• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of version 2 of the GNU General Public
6  * License as published by the Free Software Foundation.
7  */
8 #ifndef _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_LE
35 #define BPF_FROM_BE	BPF_TO_BE
36 
37 /* jmp encodings */
38 #define BPF_JNE		0x50	/* jump != */
39 #define BPF_JLT		0xa0	/* LT is unsigned, '<' */
40 #define BPF_JLE		0xb0	/* LE is unsigned, '<=' */
41 #define BPF_JSGT	0x60	/* SGT is signed '>', GT in x86 */
42 #define BPF_JSGE	0x70	/* SGE is signed '>=', GE in x86 */
43 #define BPF_JSLT	0xc0	/* SLT is signed, '<' */
44 #define BPF_JSLE	0xd0	/* SLE is signed, '<=' */
45 #define BPF_CALL	0x80	/* function call */
46 #define BPF_EXIT	0x90	/* function return */
47 
48 /* atomic op type fields (stored in immediate) */
49 #define BPF_FETCH	0x01	/* not an opcode on its own, used to build others */
50 #define BPF_XCHG	(0xe0 | BPF_FETCH)	/* atomic exchange */
51 #define BPF_CMPXCHG	(0xf0 | BPF_FETCH)	/* atomic compare-and-write */
52 
53 /* Register numbers */
54 enum {
55 	BPF_REG_0 = 0,
56 	BPF_REG_1,
57 	BPF_REG_2,
58 	BPF_REG_3,
59 	BPF_REG_4,
60 	BPF_REG_5,
61 	BPF_REG_6,
62 	BPF_REG_7,
63 	BPF_REG_8,
64 	BPF_REG_9,
65 	BPF_REG_10,
66 	__MAX_BPF_REG,
67 };
68 
69 /* BPF has 10 general purpose 64-bit registers and stack frame. */
70 #define MAX_BPF_REG	__MAX_BPF_REG
71 
72 struct bpf_insn {
73 	__u8	code;		/* opcode */
74 	__u8	dst_reg:4;	/* dest register */
75 	__u8	src_reg:4;	/* source register */
76 	__s16	off;		/* signed offset */
77 	__s32	imm;		/* signed immediate constant */
78 };
79 
80 /* Deprecated: use struct bpf_lpm_trie_key_u8 (when the "data" member is needed for
81  * byte access) or struct bpf_lpm_trie_key_hdr (when using an alternative type for
82  * the trailing flexible array member) instead.
83  */
84 struct bpf_lpm_trie_key {
85 	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
86 	__u8	data[0];	/* Arbitrary size */
87 };
88 
89 /* Header for bpf_lpm_trie_key structs */
90 struct bpf_lpm_trie_key_hdr {
91 	__u32	prefixlen;
92 };
93 
94 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry, with trailing byte array. */
95 struct bpf_lpm_trie_key_u8 {
96 	union {
97 		struct bpf_lpm_trie_key_hdr	hdr;
98 		__u32				prefixlen;
99 	};
100 	__u8	data[];		/* Arbitrary size */
101 };
102 
103 struct bpf_cgroup_storage_key {
104 	__u64	cgroup_inode_id;	/* cgroup inode id */
105 	__u32	attach_type;		/* program attach type (enum bpf_attach_type) */
106 };
107 
108 enum bpf_cgroup_iter_order {
109 	BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
110 	BPF_CGROUP_ITER_SELF_ONLY,		/* process only a single object. */
111 	BPF_CGROUP_ITER_DESCENDANTS_PRE,	/* walk descendants in pre-order. */
112 	BPF_CGROUP_ITER_DESCENDANTS_POST,	/* walk descendants in post-order. */
113 	BPF_CGROUP_ITER_ANCESTORS_UP,		/* walk ancestors upward. */
114 };
115 
116 union bpf_iter_link_info {
117 	struct {
118 		__u32	map_fd;
119 	} map;
120 	struct {
121 		enum bpf_cgroup_iter_order order;
122 
123 		/* At most one of cgroup_fd and cgroup_id can be non-zero. If
124 		 * both are zero, the walk starts from the default cgroup v2
125 		 * root. For walking v1 hierarchy, one should always explicitly
126 		 * specify cgroup_fd.
127 		 */
128 		__u32	cgroup_fd;
129 		__u64	cgroup_id;
130 	} cgroup;
131 	/* Parameters of task iterators. */
132 	struct {
133 		__u32	tid;
134 		__u32	pid;
135 		__u32	pid_fd;
136 	} task;
137 };
138 
139 /* BPF syscall commands, see bpf(2) man-page for more details. */
140 /**
141  * DOC: eBPF Syscall Preamble
142  *
143  * The operation to be performed by the **bpf**\ () system call is determined
144  * by the *cmd* argument. Each operation takes an accompanying argument,
145  * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
146  * below). The size argument is the size of the union pointed to by *attr*.
147  */
148 /**
149  * DOC: eBPF Syscall Commands
150  *
151  * BPF_MAP_CREATE
152  *	Description
153  *		Create a map and return a file descriptor that refers to the
154  *		map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
155  *		is automatically enabled for the new file descriptor.
156  *
157  *		Applying **close**\ (2) to the file descriptor returned by
158  *		**BPF_MAP_CREATE** will delete the map (but see NOTES).
159  *
160  *	Return
161  *		A new file descriptor (a nonnegative integer), or -1 if an
162  *		error occurred (in which case, *errno* is set appropriately).
163  *
164  * BPF_MAP_LOOKUP_ELEM
165  *	Description
166  *		Look up an element with a given *key* in the map referred to
167  *		by the file descriptor *map_fd*.
168  *
169  *		The *flags* argument may be specified as one of the
170  *		following:
171  *
172  *		**BPF_F_LOCK**
173  *			Look up the value of a spin-locked map without
174  *			returning the lock. This must be specified if the
175  *			elements contain a spinlock.
176  *
177  *	Return
178  *		Returns zero on success. On error, -1 is returned and *errno*
179  *		is set appropriately.
180  *
181  * BPF_MAP_UPDATE_ELEM
182  *	Description
183  *		Create or update an element (key/value pair) in a specified map.
184  *
185  *		The *flags* argument should be specified as one of the
186  *		following:
187  *
188  *		**BPF_ANY**
189  *			Create a new element or update an existing element.
190  *		**BPF_NOEXIST**
191  *			Create a new element only if it did not exist.
192  *		**BPF_EXIST**
193  *			Update an existing element.
194  *		**BPF_F_LOCK**
195  *			Update a spin_lock-ed map element.
196  *
197  *	Return
198  *		Returns zero on success. On error, -1 is returned and *errno*
199  *		is set appropriately.
200  *
201  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
202  *		**E2BIG**, **EEXIST**, or **ENOENT**.
203  *
204  *		**E2BIG**
205  *			The number of elements in the map reached the
206  *			*max_entries* limit specified at map creation time.
207  *		**EEXIST**
208  *			If *flags* specifies **BPF_NOEXIST** and the element
209  *			with *key* already exists in the map.
210  *		**ENOENT**
211  *			If *flags* specifies **BPF_EXIST** and the element with
212  *			*key* does not exist in the map.
213  *
214  * BPF_MAP_DELETE_ELEM
215  *	Description
216  *		Look up and delete an element by key in a specified map.
217  *
218  *	Return
219  *		Returns zero on success. On error, -1 is returned and *errno*
220  *		is set appropriately.
221  *
222  * BPF_MAP_GET_NEXT_KEY
223  *	Description
224  *		Look up an element by key in a specified map and return the key
225  *		of the next element. Can be used to iterate over all elements
226  *		in the map.
227  *
228  *	Return
229  *		Returns zero on success. On error, -1 is returned and *errno*
230  *		is set appropriately.
231  *
232  *		The following cases can be used to iterate over all elements of
233  *		the map:
234  *
235  *		* If *key* is not found, the operation returns zero and sets
236  *		  the *next_key* pointer to the key of the first element.
237  *		* If *key* is found, the operation returns zero and sets the
238  *		  *next_key* pointer to the key of the next element.
239  *		* If *key* is the last element, returns -1 and *errno* is set
240  *		  to **ENOENT**.
241  *
242  *		May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
243  *		**EINVAL** on error.
244  *
245  * BPF_PROG_LOAD
246  *	Description
247  *		Verify and load an eBPF program, returning a new file
248  *		descriptor associated with the program.
249  *
250  *		Applying **close**\ (2) to the file descriptor returned by
251  *		**BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
252  *
253  *		The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
254  *		automatically enabled for the new file descriptor.
255  *
256  *	Return
257  *		A new file descriptor (a nonnegative integer), or -1 if an
258  *		error occurred (in which case, *errno* is set appropriately).
259  *
260  * BPF_OBJ_PIN
261  *	Description
262  *		Pin an eBPF program or map referred by the specified *bpf_fd*
263  *		to the provided *pathname* on the filesystem.
264  *
265  *		The *pathname* argument must not contain a dot (".").
266  *
267  *		On success, *pathname* retains a reference to the eBPF object,
268  *		preventing deallocation of the object when the original
269  *		*bpf_fd* is closed. This allow the eBPF object to live beyond
270  *		**close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
271  *		process.
272  *
273  *		Applying **unlink**\ (2) or similar calls to the *pathname*
274  *		unpins the object from the filesystem, removing the reference.
275  *		If no other file descriptors or filesystem nodes refer to the
276  *		same object, it will be deallocated (see NOTES).
277  *
278  *		The filesystem type for the parent directory of *pathname* must
279  *		be **BPF_FS_MAGIC**.
280  *
281  *	Return
282  *		Returns zero on success. On error, -1 is returned and *errno*
283  *		is set appropriately.
284  *
285  * BPF_OBJ_GET
286  *	Description
287  *		Open a file descriptor for the eBPF object pinned to the
288  *		specified *pathname*.
289  *
290  *	Return
291  *		A new file descriptor (a nonnegative integer), or -1 if an
292  *		error occurred (in which case, *errno* is set appropriately).
293  *
294  * BPF_PROG_ATTACH
295  *	Description
296  *		Attach an eBPF program to a *target_fd* at the specified
297  *		*attach_type* hook.
298  *
299  *		The *attach_type* specifies the eBPF attachment point to
300  *		attach the program to, and must be one of *bpf_attach_type*
301  *		(see below).
302  *
303  *		The *attach_bpf_fd* must be a valid file descriptor for a
304  *		loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
305  *		or sock_ops type corresponding to the specified *attach_type*.
306  *
307  *		The *target_fd* must be a valid file descriptor for a kernel
308  *		object which depends on the attach type of *attach_bpf_fd*:
309  *
310  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
311  *		**BPF_PROG_TYPE_CGROUP_SKB**,
312  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
313  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
314  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
315  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
316  *		**BPF_PROG_TYPE_SOCK_OPS**
317  *
318  *			Control Group v2 hierarchy with the eBPF controller
319  *			enabled. Requires the kernel to be compiled with
320  *			**CONFIG_CGROUP_BPF**.
321  *
322  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
323  *
324  *			Network namespace (eg /proc/self/ns/net).
325  *
326  *		**BPF_PROG_TYPE_LIRC_MODE2**
327  *
328  *			LIRC device path (eg /dev/lircN). Requires the kernel
329  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
330  *
331  *		**BPF_PROG_TYPE_SK_SKB**,
332  *		**BPF_PROG_TYPE_SK_MSG**
333  *
334  *			eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
335  *
336  *	Return
337  *		Returns zero on success. On error, -1 is returned and *errno*
338  *		is set appropriately.
339  *
340  * BPF_PROG_DETACH
341  *	Description
342  *		Detach the eBPF program associated with the *target_fd* at the
343  *		hook specified by *attach_type*. The program must have been
344  *		previously attached using **BPF_PROG_ATTACH**.
345  *
346  *	Return
347  *		Returns zero on success. On error, -1 is returned and *errno*
348  *		is set appropriately.
349  *
350  * BPF_PROG_TEST_RUN
351  *	Description
352  *		Run the eBPF program associated with the *prog_fd* a *repeat*
353  *		number of times against a provided program context *ctx_in* and
354  *		data *data_in*, and return the modified program context
355  *		*ctx_out*, *data_out* (for example, packet data), result of the
356  *		execution *retval*, and *duration* of the test run.
357  *
358  *		The sizes of the buffers provided as input and output
359  *		parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
360  *		be provided in the corresponding variables *ctx_size_in*,
361  *		*ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
362  *		of these parameters are not provided (ie set to NULL), the
363  *		corresponding size field must be zero.
364  *
365  *		Some program types have particular requirements:
366  *
367  *		**BPF_PROG_TYPE_SK_LOOKUP**
368  *			*data_in* and *data_out* must be NULL.
369  *
370  *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,
371  *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
372  *
373  *			*ctx_out*, *data_in* and *data_out* must be NULL.
374  *			*repeat* must be zero.
375  *
376  *		BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
377  *
378  *	Return
379  *		Returns zero on success. On error, -1 is returned and *errno*
380  *		is set appropriately.
381  *
382  *		**ENOSPC**
383  *			Either *data_size_out* or *ctx_size_out* is too small.
384  *		**ENOTSUPP**
385  *			This command is not supported by the program type of
386  *			the program referred to by *prog_fd*.
387  *
388  * BPF_PROG_GET_NEXT_ID
389  *	Description
390  *		Fetch the next eBPF program currently loaded into the kernel.
391  *
392  *		Looks for the eBPF program with an id greater than *start_id*
393  *		and updates *next_id* on success. If no other eBPF programs
394  *		remain with ids higher than *start_id*, returns -1 and sets
395  *		*errno* to **ENOENT**.
396  *
397  *	Return
398  *		Returns zero on success. On error, or when no id remains, -1
399  *		is returned and *errno* is set appropriately.
400  *
401  * BPF_MAP_GET_NEXT_ID
402  *	Description
403  *		Fetch the next eBPF map currently loaded into the kernel.
404  *
405  *		Looks for the eBPF map with an id greater than *start_id*
406  *		and updates *next_id* on success. If no other eBPF maps
407  *		remain with ids higher than *start_id*, returns -1 and sets
408  *		*errno* to **ENOENT**.
409  *
410  *	Return
411  *		Returns zero on success. On error, or when no id remains, -1
412  *		is returned and *errno* is set appropriately.
413  *
414  * BPF_PROG_GET_FD_BY_ID
415  *	Description
416  *		Open a file descriptor for the eBPF program corresponding to
417  *		*prog_id*.
418  *
419  *	Return
420  *		A new file descriptor (a nonnegative integer), or -1 if an
421  *		error occurred (in which case, *errno* is set appropriately).
422  *
423  * BPF_MAP_GET_FD_BY_ID
424  *	Description
425  *		Open a file descriptor for the eBPF map corresponding to
426  *		*map_id*.
427  *
428  *	Return
429  *		A new file descriptor (a nonnegative integer), or -1 if an
430  *		error occurred (in which case, *errno* is set appropriately).
431  *
432  * BPF_OBJ_GET_INFO_BY_FD
433  *	Description
434  *		Obtain information about the eBPF object corresponding to
435  *		*bpf_fd*.
436  *
437  *		Populates up to *info_len* bytes of *info*, which will be in
438  *		one of the following formats depending on the eBPF object type
439  *		of *bpf_fd*:
440  *
441  *		* **struct bpf_prog_info**
442  *		* **struct bpf_map_info**
443  *		* **struct bpf_btf_info**
444  *		* **struct bpf_link_info**
445  *
446  *	Return
447  *		Returns zero on success. On error, -1 is returned and *errno*
448  *		is set appropriately.
449  *
450  * BPF_PROG_QUERY
451  *	Description
452  *		Obtain information about eBPF programs associated with the
453  *		specified *attach_type* hook.
454  *
455  *		The *target_fd* must be a valid file descriptor for a kernel
456  *		object which depends on the attach type of *attach_bpf_fd*:
457  *
458  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
459  *		**BPF_PROG_TYPE_CGROUP_SKB**,
460  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
461  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
462  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
463  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
464  *		**BPF_PROG_TYPE_SOCK_OPS**
465  *
466  *			Control Group v2 hierarchy with the eBPF controller
467  *			enabled. Requires the kernel to be compiled with
468  *			**CONFIG_CGROUP_BPF**.
469  *
470  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
471  *
472  *			Network namespace (eg /proc/self/ns/net).
473  *
474  *		**BPF_PROG_TYPE_LIRC_MODE2**
475  *
476  *			LIRC device path (eg /dev/lircN). Requires the kernel
477  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
478  *
479  *		**BPF_PROG_QUERY** always fetches the number of programs
480  *		attached and the *attach_flags* which were used to attach those
481  *		programs. Additionally, if *prog_ids* is nonzero and the number
482  *		of attached programs is less than *prog_cnt*, populates
483  *		*prog_ids* with the eBPF program ids of the programs attached
484  *		at *target_fd*.
485  *
486  *		The following flags may alter the result:
487  *
488  *		**BPF_F_QUERY_EFFECTIVE**
489  *			Only return information regarding programs which are
490  *			currently effective at the specified *target_fd*.
491  *
492  *	Return
493  *		Returns zero on success. On error, -1 is returned and *errno*
494  *		is set appropriately.
495  *
496  * BPF_RAW_TRACEPOINT_OPEN
497  *	Description
498  *		Attach an eBPF program to a tracepoint *name* to access kernel
499  *		internal arguments of the tracepoint in their raw form.
500  *
501  *		The *prog_fd* must be a valid file descriptor associated with
502  *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
503  *
504  *		No ABI guarantees are made about the content of tracepoint
505  *		arguments exposed to the corresponding eBPF program.
506  *
507  *		Applying **close**\ (2) to the file descriptor returned by
508  *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
509  *
510  *	Return
511  *		A new file descriptor (a nonnegative integer), or -1 if an
512  *		error occurred (in which case, *errno* is set appropriately).
513  *
514  * BPF_BTF_LOAD
515  *	Description
516  *		Verify and load BPF Type Format (BTF) metadata into the kernel,
517  *		returning a new file descriptor associated with the metadata.
518  *		BTF is described in more detail at
519  *		https://www.kernel.org/doc/html/latest/bpf/btf.html.
520  *
521  *		The *btf* parameter must point to valid memory providing
522  *		*btf_size* bytes of BTF binary metadata.
523  *
524  *		The returned file descriptor can be passed to other **bpf**\ ()
525  *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
526  *		associate the BTF with those objects.
527  *
528  *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
529  *		parameters to specify a *btf_log_buf*, *btf_log_size* and
530  *		*btf_log_level* which allow the kernel to return freeform log
531  *		output regarding the BTF verification process.
532  *
533  *	Return
534  *		A new file descriptor (a nonnegative integer), or -1 if an
535  *		error occurred (in which case, *errno* is set appropriately).
536  *
537  * BPF_BTF_GET_FD_BY_ID
538  *	Description
539  *		Open a file descriptor for the BPF Type Format (BTF)
540  *		corresponding to *btf_id*.
541  *
542  *	Return
543  *		A new file descriptor (a nonnegative integer), or -1 if an
544  *		error occurred (in which case, *errno* is set appropriately).
545  *
546  * BPF_TASK_FD_QUERY
547  *	Description
548  *		Obtain information about eBPF programs associated with the
549  *		target process identified by *pid* and *fd*.
550  *
551  *		If the *pid* and *fd* are associated with a tracepoint, kprobe
552  *		or uprobe perf event, then the *prog_id* and *fd_type* will
553  *		be populated with the eBPF program id and file descriptor type
554  *		of type **bpf_task_fd_type**. If associated with a kprobe or
555  *		uprobe, the  *probe_offset* and *probe_addr* will also be
556  *		populated. Optionally, if *buf* is provided, then up to
557  *		*buf_len* bytes of *buf* will be populated with the name of
558  *		the tracepoint, kprobe or uprobe.
559  *
560  *		The resulting *prog_id* may be introspected in deeper detail
561  *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
562  *
563  *	Return
564  *		Returns zero on success. On error, -1 is returned and *errno*
565  *		is set appropriately.
566  *
567  * BPF_MAP_LOOKUP_AND_DELETE_ELEM
568  *	Description
569  *		Look up an element with the given *key* in the map referred to
570  *		by the file descriptor *fd*, and if found, delete the element.
571  *
572  *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
573  *		types, the *flags* argument needs to be set to 0, but for other
574  *		map types, it may be specified as:
575  *
576  *		**BPF_F_LOCK**
577  *			Look up and delete the value of a spin-locked map
578  *			without returning the lock. This must be specified if
579  *			the elements contain a spinlock.
580  *
581  *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
582  *		implement this command as a "pop" operation, deleting the top
583  *		element rather than one corresponding to *key*.
584  *		The *key* and *key_len* parameters should be zeroed when
585  *		issuing this operation for these map types.
586  *
587  *		This command is only valid for the following map types:
588  *		* **BPF_MAP_TYPE_QUEUE**
589  *		* **BPF_MAP_TYPE_STACK**
590  *		* **BPF_MAP_TYPE_HASH**
591  *		* **BPF_MAP_TYPE_PERCPU_HASH**
592  *		* **BPF_MAP_TYPE_LRU_HASH**
593  *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**
594  *
595  *	Return
596  *		Returns zero on success. On error, -1 is returned and *errno*
597  *		is set appropriately.
598  *
599  * BPF_MAP_FREEZE
600  *	Description
601  *		Freeze the permissions of the specified map.
602  *
603  *		Write permissions may be frozen by passing zero *flags*.
604  *		Upon success, no future syscall invocations may alter the
605  *		map state of *map_fd*. Write operations from eBPF programs
606  *		are still possible for a frozen map.
607  *
608  *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
609  *
610  *	Return
611  *		Returns zero on success. On error, -1 is returned and *errno*
612  *		is set appropriately.
613  *
614  * BPF_BTF_GET_NEXT_ID
615  *	Description
616  *		Fetch the next BPF Type Format (BTF) object currently loaded
617  *		into the kernel.
618  *
619  *		Looks for the BTF object with an id greater than *start_id*
620  *		and updates *next_id* on success. If no other BTF objects
621  *		remain with ids higher than *start_id*, returns -1 and sets
622  *		*errno* to **ENOENT**.
623  *
624  *	Return
625  *		Returns zero on success. On error, or when no id remains, -1
626  *		is returned and *errno* is set appropriately.
627  *
628  * BPF_MAP_LOOKUP_BATCH
629  *	Description
630  *		Iterate and fetch multiple elements in a map.
631  *
632  *		Two opaque values are used to manage batch operations,
633  *		*in_batch* and *out_batch*. Initially, *in_batch* must be set
634  *		to NULL to begin the batched operation. After each subsequent
635  *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
636  *		*out_batch* as the *in_batch* for the next operation to
637  *		continue iteration from the current point.
638  *
639  *		The *keys* and *values* are output parameters which must point
640  *		to memory large enough to hold *count* items based on the key
641  *		and value size of the map *map_fd*. The *keys* buffer must be
642  *		of *key_size* * *count*. The *values* buffer must be of
643  *		*value_size* * *count*.
644  *
645  *		The *elem_flags* argument may be specified as one of the
646  *		following:
647  *
648  *		**BPF_F_LOCK**
649  *			Look up the value of a spin-locked map without
650  *			returning the lock. This must be specified if the
651  *			elements contain a spinlock.
652  *
653  *		On success, *count* elements from the map are copied into the
654  *		user buffer, with the keys copied into *keys* and the values
655  *		copied into the corresponding indices in *values*.
656  *
657  *		If an error is returned and *errno* is not **EFAULT**, *count*
658  *		is set to the number of successfully processed elements.
659  *
660  *	Return
661  *		Returns zero on success. On error, -1 is returned and *errno*
662  *		is set appropriately.
