 * 		Having this scratch space generic and programmable allows for2994 * 		more flexibility as the user is free to store whatever meta2995 * 		data they need.2996 *2997 * 		A call to this helper is susceptible to change the underlying2998 * 		packet buffer. Therefore, at load time, all checks on pointers2999 * 		previously done by the verifier are invalidated and must be3000 * 		performed again, if the helper is used in combination with3001 * 		direct packet access.3002 * 	Return3003 * 		0 on success, or a negative error in case of failure.3004 *3005 * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)3006 * 	Description3007 * 		Read the value of a perf event counter, and store it into *buf*3008 * 		of size *buf_size*. This helper relies on a *map* of type3009 * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event3010 * 		counter is selected when *map* is updated with perf event file3011 * 		descriptors. The *map* is an array whose size is the number of3012 * 		available CPUs, and each cell contains a value relative to one3013 * 		CPU. The value to retrieve is indicated by *flags*, that3014 * 		contains the index of the CPU to look up, masked with3015 * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to3016 * 		**BPF_F_CURRENT_CPU** to indicate that the value for the3017 * 		current CPU should be retrieved.3018 *3019 * 		This helper behaves in a way close to3020 * 		**bpf_perf_event_read**\ () helper, save that instead of3021 * 		just returning the value observed, it fills the *buf*3022 * 		structure. This allows for additional data to be retrieved: in3023 * 		particular, the enabled and running times (in *buf*\3024 * 		**->enabled** and *buf*\ **->running**, respectively) are3025 * 		copied. In general, **bpf_perf_event_read_value**\ () is3026 * 		recommended over **bpf_perf_event_read**\ (), which has some3027 * 		ABI issues and provides fewer functionalities.3028 *3029 * 		These values are interesting, because hardware PMU (Performance3030 * 		Monitoring Unit) counters are limited resources. When there are3031 * 		more PMU based perf events opened than available counters,3032 * 		kernel will multiplex these events so each event gets certain3033 * 		percentage (but not all) of the PMU time. In case that3034 * 		multiplexing happens, the number of samples or counter value3035 * 		will not reflect the case compared to when no multiplexing3036 * 		occurs. This makes comparison between different runs difficult.3037 * 		Typically, the counter value should be normalized before3038 * 		comparing to other experiments. The usual normalization is done3039 * 		as follows.3040 *3041 * 		::3042 *3043 * 			normalized_counter = counter * t_enabled / t_running3044 *3045 * 		Where t_enabled is the time enabled for event and t_running is3046 * 		the time running for event since last normalization. The3047 * 		enabled and running times are accumulated since the perf event3048 * 		open. To achieve scaling factor between two invocations of an3049 * 		eBPF program, users can use CPU id as the key (which is3050 * 		typical for perf array usage model) to remember the previous3051 * 		value and do the calculation inside the eBPF program.3052 * 	Return3053 * 		0 on success, or a negative error in case of failure.3054 *3055 * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)3056 * 	Description3057 * 		For an eBPF program attached to a perf event, retrieve the3058 * 		value of the event counter associated to *ctx* and store it in3059 * 		the structure pointed by *buf* and of size *buf_size*. Enabled3060 * 		and running times are also stored in the structure (see3061 * 		description of helper **bpf_perf_event_read_value**\ () for3062 * 		more details).3063 * 	Return3064 * 		0 on success, or a negative error in case of failure.3065 *3066 * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)3067 * 	Description3068 * 		Emulate a call to **getsockopt()** on the socket associated to3069 * 		*bpf_socket*, which must be a full socket. The *level* at3070 * 		which the option resides and the name *optname* of the option3071 * 		must be specified, see **getsockopt(2)** for more information.3072 * 		The retrieved value is stored in the structure pointed by3073 * 		*opval* and of length *optlen*.3074 *3075 * 		*bpf_socket* should be one of the following:3076 *3077 * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.3078 *		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**,3079 *		  **BPF_CGROUP_INET6_CONNECT** and **BPF_CGROUP_UNIX_CONNECT**.3080 *3081 * 		This helper actually implements a subset of **getsockopt()**.3082 * 		It supports the same set of *optname*\ s that is supported by3083 * 		the **bpf_setsockopt**\ () helper.  The exceptions are3084 * 		**TCP_BPF_*** is **bpf_setsockopt**\ () only and3085 * 		**TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.3086 * 	Return3087 * 		0 on success, or a negative error in case of failure.3088 *3089 * long bpf_override_return(struct pt_regs *regs, u64 rc)3090 * 	Description3091 * 		Used for error injection, this helper uses kprobes to override3092 * 		the return value of the probed function, and to set it to *rc*.3093 * 		The first argument is the context *regs* on which the kprobe3094 * 		works.3095 *3096 * 		This helper works by setting the PC (program counter)3097 * 		to an override function which is run in place of the original3098 * 		probed function. This means the probed function is not run at3099 * 		all. The replacement function just returns with the required3100 * 		value.3101 *3102 * 		This helper has security implications, and thus is subject to3103 * 		restrictions. It is only available if the kernel was compiled3104 * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration3105 * 		option, and in this case it only works on functions tagged with3106 * 		**ALLOW_ERROR_INJECTION** in the kernel code.3107 *3108 * 		Also, the helper is only available for the architectures having3109 * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,3110 * 		x86 architecture is the only one to support this feature.3111 * 	Return3112 * 		03113 *3114 * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)3115 * 	Description3116 * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field3117 * 		for the full TCP socket associated to *bpf_sock_ops* to3118 * 		*argval*.3119 *3120 * 		The primary use of this field is to determine if there should3121 * 		be calls to eBPF programs of type3122 * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP3123 * 		code. A program of the same type can change its value, per3124 * 		connection and as necessary, when the connection is3125 * 		established. This field is directly accessible for reading, but3126 * 		this helper must be used for updates in order to return an3127 * 		error if an eBPF program tries to set a callback that is not3128 * 		supported in the current kernel.3129 *3130 * 		*argval* is a flag array which can combine these flags:3131 *3132 * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)3133 * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)3134 * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)3135 * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)3136 *3137 * 		Therefore, this function can be used to clear a callback flag by3138 * 		setting the appropriate bit to zero. e.g. to disable the RTO3139 * 		callback:3140 *3141 * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**3142 * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**3143 *3144 * 		Here are some examples of where one could call such eBPF3145 * 		program:3146 *3147 * 		* When RTO fires.3148 * 		* When a packet is retransmitted.3149 * 		* When the connection terminates.3150 * 		* When a packet is sent.3151 * 		* When a packet is received.3152 * 	Return3153 * 		Code **-EINVAL** if the socket is not a full TCP socket;3154 * 		otherwise, a positive number containing the bits that could not3155 * 		be set is returned (which comes down to 0 if all bits were set3156 * 		as required).3157 *3158 * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)3159 * 	Description3160 * 		This helper is used in programs implementing policies at the3161 * 		socket level. If the message *msg* is allowed to pass (i.e. if3162 * 		the verdict eBPF program returns **SK_PASS**), redirect it to3163 * 		the socket referenced by *map* (of type3164 * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and3165 * 		egress interfaces can be used for redirection. The3166 * 		**BPF_F_INGRESS** value in *flags* is used to make the3167 * 		distinction (ingress path is selected if the flag is present,3168 * 		egress path otherwise). This is the only flag supported for now.3169 * 	Return3170 * 		**SK_PASS** on success, or **SK_DROP** on error.3171 *3172 * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)3173 * 	Description3174 * 		For socket policies, apply the verdict of the eBPF program to3175 * 		the next *bytes* (number of bytes) of message *msg*.3176 *3177 * 		For example, this helper can be used in the following cases:3178 *3179 * 		* A single **sendmsg**\ () or **sendfile**\ () system call3180 * 		  contains multiple logical messages that the eBPF program is3181 * 		  supposed to read and for which it should apply a verdict.3182 * 		* An eBPF program only cares to read the first *bytes* of a3183 * 		  *msg*. If the message has a large payload, then setting up3184 * 		  and calling the eBPF program repeatedly for all bytes, even3185 * 		  though the verdict is already known, would create unnecessary3186 * 		  overhead.3187 *3188 * 		When called from within an eBPF program, the helper sets a3189 * 		counter internal to the BPF infrastructure, that is used to3190 * 		apply the last verdict to the next *bytes*. If *bytes* is3191 * 		smaller than the current data being processed from a3192 * 		**sendmsg**\ () or **sendfile**\ () system call, the first3193 * 		*bytes* will be sent and the eBPF program will be re-run with3194 * 		the pointer for start of data pointing to byte number *bytes*3195 * 		**+ 1**. If *bytes* is larger than the current data being3196 * 		processed, then the eBPF verdict will be applied to multiple3197 * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are3198 * 		consumed.3199 *3200 * 		Note that if a socket closes with the internal counter holding3201 * 		a non-zero value, this is not a problem because data is not3202 * 		being buffered for *bytes* and is sent as it is received.3203 * 	Return3204 * 		03205 *3206 * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)3207 * 	Description3208 * 		For socket policies, prevent the execution of the verdict eBPF3209 * 		program for message *msg* until *bytes* (byte number) have been3210 * 		accumulated.3211 *3212 * 		This can be used when one needs a specific number of bytes3213 * 		before a verdict can be assigned, even if the data spans3214 * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme3215 * 		case would be a user calling **sendmsg**\ () repeatedly with3216 * 		1-byte long message segments. Obviously, this is bad for3217 * 		performance, but it is still valid. If the eBPF program needs3218 * 		*bytes* bytes to validate a header, this helper can be used to3219 * 		prevent the eBPF program to be called again until *bytes* have3220 * 		been accumulated.3221 * 	Return3222 * 		03223 *3224 * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)3225 * 	Description3226 * 		For socket policies, pull in non-linear data from user space3227 * 		for *msg* and set pointers *msg*\ **->data** and *msg*\3228 * 		**->data_end** to *start* and *end* bytes offsets into *msg*,3229 * 		respectively.3230 *3231 * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a3232 * 		*msg* it can only parse data that the (**data**, **data_end**)3233 * 		pointers have already consumed. For **sendmsg**\ () hooks this3234 * 		is likely the first scatterlist element. But for calls relying3235 * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will3236 * 		be the range (**0**, **0**) because the data is shared with3237 * 		user space and by default the objective is to avoid allowing3238 * 		user space to modify data while (or after) eBPF verdict is3239 * 		being decided. This helper can be used to pull in data and to3240 * 		set the start and end pointer to given values. Data will be3241 * 		copied if necessary (i.e. if data was not linear and if start3242 * 		and end pointers do not point to the same chunk).3243 *3244 * 		A call to this helper is susceptible to change the underlying3245 * 		packet buffer. Therefore, at load time, all checks on pointers3246 * 		previously done by the verifier are invalidated and must be3247 * 		performed again, if the helper is used in combination with3248 * 		direct packet access.3249 *3250 * 		All values for *flags* are reserved for future usage, and must3251 * 		be left at zero.3252 * 	Return3253 * 		0 on success, or a negative error in case of failure.3254 *3255 * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)3256 * 	Description3257 * 		Bind the socket associated to *ctx* to the address pointed by3258 * 		*addr*, of length *addr_len*. This allows for making outgoing3259 * 		connection from the desired IP address, which can be useful for3260 * 		example when all processes inside a cgroup should use one3261 * 		single IP address on a host that has multiple IP configured.3262 *3263 * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The3264 * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or3265 * 		**AF_INET6**). It's advised to pass zero port (**sin_port**3266 * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like3267 * 		behavior and lets the kernel efficiently pick up an unused3268 * 		port as long as 4-tuple is unique. Passing non-zero port might3269 * 		lead to degraded performance.3270 * 	Return3271 * 		0 on success, or a negative error in case of failure.3272 *3273 * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)3274 * 	Description3275 * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is3276 * 		possible to both shrink and grow the packet tail.3277 * 		Shrink done via *delta* being a negative integer.3278 *3279 * 		A call to this helper is susceptible to change the underlying3280 * 		packet buffer. Therefore, at load time, all checks on pointers3281 * 		previously done by the verifier are invalidated and must be3282 * 		performed again, if the helper is used in combination with3283 * 		direct packet access.3284 * 	Return3285 * 		0 on success, or a negative error in case of failure.3286 *3287 * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)3288 * 	Description3289 * 		Retrieve the XFRM state (IP transform framework, see also3290 * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.3291 *3292 * 		The retrieved value is stored in the **struct bpf_xfrm_state**3293 * 		pointed by *xfrm_state* and of length *size*.3294 *3295 * 		All values for *flags* are reserved for future usage, and must3296 * 		be left at zero.3297 *3298 * 		This helper is available only if the kernel was compiled with3299 * 		**CONFIG_XFRM** configuration option.3300 * 	Return3301 * 		0 on success, or a negative error in case of failure.3302 *3303 * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)3304 * 	Description3305 * 		Return a user or a kernel stack in bpf program provided buffer.3306 * 		To achieve this, the helper needs *ctx*, which is a pointer3307 * 		to the context on which the tracing program is executed.3308 * 		To store the stacktrace, the bpf program provides *buf* with3309 * 		a nonnegative *size*.3310 *3311 * 		The last argument, *flags*, holds the number of stack frames to3312 * 		skip (from 0 to 255), masked with3313 * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set3314 * 		the following flags:3315 *3316 * 		**BPF_F_USER_STACK**3317 * 			Collect a user space stack instead of a kernel stack.3318 * 		**BPF_F_USER_BUILD_ID**3319 * 			Collect (build_id, file_offset) instead of ips for user3320 * 			stack, only valid if **BPF_F_USER_STACK** is also3321 * 			specified.3322 *3323 * 			*file_offset* is an offset relative to the beginning3324 * 			of the executable or shared object file backing the vma3325 * 			which the *ip* falls in. It is *not* an offset relative3326 * 			to that object's base address. Accordingly, it must be3327 * 			adjusted by adding (sh_addr - sh_offset), where3328 * 			sh_{addr,offset} correspond to the executable section3329 * 			containing *file_offset* in the object, for comparisons3330 * 			to symbols' st_value to be valid.3331 *3332 * 		**bpf_get_stack**\ () can collect up to3333 * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject3334 * 		to sufficient large buffer size. Note that3335 * 		this limit can be controlled with the **sysctl** program, and3336 * 		that it should be manually increased in order to profile long3337 * 		user stacks (such as stacks for Java programs). To do so, use:3338 *3339 * 		::3340 *3341 * 			# sysctl kernel.perf_event_max_stack=<new value>3342 * 	Return3343 * 		The non-negative copied *buf* length equal to or less than3344 * 		*size* on success, or a negative error in case of failure.3345 *3346 * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)3347 * 	Description3348 * 		This helper is similar to **bpf_skb_load_bytes**\ () in that3349 * 		it provides an easy way to load *len* bytes from *offset*3350 * 		from the packet associated to *skb*, into the buffer pointed3351 * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that3352 * 		a fifth argument *start_header* exists in order to select a3353 * 		base offset to start from. *start_header* can be one of:3354 *3355 * 		**BPF_HDR_START_MAC**3356 * 			Base offset to load data from is *skb*'s mac header.3357 * 		**BPF_HDR_START_NET**3358 * 			Base offset to load data from is *skb*'s network header.3359 *3360 * 		In general, "direct packet access" is the preferred method to3361 * 		access packet data, however, this helper is in particular useful3362 * 		in socket filters where *skb*\ **->data** does not always point3363 * 		to the start of the mac header and where "direct packet access"3364 * 		is not available.3365 * 	Return3366 * 		0 on success, or a negative error in case of failure.3367 *3368 * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)3369 *	Description3370 *		Do FIB lookup in kernel tables using parameters in *params*.3371 *		If lookup is successful and result shows packet is to be3372 *		forwarded, the neighbor tables are searched for the nexthop.3373 *		If successful (ie., FIB lookup shows forwarding and nexthop3374 *		is resolved), the nexthop address is returned in ipv4_dst3375 *		or ipv6_dst based on family, smac is set to mac address of3376 *		egress device, dmac is set to nexthop mac address, rt_metric3377 *		is set to metric from route (IPv4/IPv6 only), and ifindex3378 *		is set to the device index of the nexthop from the FIB lookup.3379 *3380 *		*plen* argument is the size of the passed in struct.3381 *		*flags* argument can be a combination of one or more of the3382 *		following values:3383 *3384 *		**BPF_FIB_LOOKUP_DIRECT**3385 *			Do a direct table lookup vs full lookup using FIB3386 *			rules.3387 *		**BPF_FIB_LOOKUP_TBID**3388 *			Used with BPF_FIB_LOOKUP_DIRECT.3389 *			Use the routing table ID present in *params*->tbid3390 *			for the fib lookup.3391 *		**BPF_FIB_LOOKUP_OUTPUT**3392 *			Perform lookup from an egress perspective (default is3393 *			ingress).3394 *		**BPF_FIB_LOOKUP_SKIP_NEIGH**3395 *			Skip the neighbour table lookup. *params*->dmac3396 *			and *params*->smac will not be set as output. A common3397 *			use case is to call **bpf_redirect_neigh**\ () after3398 *			doing **bpf_fib_lookup**\ ().3399 *		**BPF_FIB_LOOKUP_SRC**3400 *			Derive and set source IP addr in *params*->ipv{4,6}_src3401 *			for the nexthop. If the src addr cannot be derived,3402 *			**BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this3403 *			case, *params*->dmac and *params*->smac are not set either.3404 *		**BPF_FIB_LOOKUP_MARK**3405 *			Use the mark present in *params*->mark for the fib lookup.3406 *			This option should not be used with BPF_FIB_LOOKUP_DIRECT,3407 *			as it only has meaning for full lookups.3408 *3409 *		*ctx* is either **struct xdp_md** for XDP programs or3410 *		**struct sk_buff** tc cls_act programs.3411 *	Return3412 *		* < 0 if any input argument is invalid3413 *		*   0 on success (packet is forwarded, nexthop neighbor exists)3414 *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the3415 *		  packet is not forwarded or needs assist from full stack3416 *3417 *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU3418 *		was exceeded and output params->mtu_result contains the MTU.3419 *3420 * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)3421 *	Description3422 *		Add an entry to, or update a sockhash *map* referencing sockets.3423 *		The *skops* is used as a new value for the entry associated to3424 *		*key*. *flags* is one of:3425 *3426 *		**BPF_NOEXIST**3427 *			The entry for *key* must not exist in the map.3428 *		**BPF_EXIST**3429 *			The entry for *key* must already exist in the map.3430 *		**BPF_ANY**3431 *			No condition on the existence of the entry for *key*.3432 *3433 *		If the *map* has eBPF programs (parser and verdict), those will3434 *		be inherited by the socket being added. If the socket is3435 *		already attached to eBPF programs, this results in an error.3436 *	Return3437 *		0 on success, or a negative error in case of failure.3438 *3439 * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)3440 *	Description3441 *		This helper is used in programs implementing policies at the3442 *		socket level. If the message *msg* is allowed to pass (i.e. if3443 *		the verdict eBPF program returns **SK_PASS**), redirect it to3444 *		the socket referenced by *map* (of type3445 *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and3446 *		egress interfaces can be used for redirection. The3447 *		**BPF_F_INGRESS** value in *flags* is used to make the3448 *		distinction (ingress path is selected if the flag is present,3449 *		egress path otherwise). This is the only flag supported for now.3450 *	Return3451 *		**SK_PASS** on success, or **SK_DROP** on error.3452 *3453 * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)3454 *	Description3455 *		This helper is used in programs implementing policies at the3456 *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.3457 *		if the verdict eBPF program returns **SK_PASS**), redirect it3458 *		to the socket referenced by *map* (of type3459 *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and3460 *		egress interfaces can be used for redirection. The3461 *		**BPF_F_INGRESS** value in *flags* is used to make the3462 *		distinction (ingress path is selected if the flag is present,3463 *		egress otherwise). This is the only flag supported for now.3464 *	Return3465 *		**SK_PASS** on success, or **SK_DROP** on error.3466 *3467 * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)3468 *	Description3469 *		Encapsulate the packet associated to *skb* within a Layer 33470 *		protocol header. This header is provided in the buffer at3471 *		address *hdr*, with *len* its size in bytes. *type* indicates3472 *		the protocol of the header and can be one of:3473 *3474 *		**BPF_LWT_ENCAP_SEG6**3475 *			IPv6 encapsulation with Segment Routing Header3476 *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,3477 *			the IPv6 header is computed by the kernel.3478 *		**BPF_LWT_ENCAP_SEG6_INLINE**3479 *			Only works if *skb* contains an IPv6 packet. Insert a3480 *			Segment Routing Header (**struct ipv6_sr_hdr**) inside3481 *			the IPv6 header.3482 *		**BPF_LWT_ENCAP_IP**3483 *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header3484 *			must be IPv4 or IPv6, followed by zero or more3485 *			additional headers, up to **LWT_BPF_MAX_HEADROOM**3486 *			total bytes in all prepended headers. Please note that3487 *			if **skb_is_gso**\ (*skb*) is true, no more than two3488 *			headers can be prepended, and the inner header, if3489 *			present, should be either GRE or UDP/GUE.3490 *3491 *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs3492 *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can3493 *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and3494 *		**BPF_PROG_TYPE_LWT_XMIT**.3495 *3496 * 		A call to this helper is susceptible to change the underlying3497 * 		packet buffer. Therefore, at load time, all checks on pointers3498 * 		previously done by the verifier are invalidated and must be3499 * 		performed again, if the helper is used in combination with3500 * 		direct packet access.3501 *	Return3502 * 		0 on success, or a negative error in case of failure.3503 *3504 * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)3505 *	Description3506 *		Store *len* bytes from address *from* into the packet3507 *		associated to *skb*, at *offset*. Only the flags, tag and TLVs3508 *		inside the outermost IPv6 Segment Routing Header can be3509 *		modified through this helper.3510 *3511 * 		A call to this helper is susceptible to change the underlying3512 * 		packet buffer. Therefore, at load time, all checks on pointers3513 * 		previously done by the verifier are invalidated and must be3514 * 		performed again, if the helper is used in combination with3515 * 		direct packet access.3516 *	Return3517 * 		0 on success, or a negative error in case of failure.3518 *3519 * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)3520 *	Description3521 *		Adjust the size allocated to TLVs in the outermost IPv63522 *		Segment Routing Header contained in the packet associated to3523 *		*skb*, at position *offset* by *delta* bytes. Only offsets3524 *		after the segments are accepted. *delta* can be as well3525 *		positive (growing) as negative (shrinking).3526 *3527 * 		A call to this helper is susceptible to change the underlying3528 * 		packet buffer. Therefore, at load time, all checks on pointers3529 * 		previously done by the verifier are invalidated and must be3530 * 		performed again, if the helper is used in combination with3531 * 		direct packet access.3532 *	Return3533 * 		0 on success, or a negative error in case of failure.3534 *3535 * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)3536 *	Description3537 *		Apply an IPv6 Segment Routing action of type *action* to the3538 *		packet associated to *skb*. Each action takes a parameter3539 *		contained at address *param*, and of length *param_len* bytes.3540 *		*action* can be one of:3541 *3542 *		**SEG6_LOCAL_ACTION_END_X**3543 *			End.X action: Endpoint with Layer-3 cross-connect.3544 *			Type of *param*: **struct in6_addr**.3545 *		**SEG6_LOCAL_ACTION_END_T**3546 *			End.T action: Endpoint with specific IPv6 table lookup.3547 *			Type of *param*: **int**.3548 *		**SEG6_LOCAL_ACTION_END_B6**3549 *			End.B6 action: Endpoint bound to an SRv6 policy.3550 *			Type of *param*: **struct ipv6_sr_hdr**.3551 *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**3552 *			End.B6.Encap action: Endpoint bound to an SRv63553 *			encapsulation policy.3554 *			Type of *param*: **struct ipv6_sr_hdr**.3555 *3556 * 		A call to this helper is susceptible to change the underlying3557 * 		packet buffer. Therefore, at load time, all checks on pointers3558 * 		previously done by the verifier are invalidated and must be3559 * 		performed again, if the helper is used in combination with3560 * 		direct packet access.3561 *	Return3562 * 		0 on success, or a negative error in case of failure.3563 *3564 * long bpf_rc_repeat(void *ctx)3565 *	Description3566 *		This helper is used in programs implementing IR decoding, to3567 *		report a successfully decoded repeat key message. This delays3568 *		the generation of a key up event for previously generated3569 *		key down event.3570 *3571 *		Some IR protocols like NEC have a special IR message for3572 *		repeating last button, for when a button is held down.3573 *3574 *		The *ctx* should point to the lirc sample as passed into3575 *		the program.3576 *3577 *		This helper is only available is the kernel was compiled with3578 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to3579 *		"**y**".3580 *	Return3581 *		03582 *3583 * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)3584 *	Description3585 *		This helper is used in programs implementing IR decoding, to3586 *		report a successfully decoded key press with *scancode*,3587 *		*toggle* value in the given *protocol*. The scancode will be3588 *		translated to a keycode using the rc keymap, and reported as3589 *		an input key down event. After a period a key up event is3590 *		generated. This period can be extended by calling either3591 *		**bpf_rc_keydown**\ () again with the same values, or calling3592 *		**bpf_rc_repeat**\ ().3593 *3594 *		Some protocols include a toggle bit, in case the button was3595 *		released and pressed again between consecutive scancodes.3596 *3597 *		The *ctx* should point to the lirc sample as passed into3598 *		the program.3599 *3600 *		The *protocol* is the decoded protocol number (see3601 *		**enum rc_proto** for some predefined values).3602 *3603 *		This helper is only available is the kernel was compiled with3604 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to3605 *		"**y**".3606 *	Return3607 *		03608 *3609 * u64 bpf_skb_cgroup_id(struct sk_buff *skb)3610 * 	Description3611 * 		Return the cgroup v2 id of the socket associated with the *skb*.3612 * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()3613 * 		helper for cgroup v1 by providing a tag resp. identifier that3614 * 		can be matched on or used for map lookups e.g. to implement3615 * 		policy. The cgroup v2 id of a given path in the hierarchy is3616 * 		exposed in user space through the f_handle API in order to get3617 * 		to the same 64-bit id.3618 *3619 * 		This helper can be used on TC egress path, but not on ingress,3620 * 		and is available only if the kernel was compiled with the3621 * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.3622 * 	Return3623 * 		The id is returned or 0 in case the id could not be retrieved.3624 *3625 * u64 bpf_get_current_cgroup_id(void)3626 * 	Description3627 * 		Get the current cgroup id based on the cgroup within which3628 * 		the current task is running.3629 * 	Return3630 * 		A 64-bit integer containing the current cgroup id based3631 * 		on the cgroup within which the current task is running.3632 *3633 * void *bpf_get_local_storage(void *map, u64 flags)3634 *	Description3635 *		Get the pointer to the local storage area.3636 *		The type and the size of the local storage is defined3637 *		by the *map* argument.3638 *		The *flags* meaning is specific for each map type,3639 *		and has to be 0 for cgroup local storage.3640 *3641 *		Depending on the BPF program type, a local storage area3642 *		can be shared between multiple instances of the BPF program,3643 *		running simultaneously.3644 *3645 *		A user should care about the synchronization by himself.3646 *		For example, by using the **BPF_ATOMIC** instructions to alter3647 *		the shared data.3648 *	Return3649 *		A pointer to the local storage area.3650 *3651 * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)3652 *	Description3653 *		Select a **SO_REUSEPORT** socket from a3654 *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.3655 *		It checks the selected socket is matching the incoming3656 *		request in the socket buffer.3657 *	Return3658 *		0 on success, or a negative error in case of failure.3659 *3660 * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)3661 *	Description3662 *		Return id of cgroup v2 that is ancestor of cgroup associated3663 *		with the *skb* at the *ancestor_level*.  The root cgroup is at3664 *		*ancestor_level* zero and each step down the hierarchy3665 *		increments the level. If *ancestor_level* == level of cgroup3666 *		associated with *skb*, then return value will be same as that3667 *		of **bpf_skb_cgroup_id**\ ().3668 *3669 *		The helper is useful to implement policies based on cgroups3670 *		that are upper in hierarchy than immediate cgroup associated3671 *		with *skb*.3672 *3673 *		The format of returned id and helper limitations are same as in3674 *		**bpf_skb_cgroup_id**\ ().3675 *	Return3676 *		The id is returned or 0 in case the id could not be retrieved.3677 *3678 * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)3679 *	Description3680 *		Look for TCP socket matching *tuple*, optionally in a child3681 *		network namespace *netns*. The return value must be checked,3682 *		and if non-**NULL**, released via **bpf_sk_release**\ ().3683 *3684 *		The *ctx* should point to the context of the program, such as3685 *		the skb or socket (depending on the hook in use). This is used3686 *		to determine the base network namespace for the lookup.3687 *3688 *		*tuple_size* must be one of:3689 *3690 *		**sizeof**\ (*tuple*\ **->ipv4**)3691 *			Look for an IPv4 socket.3692 *		**sizeof**\ (*tuple*\ **->ipv6**)3693 *			Look for an IPv6 socket.3694 *3695 *		If the *netns* is a negative signed 32-bit integer, then the3696 *		socket lookup table in the netns associated with the *ctx*3697 *		will be used. For the TC hooks, this is the netns of the device3698 *		in the skb. For socket hooks, this is the netns of the socket.3699 *		If *netns* is any other signed 32-bit value greater than or3700 *		equal to zero then it specifies the ID of the netns relative to3701 *		the netns associated with the *ctx*. *netns* values beyond the3702 *		range of 32-bit integers are reserved for future use.3703 *3704 *		All values for *flags* are reserved for future usage, and must3705 *		be left at zero.3706 *3707 *		This helper is available only if the kernel was compiled with3708 *		**CONFIG_NET** configuration option.3709 *	Return3710 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.3711 *		For sockets with reuseport option, the **struct bpf_sock**3712 *		result is from *reuse*\ **->socks**\ [] using the hash of the3713 *		tuple.3714 *3715 * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)3716 *	Description3717 *		Look for UDP socket matching *tuple*, optionally in a child3718 *		network namespace *netns*. The return value must be checked,3719 *		and if non-**NULL**, released via **bpf_sk_release**\ ().3720 *3721 *		The *ctx* should point to the context of the program, such as3722 *		the skb or socket (depending on the hook in use). This is used3723 *		to determine the base network namespace for the lookup.3724 *3725 *		*tuple_size* must be one of:3726 *3727 *		**sizeof**\ (*tuple*\ **->ipv4**)3728 *			Look for an IPv4 socket.3729 *		**sizeof**\ (*tuple*\ **->ipv6**)3730 *			Look for an IPv6 socket.3731 *3732 *		If the *netns* is a negative signed 32-bit integer, then the3733 *		socket lookup table in the netns associated with the *ctx*3734 *		will be used. For the TC hooks, this is the netns of the device3735 *		in the skb. For socket hooks, this is the netns of the socket.3736 *		If *netns* is any other signed 32-bit value greater than or3737 *		equal to zero then it specifies the ID of the netns relative to3738 *		the netns associated with the *ctx*. *netns* values beyond the3739 *		range of 32-bit integers are reserved for future use.3740 *3741 *		All values for *flags* are reserved for future usage, and must3742 *		be left at zero.3743 *3744 *		This helper is available only if the kernel was compiled with3745 *		**CONFIG_NET** configuration option.3746 *	Return3747 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.3748 *		For sockets with reuseport option, the **struct bpf_sock**3749 *		result is from *reuse*\ **->socks**\ [] using the hash of the3750 *		tuple.3751 *3752 * long bpf_sk_release(void *sock)3753 *	Description3754 *		Release the reference held by *sock*. *sock* must be a3755 *		non-**NULL** pointer that was returned from3756 *		**bpf_sk_lookup_xxx**\ ().3757 *	Return3758 *		0 on success, or a negative error in case of failure.3759 *3760 * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)3761 * 	Description3762 * 		Push an element *value* in *map*. *flags* is one of:3763 *3764 * 		**BPF_EXIST**3765 * 			If the queue/stack is full, the oldest element is3766 * 			removed to make room for this.3767 * 	Return3768 * 		0 on success, or a negative error in case of failure.3769 *3770 * long bpf_map_pop_elem(struct bpf_map *map, void *value)3771 * 	Description3772 * 		Pop an element from *map*.3773 * 	Return3774 * 		0 on success, or a negative error in case of failure.3775 *3776 * long bpf_map_peek_elem(struct bpf_map *map, void *value)3777 * 	Description3778 * 		Get an element from *map* without removing it.3779 * 	Return3780 * 		0 on success, or a negative error in case of failure.3781 *3782 * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)3783 *	Description3784 *		For socket policies, insert *len* bytes into *msg* at offset3785 *		*start*.3786 *3787 *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a3788 *		*msg* it may want to insert metadata or options into the *msg*.3789 *		This can later be read and used by any of the lower layer BPF3790 *		hooks.3791 *3792 *		This helper may fail if under memory pressure (a malloc3793 *		fails) in these cases BPF programs will get an appropriate3794 *		error and BPF programs will need to handle them.3795 *	Return3796 *		0 on success, or a negative error in case of failure.3797 *3798 * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)3799 *	Description3800 *		Will remove *len* bytes from a *msg* starting at byte *start*.3801 *		This may result in **ENOMEM** errors under certain situations if3802 *		an allocation and copy are required due to a full ring buffer.3803 *		However, the helper will try to avoid doing the allocation3804 *		if possible. Other errors can occur if input parameters are3805 *		invalid either due to *start* byte not being valid part of *msg*3806 *		payload and/or *pop* value being to large.3807 *	Return3808 *		0 on success, or a negative error in case of failure.3809 *3810 * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)3811 *	Description3812 *		This helper is used in programs implementing IR decoding, to3813 *		report a successfully decoded pointer movement.3814 *3815 *		The *ctx* should point to the lirc sample as passed into3816 *		the program.3817 *3818 *		This helper is only available is the kernel was compiled with3819 *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to3820 *		"**y**".3821 *	Return3822 *		03823 *3824 * long bpf_spin_lock(struct bpf_spin_lock *lock)3825 *	Description3826 *		Acquire a spinlock represented by the pointer *lock*, which is3827 *		stored as part of a value of a map. Taking the lock allows to3828 *		safely update the rest of the fields in that value. The3829 *		spinlock can (and must) later be released with a call to3830 *		**bpf_spin_unlock**\ (\ *lock*\ ).3831 *3832 *		Spinlocks in BPF programs come with a number of restrictions3833 *		and constraints:3834 *3835 *		* **bpf_spin_lock** objects are only allowed inside maps of3836 *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this3837 *		  list could be extended in the future).3838 *		* BTF description of the map is mandatory.3839 *		* The BPF program can take ONE lock at a time, since taking two3840 *		  or more could cause dead locks.3841 *		* Only one **struct bpf_spin_lock** is allowed per map element.3842 *		* When the lock is taken, calls (either BPF to BPF or helpers)3843 *		  are not allowed.3844 *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not3845 *		  allowed inside a spinlock-ed region.3846 *		* The BPF program MUST call **bpf_spin_unlock**\ () to release3847 *		  the lock, on all execution paths, before it returns.3848 *		* The BPF program can access **struct bpf_spin_lock** only via3849 *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()3850 *		  helpers. Loading or storing data into the **struct3851 *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.3852 *		* To use the **bpf_spin_lock**\ () helper, the BTF description3853 *		  of the map value must be a struct and have **struct3854 *		  bpf_spin_lock** *anyname*\ **;** field at the top level.3855 *		  Nested lock inside another struct is not allowed.3856 *		* The **struct bpf_spin_lock** *lock* field in a map value must3857 *		  be aligned on a multiple of 4 bytes in that value.3858 *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy3859 *		  the **bpf_spin_lock** field to user space.3860 *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from3861 *		  a BPF program, do not update the **bpf_spin_lock** field.3862 *		* **bpf_spin_lock** cannot be on the stack or inside a3863 *		  networking packet (it can only be inside of a map values).3864 *		* **bpf_spin_lock** is available to root only.3865 *		* Tracing programs and socket filter programs cannot use3866 *		  **bpf_spin_lock**\ () due to insufficient preemption checks3867 *		  (but this may change in the future).3868 *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.3869 *	Return3870 *		03871 *3872 * long bpf_spin_unlock(struct bpf_spin_lock *lock)3873 *	Description3874 *		Release the *lock* previously locked by a call to3875 *		**bpf_spin_lock**\ (\ *lock*\ ).3876 *	Return3877 *		03878 *3879 * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)3880 *	Description3881 *		This helper gets a **struct bpf_sock** pointer such3882 *		that all the fields in this **bpf_sock** can be accessed.3883 *	Return3884 *		A **struct bpf_sock** pointer on success, or **NULL** in3885 *		case of failure.3886 *3887 * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)3888 *	Description3889 *		This helper gets a **struct bpf_tcp_sock** pointer from a3890 *		**struct bpf_sock** pointer.3891 *	Return3892 *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in3893 *		case of failure.3894 *3895 * long bpf_skb_ecn_set_ce(struct sk_buff *skb)3896 *	Description3897 *		Set ECN (Explicit Congestion Notification) field of IP header3898 *		to **CE** (Congestion Encountered) if current value is **ECT**3899 *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv63900 *		and IPv4.3901 *	Return3902 *		1 if the **CE** flag is set (either by the current helper call3903 *		or because it was already present), 0 if it is not set.3904 *3905 * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)3906 *	Description3907 *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.3908 *		**bpf_sk_release**\ () is unnecessary and not allowed.3909 *	Return3910 *		A **struct bpf_sock** pointer on success, or **NULL** in3911 *		case of failure.3912 *3913 * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)3914 *	Description3915 *		Look for TCP socket matching *tuple*, optionally in a child3916 *		network namespace *netns*. The return value must be checked,3917 *		and if non-**NULL**, released via **bpf_sk_release**\ ().3918 *3919 *		This function is identical to **bpf_sk_lookup_tcp**\ (), except3920 *		that it also returns timewait or request sockets. Use3921 *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the3922 *		full structure.3923 *3924 *		This helper is available only if the kernel was compiled with3925 *		**CONFIG_NET** configuration option.3926 *	Return3927 *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.3928 *		For sockets with reuseport option, the **struct bpf_sock**3929 *		result is from *reuse*\ **->socks**\ [] using the hash of the3930 *		tuple.3931 *3932 * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)3933 * 	Description3934 * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for3935 * 		the listening socket in *sk*.3936 *3937 * 		*iph* points to the start of the IPv4 or IPv6 header, while3938 * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or3939 * 		**sizeof**\ (**struct ipv6hdr**).3940 *3941 * 		*th* points to the start of the TCP header, while *th_len*3942 *		contains the length of the TCP header (at least3943 *		**sizeof**\ (**struct tcphdr**)).3944 * 	Return3945 * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative3946 * 		error otherwise.3947 *3948 * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)3949 *	Description3950 *		Get name of sysctl in /proc/sys/ and copy it into provided by3951 *		program buffer *buf* of size *buf_len*.3952 *3953 *		The buffer is always NUL terminated, unless it's zero-sized.3954 *3955 *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is3956 *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name3957 *		only (e.g. "tcp_mem").3958 *	Return3959 *		Number of character copied (not including