663  *
664  *		May set *errno* to **ENOSPC** to indicate that *keys* or
665  *		*values* is too small to dump an entire bucket during
666  *		iteration of a hash-based map type.
667  *
668  * BPF_MAP_LOOKUP_AND_DELETE_BATCH
669  *	Description
670  *		Iterate and delete all elements in a map.
671  *
672  *		This operation has the same behavior as
673  *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:
674  *
675  *		* Every element that is successfully returned is also deleted
676  *		  from the map. This is at least *count* elements. Note that
677  *		  *count* is both an input and an output parameter.
678  *		* Upon returning with *errno* set to **EFAULT**, up to
679  *		  *count* elements may be deleted without returning the keys
680  *		  and values of the deleted elements.
681  *
682  *	Return
683  *		Returns zero on success. On error, -1 is returned and *errno*
684  *		is set appropriately.
685  *
686  * BPF_MAP_UPDATE_BATCH
687  *	Description
688  *		Update multiple elements in a map by *key*.
689  *
690  *		The *keys* and *values* are input parameters which must point
691  *		to memory large enough to hold *count* items based on the key
692  *		and value size of the map *map_fd*. The *keys* buffer must be
693  *		of *key_size* * *count*. The *values* buffer must be of
694  *		*value_size* * *count*.
695  *
696  *		Each element specified in *keys* is sequentially updated to the
697  *		value in the corresponding index in *values*. The *in_batch*
698  *		and *out_batch* parameters are ignored and should be zeroed.
699  *
700  *		The *elem_flags* argument should be specified as one of the
701  *		following:
702  *
703  *		**BPF_ANY**
704  *			Create new elements or update a existing elements.
705  *		**BPF_NOEXIST**
706  *			Create new elements only if they do not exist.
707  *		**BPF_EXIST**
708  *			Update existing elements.
709  *		**BPF_F_LOCK**
710  *			Update spin_lock-ed map elements. This must be
711  *			specified if the map value contains a spinlock.
712  *
713  *		On success, *count* elements from the map are updated.
714  *
715  *		If an error is returned and *errno* is not **EFAULT**, *count*
716  *		is set to the number of successfully processed elements.
717  *
718  *	Return
719  *		Returns zero on success. On error, -1 is returned and *errno*
720  *		is set appropriately.
721  *
722  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
723  *		**E2BIG**. **E2BIG** indicates that the number of elements in
724  *		the map reached the *max_entries* limit specified at map
725  *		creation time.
726  *
727  *		May set *errno* to one of the following error codes under
728  *		specific circumstances:
729  *
730  *		**EEXIST**
731  *			If *flags* specifies **BPF_NOEXIST** and the element
732  *			with *key* already exists in the map.
733  *		**ENOENT**
734  *			If *flags* specifies **BPF_EXIST** and the element with
735  *			*key* does not exist in the map.
736  *
737  * BPF_MAP_DELETE_BATCH
738  *	Description
739  *		Delete multiple elements in a map by *key*.
740  *
741  *		The *keys* parameter is an input parameter which must point
742  *		to memory large enough to hold *count* items based on the key
743  *		size of the map *map_fd*, that is, *key_size* * *count*.
744  *
745  *		Each element specified in *keys* is sequentially deleted. The
746  *		*in_batch*, *out_batch*, and *values* parameters are ignored
747  *		and should be zeroed.
748  *
749  *		The *elem_flags* argument may be specified as one of the
750  *		following:
751  *
752  *		**BPF_F_LOCK**
753  *			Look up the value of a spin-locked map without
754  *			returning the lock. This must be specified if the
755  *			elements contain a spinlock.
756  *
757  *		On success, *count* elements from the map are updated.
758  *
759  *		If an error is returned and *errno* is not **EFAULT**, *count*
760  *		is set to the number of successfully processed elements. If
761  *		*errno* is **EFAULT**, up to *count* elements may be been
762  *		deleted.
763  *
764  *	Return
765  *		Returns zero on success. On error, -1 is returned and *errno*
766  *		is set appropriately.
767  *
768  * BPF_LINK_CREATE
769  *	Description
770  *		Attach an eBPF program to a *target_fd* at the specified
771  *		*attach_type* hook and return a file descriptor handle for
772  *		managing the link.
773  *
774  *	Return
775  *		A new file descriptor (a nonnegative integer), or -1 if an
776  *		error occurred (in which case, *errno* is set appropriately).
777  *
778  * BPF_LINK_UPDATE
779  *	Description
780  *		Update the eBPF program in the specified *link_fd* to
781  *		*new_prog_fd*.
782  *
783  *	Return
784  *		Returns zero on success. On error, -1 is returned and *errno*
785  *		is set appropriately.
786  *
787  * BPF_LINK_GET_FD_BY_ID
788  *	Description
789  *		Open a file descriptor for the eBPF Link corresponding to
790  *		*link_id*.
791  *
792  *	Return
793  *		A new file descriptor (a nonnegative integer), or -1 if an
794  *		error occurred (in which case, *errno* is set appropriately).
795  *
796  * BPF_LINK_GET_NEXT_ID
797  *	Description
798  *		Fetch the next eBPF link currently loaded into the kernel.
799  *
800  *		Looks for the eBPF link with an id greater than *start_id*
801  *		and updates *next_id* on success. If no other eBPF links
802  *		remain with ids higher than *start_id*, returns -1 and sets
803  *		*errno* to **ENOENT**.
804  *
805  *	Return
806  *		Returns zero on success. On error, or when no id remains, -1
807  *		is returned and *errno* is set appropriately.
808  *
809  * BPF_ENABLE_STATS
810  *	Description
811  *		Enable eBPF runtime statistics gathering.
812  *
813  *		Runtime statistics gathering for the eBPF runtime is disabled
814  *		by default to minimize the corresponding performance overhead.
815  *		This command enables statistics globally.
816  *
817  *		Multiple programs may independently enable statistics.
818  *		After gathering the desired statistics, eBPF runtime statistics
819  *		may be disabled again by calling **close**\ (2) for the file
820  *		descriptor returned by this function. Statistics will only be
821  *		disabled system-wide when all outstanding file descriptors
822  *		returned by prior calls for this subcommand are closed.
823  *
824  *	Return
825  *		A new file descriptor (a nonnegative integer), or -1 if an
826  *		error occurred (in which case, *errno* is set appropriately).
827  *
828  * BPF_ITER_CREATE
829  *	Description
830  *		Create an iterator on top of the specified *link_fd* (as
831  *		previously created using **BPF_LINK_CREATE**) and return a
832  *		file descriptor that can be used to trigger the iteration.
833  *
834  *		If the resulting file descriptor is pinned to the filesystem
835  *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
836  *		for that path will trigger the iterator to read kernel state
837  *		using the eBPF program attached to *link_fd*.
838  *
839  *	Return
840  *		A new file descriptor (a nonnegative integer), or -1 if an
841  *		error occurred (in which case, *errno* is set appropriately).
842  *
843  * BPF_LINK_DETACH
844  *	Description
845  *		Forcefully detach the specified *link_fd* from its
846  *		corresponding attachment point.
847  *
848  *	Return
849  *		Returns zero on success. On error, -1 is returned and *errno*
850  *		is set appropriately.
851  *
852  * BPF_PROG_BIND_MAP
853  *	Description
854  *		Bind a map to the lifetime of an eBPF program.
855  *
856  *		The map identified by *map_fd* is bound to the program
857  *		identified by *prog_fd* and only released when *prog_fd* is
858  *		released. This may be used in cases where metadata should be
859  *		associated with a program which otherwise does not contain any
860  *		references to the map (for example, embedded in the eBPF
861  *		program instructions).
862  *
863  *	Return
864  *		Returns zero on success. On error, -1 is returned and *errno*
865  *		is set appropriately.
866  *
867  * NOTES
868  *	eBPF objects (maps and programs) can be shared between processes.
869  *
870  *	* After **fork**\ (2), the child inherits file descriptors
871  *	  referring to the same eBPF objects.
872  *	* File descriptors referring to eBPF objects can be transferred over
873  *	  **unix**\ (7) domain sockets.
874  *	* File descriptors referring to eBPF objects can be duplicated in the
875  *	  usual way, using **dup**\ (2) and similar calls.
876  *	* File descriptors referring to eBPF objects can be pinned to the
877  *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
878  *
879  *	An eBPF object is deallocated only after all file descriptors referring
880  *	to the object have been closed and no references remain pinned to the
881  *	filesystem or attached (for example, bound to a program or device).
882  */
883 enum bpf_cmd {
884 	BPF_MAP_CREATE,
885 	BPF_MAP_LOOKUP_ELEM,
886 	BPF_MAP_UPDATE_ELEM,
887 	BPF_MAP_DELETE_ELEM,
888 	BPF_MAP_GET_NEXT_KEY,
889 	BPF_PROG_LOAD,
890 	BPF_OBJ_PIN,
891 	BPF_OBJ_GET,
892 	BPF_PROG_ATTACH,
893 	BPF_PROG_DETACH,
894 	BPF_PROG_TEST_RUN,
895 	BPF_PROG_RUN = BPF_PROG_TEST_RUN,
896 	BPF_PROG_GET_NEXT_ID,
897 	BPF_MAP_GET_NEXT_ID,
898 	BPF_PROG_GET_FD_BY_ID,
899 	BPF_MAP_GET_FD_BY_ID,
900 	BPF_OBJ_GET_INFO_BY_FD,
901 	BPF_PROG_QUERY,
902 	BPF_RAW_TRACEPOINT_OPEN,
903 	BPF_BTF_LOAD,
904 	BPF_BTF_GET_FD_BY_ID,
905 	BPF_TASK_FD_QUERY,
906 	BPF_MAP_LOOKUP_AND_DELETE_ELEM,
907 	BPF_MAP_FREEZE,
908 	BPF_BTF_GET_NEXT_ID,
909 	BPF_MAP_LOOKUP_BATCH,
910 	BPF_MAP_LOOKUP_AND_DELETE_BATCH,
911 	BPF_MAP_UPDATE_BATCH,
912 	BPF_MAP_DELETE_BATCH,
913 	BPF_LINK_CREATE,
914 	BPF_LINK_UPDATE,
915 	BPF_LINK_GET_FD_BY_ID,
916 	BPF_LINK_GET_NEXT_ID,
917 	BPF_ENABLE_STATS,
918 	BPF_ITER_CREATE,
919 	BPF_LINK_DETACH,
920 	BPF_PROG_BIND_MAP,
921 };
922 
923 enum bpf_map_type {
924 	BPF_MAP_TYPE_UNSPEC,
925 	BPF_MAP_TYPE_HASH,
926 	BPF_MAP_TYPE_ARRAY,
927 	BPF_MAP_TYPE_PROG_ARRAY,
928 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
929 	BPF_MAP_TYPE_PERCPU_HASH,
930 	BPF_MAP_TYPE_PERCPU_ARRAY,
931 	BPF_MAP_TYPE_STACK_TRACE,
932 	BPF_MAP_TYPE_CGROUP_ARRAY,
933 	BPF_MAP_TYPE_LRU_HASH,
934 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
935 	BPF_MAP_TYPE_LPM_TRIE,
936 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
937 	BPF_MAP_TYPE_HASH_OF_MAPS,
938 	BPF_MAP_TYPE_DEVMAP,
939 	BPF_MAP_TYPE_SOCKMAP,
940 	BPF_MAP_TYPE_CPUMAP,
941 	BPF_MAP_TYPE_XSKMAP,
942 	BPF_MAP_TYPE_SOCKHASH,
943 	BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
944 	/* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
945 	 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
946 	 * both cgroup-attached and other progs and supports all functionality
947 	 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
948 	 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
949 	 */
950 	BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
951 	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
952 	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
953 	BPF_MAP_TYPE_QUEUE,
954 	BPF_MAP_TYPE_STACK,
955 	BPF_MAP_TYPE_SK_STORAGE,
956 	BPF_MAP_TYPE_DEVMAP_HASH,
957 	BPF_MAP_TYPE_STRUCT_OPS,
958 	BPF_MAP_TYPE_RINGBUF,
959 	BPF_MAP_TYPE_INODE_STORAGE,
960 	BPF_MAP_TYPE_TASK_STORAGE,
961 	BPF_MAP_TYPE_BLOOM_FILTER,
962 	BPF_MAP_TYPE_USER_RINGBUF,
963 	BPF_MAP_TYPE_CGRP_STORAGE,
964 };
965 
966 /* Note that tracing related programs such as
967  * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
968  * are not subject to a stable API since kernel internal data
969  * structures can change from release to release and may
970  * therefore break existing tracing BPF programs. Tracing BPF
971  * programs correspond to /a/ specific kernel which is to be
972  * analyzed, and not /a/ specific kernel /and/ all future ones.
973  */
974 enum bpf_prog_type {
975 	BPF_PROG_TYPE_UNSPEC,
976 	BPF_PROG_TYPE_SOCKET_FILTER,
977 	BPF_PROG_TYPE_KPROBE,
978 	BPF_PROG_TYPE_SCHED_CLS,
979 	BPF_PROG_TYPE_SCHED_ACT,
980 	BPF_PROG_TYPE_TRACEPOINT,
981 	BPF_PROG_TYPE_XDP,
982 	BPF_PROG_TYPE_PERF_EVENT,
983 	BPF_PROG_TYPE_CGROUP_SKB,
984 	BPF_PROG_TYPE_CGROUP_SOCK,
985 	BPF_PROG_TYPE_LWT_IN,
986 	BPF_PROG_TYPE_LWT_OUT,
987 	BPF_PROG_TYPE_LWT_XMIT,
988 	BPF_PROG_TYPE_SOCK_OPS,
989 	BPF_PROG_TYPE_SK_SKB,
990 	BPF_PROG_TYPE_CGROUP_DEVICE,
991 	BPF_PROG_TYPE_SK_MSG,
992 	BPF_PROG_TYPE_RAW_TRACEPOINT,
993 	BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
994 	BPF_PROG_TYPE_LWT_SEG6LOCAL,
995 	BPF_PROG_TYPE_LIRC_MODE2,
996 	BPF_PROG_TYPE_SK_REUSEPORT,
997 	BPF_PROG_TYPE_FLOW_DISSECTOR,
998 	BPF_PROG_TYPE_CGROUP_SYSCTL,
999 	BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
1000 	BPF_PROG_TYPE_CGROUP_SOCKOPT,
1001 	BPF_PROG_TYPE_TRACING,
1002 	BPF_PROG_TYPE_STRUCT_OPS,
1003 	BPF_PROG_TYPE_EXT,
1004 	BPF_PROG_TYPE_LSM,
1005 	BPF_PROG_TYPE_SK_LOOKUP,
1006 	BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
1007 	BPF_PROG_TYPE_NETFILTER,
1008 };
1009 
1010 enum bpf_attach_type {
1011 	BPF_CGROUP_INET_INGRESS,
1012 	BPF_CGROUP_INET_EGRESS,
1013 	BPF_CGROUP_INET_SOCK_CREATE,
1014 	BPF_CGROUP_SOCK_OPS,
1015 	BPF_SK_SKB_STREAM_PARSER,
1016 	BPF_SK_SKB_STREAM_VERDICT,
1017 	BPF_CGROUP_DEVICE,
1018 	BPF_SK_MSG_VERDICT,
1019 	BPF_CGROUP_INET4_BIND,
1020 	BPF_CGROUP_INET6_BIND,
1021 	BPF_CGROUP_INET4_CONNECT,
1022 	BPF_CGROUP_INET6_CONNECT,
1023 	BPF_CGROUP_INET4_POST_BIND,
1024 	BPF_CGROUP_INET6_POST_BIND,
1025 	BPF_CGROUP_UDP4_SENDMSG,
1026 	BPF_CGROUP_UDP6_SENDMSG,
1027 	BPF_LIRC_MODE2,
1028 	BPF_FLOW_DISSECTOR,
1029 	BPF_CGROUP_SYSCTL,
1030 	BPF_CGROUP_UDP4_RECVMSG,
1031 	BPF_CGROUP_UDP6_RECVMSG,
1032 	BPF_CGROUP_GETSOCKOPT,
1033 	BPF_CGROUP_SETSOCKOPT,
1034 	BPF_TRACE_RAW_TP,
1035 	BPF_TRACE_FENTRY,
1036 	BPF_TRACE_FEXIT,
1037 	BPF_MODIFY_RETURN,
1038 	BPF_LSM_MAC,
1039 	BPF_TRACE_ITER,
1040 	BPF_CGROUP_INET4_GETPEERNAME,
1041 	BPF_CGROUP_INET6_GETPEERNAME,
1042 	BPF_CGROUP_INET4_GETSOCKNAME,
1043 	BPF_CGROUP_INET6_GETSOCKNAME,
1044 	BPF_XDP_DEVMAP,
1045 	BPF_CGROUP_INET_SOCK_RELEASE,
1046 	BPF_XDP_CPUMAP,
1047 	BPF_SK_LOOKUP,
1048 	BPF_XDP,
1049 	BPF_SK_SKB_VERDICT,
1050 	BPF_SK_REUSEPORT_SELECT,
1051 	BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1052 	BPF_PERF_EVENT,
1053 	BPF_TRACE_KPROBE_MULTI,
1054 	BPF_LSM_CGROUP,
1055 	BPF_STRUCT_OPS,
1056 	BPF_NETFILTER,
1057 	BPF_TCX_INGRESS,
1058 	BPF_TCX_EGRESS,
1059 	BPF_TRACE_UPROBE_MULTI,
1060 	__MAX_BPF_ATTACH_TYPE
1061 };
1062 
1063 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1064 
1065 enum bpf_link_type {
1066 	BPF_LINK_TYPE_UNSPEC = 0,
1067 	BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1068 	BPF_LINK_TYPE_TRACING = 2,
1069 	BPF_LINK_TYPE_CGROUP = 3,
1070 	BPF_LINK_TYPE_ITER = 4,
1071 	BPF_LINK_TYPE_NETNS = 5,
1072 	BPF_LINK_TYPE_XDP = 6,
1073 	BPF_LINK_TYPE_PERF_EVENT = 7,
1074 	BPF_LINK_TYPE_KPROBE_MULTI = 8,
1075 	BPF_LINK_TYPE_STRUCT_OPS = 9,
1076 	BPF_LINK_TYPE_NETFILTER = 10,
1077 	BPF_LINK_TYPE_TCX = 11,
1078 	BPF_LINK_TYPE_UPROBE_MULTI = 12,
1079 	MAX_BPF_LINK_TYPE,
1080 };
1081 
1082 enum bpf_perf_event_type {
1083 	BPF_PERF_EVENT_UNSPEC = 0,
1084 	BPF_PERF_EVENT_UPROBE = 1,
1085 	BPF_PERF_EVENT_URETPROBE = 2,
1086 	BPF_PERF_EVENT_KPROBE = 3,
1087 	BPF_PERF_EVENT_KRETPROBE = 4,
1088 	BPF_PERF_EVENT_TRACEPOINT = 5,
1089 	BPF_PERF_EVENT_EVENT = 6,
1090 };
1091 
1092 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1093  *
1094  * NONE(default): No further bpf programs allowed in the subtree.
1095  *
1096  * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1097  * the program in this cgroup yields to sub-cgroup program.
1098  *
1099  * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1100  * that cgroup program gets run in addition to the program in this cgroup.
1101  *
1102  * Only one program is allowed to be attached to a cgroup with
1103  * NONE or BPF_F_ALLOW_OVERRIDE flag.
1104  * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1105  * release old program and attach the new one. Attach flags has to match.
1106  *
1107  * Multiple programs are allowed to be attached to a cgroup with
1108  * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1109  * (those that were attached first, run first)
1110  * The programs of sub-cgroup are executed first, then programs of
1111  * this cgroup and then programs of parent cgroup.
1112  * When children program makes decision (like picking TCP CA or sock bind)
1113  * parent program has a chance to override it.
1114  *
1115  * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1116  * programs for a cgroup. Though it's possible to replace an old program at
1117  * any position by also specifying BPF_F_REPLACE flag and position itself in
1118  * replace_bpf_fd attribute. Old program at this position will be released.
1119  *
1120  * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1121  * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1122  * Ex1:
1123  * cgrp1 (MULTI progs A, B) ->
1124  *    cgrp2 (OVERRIDE prog C) ->
1125  *      cgrp3 (MULTI prog D) ->
1126  *        cgrp4 (OVERRIDE prog E) ->
1127  *          cgrp5 (NONE prog F)
1128  * the event in cgrp5 triggers execution of F,D,A,B in that order.
1129  * if prog F is detached, the execution is E,D,A,B
1130  * if prog F and D are detached, the execution is E,A,B
1131  * if prog F, E and D are detached, the execution is C,A,B
1132  *
1133  * All eligible programs are executed regardless of return code from
1134  * earlier programs.
1135  */
1136 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
1137 #define BPF_F_ALLOW_MULTI	(1U << 1)
1138 /* Generic attachment flags. */
1139 #define BPF_F_REPLACE		(1U << 2)
1140 #define BPF_F_BEFORE		(1U << 3)
1141 #define BPF_F_AFTER		(1U << 4)
1142 #define BPF_F_ID		(1U << 5)
1143 #define BPF_F_LINK		BPF_F_LINK /* 1 << 13 */
1144 
1145 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1146  * verifier will perform strict alignment checking as if the kernel
1147  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1148  * and NET_IP_ALIGN defined to 2.
1149  */
1150 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
1151 
1152 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROG_LOAD command, the
1153  * verifier will allow any alignment whatsoever.  On platforms
1154  * with strict alignment requirements for loads ands stores (such
1155  * as sparc and mips) the verifier validates that all loads and
1156  * stores provably follow this requirement.  This flag turns that
1157  * checking and enforcement off.
1158  *
1159  * It is mostly used for testing when we want to validate the
1160  * context and memory access aspects of the verifier, but because
1161  * of an unaligned access the alignment check would trigger before
1162  * the one we are interested in.
1163  */
1164 #define BPF_F_ANY_ALIGNMENT	(1U << 1)
1165 
1166 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1167  * Verifier does sub-register def/use analysis and identifies instructions whose
1168  * def only matters for low 32-bit, high 32-bit is never referenced later
1169  * through implicit zero extension. Therefore verifier notifies JIT back-ends
1170  * that it is safe to ignore clearing high 32-bit for these instructions. This
1171  * saves some back-ends a lot of code-gen. However such optimization is not
1172  * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1173  * hence hasn't used verifier's analysis result. But, we really want to have a
1174  * way to be able to verify the correctness of the described optimization on
1175  * x86_64 on which testsuites are frequently exercised.
1176  *
1177  * So, this flag is introduced. Once it is set, verifier will randomize high
1178  * 32-bit for those instructions who has been identified as safe to ignore them.
1179  * Then, if verifier is not doing correct analysis, such randomization will
1180  * regress tests to expose bugs.
1181  */
1182 #define BPF_F_TEST_RND_HI32	(1U << 2)
1183 
1184 /* The verifier internal test flag. Behavior is undefined */
1185 #define BPF_F_TEST_STATE_FREQ	(1U << 3)
1186 
1187 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1188  * restrict map and helper usage for such programs. Sleepable BPF programs can
1189  * only be attached to hooks where kernel execution context allows sleeping.
1190  * Such programs are allowed to use helpers that may sleep like
1191  * bpf_copy_from_user().
1192  */
1193 #define BPF_F_SLEEPABLE		(1U << 4)
1194 
1195 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1196  * fully support xdp frags.
1197  */
1198 #define BPF_F_XDP_HAS_FRAGS	(1U << 5)
1199 
1200 /* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
1201  * program becomes device-bound but can access XDP metadata.
1202  */
1203 #define BPF_F_XDP_DEV_BOUND_ONLY	(1U << 6)
1204 
1205 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1206  * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1207  */
1208 enum {
1209 	BPF_F_KPROBE_MULTI_RETURN = (1U << 0)
1210 };
1211 
1212 /* link_create.uprobe_multi.flags used in LINK_CREATE command for
1213  * BPF_TRACE_UPROBE_MULTI attach type to create return probe.
1214  */
1215 enum {
1216 	BPF_F_UPROBE_MULTI_RETURN = (1U << 0)
1217 };
1218 
1219 /* link_create.netfilter.flags used in LINK_CREATE command for
1220  * BPF_PROG_TYPE_NETFILTER to enable IP packet defragmentation.
1221  */
1222 #define BPF_F_NETFILTER_IP_DEFRAG (1U << 0)
1223 
1224 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1225  * the following extensions:
1226  *
1227  * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
1228  * insn[0].imm:      map fd or fd_idx
1229  * insn[1].imm:      0
1230  * insn[0].off:      0
1231  * insn[1].off:      0
1232  * ldimm64 rewrite:  address of map
1233  * verifier type:    CONST_PTR_TO_MAP
1234  */
1235 #define BPF_PSEUDO_MAP_FD	1
1236 #define BPF_PSEUDO_MAP_IDX	5
1237 
1238 /* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
1239  * insn[0].imm:      map fd or fd_idx
1240  * insn[1].imm:      offset into value
1241  * insn[0].off:      0
1242  * insn[1].off:      0
1243  * ldimm64 rewrite:  address of map[0]+offset
1244  * verifier type:    PTR_TO_MAP_VALUE
1245  */
1246 #define BPF_PSEUDO_MAP_VALUE		2
1247 #define BPF_PSEUDO_MAP_IDX_VALUE	6
1248 
1249 /* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
1250  * insn[0].imm:      kernel btd id of VAR
1251  * insn[1].imm:      0
1252  * insn[0].off:      0
1253  * insn[1].off:      0
1254  * ldimm64 rewrite:  address of the kernel variable
1255  * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1256  *                   is struct/union.
1257  */
1258 #define BPF_PSEUDO_BTF_ID	3
1259 /* insn[0].src_reg:  BPF_PSEUDO_FUNC
1260  * insn[0].imm:      insn offset to the func
1261  * insn[1].imm:      0
1262  * insn[0].off:      0
1263  * insn[1].off:      0
1264  * ldimm64 rewrite:  address of the function
1265  * verifier type:    PTR_TO_FUNC.
1266  */
1267 #define BPF_PSEUDO_FUNC		4
1268 
1269 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1270  * offset to another bpf function
1271  */
1272 #define BPF_PSEUDO_CALL		1
1273 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1274  * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1275  */
1276 #define BPF_PSEUDO_KFUNC_CALL	2
1277 
1278 /* flags for BPF_MAP_UPDATE_ELEM command */
1279 enum {
1280 	BPF_ANY		= 0, /* create new element or update existing */
1281 	BPF_NOEXIST	= 1, /* create new element if it didn't exist */
1282 	BPF_EXIST	= 2, /* update existing element */
1283 	BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */
1284 };
1285 
1286 /* flags for BPF_MAP_CREATE command */
1287 enum {
1288 	BPF_F_NO_PREALLOC	= (1U << 0),
1289 /* Instead of having one common LRU list in the
1290  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1291  * which can scale and perform better.
1292  * Note, the LRU nodes (including free nodes) cannot be moved
1293  * across different LRU lists.
1294  */
1295 	BPF_F_NO_COMMON_LRU	= (1U << 1),
1296 /* Specify numa node during map creation */
1297 	BPF_F_NUMA_NODE		= (1U << 2),
1298 
1299 /* Flags for accessing BPF object from syscall side. */
1300 	BPF_F_RDONLY		= (1U << 3),
1301 	BPF_F_WRONLY		= (1U << 4),
1302 
1303 /* Flag for stack_map, store build_id+offset instead of pointer */
1304 	BPF_F_STACK_BUILD_ID	= (1U << 5),
1305 
1306 /* Zero-initialize hash function seed. This should only be used for testing. */
1307 	BPF_F_ZERO_SEED		= (1U << 6),
1308 
1309 /* Flags for accessing BPF object from program side. */
1310 	BPF_F_RDONLY_PROG	= (1U << 7),
1311 	BPF_F_WRONLY_PROG	= (1U << 8),
1312 
1313 /* Clone map from listener for newly accepted socket */
1314 	BPF_F_CLONE		= (1U << 9),
1315 
1316 /* Enable memory-mapping BPF map */
1317 	BPF_F_MMAPABLE		= (1U << 10),
1318 
1319 /* Share perf_event among processes */
1320 	BPF_F_PRESERVE_ELEMS	= (1U << 11),
1321 
1322 /* Create a map that is suitable to be an inner map with dynamic max entries */
1323 	BPF_F_INNER_MAP		= (1U << 12),
1324 
1325 /* Create a map that will be registered/unregesitered by the backed bpf_link */
1326 	BPF_F_LINK		= (1U << 13),
1327 
1328 /* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1329 	BPF_F_PATH_FD		= (1U << 14),
1330 };
1331 
1332 /* Flags for BPF_PROG_QUERY. */
1333 
1334 /* Query effective (directly attached + inherited from ancestor cgroups)
1335  * programs that will be executed for events within a cgroup.
1336  * attach_flags with this flag are always returned 0.
1337  */
1338 #define BPF_F_QUERY_EFFECTIVE	(1U << 0)
1339 
1340 /* Flags for BPF_PROG_TEST_RUN */
1341 
1342 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1343 #define BPF_F_TEST_RUN_ON_CPU	(1U << 0)
1344 /* If set, XDP frames will be transmitted after processing */
1345 #define BPF_F_TEST_XDP_LIVE_FRAMES	(1U << 1)
1346 
1347 /* type for BPF_ENABLE_STATS */
1348 enum bpf_stats_type {
1349 	/* enabled run_time_ns and run_cnt */
1350 	BPF_STATS_RUN_TIME = 0,
1351 };
1352 
1353 enum bpf_stack_build_id_status {
1354 	/* user space need an empty entry to identify end of a trace */
1355 	BPF_STACK_BUILD_ID_EMPTY = 0,
1356 	/* with valid build_id and offset */
1357 	BPF_STACK_BUILD_ID_VALID = 1,
1358 	/* couldn't get build_id, fallback to ip */
1359 	BPF_STACK_BUILD_ID_IP = 2,
1360 };
1361 
1362 #define BPF_BUILD_ID_SIZE 20
1363 struct bpf_stack_build_id {
1364 	__s32		status;
1365 	unsigned char	build_id[BPF_BUILD_ID_SIZE];
1366 	union {
1367 		__u64	offset;
1368 		__u64	ip;
1369 	};
1370 };
1371 
1372 #define BPF_OBJ_NAME_LEN 16U
1373 
1374 union bpf_attr {
1375 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
1376 		__u32	map_type;	/* one of enum bpf_map_type */
1377 		__u32	key_size;	/* size of key in bytes */
1378 		__u32	value_size;	/* size of value in bytes */
1379 		__u32	max_entries;	/* max number of entries in a map */
1380 		__u32	map_flags;	/* BPF_MAP_CREATE related
1381 					 * flags defined above.
1382 					 */
1383 		__u32	inner_map_fd;	/* fd pointing to the inner map */
1384 		__u32	numa_node;	/* numa node (effective only if
1385 					 * BPF_F_NUMA_NODE is set).
1386 					 */
1387 		char	map_name[BPF_OBJ_NAME_LEN];
1388 		__u32	map_ifindex;	/* ifindex of netdev to create on */
1389 		__u32	btf_fd;		/* fd pointing to a BTF type data */
1390 		__u32	btf_key_type_id;	/* BTF type_id of the key */
1391 		__u32	btf_value_type_id;	/* BTF type_id of the value */
1392 		__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1393 						   * struct stored as the
1394 						   * map value
1395 						   */
1396 		/* Any per-map-type extra fields
1397 		 *
1398 		 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1399 		 * number of hash functions (if 0, the bloom filter will default
1400 		 * to using 5 hash functions).
1401 		 */
1402 		__u64	map_extra;
1403 	};
1404 
1405 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1406 		__u32		map_fd;
1407 		__aligned_u64	key;
1408 		union {
1409 			__aligned_u64 value;
1410 			__aligned_u64 next_key;
1411 		};
1412 		__u64		flags;
1413 	};
1414 
1415 	struct { /* struct used by BPF_MAP_*_BATCH commands */
1416 		__aligned_u64	in_batch;	/* start batch,
1417 						 * NULL to start from beginning
1418 						 */
1419 		__aligned_u64	out_batch;	/* output: next start batch */
1420 		__aligned_u64	keys;
1421 		__aligned_u64	values;
1422 		__u32		count;		/* input/output:
1423 						 * input: # of key/value
1424 						 * elements
1425 						 * output: # of filled elements
1426 						 */
1427 		__u32		map_fd;
1428 		__u64		elem_flags;
1429 		__u64		flags;
1430 	} batch;
1431 
1432 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
1433 		__u32		prog_type;	/* one of enum bpf_prog_type */
1434 		__u32		insn_cnt;
1435 		__aligned_u64	insns;
1436 		__aligned_u64	license;
1437 		__u32		log_level;	/* verbosity level of verifier */
1438 		__u32		log_size;	/* size of user buffer */
1439 		__aligned_u64	log_buf;	/* user supplied buffer */
1440 		__u32		kern_version;	/* not used */
1441 		__u32		prog_flags;
1442 		char		prog_name[BPF_OBJ_NAME_LEN];
1443 		__u32		prog_ifindex;	/* ifindex of netdev to prep for */
1444 		/* For some prog types expected attach type must be known at
1445 		 * load time to verify attach type specific parts of prog
1446 		 * (context accesses, allowed helpers, etc).
1447 		 */
1448 		__u32		expected_attach_type;
1449 		__u32		prog_btf_fd;	/* fd pointing to BTF type data */
1450 		__u32		func_info_rec_size;	/* userspace bpf_func_info size */
1451 		__aligned_u64	func_info;	/* func info */
1452 		__u32		func_info_cnt;	/* number of bpf_func_info records */
1453 		__u32		line_info_rec_size;	/* userspace bpf_line_info size */
1454 		__aligned_u64	line_info;	/* line info */
1455 		__u32		line_info_cnt;	/* number of bpf_line_info records */
1456 		__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */
1457 		union {
1458 			/* valid prog_fd to attach to bpf prog */
1459 			__u32		attach_prog_fd;
1460 			/* or valid module BTF object fd or 0 to attach to vmlinux */
1461 			__u32		attach_btf_obj_fd;
1462 		};
1463 		__u32		core_relo_cnt;	/* number of bpf_core_relo */
1464 		__aligned_u64	fd_array;	/* array of FDs */
1465 		__aligned_u64	core_relos;
1466 		__u32		core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1467 		/* output: actual total log contents size (including termintaing zero).
1468 		 * It could be both larger than original log_size (if log was
1469 		 * truncated), or smaller (if log buffer wasn't filled completely).
1470 		 */
1471 		__u32		log_true_size;
1472 	};
1473 
1474 	struct { /* anonymous struct used by BPF_OBJ_* commands */
1475 		__aligned_u64	pathname;
1476 		__u32		bpf_fd;
1477 		__u32		file_flags;
1478 		/* Same as dirfd in openat() syscall; see openat(2)
1479 		 * manpage for details of path FD and pathname semantics;
1480 		 * path_fd should accompanied by BPF_F_PATH_FD flag set in
1481 		 * file_flags field, otherwise it should be set to zero;
1482 		 * if BPF_F_PATH_FD flag is not set, AT_FDCWD is assumed.
1483 		 */
1484 		__s32		path_fd;
1485 	};
1486 
1487 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1488 		union {
1489 			__u32	target_fd;	/* target object to attach to or ... */
1490 			__u32	target_ifindex;	/* target ifindex */
1491 		};
1492 		__u32		attach_bpf_fd;
1493 		__u32		attach_type;
1494 		__u32		attach_flags;
1495 		__u32		replace_bpf_fd;
1496 		union {
1497 			__u32	relative_fd;
1498 			__u32	relative_id;
1499 		};
1500 		__u64		expected_revision;
1501 	};
1502 
1503 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1504 		__u32		prog_fd;
1505 		__u32		retval;
1506 		__u32		data_size_in;	/* input: len of data_in */
1507 		__u32		data_size_out;	/* input/output: len of data_out
1508 						 *   returns ENOSPC if data_out
1509 						 *   is too small.
1510 						 */
1511 		__aligned_u64	data_in;
1512 		__aligned_u64	data_out;
1513 		__u32		repeat;
1514 		__u32		duration;
1515 		__u32		ctx_size_in;	/* input: len of ctx_in */
1516 		__u32		ctx_size_out;	/* input/output: len of ctx_out
1517 						 *   returns ENOSPC if ctx_out
1518 						 *   is too small.
1519 						 */
1520 		__aligned_u64	ctx_in;
1521 		__aligned_u64	ctx_out;
1522 		__u32		flags;
1523 		__u32		cpu;
1524 		__u32		batch_size;
1525 	} test;
1526 
1527 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
1528 		union {
1529 			__u32		start_id;
1530 			__u32		prog_id;
1531 			__u32		map_id;
1532 			__u32		btf_id;
1533 			__u32		link_id;
1534 		};
1535 		__u32		next_id;
1536 		__u32		open_flags;
1537 	};
1538 
1539 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1540 		__u32		bpf_fd;
1541 		__u32		info_len;
1542 		__aligned_u64	info;
1543 	} info;
1544 
1545 	struct { /* anonymous struct used by BPF_PROG_QUERY command */
1546 		union {
1547 			__u32	target_fd;	/* target object to query or ... */
1548 			__u32	target_ifindex;	/* target ifindex */
1549 		};
1550 		__u32		attach_type;
1551 		__u32		query_flags;
1552 		__u32		attach_flags;
1553 		__aligned_u64	prog_ids;
1554 		union {
1555 			__u32	prog_cnt;
1556 			__u32	count;
1557 		};
1558 		__u32		:32;
1559 		/* output: per-program attach_flags.
1560 		 * not allowed to be set during effective query.
1561 		 */
1562 		__aligned_u64	prog_attach_flags;
1563 		__aligned_u64	link_ids;
1564 		__aligned_u64	link_attach_flags;
1565 		__u64		revision;
1566 	} query;
1567 
1568 	struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1569 		__u64 name;
1570 		__u32 prog_fd;
1571 	} raw_tracepoint;
1572 
1573 	struct { /* anonymous struct for BPF_BTF_LOAD */
1574 		__aligned_u64	btf;
1575 		__aligned_u64	btf_log_buf;
1576 		__u32		btf_size;
1577 		__u32		btf_log_size;
1578 		__u32		btf_log_level;
1579 		/* output: actual total log contents size (including termintaing zero).
1580 		 * It could be both larger than original log_size (if log was
1581 		 * truncated), or smaller (if log buffer wasn't filled completely).
1582 		 */
1583 		__u32		btf_log_true_size;
1584 	};
1585 
1586 	struct {
1587 		__u32		pid;		/* input: pid */
1588 		__u32		fd;		/* input: fd */
1589 		__u32		flags;		/* input: flags */
1590 		__u32		buf_len;	/* input/output: buf len */
1591 		__aligned_u64	buf;		/* input/output:
1592 						 *   tp_name for tracepoint
1593 						 *   symbol for kprobe
1594 						 *   filename for uprobe
1595 						 */
1596 		__u32		prog_id;	/* output: prod_id */
1597 		__u32		fd_type;	/* output: BPF_FD_TYPE_* */
1598 		__u64		probe_offset;	/* output: probe_offset */
1599 		__u64		probe_addr;	/* output: probe_addr */
1600 	} task_fd_query;
1601 
1602 	struct { /* struct used by BPF_LINK_CREATE command */
1603 		union {
1604 			__u32		prog_fd;	/* eBPF program to attach */
1605 			__u32		map_fd;		/* struct_ops to attach */
1606 		};
1607 		union {
1608 			__u32	target_fd;	/* target object to attach to or ... */
1609 			__u32	target_ifindex; /* target ifindex */
1610 		};
1611 		__u32		attach_type;	/* attach type */
1612 		__u32		flags;		/* extra flags */
1613 		union {
1614 			__u32	target_btf_id;	/* btf_id of target to attach to */
1615 			struct {
1616 				__aligned_u64	iter_info;	/* extra bpf_iter_link_info */
1617 				__u32		iter_info_len;	/* iter_info length */
1618 			};
1619 			struct {
1620 				/* black box user-provided value passed through
1621 				 * to BPF program at the execution time and
1622 				 * accessible through bpf_get_attach_cookie() BPF helper
1623 				 */
1624 				__u64		bpf_cookie;
1625 			} perf_event;
1626 			struct {
1627 				__u32		flags;
1628 				__u32		cnt;
1629 				__aligned_u64	syms;
1630 				__aligned_u64	addrs;
1631 				__aligned_u64	cookies;
1632 			} kprobe_multi;
1633 			struct {
1634 				/* this is overlaid with the target_btf_id above. */
1635 				__u32		target_btf_id;
1636 				/* black box user-provided value passed through
1637 				 * to BPF program at the execution time and
1638 				 * accessible through bpf_get_attach_cookie() BPF helper
1639 				 */
1640 				__u64		cookie;
1641 			} tracing;
1642 			struct {
1643 				__u32		pf;
1644 				__u32		hooknum;
1645 				__s32		priority;
1646 				__u32		flags;
1647 			} netfilter;
1648 			struct {
1649 				union {
1650 					__u32	relative_fd;
1651 					__u32	relative_id;
1652 				};
1653 				__u64		expected_revision;
1654 			} tcx;
1655 			struct {
1656 				__aligned_u64	path;
1657 				__aligned_u64	offsets;
1658 				__aligned_u64	ref_ctr_offsets;
1659 				__aligned_u64	cookies;
1660 				__u32		cnt;
1661 				__u32		flags;
1662 				__u32		pid;
1663 			} uprobe_multi;
1664 		};
1665 	} link_create;
1666 
1667 	struct { /* struct used by BPF_LINK_UPDATE command */
1668 		__u32		link_fd;	/* link fd */
1669 		union {
1670 			/* new program fd to update link with */
1671 			__u32		new_prog_fd;
1672 			/* new struct_ops map fd to update link with */
1673 			__u32           new_map_fd;
1674 		};
1675 		__u32		flags;		/* extra flags */
1676 		union {
1677 			/* expected link's program fd; is specified only if
1678 			 * BPF_F_REPLACE flag is set in flags.
1679 			 */
1680 			__u32		old_prog_fd;
1681 			/* expected link's map fd; is specified only
1682 			 * if BPF_F_REPLACE flag is set.
1683 			 */
1684 			__u32           old_map_fd;
1685 		};
1686 	} link_update;
1687 
1688 	struct {
1689 		__u32		link_fd;
1690 	} link_detach;
1691 
1692 	struct { /* struct used by BPF_ENABLE_STATS command */
1693 		__u32		type;
1694 	} enable_stats;
1695 
1696 	struct { /* struct used by BPF_ITER_CREATE command */
1697 		__u32		link_fd;
1698 		__u32		flags;
1699 	} iter_create;
1700 
1701 	struct { /* struct used by BPF_PROG_BIND_MAP command */
1702 		__u32		prog_fd;
1703 		__u32		map_fd;
1704 		__u32		flags;		/* extra flags */
1705 	} prog_bind_map;
1706 
1707 } __attribute__((aligned(8)));
1708 
1709 /* The description below is an attempt at providing documentation to eBPF
1710  * developers about the multiple available eBPF helper functions. It can be
1711  * parsed and used to produce a manual page. The workflow is the following,
1712  * and requires the rst2man utility:
1713  *
1714  *     $ ./scripts/bpf_doc.py \
1715  *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1716  *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1717  *     $ man /tmp/bpf-helpers.7
1718  *
1719  * Note that in order to produce this external documentation, some RST
1720  * formatting is used in the descriptions to get "bold" and "italics" in
1721  * manual pages. Also note that the few trailing white spaces are
1722  * intentional, removing them would break paragraphs for rst2man.
1723  *
1724  * Start of BPF helper function descriptions:
1725  *
1726  * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1727  * 	Description
1728  * 		Perform a lookup in *map* for an entry associated to *key*.
1729  * 	Return
1730  * 		Map value associated to *key*, or **NULL** if no entry was
1731  * 		found.
1732  *
1733  * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1734  * 	Description
1735  * 		Add or update the value of the entry associated to *key* in
1736  * 		*map* with *value*. *flags* is one of:
1737  *
1738  * 		**BPF_NOEXIST**
1739  * 			The entry for *key* must not exist in the map.
1740  * 		**BPF_EXIST**
1741  * 			The entry for *key* must already exist in the map.
1742  * 		**BPF_ANY**
1743  * 			No condition on the existence of the entry for *key*.
1744  *
1745  * 		Flag value **BPF_NOEXIST** cannot be used for maps of types
1746  * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
1747  * 		elements always exist), the helper would return an error.
1748  * 	Return
1749  * 		0 on success, or a negative error in case of failure.
1750  *
1751  * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1752  * 	Description
1753  * 		Delete entry with *key* from *map*.
1754  * 	Return
1755  * 		0 on success, or a negative error in case of failure.
1756  *
1757  * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1758  * 	Description
1759  * 		For tracing programs, safely attempt to read *size* bytes from
1760  * 		kernel space address *unsafe_ptr* and store the data in *dst*.
1761  *
1762  * 		Generally, use **bpf_probe_read_user**\ () or
1763  * 		**bpf_probe_read_kernel**\ () instead.
1764  * 	Return
1765  * 		0 on success, or a negative error in case of failure.
1766  *
1767  * u64 bpf_ktime_get_ns(void)
1768  * 	Description
1769  * 		Return the time elapsed since system boot, in nanoseconds.
1770  * 		Does not include time the system was suspended.
1771  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1772  * 	Return
1773  * 		Current *ktime*.
1774  *
1775  * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1776  * 	Description
1777  * 		This helper is a "printk()-like" facility for debugging. It
1778  * 		prints a message defined by format *fmt* (of size *fmt_size*)
1779  * 		to file *\/sys/kernel/tracing/trace* from TraceFS, if
1780  * 		available. It can take up to three additional **u64**
1781  * 		arguments (as an eBPF helpers, the total number of arguments is
1782  * 		limited to five).
1783  *
1784  * 		Each time the helper is called, it appends a line to the trace.
1785  * 		Lines are discarded while *\/sys/kernel/tracing/trace* is
1786  * 		open, use *\/sys/kernel/tracing/trace_pipe* to avoid this.
1787  * 		The format of the trace is customizable, and the exact output
1788  * 		one will get depends on the options set in
1789  * 		*\/sys/kernel/tracing/trace_options* (see also the
1790  * 		*README* file under the same directory). However, it usually
1791  * 		defaults to something like:
1792  *
1793  * 		::
1794  *
1795  * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1796  *
1797  * 		In the above:
1798  *
1799  * 			* ``telnet`` is the name of the current task.
1800  * 			* ``470`` is the PID of the current task.
1801  * 			* ``001`` is the CPU number on which the task is
1802  * 			  running.
1803  * 			* In ``.N..``, each character refers to a set of
1804  * 			  options (whether irqs are enabled, scheduling
1805  * 			  options, whether hard/softirqs are running, level of
1806  * 			  preempt_disabled respectively). **N** means that
1807  * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1808  * 			  are set.
1809  * 			* ``419421.045894`` is a timestamp.
1810  * 			* ``0x00000001`` is a fake value used by BPF for the
1811  * 			  instruction pointer register.
1812  * 			* ``<formatted msg>`` is the message formatted with
1813  * 			  *fmt*.