the trailing NUL).3960 *3961 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain3962 *		truncated name in this case).3963 *3964 * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)3965 *	Description3966 *		Get current value of sysctl as it is presented in /proc/sys3967 *		(incl. newline, etc), and copy it as a string into provided3968 *		by program buffer *buf* of size *buf_len*.3969 *3970 *		The whole value is copied, no matter what file position user3971 *		space issued e.g. sys_read at.3972 *3973 *		The buffer is always NUL terminated, unless it's zero-sized.3974 *	Return3975 *		Number of character copied (not including the trailing NUL).3976 *3977 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain3978 *		truncated name in this case).3979 *3980 *		**-EINVAL** if current value was unavailable, e.g. because3981 *		sysctl is uninitialized and read returns -EIO for it.3982 *3983 * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)3984 *	Description3985 *		Get new value being written by user space to sysctl (before3986 *		the actual write happens) and copy it as a string into3987 *		provided by program buffer *buf* of size *buf_len*.3988 *3989 *		User space may write new value at file position > 0.3990 *3991 *		The buffer is always NUL terminated, unless it's zero-sized.3992 *	Return3993 *		Number of character copied (not including the trailing NUL).3994 *3995 *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain3996 *		truncated name in this case).3997 *3998 *		**-EINVAL** if sysctl is being read.3999 *4000 * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)4001 *	Description4002 *		Override new value being written by user space to sysctl with4003 *		value provided by program in buffer *buf* of size *buf_len*.4004 *4005 *		*buf* should contain a string in same form as provided by user4006 *		space on sysctl write.4007 *4008 *		User space may write new value at file position > 0. To override4009 *		the whole sysctl value file position should be set to zero.4010 *	Return4011 *		0 on success.4012 *4013 *		**-E2BIG** if the *buf_len* is too big.4014 *4015 *		**-EINVAL** if sysctl is being read.4016 *4017 * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)4018 *	Description4019 *		Convert the initial part of the string from buffer *buf* of4020 *		size *buf_len* to a long integer according to the given base4021 *		and save the result in *res*.4022 *4023 *		The string may begin with an arbitrary amount of white space4024 *		(as determined by **isspace**\ (3)) followed by a single4025 *		optional '**-**' sign.4026 *4027 *		Five least significant bits of *flags* encode base, other bits4028 *		are currently unused.4029 *4030 *		Base must be either 8, 10, 16 or 0 to detect it automatically4031 *		similar to user space **strtol**\ (3).4032 *	Return4033 *		Number of characters consumed on success. Must be positive but4034 *		no more than *buf_len*.4035 *4036 *		**-EINVAL** if no valid digits were found or unsupported base4037 *		was provided.4038 *4039 *		**-ERANGE** if resulting value was out of range.4040 *4041 * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)4042 *	Description4043 *		Convert the initial part of the string from buffer *buf* of4044 *		size *buf_len* to an unsigned long integer according to the4045 *		given base and save the result in *res*.4046 *4047 *		The string may begin with an arbitrary amount of white space4048 *		(as determined by **isspace**\ (3)).4049 *4050 *		Five least significant bits of *flags* encode base, other bits4051 *		are currently unused.4052 *4053 *		Base must be either 8, 10, 16 or 0 to detect it automatically4054 *		similar to user space **strtoul**\ (3).4055 *	Return4056 *		Number of characters consumed on success. Must be positive but4057 *		no more than *buf_len*.4058 *4059 *		**-EINVAL** if no valid digits were found or unsupported base4060 *		was provided.4061 *4062 *		**-ERANGE** if resulting value was out of range.4063 *4064 * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)4065 *	Description4066 *		Get a bpf-local-storage from a *sk*.4067 *4068 *		Logically, it could be thought of getting the value from4069 *		a *map* with *sk* as the **key**.  From this4070 *		perspective,  the usage is not much different from4071 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this4072 *		helper enforces the key must be a full socket and the map must4073 *		be a **BPF_MAP_TYPE_SK_STORAGE** also.4074 *4075 *		Underneath, the value is stored locally at *sk* instead of4076 *		the *map*.  The *map* is used as the bpf-local-storage4077 *		"type". The bpf-local-storage "type" (i.e. the *map*) is4078 *		searched against all bpf-local-storages residing at *sk*.4079 *4080 *		*sk* is a kernel **struct sock** pointer for LSM program.4081 *		*sk* is a **struct bpf_sock** pointer for other program types.4082 *4083 *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be4084 *		used such that a new bpf-local-storage will be4085 *		created if one does not exist.  *value* can be used4086 *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify4087 *		the initial value of a bpf-local-storage.  If *value* is4088 *		**NULL**, the new bpf-local-storage will be zero initialized.4089 *	Return4090 *		A bpf-local-storage pointer is returned on success.4091 *4092 *		**NULL** if not found or there was an error in adding4093 *		a new bpf-local-storage.4094 *4095 * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)4096 *	Description4097 *		Delete a bpf-local-storage from a *sk*.4098 *	Return4099 *		0 on success.4100 *4101 *		**-ENOENT** if the bpf-local-storage cannot be found.4102 *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).4103 *4104 * long bpf_send_signal(u32 sig)4105 *	Description4106 *		Send signal *sig* to the process of the current task.4107 *		The signal may be delivered to any of this process's threads.4108 *	Return4109 *		0 on success or successfully queued.4110 *4111 *		**-EBUSY** if work queue under nmi is full.4112 *4113 *		**-EINVAL** if *sig* is invalid.4114 *4115 *		**-EPERM** if no permission to send the *sig*.4116 *4117 *		**-EAGAIN** if bpf program can try again.4118 *4119 * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)4120 *	Description4121 *		Try to issue a SYN cookie for the packet with corresponding4122 *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.4123 *4124 *		*iph* points to the start of the IPv4 or IPv6 header, while4125 *		*iph_len* contains **sizeof**\ (**struct iphdr**) or4126 *		**sizeof**\ (**struct ipv6hdr**).4127 *4128 *		*th* points to the start of the TCP header, while *th_len*4129 *		contains the length of the TCP header with options (at least4130 *		**sizeof**\ (**struct tcphdr**)).4131 *	Return4132 *		On success, lower 32 bits hold the generated SYN cookie in4133 *		followed by 16 bits which hold the MSS value for that cookie,4134 *		and the top 16 bits are unused.4135 *4136 *		On failure, the returned value is one of the following:4137 *4138 *		**-EINVAL** SYN cookie cannot be issued due to error4139 *4140 *		**-ENOENT** SYN cookie should not be issued (no SYN flood)4141 *4142 *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies4143 *4144 *		**-EPROTONOSUPPORT** IP packet version is not 4 or 64145 *4146 * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)4147 * 	Description4148 * 		Write raw *data* blob into a special BPF perf event held by4149 * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf4150 * 		event must have the following attributes: **PERF_SAMPLE_RAW**4151 * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and4152 * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.4153 *4154 * 		The *flags* are used to indicate the index in *map* for which4155 * 		the value must be put, masked with **BPF_F_INDEX_MASK**.4156 * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**4157 * 		to indicate that the index of the current CPU core should be4158 * 		used.4159 *4160 * 		The value to write, of *size*, is passed through eBPF stack and4161 * 		pointed by *data*.4162 *4163 * 		*ctx* is a pointer to in-kernel struct sk_buff.4164 *4165 * 		This helper is similar to **bpf_perf_event_output**\ () but4166 * 		restricted to raw_tracepoint bpf programs.4167 * 	Return4168 * 		0 on success, or a negative error in case of failure.4169 *4170 * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)4171 * 	Description4172 * 		Safely attempt to read *size* bytes from user space address4173 * 		*unsafe_ptr* and store the data in *dst*.4174 * 	Return4175 * 		0 on success, or a negative error in case of failure.4176 *4177 * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)4178 * 	Description4179 * 		Safely attempt to read *size* bytes from kernel space address4180 * 		*unsafe_ptr* and store the data in *dst*.4181 * 	Return4182 * 		0 on success, or a negative error in case of failure.4183 *4184 * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)4185 * 	Description4186 * 		Copy a NUL terminated string from an unsafe user address4187 * 		*unsafe_ptr* to *dst*. The *size* should include the4188 * 		terminating NUL byte. In case the string length is smaller than4189 * 		*size*, the target is not padded with further NUL bytes. If the4190 * 		string length is larger than *size*, just *size*-1 bytes are4191 * 		copied and the last byte is set to NUL.4192 *4193 * 		On success, returns the number of bytes that were written,4194 * 		including the terminal NUL. This makes this helper useful in4195 * 		tracing programs for reading strings, and more importantly to4196 * 		get its length at runtime. See the following snippet:4197 *4198 * 		::4199 *4200 * 			SEC("kprobe/sys_open")4201 * 			void bpf_sys_open(struct pt_regs *ctx)4202 * 			{4203 * 			        char buf[PATHLEN]; // PATHLEN is defined to 2564204 * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),4205 * 				                                  ctx->di);4206 *4207 * 				// Consume buf, for example push it to4208 * 				// userspace via bpf_perf_event_output(); we4209 * 				// can use res (the string length) as event4210 * 				// size, after checking its boundaries.4211 * 			}4212 *4213 * 		In comparison, using **bpf_probe_read_user**\ () helper here4214 * 		instead to read the string would require to estimate the length4215 * 		at compile time, and would often result in copying more memory4216 * 		than necessary.4217 *4218 * 		Another useful use case is when parsing individual process4219 * 		arguments or individual environment variables navigating4220 * 		*current*\ **->mm->arg_start** and *current*\4221 * 		**->mm->env_start**: using this helper and the return value,4222 * 		one can quickly iterate at the right offset of the memory area.4223 * 	Return4224 * 		On success, the strictly positive length of the output string,4225 * 		including the trailing NUL character. On error, a negative4226 * 		value.4227 *4228 * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)4229 * 	Description4230 * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*4231 * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.4232 * 	Return4233 * 		On success, the strictly positive length of the string, including4234 * 		the trailing NUL character. On error, a negative value.4235 *4236 * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)4237 *	Description4238 *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.4239 *		*rcv_nxt* is the ack_seq to be sent out.4240 *	Return4241 *		0 on success, or a negative error in case of failure.4242 *4243 * long bpf_send_signal_thread(u32 sig)4244 *	Description4245 *		Send signal *sig* to the thread corresponding to the current task.4246 *	Return4247 *		0 on success or successfully queued.4248 *4249 *		**-EBUSY** if work queue under nmi is full.4250 *4251 *		**-EINVAL** if *sig* is invalid.4252 *4253 *		**-EPERM** if no permission to send the *sig*.4254 *4255 *		**-EAGAIN** if bpf program can try again.4256 *4257 * u64 bpf_jiffies64(void)4258 *	Description4259 *		Obtain the 64bit jiffies4260 *	Return4261 *		The 64 bit jiffies4262 *4263 * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)4264 *	Description4265 *		For an eBPF program attached to a perf event, retrieve the4266 *		branch records (**struct perf_branch_entry**) associated to *ctx*4267 *		and store it in the buffer pointed by *buf* up to size4268 *		*size* bytes.4269 *	Return4270 *		On success, number of bytes written to *buf*. On error, a4271 *		negative value.4272 *4273 *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to4274 *		instead return the number of bytes required to store all the4275 *		branch entries. If this flag is set, *buf* may be NULL.4276 *4277 *		**-EINVAL** if arguments invalid or **size** not a multiple4278 *		of **sizeof**\ (**struct perf_branch_entry**\ ).4279 *4280 *		**-ENOENT** if architecture does not support branch records.4281 *4282 * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)4283 *	Description4284 *		Returns 0 on success, values for *pid* and *tgid* as seen from the current4285 *		*namespace* will be returned in *nsdata*.4286 *	Return4287 *		0 on success, or one of the following in case of failure:4288 *4289 *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number4290 *              with nsfs of current task, or if dev conversion to dev_t lost high bits.4291 *4292 *		**-ENOENT** if pidns does not exists for the current task.4293 *4294 * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)4295 *	Description4296 *		Write raw *data* blob into a special BPF perf event held by4297 *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf4298 *		event must have the following attributes: **PERF_SAMPLE_RAW**4299 *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and4300 *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.4301 *4302 *		The *flags* are used to indicate the index in *map* for which4303 *		the value must be put, masked with **BPF_F_INDEX_MASK**.4304 *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**4305 *		to indicate that the index of the current CPU core should be4306 *		used.4307 *4308 *		The value to write, of *size*, is passed through eBPF stack and4309 *		pointed by *data*.4310 *4311 *		*ctx* is a pointer to in-kernel struct xdp_buff.4312 *4313 *		This helper is similar to **bpf_perf_eventoutput**\ () but4314 *		restricted to raw_tracepoint bpf programs.4315 *	Return4316 *		0 on success, or a negative error in case of failure.4317 *4318 * u64 bpf_get_netns_cookie(void *ctx)4319 * 	Description4320 * 		Retrieve the cookie (generated by the kernel) of the network4321 * 		namespace the input *ctx* is associated with. The network4322 * 		namespace cookie remains stable for its lifetime and provides4323 * 		a global identifier that can be assumed unique. If *ctx* is4324 * 		NULL, then the helper returns the cookie for the initial4325 * 		network namespace. The cookie itself is very similar to that4326 * 		of **bpf_get_socket_cookie**\ () helper, but for network4327 * 		namespaces instead of sockets.4328 * 	Return4329 * 		A 8-byte long opaque number.4330 *4331 * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)4332 * 	Description4333 * 		Return id of cgroup v2 that is ancestor of the cgroup associated4334 * 		with the current task at the *ancestor_level*. The root cgroup4335 * 		is at *ancestor_level* zero and each step down the hierarchy4336 * 		increments the level. If *ancestor_level* == level of cgroup4337 * 		associated with the current task, then return value will be the4338 * 		same as that of **bpf_get_current_cgroup_id**\ ().4339 *4340 * 		The helper is useful to implement policies based on cgroups4341 * 		that are upper in hierarchy than immediate cgroup associated4342 * 		with the current task.4343 *4344 * 		The format of returned id and helper limitations are same as in4345 * 		**bpf_get_current_cgroup_id**\ ().4346 * 	Return4347 * 		The id is returned or 0 in case the id could not be retrieved.4348 *4349 * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)4350 *	Description4351 *		Helper is overloaded depending on BPF program type. This4352 *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and4353 *		**BPF_PROG_TYPE_SCHED_ACT** programs.4354 *4355 *		Assign the *sk* to the *skb*. When combined with appropriate4356 *		routing configuration to receive the packet towards the socket,4357 *		will cause *skb* to be delivered to the specified socket.4358 *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),4359 *		**bpf_clone_redirect**\ () or other methods outside of BPF may4360 *		interfere with successful delivery to the socket.4361 *4362 *		This operation is only valid from TC ingress path.4363 *4364 *		The *flags* argument must be zero.4365 *	Return4366 *		0 on success, or a negative error in case of failure:4367 *4368 *		**-EINVAL** if specified *flags* are not supported.4369 *4370 *		**-ENOENT** if the socket is unavailable for assignment.4371 *4372 *		**-ENETUNREACH** if the socket is unreachable (wrong netns).4373 *4374 *		**-EOPNOTSUPP** if the operation is not supported, for example4375 *		a call from outside of TC ingress.4376 *4377 * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)4378 *	Description4379 *		Helper is overloaded depending on BPF program type. This4380 *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.4381 *4382 *		Select the *sk* as a result of a socket lookup.4383 *4384 *		For the operation to succeed passed socket must be compatible4385 *		with the packet description provided by the *ctx* object.4386 *4387 *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must4388 *		be an exact match. While IP family (**AF_INET** or4389 *		**AF_INET6**) must be compatible, that is IPv6 sockets4390 *		that are not v6-only can be selected for IPv4 packets.4391 *4392 *		Only TCP listeners and UDP unconnected sockets can be4393 *		selected. *sk* can also be NULL to reset any previous4394 *		selection.4395 *4396 *		*flags* argument can combination of following values:4397 *4398 *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous4399 *		  socket selection, potentially done by a BPF program4400 *		  that ran before us.4401 *4402 *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip4403 *		  load-balancing within reuseport group for the socket4404 *		  being selected.4405 *4406 *		On success *ctx->sk* will point to the selected socket.4407 *4408 *	Return4409 *		0 on success, or a negative errno in case of failure.4410 *4411 *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is4412 *		  not compatible with packet family (*ctx->family*).4413 *4414 *		* **-EEXIST** if socket has been already selected,4415 *		  potentially by another program, and4416 *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.4417 *4418 *		* **-EINVAL** if unsupported flags were specified.4419 *4420 *		* **-EPROTOTYPE** if socket L4 protocol4421 *		  (*sk->protocol*) doesn't match packet protocol4422 *		  (*ctx->protocol*).4423 *4424 *		* **-ESOCKTNOSUPPORT** if socket is not in allowed4425 *		  state (TCP listening or UDP unconnected).4426 *4427 * u64 bpf_ktime_get_boot_ns(void)4428 * 	Description4429 * 		Return the time elapsed since system boot, in nanoseconds.4430 * 		Does include the time the system was suspended.4431 * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)4432 * 	Return4433 * 		Current *ktime*.4434 *4435 * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)4436 * 	Description4437 * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print4438 * 		out the format string.4439 * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for4440 * 		the format string itself. The *data* and *data_len* are format string4441 * 		arguments. The *data* are a **u64** array and corresponding format string4442 * 		values are stored in the array. For strings and pointers where pointees4443 * 		are accessed, only the pointer values are stored in the *data* array.4444 * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.4445 *4446 *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.4447 *		Reading kernel memory may fail due to either invalid address or4448 *		valid address but requiring a major memory fault. If reading kernel memory4449 *		fails, the string for **%s** will be an empty string, and the ip4450 *		address for **%p{i,I}{4,6}** will be 0. Not returning error to4451 *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.4452 * 	Return4453 * 		0 on success, or a negative error in case of failure:4454 *4455 *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again4456 *		by returning 1 from bpf program.4457 *4458 *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.4459 *4460 *		**-E2BIG** if *fmt* contains too many format specifiers.4461 *4462 *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.4463 *4464 * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)4465 * 	Description4466 * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.4467 * 		The *m* represents the seq_file. The *data* and *len* represent the4468 * 		data to write in bytes.4469 * 	Return4470 * 		0 on success, or a negative error in case of failure:4471 *4472 *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.4473 *4474 * u64 bpf_sk_cgroup_id(void *sk)4475 *	Description4476 *		Return the cgroup v2 id of the socket *sk*.4477 *4478 *		*sk* must be a non-**NULL** pointer to a socket, e.g. one4479 *		returned from **bpf_sk_lookup_xxx**\ (),4480 *		**bpf_sk_fullsock**\ (), etc. The format of returned id is4481 *		same as in **bpf_skb_cgroup_id**\ ().4482 *4483 *		This helper is available only if the kernel was compiled with4484 *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.4485 *	Return4486 *		The id is returned or 0 in case the id could not be retrieved.4487 *4488 * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)4489 *	Description4490 *		Return id of cgroup v2 that is ancestor of cgroup associated4491 *		with the *sk* at the *ancestor_level*.  The root cgroup is at4492 *		*ancestor_level* zero and each step down the hierarchy4493 *		increments the level. If *ancestor_level* == level of cgroup4494 *		associated with *sk*, then return value will be same as that4495 *		of **bpf_sk_cgroup_id**\ ().4496 *4497 *		The helper is useful to implement policies based on cgroups4498 *		that are upper in hierarchy than immediate cgroup associated4499 *		with *sk*.4500 *4501 *		The format of returned id and helper limitations are same as in4502 *		**bpf_sk_cgroup_id**\ ().4503 *	Return4504 *		The id is returned or 0 in case the id could not be retrieved.4505 *4506 * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)4507 * 	Description4508 * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.4509 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification4510 * 		of new data availability is sent.4511 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification4512 * 		of new data availability is sent unconditionally.4513 * 		If **0** is specified in *flags*, an adaptive notification4514 * 		of new data availability is sent.4515 *4516 * 		An adaptive notification is a notification sent whenever the user-space4517 * 		process has caught up and consumed all available payloads. In case the user-space4518 * 		process is still processing a previous payload, then no notification is needed4519 * 		as it will process the newly added payload automatically.4520 * 	Return4521 * 		0 on success, or a negative error in case of failure.4522 *4523 * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)4524 * 	Description4525 * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.4526 * 		*flags* must be 0.4527 * 	Return4528 * 		Valid pointer with *size* bytes of memory available; NULL,4529 * 		otherwise.4530 *4531 * void bpf_ringbuf_submit(void *data, u64 flags)4532 * 	Description4533 * 		Submit reserved ring buffer sample, pointed to by *data*.4534 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification4535 * 		of new data availability is sent.4536 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification4537 * 		of new data availability is sent unconditionally.4538 * 		If **0** is specified in *flags*, an adaptive notification4539 * 		of new data availability is sent.4540 *4541 * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.4542 * 	Return4543 * 		Nothing. Always succeeds.4544 *4545 * void bpf_ringbuf_discard(void *data, u64 flags)4546 * 	Description4547 * 		Discard reserved ring buffer sample, pointed to by *data*.4548 * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification4549 * 		of new data availability is sent.4550 * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification4551 * 		of new data availability is sent unconditionally.4552 * 		If **0** is specified in *flags*, an adaptive notification4553 * 		of new data availability is sent.4554 *4555 * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.4556 * 	Return4557 * 		Nothing. Always succeeds.4558 *4559 * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)4560 *	Description4561 *		Query various characteristics of provided ring buffer. What4562 *		exactly is queries is determined by *flags*:4563 *4564 *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.4565 *		* **BPF_RB_RING_SIZE**: The size of ring buffer.4566 *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).4567 *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).4568 *4569 *		Data returned is just a momentary snapshot of actual values4570 *		and could be inaccurate, so this facility should be used to4571 *		power heuristics and for reporting, not to make 100% correct4572 *		calculation.4573 *	Return4574 *		Requested value, or 0, if *flags* are not recognized.4575 *4576 * long bpf_csum_level(struct sk_buff *skb, u64 level)4577 * 	Description4578 * 		Change the skbs checksum level by one layer up or down, or4579 * 		reset it entirely to none in order to have the stack perform4580 * 		checksum validation. The level is applicable to the following4581 * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of4582 * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |4583 * 		through **bpf_skb_adjust_room**\ () helper with passing in4584 * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call4585 * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since4586 * 		the UDP header is removed. Similarly, an encap of the latter4587 * 		into the former could be accompanied by a helper call to4588 * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the4589 * 		skb is still intended to be processed in higher layers of the4590 * 		stack instead of just egressing at tc.4591 *4592 * 		There are three supported level settings at this time:4593 *4594 * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs4595 * 		  with CHECKSUM_UNNECESSARY.4596 * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs4597 * 		  with CHECKSUM_UNNECESSARY.4598 * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and4599 * 		  sets CHECKSUM_NONE to force checksum validation by the stack.4600 * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current4601 * 		  skb->csum_level.4602 * 	Return4603 * 		0 on success, or a negative error in case of failure. In the4604 * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level4605 * 		is returned or the error code -EACCES in case the skb is not4606 * 		subject to CHECKSUM_UNNECESSARY.4607 *4608 * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)4609 *	Description4610 *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.4611 *	Return4612 *		*sk* if casting is valid, or **NULL** otherwise.4613 *4614 * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)4615 *	Description4616 *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.4617 *	Return4618 *		*sk* if casting is valid, or **NULL** otherwise.4619 *4620 * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)4621 * 	Description4622 *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.4623 *	Return4624 *		*sk* if casting is valid, or **NULL** otherwise.4625 *4626 * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)4627 * 	Description4628 *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.4629 *	Return4630 *		*sk* if casting is valid, or **NULL** otherwise.4631 *4632 * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)4633 * 	Description4634 *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.4635 *	Return4636 *		*sk* if casting is valid, or **NULL** otherwise.4637 *4638 * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)4639 *	Description4640 *		Return a user or a kernel stack in bpf program provided buffer.4641 *		Note: the user stack will only be populated if the *task* is4642 *		the current task; all other tasks will return -EOPNOTSUPP.4643 *		To achieve this, the helper needs *task*, which is a valid4644 *		pointer to **struct task_struct**. To store the stacktrace, the4645 *		bpf program provides *buf* with a nonnegative *size*.4646 *4647 *		The last argument, *flags*, holds the number of stack frames to4648 *		skip (from 0 to 255), masked with4649 *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set4650 *		the following flags:4651 *4652 *		**BPF_F_USER_STACK**4653 *			Collect a user space stack instead of a kernel stack.4654 *			The *task* must be the current task.4655 *		**BPF_F_USER_BUILD_ID**4656 *			Collect buildid+offset instead of ips for user stack,4657 *			only valid if **BPF_F_USER_STACK** is also specified.4658 *4659 *		**bpf_get_task_stack**\ () can collect up to4660 *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject4661 *		to sufficient large buffer size. Note that4662 *		this limit can be controlled with the **sysctl** program, and4663 *		that it should be manually increased in order to profile long4664 *		user stacks (such as stacks for Java programs). To do so, use:4665 *4666 *		::4667 *4668 *			# sysctl kernel.perf_event_max_stack=<new value>4669 *	Return4670 * 		The non-negative copied *buf* length equal to or less than4671 * 		*size* on success, or a negative error in case of failure.4672 *4673 * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)4674 *	Description4675 *		Load header option.  Support reading a particular TCP header4676 *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).4677 *4678 *		If *flags* is 0, it will search the option from the4679 *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**4680 *		has details on what skb_data contains under different4681 *		*skops*\ **->op**.4682 *4683 *		The first byte of the *searchby_res* specifies the4684 *		kind that it wants to search.4685 *4686 *		If the searching kind is an experimental kind4687 *		(i.e. 253 or 254 according to RFC6994).  It also4688 *		needs to specify the "magic" which is either4689 *		2 bytes or 4 bytes.  It then also needs to4690 *		specify the size of the magic by using4691 *		the 2nd byte which is "kind-length" of a TCP4692 *		header option and the "kind-length" also4693 *		includes the first 2 bytes "kind" and "kind-length"4694 *		itself as a normal TCP header option also does.4695 *4696 *		For example, to search experimental kind 254 with4697 *		2 byte magic 0xeB9F, the searchby_res should be4698 *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].4699 *4700 *		To search for the standard window scale option (3),4701 *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].4702 *		Note, kind-length must be 0 for regular option.4703 *4704 *		Searching for No-Op (0) and End-of-Option-List (1) are4705 *		not supported.4706 *4707 *		*len* must be at least 2 bytes which is the minimal size4708 *		of a header option.4709 *4710 *		Supported flags:4711 *4712 *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the4713 *		  saved_syn packet or the just-received syn packet.4714 *4715 *	Return4716 *		> 0 when found, the header option is copied to *searchby_res*.4717 *		The return value is the total length copied. On failure, a4718 *		negative error code is returned:4719 *4720 *		**-EINVAL** if a parameter is invalid.4721 *4722 *		**-ENOMSG** if the option is not found.4723 *4724 *		**-ENOENT** if no syn packet is available when4725 *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.4726 *4727 *		**-ENOSPC** if there is not enough space.  Only *len* number of4728 *		bytes are copied.4729 *4730 *		**-EFAULT** on failure to parse the header options in the4731 *		packet.4732 *4733 *		**-EPERM** if the helper cannot be used under the current4734 *		*skops*\ **->op**.4735 *4736 * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)4737 *	Description4738 *		Store header option.  The data will be copied4739 *		from buffer *from* with length *len* to the TCP header.4740 *4741 *		The buffer *from* should have the whole option that4742 *		includes the kind, kind-length, and the actual4743 *		option data.  The *len* must be at least kind-length4744 *		long.  The kind-length does not have to be 4 byte4745 *		aligned.  The kernel will take care of the padding4746 *		and setting the 4 bytes aligned value to th->doff.4747 *4748 *		This helper will check for duplicated option4749 *		by searching the same option in the outgoing skb.4750 *4751 *		This helper can only be called during4752 *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.4753 *4754 *	Return4755 *		0 on success, or negative error in case of failure:4756 *4757 *		**-EINVAL** If param is invalid.4758 *4759 *		**-ENOSPC** if there is not enough space in the header.4760 *		Nothing has been written4761 *4762 *		**-EEXIST** if the option already exists.4763 *4764 *		**-EFAULT** on failure to parse the existing header options.4765 *4766 *		**-EPERM** if the helper cannot be used under the current4767 *		*skops*\ **->op**.4768 *4769 * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)4770 *	Description4771 *		Reserve *len* bytes for the bpf header option.  The4772 *		space will be used by **bpf_store_hdr_opt**\ () later in4773 *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.4774 *4775 *		If **bpf_reserve_hdr_opt**\ () is called multiple times,4776 *		the total number of bytes will be reserved.4777 *4778 *		This helper can only be called during4779 *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.4780 *4781 *	Return4782 *		0 on success, or negative error in case of failure:4783 *4784 *		**-EINVAL** if a parameter is invalid.4785 *4786 *		**-ENOSPC** if there is not enough space in the header.4787 *4788 *		**-EPERM** if the helper cannot be used under the current4789 *		*skops*\ **->op**.4790 *4791 * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)4792 *	Description4793 *		Get a bpf_local_storage from an *inode*.4794 *4795 *		Logically, it could be thought of as getting the value from4796 *		a *map* with *inode* as the **key**.  From this4797 *		perspective,  the usage is not much different from4798 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this4799 *		helper enforces the key must be an inode and the map must also4800 *		be a **BPF_MAP_TYPE_INODE_STORAGE**.4801 *4802 *		Underneath, the value is stored locally at *inode* instead of4803 *		the *map*.  The *map* is used as the bpf-local-storage4804 *		"type". The bpf-local-storage "type" (i.e. the *map*) is4805 *		searched against all bpf_local_storage residing at *inode*.4806 *4807 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be4808 *		used such that a new bpf_local_storage will be4809 *		created if one does not exist.  *value* can be used4810 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify4811 *		the initial value of a bpf_local_storage.  If *value* is4812 *		**NULL**, the new bpf_local_storage will be zero initialized.4813 *	Return4814 *		A bpf_local_storage pointer is returned on success.4815 *4816 *		**NULL** if not found or there was an error in adding4817 *		a new bpf_local_storage.4818 *4819 * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)4820 *	Description4821 *		Delete a bpf_local_storage from an *inode*.4822 *	Return4823 *		0 on success.4824 *4825 *		**-ENOENT** if the bpf_local_storage cannot be found.4826 *4827 * long bpf_d_path(struct path *path, char *buf, u32 sz)4828 *	Description4829 *		Return full path for given **struct path** object, which4830 *		needs to be the kernel BTF *path* object. The path is4831 *		returned in the provided buffer *buf* of size *sz* and4832 *		is zero terminated.4833 *4834 *	Return4835 *		On success, the strictly positive length of the string,4836 *		including the trailing NUL character. On error, a negative4837 *		value.4838 *4839 * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)4840 * 	Description4841 * 		Read *size* bytes from user space address *user_ptr* and store4842 * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().4843 * 	Return4844 * 		0 on success, or a negative error in case of failure.4845 *4846 * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)4847 *	Description4848 *		Use BTF to store a string representation of *ptr*->ptr in *str*,4849 *		using *ptr*->type_id.  This value should specify the type4850 *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)4851 *		can be used to look up vmlinux BTF type ids. Traversing the4852 *		data structure using BTF, the type information and values are4853 *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of4854 *		the pointer data is carried out to avoid kernel crashes during4855 *		operation.  Smaller types can use string space on the stack;4856 *		larger programs can use map data to store the string4857 *		representation.4858 *4859 *		The string can be subsequently shared with userspace via4860 *		bpf_perf_event_output() or ring buffer interfaces.4861 *		bpf_trace_printk() is to be avoided as it places too small4862 *		a limit on string size to be useful.4863 *4864 *		*flags* is a combination of4865 *4866 *		**BTF_F_COMPACT**4867 *			no formatting around type information4868 *		**BTF_F_NONAME**4869 *			no struct/union member names/types4870 *		**BTF_F_PTR_RAW**4871 *			show raw (unobfuscated) pointer values;4872 *			equivalent to printk specifier %px.4873 *		**BTF_F_ZERO**4874 *			show zero-valued struct/union members; they4875 *			are not displayed by default4876 *4877 *	Return4878 *		The number of bytes that were written (or would have been4879 *		written if output had to be truncated due to string size),4880 *		or a negative error in cases of failure.4881 *4882 * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)4883 *	Description4884 *		Use BTF to write to seq_write a string representation of4885 *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().4886 *		*flags* are identical to those used for bpf_snprintf_btf.4887 *	Return4888 *		0 on success or a negative error in case of failure.4889 *4890 * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)4891 * 	Description4892 * 		See **bpf_get_cgroup_classid**\ () for the main description.4893 * 		This helper differs from **bpf_get_cgroup_classid**\ () in that4894 * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s4895 * 		associated socket instead of the current process.4896 * 	Return4897 * 		The id is returned or 0 in case the id could not be retrieved.4898 *4899 * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)4900 * 	Description4901 * 		Redirect the packet to another net device of index *ifindex*4902 * 		and fill in L2 addresses from neighboring subsystem. This helper4903 * 		is somewhat similar to **bpf_redirect**\ (), except that it4904 * 		populates L2 addresses as well, meaning, internally, the helper4905 * 		relies on the neighbor lookup for the L2 address of the nexthop.4906 *4907 * 		The helper will perform a FIB lookup based on the skb's4908 * 		networking header to get the address of the next hop, unless4909 * 		this is supplied by the caller in the *params* argument. The4910 * 		*plen* argument indicates the len of *params* and should be set4911 * 		to 0 if *params* is NULL.4912 *4913 * 		The *flags* argument is reserved and must be 0. The helper is4914 * 		currently only supported for tc BPF program types, and enabled4915 * 		for IPv4 and IPv6 protocols.4916 * 	Return4917 * 		The helper returns **TC_ACT_REDIRECT** on success or4918 * 		**TC_ACT_SHOT** on error.4919 *4920 * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)4921 *     Description4922 *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a4923 *             pointer to the percpu kernel variable on *cpu*. A ksym is an4924 *             extern variable decorated with '__ksym'. For ksym, there is a4925 *             global var (either static or global) defined of the same name4926 *             in the kernel. The ksym is percpu if the global var is percpu.4927 *             The returned pointer points to the global percpu var on *cpu*.4928 *4929 *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the4930 *             kernel, except that bpf_per_cpu_ptr() may return NULL. This4931 *             happens if *cpu* is larger than nr_cpu_ids. The caller of4932 *             bpf_per_cpu_ptr() must check the returned value.4933 *     Return4934 *             A pointer pointing to the kernel percpu variable on *cpu*, or4935 *             NULL, if *cpu* is invalid.4936 *4937 * void *bpf_this_cpu_ptr(const void *percpu_ptr)4938 *	Description4939 *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a4940 *		pointer to the percpu kernel variable on this cpu. See the4941 *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().4942 *4943 *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in4944 *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would4945 *		never return NULL.4946 *	Return4947 *		A pointer pointing to the kernel percpu variable on this cpu.4948 *4949 * long bpf_redirect_peer(u32 ifindex, u64 flags)4950 * 	Description4951 * 		Redirect the packet to another net device of index *ifindex*.4952 * 		This helper is somewhat similar to **bpf_redirect**\ (), except4953 * 		that the redirection happens to the *ifindex*' peer device and4954 * 		the netns switch takes place from ingress to ingress without4955 * 		going through the CPU's backlog queue.4956 *4957 * 		The *flags* argument is reserved and must be 0. The helper is4958 * 		currently only supported for tc BPF program types at the4959 * 		ingress hook and for veth and netkit target device types. The4960 * 		peer device must reside in a different network namespace.4961 * 	Return4962 * 		The helper returns **TC_ACT_REDIRECT** on success or4963 * 		**TC_ACT_SHOT** on error.4964 *4965 * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)4966 *	Description4967 *		Get a bpf_local_storage from the *task*.4968 *4969 *		Logically, it could be thought of as getting the value from4970 *		a *map* with *task* as the **key**.  From this4971 *		perspective,  the usage is not much different from4972 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this4973 *		helper enforces the key must be a task_struct and the map must also4974 *		be a **BPF_MAP_TYPE_TASK_STORAGE**.4975 *4976 *		Underneath, the value is stored locally at *task* instead of4977 *		the *map*.  The *map* is used as the bpf-local-storage4978 *		"type". The bpf-local-storage "type" (i.e. the *map*) is4979 *		searched against all bpf_local_storage residing at *task*.4980 *4981 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be4982 *		used such that a new bpf_local_storage will be4983 *		created if one does not exist.  *value* can be used4984 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify4985 *		the initial value of a bpf_local_storage.  