1814  *
1815  * 		The conversion specifiers supported by *fmt* are similar, but
1816  * 		more limited than for printk(). They are **%d**, **%i**,
1817  * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1818  * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1819  * 		of field, padding with zeroes, etc.) is available, and the
1820  * 		helper will return **-EINVAL** (but print nothing) if it
1821  * 		encounters an unknown specifier.
1822  *
1823  * 		Also, note that **bpf_trace_printk**\ () is slow, and should
1824  * 		only be used for debugging purposes. For this reason, a notice
1825  * 		block (spanning several lines) is printed to kernel logs and
1826  * 		states that the helper should not be used "for production use"
1827  * 		the first time this helper is used (or more precisely, when
1828  * 		**trace_printk**\ () buffers are allocated). For passing values
1829  * 		to user space, perf events should be preferred.
1830  * 	Return
1831  * 		The number of bytes written to the buffer, or a negative error
1832  * 		in case of failure.
1833  *
1834  * u32 bpf_get_prandom_u32(void)
1835  * 	Description
1836  * 		Get a pseudo-random number.
1837  *
1838  * 		From a security point of view, this helper uses its own
1839  * 		pseudo-random internal state, and cannot be used to infer the
1840  * 		seed of other random functions in the kernel. However, it is
1841  * 		essential to note that the generator used by the helper is not
1842  * 		cryptographically secure.
1843  * 	Return
1844  * 		A random 32-bit unsigned value.
1845  *
1846  * u32 bpf_get_smp_processor_id(void)
1847  * 	Description
1848  * 		Get the SMP (symmetric multiprocessing) processor id. Note that
1849  * 		all programs run with migration disabled, which means that the
1850  * 		SMP processor id is stable during all the execution of the
1851  * 		program.
1852  * 	Return
1853  * 		The SMP id of the processor running the program.
1854  *
1855  * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1856  * 	Description
1857  * 		Store *len* bytes from address *from* into the packet
1858  * 		associated to *skb*, at *offset*. *flags* are a combination of
1859  * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1860  * 		checksum for the packet after storing the bytes) and
1861  * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1862  * 		**->swhash** and *skb*\ **->l4hash** to 0).
1863  *
1864  * 		A call to this helper is susceptible to change the underlying
1865  * 		packet buffer. Therefore, at load time, all checks on pointers
1866  * 		previously done by the verifier are invalidated and must be
1867  * 		performed again, if the helper is used in combination with
1868  * 		direct packet access.
1869  * 	Return
1870  * 		0 on success, or a negative error in case of failure.
1871  *
1872  * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1873  * 	Description
1874  * 		Recompute the layer 3 (e.g. IP) checksum for the packet
1875  * 		associated to *skb*. Computation is incremental, so the helper
1876  * 		must know the former value of the header field that was
1877  * 		modified (*from*), the new value of this field (*to*), and the
1878  * 		number of bytes (2 or 4) for this field, stored in *size*.
1879  * 		Alternatively, it is possible to store the difference between
1880  * 		the previous and the new values of the header field in *to*, by
1881  * 		setting *from* and *size* to 0. For both methods, *offset*
1882  * 		indicates the location of the IP checksum within the packet.
1883  *
1884  * 		This helper works in combination with **bpf_csum_diff**\ (),
1885  * 		which does not update the checksum in-place, but offers more
1886  * 		flexibility and can handle sizes larger than 2 or 4 for the
1887  * 		checksum to update.
1888  *
1889  * 		A call to this helper is susceptible to change the underlying
1890  * 		packet buffer. Therefore, at load time, all checks on pointers
1891  * 		previously done by the verifier are invalidated and must be
1892  * 		performed again, if the helper is used in combination with
1893  * 		direct packet access.
1894  * 	Return
1895  * 		0 on success, or a negative error in case of failure.
1896  *
1897  * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1898  * 	Description
1899  * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1900  * 		packet associated to *skb*. Computation is incremental, so the
1901  * 		helper must know the former value of the header field that was
1902  * 		modified (*from*), the new value of this field (*to*), and the
1903  * 		number of bytes (2 or 4) for this field, stored on the lowest
1904  * 		four bits of *flags*. Alternatively, it is possible to store
1905  * 		the difference between the previous and the new values of the
1906  * 		header field in *to*, by setting *from* and the four lowest
1907  * 		bits of *flags* to 0. For both methods, *offset* indicates the
1908  * 		location of the IP checksum within the packet. In addition to
1909  * 		the size of the field, *flags* can be added (bitwise OR) actual
1910  * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1911  * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1912  * 		for updates resulting in a null checksum the value is set to
1913  * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1914  * 		the checksum is to be computed against a pseudo-header.
1915  *
1916  * 		This helper works in combination with **bpf_csum_diff**\ (),
1917  * 		which does not update the checksum in-place, but offers more
1918  * 		flexibility and can handle sizes larger than 2 or 4 for the
1919  * 		checksum to update.
1920  *
1921  * 		A call to this helper is susceptible to change the underlying
1922  * 		packet buffer. Therefore, at load time, all checks on pointers
1923  * 		previously done by the verifier are invalidated and must be
1924  * 		performed again, if the helper is used in combination with
1925  * 		direct packet access.
1926  * 	Return
1927  * 		0 on success, or a negative error in case of failure.
1928  *
1929  * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1930  * 	Description
1931  * 		This special helper is used to trigger a "tail call", or in
1932  * 		other words, to jump into another eBPF program. The same stack
1933  * 		frame is used (but values on stack and in registers for the
1934  * 		caller are not accessible to the callee). This mechanism allows
1935  * 		for program chaining, either for raising the maximum number of
1936  * 		available eBPF instructions, or to execute given programs in
1937  * 		conditional blocks. For security reasons, there is an upper
1938  * 		limit to the number of successive tail calls that can be
1939  * 		performed.
1940  *
1941  * 		Upon call of this helper, the program attempts to jump into a
1942  * 		program referenced at index *index* in *prog_array_map*, a
1943  * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1944  * 		*ctx*, a pointer to the context.
1945  *
1946  * 		If the call succeeds, the kernel immediately runs the first
1947  * 		instruction of the new program. This is not a function call,
1948  * 		and it never returns to the previous program. If the call
1949  * 		fails, then the helper has no effect, and the caller continues
1950  * 		to run its subsequent instructions. A call can fail if the
1951  * 		destination program for the jump does not exist (i.e. *index*
1952  * 		is superior to the number of entries in *prog_array_map*), or
1953  * 		if the maximum number of tail calls has been reached for this
1954  * 		chain of programs. This limit is defined in the kernel by the
1955  * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1956  *		which is currently set to 33.
1957  * 	Return
1958  * 		0 on success, or a negative error in case of failure.
1959  *
1960  * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1961  * 	Description
1962  * 		Clone and redirect the packet associated to *skb* to another
1963  * 		net device of index *ifindex*. Both ingress and egress
1964  * 		interfaces can be used for redirection. The **BPF_F_INGRESS**
1965  * 		value in *flags* is used to make the distinction (ingress path
1966  * 		is selected if the flag is present, egress path otherwise).
1967  * 		This is the only flag supported for now.
1968  *
1969  * 		In comparison with **bpf_redirect**\ () helper,
1970  * 		**bpf_clone_redirect**\ () has the associated cost of
1971  * 		duplicating the packet buffer, but this can be executed out of
1972  * 		the eBPF program. Conversely, **bpf_redirect**\ () is more
1973  * 		efficient, but it is handled through an action code where the
1974  * 		redirection happens only after the eBPF program has returned.
1975  *
1976  * 		A call to this helper is susceptible to change the underlying
1977  * 		packet buffer. Therefore, at load time, all checks on pointers
1978  * 		previously done by the verifier are invalidated and must be
1979  * 		performed again, if the helper is used in combination with
1980  * 		direct packet access.
1981  * 	Return
1982  * 		0 on success, or a negative error in case of failure. Positive
1983  * 		error indicates a potential drop or congestion in the target
1984  * 		device. The particular positive error codes are not defined.
1985  *
1986  * u64 bpf_get_current_pid_tgid(void)
1987  * 	Description
1988  * 		Get the current pid and tgid.
1989  * 	Return
1990  * 		A 64-bit integer containing the current tgid and pid, and
1991  * 		created as such:
1992  * 		*current_task*\ **->tgid << 32 \|**
1993  * 		*current_task*\ **->pid**.
1994  *
1995  * u64 bpf_get_current_uid_gid(void)
1996  * 	Description
1997  * 		Get the current uid and gid.
1998  * 	Return
1999  * 		A 64-bit integer containing the current GID and UID, and
2000  * 		created as such: *current_gid* **<< 32 \|** *current_uid*.
2001  *
2002  * long bpf_get_current_comm(void *buf, u32 size_of_buf)
2003  * 	Description
2004  * 		Copy the **comm** attribute of the current task into *buf* of
2005  * 		*size_of_buf*. The **comm** attribute contains the name of
2006  * 		the executable (excluding the path) for the current task. The
2007  * 		*size_of_buf* must be strictly positive. On success, the
2008  * 		helper makes sure that the *buf* is NUL-terminated. On failure,
2009  * 		it is filled with zeroes.
2010  * 	Return
2011  * 		0 on success, or a negative error in case of failure.
2012  *
2013  * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
2014  * 	Description
2015  * 		Retrieve the classid for the current task, i.e. for the net_cls
2016  * 		cgroup to which *skb* belongs.
2017  *
2018  * 		This helper can be used on TC egress path, but not on ingress.
2019  *
2020  * 		The net_cls cgroup provides an interface to tag network packets
2021  * 		based on a user-provided identifier for all traffic coming from
2022  * 		the tasks belonging to the related cgroup. See also the related
2023  * 		kernel documentation, available from the Linux sources in file
2024  * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.
2025  *
2026  * 		The Linux kernel has two versions for cgroups: there are
2027  * 		cgroups v1 and cgroups v2. Both are available to users, who can
2028  * 		use a mixture of them, but note that the net_cls cgroup is for
2029  * 		cgroup v1 only. This makes it incompatible with BPF programs
2030  * 		run on cgroups, which is a cgroup-v2-only feature (a socket can
2031  * 		only hold data for one version of cgroups at a time).
2032  *
2033  * 		This helper is only available is the kernel was compiled with
2034  * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
2035  * 		"**y**" or to "**m**".
2036  * 	Return
2037  * 		The classid, or 0 for the default unconfigured classid.
2038  *
2039  * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
2040  * 	Description
2041  * 		Push a *vlan_tci* (VLAN tag control information) of protocol
2042  * 		*vlan_proto* to the packet associated to *skb*, then update
2043  * 		the checksum. Note that if *vlan_proto* is different from
2044  * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
2045  * 		be **ETH_P_8021Q**.
2046  *
2047  * 		A call to this helper is susceptible to change the underlying
2048  * 		packet buffer. Therefore, at load time, all checks on pointers
2049  * 		previously done by the verifier are invalidated and must be
2050  * 		performed again, if the helper is used in combination with
2051  * 		direct packet access.
2052  * 	Return
2053  * 		0 on success, or a negative error in case of failure.
2054  *
2055  * long bpf_skb_vlan_pop(struct sk_buff *skb)
2056  * 	Description
2057  * 		Pop a VLAN header from the packet associated to *skb*.
2058  *
2059  * 		A call to this helper is susceptible to change the underlying
2060  * 		packet buffer. Therefore, at load time, all checks on pointers
2061  * 		previously done by the verifier are invalidated and must be
2062  * 		performed again, if the helper is used in combination with
2063  * 		direct packet access.
2064  * 	Return
2065  * 		0 on success, or a negative error in case of failure.
2066  *
2067  * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2068  * 	Description
2069  * 		Get tunnel metadata. This helper takes a pointer *key* to an
2070  * 		empty **struct bpf_tunnel_key** of **size**, that will be
2071  * 		filled with tunnel metadata for the packet associated to *skb*.
2072  * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
2073  * 		indicates that the tunnel is based on IPv6 protocol instead of
2074  * 		IPv4.
2075  *
2076  * 		The **struct bpf_tunnel_key** is an object that generalizes the
2077  * 		principal parameters used by various tunneling protocols into a
2078  * 		single struct. This way, it can be used to easily make a
2079  * 		decision based on the contents of the encapsulation header,
2080  * 		"summarized" in this struct. In particular, it holds the IP
2081  * 		address of the remote end (IPv4 or IPv6, depending on the case)
2082  * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
2083  * 		this struct exposes the *key*\ **->tunnel_id**, which is
2084  * 		generally mapped to a VNI (Virtual Network Identifier), making
2085  * 		it programmable together with the **bpf_skb_set_tunnel_key**\
2086  * 		() helper.
2087  *
2088  * 		Let's imagine that the following code is part of a program
2089  * 		attached to the TC ingress interface, on one end of a GRE
2090  * 		tunnel, and is supposed to filter out all messages coming from
2091  * 		remote ends with IPv4 address other than 10.0.0.1:
2092  *
2093  * 		::
2094  *
2095  * 			int ret;
2096  * 			struct bpf_tunnel_key key = {};
2097  *
2098  * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
2099  * 			if (ret < 0)
2100  * 				return TC_ACT_SHOT;	// drop packet
2101  *
2102  * 			if (key.remote_ipv4 != 0x0a000001)
2103  * 				return TC_ACT_SHOT;	// drop packet
2104  *
2105  * 			return TC_ACT_OK;		// accept packet
2106  *
2107  * 		This interface can also be used with all encapsulation devices
2108  * 		that can operate in "collect metadata" mode: instead of having
2109  * 		one network device per specific configuration, the "collect
2110  * 		metadata" mode only requires a single device where the
2111  * 		configuration can be extracted from this helper.
2112  *
2113  * 		This can be used together with various tunnels such as VXLan,
2114  * 		Geneve, GRE or IP in IP (IPIP).
2115  * 	Return
2116  * 		0 on success, or a negative error in case of failure.
2117  *
2118  * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
2119  * 	Description
2120  * 		Populate tunnel metadata for packet associated to *skb.* The
2121  * 		tunnel metadata is set to the contents of *key*, of *size*. The
2122  * 		*flags* can be set to a combination of the following values:
2123  *
2124  * 		**BPF_F_TUNINFO_IPV6**
2125  * 			Indicate that the tunnel is based on IPv6 protocol
2126  * 			instead of IPv4.
2127  * 		**BPF_F_ZERO_CSUM_TX**
2128  * 			For IPv4 packets, add a flag to tunnel metadata
2129  * 			indicating that checksum computation should be skipped
2130  * 			and checksum set to zeroes.
2131  * 		**BPF_F_DONT_FRAGMENT**
2132  * 			Add a flag to tunnel metadata indicating that the
2133  * 			packet should not be fragmented.
2134  * 		**BPF_F_SEQ_NUMBER**
2135  * 			Add a flag to tunnel metadata indicating that a
2136  * 			sequence number should be added to tunnel header before
2137  * 			sending the packet. This flag was added for GRE
2138  * 			encapsulation, but might be used with other protocols
2139  * 			as well in the future.
2140  * 		**BPF_F_NO_TUNNEL_KEY**
2141  * 			Add a flag to tunnel metadata indicating that no tunnel
2142  * 			key should be set in the resulting tunnel header.
2143  *
2144  * 		Here is a typical usage on the transmit path:
2145  *
2146  * 		::
2147  *
2148  * 			struct bpf_tunnel_key key;
2149  * 			     populate key ...
2150  * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2151  * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2152  *
2153  * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()
2154  * 		helper for additional information.
2155  * 	Return
2156  * 		0 on success, or a negative error in case of failure.
2157  *
2158  * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2159  * 	Description
2160  * 		Read the value of a perf event counter. This helper relies on a
2161  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2162  * 		the perf event counter is selected when *map* is updated with
2163  * 		perf event file descriptors. The *map* is an array whose size
2164  * 		is the number of available CPUs, and each cell contains a value
2165  * 		relative to one CPU. The value to retrieve is indicated by
2166  * 		*flags*, that contains the index of the CPU to look up, masked
2167  * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2168  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2169  * 		current CPU should be retrieved.
2170  *
2171  * 		Note that before Linux 4.13, only hardware perf event can be
2172  * 		retrieved.
2173  *
2174  * 		Also, be aware that the newer helper
2175  * 		**bpf_perf_event_read_value**\ () is recommended over
2176  * 		**bpf_perf_event_read**\ () in general. The latter has some ABI
2177  * 		quirks where error and counter value are used as a return code
2178  * 		(which is wrong to do since ranges may overlap). This issue is
2179  * 		fixed with **bpf_perf_event_read_value**\ (), which at the same
2180  * 		time provides more features over the **bpf_perf_event_read**\
2181  * 		() interface. Please refer to the description of
2182  * 		**bpf_perf_event_read_value**\ () for details.
2183  * 	Return
2184  * 		The value of the perf event counter read from the map, or a
2185  * 		negative error code in case of failure.
2186  *
2187  * long bpf_redirect(u32 ifindex, u64 flags)
2188  * 	Description
2189  * 		Redirect the packet to another net device of index *ifindex*.
2190  * 		This helper is somewhat similar to **bpf_clone_redirect**\
2191  * 		(), except that the packet is not cloned, which provides
2192  * 		increased performance.
2193  *
2194  * 		Except for XDP, both ingress and egress interfaces can be used
2195  * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used
2196  * 		to make the distinction (ingress path is selected if the flag
2197  * 		is present, egress path otherwise). Currently, XDP only
2198  * 		supports redirection to the egress interface, and accepts no
2199  * 		flag at all.
2200  *
2201  * 		The same effect can also be attained with the more generic
2202  * 		**bpf_redirect_map**\ (), which uses a BPF map to store the
2203  * 		redirect target instead of providing it directly to the helper.
2204  * 	Return
2205  * 		For XDP, the helper returns **XDP_REDIRECT** on success or
2206  * 		**XDP_ABORTED** on error. For other program types, the values
2207  * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2208  * 		error.
2209  *
2210  * u32 bpf_get_route_realm(struct sk_buff *skb)
2211  * 	Description
2212  * 		Retrieve the realm or the route, that is to say the
2213  * 		**tclassid** field of the destination for the *skb*. The
2214  * 		identifier retrieved is a user-provided tag, similar to the
2215  * 		one used with the net_cls cgroup (see description for
2216  * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is
2217  * 		held by a route (a destination entry), not by a task.
2218  *
2219  * 		Retrieving this identifier works with the clsact TC egress hook
2220  * 		(see also **tc-bpf(8)**), or alternatively on conventional
2221  * 		classful egress qdiscs, but not on TC ingress path. In case of
2222  * 		clsact TC egress hook, this has the advantage that, internally,
2223  * 		the destination entry has not been dropped yet in the transmit
2224  * 		path. Therefore, the destination entry does not need to be
2225  * 		artificially held via **netif_keep_dst**\ () for a classful
2226  * 		qdisc until the *skb* is freed.
2227  *
2228  * 		This helper is available only if the kernel was compiled with
2229  * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.
2230  * 	Return
2231  * 		The realm of the route for the packet associated to *skb*, or 0
2232  * 		if none was found.
2233  *
2234  * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2235  * 	Description
2236  * 		Write raw *data* blob into a special BPF perf event held by
2237  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2238  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
2239  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2240  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2241  *
2242  * 		The *flags* are used to indicate the index in *map* for which
2243  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
2244  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2245  * 		to indicate that the index of the current CPU core should be
2246  * 		used.
2247  *
2248  * 		The value to write, of *size*, is passed through eBPF stack and
2249  * 		pointed by *data*.
2250  *
2251  * 		The context of the program *ctx* needs also be passed to the
2252  * 		helper.
2253  *
2254  * 		On user space, a program willing to read the values needs to
2255  * 		call **perf_event_open**\ () on the perf event (either for
2256  * 		one or for all CPUs) and to store the file descriptor into the
2257  * 		*map*. This must be done before the eBPF program can send data
2258  * 		into it. An example is available in file
2259  * 		*samples/bpf/trace_output_user.c* in the Linux kernel source
2260  * 		tree (the eBPF program counterpart is in
2261  * 		*samples/bpf/trace_output_kern.c*).
2262  *
2263  * 		**bpf_perf_event_output**\ () achieves better performance
2264  * 		than **bpf_trace_printk**\ () for sharing data with user
2265  * 		space, and is much better suitable for streaming data from eBPF
2266  * 		programs.
2267  *
2268  * 		Note that this helper is not restricted to tracing use cases
2269  * 		and can be used with programs attached to TC or XDP as well,
2270  * 		where it allows for passing data to user space listeners. Data
2271  * 		can be:
2272  *
2273  * 		* Only custom structs,
2274  * 		* Only the packet payload, or
2275  * 		* A combination of both.
2276  * 	Return
2277  * 		0 on success, or a negative error in case of failure.
2278  *
2279  * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2280  * 	Description
2281  * 		This helper was provided as an easy way to load data from a
2282  * 		packet. It can be used to load *len* bytes from *offset* from
2283  * 		the packet associated to *skb*, into the buffer pointed by
2284  * 		*to*.
2285  *
2286  * 		Since Linux 4.7, usage of this helper has mostly been replaced
2287  * 		by "direct packet access", enabling packet data to be
2288  * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2289  * 		pointing respectively to the first byte of packet data and to
2290  * 		the byte after the last byte of packet data. However, it
2291  * 		remains useful if one wishes to read large quantities of data
2292  * 		at once from a packet into the eBPF stack.
2293  * 	Return
2294  * 		0 on success, or a negative error in case of failure.
2295  *
2296  * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2297  * 	Description
2298  * 		Walk a user or a kernel stack and return its id. To achieve
2299  * 		this, the helper needs *ctx*, which is a pointer to the context
2300  * 		on which the tracing program is executed, and a pointer to a
2301  * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2302  *
2303  * 		The last argument, *flags*, holds the number of stack frames to
2304  * 		skip (from 0 to 255), masked with
2305  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2306  * 		a combination of the following flags:
2307  *
2308  * 		**BPF_F_USER_STACK**
2309  * 			Collect a user space stack instead of a kernel stack.
2310  * 		**BPF_F_FAST_STACK_CMP**
2311  * 			Compare stacks by hash only.
2312  * 		**BPF_F_REUSE_STACKID**
2313  * 			If two different stacks hash into the same *stackid*,
2314  * 			discard the old one.
2315  *
2316  * 		The stack id retrieved is a 32 bit long integer handle which
2317  * 		can be further combined with other data (including other stack
2318  * 		ids) and used as a key into maps. This can be useful for
2319  * 		generating a variety of graphs (such as flame graphs or off-cpu
2320  * 		graphs).
2321  *
2322  * 		For walking a stack, this helper is an improvement over
2323  * 		**bpf_probe_read**\ (), which can be used with unrolled loops
2324  * 		but is not efficient and consumes a lot of eBPF instructions.
2325  * 		Instead, **bpf_get_stackid**\ () can collect up to
2326  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2327  * 		this limit can be controlled with the **sysctl** program, and
2328  * 		that it should be manually increased in order to profile long
2329  * 		user stacks (such as stacks for Java programs). To do so, use:
2330  *
2331  * 		::
2332  *
2333  * 			# sysctl kernel.perf_event_max_stack=<new value>
2334  * 	Return
2335  * 		The positive or null stack id on success, or a negative error
2336  * 		in case of failure.
2337  *
2338  * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2339  * 	Description
2340  * 		Compute a checksum difference, from the raw buffer pointed by
2341  * 		*from*, of length *from_size* (that must be a multiple of 4),
2342  * 		towards the raw buffer pointed by *to*, of size *to_size*
2343  * 		(same remark). An optional *seed* can be added to the value
2344  * 		(this can be cascaded, the seed may come from a previous call
2345  * 		to the helper).