If *value* is4986 *		**NULL**, the new bpf_local_storage will be zero initialized.4987 *	Return4988 *		A bpf_local_storage pointer is returned on success.4989 *4990 *		**NULL** if not found or there was an error in adding4991 *		a new bpf_local_storage.4992 *4993 * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)4994 *	Description4995 *		Delete a bpf_local_storage from a *task*.4996 *	Return4997 *		0 on success.4998 *4999 *		**-ENOENT** if the bpf_local_storage cannot be found.5000 *5001 * struct task_struct *bpf_get_current_task_btf(void)5002 *	Description5003 *		Return a BTF pointer to the "current" task.5004 *		This pointer can also be used in helpers that accept an5005 *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.5006 *	Return5007 *		Pointer to the current task.5008 *5009 * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)5010 *	Description5011 *		Set or clear certain options on *bprm*:5012 *5013 *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit5014 *		which sets the **AT_SECURE** auxv for glibc. The bit5015 *		is cleared if the flag is not specified.5016 *	Return5017 *		**-EINVAL** if invalid *flags* are passed, zero otherwise.5018 *5019 * u64 bpf_ktime_get_coarse_ns(void)5020 * 	Description5021 * 		Return a coarse-grained version of the time elapsed since5022 * 		system boot, in nanoseconds. Does not include time the system5023 * 		was suspended.5024 *5025 * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)5026 * 	Return5027 * 		Current *ktime*.5028 *5029 * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)5030 *	Description5031 *		Returns the stored IMA hash of the *inode* (if it's available).5032 *		If the hash is larger than *size*, then only *size*5033 *		bytes will be copied to *dst*5034 *	Return5035 *		The **hash_algo** is returned on success,5036 *		**-EOPNOTSUPP** if IMA is disabled or **-EINVAL** if5037 *		invalid arguments are passed.5038 *5039 * struct socket *bpf_sock_from_file(struct file *file)5040 *	Description5041 *		If the given file represents a socket, returns the associated5042 *		socket.5043 *	Return5044 *		A pointer to a struct socket on success or NULL if the file is5045 *		not a socket.5046 *5047 * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)5048 *	Description5049 *		Check packet size against exceeding MTU of net device (based5050 *		on *ifindex*).  This helper will likely be used in combination5051 *		with helpers that adjust/change the packet size.5052 *5053 *		The argument *len_diff* can be used for querying with a planned5054 *		size change. This allows to check MTU prior to changing packet5055 *		ctx. Providing a *len_diff* adjustment that is larger than the5056 *		actual packet size (resulting in negative packet size) will in5057 *		principle not exceed the MTU, which is why it is not considered5058 *		a failure.  Other BPF helpers are needed for performing the5059 *		planned size change; therefore the responsibility for catching5060 *		a negative packet size belongs in those helpers.5061 *5062 *		Specifying *ifindex* zero means the MTU check is performed5063 *		against the current net device.  This is practical if this isn't5064 *		used prior to redirect.5065 *5066 *		On input *mtu_len* must be a valid pointer, else verifier will5067 *		reject BPF program.  If the value *mtu_len* is initialized to5068 *		zero then the ctx packet size is use.  When value *mtu_len* is5069 *		provided as input this specify the L3 length that the MTU check5070 *		is done against. Remember XDP and TC length operate at L2, but5071 *		this value is L3 as this correlate to MTU and IP-header tot_len5072 *		values which are L3 (similar behavior as bpf_fib_lookup).5073 *5074 *		The Linux kernel route table can configure MTUs on a more5075 *		specific per route level, which is not provided by this helper.5076 *		For route level MTU checks use the **bpf_fib_lookup**\ ()5077 *		helper.5078 *5079 *		*ctx* is either **struct xdp_md** for XDP programs or5080 *		**struct sk_buff** for tc cls_act programs.5081 *5082 *		The *flags* argument can be a combination of one or more of the5083 *		following values:5084 *5085 *		**BPF_MTU_CHK_SEGS**5086 *			This flag will only works for *ctx* **struct sk_buff**.5087 *			If packet context contains extra packet segment buffers5088 *			(often knows as GSO skb), then MTU check is harder to5089 *			check at this point, because in transmit path it is5090 *			possible for the skb packet to get re-segmented5091 *			(depending on net device features).  This could still be5092 *			a MTU violation, so this flag enables performing MTU5093 *			check against segments, with a different violation5094 *			return code to tell it apart. Check cannot use len_diff.5095 *5096 *		On return *mtu_len* pointer contains the MTU value of the net5097 *		device.  Remember the net device configured MTU is the L3 size,5098 *		which is returned here and XDP and TC length operate at L2.5099 *		Helper take this into account for you, but remember when using5100 *		MTU value in your BPF-code.5101 *5102 *	Return5103 *		* 0 on success, and populate MTU value in *mtu_len* pointer.5104 *5105 *		* < 0 if any input argument is invalid (*mtu_len* not updated)5106 *5107 *		MTU violations return positive values, but also populate MTU5108 *		value in *mtu_len* pointer, as this can be needed for5109 *		implementing PMTU handing:5110 *5111 *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**5112 *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**5113 *5114 * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)5115 *	Description5116 *		For each element in **map**, call **callback_fn** function with5117 *		**map**, **callback_ctx** and other map-specific parameters.5118 *		The **callback_fn** should be a static function and5119 *		the **callback_ctx** should be a pointer to the stack.5120 *		The **flags** is used to control certain aspects of the helper.5121 *		Currently, the **flags** must be 0.5122 *5123 *		The following are a list of supported map types and their5124 *		respective expected callback signatures:5125 *5126 *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,5127 *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,5128 *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY5129 *5130 *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);5131 *5132 *		For per_cpu maps, the map_value is the value on the cpu where the5133 *		bpf_prog is running.5134 *5135 *		If **callback_fn** return 0, the helper will continue to the next5136 *		element. If return value is 1, the helper will skip the rest of5137 *		elements and return. Other return values are not used now.5138 *5139 *	Return5140 *		The number of traversed map elements for success, **-EINVAL** for5141 *		invalid **flags**.5142 *5143 * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)5144 *	Description5145 *		Outputs a string into the **str** buffer of size **str_size**5146 *		based on a format string stored in a read-only map pointed by5147 *		**fmt**.5148 *5149 *		Each format specifier in **fmt** corresponds to one u64 element5150 *		in the **data** array. For strings and pointers where pointees5151 *		are accessed, only the pointer values are stored in the *data*5152 *		array. The *data_len* is the size of *data* in bytes - must be5153 *		a multiple of 8.5154 *5155 *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel5156 *		memory. Reading kernel memory may fail due to either invalid5157 *		address or valid address but requiring a major memory fault. If5158 *		reading kernel memory fails, the string for **%s** will be an5159 *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.5160 *		Not returning error to bpf program is consistent with what5161 *		**bpf_trace_printk**\ () does for now.5162 *5163 *	Return5164 *		The strictly positive length of the formatted string, including5165 *		the trailing zero character. If the return value is greater than5166 *		**str_size**, **str** contains a truncated string, guaranteed to5167 *		be zero-terminated except when **str_size** is 0.5168 *5169 *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.5170 *5171 * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)5172 * 	Description5173 * 		Execute bpf syscall with given arguments.5174 * 	Return5175 * 		A syscall result.5176 *5177 * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)5178 * 	Description5179 * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.5180 * 	Return5181 * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.5182 *5183 * long bpf_sys_close(u32 fd)5184 * 	Description5185 * 		Execute close syscall for given FD.5186 * 	Return5187 * 		A syscall result.5188 *5189 * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)5190 *	Description5191 *		Initialize the timer.5192 *		First 4 bits of *flags* specify clockid.5193 *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.5194 *		All other bits of *flags* are reserved.5195 *		The verifier will reject the program if *timer* is not from5196 *		the same *map*.5197 *	Return5198 *		0 on success.5199 *		**-EBUSY** if *timer* is already initialized.5200 *		**-EINVAL** if invalid *flags* are passed.5201 *		**-EPERM** if *timer* is in a map that doesn't have any user references.5202 *		The user space should either hold a file descriptor to a map with timers5203 *		or pin such map in bpffs. When map is unpinned or file descriptor is5204 *		closed all timers in the map will be cancelled and freed.5205 *5206 * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)5207 *	Description5208 *		Configure the timer to call *callback_fn* static function.5209 *	Return5210 *		0 on success.5211 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.5212 *		**-EPERM** if *timer* is in a map that doesn't have any user references.5213 *		The user space should either hold a file descriptor to a map with timers5214 *		or pin such map in bpffs. When map is unpinned or file descriptor is5215 *		closed all timers in the map will be cancelled and freed.5216 *5217 * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)5218 *	Description5219 *		Set timer expiration N nanoseconds from the current time. The5220 *		configured callback will be invoked in soft irq context on some cpu5221 *		and will not repeat unless another bpf_timer_start() is made.5222 *		In such case the next invocation can migrate to a different cpu.5223 *		Since struct bpf_timer is a field inside map element the map5224 *		owns the timer. The bpf_timer_set_callback() will increment refcnt5225 *		of BPF program to make sure that callback_fn code stays valid.5226 *		When user space reference to a map reaches zero all timers5227 *		in a map are cancelled and corresponding program's refcnts are5228 *		decremented. This is done to make sure that Ctrl-C of a user5229 *		process doesn't leave any timers running. If map is pinned in5230 *		bpffs the callback_fn can re-arm itself indefinitely.5231 *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands5232 *		cancel and free the timer in the given map element.5233 *		The map can contain timers that invoke callback_fn-s from different5234 *		programs. The same callback_fn can serve different timers from5235 *		different maps if key/value layout matches across maps.5236 *		Every bpf_timer_set_callback() can have different callback_fn.5237 *5238 *		*flags* can be one of:5239 *5240 *		**BPF_F_TIMER_ABS**5241 *			Start the timer in absolute expire value instead of the5242 *			default relative one.5243 *		**BPF_F_TIMER_CPU_PIN**5244 *			Timer will be pinned to the CPU of the caller.5245 *5246 *	Return5247 *		0 on success.5248 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier5249 *		or invalid *flags* are passed.5250 *5251 * long bpf_timer_cancel(struct bpf_timer *timer)5252 *	Description5253 *		Cancel the timer and wait for callback_fn to finish if it was running.5254 *	Return5255 *		0 if the timer was not active.5256 *		1 if the timer was active.5257 *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.5258 *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its5259 *		own timer which would have led to a deadlock otherwise.5260 *5261 * u64 bpf_get_func_ip(void *ctx)5262 * 	Description5263 * 		Get address of the traced function (for tracing and kprobe programs).5264 *5265 * 		When called for kprobe program attached as uprobe it returns5266 * 		probe address for both entry and return uprobe.5267 *5268 * 	Return5269 * 		Address of the traced function for kprobe.5270 * 		0 for kprobes placed within the function (not at the entry).5271 * 		Address of the probe for uprobe and return uprobe.5272 *5273 * u64 bpf_get_attach_cookie(void *ctx)5274 * 	Description5275 * 		Get bpf_cookie value provided (optionally) during the program5276 * 		attachment. It might be different for each individual5277 * 		attachment, even if BPF program itself is the same.5278 * 		Expects BPF program context *ctx* as a first argument.5279 *5280 * 		Supported for the following program types:5281 *			- kprobe/uprobe;5282 *			- tracepoint;5283 *			- perf_event.5284 * 	Return5285 *		Value specified by user at BPF link creation/attachment time5286 *		or 0, if it was not specified.5287 *5288 * long bpf_task_pt_regs(struct task_struct *task)5289 *	Description5290 *		Get the struct pt_regs associated with **task**.5291 *	Return5292 *		A pointer to struct pt_regs.5293 *5294 * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)5295 *	Description5296 *		Get branch trace from hardware engines like Intel LBR. The5297 *		hardware engine is stopped shortly after the helper is5298 *		called. Therefore, the user need to filter branch entries5299 *		based on the actual use case. To capture branch trace5300 *		before the trigger point of the BPF program, the helper5301 *		should be called at the beginning of the BPF program.5302 *5303 *		The data is stored as struct perf_branch_entry into output5304 *		buffer *entries*. *size* is the size of *entries* in bytes.5305 *		*flags* is reserved for now and must be zero.5306 *5307 *	Return5308 *		On success, number of bytes written to *buf*. On error, a5309 *		negative value.5310 *5311 *		**-EINVAL** if *flags* is not zero.5312 *5313 *		**-ENOENT** if architecture does not support branch records.5314 *5315 * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)5316 *	Description5317 *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u645318 *		to format and can handle more format args as a result.5319 *5320 *		Arguments are to be used as in **bpf_seq_printf**\ () helper.5321 *	Return5322 *		The number of bytes written to the buffer, or a negative error5323 *		in case of failure.5324 *5325 * struct unix_sock *bpf_skc_to_unix_sock(void *sk)5326 * 	Description5327 *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.5328 *	Return5329 *		*sk* if casting is valid, or **NULL** otherwise.5330 *5331 * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)5332 *	Description5333 *		Get the address of a kernel symbol, returned in *res*. *res* is5334 *		set to 0 if the symbol is not found.5335 *	Return5336 *		On success, zero. On error, a negative value.5337 *5338 *		**-EINVAL** if *flags* is not zero.5339 *5340 *		**-EINVAL** if string *name* is not the same size as *name_sz*.5341 *5342 *		**-ENOENT** if symbol is not found.5343 *5344 *		**-EPERM** if caller does not have permission to obtain kernel address.5345 *5346 * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)5347 *	Description5348 *		Find vma of *task* that contains *addr*, call *callback_fn*5349 *		function with *task*, *vma*, and *callback_ctx*.5350 *		The *callback_fn* should be a static function and5351 *		the *callback_ctx* should be a pointer to the stack.5352 *		The *flags* is used to control certain aspects of the helper.5353 *		Currently, the *flags* must be 0.5354 *5355 *		The expected callback signature is5356 *5357 *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);5358 *5359 *	Return5360 *		0 on success.5361 *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.5362 *		**-EBUSY** if failed to try lock mmap_lock.5363 *		**-EINVAL** for invalid **flags**.5364 *5365 * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)5366 *	Description5367 *		For **nr_loops**, call **callback_fn** function5368 *		with **callback_ctx** as the context parameter.5369 *		The **callback_fn** should be a static function and5370 *		the **callback_ctx** should be a pointer to the stack.5371 *		The **flags** is used to control certain aspects of the helper.5372 *		Currently, the **flags** must be 0. Currently, nr_loops is5373 *		limited to 1 << 23 (~8 million) loops.5374 *5375 *		long (\*callback_fn)(u32 index, void \*ctx);5376 *5377 *		where **index** is the current index in the loop. The index5378 *		is zero-indexed.5379 *5380 *		If **callback_fn** returns 0, the helper will continue to the next5381 *		loop. If return value is 1, the helper will skip the rest of5382 *		the loops and return. Other return values are not used now,5383 *		and will be rejected by the verifier.5384 *5385 *	Return5386 *		The number of loops performed, **-EINVAL** for invalid **flags**,5387 *		**-E2BIG** if **nr_loops** exceeds the maximum number of loops.5388 *5389 * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)5390 *	Description5391 *		Do strncmp() between **s1** and **s2**. **s1** doesn't need5392 *		to be null-terminated and **s1_sz** is the maximum storage5393 *		size of **s1**. **s2** must be a read-only string.5394 *	Return5395 *		An integer less than, equal to, or greater than zero5396 *		if the first **s1_sz** bytes of **s1** is found to be5397 *		less than, to match, or be greater than **s2**.5398 *5399 * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)5400 *	Description5401 *		Get **n**-th argument register (zero based) of the traced function (for tracing programs)5402 *		returned in **value**.5403 *5404 *	Return5405 *		0 on success.5406 *		**-EINVAL** if n >= argument register count of traced function.5407 *5408 * long bpf_get_func_ret(void *ctx, u64 *value)5409 *	Description5410 *		Get return value of the traced function (for tracing programs)5411 *		in **value**.5412 *5413 *	Return5414 *		0 on success.5415 *		**-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.5416 *5417 * long bpf_get_func_arg_cnt(void *ctx)5418 *	Description5419 *		Get number of registers of the traced function (for tracing programs) where5420 *		function arguments are stored in these registers.5421 *5422 *	Return5423 *		The number of argument registers of the traced function.5424 *5425 * int bpf_get_retval(void)5426 *	Description5427 *		Get the BPF program's return value that will be returned to the upper layers.5428 *5429 *		This helper is currently supported by cgroup programs and only by the hooks5430 *		where BPF program's return value is returned to the userspace via errno.5431 *	Return5432 *		The BPF program's return value.5433 *5434 * int bpf_set_retval(int retval)5435 *	Description5436 *		Set the BPF program's return value that will be returned to the upper layers.5437 *5438 *		This helper is currently supported by cgroup programs and only by the hooks5439 *		where BPF program's return value is returned to the userspace via errno.5440 *5441 *		Note that there is the following corner case where the program exports an error5442 *		via bpf_set_retval but signals success via 'return 1':5443 *5444 *			bpf_set_retval(-EPERM);5445 *			return 1;5446 *5447 *		In this case, the BPF program's return value will use helper's -EPERM. This5448 *		still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.5449 *5450 *	Return5451 *		0 on success, or a negative error in case of failure.5452 *5453 * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)5454 *	Description5455 *		Get the total size of a given xdp buff (linear and paged area)5456 *	Return5457 *		The total size of a given xdp buffer.5458 *5459 * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)5460 *	Description5461 *		This helper is provided as an easy way to load data from a5462 *		xdp buffer. It can be used to load *len* bytes from *offset* from5463 *		the frame associated to *xdp_md*, into the buffer pointed by5464 *		*buf*.5465 *	Return5466 *		0 on success, or a negative error in case of failure.5467 *5468 * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)5469 *	Description5470 *		Store *len* bytes from buffer *buf* into the frame5471 *		associated to *xdp_md*, at *offset*.5472 *	Return5473 *		0 on success, or a negative error in case of failure.5474 *5475 * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)5476 *	Description5477 *		Read *size* bytes from user space address *user_ptr* in *tsk*'s5478 *		address space, and stores the data in *dst*. *flags* is not5479 *		used yet and is provided for future extensibility. This helper5480 *		can only be used by sleepable programs.5481 *	Return5482 *		0 on success, or a negative error in case of failure. On error5483 *		*dst* buffer is zeroed out.5484 *5485 * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)5486 *	Description5487 *		Change the __sk_buff->tstamp_type to *tstamp_type*5488 *		and set *tstamp* to the __sk_buff->tstamp together.5489 *5490 *		If there is no need to change the __sk_buff->tstamp_type,5491 *		the tstamp value can be directly written to __sk_buff->tstamp5492 *		instead.5493 *5494 *		BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that5495 *		will be kept during bpf_redirect_*().  A non zero5496 *		*tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO5497 *		*tstamp_type*.5498 *5499 *		A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used5500 *		with a zero *tstamp*.5501 *5502 *		Only IPv4 and IPv6 skb->protocol are supported.5503 *5504 *		This function is most useful when it needs to set a5505 *		mono delivery time to __sk_buff->tstamp and then5506 *		bpf_redirect_*() to the egress of an iface.  