2346  *
2347  * 		This is flexible enough to be used in several ways:
2348  *
2349  * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to
2350  * 		  checksum, it can be used when pushing new data.
2351  * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to
2352  * 		  checksum, it can be used when removing data from a packet.
2353  * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2354  * 		  can be used to compute a diff. Note that *from_size* and
2355  * 		  *to_size* do not need to be equal.
2356  *
2357  * 		This helper can be used in combination with
2358  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2359  * 		which one can feed in the difference computed with
2360  * 		**bpf_csum_diff**\ ().
2361  * 	Return
2362  * 		The checksum result, or a negative error code in case of
2363  * 		failure.
2364  *
2365  * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2366  * 	Description
2367  * 		Retrieve tunnel options metadata for the packet associated to
2368  * 		*skb*, and store the raw tunnel option data to the buffer *opt*
2369  * 		of *size*.
2370  *
2371  * 		This helper can be used with encapsulation devices that can
2372  * 		operate in "collect metadata" mode (please refer to the related
2373  * 		note in the description of **bpf_skb_get_tunnel_key**\ () for
2374  * 		more details). A particular example where this can be used is
2375  * 		in combination with the Geneve encapsulation protocol, where it
2376  * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2377  * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from
2378  * 		the eBPF program. This allows for full customization of these
2379  * 		headers.
2380  * 	Return
2381  * 		The size of the option data retrieved.
2382  *
2383  * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2384  * 	Description
2385  * 		Set tunnel options metadata for the packet associated to *skb*
2386  * 		to the option data contained in the raw buffer *opt* of *size*.
2387  *
2388  * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2389  * 		helper for additional information.
2390  * 	Return
2391  * 		0 on success, or a negative error in case of failure.
2392  *
2393  * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2394  * 	Description
2395  * 		Change the protocol of the *skb* to *proto*. Currently
2396  * 		supported are transition from IPv4 to IPv6, and from IPv6 to
2397  * 		IPv4. The helper takes care of the groundwork for the
2398  * 		transition, including resizing the socket buffer. The eBPF
2399  * 		program is expected to fill the new headers, if any, via
2400  * 		**skb_store_bytes**\ () and to recompute the checksums with
2401  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2402  * 		(). The main case for this helper is to perform NAT64
2403  * 		operations out of an eBPF program.
2404  *
2405  * 		Internally, the GSO type is marked as dodgy so that headers are
2406  * 		checked and segments are recalculated by the GSO/GRO engine.
2407  * 		The size for GSO target is adapted as well.
2408  *
2409  * 		All values for *flags* are reserved for future usage, and must
2410  * 		be left at zero.
2411  *
2412  * 		A call to this helper is susceptible to change the underlying
2413  * 		packet buffer. Therefore, at load time, all checks on pointers
2414  * 		previously done by the verifier are invalidated and must be
2415  * 		performed again, if the helper is used in combination with
2416  * 		direct packet access.
2417  * 	Return
2418  * 		0 on success, or a negative error in case of failure.
2419  *
2420  * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2421  * 	Description
2422  * 		Change the packet type for the packet associated to *skb*. This
2423  * 		comes down to setting *skb*\ **->pkt_type** to *type*, except
2424  * 		the eBPF program does not have a write access to *skb*\
2425  * 		**->pkt_type** beside this helper. Using a helper here allows
2426  * 		for graceful handling of errors.
2427  *
2428  * 		The major use case is to change incoming *skb*s to
2429  * 		**PACKET_HOST** in a programmatic way instead of having to
2430  * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2431  * 		example.
2432  *
2433  * 		Note that *type* only allows certain values. At this time, they
2434  * 		are:
2435  *
2436  * 		**PACKET_HOST**
2437  * 			Packet is for us.
2438  * 		**PACKET_BROADCAST**
2439  * 			Send packet to all.
2440  * 		**PACKET_MULTICAST**
2441  * 			Send packet to group.
2442  * 		**PACKET_OTHERHOST**
2443  * 			Send packet to someone else.
2444  * 	Return
2445  * 		0 on success, or a negative error in case of failure.
2446  *
2447  * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2448  * 	Description
2449  * 		Check whether *skb* is a descendant of the cgroup2 held by
2450  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2451  * 	Return
2452  * 		The return value depends on the result of the test, and can be:
2453  *
2454  * 		* 0, if the *skb* failed the cgroup2 descendant test.
2455  * 		* 1, if the *skb* succeeded the cgroup2 descendant test.
2456  * 		* A negative error code, if an error occurred.
2457  *
2458  * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2459  * 	Description
2460  * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2461  * 		not set, in particular if the hash was cleared due to mangling,
2462  * 		recompute this hash. Later accesses to the hash can be done
2463  * 		directly with *skb*\ **->hash**.
2464  *
2465  * 		Calling **bpf_set_hash_invalid**\ (), changing a packet
2466  * 		prototype with **bpf_skb_change_proto**\ (), or calling
2467  * 		**bpf_skb_store_bytes**\ () with the
2468  * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2469  * 		the hash and to trigger a new computation for the next call to
2470  * 		**bpf_get_hash_recalc**\ ().
2471  * 	Return
2472  * 		The 32-bit hash.
2473  *
2474  * u64 bpf_get_current_task(void)
2475  * 	Description
2476  * 		Get the current task.
2477  * 	Return
2478  * 		A pointer to the current task struct.
2479  *
2480  * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2481  * 	Description
2482  * 		Attempt in a safe way to write *len* bytes from the buffer
2483  * 		*src* to *dst* in memory. It only works for threads that are in
2484  * 		user context, and *dst* must be a valid user space address.
2485  *
2486  * 		This helper should not be used to implement any kind of
2487  * 		security mechanism because of TOC-TOU attacks, but rather to
2488  * 		debug, divert, and manipulate execution of semi-cooperative
2489  * 		processes.
2490  *
2491  * 		Keep in mind that this feature is meant for experiments, and it
2492  * 		has a risk of crashing the system and running programs.
2493  * 		Therefore, when an eBPF program using this helper is attached,
2494  * 		a warning including PID and process name is printed to kernel
2495  * 		logs.
2496  * 	Return
2497  * 		0 on success, or a negative error in case of failure.
2498  *
2499  * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2500  * 	Description
2501  * 		Check whether the probe is being run is the context of a given
2502  * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2503  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2504  * 	Return
2505  * 		The return value depends on the result of the test, and can be:
2506  *
2507  *		* 1, if current task belongs to the cgroup2.
2508  *		* 0, if current task does not belong to the cgroup2.
2509  * 		* A negative error code, if an error occurred.
2510  *
2511  * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2512  * 	Description
2513  * 		Resize (trim or grow) the packet associated to *skb* to the
2514  * 		new *len*. The *flags* are reserved for future usage, and must
2515  * 		be left at zero.
2516  *
2517  * 		The basic idea is that the helper performs the needed work to
2518  * 		change the size of the packet, then the eBPF program rewrites
2519  * 		the rest via helpers like **bpf_skb_store_bytes**\ (),
2520  * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2521  * 		and others. This helper is a slow path utility intended for
2522  * 		replies with control messages. And because it is targeted for
2523  * 		slow path, the helper itself can afford to be slow: it
2524  * 		implicitly linearizes, unclones and drops offloads from the
2525  * 		*skb*.
2526  *
2527  * 		A call to this helper is susceptible to change the underlying
2528  * 		packet buffer. Therefore, at load time, all checks on pointers
2529  * 		previously done by the verifier are invalidated and must be
2530  * 		performed again, if the helper is used in combination with
2531  * 		direct packet access.
2532  * 	Return
2533  * 		0 on success, or a negative error in case of failure.
2534  *
2535  * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2536  * 	Description
2537  * 		Pull in non-linear data in case the *skb* is non-linear and not
2538  * 		all of *len* are part of the linear section. Make *len* bytes
2539  * 		from *skb* readable and writable. If a zero value is passed for
2540  *		*len*, then all bytes in the linear part of *skb* will be made
2541  *		readable and writable.
2542  *
2543  * 		This helper is only needed for reading and writing with direct
2544  * 		packet access.
2545  *
2546  * 		For direct packet access, testing that offsets to access
2547  * 		are within packet boundaries (test on *skb*\ **->data_end**) is
2548  * 		susceptible to fail if offsets are invalid, or if the requested
2549  * 		data is in non-linear parts of the *skb*. On failure the
2550  * 		program can just bail out, or in the case of a non-linear
2551  * 		buffer, use a helper to make the data available. The
2552  * 		**bpf_skb_load_bytes**\ () helper is a first solution to access
2553  * 		the data. Another one consists in using **bpf_skb_pull_data**
2554  * 		to pull in once the non-linear parts, then retesting and
2555  * 		eventually access the data.
2556  *
2557  * 		At the same time, this also makes sure the *skb* is uncloned,
2558  * 		which is a necessary condition for direct write. As this needs
2559  * 		to be an invariant for the write part only, the verifier
2560  * 		detects writes and adds a prologue that is calling
2561  * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from
2562  * 		the very beginning in case it is indeed cloned.
2563  *
2564  * 		A call to this helper is susceptible to change the underlying
2565  * 		packet buffer. Therefore, at load time, all checks on pointers
2566  * 		previously done by the verifier are invalidated and must be
2567  * 		performed again, if the helper is used in combination with
2568  * 		direct packet access.
2569  * 	Return
2570  * 		0 on success, or a negative error in case of failure.
2571  *
2572  * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2573  * 	Description
2574  * 		Add the checksum *csum* into *skb*\ **->csum** in case the
2575  * 		driver has supplied a checksum for the entire packet into that
2576  * 		field. Return an error otherwise. This helper is intended to be
2577  * 		used in combination with **bpf_csum_diff**\ (), in particular
2578  * 		when the checksum needs to be updated after data has been
2579  * 		written into the packet through direct packet access.
2580  * 	Return
2581  * 		The checksum on success, or a negative error code in case of
2582  * 		failure.
2583  *
2584  * void bpf_set_hash_invalid(struct sk_buff *skb)
2585  * 	Description
2586  * 		Invalidate the current *skb*\ **->hash**. It can be used after
2587  * 		mangling on headers through direct packet access, in order to
2588  * 		indicate that the hash is outdated and to trigger a
2589  * 		recalculation the next time the kernel tries to access this
2590  * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.
2591  * 	Return
2592  * 		void.
2593  *
2594  * long bpf_get_numa_node_id(void)
2595  * 	Description
2596  * 		Return the id of the current NUMA node. The primary use case
2597  * 		for this helper is the selection of sockets for the local NUMA
2598  * 		node, when the program is attached to sockets using the
2599  * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2600  * 		but the helper is also available to other eBPF program types,
2601  * 		similarly to **bpf_get_smp_processor_id**\ ().
2602  * 	Return
2603  * 		The id of current NUMA node.
2604  *
2605  * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2606  * 	Description
2607  * 		Grows headroom of packet associated to *skb* and adjusts the
2608  * 		offset of the MAC header accordingly, adding *len* bytes of
2609  * 		space. It automatically extends and reallocates memory as
2610  * 		required.
2611  *
2612  * 		This helper can be used on a layer 3 *skb* to push a MAC header
2613  * 		for redirection into a layer 2 device.
2614  *
2615  * 		All values for *flags* are reserved for future usage, and must
2616  * 		be left at zero.
2617  *
2618  * 		A call to this helper is susceptible to change the underlying
2619  * 		packet buffer. Therefore, at load time, all checks on pointers
2620  * 		previously done by the verifier are invalidated and must be
2621  * 		performed again, if the helper is used in combination with
2622  * 		direct packet access.
2623  * 	Return
2624  * 		0 on success, or a negative error in case of failure.
2625  *
2626  * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2627  * 	Description
2628  * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2629  * 		it is possible to use a negative value for *delta*. This helper
2630  * 		can be used to prepare the packet for pushing or popping
2631  * 		headers.
2632  *
2633  * 		A call to this helper is susceptible to change the underlying
2634  * 		packet buffer. Therefore, at load time, all checks on pointers
2635  * 		previously done by the verifier are invalidated and must be
2636  * 		performed again, if the helper is used in combination with
2637  * 		direct packet access.
2638  * 	Return
2639  * 		0 on success, or a negative error in case of failure.
2640  *
2641  * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2642  * 	Description
2643  * 		Copy a NUL terminated string from an unsafe kernel address
2644  * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2645  * 		more details.
2646  *
2647  * 		Generally, use **bpf_probe_read_user_str**\ () or
2648  * 		**bpf_probe_read_kernel_str**\ () instead.
2649  * 	Return
2650  * 		On success, the strictly positive length of the string,
2651  * 		including the trailing NUL character. On error, a negative
2652  * 		value.
2653  *
2654  * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2655  * 	Description
2656  * 		If the **struct sk_buff** pointed by *skb* has a known socket,
2657  * 		retrieve the cookie (generated by the kernel) of this socket.
2658  * 		If no cookie has been set yet, generate a new cookie. Once
2659  * 		generated, the socket cookie remains stable for the life of the
2660  * 		socket. This helper can be useful for monitoring per socket
2661  * 		networking traffic statistics as it provides a global socket
2662  * 		identifier that can be assumed unique.
2663  * 	Return
2664  * 		A 8-byte long unique number on success, or 0 if the socket
2665  * 		field is missing inside *skb*.
2666  *
2667  * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2668  * 	Description
2669  * 		Equivalent to bpf_get_socket_cookie() helper that accepts
2670  * 		*skb*, but gets socket from **struct bpf_sock_addr** context.
2671  * 	Return
2672  * 		A 8-byte long unique number.
2673  *
2674  * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2675  * 	Description
2676  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2677  * 		*skb*, but gets socket from **struct bpf_sock_ops** context.
2678  * 	Return
2679  * 		A 8-byte long unique number.
2680  *
2681  * u64 bpf_get_socket_cookie(struct sock *sk)
2682  * 	Description
2683  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2684  * 		*sk*, but gets socket from a BTF **struct sock**. This helper
2685  * 		also works for sleepable programs.
2686  * 	Return
2687  * 		A 8-byte long unique number or 0 if *sk* is NULL.
2688  *
2689  * u32 bpf_get_socket_uid(struct sk_buff *skb)
2690  * 	Description
2691  * 		Get the owner UID of the socked associated to *skb*.
2692  * 	Return
2693  * 		The owner UID of the socket associated to *skb*. If the socket
2694  * 		is **NULL**, or if it is not a full socket (i.e. if it is a
2695  * 		time-wait or a request socket instead), **overflowuid** value
2696  * 		is returned (note that **overflowuid** might also be the actual
2697  * 		UID value for the socket).
2698  *
2699  * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2700  * 	Description
2701  * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)
2702  * 		to value *hash*.
2703  * 	Return
2704  * 		0
2705  *
2706  * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2707  * 	Description
2708  * 		Emulate a call to **setsockopt()** on the socket associated to
2709  * 		*bpf_socket*, which must be a full socket. The *level* at
2710  * 		which the option resides and the name *optname* of the option
2711  * 		must be specified, see **setsockopt(2)** for more information.
2712  * 		The option value of length *optlen* is pointed by *optval*.
2713  *
2714  * 		*bpf_socket* should be one of the following:
2715  *
2716  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2717  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2718  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2719  *
2720  * 		This helper actually implements a subset of **setsockopt()**.
2721  * 		It supports the following *level*\ s:
2722  *
2723  * 		* **SOL_SOCKET**, which supports the following *optname*\ s:
2724  * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2725  * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2726  * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2727  * 		  **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2728  * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:
2729  * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,
2730  * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2731  * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2732  * 		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2733  * 		  **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2734  * 		  **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2735  * 		  **TCP_BPF_RTO_MIN**.
2736  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2737  * 		* **IPPROTO_IPV6**, which supports the following *optname*\ s:
2738  * 		  **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2739  * 	Return
2740  * 		0 on success, or a negative error in case of failure.
2741  *
2742  * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2743  * 	Description
2744  * 		Grow or shrink the room for data in the packet associated to
2745  * 		*skb* by *len_diff*, and according to the selected *mode*.
2746  *
2747  * 		By default, the helper will reset any offloaded checksum
2748  * 		indicator of the skb to CHECKSUM_NONE. This can be avoided
2749  * 		by the following flag:
2750  *
2751  * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2752  * 		  checksum data of the skb to CHECKSUM_NONE.
2753  *
2754  *		There are two supported modes at this time:
2755  *
2756  *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2757  * 		  (room space is added or removed between the layer 2 and
2758  * 		  layer 3 headers).
2759  *
2760  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2761  * 		  (room space is added or removed between the layer 3 and
2762  * 		  layer 4 headers).
2763  *
2764  *		The following flags are supported at this time:
2765  *
2766  *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2767  *		  Adjusting mss in this way is not allowed for datagrams.
2768  *
2769  *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2770  *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2771  *		  Any new space is reserved to hold a tunnel header.
2772  *		  Configure skb offsets and other fields accordingly.
2773  *
2774  *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2775  *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2776  *		  Use with ENCAP_L3 flags to further specify the tunnel type.
2777  *
2778  *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2779  *		  Use with ENCAP_L3/L4 flags to further specify the tunnel
2780  *		  type; *len* is the length of the inner MAC header.
2781  *
2782  *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2783  *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2784  *		  L2 type as Ethernet.
2785  *
2786  *		* **BPF_F_ADJ_ROOM_DECAP_L3_IPV4**,
2787  *		  **BPF_F_ADJ_ROOM_DECAP_L3_IPV6**:
2788  *		  Indicate the new IP header version after decapsulating the outer
2789  *		  IP header. Used when the inner and outer IP versions are different.
2790  *
2791  * 		A call to this helper is susceptible to change the underlying
2792  * 		packet buffer. Therefore, at load time, all checks on pointers
2793  * 		previously done by the verifier are invalidated and must be
2794  * 		performed again, if the helper is used in combination with
2795  * 		direct packet access.
2796  * 	Return
2797  * 		0 on success, or a negative error in case of failure.
2798  *
2799  * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2800  * 	Description
2801  * 		Redirect the packet to the endpoint referenced by *map* at
2802  * 		index *key*. Depending on its type, this *map* can contain
2803  * 		references to net devices (for forwarding packets through other
2804  * 		ports), or to CPUs (for redirecting XDP frames to another CPU;
2805  * 		but this is only implemented for native XDP (with driver
2806  * 		support) as of this writing).
2807  *
2808  * 		The lower two bits of *flags* are used as the return code if
2809  * 		the map lookup fails. This is so that the return value can be
2810  * 		one of the XDP program return codes up to **XDP_TX**, as chosen
2811  * 		by the caller. The higher bits of *flags* can be set to
2812  * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2813  *
2814  * 		With BPF_F_BROADCAST the packet will be broadcasted to all the
2815  * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2816  * 		interface will be excluded when do broadcasting.
2817  *
2818  * 		See also **bpf_redirect**\ (), which only supports redirecting
2819  * 		to an ifindex, but doesn't require a map to do so.
2820  * 	Return
2821  * 		**XDP_REDIRECT** on success, or the value of the two lower bits
2822  * 		of the *flags* argument on error.
2823  *
2824  * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2825  * 	Description
2826  * 		Redirect the packet to the socket referenced by *map* (of type
2827  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2828  * 		egress interfaces can be used for redirection. The
2829  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2830  * 		distinction (ingress path is selected if the flag is present,
2831  * 		egress path otherwise). This is the only flag supported for now.
2832  * 	Return
2833  * 		**SK_PASS** on success, or **SK_DROP** on error.
2834  *
2835  * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2836  * 	Description
2837  * 		Add an entry to, or update a *map* referencing sockets. The
2838  * 		*skops* is used as a new value for the entry associated to
2839  * 		*key*. *flags* is one of:
2840  *
2841  * 		**BPF_NOEXIST**
2842  * 			The entry for *key* must not exist in the map.
2843  * 		**BPF_EXIST**
2844  * 			The entry for *key* must already exist in the map.
2845  * 		**BPF_ANY**
2846  * 			No condition on the existence of the entry for *key*.
2847  *
2848  * 		If the *map* has eBPF programs (parser and verdict), those will
2849  * 		be inherited by the socket being added. If the socket is
2850  * 		already attached to eBPF programs, this results in an error.
2851  * 	Return
2852  * 		0 on success, or a negative error in case of failure.
2853  *
2854  * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2855  * 	Description
2856  * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by
2857  * 		*delta* (which can be positive or negative). Note that this
2858  * 		operation modifies the address stored in *xdp_md*\ **->data**,
2859  * 		so the latter must be loaded only after the helper has been
2860  * 		called.
2861  *
2862  * 		The use of *xdp_md*\ **->data_meta** is optional and programs
2863  * 		are not required to use it. The rationale is that when the
2864  * 		packet is processed with XDP (e.g. as DoS filter), it is
2865  * 		possible to push further meta data along with it before passing
2866  * 		to the stack, and to give the guarantee that an ingress eBPF
2867  * 		program attached as a TC classifier on the same device can pick
2868  * 		this up for further post-processing. Since TC works with socket
2869  * 		buffers, it remains possible to set from XDP the **mark** or
2870  * 		**priority** pointers, or other pointers for the socket buffer.
2871  * 		Having this scratch space generic and programmable allows for
2872  * 		more flexibility as the user is free to store whatever meta
2873  * 		data they need.
2874  *
2875  * 		A call to this helper is susceptible to change the underlying
2876  * 		packet buffer. Therefore, at load time, all checks on pointers
2877  * 		previously done by the verifier are invalidated and must be
2878  * 		performed again, if the helper is used in combination with
2879  * 		direct packet access.
2880  * 	Return
2881  * 		0 on success, or a negative error in case of failure.
2882  *
2883  * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2884  * 	Description
2885  * 		Read the value of a perf event counter, and store it into *buf*
2886  * 		of size *buf_size*. This helper relies on a *map* of type
2887  * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2888  * 		counter is selected when *map* is updated with perf event file
2889  * 		descriptors. The *map* is an array whose size is the number of
2890  * 		available CPUs, and each cell contains a value relative to one
2891  * 		CPU. The value to retrieve is indicated by *flags*, that
2892  * 		contains the index of the CPU to look up, masked with
2893  * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2894  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2895  * 		current CPU should be retrieved.
2896  *
2897  * 		This helper behaves in a way close to
2898  * 		**bpf_perf_event_read**\ () helper, save that instead of
2899  * 		just returning the value observed, it fills the *buf*
2900  * 		structure. This allows for additional data to be retrieved: in
2901  * 		particular, the enabled and running times (in *buf*\
2902  * 		**->enabled** and *buf*\ **->running**, respectively) are
2903  * 		copied. In general, **bpf_perf_event_read_value**\ () is
2904  * 		recommended over **bpf_perf_event_read**\ (), which has some
2905  * 		ABI issues and provides fewer functionalities.
2906  *
2907  * 		These values are interesting, because hardware PMU (Performance
2908  * 		Monitoring Unit) counters are limited resources. When there are
2909  * 		more PMU based perf events opened than available counters,
2910  * 		kernel will multiplex these events so each event gets certain
2911  * 		percentage (but not all) of the PMU time. In case that
2912  * 		multiplexing happens, the number of samples or counter value
2913  * 		will not reflect the case compared to when no multiplexing
2914  * 		occurs. This makes comparison between different runs difficult.
2915  * 		Typically, the counter value should be normalized before
2916  * 		comparing to other experiments. The usual normalization is done
2917  * 		as follows.
2918  *
2919  * 		::
2920  *
2921  * 			normalized_counter = counter * t_enabled / t_running
2922  *
2923  * 		Where t_enabled is the time enabled for event and t_running is
2924  * 		the time running for event since last normalization. The
2925  * 		enabled and running times are accumulated since the perf event
2926  * 		open. To achieve scaling factor between two invocations of an
2927  * 		eBPF program, users can use CPU id as the key (which is
2928  * 		typical for perf array usage model) to remember the previous
2929  * 		value and do the calculation inside the eBPF program.