For example,5507 *		changing the (rcv) timestamp in __sk_buff->tstamp at5508 *		ingress to a mono delivery time and then bpf_redirect_*()5509 *		to sch_fq@phy-dev.5510 *	Return5511 *		0 on success.5512 *		**-EINVAL** for invalid input5513 *		**-EOPNOTSUPP** for unsupported protocol5514 *5515 * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)5516 *	Description5517 *		Returns a calculated IMA hash of the *file*.5518 *		If the hash is larger than *size*, then only *size*5519 *		bytes will be copied to *dst*5520 *	Return5521 *		The **hash_algo** is returned on success,5522 *		**-EOPNOTSUPP** if the hash calculation failed or **-EINVAL** if5523 *		invalid arguments are passed.5524 *5525 * void *bpf_kptr_xchg(void *dst, void *ptr)5526 *	Description5527 *		Exchange kptr at pointer *dst* with *ptr*, and return the old value.5528 *		*dst* can be map value or local kptr. *ptr* can be NULL, otherwise5529 *		it must be a referenced pointer which will be released when this helper5530 *		is called.5531 *	Return5532 *		The old value of kptr (which can be NULL). The returned pointer5533 *		if not NULL, is a reference which must be released using its5534 *		corresponding release function, or moved into a BPF map before5535 *		program exit.5536 *5537 * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)5538 * 	Description5539 * 		Perform a lookup in *percpu map* for an entry associated to5540 * 		*key* on *cpu*.5541 * 	Return5542 * 		Map value associated to *key* on *cpu*, or **NULL** if no entry5543 * 		was found or *cpu* is invalid.5544 *5545 * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)5546 *	Description5547 *		Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.5548 *	Return5549 *		*sk* if casting is valid, or **NULL** otherwise.5550 *5551 * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)5552 *	Description5553 *		Get a dynptr to local memory *data*.5554 *5555 *		*data* must be a ptr to a map value.5556 *		The maximum *size* supported is DYNPTR_MAX_SIZE.5557 *		*flags* is currently unused.5558 *	Return5559 *		0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,5560 *		-EINVAL if flags is not 0.5561 *5562 * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)5563 *	Description5564 *		Reserve *size* bytes of payload in a ring buffer *ringbuf*5565 *		through the dynptr interface. *flags* must be 0.5566 *5567 *		Please note that a corresponding bpf_ringbuf_submit_dynptr or5568 *		bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the5569 *		reservation fails. This is enforced by the verifier.5570 *	Return5571 *		0 on success, or a negative error in case of failure.5572 *5573 * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)5574 *	Description5575 *		Submit reserved ring buffer sample, pointed to by *data*,5576 *		through the dynptr interface. This is a no-op if the dynptr is5577 *		invalid/null.5578 *5579 *		For more information on *flags*, please see5580 *		'bpf_ringbuf_submit'.5581 *	Return5582 *		Nothing. Always succeeds.5583 *5584 * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)5585 *	Description5586 *		Discard reserved ring buffer sample through the dynptr5587 *		interface. This is a no-op if the dynptr is invalid/null.5588 *5589 *		For more information on *flags*, please see5590 *		'bpf_ringbuf_discard'.5591 *	Return5592 *		Nothing. Always succeeds.5593 *5594 * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)5595 *	Description5596 *		Read *len* bytes from *src* into *dst*, starting from *offset*5597 *		into *src*.5598 *		*flags* is currently unused.5599 *	Return5600 *		0 on success, -E2BIG if *offset* + *len* exceeds the length5601 *		of *src*'s data, -EINVAL if *src* is an invalid dynptr or if5602 *		*flags* is not 0.5603 *5604 * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)5605 *	Description5606 *		Write *len* bytes from *src* into *dst*, starting from *offset*5607 *		into *dst*.5608 *5609 *		*flags* must be 0 except for skb-type dynptrs.5610 *5611 *		For skb-type dynptrs:5612 *		    *  All data slices of the dynptr are automatically5613 *		       invalidated after **bpf_dynptr_write**\ (). This is5614 *		       because writing may pull the skb and change the5615 *		       underlying packet buffer.5616 *5617 *		    *  For *flags*, please see the flags accepted by5618 *		       **bpf_skb_store_bytes**\ ().5619 *	Return5620 *		0 on success, -E2BIG if *offset* + *len* exceeds the length5621 *		of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*5622 *		is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,5623 *		other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().5624 *5625 * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)5626 *	Description5627 *		Get a pointer to the underlying dynptr data.5628 *5629 *		*len* must be a statically known value. The returned data slice5630 *		is invalidated whenever the dynptr is invalidated.5631 *5632 *		skb and xdp type dynptrs may not use bpf_dynptr_data. They should5633 *		instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.5634 *	Return5635 *		Pointer to the underlying dynptr data, NULL if the dynptr is5636 *		read-only, if the dynptr is invalid, or if the offset and length5637 *		is out of bounds.5638 *5639 * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)5640 *	Description5641 *		Try to issue a SYN cookie for the packet with corresponding5642 *		IPv4/TCP headers, *iph* and *th*, without depending on a5643 *		listening socket.5644 *5645 *		*iph* points to the IPv4 header.5646 *5647 *		*th* points to the start of the TCP header, while *th_len*5648 *		contains the length of the TCP header (at least5649 *		**sizeof**\ (**struct tcphdr**)).5650 *	Return5651 *		On success, lower 32 bits hold the generated SYN cookie in5652 *		followed by 16 bits which hold the MSS value for that cookie,5653 *		and the top 16 bits are unused.5654 *5655 *		On failure, the returned value is one of the following:5656 *5657 *		**-EINVAL** if *th_len* is invalid.5658 *5659 * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)5660 *	Description5661 *		Try to issue a SYN cookie for the packet with corresponding5662 *		IPv6/TCP headers, *iph* and *th*, without depending on a5663 *		listening socket.5664 *5665 *		*iph* points to the IPv6 header.5666 *5667 *		*th* points to the start of the TCP header, while *th_len*5668 *		contains the length of the TCP header (at least5669 *		**sizeof**\ (**struct tcphdr**)).5670 *	Return5671 *		On success, lower 32 bits hold the generated SYN cookie in5672 *		followed by 16 bits which hold the MSS value for that cookie,5673 *		and the top 16 bits are unused.5674 *5675 *		On failure, the returned value is one of the following:5676 *5677 *		**-EINVAL** if *th_len* is invalid.5678 *5679 *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.5680 *5681 * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)5682 *	Description5683 *		Check whether *iph* and *th* contain a valid SYN cookie ACK5684 *		without depending on a listening socket.5685 *5686 *		*iph* points to the IPv4 header.5687 *5688 *		*th* points to the TCP header.5689 *	Return5690 *		0 if *iph* and *th* are a valid SYN cookie ACK.5691 *5692 *		On failure, the returned value is one of the following:5693 *5694 *		**-EACCES** if the SYN cookie is not valid.5695 *5696 * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)5697 *	Description5698 *		Check whether *iph* and *th* contain a valid SYN cookie ACK5699 *		without depending on a listening socket.5700 *5701 *		*iph* points to the IPv6 header.5702 *5703 *		*th* points to the TCP header.5704 *	Return5705 *		0 if *iph* and *th* are a valid SYN cookie ACK.5706 *5707 *		On failure, the returned value is one of the following:5708 *5709 *		**-EACCES** if the SYN cookie is not valid.5710 *5711 *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.5712 *5713 * u64 bpf_ktime_get_tai_ns(void)5714 *	Description5715 *		A nonsettable system-wide clock derived from wall-clock time but5716 *		ignoring leap seconds.  This clock does not experience5717 *		discontinuities and backwards jumps caused by NTP inserting leap5718 *		seconds as CLOCK_REALTIME does.5719 *5720 *		See: **clock_gettime**\ (**CLOCK_TAI**)5721 *	Return5722 *		Current *ktime*.5723 *5724 * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)5725 *	Description5726 *		Drain samples from the specified user ring buffer, and invoke5727 *		the provided callback for each such sample:5728 *5729 *		long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);5730 *5731 *		If **callback_fn** returns 0, the helper will continue to try5732 *		and drain the next sample, up to a maximum of5733 *		BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,5734 *		the helper will skip the rest of the samples and return. Other5735 *		return values are not used now, and will be rejected by the5736 *		verifier.5737 *	Return5738 *		The number of drained samples if no error was encountered while5739 *		draining samples, or 0 if no samples were present in the ring5740 *		buffer. If a user-space producer was epoll-waiting on this map,5741 *		and at least one sample was drained, they will receive an event5742 *		notification notifying them of available space in the ring5743 *		buffer. If the BPF_RB_NO_WAKEUP flag is passed to this5744 *		function, no wakeup notification will be sent. If the5745 *		BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will5746 *		be sent even if no sample was drained.5747 *5748 *		On failure, the returned value is one of the following:5749 *5750 *		**-EBUSY** if the ring buffer is contended, and another calling5751 *		context was concurrently draining the ring buffer.5752 *5753 *		**-EINVAL** if user-space is not properly tracking the ring5754 *		buffer due to the producer position not being aligned to 85755 *		bytes, a sample not being aligned to 8 bytes, or the producer5756 *		position not matching the advertised length of a sample.5757 *5758 *		**-E2BIG** if user-space has tried to publish a sample which is5759 *		larger than the size of the ring buffer, or which cannot fit5760 *		within a struct bpf_dynptr.5761 *5762 * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)5763 *	Description5764 *		Get a bpf_local_storage from the *cgroup*.5765 *5766 *		Logically, it could be thought of as getting the value from5767 *		a *map* with *cgroup* as the **key**.  From this5768 *		perspective,  the usage is not much different from5769 *		**bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this5770 *		helper enforces the key must be a cgroup struct and the map must also5771 *		be a **BPF_MAP_TYPE_CGRP_STORAGE**.5772 *5773 *		In reality, the local-storage value is embedded directly inside of the5774 *		*cgroup* object itself, rather than being located in the5775 *		**BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is5776 *		queried for some *map* on a *cgroup* object, the kernel will perform an5777 *		O(n) iteration over all of the live local-storage values for that5778 *		*cgroup* object until the local-storage value for the *map* is found.5779 *5780 *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be5781 *		used such that a new bpf_local_storage will be5782 *		created if one does not exist.  *value* can be used5783 *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify5784 *		the initial value of a bpf_local_storage.  If *value* is5785 *		**NULL**, the new bpf_local_storage will be zero initialized.5786 *	Return5787 *		A bpf_local_storage pointer is returned on success.5788 *5789 *		**NULL** if not found or there was an error in adding5790 *		a new bpf_local_storage.5791 *5792 * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)5793 *	Description5794 *		Delete a bpf_local_storage from a *cgroup*.5795 *	Return5796 *		0 on success.5797 *5798 *		**-ENOENT** if the bpf_local_storage cannot be found.5799 */5800#define ___BPF_FUNC_MAPPER(FN, ctx...)			\5801	FN(unspec, 0, ##ctx)				\5802	FN(map_lookup_elem, 1, ##ctx)			\5803	FN(map_update_elem, 2, ##ctx)			\5804	FN(map_delete_elem, 3, ##ctx)			\5805	FN(probe_read, 4, ##ctx)			\5806	FN(ktime_get_ns, 5, ##ctx)			\5807	FN(trace_printk, 6, ##ctx)			\5808	FN(get_prandom_u32, 7, ##ctx)			\5809	FN(get_smp_processor_id, 8, ##ctx)		\5810	FN(skb_store_bytes, 9, ##ctx)			\5811	FN(l3_csum_replace, 10, ##ctx)			\5812	FN(l4_csum_replace, 11, ##ctx)			\5813	FN(tail_call, 12, ##ctx)			\5814	FN(clone_redirect, 13, ##ctx)			\5815	FN(get_current_pid_tgid, 14, ##ctx)		\5816	FN(get_current_uid_gid, 15, ##ctx)		\5817	FN(get_current_comm, 16, ##ctx)			\5818	FN(get_cgroup_classid, 17, ##ctx)		\5819	FN(skb_vlan_push, 18, ##ctx)			\5820	FN(skb_vlan_pop, 19, ##ctx)			\5821	FN(skb_get_tunnel_key, 20, ##ctx)		\5822	FN(skb_set_tunnel_key, 21, ##ctx)		\5823	FN(perf_event_read, 22, ##ctx)			\5824	FN(redirect, 23, ##ctx)				\5825	FN(get_route_realm, 24, ##ctx)			\5826	FN(perf_event_output, 25, ##ctx)		\5827	FN(skb_load_bytes, 26, ##ctx)			\5828	FN(get_stackid, 27, ##ctx)			\5829	FN(csum_diff, 28, ##ctx)			\5830	FN(skb_get_tunnel_opt, 29, ##ctx)		\5831	FN(skb_set_tunnel_opt, 30, ##ctx)		\5832	FN(skb_change_proto, 31, ##ctx)			\5833	FN(skb_change_type, 32, ##ctx)			\5834	FN(skb_under_cgroup, 33, ##ctx)			\5835	FN(get_hash_recalc, 34, ##ctx)			\5836	FN(get_current_task, 35, ##ctx)			\5837	FN(probe_write_user, 36, ##ctx)			\5838	FN(current_task_under_cgroup, 37, ##ctx)	\5839	FN(skb_change_tail, 38, ##ctx)			\5840	FN(skb_pull_data, 39, ##ctx)			\5841	FN(csum_update, 40, ##ctx)			\5842	FN(set_hash_invalid, 41, ##ctx)			\5843	FN(get_numa_node_id, 42, ##ctx)			\5844	FN(skb_change_head, 43, ##ctx)			\5845	FN(xdp_adjust_head, 44, ##ctx)			\5846	FN(probe_read_str, 45, ##ctx)			\5847	FN(get_socket_cookie, 46, ##ctx)		\5848	FN(get_socket_uid, 47, ##ctx)			\5849	FN(set_hash, 48, ##ctx)				\5850	FN(setsockopt, 49, ##ctx)			\5851	FN(skb_adjust_room, 50, ##ctx)			\5852	FN(redirect_map, 51, ##ctx)			\5853	FN(sk_redirect_map, 52, ##ctx)			\5854	FN(sock_map_update, 53, ##ctx)			\5855	FN(xdp_adjust_meta, 54, ##ctx)			\5856	FN(perf_event_read_value, 55, ##ctx)		\5857	FN(perf_prog_read_value, 56, ##ctx)		\5858	FN(getsockopt, 57, ##ctx)			\5859	FN(override_return, 58, ##ctx)			\5860	FN(sock_ops_cb_flags_set, 59, ##ctx)		\5861	FN(msg_redirect_map, 60, ##ctx)			\5862	FN(msg_apply_bytes, 61, ##ctx)			\5863	FN(msg_cork_bytes, 62, ##ctx)			\5864	FN(msg_pull_data, 63, ##ctx)			\5865	FN(bind, 64, ##ctx)				\5866	FN(xdp_adjust_tail, 65, ##ctx)			\5867	FN(skb_get_xfrm_state, 66, ##ctx)		\5868	FN(get_stack, 67, ##ctx)			\5869	FN(skb_load_bytes_relative, 68, ##ctx)		\5870	FN(fib_lookup, 69, ##ctx)			\5871	FN(sock_hash_update, 70, ##ctx)			\5872	FN(msg_redirect_hash, 71, ##ctx)		\5873	FN(sk_redirect_hash, 72, ##ctx)			\5874	FN(lwt_push_encap, 73, ##ctx)			\5875	FN(lwt_seg6_store_bytes, 74, ##ctx)		\5876	FN(lwt_seg6_adjust_srh, 75, ##ctx)		\5877	FN(lwt_seg6_action, 76, ##ctx)			\5878	FN(rc_repeat, 77, ##ctx)			\5879	FN(rc_keydown, 78, ##ctx)			\5880	FN(skb_cgroup_id, 79, ##ctx)			\5881	FN(get_current_cgroup_id, 80, ##ctx)		\5882	FN(get_local_storage, 81, ##ctx)		\5883	FN(sk_select_reuseport, 82, ##ctx)		\5884	FN(skb_ancestor_cgroup_id, 83, ##ctx)		\5885	FN(sk_lookup_tcp, 84, ##ctx)			\5886	FN(sk_lookup_udp, 85, ##ctx)			\5887	FN(sk_release, 86, ##ctx)			\5888	FN(map_push_elem, 87, ##ctx)			\5889	FN(map_pop_elem, 88, ##ctx)			\5890	FN(map_peek_elem, 89, ##ctx)			\5891	FN(msg_push_data, 90, ##ctx)			\5892	FN(msg_pop_data, 91, ##ctx)			\5893	FN(rc_pointer_rel, 92, ##ctx)			\5894	FN(spin_lock, 93, ##ctx)			\5895	FN(spin_unlock, 94, ##ctx)			\5896	FN(sk_fullsock, 95, ##ctx)			\5897	FN(tcp_sock, 96, ##ctx)				\5898	FN(skb_ecn_set_ce, 97, ##ctx)			\5899	FN(get_listener_sock, 98, ##ctx)		\5900	FN(skc_lookup_tcp, 99, ##ctx)			\5901	FN(tcp_check_syncookie, 100, ##ctx)		\5902	FN(sysctl_get_name, 101, ##ctx)			\5903	FN(sysctl_get_current_value, 102, ##ctx)	\5904	FN(sysctl_get_new_value, 103, ##ctx)		\5905	FN(sysctl_set_new_value, 104, ##ctx)		\5906	FN(strtol, 105, ##ctx)				\5907	FN(strtoul, 106, ##ctx)				\5908	FN(sk_storage_get, 107, ##ctx)			\5909	FN(sk_storage_delete, 108, ##ctx)		\5910	FN(send_signal, 109, ##ctx)			\5911	FN(tcp_gen_syncookie, 110, ##ctx)		\5912	FN(skb_output, 111, ##ctx)			\5913	FN(probe_read_user, 112, ##ctx)			\5914	FN(probe_read_kernel, 113, ##ctx)		\5915	FN(probe_read_user_str, 114, ##ctx)		\5916	FN(probe_read_kernel_str, 115, ##ctx)		\5917	FN(tcp_send_ack, 116, ##ctx)			\5918	FN(send_signal_thread, 117, ##ctx)		\5919	FN(jiffies64, 118, ##ctx)			\5920	FN(read_branch_records, 119, ##ctx)		\5921	FN(get_ns_current_pid_tgid, 120, ##ctx)		\5922	FN(xdp_output, 121, ##ctx)			\5923	FN(get_netns_cookie, 122, ##ctx)		\5924	FN(get_current_ancestor_cgroup_id, 123, ##ctx)	\5925	FN(sk_assign, 124, ##ctx)			\5926	FN(ktime_get_boot_ns, 125, ##ctx)		\5927	FN(seq_printf, 126, ##ctx)			\5928	FN(seq_write, 127, ##ctx)			\5929	FN(sk_cgroup_id, 128, ##ctx)			\5930	FN(sk_ancestor_cgroup_id, 129, ##ctx)		\5931	FN(ringbuf_output, 130, ##ctx)			\5932	FN(ringbuf_reserve, 131, ##ctx)			\5933	FN(ringbuf_submit, 132, ##ctx)			\5934	FN(ringbuf_discard, 133, ##ctx)			\5935	FN(ringbuf_query, 134, ##ctx)			\5936	FN(csum_level, 135, ##ctx)			\5937	FN(skc_to_tcp6_sock, 136, ##ctx)		\5938	FN(skc_to_tcp_sock, 137, ##ctx)			\5939	FN(skc_to_tcp_timewait_sock, 138, ##ctx)	\5940	FN(skc_to_tcp_request_sock, 139, ##ctx)		\5941	FN(skc_to_udp6_sock, 140, ##ctx)		\5942	FN(get_task_stack, 141, ##ctx)			\5943	FN(load_hdr_opt, 142, ##ctx)			\5944	FN(store_hdr_opt, 143, ##ctx)			\5945	FN(reserve_hdr_opt, 144, ##ctx)			\5946	FN(inode_storage_get, 145, ##ctx)		\5947	FN(inode_storage_delete, 146, ##ctx)		\5948	FN(d_path, 147, ##ctx)				\5949	FN(copy_from_user, 148, ##ctx)			\5950	FN(snprintf_btf, 149, ##ctx)			\5951	FN(seq_printf_btf, 150, ##ctx)			\5952	FN(skb_cgroup_classid, 151, ##ctx)		\5953	FN(redirect_neigh, 152, ##ctx)			\5954	FN(per_cpu_ptr, 153, ##ctx)			\5955	FN(this_cpu_ptr, 154, ##ctx)			\5956	FN(redirect_peer, 155, ##ctx)			\5957	FN(task_storage_get, 156, ##ctx)		\5958	FN(task_storage_delete, 157, ##ctx)		\5959	FN(get_current_task_btf, 158, ##ctx)		\5960	FN(bprm_opts_set, 159, ##ctx)			\5961	FN(ktime_get_coarse_ns, 160, ##ctx)		\5962	FN(ima_inode_hash, 161, ##ctx)			\5963	FN(sock_from_file, 162, ##ctx)			\5964	FN(check_mtu, 163, ##ctx)			\5965	FN(for_each_map_elem, 164, ##ctx)		\5966	FN(snprintf, 165, ##ctx)			\5967	FN(sys_bpf, 166, ##ctx)				\5968	FN(btf_find_by_name_kind, 167, ##ctx)		\5969	FN(sys_close, 168, ##ctx)			\5970	FN(timer_init, 169, ##ctx)			\5971	FN(timer_set_callback, 170, ##ctx)		\5972	FN(timer_start, 171, ##ctx)			\5973	FN(timer_cancel, 172, ##ctx)			\5974	FN(get_func_ip, 173, ##ctx)			\5975	FN(get_attach_cookie, 174, ##ctx)		\5976	FN(task_pt_regs, 175, ##ctx)			\5977	FN(get_branch_snapshot, 176, ##ctx)		\5978	FN(trace_vprintk, 177, ##ctx)			\5979	FN(skc_to_unix_sock, 178, ##ctx)		\5980	FN(kallsyms_lookup_name, 179, ##ctx)		\5981	FN(find_vma, 180, ##ctx)			\5982	FN(loop, 181, ##ctx)				\5983	FN(strncmp, 182, ##ctx)				\5984	FN(get_func_arg, 183, ##ctx)			\5985	FN(get_func_ret, 184, ##ctx)			\5986	FN(get_func_arg_cnt, 185, ##ctx)		\5987	FN(get_retval, 186, ##ctx)			\5988	FN(set_retval, 187, ##ctx)			\5989	FN(xdp_get_buff_len, 188, ##ctx)		\5990	FN(xdp_load_bytes, 189, ##ctx)			\5991	FN(xdp_store_bytes, 190, ##ctx)			\5992	FN(copy_from_user_task, 191, ##ctx)		\5993	FN(skb_set_tstamp, 192, ##ctx)			\5994	FN(ima_file_hash, 193, ##ctx)			\5995	FN(kptr_xchg, 194, ##ctx)			\5996	FN(map_lookup_percpu_elem, 195, ##ctx)		\5997	FN(skc_to_mptcp_sock, 196, ##ctx)		\5998	FN(dynptr_from_mem, 197, ##ctx)			\5999	FN(ringbuf_reserve_dynptr, 198, ##ctx)		\6000	FN(ringbuf_submit_dynptr, 199, ##ctx)		\6001	FN(ringbuf_discard_dynptr, 200, ##ctx)		\6002	FN(dynptr_read, 201, ##ctx)			\6003	FN(dynptr_write, 202, ##ctx)			\6004	FN(dynptr_data, 203, ##ctx)			\6005	FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx)	\6006	FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx)	\6007	FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx)	\6008	FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)	\6009	FN(ktime_get_tai_ns, 208, ##ctx)		\6010	FN(user_ringbuf_drain, 209, ##ctx)		\6011	FN(cgrp_storage_get, 210, ##ctx)		\6012	FN(cgrp_storage_delete, 211, ##ctx)		\6013	/* */6014 6015/* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't6016 * know or care about integer value that is now passed as second argument6017 */6018#define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),6019#define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)6020 6021/* integer value in 'imm' field of BPF_CALL instruction selects which helper6022 * function eBPF program intends to call6023 */6024#define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,6025enum bpf_func_id {6026	___BPF_FUNC_MAPPER(__BPF_ENUM_FN)6027	__BPF_FUNC_MAX_ID,6028};6029#undef __BPF_ENUM_FN6030 6031/* All flags used by eBPF helper functions, placed here. */6032 6033/* BPF_FUNC_skb_store_bytes flags. */6034enum {6035	BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),6036	BPF_F_INVALIDATE_HASH		= (1ULL << 1),6037};6038 6039/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.6040 * First 4 bits are for passing the header field size.6041 */6042enum {6043	BPF_F_HDR_FIELD_MASK		= 0xfULL,6044};6045 6046/* BPF_FUNC_l4_csum_replace flags. */6047enum {6048	BPF_F_PSEUDO_HDR		= (1ULL << 4),6049	BPF_F_MARK_MANGLED_0		= (1ULL << 5),6050	BPF_F_MARK_ENFORCE		= (1ULL << 6),6051};6052 6053/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */6054enum {6055	BPF_F_TUNINFO_IPV6		= (1ULL << 0),6056};6057 6058/* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */6059enum {6060	BPF_F_SKIP_FIELD_MASK		= 0xffULL,6061	BPF_F_USER_STACK		= (1ULL << 8),6062/* flags used by BPF_FUNC_get_stackid only. */6063	BPF_F_FAST_STACK_CMP		= (1ULL << 9),6064	BPF_F_REUSE_STACKID		= (1ULL << 10),6065/* flags used by BPF_FUNC_get_stack only. */6066	BPF_F_USER_BUILD_ID		= (1ULL << 11),6067};6068 6069/* BPF_FUNC_skb_set_tunnel_key flags. */6070enum {6071	BPF_F_ZERO_CSUM_TX		= (1ULL << 1),6072	BPF_F_DONT_FRAGMENT		= (1ULL << 2),6073	BPF_F_SEQ_NUMBER		= (1ULL << 3),6074	BPF_F_NO_TUNNEL_KEY		= (1ULL << 4),6075};6076 6077/* BPF_FUNC_skb_get_tunnel_key flags. */6078enum {6079	BPF_F_TUNINFO_FLAGS		= (1ULL << 4),6080};6081 6082/* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and6083 * BPF_FUNC_perf_event_read_value flags.6084 */6085enum {6086	BPF_F_INDEX_MASK		= 0xffffffffULL,6087	BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,6088/* BPF_FUNC_perf_event_output for sk_buff input context. */6089	BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),6090};6091 6092/* Current network namespace */6093enum {6094	BPF_F_CURRENT_NETNS		= (-1L),6095};6096 6097/* BPF_FUNC_csum_level level values. */6098enum {6099	BPF_CSUM_LEVEL_QUERY,6100	BPF_CSUM_LEVEL_INC,6101	BPF_CSUM_LEVEL_DEC,6102	BPF_CSUM_LEVEL_RESET,6103};6104 6105/* BPF_FUNC_skb_adjust_room flags. */6106enum {6107	BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),6108	BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),6109	BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),6110	BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),6111	BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),6112	BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),6113	BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),6114	BPF_F_ADJ_ROOM_DECAP_L3_IPV4	= (1ULL << 7),6115	BPF_F_ADJ_ROOM_DECAP_L3_IPV6	= (1ULL << 8),6116};6117 6118enum {6119	BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,6120	BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,6121};6122 6123#define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \6124					  BPF_ADJ_ROOM_ENCAP_L2_MASK) \6125					 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)6126 6127/* BPF_FUNC_sysctl_get_name flags. */6128enum {6129	BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),6130};6131 6132/* BPF_FUNC_<kernel_obj>_storage_get flags */6133enum {6134	BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),6135	/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility6136	 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.6137	 */6138	BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,6139};6140 6141/* BPF_FUNC_read_branch_records flags. */6142enum {6143	BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),6144};6145 6146/* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and6147 * BPF_FUNC_bpf_ringbuf_output flags.6148 */6149enum {6150	BPF_RB_NO_WAKEUP		= (1ULL << 0),6151	BPF_RB_FORCE_WAKEUP		= (1ULL << 1),6152};6153 6154/* BPF_FUNC_bpf_ringbuf_query flags */6155enum {6156	BPF_RB_AVAIL_DATA = 0,6157	BPF_RB_RING_SIZE = 1,6158	BPF_RB_CONS_POS = 2,6159	BPF_RB_PROD_POS = 3,6160};6161 6162/* BPF ring buffer constants */6163enum {6164	BPF_RINGBUF_BUSY_BIT		= (1U << 31),6165	BPF_RINGBUF_DISCARD_BIT		= (1U << 30),6166	BPF_RINGBUF_HDR_SZ		= 8,6167};6168 6169/* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */6170enum {6171	BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),6172	BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),6173};6174 6175/* Mode for BPF_FUNC_skb_adjust_room helper. */6176enum bpf_adj_room_mode {6177	BPF_ADJ_ROOM_NET,6178	BPF_ADJ_ROOM_MAC,6179};6180 6181/* Mode for BPF_FUNC_skb_load_bytes_relative helper. */6182enum bpf_hdr_start_off {6183	BPF_HDR_START_MAC,6184	BPF_HDR_START_NET,6185};6186 6187/* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */6188enum bpf_lwt_encap_mode {6189	BPF_LWT_ENCAP_SEG6,6190	BPF_LWT_ENCAP_SEG6_INLINE,6191	BPF_LWT_ENCAP_IP,6192};6193 6194/* Flags for bpf_bprm_opts_set helper */6195enum {6196	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),6197};6198 6199/* Flags for bpf_redirect and bpf_redirect_map helpers */6200enum {6201	BPF_F_INGRESS		= (1ULL << 0), /* used for skb path */6202	BPF_F_BROADCAST		= (1ULL << 3), /* used for XDP path */6203	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4), /* used for XDP path */6204#define BPF_F_REDIRECT_FLAGS (BPF_F_INGRESS | BPF_F_BROADCAST | BPF_F_EXCLUDE_INGRESS)6205};6206 6207#define __bpf_md_ptr(type, name)	\6208union {					\6209	type name;			\6210	__u64 :64;			\6211} __attribute__((aligned(8)))6212 6213/* The enum used in skb->tstamp_type. It specifies the clock type6214 * of the time stored in the skb->tstamp.6215 */6216enum {6217	BPF_SKB_TSTAMP_UNSPEC = 0,		/* DEPRECATED */6218	BPF_SKB_TSTAMP_DELIVERY_MONO = 1,	/* DEPRECATED */6219	BPF_SKB_CLOCK_REALTIME = 0,6220	BPF_SKB_CLOCK_MONOTONIC = 1,6221	BPF_SKB_CLOCK_TAI = 2,6222	/* For any future BPF_SKB_CLOCK_* that the bpf prog cannot handle,6223	 * the bpf prog can try to deduce it by ingress/egress/skb->sk->sk_clockid.6224	 */6225};6226 6227/* user accessible mirror of in-kernel sk_buff.6228 * new fields can only be added to the end of this structure6229 */6230struct __sk_buff {6231	__u32 len;6232	__u32 pkt_type;6233	__u32 mark;6234	__u32 queue_mapping;6235	__u32 protocol;6236	__u32 vlan_present;6237	__u32 vlan_tci;6238	__u32 vlan_proto;6239	__u32 priority;6240	__u32 ingress_ifindex;6241	__u32 ifindex;6242	__u32 tc_index;6243	__u32 cb[5];6244	__u32 hash;6245	__u32 tc_classid;6246	__u32 data;6247	__u32 data_end;6248	__u32 napi_id;6249 6250	/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */6251	__u32 family;6252	__u32 remote_ip4;	/* Stored in network byte order */6253	__u32 local_ip4;	/* Stored in network byte order */6254	__u32 remote_ip6[4];	/* Stored in network byte order */6255	__u32 local_ip6[4];	/* Stored in network byte order */6256	__u32 remote_port;	/* Stored in network byte order */6257	__u32 local_port;	/* stored in host byte order */6258	/* ... here. */6259 6260	__u32 data_meta;6261	__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);6262	__u64 tstamp;6263	__u32 wire_len;6264	__u32 gso_segs;6265	__bpf_md_ptr(struct bpf_sock *, sk);6266	__u32 gso_size;6267	__u8  tstamp_type;6268	__u32 :24;		/* Padding, future use. */6269	__u64 hwtstamp;6270};6271 6272struct bpf_tunnel_key {6273	__u32 tunnel_id;6274	union {6275		__u32 remote_ipv4;6276		__u32 remote_ipv6[4];6277	};6278	__u8 tunnel_tos;6279	__u8 tunnel_ttl;6280	union {6281		__u16 tunnel_ext;	/* compat */6282		__be16 tunnel_flags;6283	};6284	__u32 tunnel_label;6285	union {6286		__u32 local_ipv4;6287		__u32 local_ipv6[4];6288	};6289};6290 6291/* user accessible mirror of in-kernel xfrm_state.6292 * new fields can only be added to the end of this structure6293 */6294struct bpf_xfrm_state {6295	__u32 reqid;6296	__u32 spi;	/* Stored in network byte order */6297	__u16 family;6298	__u16 ext;	/* Padding, future use. */6299	union {6300		__u32 remote_ipv4;	/* Stored in network byte order */6301		__u32 remote_ipv6[4];	/* Stored in network byte order */6302	};6303};6304 6305/* Generic BPF return codes which all BPF program types may support.6306 * The values are binary compatible with their TC_ACT_* counter-part to6307 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT6308 * programs.6309 *6310 * XDP is handled seprately, see XDP_*.6311 */6312enum bpf_ret_code {6313	BPF_OK = 0,6314	/* 1 reserved */6315	BPF_DROP = 2,6316	/* 3-6 reserved */6317	BPF_REDIRECT = 7,6318	/* >127 are reserved for prog type specific return codes.6319	 *6320	 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and6321	 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been6322	 *    changed and should be routed based on its new L3 header.6323	 *    (This is an L3 redirect, as opposed to L2 redirect6324	 *    represented by BPF_REDIRECT above).6325	 */6326	BPF_LWT_REROUTE = 128,6327	/* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR6328	 *   to indicate that no custom dissection was performed, and6329	 *   fallback to standard dissector is requested.6330	 */6331	BPF_FLOW_DISSECTOR_CONTINUE = 129,6332};6333 6334struct bpf_sock {6335	__u32 bound_dev_if;6336	__u32 family;6337	__u32 type;6338	__u32 protocol;6339	__u32 mark;6340	__u32 priority;6341	/* IP address also allows 1 and 2 bytes access */6342	__u32 src_ip4;6343	__u32 src_ip6[4];6344	__u32 src_port;		/* host byte order */6345	__be16 dst_port;	/* network byte order */6346	__u16 :16;		/* zero padding */6347	__u32 dst_ip4;6348	__u32 dst_ip6[4];6349	__u32 state;6350	__s32 rx_queue_mapping;6351};6352 6353struct bpf_tcp_sock {6354	__u32 snd_cwnd;		/* Sending congestion window		*/6355	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */6356	__u32 rtt_min;6357	__u32 snd_ssthresh;	/* Slow start size threshold		*/6358	__u32 rcv_nxt;		/* What we want to receive next		*/6359	__u32 snd_nxt;		/* Next sequence we send		*/6360	__u32 snd_una;		/* First byte we want an ack for	*/6361	__u32 mss_cache;	/* Cached effective mss, not including SACKS */6362	__u32 ecn_flags;	/* ECN status bits.			*/6363	__u32 rate_delivered;	/* saved rate sample: packets delivered */6364	__u32 rate_interval_us;	/* saved rate sample: time elapsed */6365	__u32 packets_out;	/* Packets which are "in flight"	*/6366	__u32 retrans_out;	/* Retransmitted packets out		*/6367	__u32 total_retrans;	/* Total retransmits for entire connection */6368	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn6369				 * total number of segments in.6370				 */6371	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn6372				 * total number of data segments in.6373				 */6374	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut6375				 * The total number of segments sent.6376				 */6377	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut6378				 * total number of data segments sent.6379				 */6380	__u32 lost_out;		/* Lost packets			*/6381	__u32 sacked_out;	/* SACK'd packets			*/6382	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived6383				 * sum(delta(rcv_nxt)), or how many bytes6384				 * were acked.6385				 */6386	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked6387				 * sum(delta(snd_una)), or how many bytes6388				 * were acked.6389				 */6390	__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups6391				 * total number of DSACK blocks received6392				 */6393	__u32 delivered;	/* Total data packets delivered incl. rexmits */6394	__u32 delivered_ce;	/* Like the above but only ECE marked packets */6395	__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */6396};6397 6398struct bpf_sock_tuple {6399	union {6400		struct {6401			__be32 saddr;6402			__be32 daddr;6403			__be16 sport;6404			__be16 dport;6405		} ipv4;6406		struct {6407			__be32 saddr[4];6408			__be32 daddr[4];6409			__be16 sport;6410			__be16 dport;6411		} ipv6;6412	};6413};6414 6415/* (Simplified) user return codes for tcx prog type.6416 * A valid tcx program must return one of these defined values. All other6417 * return codes are reserved for future use. Must remain compatible with6418 * their TC_ACT_* counter-parts. For compatibility in behavior, unknown6419 * return codes are mapped to TCX_NEXT.6420 */6421enum tcx_action_base {6422	TCX_NEXT	= -1,6423	TCX_PASS	= 0,6424	TCX_DROP	= 2,6425	TCX_REDIRECT	= 7,6426};6427 6428struct bpf_xdp_sock {6429	__u32 queue_id;6430};6431 6432#define XDP_PACKET_HEADROOM 2566433 6434/* User return codes for XDP prog type.6435 * A valid XDP program must return one of these defined values. All other6436 * return codes are reserved for future use. Unknown return codes will6437 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().6438 */6439enum xdp_action {6440	XDP_ABORTED = 0,6441	XDP_DROP,6442	XDP_PASS,6443	XDP_TX,6444	XDP_REDIRECT,6445};6446 6447/* user accessible metadata for XDP packet hook6448 * new fields must be added to the end of this structure6449 */6450struct xdp_md {6451	__u32 data;6452	__u32 data_end;6453	__u32 data_meta;6454	/* Below access go through struct xdp_rxq_info */6455	__u32 ingress_ifindex; /* rxq->dev->ifindex */6456	__u32 rx_queue_index;  /* rxq->queue_index  */6457 6458	__u32 egress_ifindex;  /* txq->dev->ifindex */6459};6460 6461/* DEVMAP map-value layout6462 *6463 * The struct data-layout of map-value is a configuration interface.6464 * New members can only be added to the end of this structure.6465 */6466struct bpf_devmap_val {6467	__u32 ifindex;   /* device index */6468	union {6469		int   fd;  /* prog fd on map write */6470		__u32 id;  /* prog id on map read */6471	} bpf_prog;6472};6473 6474/* CPUMAP map-value layout6475 *6476 * The struct data-layout of map-value is a configuration interface.6477 * New members can only be added to the end of this structure.6478 */6479struct bpf_cpumap_val {6480	__u32 qsize;	/* queue size to remote target CPU */6481	union {6482		int   fd;	/* prog fd on map write */6483		__u32 id;	/* prog id on map read */6484	} bpf_prog;6485};6486 6487enum sk_action {6488	SK_DROP = 0,6489	SK_PASS,6490};6491 6492/* user accessible metadata for SK_MSG packet hook, new fields must6493 * be added to the end of this structure6494 */6495struct sk_msg_md {6496	__bpf_md_ptr(void *, data);6497	__bpf_md_ptr(void *, data_end);6498 6499	__u32 family;6500	__u32 remote_ip4;	/* Stored in network byte order */6501	__u32 local_ip4;	/* Stored in network byte order */6502	__u32 remote_ip6[4];	/* Stored in network byte order */6503	__u32 local_ip6[4];	/* Stored in network byte order */6504	__u32 remote_port;	/* Stored in network byte order */6505	__u32 local_port;	/* stored in host byte order */6506	__u32 size;		/* Total size of sk_msg */6507 6508	__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */6509};6510 6511struct sk_reuseport_md {6512	/*6513	 * Start of directly accessible data. It begins from6514	 * the tcp/udp header.6515	 */6516	__bpf_md_ptr(void *, data);6517	/* End of directly accessible data */6518	__bpf_md_ptr(void *, data_end);6519	/*6520	 * Total length of packet (starting from the tcp/udp header).6521	 * Note that the directly accessible bytes (data_end - data)6522	 * could be less than this "len".  Those bytes could be6523	 * indirectly read by a helper "bpf_skb_load_bytes()".6524	 */6525	__u32 len;6526	/*6527	 * Eth protocol in the mac header (network byte order). e.g.6528	 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)6529	 */6530	__u32 eth_protocol;6531	__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */6532	__u32 bind_inany;	/* Is sock bound to an INANY address? */6533	__u32 hash;		/* A hash of the packet 4 tuples */6534	/* When reuse->migrating_sk is NULL, it is selecting a sk for the6535	 * new incoming connection request (e.g. selecting a listen sk for6536	 * the received SYN in the TCP case).  reuse->sk is one of the sk6537	 * in the reuseport group. The bpf prog can use reuse->sk to learn6538	 * the local listening ip/port without looking into the skb.6539	 *6540	 * When reuse->migrating_sk is not NULL, reuse->sk is closed and6541	 * reuse->migrating_sk is the socket that needs to be migrated6542	 * to another listening socket.  migrating_sk could be a fullsock6543	 * sk that is fully established or a reqsk that is in-the-middle6544	 * of 3-way handshake.6545	 */6546	__bpf_md_ptr(struct bpf_sock *, sk);6547	__bpf_md_ptr(struct bpf_sock *, migrating_sk);6548};6549 6550#define BPF_TAG_SIZE	86551 6552struct bpf_prog_info {6553	__u32 type;6554	__u32 id;6555	__u8  tag[BPF_TAG_SIZE];6556	__u32 jited_prog_len;6557	__u32 xlated_prog_len;6558	__aligned_u64 jited_prog_insns;6559	__aligned_u64 xlated_prog_insns;6560	__u64 load_time;	/* ns since boottime */6561	__u32 created_by_uid;6562	__u32 nr_map_ids;6563	__aligned_u64 map_ids;6564	char name[BPF_OBJ_NAME_LEN];6565	__u32 ifindex;6566	__u32 gpl_compatible:1;6567	__u32 :31; /* alignment pad */6568	__u64 netns_dev;6569	__u64 netns_ino;6570	__u32 nr_jited_ksyms;6571	__u32 nr_jited_func_lens;6572	__aligned_u64 jited_ksyms;6573	__aligned_u64 jited_func_lens;6574	__u32 btf_id;6575	__u32 func_info_rec_size;6576	__aligned_u64 func_info;6577	__u32 nr_func_info;6578	__u32 nr_line_info;6579	__aligned_u64 line_info;6580	__aligned_u64 jited_line_info;6581	__u32 nr_jited_line_info;6582	__u32 line_info_rec_size;6583	__u32 jited_line_info_rec_size;6584	__u32 nr_prog_tags;6585	__aligned_u64 prog_tags;6586	__u64 run_time_ns;6587	__u64 run_cnt;6588	__u64 recursion_misses;6589	__u32 verified_insns;6590	__u32 attach_btf_obj_id;6591	__u32 attach_btf_id;6592} __attribute__((aligned(8)));6593 6594struct bpf_map_info {6595	__u32 type;6596	__u32 id;6597	__u32 key_size;6598	__u32 value_size;6599	__u32 max_entries;6600	__u32 map_flags;6601	char  name[BPF_OBJ_NAME_LEN];6602	__u32 ifindex;6603	__u32 btf_vmlinux_value_type_id;6604	__u64 netns_dev;6605	__u64 netns_ino;6606	__u32 btf_id;6607	__u32 btf_key_type_id;6608	__u32 btf_value_type_id;6609	__u32 btf_vmlinux_id;6610	__u64 map_extra;6611} __attribute__((aligned(8)));6612 6613struct bpf_btf_info {6614	__aligned_u64 btf;6615	__u32 btf_size;6616	__u32 id;6617	__aligned_u64 name;6618	__u32 name_len;6619	__u32 kernel_btf;6620} __attribute__((aligned(8)));6621 6622struct bpf_link_info {6623	__u32 type;6624	__u32 id;6625	__u32 prog_id;6626	union {6627		struct {6628			__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */6629			__u32 tp_name_len;     /* in/out: tp_name buffer len */6630		} raw_tracepoint;6631		struct {6632			__u32 attach_type;6633			__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */6634			__u32 target_btf_id; /* BTF type id inside the object */6635		} tracing;6636		struct {6637			__u64 cgroup_id;6638			__u32 attach_type;6639		} cgroup;6640		struct {6641			__aligned_u64 target_name; /* in/out: target_name buffer ptr */6642			__u32 target_name_len;	   /* in/out: target_name buffer len */6643 6644			/* If the iter specific field is 32 bits, it can be put6645			 * in the first or second union. Otherwise it should be6646			 * put in the second union.6647			 */6648			union {6649				struct {6650					__u32 map_id;6651				} map;6652			};6653			union {6654				struct {6655					__u64 cgroup_id;6656					__u32 order;6657				} cgroup;6658				struct {6659					__u32 tid;6660					__u32 pid;6661				} task;6662			};6663		} iter;6664		struct  {6665			__u32 netns_ino;6666			__u32 attach_type;6667		} netns;6668		struct {6669			__u32 ifindex;6670		} xdp;6671		struct {6672			__u32 map_id;6673		} struct_ops;6674		struct {6675			__u32 pf;6676			__u32 hooknum;6677			__s32 priority;6678			__u32 flags;6679		} netfilter;6680		struct {6681			__aligned_u64 addrs;6682			__u32 count; /* in/out: kprobe_multi function count */6683			__u32 flags;6684			__u64 missed;6685			__aligned_u64 cookies;6686		} kprobe_multi;6687		struct {6688			__aligned_u64 path;6689			__aligned_u64 offsets;6690			__aligned_u64 ref_ctr_offsets;6691			__aligned_u64 cookies;6692			__u32 path_size; /* in/out: real path size on success, including zero byte */6693			__u32 count; /* in/out: uprobe_multi offsets/ref_ctr_offsets/cookies count */6694			__u32 flags;6695			__u32 pid;6696		} uprobe_multi;6697		struct {6698			__u32 type; /* enum bpf_perf_event_type */6699			__u32 :32;6700			union {6701				struct {6702					__aligned_u64 file_name; /* in/out */6703					__u32 name_len;6704					__u32 offset; /* offset from file_name */6705					__u64 cookie;6706				} uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */6707				struct {6708					__aligned_u64 func_name; /* in/out */6709					__u32 name_len;6710					__u32 offset; /* offset from func_name */6711					__u64 addr;6712					__u64 missed;6713					__u64 cookie;6714				} kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */6715				struct {6716					__aligned_u64 tp_name;   /* in/out */6717					__u32 name_len;6718					__u32 :32;6719					__u64 cookie;6720				} tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */6721				struct {6722					__u64 config;6723					__u32 type;6724					__u32 :32;6725					__u64 cookie;6726				} event; /* BPF_PERF_EVENT_EVENT */6727			};6728		} perf_event;6729		struct {6730			__u32 ifindex;6731			__u32 attach_type;6732		} tcx;6733		struct {6734			__u32 ifindex;6735			__u32 attach_type;6736		} netkit;6737		struct {6738			__u32 map_id;6739			__u32 attach_type;6740		} sockmap;6741	};6742} __attribute__((aligned(8)));6743 6744/* User bpf_sock_addr struct to access socket fields and sockaddr struct passed6745 * by user and intended to be used by socket (e.g. to bind to, depends on6746 * attach type).6747 */6748struct bpf_sock_addr {6749	__u32 user_family;	/* Allows 4-byte read, but no write. */6750	__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.6751				 * Stored in network byte order.6752				 */6753	__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.6754				 * Stored in network byte order.6755				 */6756	__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.6757				 * Stored in network byte order6758				 */6759	__u32 family;		/* Allows 4-byte read, but no write */6760	__u32 type;		/* Allows 4-byte read, but no write */6761	__u32 protocol;		/* Allows 4-byte read, but no write */6762	__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.6763				 * Stored in network byte order.6764				 */6765	__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.6766				 * Stored in network byte order.6767				 */6768	__bpf_md_ptr(struct bpf_sock *, sk);6769};6770 6771/* User bpf_sock_ops struct to access socket values and specify request ops6772 * and their replies.6773 * Some of this fields are in network (bigendian) byte order and may need6774 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).6775 * New fields can only be added at the end of this structure6776 */6777struct bpf_sock_ops {6778	__u32 op;6779	union {6780		__u32 args[4];		/* Optionally passed to bpf program */6781		__u32 reply;		/* Returned by bpf program	    */6782		__u32 replylong[4];	/* Optionally returned by bpf prog  */6783	};6784	__u32 family;6785	__u32 remote_ip4;	/* Stored in network byte order */6786	__u32 local_ip4;	/* Stored in network byte order */6787	__u32 remote_ip6[4];	/* Stored in network byte order */6788	__u32 local_ip6[4];	/* Stored in network byte order */6789	__u32 remote_port;	/* Stored in network byte order */6790	__u32 local_port;	/* stored in host byte order */6791	__u32 is_fullsock;	/* Some TCP fields are only valid if6792				 * there is a full socket. If not, the6793				 * fields read as zero.6794				 */6795	__u32 snd_cwnd;6796	__u32 srtt_us;		/* Averaged RTT << 3 in usecs */6797	__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */6798	__u32 state;6799	__u32 rtt_min;6800	__u32 snd_ssthresh;6801	__u32 rcv_nxt;6802	__u32 snd_nxt;6803	__u32 snd_una;6804	__u32 mss_cache;6805	__u32 ecn_flags;6806	__u32 rate_delivered;6807	__u32 rate_interval_us;6808	__u32 packets_out;6809	__u32 retrans_out;6810	__u32 total_retrans;6811	__u32 segs_in;6812	__u32 data_segs_in;6813	__u32 segs_out;6814	__u32 data_segs_out;6815	__u32 lost_out;6816	__u32 sacked_out;6817	__u32 sk_txhash;6818	__u64 bytes_received;6819	__u64 bytes_acked;6820	__bpf_md_ptr(struct bpf_sock *, sk);6821	/* [skb_data, skb_data_end) covers the whole TCP header.6822	 *6823	 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received6824	 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the6825	 *                                header has not been written.6826	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have6827	 *				  been written so far.6828	 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes6829	 *					the 3WHS.6830	 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes6831	 *					the 3WHS.6832	 *6833	 * bpf_load_hdr_opt() can also be used to read a particular option.6834	 */6835	__bpf_md_ptr(void *, skb_data);6836	__bpf_md_ptr(void *, skb_data_end);6837	__u32 skb_len;		/* The total length of a packet.6838				 * It includes the header, options,6839				 * and payload.6840				 */6841	__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides6842				 * an easy way to check for tcp_flags6843				 * without parsing skb_data.6844				 *6845				 * In particular, the skb_tcp_flags6846				 * will still be available in6847				 * BPF_SOCK_OPS_HDR_OPT_LEN even though6848				 * the outgoing header has not6849				 * been written yet.6850				 */6851	__u64 skb_hwtstamp;6852};6853 6854/* Definitions for bpf_sock_ops_cb_flags */6855enum {6856	BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),6857	BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),6858	BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),6859	BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),6860	/* Call bpf for all received TCP headers.  