2930  * 	Return
2931  * 		0 on success, or a negative error in case of failure.
2932  *
2933  * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2934  * 	Description
2935  * 		For an eBPF program attached to a perf event, retrieve the
2936  * 		value of the event counter associated to *ctx* and store it in
2937  * 		the structure pointed by *buf* and of size *buf_size*. Enabled
2938  * 		and running times are also stored in the structure (see
2939  * 		description of helper **bpf_perf_event_read_value**\ () for
2940  * 		more details).
2941  * 	Return
2942  * 		0 on success, or a negative error in case of failure.
2943  *
2944  * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2945  * 	Description
2946  * 		Emulate a call to **getsockopt()** on the socket associated to
2947  * 		*bpf_socket*, which must be a full socket. The *level* at
2948  * 		which the option resides and the name *optname* of the option
2949  * 		must be specified, see **getsockopt(2)** for more information.
2950  * 		The retrieved value is stored in the structure pointed by
2951  * 		*opval* and of length *optlen*.
2952  *
2953  * 		*bpf_socket* should be one of the following:
2954  *
2955  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2956  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2957  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2958  *
2959  * 		This helper actually implements a subset of **getsockopt()**.
2960  * 		It supports the same set of *optname*\ s that is supported by
2961  * 		the **bpf_setsockopt**\ () helper.  The exceptions are
2962  * 		**TCP_BPF_*** is **bpf_setsockopt**\ () only and
2963  * 		**TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
2964  * 	Return
2965  * 		0 on success, or a negative error in case of failure.
2966  *
2967  * long bpf_override_return(struct pt_regs *regs, u64 rc)
2968  * 	Description
2969  * 		Used for error injection, this helper uses kprobes to override
2970  * 		the return value of the probed function, and to set it to *rc*.
2971  * 		The first argument is the context *regs* on which the kprobe
2972  * 		works.
2973  *
2974  * 		This helper works by setting the PC (program counter)
2975  * 		to an override function which is run in place of the original
2976  * 		probed function. This means the probed function is not run at
2977  * 		all. The replacement function just returns with the required
2978  * 		value.
2979  *
2980  * 		This helper has security implications, and thus is subject to
2981  * 		restrictions. It is only available if the kernel was compiled
2982  * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2983  * 		option, and in this case it only works on functions tagged with
2984  * 		**ALLOW_ERROR_INJECTION** in the kernel code.
2985  *
2986  * 		Also, the helper is only available for the architectures having
2987  * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2988  * 		x86 architecture is the only one to support this feature.
2989  * 	Return
2990  * 		0
2991  *
2992  * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2993  * 	Description
2994  * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2995  * 		for the full TCP socket associated to *bpf_sock_ops* to
2996  * 		*argval*.
2997  *
2998  * 		The primary use of this field is to determine if there should
2999  * 		be calls to eBPF programs of type
3000  * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
3001  * 		code. A program of the same type can change its value, per
3002  * 		connection and as necessary, when the connection is
3003  * 		established. This field is directly accessible for reading, but
3004  * 		this helper must be used for updates in order to return an
3005  * 		error if an eBPF program tries to set a callback that is not
3006  * 		supported in the current kernel.
3007  *
3008  * 		*argval* is a flag array which can combine these flags:
3009  *
3010  * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
3011  * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
3012  * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
3013  * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
3014  *
3015  * 		Therefore, this function can be used to clear a callback flag by
3016  * 		setting the appropriate bit to zero. e.g. to disable the RTO
3017  * 		callback:
3018  *
3019  * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**
3020  * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
3021  *
3022  * 		Here are some examples of where one could call such eBPF
3023  * 		program:
3024  *
3025  * 		* When RTO fires.
3026  * 		* When a packet is retransmitted.
3027  * 		* When the connection terminates.
3028  * 		* When a packet is sent.
3029  * 		* When a packet is received.
3030  * 	Return
3031  * 		Code **-EINVAL** if the socket is not a full TCP socket;
3032  * 		otherwise, a positive number containing the bits that could not
3033  * 		be set is returned (which comes down to 0 if all bits were set
3034  * 		as required).
3035  *
3036  * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
3037  * 	Description
3038  * 		This helper is used in programs implementing policies at the
3039  * 		socket level. If the message *msg* is allowed to pass (i.e. if
3040  * 		the verdict eBPF program returns **SK_PASS**), redirect it to
3041  * 		the socket referenced by *map* (of type
3042  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
3043  * 		egress interfaces can be used for redirection. The
3044  * 		**BPF_F_INGRESS** value in *flags* is used to make the
3045  * 		distinction (ingress path is selected if the flag is present,
3046  * 		egress path otherwise). This is the only flag supported for now.
3047  * 	Return
3048  * 		**SK_PASS** on success, or **SK_DROP** on error.
3049  *
3050  * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
3051  * 	Description
3052  * 		For socket policies, apply the verdict of the eBPF program to
3053  * 		the next *bytes* (number of bytes) of message *msg*.
3054  *
3055  * 		For example, this helper can be used in the following cases:
3056  *
3057  * 		* A single **sendmsg**\ () or **sendfile**\ () system call
3058  * 		  contains multiple logical messages that the eBPF program is
3059  * 		  supposed to read and for which it should apply a verdict.
3060  * 		* An eBPF program only cares to read the first *bytes* of a
3061  * 		  *msg*. If the message has a large payload, then setting up
3062  * 		  and calling the eBPF program repeatedly for all bytes, even
3063  * 		  though the verdict is already known, would create unnecessary
3064  * 		  overhead.
3065  *
3066  * 		When called from within an eBPF program, the helper sets a
3067  * 		counter internal to the BPF infrastructure, that is used to
3068  * 		apply the last verdict to the next *bytes*. If *bytes* is
3069  * 		smaller than the current data being processed from a
3070  * 		**sendmsg**\ () or **sendfile**\ () system call, the first
3071  * 		*bytes* will be sent and the eBPF program will be re-run with
3072  * 		the pointer for start of data pointing to byte number *bytes*
3073  * 		**+ 1**. If *bytes* is larger than the current data being
3074  * 		processed, then the eBPF verdict will be applied to multiple
3075  * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are
3076  * 		consumed.
3077  *
3078  * 		Note that if a socket closes with the internal counter holding
3079  * 		a non-zero value, this is not a problem because data is not
3080  * 		being buffered for *bytes* and is sent as it is received.
3081  * 	Return
3082  * 		0
3083  *
3084  * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
3085  * 	Description
3086  * 		For socket policies, prevent the execution of the verdict eBPF
3087  * 		program for message *msg* until *bytes* (byte number) have been
3088  * 		accumulated.
3089  *
3090  * 		This can be used when one needs a specific number of bytes
3091  * 		before a verdict can be assigned, even if the data spans
3092  * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
3093  * 		case would be a user calling **sendmsg**\ () repeatedly with
3094  * 		1-byte long message segments. Obviously, this is bad for
3095  * 		performance, but it is still valid. If the eBPF program needs
3096  * 		*bytes* bytes to validate a header, this helper can be used to
3097  * 		prevent the eBPF program to be called again until *bytes* have
3098  * 		been accumulated.
3099  * 	Return
3100  * 		0
3101  *
3102  * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
3103  * 	Description
3104  * 		For socket policies, pull in non-linear data from user space
3105  * 		for *msg* and set pointers *msg*\ **->data** and *msg*\
3106  * 		**->data_end** to *start* and *end* bytes offsets into *msg*,
3107  * 		respectively.
3108  *
3109  * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3110  * 		*msg* it can only parse data that the (**data**, **data_end**)
3111  * 		pointers have already consumed. For **sendmsg**\ () hooks this
3112  * 		is likely the first scatterlist element. But for calls relying
3113  * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will
3114  * 		be the range (**0**, **0**) because the data is shared with
3115  * 		user space and by default the objective is to avoid allowing
3116  * 		user space to modify data while (or after) eBPF verdict is
3117  * 		being decided. This helper can be used to pull in data and to
3118  * 		set the start and end pointer to given values. Data will be
3119  * 		copied if necessary (i.e. if data was not linear and if start
3120  * 		and end pointers do not point to the same chunk).
3121  *
3122  * 		A call to this helper is susceptible to change the underlying
3123  * 		packet buffer. Therefore, at load time, all checks on pointers
3124  * 		previously done by the verifier are invalidated and must be
3125  * 		performed again, if the helper is used in combination with
3126  * 		direct packet access.
3127  *
3128  * 		All values for *flags* are reserved for future usage, and must
3129  * 		be left at zero.
3130  * 	Return
3131  * 		0 on success, or a negative error in case of failure.
3132  *
3133  * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
3134  * 	Description
3135  * 		Bind the socket associated to *ctx* to the address pointed by
3136  * 		*addr*, of length *addr_len*. This allows for making outgoing
3137  * 		connection from the desired IP address, which can be useful for
3138  * 		example when all processes inside a cgroup should use one
3139  * 		single IP address on a host that has multiple IP configured.
3140  *
3141  * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3142  * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3143  * 		**AF_INET6**). It's advised to pass zero port (**sin_port**
3144  * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3145  * 		behavior and lets the kernel efficiently pick up an unused
3146  * 		port as long as 4-tuple is unique. Passing non-zero port might
3147  * 		lead to degraded performance.
3148  * 	Return
3149  * 		0 on success, or a negative error in case of failure.
3150  *
3151  * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3152  * 	Description
3153  * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3154  * 		possible to both shrink and grow the packet tail.
3155  * 		Shrink done via *delta* being a negative integer.
3156  *
3157  * 		A call to this helper is susceptible to change the underlying
3158  * 		packet buffer. Therefore, at load time, all checks on pointers
3159  * 		previously done by the verifier are invalidated and must be
3160  * 		performed again, if the helper is used in combination with
3161  * 		direct packet access.
3162  * 	Return
3163  * 		0 on success, or a negative error in case of failure.
3164  *
3165  * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3166  * 	Description
3167  * 		Retrieve the XFRM state (IP transform framework, see also
3168  * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3169  *
3170  * 		The retrieved value is stored in the **struct bpf_xfrm_state**
3171  * 		pointed by *xfrm_state* and of length *size*.
3172  *
3173  * 		All values for *flags* are reserved for future usage, and must
3174  * 		be left at zero.
3175  *
3176  * 		This helper is available only if the kernel was compiled with
3177  * 		**CONFIG_XFRM** configuration option.
3178  * 	Return
3179  * 		0 on success, or a negative error in case of failure.
3180  *
3181  * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3182  * 	Description
3183  * 		Return a user or a kernel stack in bpf program provided buffer.
3184  * 		To achieve this, the helper needs *ctx*, which is a pointer
3185  * 		to the context on which the tracing program is executed.
3186  * 		To store the stacktrace, the bpf program provides *buf* with
3187  * 		a nonnegative *size*.
3188  *
3189  * 		The last argument, *flags*, holds the number of stack frames to
3190  * 		skip (from 0 to 255), masked with
3191  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3192  * 		the following flags:
3193  *
3194  * 		**BPF_F_USER_STACK**
3195  * 			Collect a user space stack instead of a kernel stack.
3196  * 		**BPF_F_USER_BUILD_ID**
3197  * 			Collect (build_id, file_offset) instead of ips for user
3198  * 			stack, only valid if **BPF_F_USER_STACK** is also
3199  * 			specified.
3200  *
3201  * 			*file_offset* is an offset relative to the beginning
3202  * 			of the executable or shared object file backing the vma
3203  * 			which the *ip* falls in. It is *not* an offset relative
3204  * 			to that object's base address. Accordingly, it must be
3205  * 			adjusted by adding (sh_addr - sh_offset), where
3206  * 			sh_{addr,offset} correspond to the executable section
3207  * 			containing *file_offset* in the object, for comparisons
3208  * 			to symbols' st_value to be valid.
3209  *
3210  * 		**bpf_get_stack**\ () can collect up to
3211  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3212  * 		to sufficient large buffer size. Note that
3213  * 		this limit can be controlled with the **sysctl** program, and
3214  * 		that it should be manually increased in order to profile long
3215  * 		user stacks (such as stacks for Java programs). To do so, use:
3216  *
3217  * 		::
3218  *
3219  * 			# sysctl kernel.perf_event_max_stack=<new value>
3220  * 	Return
3221  * 		The non-negative copied *buf* length equal to or less than
3222  * 		*size* on success, or a negative error in case of failure.
3223  *
3224  * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3225  * 	Description
3226  * 		This helper is similar to **bpf_skb_load_bytes**\ () in that
3227  * 		it provides an easy way to load *len* bytes from *offset*
3228  * 		from the packet associated to *skb*, into the buffer pointed
3229  * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3230  * 		a fifth argument *start_header* exists in order to select a
3231  * 		base offset to start from. *start_header* can be one of:
3232  *
3233  * 		**BPF_HDR_START_MAC**
3234  * 			Base offset to load data from is *skb*'s mac header.
3235  * 		**BPF_HDR_START_NET**
3236  * 			Base offset to load data from is *skb*'s network header.
3237  *
3238  * 		In general, "direct packet access" is the preferred method to
3239  * 		access packet data, however, this helper is in particular useful
3240  * 		in socket filters where *skb*\ **->data** does not always point
3241  * 		to the start of the mac header and where "direct packet access"
3242  * 		is not available.
3243  * 	Return
3244  * 		0 on success, or a negative error in case of failure.
3245  *
3246  * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3247  *	Description
3248  *		Do FIB lookup in kernel tables using parameters in *params*.
3249  *		If lookup is successful and result shows packet is to be
3250  *		forwarded, the neighbor tables are searched for the nexthop.
3251  *		If successful (ie., FIB lookup shows forwarding and nexthop
3252  *		is resolved), the nexthop address is returned in ipv4_dst
3253  *		or ipv6_dst based on family, smac is set to mac address of
3254  *		egress device, dmac is set to nexthop mac address, rt_metric
3255  *		is set to metric from route (IPv4/IPv6 only), and ifindex
3256  *		is set to the device index of the nexthop from the FIB lookup.
3257  *
3258  *		*plen* argument is the size of the passed in struct.
3259  *		*flags* argument can be a combination of one or more of the
3260  *		following values:
3261  *
3262  *		**BPF_FIB_LOOKUP_DIRECT**
3263  *			Do a direct table lookup vs full lookup using FIB
3264  *			rules.
3265  *		**BPF_FIB_LOOKUP_TBID**
3266  *			Used with BPF_FIB_LOOKUP_DIRECT.
3267  *			Use the routing table ID present in *params*->tbid
3268  *			for the fib lookup.
3269  *		**BPF_FIB_LOOKUP_OUTPUT**
3270  *			Perform lookup from an egress perspective (default is
3271  *			ingress).
3272  *		**BPF_FIB_LOOKUP_SKIP_NEIGH**
3273  *			Skip the neighbour table lookup. *params*->dmac
3274  *			and *params*->smac will not be set as output. A common
3275  *			use case is to call **bpf_redirect_neigh**\ () after
3276  *			doing **bpf_fib_lookup**\ ().
3277  *		**BPF_FIB_LOOKUP_SRC**
3278  *			Derive and set source IP addr in *params*->ipv{4,6}_src
3279  *			for the nexthop. If the src addr cannot be derived,
3280  *			**BPF_FIB_LKUP_RET_NO_SRC_ADDR** is returned. In this
3281  *			case, *params*->dmac and *params*->smac are not set either.
3282  *
3283  *		*ctx* is either **struct xdp_md** for XDP programs or
3284  *		**struct sk_buff** tc cls_act programs.
3285  *	Return
3286  *		* < 0 if any input argument is invalid
3287  *		*   0 on success (packet is forwarded, nexthop neighbor exists)
3288  *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3289  *		  packet is not forwarded or needs assist from full stack
3290  *
3291  *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3292  *		was exceeded and output params->mtu_result contains the MTU.
3293  *
3294  * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3295  *	Description
3296  *		Add an entry to, or update a sockhash *map* referencing sockets.
3297  *		The *skops* is used as a new value for the entry associated to
3298  *		*key*. *flags* is one of:
3299  *
3300  *		**BPF_NOEXIST**
3301  *			The entry for *key* must not exist in the map.
3302  *		**BPF_EXIST**
3303  *			The entry for *key* must already exist in the map.
3304  *		**BPF_ANY**
3305  *			No condition on the existence of the entry for *key*.
3306  *
3307  *		If the *map* has eBPF programs (parser and verdict), those will
3308  *		be inherited by the socket being added. If the socket is
3309  *		already attached to eBPF programs, this results in an error.
3310  *	Return
3311  *		0 on success, or a negative error in case of failure.
3312  *
3313  * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3314  *	Description
3315  *		This helper is used in programs implementing policies at the
3316  *		socket level. If the message *msg* is allowed to pass (i.e. if
3317  *		the verdict eBPF program returns **SK_PASS**), redirect it to
3318  *		the socket referenced by *map* (of type
3319  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3320  *		egress interfaces can be used for redirection. The
3321  *		**BPF_F_INGRESS** value in *flags* is used to make the
3322  *		distinction (ingress path is selected if the flag is present,
3323  *		egress path otherwise). This is the only flag supported for now.
3324  *	Return
3325  *		**SK_PASS** on success, or **SK_DROP** on error.
3326  *
3327  * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3328  *	Description
3329  *		This helper is used in programs implementing policies at the
3330  *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3331  *		if the verdict eBPF program returns **SK_PASS**), redirect it
3332  *		to the socket referenced by *map* (of type
3333  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3334  *		egress interfaces can be used for redirection. The
3335  *		**BPF_F_INGRESS** value in *flags* is used to make the
3336  *		distinction (ingress path is selected if the flag is present,
3337  *		egress otherwise). This is the only flag supported for now.
3338  *	Return
3339  *		**SK_PASS** on success, or **SK_DROP** on error.
3340  *
3341  * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3342  *	Description
3343  *		Encapsulate the packet associated to *skb* within a Layer 3
3344  *		protocol header. This header is provided in the buffer at
3345  *		address *hdr*, with *len* its size in bytes. *type* indicates
3346  *		the protocol of the header and can be one of:
3347  *
3348  *		**BPF_LWT_ENCAP_SEG6**
3349  *			IPv6 encapsulation with Segment Routing Header
3350  *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3351  *			the IPv6 header is computed by the kernel.
3352  *		**BPF_LWT_ENCAP_SEG6_INLINE**
3353  *			Only works if *skb* contains an IPv6 packet. Insert a
3354  *			Segment Routing Header (**struct ipv6_sr_hdr**) inside
3355  *			the IPv6 header.
3356  *		**BPF_LWT_ENCAP_IP**
3357  *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3358  *			must be IPv4 or IPv6, followed by zero or more
3359  *			additional headers, up to **LWT_BPF_MAX_HEADROOM**
3360  *			total bytes in all prepended headers. Please note that
3361  *			if **skb_is_gso**\ (*skb*) is true, no more than two
3362  *			headers can be prepended, and the inner header, if
3363  *			present, should be either GRE or UDP/GUE.
3364  *
3365  *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3366  *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3367  *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3368  *		**BPF_PROG_TYPE_LWT_XMIT**.
3369  *
3370  * 		A call to this helper is susceptible to change the underlying
3371  * 		packet buffer. Therefore, at load time, all checks on pointers
3372  * 		previously done by the verifier are invalidated and must be
3373  * 		performed again, if the helper is used in combination with
3374  * 		direct packet access.
3375  *	Return
3376  * 		0 on success, or a negative error in case of failure.
3377  *
3378  * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3379  *	Description
3380  *		Store *len* bytes from address *from* into the packet
3381  *		associated to *skb*, at *offset*. Only the flags, tag and TLVs
3382  *		inside the outermost IPv6 Segment Routing Header can be
3383  *		modified through this helper.
3384  *
3385  * 		A call to this helper is susceptible to change the underlying
3386  * 		packet buffer. Therefore, at load time, all checks on pointers
3387  * 		previously done by the verifier are invalidated and must be
3388  * 		performed again, if the helper is used in combination with
3389  * 		direct packet access.
3390  *	Return
3391  * 		0 on success, or a negative error in case of failure.
3392  *
3393  * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3394  *	Description
3395  *		Adjust the size allocated to TLVs in the outermost IPv6
3396  *		Segment Routing Header contained in the packet associated to
3397  *		*skb*, at position *offset* by *delta* bytes. Only offsets
3398  *		after the segments are accepted. *delta* can be as well
3399  *		positive (growing) as negative (shrinking).
3400  *
3401  * 		A call to this helper is susceptible to change the underlying
3402  * 		packet buffer. Therefore, at load time, all checks on pointers
3403  * 		previously done by the verifier are invalidated and must be
3404  * 		performed again, if the helper is used in combination with
3405  * 		direct packet access.
3406  *	Return
3407  * 		0 on success, or a negative error in case of failure.
3408  *
3409  * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3410  *	Description
3411  *		Apply an IPv6 Segment Routing action of type *action* to the
3412  *		packet associated to *skb*. Each action takes a parameter
3413  *		contained at address *param*, and of length *param_len* bytes.
3414  *		*action* can be one of:
3415  *
3416  *		**SEG6_LOCAL_ACTION_END_X**
3417  *			End.X action: Endpoint with Layer-3 cross-connect.
3418  *			Type of *param*: **struct in6_addr**.
3419  *		**SEG6_LOCAL_ACTION_END_T**
3420  *			End.T action: Endpoint with specific IPv6 table lookup.
3421  *			Type of *param*: **int**.
3422  *		**SEG6_LOCAL_ACTION_END_B6**
3423  *			End.B6 action: Endpoint bound to an SRv6 policy.
3424  *			Type of *param*: **struct ipv6_sr_hdr**.
3425  *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**
3426  *			End.B6.Encap action: Endpoint bound to an SRv6
3427  *			encapsulation policy.
3428  *			Type of *param*: **struct ipv6_sr_hdr**.
3429  *
3430  * 		A call to this helper is susceptible to change the underlying
3431  * 		packet buffer. Therefore, at load time, all checks on pointers
3432  * 		previously done by the verifier are invalidated and must be
3433  * 		performed again, if the helper is used in combination with
3434  * 		direct packet access.
3435  *	Return
3436  * 		0 on success, or a negative error in case of failure.
3437  *
3438  * long bpf_rc_repeat(void *ctx)
3439  *	Description
3440  *		This helper is used in programs implementing IR decoding, to
3441  *		report a successfully decoded repeat key message. This delays
3442  *		the generation of a key up event for previously generated
3443  *		key down event.
3444  *
3445  *		Some IR protocols like NEC have a special IR message for
3446  *		repeating last button, for when a button is held down.
3447  *
3448  *		The *ctx* should point to the lirc sample as passed into
3449  *		the program.
3450  *
3451  *		This helper is only available is the kernel was compiled with
3452  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3453  *		"**y**".
3454  *	Return
3455  *		0
3456  *
3457  * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3458  *	Description
3459  *		This helper is used in programs implementing IR decoding, to
3460  *		report a successfully decoded key press with *scancode*,
3461  *		*toggle* value in the given *protocol*. The scancode will be
3462  *		translated to a keycode using the rc keymap, and reported as
3463  *		an input key down event. After a period a key up event is
3464  *		generated. This period can be extended by calling either
3465  *		**bpf_rc_keydown**\ () again with the same values, or calling
3466  *		**bpf_rc_repeat**\ ().
3467  *
3468  *		Some protocols include a toggle bit, in case the button was
3469  *		released and pressed again between consecutive scancodes.
3470  *
3471  *		The *ctx* should point to the lirc sample as passed into
3472  *		the program.
3473  *
3474  *		The *protocol* is the decoded protocol number (see
3475  *		**enum rc_proto** for some predefined values).
3476  *
3477  *		This helper is only available is the kernel was compiled with
3478  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3479  *		"**y**".
3480  *	Return
3481  *		0
3482  *
3483  * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3484  * 	Description
3485  * 		Return the cgroup v2 id of the socket associated with the *skb*.
3486  * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3487  * 		helper for cgroup v1 by providing a tag resp. identifier that
3488  * 		can be matched on or used for map lookups e.g. to implement
3489  * 		policy. The cgroup v2 id of a given path in the hierarchy is
3490  * 		exposed in user space through the f_handle API in order to get
3491  * 		to the same 64-bit id.
3492  *
3493  * 		This helper can be used on TC egress path, but not on ingress,
3494  * 		and is available only if the kernel was compiled with the
3495  * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.
3496  * 	Return
3497  * 		The id is returned or 0 in case the id could not be retrieved.
3498  *
3499  * u64 bpf_get_current_cgroup_id(void)
3500  * 	Description
3501  * 		Get the current cgroup id based on the cgroup within which
3502  * 		the current task is running.
3503  * 	Return
3504  * 		A 64-bit integer containing the current cgroup id based
3505  * 		on the cgroup within which the current task is running.
3506  *
3507  * void *bpf_get_local_storage(void *map, u64 flags)
3508  *	Description
3509  *		Get the pointer to the local storage area.
3510  *		The type and the size of the local storage is defined
3511  *		by the *map* argument.
3512  *		The *flags* meaning is specific for each map type,
3513  *		and has to be 0 for cgroup local storage.
3514  *
3515  *		Depending on the BPF program type, a local storage area
3516  *		can be shared between multiple instances of the BPF program,
3517  *		running simultaneously.
3518  *
3519  *		A user should care about the synchronization by himself.
3520  *		For example, by using the **BPF_ATOMIC** instructions to alter
3521  *		the shared data.
3522  *	Return
3523  *		A pointer to the local storage area.
3524  *
3525  * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3526  *	Description
3527  *		Select a **SO_REUSEPORT** socket from a
3528  *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3529  *		It checks the selected socket is matching the incoming
3530  *		request in the socket buffer.
3531  *	Return
3532  *		0 on success, or a negative error in case of failure.
3533  *
3534  * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3535  *	Description
3536  *		Return id of cgroup v2 that is ancestor of cgroup associated
3537  *		with the *skb* at the *ancestor_level*.  The root cgroup is at
3538  *		*ancestor_level* zero and each step down the hierarchy
3539  *		increments the level. If *ancestor_level* == level of cgroup
3540  *		associated with *skb*, then return value will be same as that
3541  *		of **bpf_skb_cgroup_id**\ ().
3542  *
3543  *		The helper is useful to implement policies based on cgroups
3544  *		that are upper in hierarchy than immediate cgroup associated
3545  *		with *skb*.
3546  *
3547  *		The format of returned id and helper limitations are same as in
3548  *		**bpf_skb_cgroup_id**\ ().
3549  *	Return
3550  *		The id is returned or 0 in case the id could not be retrieved.
3551  *
3552  * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3553  *	Description
3554  *		Look for TCP socket matching *tuple*, optionally in a child
3555  *		network namespace *netns*. The return value must be checked,
3556  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3557  *
3558  *		The *ctx* should point to the context of the program, such as
3559  *		the skb or socket (depending on the hook in use). This is used
3560  *		to determine the base network namespace for the lookup.
3561  *
3562  *		*tuple_size* must be one of:
3563  *
3564  *		**sizeof**\ (*tuple*\ **->ipv4**)
3565  *			Look for an IPv4 socket.
3566  *		**sizeof**\ (*tuple*\ **->ipv6**)
3567  *			Look for an IPv6 socket.
3568  *
3569  *		If the *netns* is a negative signed 32-bit integer, then the
3570  *		socket lookup table in the netns associated with the *ctx*
3571  *		will be used. For the TC hooks, this is the netns of the device
3572  *		in the skb. For socket hooks, this is the netns of the socket.
3573  *		If *netns* is any other signed 32-bit value greater than or
3574  *		equal to zero then it specifies the ID of the netns relative to
3575  *		the netns associated with the *ctx*. *netns* values beyond the
3576  *		range of 32-bit integers are reserved for future use.
3577  *
3578  *		All values for *flags* are reserved for future usage, and must
3579  *		be left at zero.
3580  *
3581  *		This helper is available only if the kernel was compiled with
3582  *		**CONFIG_NET** configuration option.
3583  *	Return
3584  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3585  *		For sockets with reuseport option, the **struct bpf_sock**
3586  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3587  *		tuple.
3588  *
3589  * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3590  *	Description
3591  *		Look for UDP socket matching *tuple*, optionally in a child
3592  *		network namespace *netns*. The return value must be checked,
3593  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3594  *
3595  *		The *ctx* should point to the context of the program, such as
3596  *		the skb or socket (depending on the hook in use). This is used
3597  *		to determine the base network namespace for the lookup.
3598  *
3599  *		*tuple_size* must be one of:
3600  *
3601  *		**sizeof**\ (*tuple*\ **->ipv4**)
3602  *			Look for an IPv4 socket.
3603  *		**sizeof**\ (*tuple*\ **->ipv6**)
3604  *			Look for an IPv6 socket.
3605  *
3606  *		If the *netns* is a negative signed 32-bit integer, then the
3607  *		socket lookup table in the netns associated with the *ctx*
3608  *		will be used. For the TC hooks, this is the netns of the device
3609  *		in the skb. For socket hooks, this is the netns of the socket.
3610  *		If *netns* is any other signed 32-bit value greater than or
3611  *		equal to zero then it specifies the ID of the netns relative to
3612  *		the netns associated with the *ctx*. *netns* values beyond the
3613  *		range of 32-bit integers are reserved for future use.
3614  *
3615  *		All values for *flags* are reserved for future usage, and must
3616  *		be left at zero.
3617  *
3618  *		This helper is available only if the kernel was compiled with
3619  *		**CONFIG_NET** configuration option.
3620  *	Return
3621  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3622  *		For sockets with reuseport option, the **struct bpf_sock**
3623  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3624  *		tuple.
3625  *
3626  * long bpf_sk_release(void *sock)
3627  *	Description
3628  *		Release the reference held by *sock*. *sock* must be a
3629  *		non-**NULL** pointer that was returned from
3630  *		**bpf_sk_lookup_xxx**\ ().
3631  *	Return
3632  *		0 on success, or a negative error in case of failure.
3633  *
3634  * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3635  * 	Description
3636  * 		Push an element *value* in *map*. *flags* is one of:
3637  *
3638  * 		**BPF_EXIST**
3639  * 			If the queue/stack is full, the oldest element is
3640  * 			removed to make room for this.
3641  * 	Return
3642  * 		0 on success, or a negative error in case of failure.
3643  *
3644  * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3645  * 	Description
3646  * 		Pop an element from *map*.
3647  * 	Return
3648  * 		0 on success, or a negative error in case of failure.
3649  *
3650  * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3651  * 	Description
3652  * 		Get an element from *map* without removing it.
3653  * 	Return
3654  * 		0 on success, or a negative error in case of failure.
3655  *
3656  * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3657  *	Description
3658  *		For socket policies, insert *len* bytes into *msg* at offset
3659  *		*start*.
3660  *
3661  *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3662  *		*msg* it may want to insert metadata or options into the *msg*.
3663  *		This can later be read and used by any of the lower layer BPF
3664  *		hooks.
3665  *
3666  *		This helper may fail if under memory pressure (a malloc
3667  *		fails) in these cases BPF programs will get an appropriate
3668  *		error and BPF programs will need to handle them.
3669  *	Return
3670  *		0 on success, or a negative error in case of failure.
3671  *
3672  * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3673  *	Description
3674  *		Will remove *len* bytes from a *msg* starting at byte *start*.
3675  *		This may result in **ENOMEM** errors under certain situations if
3676  *		an allocation and copy are required due to a full ring buffer.
3677  *		However, the helper will try to avoid doing the allocation
3678  *		if possible. Other errors can occur if input parameters are
3679  *		invalid either due to *start* byte not being valid part of *msg*
3680  *		payload and/or *pop* value being to large.
3681  *	Return
3682  *		0 on success, or a negative error in case of failure.
3683  *
3684  * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3685  *	Description
3686  *		This helper is used in programs implementing IR decoding, to
3687  *		report a successfully decoded pointer movement.
3688  *
3689  *		The *ctx* should point to the lirc sample as passed into
3690  *		the program.
3691  *
3692  *		This helper is only available is the kernel was compiled with
3693  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3694  *		"**y**".
3695  *	Return
3696  *		0
3697  *
3698  * long bpf_spin_lock(struct bpf_spin_lock *lock)
3699  *	Description
3700  *		Acquire a spinlock represented by the pointer *lock*, which is
3701  *		stored as part of a value of a map. Taking the lock allows to
3702  *		safely update the rest of the fields in that value. The
3703  *		spinlock can (and must) later be released with a call to
3704  *		**bpf_spin_unlock**\ (\ *lock*\ ).
3705  *
3706  *		Spinlocks in BPF programs come with a number of restrictions
3707  *		and constraints:
3708  *
3709  *		* **bpf_spin_lock** objects are only allowed inside maps of
3710  *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3711  *		  list could be extended in the future).
3712  *		* BTF description of the map is mandatory.
3713  *		* The BPF program can take ONE lock at a time, since taking two
3714  *		  or more could cause dead locks.
3715  *		* Only one **struct bpf_spin_lock** is allowed per map element.
3716  *		* When the lock is taken, calls (either BPF to BPF or helpers)
3717  *		  are not allowed.
3718  *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3719  *		  allowed inside a spinlock-ed region.
3720  *		* The BPF program MUST call **bpf_spin_unlock**\ () to release
3721  *		  the lock, on all execution paths, before it returns.
3722  *		* The BPF program can access **struct bpf_spin_lock** only via
3723  *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3724  *		  helpers. Loading or storing data into the **struct
3725  *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3726  *		* To use the **bpf_spin_lock**\ () helper, the BTF description
3727  *		  of the map value must be a struct and have **struct
3728  *		  bpf_spin_lock** *anyname*\ **;** field at the top level.
3729  *		  Nested lock inside another struct is not allowed.
3730  *		* The **struct bpf_spin_lock** *lock* field in a map value must
3731  *		  be aligned on a multiple of 4 bytes in that value.
3732  *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3733  *		  the **bpf_spin_lock** field to user space.
3734  *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3735  *		  a BPF program, do not update the **bpf_spin_lock** field.
3736  *		* **bpf_spin_lock** cannot be on the stack or inside a
3737  *		  networking packet (it can only be inside of a map values).
3738  *		* **bpf_spin_lock** is available to root only.
3739  *		* Tracing programs and socket filter programs cannot use
3740  *		  **bpf_spin_lock**\ () due to insufficient preemption checks
3741  *		  (but this may change in the future).
3742  *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3743  *	Return
3744  *		0
3745  *
3746  * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3747  *	Description
3748  *		Release the *lock* previously locked by a call to
3749  *		**bpf_spin_lock**\ (\ *lock*\ ).
3750  *	Return
3751  *		0
3752  *
3753  * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3754  *	Description
3755  *		This helper gets a **struct bpf_sock** pointer such
3756  *		that all the fields in this **bpf_sock** can be accessed.
3757  *	Return
3758  *		A **struct bpf_sock** pointer on success, or **NULL** in
3759  *		case of failure.
3760  *
3761  * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3762  *	Description
3763  *		This helper gets a **struct bpf_tcp_sock** pointer from a
3764  *		**struct bpf_sock** pointer.
3765  *	Return
3766  *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3767  *		case of failure.
3768  *
3769  * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3770  *	Description
3771  *		Set ECN (Explicit Congestion Notification) field of IP header
3772  *		to **CE** (Congestion Encountered) if current value is **ECT**
3773  *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3774  *		and IPv4.
3775  *	Return
3776  *		1 if the **CE** flag is set (either by the current helper call
3777  *		or because it was already present), 0 if it is not set.
3778  *
3779  * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3780  *	Description
3781  *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3782  *		**bpf_sk_release**\ () is unnecessary and not allowed.
3783  *	Return
3784  *		A **struct bpf_sock** pointer on success, or **NULL** in
3785  *		case of failure.
3786  *
3787  * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3788  *	Description
3789  *		Look for TCP socket matching *tuple*, optionally in a child
3790  *		network namespace *netns*. The return value must be checked,
3791  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3792  *
3793  *		This function is identical to **bpf_sk_lookup_tcp**\ (), except
3794  *		that it also returns timewait or request sockets. Use
3795  *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3796  *		full structure.
3797  *
3798  *		This helper is available only if the kernel was compiled with
3799  *		**CONFIG_NET** configuration option.
3800  *	Return
3801  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3802  *		For sockets with reuseport option, the **struct bpf_sock**
3803  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3804  *		tuple.
3805  *
3806  * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3807  * 	Description
3808  * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for
3809  * 		the listening socket in *sk*.
3810  *
3811  * 		*iph* points to the start of the IPv4 or IPv6 header, while
3812  * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3813  * 		**sizeof**\ (**struct ipv6hdr**).
3814  *
3815  * 		*th* points to the start of the TCP header, while *th_len*
3816  *		contains the length of the TCP header (at least
3817  *		**sizeof**\ (**struct tcphdr**)).
3818  * 	Return
3819  * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3820  * 		error otherwise.
3821  *
3822  * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3823  *	Description
3824  *		Get name of sysctl in /proc/sys/ and copy it into provided by
3825  *		program buffer *buf* of size *buf_len*.
3826  *
3827  *		The buffer is always NUL terminated, unless it's zero-sized.
3828  *
3829  *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3830  *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3831  *		only (e.g. "tcp_mem").
3832  *	Return
3833  *		Number of character copied (not including the trailing NUL).
3834  *
3835  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3836  *		truncated name in this case).
3837  *
3838  * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3839  *	Description
3840  *		Get current value of sysctl as it is presented in /proc/sys
3841  *		(incl. newline, etc), and copy it as a string into provided
3842  *		by program buffer *buf* of size *buf_len*.
3843  *
3844  *		The whole value is copied, no matter what file position user
3845  *		space issued e.g. sys_read at.
3846  *
3847  *		The buffer is always NUL terminated, unless it's zero-sized.
3848  *	Return
3849  *		Number of character copied (not including the trailing NUL).
3850  *
3851  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3852  *		truncated name in this case).
3853  *
3854  *		**-EINVAL** if current value was unavailable, e.g. because
3855  *		sysctl is uninitialized and read returns -EIO for it.
3856  *
3857  * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3858  *	Description
3859  *		Get new value being written by user space to sysctl (before
3860  *		the actual write happens) and copy it as a string into
3861  *		provided by program buffer *buf* of size *buf_len*.
3862  *
3863  *		User space may write new value at file position > 0.
3864  *
3865  *		The buffer is always NUL terminated, unless it's zero-sized.
3866  *	Return
3867  *		Number of character copied (not including the trailing NUL).
3868  *
3869  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3870  *		truncated name in this case).
3871  *
3872  *		**-EINVAL** if sysctl is being read.
3873  *
3874  * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3875  *	Description
3876  *		Override new value being written by user space to sysctl with
3877  *		value provided by program in buffer *buf* of size *buf_len*.
3878  *
3879  *		*buf* should contain a string in same form as provided by user
3880  *		space on sysctl write.
3881  *
3882  *		User space may write new value at file position > 0. To override
3883  *		the whole sysctl value file position should be set to zero.
3884  *	Return
3885  *		0 on success.
3886  *
3887  *		**-E2BIG** if the *buf_len* is too big.
3888  *
3889  *		**-EINVAL** if sysctl is being read.
3890  *
3891  * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3892  *	Description
3893  *		Convert the initial part of the string from buffer *buf* of
3894  *		size *buf_len* to a long integer according to the given base
3895  *		and save the result in *res*.
3896  *
3897  *		The string may begin with an arbitrary amount of white space
3898  *		(as determined by **isspace**\ (3)) followed by a single
3899  *		optional '**-**' sign.
3900  *
3901  *		Five least significant bits of *flags* encode base, other bits
3902  *		are currently unused.
3903  *
3904  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3905  *		similar to user space **strtol**\ (3).
3906  *	Return
3907  *		Number of characters consumed on success. Must be positive but
3908  *		no more than *buf_len*.
3909  *
3910  *		**-EINVAL** if no valid digits were found or unsupported base
3911  *		was provided.
3912  *
3913  *		**-ERANGE** if resulting value was out of range.
3914  *
3915  * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3916  *	Description
3917  *		Convert the initial part of the string from buffer *buf* of
3918  *		size *buf_len* to an unsigned long integer according to the
3919  *		given base and save the result in *res*.
3920  *
3921  *		The string may begin with an arbitrary amount of white space
3922  *		(as determined by **isspace**\ (3)).
3923  *
3924  *		Five least significant bits of *flags* encode base, other bits
3925  *		are currently unused.
3926  *
3927  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3928  *		similar to user space **strtoul**\ (3).
3929  *	Return
3930  *		Number of characters consumed on success. Must be positive but
3931  *		no more than *buf_len*.
3932  *
3933  *		**-EINVAL** if no valid digits were found or unsupported base
3934  *		was provided.
3935  *
3936  *		**-ERANGE** if resulting value was out of range.
3937  *
3938  * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3939  *	Description
3940  *		Get a bpf-local-storage from a *sk*.
3941  *
3942  *		Logically, it could be thought of getting the value from
3943  *		a *map* with *sk* as the **key**.  From this
3944  *		perspective,  the usage is not much different from
3945  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3946  *		helper enforces the key must be a full socket and the map must
3947  *		be a **BPF_MAP_TYPE_SK_STORAGE** also.
3948  *
3949  *		Underneath, the value is stored locally at *sk* instead of
3950  *		the *map*.  The *map* is used as the bpf-local-storage
3951  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
3952  *		searched against all bpf-local-storages residing at *sk*.
3953  *
3954  *		*sk* is a kernel **struct sock** pointer for LSM program.
3955  *		*sk* is a **struct bpf_sock** pointer for other program types.
3956  *
3957  *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3958  *		used such that a new bpf-local-storage will be
3959  *		created if one does not exist.  *value* can be used
3960  *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3961  *		the initial value of a bpf-local-storage.  If *value* is
3962  *		**NULL**, the new bpf-local-storage will be zero initialized.
3963  *	Return
3964  *		A bpf-local-storage pointer is returned on success.
3965  *
3966  *		**NULL** if not found or there was an error in adding
3967  *		a new bpf-local-storage.
3968  *
3969  * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3970  *	Description
3971  *		Delete a bpf-local-storage from a *sk*.
3972  *	Return
3973  *		0 on success.
3974  *
3975  *		**-ENOENT** if the bpf-local-storage cannot be found.
3976  *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3977  *
3978  * long bpf_send_signal(u32 sig)
3979  *	Description
3980  *		Send signal *sig* to the process of the current task.
3981  *		The signal may be delivered to any of this process's threads.
3982  *	Return
3983  *		0 on success or successfully queued.
3984  *
3985  *		**-EBUSY** if work queue under nmi is full.
3986  *
3987  *		**-EINVAL** if *sig* is invalid.
3988  *
3989  *		**-EPERM** if no permission to send the *sig*.
3990  *
3991  *		**-EAGAIN** if bpf program can try again.
3992  *
3993  * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3994  *	Description
3995  *		Try to issue a SYN cookie for the packet with corresponding
3996  *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3997  *
3998  *		*iph* points to the start of the IPv4 or IPv6 header, while
3999  *		*iph_len* contains **sizeof**\ (**struct iphdr**) or
4000  *		**sizeof**\ (**struct ipv6hdr**).
4001  *
4002  *		*th* points to the start of the TCP header, while *th_len*
4003  *		contains the length of the TCP header with options (at least
4004  *		**sizeof**\ (**struct tcphdr**)).
4005  *	Return
4006  *		On success, lower 32 bits hold the generated SYN cookie in
4007  *		followed by 16 bits which hold the MSS value for that cookie,
4008  *		and the top 16 bits are unused.
4009  *
4010  *		On failure, the returned value is one of the following:
4011  *
4012  *		**-EINVAL** SYN cookie cannot be issued due to error
4013  *
4014  *		**-ENOENT** SYN cookie should not be issued (no SYN flood)
4015  *
4016  *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies
4017  *
4018  *		**-EPROTONOSUPPORT** IP packet version is not 4 or 6
4019  *
4020  * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4021  * 	Description
4022  * 		Write raw *data* blob into a special BPF perf event held by
4023  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4024  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
4025  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4026  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4027  *
4028  * 		The *flags* are used to indicate the index in *map* for which
4029  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
4030  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4031  * 		to indicate that the index of the current CPU core should be
4032  * 		used.
4033  *
4034  * 		The value to write, of *size*, is passed through eBPF stack and
4035  * 		pointed by *data*.
4036  *
4037  * 		*ctx* is a pointer to in-kernel struct sk_buff.
4038  *
4039  * 		This helper is similar to **bpf_perf_event_output**\ () but
4040  * 		restricted to raw_tracepoint bpf programs.
4041  * 	Return
4042  * 		0 on success, or a negative error in case of failure.
4043  *
4044  * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
4045  * 	Description
4046  * 		Safely attempt to read *size* bytes from user space address
4047  * 		*unsafe_ptr* and store the data in *dst*.
4048  * 	Return
4049  * 		0 on success, or a negative error in case of failure.
4050  *
4051  * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
4052  * 	Description
4053  * 		Safely attempt to read *size* bytes from kernel space address
4054  * 		*unsafe_ptr* and store the data in *dst*.
4055  * 	Return
4056  * 		0 on success, or a negative error in case of failure.
4057  *
4058  * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
4059  * 	Description
4060  * 		Copy a NUL terminated string from an unsafe user address
4061  * 		*unsafe_ptr* to *dst*. The *size* should include the
4062  * 		terminating NUL byte. In case the string length is smaller than
4063  * 		*size*, the target is not padded with further NUL bytes. If the
4064  * 		string length is larger than *size*, just *size*-1 bytes are
4065  * 		copied and the last byte is set to NUL.
4066  *
4067  * 		On success, returns the number of bytes that were written,
4068  * 		including the terminal NUL. This makes this helper useful in
4069  * 		tracing programs for reading strings, and more importantly to
4070  * 		get its length at runtime. See the following snippet:
4071  *
4072  * 		::
4073  *
4074  * 			SEC("kprobe/sys_open")
4075  * 			void bpf_sys_open(struct pt_regs *ctx)
4076  * 			{
4077  * 			        char buf[PATHLEN]; // PATHLEN is defined to 256
4078  * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),
4079  * 				                                  ctx->di);
4080  *
4081  * 				// Consume buf, for example push it to
4082  * 				// userspace via bpf_perf_event_output(); we
4083  * 				// can use res (the string length) as event
4084  * 				// size, after checking its boundaries.
4085  * 			}
4086  *
4087  * 		In comparison, using **bpf_probe_read_user**\ () helper here
4088  * 		instead to read the string would require to estimate the length
4089  * 		at compile time, and would often result in copying more memory
4090  * 		than necessary.
4091  *
4092  * 		Another useful use case is when parsing individual process
4093  * 		arguments or individual environment variables navigating
4094  * 		*current*\ **->mm->arg_start** and *current*\
4095  * 		**->mm->env_start**: using this helper and the return value,
4096  * 		one can quickly iterate at the right offset of the memory area.
4097  * 	Return
4098  * 		On success, the strictly positive length of the output string,
4099  * 		including the trailing NUL character. On error, a negative
4100  * 		value.
4101  *
4102  * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
4103  * 	Description
4104  * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
4105  * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
4106  * 	Return
4107  * 		On success, the strictly positive length of the string, including
4108  * 		the trailing NUL character. On error, a negative value.