The bpf prog will be6861	 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB6862	 *6863	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB6864	 * for the header option related helpers that will be useful6865	 * to the bpf programs.6866	 *6867	 * It could be used at the client/active side (i.e. connect() side)6868	 * when the server told it that the server was in syncookie6869	 * mode and required the active side to resend the bpf-written6870	 * options.  The active side can keep writing the bpf-options until6871	 * it received a valid packet from the server side to confirm6872	 * the earlier packet (and options) has been received.  The later6873	 * example patch is using it like this at the active side when the6874	 * server is in syncookie mode.6875	 *6876	 * The bpf prog will usually turn this off in the common cases.6877	 */6878	BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),6879	/* Call bpf when kernel has received a header option that6880	 * the kernel cannot handle.  The bpf prog will be called under6881	 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.6882	 *6883	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB6884	 * for the header option related helpers that will be useful6885	 * to the bpf programs.6886	 */6887	BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),6888	/* Call bpf when the kernel is writing header options for the6889	 * outgoing packet.  The bpf prog will first be called6890	 * to reserve space in a skb under6891	 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then6892	 * the bpf prog will be called to write the header option(s)6893	 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.6894	 *6895	 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB6896	 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option6897	 * related helpers that will be useful to the bpf programs.6898	 *6899	 * The kernel gets its chance to reserve space and write6900	 * options first before the BPF program does.6901	 */6902	BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),6903/* Mask of all currently supported cb flags */6904	BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,6905};6906 6907/* List of known BPF sock_ops operators.6908 * New entries can only be added at the end6909 */6910enum {6911	BPF_SOCK_OPS_VOID,6912	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or6913					 * -1 if default value should be used6914					 */6915	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized6916					 * window (in packets) or -1 if default6917					 * value should be used6918					 */6919	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an6920					 * active connection is initialized6921					 */6922	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an6923						 * active connection is6924						 * established6925						 */6926	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a6927						 * passive connection is6928						 * established6929						 */6930	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control6931					 * needs ECN6932					 */6933	BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is6934					 * based on the path and may be6935					 * dependent on the congestion control6936					 * algorithm. In general it indicates6937					 * a congestion threshold. RTTs above6938					 * this indicate congestion6939					 */6940	BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.6941					 * Arg1: value of icsk_retransmits6942					 * Arg2: value of icsk_rto6943					 * Arg3: whether RTO has expired6944					 */6945	BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.6946					 * Arg1: sequence number of 1st byte6947					 * Arg2: # segments6948					 * Arg3: return value of6949					 *       tcp_transmit_skb (0 => success)6950					 */6951	BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.6952					 * Arg1: old_state6953					 * Arg2: new_state6954					 */6955	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after6956					 * socket transition to LISTEN state.6957					 */6958	BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.6959					 * Arg1: measured RTT input (mrtt)6960					 * Arg2: updated srtt6961					 */6962	BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.6963					 * It will be called to handle6964					 * the packets received at6965					 * an already established6966					 * connection.6967					 *6968					 * sock_ops->skb_data:6969					 * Referring to the received skb.6970					 * It covers the TCP header only.6971					 *6972					 * bpf_load_hdr_opt() can also6973					 * be used to search for a6974					 * particular option.6975					 */6976	BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the6977					 * header option later in6978					 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.6979					 * Arg1: bool want_cookie. (in6980					 *       writing SYNACK only)6981					 *6982					 * sock_ops->skb_data:6983					 * Not available because no header has6984					 * been	written yet.6985					 *6986					 * sock_ops->skb_tcp_flags:6987					 * The tcp_flags of the6988					 * outgoing skb. (e.g. SYN, ACK, FIN).6989					 *6990					 * bpf_reserve_hdr_opt() should6991					 * be used to reserve space.6992					 */6993	BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options6994					 * Arg1: bool want_cookie. (in6995					 *       writing SYNACK only)6996					 *6997					 * sock_ops->skb_data:6998					 * Referring to the outgoing skb.6999					 * It covers the TCP header7000					 * that has already been written7001					 * by the kernel and the7002					 * earlier bpf-progs.7003					 *7004					 * sock_ops->skb_tcp_flags:7005					 * The tcp_flags of the outgoing7006					 * skb. (e.g. SYN, ACK, FIN).7007					 *7008					 * bpf_store_hdr_opt() should7009					 * be used to write the7010					 * option.7011					 *7012					 * bpf_load_hdr_opt() can also7013					 * be used to search for a7014					 * particular option that7015					 * has already been written7016					 * by the kernel or the7017					 * earlier bpf-progs.7018					 */7019};7020 7021/* List of TCP states. There is a build check in net/ipv4/tcp.c to detect7022 * changes between the TCP and BPF versions. Ideally this should never happen.7023 * If it does, we need to add code to convert them before calling7024 * the BPF sock_ops function.7025 */7026enum {7027	BPF_TCP_ESTABLISHED = 1,7028	BPF_TCP_SYN_SENT,7029	BPF_TCP_SYN_RECV,7030	BPF_TCP_FIN_WAIT1,7031	BPF_TCP_FIN_WAIT2,7032	BPF_TCP_TIME_WAIT,7033	BPF_TCP_CLOSE,7034	BPF_TCP_CLOSE_WAIT,7035	BPF_TCP_LAST_ACK,7036	BPF_TCP_LISTEN,7037	BPF_TCP_CLOSING,	/* Now a valid state */7038	BPF_TCP_NEW_SYN_RECV,7039	BPF_TCP_BOUND_INACTIVE,7040 7041	BPF_TCP_MAX_STATES	/* Leave at the end! */7042};7043 7044enum {7045	TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */7046	TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */7047	TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */7048	TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */7049	/* Copy the SYN pkt to optval7050	 *7051	 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the7052	 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit7053	 * to only getting from the saved_syn.  It can either get the7054	 * syn packet from:7055	 *7056	 * 1. the just-received SYN packet (only available when writing the7057	 *    SYNACK).  It will be useful when it is not necessary to7058	 *    save the SYN packet for latter use.  It is also the only way7059	 *    to get the SYN during syncookie mode because the syn7060	 *    packet cannot be saved during syncookie.7061	 *7062	 * OR7063	 *7064	 * 2. the earlier saved syn which was done by7065	 *    bpf_setsockopt(TCP_SAVE_SYN).7066	 *7067	 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the7068	 * SYN packet is obtained.7069	 *7070	 * If the bpf-prog does not need the IP[46] header,  the7071	 * bpf-prog can avoid parsing the IP header by using7072	 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both7073	 * IP[46] and TCP header by using TCP_BPF_SYN_IP.7074	 *7075	 *      >0: Total number of bytes copied7076	 * -ENOSPC: Not enough space in optval. Only optlen number of7077	 *          bytes is copied.7078	 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt7079	 *	    is not saved by setsockopt(TCP_SAVE_SYN).7080	 */7081	TCP_BPF_SYN		= 1005, /* Copy the TCP header */7082	TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */7083	TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */7084	TCP_BPF_SOCK_OPS_CB_FLAGS = 1008, /* Get or Set TCP sock ops flags */7085};7086 7087enum {7088	BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),7089};7090 7091/* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and7092 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.7093 */7094enum {7095	BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the7096						 * total option spaces7097						 * required for an established7098						 * sk in order to calculate the7099						 * MSS.  No skb is actually7100						 * sent.7101						 */7102	BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode7103						 * when sending a SYN.7104						 */7105};7106 7107struct bpf_perf_event_value {7108	__u64 counter;7109	__u64 enabled;7110	__u64 running;7111};7112 7113enum {7114	BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),7115	BPF_DEVCG_ACC_READ	= (1ULL << 1),7116	BPF_DEVCG_ACC_WRITE	= (1ULL << 2),7117};7118 7119enum {7120	BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),7121	BPF_DEVCG_DEV_CHAR	= (1ULL << 1),7122};7123 7124struct bpf_cgroup_dev_ctx {7125	/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */7126	__u32 access_type;7127	__u32 major;7128	__u32 minor;7129};7130 7131struct bpf_raw_tracepoint_args {7132	__u64 args[0];7133};7134 7135/* DIRECT:  Skip the FIB rules and go to FIB table associated with device7136 * OUTPUT:  Do lookup from egress perspective; default is ingress7137 */7138enum {7139	BPF_FIB_LOOKUP_DIRECT  = (1U << 0),7140	BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),7141	BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),7142	BPF_FIB_LOOKUP_TBID    = (1U << 3),7143	BPF_FIB_LOOKUP_SRC     = (1U << 4),7144	BPF_FIB_LOOKUP_MARK    = (1U << 5),7145};7146 7147enum {7148	BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */7149	BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */7150	BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */7151	BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */7152	BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */7153	BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */7154	BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */7155	BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */7156	BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */7157	BPF_FIB_LKUP_RET_NO_SRC_ADDR,  /* failed to derive IP src addr */7158};7159 7160struct bpf_fib_lookup {7161	/* input:  network family for lookup (AF_INET, AF_INET6)7162	 * output: network family of egress nexthop7163	 */7164	__u8	family;7165 7166	/* set if lookup is to consider L4 data - e.g., FIB rules */7167	__u8	l4_protocol;7168	__be16	sport;7169	__be16	dport;7170 7171	union {	/* used for MTU check */7172		/* input to lookup */7173		__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */7174 7175		/* output: MTU value */7176		__u16	mtu_result;7177	} __attribute__((packed, aligned(2)));7178	/* input: L3 device index for lookup7179	 * output: device index from FIB lookup7180	 */7181	__u32	ifindex;7182 7183	union {7184		/* inputs to lookup */7185		__u8	tos;		/* AF_INET  */7186		__be32	flowinfo;	/* AF_INET6, flow_label + priority */7187 7188		/* output: metric of fib result (IPv4/IPv6 only) */7189		__u32	rt_metric;7190	};7191 7192	/* input: source address to consider for lookup7193	 * output: source address result from lookup7194	 */7195	union {7196		__be32		ipv4_src;7197		__u32		ipv6_src[4];  /* in6_addr; network order */7198	};7199 7200	/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in7201	 * network header. output: bpf_fib_lookup sets to gateway address7202	 * if FIB lookup returns gateway route7203	 */7204	union {7205		__be32		ipv4_dst;7206		__u32		ipv6_dst[4];  /* in6_addr; network order */7207	};7208 7209	union {7210		struct {7211			/* output */7212			__be16	h_vlan_proto;7213			__be16	h_vlan_TCI;7214		};7215		/* input: when accompanied with the7216		 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a7217		 * specific routing table to use for the fib lookup.7218		 */7219		__u32	tbid;7220	};7221 7222	union {7223		/* input */7224		struct {7225			__u32	mark;   /* policy routing */7226			/* 2 4-byte holes for input */7227		};7228 7229		/* output: source and dest mac */7230		struct {7231			__u8	smac[6];	/* ETH_ALEN */7232			__u8	dmac[6];	/* ETH_ALEN */7233		};7234	};7235};7236 7237struct bpf_redir_neigh {7238	/* network family for lookup (AF_INET, AF_INET6) */7239	__u32 nh_family;7240	/* network address of nexthop; skips fib lookup to find gateway */7241	union {7242		__be32		ipv4_nh;7243		__u32		ipv6_nh[4];  /* in6_addr; network order */7244	};7245};7246 7247/* bpf_check_mtu flags*/7248enum  bpf_check_mtu_flags {7249	BPF_MTU_CHK_SEGS  = (1U << 0),7250};7251 7252enum bpf_check_mtu_ret {7253	BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */7254	BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */7255	BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */7256};7257 7258enum bpf_task_fd_type {7259	BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */7260	BPF_FD_TYPE_TRACEPOINT,		/* tp name */7261	BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */7262	BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */7263	BPF_FD_TYPE_UPROBE,		/* filename + offset */7264	BPF_FD_TYPE_URETPROBE,		/* filename + offset */7265};7266 7267enum {7268	BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),7269	BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),7270	BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),7271};7272 7273struct bpf_flow_keys {7274	__u16	nhoff;7275	__u16	thoff;7276	__u16	addr_proto;			/* ETH_P_* of valid addrs */7277	__u8	is_frag;7278	__u8	is_first_frag;7279	__u8	is_encap;7280	__u8	ip_proto;7281	__be16	n_proto;7282	__be16	sport;7283	__be16	dport;7284	union {7285		struct {7286			__be32	ipv4_src;7287			__be32	ipv4_dst;7288		};7289		struct {7290			__u32	ipv6_src[4];	/* in6_addr; network order */7291			__u32	ipv6_dst[4];	/* in6_addr; network order */7292		};7293	};7294	__u32	flags;7295	__be32	flow_label;7296};7297 7298struct bpf_func_info {7299	__u32	insn_off;7300	__u32	type_id;7301};7302 7303#define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)7304#define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)7305 7306struct bpf_line_info {7307	__u32	insn_off;7308	__u32	file_name_off;7309	__u32	line_off;7310	__u32	line_col;7311};7312 7313struct bpf_spin_lock {7314	__u32	val;7315};7316 7317struct bpf_timer {7318	__u64 __opaque[2];7319} __attribute__((aligned(8)));7320 7321struct bpf_wq {7322	__u64 __opaque[2];7323} __attribute__((aligned(8)));7324 7325struct bpf_dynptr {7326	__u64 __opaque[2];7327} __attribute__((aligned(8)));7328 7329struct bpf_list_head {7330	__u64 __opaque[2];7331} __attribute__((aligned(8)));7332 7333struct bpf_list_node {7334	__u64 __opaque[3];7335} __attribute__((aligned(8)));7336 7337struct bpf_rb_root {7338	__u64 __opaque[2];7339} __attribute__((aligned(8)));7340 7341struct bpf_rb_node {7342	__u64 __opaque[4];7343} __attribute__((aligned(8)));7344 7345struct bpf_refcount {7346	__u32 __opaque[1];7347} __attribute__((aligned(4)));7348 7349struct bpf_sysctl {7350	__u32	write;		/* Sysctl is being read (= 0) or written (= 1).7351				 * Allows 1,2,4-byte read, but no write.7352				 */7353	__u32	file_pos;	/* Sysctl file position to read from, write to.7354				 * Allows 1,2,4-byte read an 4-byte write.7355				 */7356};7357 7358struct bpf_sockopt {7359	__bpf_md_ptr(struct bpf_sock *, sk);7360	__bpf_md_ptr(void *, optval);7361	__bpf_md_ptr(void *, optval_end);7362 7363	__s32	level;7364	__s32	optname;7365	__s32	optlen;7366	__s32	retval;7367};7368 7369struct bpf_pidns_info {7370	__u32 pid;7371	__u32 tgid;7372};7373 7374/* User accessible data for SK_LOOKUP programs. Add new fields at the end. */7375struct bpf_sk_lookup {7376	union {7377		__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */7378		__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */7379	};7380 7381	__u32 family;		/* Protocol family (AF_INET, AF_INET6) */7382	__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */7383	__u32 remote_ip4;	/* Network byte order */7384	__u32 remote_ip6[4];	/* Network byte order */7385	__be16 remote_port;	/* Network byte order */7386	__u16 :16;		/* Zero padding */7387	__u32 local_ip4;	/* Network byte order */7388	__u32 local_ip6[4];	/* Network byte order */7389	__u32 local_port;	/* Host byte order */7390	__u32 ingress_ifindex;		/* The arriving interface. Determined by inet_iif. */7391};7392 7393/*7394 * struct btf_ptr is used for typed pointer representation; the7395 * type id is used to render the pointer data as the appropriate type7396 * via the bpf_snprintf_btf() helper described above.  A flags field -7397 * potentially to specify additional details about the BTF pointer7398 * (rather than its mode of display) - is included for future use.7399 * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.7400 */7401struct btf_ptr {7402	void *ptr;7403	__u32 type_id;7404	__u32 flags;		/* BTF ptr flags; unused at present. */7405};7406 7407/*7408 * Flags to control bpf_snprintf_btf() behaviour.7409 *     - BTF_F_COMPACT: no formatting around type information7410 *     - BTF_F_NONAME: no struct/union member names/types7411 *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;7412 *       equivalent to %px.7413 *     - BTF_F_ZERO: show zero-valued struct/union members; they7414 *       are not displayed by default7415 */7416enum {7417	BTF_F_COMPACT	=	(1ULL << 0),7418	BTF_F_NONAME	=	(1ULL << 1),7419	BTF_F_PTR_RAW	=	(1ULL << 2),7420	BTF_F_ZERO	=	(1ULL << 3),7421};7422 7423/* bpf_core_relo_kind encodes which aspect of captured field/type/enum value7424 * has to be adjusted by relocations. It is emitted by llvm and passed to7425 * libbpf and later to the kernel.7426 */7427enum bpf_core_relo_kind {7428	BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */7429	BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */7430	BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */7431	BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */7432	BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */7433	BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */7434	BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */7435	BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */7436	BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */7437	BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */7438	BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */7439	BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */7440	BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */7441};7442 7443/*7444 * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf7445 * and from libbpf to the kernel.7446 *7447 * CO-RE relocation captures the following data:7448 * - insn_off - instruction offset (in bytes) within a BPF program that needs7449 *   its insn->imm field to be relocated with actual field info;7450 * - type_id - BTF type ID of the "root" (containing) entity of a relocatable7451 *   type or field;7452 * - access_str_off - offset into corresponding .BTF string section. String7453 *   interpretation depends on specific relocation kind:7454 *     - for field-based relocations, string encodes an accessed field using7455 *       a sequence of field and array indices, separated by colon (:). It's7456 *       conceptually very close to LLVM's getelementptr ([0]) instruction's7457 *       arguments for identifying offset to a field.7458 *     - for type-based relocations, strings is expected to be just "0";7459 *     - for enum value-based relocations, string contains an index of enum7460 *       value within its enum type;7461 * - kind - one of enum bpf_core_relo_kind;7462 *7463 * Example:7464 *   struct sample {7465 *       int a;7466 *       struct {7467 *           int b[10];7468 *       };7469 *   };7470 *7471 *   struct sample *s = ...;7472 *   int *x = &s->a;     // encoded as "0:0" (a is field #0)7473 *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,7474 *                       // b is field #0 inside anon struct, accessing elem #5)7475 *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)7476 *7477 * type_id for all relocs in this example will capture BTF type id of7478 * `struct sample`.7479 *7480 * Such relocation is emitted when using __builtin_preserve_access_index()7481 * Clang built-in, passing expression that captures field address, e.g.:7482 *7483 * bpf_probe_read(&dst, sizeof(dst),7484 *		  __builtin_preserve_access_index(&src->a.b.c));7485 *7486 * In this case Clang will emit field relocation recording necessary data to7487 * be able to find offset of embedded `a.b.c` field within `src` struct.7488 *7489 * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction7490 */7491struct bpf_core_relo {7492	__u32 insn_off;7493	__u32 type_id;7494	__u32 access_str_off;7495	enum bpf_core_relo_kind kind;7496};7497 7498/*7499 * Flags to control bpf_timer_start() behaviour.7500 *     - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is7501 *       relative to current time.7502 *     - BPF_F_TIMER_CPU_PIN: Timer will be pinned to the CPU of the caller.7503 */7504enum {7505	BPF_F_TIMER_ABS = (1ULL << 0),7506	BPF_F_TIMER_CPU_PIN = (1ULL << 1),7507};7508 7509/* BPF numbers iterator state */7510struct bpf_iter_num {7511	/* opaque iterator state; having __u64 here allows to preserve correct7512	 * alignment requirements in vmlinux.h, generated from BTF7513	 */7514	__u64 __opaque[1];7515} __attribute__((aligned(8)));7516 7517/*7518 * Flags to control BPF kfunc behaviour.7519 *     - BPF_F_PAD_ZEROS: Pad destination buffer with zeros. (See the respective7520 *       helper documentation for details.)7521 */7522enum bpf_kfunc_flags {7523	BPF_F_PAD_ZEROS = (1ULL << 0),7524};7525 7526#endif /* _UAPI__LINUX_BPF_H__ */7527