4109  *
4110  * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
4111  *	Description
4112  *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
4113  *		*rcv_nxt* is the ack_seq to be sent out.
4114  *	Return
4115  *		0 on success, or a negative error in case of failure.
4116  *
4117  * long bpf_send_signal_thread(u32 sig)
4118  *	Description
4119  *		Send signal *sig* to the thread corresponding to the current task.
4120  *	Return
4121  *		0 on success or successfully queued.
4122  *
4123  *		**-EBUSY** if work queue under nmi is full.
4124  *
4125  *		**-EINVAL** if *sig* is invalid.
4126  *
4127  *		**-EPERM** if no permission to send the *sig*.
4128  *
4129  *		**-EAGAIN** if bpf program can try again.
4130  *
4131  * u64 bpf_jiffies64(void)
4132  *	Description
4133  *		Obtain the 64bit jiffies
4134  *	Return
4135  *		The 64 bit jiffies
4136  *
4137  * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
4138  *	Description
4139  *		For an eBPF program attached to a perf event, retrieve the
4140  *		branch records (**struct perf_branch_entry**) associated to *ctx*
4141  *		and store it in the buffer pointed by *buf* up to size
4142  *		*size* bytes.
4143  *	Return
4144  *		On success, number of bytes written to *buf*. On error, a
4145  *		negative value.
4146  *
4147  *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
4148  *		instead return the number of bytes required to store all the
4149  *		branch entries. If this flag is set, *buf* may be NULL.
4150  *
4151  *		**-EINVAL** if arguments invalid or **size** not a multiple
4152  *		of **sizeof**\ (**struct perf_branch_entry**\ ).
4153  *
4154  *		**-ENOENT** if architecture does not support branch records.
4155  *
4156  * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4157  *	Description
4158  *		Returns 0 on success, values for *pid* and *tgid* as seen from the current
4159  *		*namespace* will be returned in *nsdata*.
4160  *	Return
4161  *		0 on success, or one of the following in case of failure:
4162  *
4163  *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number
4164  *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
4165  *
4166  *		**-ENOENT** if pidns does not exists for the current task.
4167  *
4168  * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4169  *	Description
4170  *		Write raw *data* blob into a special BPF perf event held by
4171  *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4172  *		event must have the following attributes: **PERF_SAMPLE_RAW**
4173  *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4174  *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4175  *
4176  *		The *flags* are used to indicate the index in *map* for which
4177  *		the value must be put, masked with **BPF_F_INDEX_MASK**.
4178  *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4179  *		to indicate that the index of the current CPU core should be
4180  *		used.
4181  *
4182  *		The value to write, of *size*, is passed through eBPF stack and
4183  *		pointed by *data*.
4184  *
4185  *		*ctx* is a pointer to in-kernel struct xdp_buff.
4186  *
4187  *		This helper is similar to **bpf_perf_eventoutput**\ () but
4188  *		restricted to raw_tracepoint bpf programs.
4189  *	Return
4190  *		0 on success, or a negative error in case of failure.
4191  *
4192  * u64 bpf_get_netns_cookie(void *ctx)
4193  * 	Description
4194  * 		Retrieve the cookie (generated by the kernel) of the network
4195  * 		namespace the input *ctx* is associated with. The network
4196  * 		namespace cookie remains stable for its lifetime and provides
4197  * 		a global identifier that can be assumed unique. If *ctx* is
4198  * 		NULL, then the helper returns the cookie for the initial
4199  * 		network namespace. The cookie itself is very similar to that
4200  * 		of **bpf_get_socket_cookie**\ () helper, but for network
4201  * 		namespaces instead of sockets.
4202  * 	Return
4203  * 		A 8-byte long opaque number.
4204  *
4205  * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4206  * 	Description
4207  * 		Return id of cgroup v2 that is ancestor of the cgroup associated
4208  * 		with the current task at the *ancestor_level*. The root cgroup
4209  * 		is at *ancestor_level* zero and each step down the hierarchy
4210  * 		increments the level. If *ancestor_level* == level of cgroup
4211  * 		associated with the current task, then return value will be the
4212  * 		same as that of **bpf_get_current_cgroup_id**\ ().
4213  *
4214  * 		The helper is useful to implement policies based on cgroups
4215  * 		that are upper in hierarchy than immediate cgroup associated
4216  * 		with the current task.
4217  *
4218  * 		The format of returned id and helper limitations are same as in
4219  * 		**bpf_get_current_cgroup_id**\ ().
4220  * 	Return
4221  * 		The id is returned or 0 in case the id could not be retrieved.
4222  *
4223  * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4224  *	Description
4225  *		Helper is overloaded depending on BPF program type. This
4226  *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4227  *		**BPF_PROG_TYPE_SCHED_ACT** programs.
4228  *
4229  *		Assign the *sk* to the *skb*. When combined with appropriate
4230  *		routing configuration to receive the packet towards the socket,
4231  *		will cause *skb* to be delivered to the specified socket.
4232  *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),
4233  *		**bpf_clone_redirect**\ () or other methods outside of BPF may
4234  *		interfere with successful delivery to the socket.
4235  *
4236  *		This operation is only valid from TC ingress path.
4237  *
4238  *		The *flags* argument must be zero.
4239  *	Return
4240  *		0 on success, or a negative error in case of failure:
4241  *
4242  *		**-EINVAL** if specified *flags* are not supported.
4243  *
4244  *		**-ENOENT** if the socket is unavailable for assignment.
4245  *
4246  *		**-ENETUNREACH** if the socket is unreachable (wrong netns).
4247  *
4248  *		**-EOPNOTSUPP** if the operation is not supported, for example
4249  *		a call from outside of TC ingress.
4250  *
4251  * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4252  *	Description
4253  *		Helper is overloaded depending on BPF program type. This
4254  *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4255  *
4256  *		Select the *sk* as a result of a socket lookup.
4257  *
4258  *		For the operation to succeed passed socket must be compatible
4259  *		with the packet description provided by the *ctx* object.
4260  *
4261  *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4262  *		be an exact match. While IP family (**AF_INET** or
4263  *		**AF_INET6**) must be compatible, that is IPv6 sockets
4264  *		that are not v6-only can be selected for IPv4 packets.
4265  *
4266  *		Only TCP listeners and UDP unconnected sockets can be
4267  *		selected. *sk* can also be NULL to reset any previous
4268  *		selection.
4269  *
4270  *		*flags* argument can combination of following values:
4271  *
4272  *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4273  *		  socket selection, potentially done by a BPF program
4274  *		  that ran before us.
4275  *
4276  *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4277  *		  load-balancing within reuseport group for the socket
4278  *		  being selected.
4279  *
4280  *		On success *ctx->sk* will point to the selected socket.
4281  *
4282  *	Return
4283  *		0 on success, or a negative errno in case of failure.
4284  *
4285  *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is
4286  *		  not compatible with packet family (*ctx->family*).
4287  *
4288  *		* **-EEXIST** if socket has been already selected,
4289  *		  potentially by another program, and
4290  *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4291  *
4292  *		* **-EINVAL** if unsupported flags were specified.
4293  *
4294  *		* **-EPROTOTYPE** if socket L4 protocol
4295  *		  (*sk->protocol*) doesn't match packet protocol
4296  *		  (*ctx->protocol*).
4297  *
4298  *		* **-ESOCKTNOSUPPORT** if socket is not in allowed
4299  *		  state (TCP listening or UDP unconnected).
4300  *
4301  * u64 bpf_ktime_get_boot_ns(void)
4302  * 	Description
4303  * 		Return the time elapsed since system boot, in nanoseconds.
4304  * 		Does include the time the system was suspended.
4305  * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4306  * 	Return
4307  * 		Current *ktime*.
4308  *
4309  * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4310  * 	Description
4311  * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4312  * 		out the format string.
4313  * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4314  * 		the format string itself. The *data* and *data_len* are format string
4315  * 		arguments. The *data* are a **u64** array and corresponding format string
4316  * 		values are stored in the array. For strings and pointers where pointees
4317  * 		are accessed, only the pointer values are stored in the *data* array.
4318  * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4319  *
4320  *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4321  *		Reading kernel memory may fail due to either invalid address or
4322  *		valid address but requiring a major memory fault. If reading kernel memory
4323  *		fails, the string for **%s** will be an empty string, and the ip
4324  *		address for **%p{i,I}{4,6}** will be 0. Not returning error to
4325  *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4326  * 	Return
4327  * 		0 on success, or a negative error in case of failure:
4328  *
4329  *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again
4330  *		by returning 1 from bpf program.
4331  *
4332  *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4333  *
4334  *		**-E2BIG** if *fmt* contains too many format specifiers.
4335  *
4336  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4337  *
4338  * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4339  * 	Description
4340  * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4341  * 		The *m* represents the seq_file. The *data* and *len* represent the
4342  * 		data to write in bytes.
4343  * 	Return
4344  * 		0 on success, or a negative error in case of failure:
4345  *
4346  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4347  *
4348  * u64 bpf_sk_cgroup_id(void *sk)
4349  *	Description
4350  *		Return the cgroup v2 id of the socket *sk*.
4351  *
4352  *		*sk* must be a non-**NULL** pointer to a socket, e.g. one
4353  *		returned from **bpf_sk_lookup_xxx**\ (),
4354  *		**bpf_sk_fullsock**\ (), etc. The format of returned id is
4355  *		same as in **bpf_skb_cgroup_id**\ ().
4356  *
4357  *		This helper is available only if the kernel was compiled with
4358  *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4359  *	Return
4360  *		The id is returned or 0 in case the id could not be retrieved.
4361  *
4362  * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4363  *	Description
4364  *		Return id of cgroup v2 that is ancestor of cgroup associated
4365  *		with the *sk* at the *ancestor_level*.  The root cgroup is at
4366  *		*ancestor_level* zero and each step down the hierarchy
4367  *		increments the level. If *ancestor_level* == level of cgroup
4368  *		associated with *sk*, then return value will be same as that
4369  *		of **bpf_sk_cgroup_id**\ ().
4370  *
4371  *		The helper is useful to implement policies based on cgroups
4372  *		that are upper in hierarchy than immediate cgroup associated
4373  *		with *sk*.
4374  *
4375  *		The format of returned id and helper limitations are same as in
4376  *		**bpf_sk_cgroup_id**\ ().
4377  *	Return
4378  *		The id is returned or 0 in case the id could not be retrieved.
4379  *
4380  * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4381  * 	Description
4382  * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4383  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4384  * 		of new data availability is sent.
4385  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4386  * 		of new data availability is sent unconditionally.
4387  * 		If **0** is specified in *flags*, an adaptive notification
4388  * 		of new data availability is sent.
4389  *
4390  * 		An adaptive notification is a notification sent whenever the user-space
4391  * 		process has caught up and consumed all available payloads. In case the user-space
4392  * 		process is still processing a previous payload, then no notification is needed
4393  * 		as it will process the newly added payload automatically.
4394  * 	Return
4395  * 		0 on success, or a negative error in case of failure.
4396  *
4397  * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4398  * 	Description
4399  * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4400  * 		*flags* must be 0.
4401  * 	Return
4402  * 		Valid pointer with *size* bytes of memory available; NULL,
4403  * 		otherwise.
4404  *
4405  * void bpf_ringbuf_submit(void *data, u64 flags)
4406  * 	Description
4407  * 		Submit reserved ring buffer sample, pointed to by *data*.
4408  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4409  * 		of new data availability is sent.
4410  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4411  * 		of new data availability is sent unconditionally.
4412  * 		If **0** is specified in *flags*, an adaptive notification
4413  * 		of new data availability is sent.
4414  *
4415  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4416  * 	Return
4417  * 		Nothing. Always succeeds.
4418  *
4419  * void bpf_ringbuf_discard(void *data, u64 flags)
4420  * 	Description
4421  * 		Discard reserved ring buffer sample, pointed to by *data*.
4422  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4423  * 		of new data availability is sent.
4424  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4425  * 		of new data availability is sent unconditionally.
4426  * 		If **0** is specified in *flags*, an adaptive notification
4427  * 		of new data availability is sent.
4428  *
4429  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4430  * 	Return
4431  * 		Nothing. Always succeeds.
4432  *
4433  * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4434  *	Description
4435  *		Query various characteristics of provided ring buffer. What
4436  *		exactly is queries is determined by *flags*:
4437  *
4438  *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4439  *		* **BPF_RB_RING_SIZE**: The size of ring buffer.
4440  *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4441  *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4442  *
4443  *		Data returned is just a momentary snapshot of actual values
4444  *		and could be inaccurate, so this facility should be used to
4445  *		power heuristics and for reporting, not to make 100% correct
4446  *		calculation.
4447  *	Return
4448  *		Requested value, or 0, if *flags* are not recognized.
4449  *
4450  * long bpf_csum_level(struct sk_buff *skb, u64 level)
4451  * 	Description
4452  * 		Change the skbs checksum level by one layer up or down, or
4453  * 		reset it entirely to none in order to have the stack perform
4454  * 		checksum validation. The level is applicable to the following
4455  * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4456  * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4457  * 		through **bpf_skb_adjust_room**\ () helper with passing in
4458  * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call
4459  * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4460  * 		the UDP header is removed. Similarly, an encap of the latter
4461  * 		into the former could be accompanied by a helper call to
4462  * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4463  * 		skb is still intended to be processed in higher layers of the
4464  * 		stack instead of just egressing at tc.
4465  *
4466  * 		There are three supported level settings at this time:
4467  *
4468  * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4469  * 		  with CHECKSUM_UNNECESSARY.
4470  * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4471  * 		  with CHECKSUM_UNNECESSARY.
4472  * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4473  * 		  sets CHECKSUM_NONE to force checksum validation by the stack.
4474  * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4475  * 		  skb->csum_level.
4476  * 	Return
4477  * 		0 on success, or a negative error in case of failure. In the
4478  * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4479  * 		is returned or the error code -EACCES in case the skb is not
4480  * 		subject to CHECKSUM_UNNECESSARY.
4481  *
4482  * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4483  *	Description
4484  *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4485  *	Return
4486  *		*sk* if casting is valid, or **NULL** otherwise.
4487  *
4488  * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4489  *	Description
4490  *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4491  *	Return
4492  *		*sk* if casting is valid, or **NULL** otherwise.
4493  *
4494  * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4495  * 	Description
4496  *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4497  *	Return
4498  *		*sk* if casting is valid, or **NULL** otherwise.
4499  *
4500  * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4501  * 	Description
4502  *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4503  *	Return
4504  *		*sk* if casting is valid, or **NULL** otherwise.
4505  *
4506  * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4507  * 	Description
4508  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4509  *	Return
4510  *		*sk* if casting is valid, or **NULL** otherwise.
4511  *
4512  * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4513  *	Description
4514  *		Return a user or a kernel stack in bpf program provided buffer.
4515  *		Note: the user stack will only be populated if the *task* is
4516  *		the current task; all other tasks will return -EOPNOTSUPP.
4517  *		To achieve this, the helper needs *task*, which is a valid
4518  *		pointer to **struct task_struct**. To store the stacktrace, the
4519  *		bpf program provides *buf* with a nonnegative *size*.
4520  *
4521  *		The last argument, *flags*, holds the number of stack frames to
4522  *		skip (from 0 to 255), masked with
4523  *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4524  *		the following flags:
4525  *
4526  *		**BPF_F_USER_STACK**
4527  *			Collect a user space stack instead of a kernel stack.
4528  *			The *task* must be the current task.
4529  *		**BPF_F_USER_BUILD_ID**
4530  *			Collect buildid+offset instead of ips for user stack,
4531  *			only valid if **BPF_F_USER_STACK** is also specified.
4532  *
4533  *		**bpf_get_task_stack**\ () can collect up to
4534  *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4535  *		to sufficient large buffer size. Note that
4536  *		this limit can be controlled with the **sysctl** program, and
4537  *		that it should be manually increased in order to profile long
4538  *		user stacks (such as stacks for Java programs). To do so, use:
4539  *
4540  *		::
4541  *
4542  *			# sysctl kernel.perf_event_max_stack=<new value>
4543  *	Return
4544  * 		The non-negative copied *buf* length equal to or less than
4545  * 		*size* on success, or a negative error in case of failure.
4546  *
4547  * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4548  *	Description
4549  *		Load header option.  Support reading a particular TCP header
4550  *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4551  *
4552  *		If *flags* is 0, it will search the option from the
4553  *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
4554  *		has details on what skb_data contains under different
4555  *		*skops*\ **->op**.
4556  *
4557  *		The first byte of the *searchby_res* specifies the
4558  *		kind that it wants to search.
4559  *
4560  *		If the searching kind is an experimental kind
4561  *		(i.e. 253 or 254 according to RFC6994).  It also
4562  *		needs to specify the "magic" which is either
4563  *		2 bytes or 4 bytes.  It then also needs to
4564  *		specify the size of the magic by using
4565  *		the 2nd byte which is "kind-length" of a TCP
4566  *		header option and the "kind-length" also
4567  *		includes the first 2 bytes "kind" and "kind-length"
4568  *		itself as a normal TCP header option also does.
4569  *
4570  *		For example, to search experimental kind 254 with
4571  *		2 byte magic 0xeB9F, the searchby_res should be
4572  *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4573  *
4574  *		To search for the standard window scale option (3),
4575  *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4576  *		Note, kind-length must be 0 for regular option.
4577  *
4578  *		Searching for No-Op (0) and End-of-Option-List (1) are
4579  *		not supported.
4580  *
4581  *		*len* must be at least 2 bytes which is the minimal size
4582  *		of a header option.
4583  *
4584  *		Supported flags:
4585  *
4586  *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4587  *		  saved_syn packet or the just-received syn packet.
4588  *
4589  *	Return
4590  *		> 0 when found, the header option is copied to *searchby_res*.
4591  *		The return value is the total length copied. On failure, a
4592  *		negative error code is returned:
4593  *
4594  *		**-EINVAL** if a parameter is invalid.
4595  *
4596  *		**-ENOMSG** if the option is not found.
4597  *
4598  *		**-ENOENT** if no syn packet is available when
4599  *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4600  *
4601  *		**-ENOSPC** if there is not enough space.  Only *len* number of
4602  *		bytes are copied.
4603  *
4604  *		**-EFAULT** on failure to parse the header options in the
4605  *		packet.
4606  *
4607  *		**-EPERM** if the helper cannot be used under the current
4608  *		*skops*\ **->op**.
4609  *
4610  * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4611  *	Description
4612  *		Store header option.  The data will be copied
4613  *		from buffer *from* with length *len* to the TCP header.
4614  *
4615  *		The buffer *from* should have the whole option that
4616  *		includes the kind, kind-length, and the actual
4617  *		option data.  The *len* must be at least kind-length
4618  *		long.  The kind-length does not have to be 4 byte
4619  *		aligned.  The kernel will take care of the padding
4620  *		and setting the 4 bytes aligned value to th->doff.
4621  *
4622  *		This helper will check for duplicated option
4623  *		by searching the same option in the outgoing skb.
4624  *
4625  *		This helper can only be called during
4626  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4627  *
4628  *	Return
4629  *		0 on success, or negative error in case of failure:
4630  *
4631  *		**-EINVAL** If param is invalid.
4632  *
4633  *		**-ENOSPC** if there is not enough space in the header.
4634  *		Nothing has been written
4635  *
4636  *		**-EEXIST** if the option already exists.
4637  *
4638  *		**-EFAULT** on failure to parse the existing header options.
4639  *
4640  *		**-EPERM** if the helper cannot be used under the current
4641  *		*skops*\ **->op**.
4642  *
4643  * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4644  *	Description
4645  *		Reserve *len* bytes for the bpf header option.  The
4646  *		space will be used by **bpf_store_hdr_opt**\ () later in
4647  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4648  *
4649  *		If **bpf_reserve_hdr_opt**\ () is called multiple times,
4650  *		the total number of bytes will be reserved.
4651  *
4652  *		This helper can only be called during
4653  *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4654  *
4655  *	Return
4656  *		0 on success, or negative error in case of failure:
4657  *
4658  *		**-EINVAL** if a parameter is invalid.
4659  *
4660  *		**-ENOSPC** if there is not enough space in the header.
4661  *
4662  *		**-EPERM** if the helper cannot be used under the current
4663  *		*skops*\ **->op**.
4664  *
4665  * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4666  *	Description
4667  *		Get a bpf_local_storage from an *inode*.
4668  *
4669  *		Logically, it could be thought of as getting the value from
4670  *		a *map* with *inode* as the **key**.  From this
4671  *		perspective,  the usage is not much different from
4672  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4673  *		helper enforces the key must be an inode and the map must also
4674  *		be a **BPF_MAP_TYPE_INODE_STORAGE**.
4675  *
4676  *		Underneath, the value is stored locally at *inode* instead of
4677  *		the *map*.  The *map* is used as the bpf-local-storage
4678  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4679  *		searched against all bpf_local_storage residing at *inode*.
4680  *
4681  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4682  *		used such that a new bpf_local_storage will be
4683  *		created if one does not exist.  *value* can be used
4684  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4685  *		the initial value of a bpf_local_storage.  If *value* is
4686  *		**NULL**, the new bpf_local_storage will be zero initialized.
4687  *	Return
4688  *		A bpf_local_storage pointer is returned on success.
4689  *
4690  *		**NULL** if not found or there was an error in adding
4691  *		a new bpf_local_storage.
4692  *
4693  * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4694  *	Description
4695  *		Delete a bpf_local_storage from an *inode*.
4696  *	Return
4697  *		0 on success.
4698  *
4699  *		**-ENOENT** if the bpf_local_storage cannot be found.
4700  *
4701  * long bpf_d_path(struct path *path, char *buf, u32 sz)
4702  *	Description
4703  *		Return full path for given **struct path** object, which
4704  *		needs to be the kernel BTF *path* object. The path is
4705  *		returned in the provided buffer *buf* of size *sz* and
4706  *		is zero terminated.
4707  *
4708  *	Return
4709  *		On success, the strictly positive length of the string,
4710  *		including the trailing NUL character. On error, a negative
4711  *		value.
4712  *
4713  * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4714  * 	Description
4715  * 		Read *size* bytes from user space address *user_ptr* and store
4716  * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4717  * 	Return
4718  * 		0 on success, or a negative error in case of failure.
4719  *
4720  * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4721  *	Description
4722  *		Use BTF to store a string representation of *ptr*->ptr in *str*,
4723  *		using *ptr*->type_id.  This value should specify the type
4724  *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4725  *		can be used to look up vmlinux BTF type ids. Traversing the
4726  *		data structure using BTF, the type information and values are
4727  *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
4728  *		the pointer data is carried out to avoid kernel crashes during
4729  *		operation.  Smaller types can use string space on the stack;
4730  *		larger programs can use map data to store the string
4731  *		representation.
4732  *
4733  *		The string can be subsequently shared with userspace via
4734  *		bpf_perf_event_output() or ring buffer interfaces.
4735  *		bpf_trace_printk() is to be avoided as it places too small
4736  *		a limit on string size to be useful.
4737  *
4738  *		*flags* is a combination of
4739  *
4740  *		**BTF_F_COMPACT**
4741  *			no formatting around type information
4742  *		**BTF_F_NONAME**
4743  *			no struct/union member names/types
4744  *		**BTF_F_PTR_RAW**
4745  *			show raw (unobfuscated) pointer values;
4746  *			equivalent to printk specifier %px.
4747  *		**BTF_F_ZERO**
4748  *			show zero-valued struct/union members; they
4749  *			are not displayed by default
4750  *
4751  *	Return
4752  *		The number of bytes that were written (or would have been
4753  *		written if output had to be truncated due to string size),
4754  *		or a negative error in cases of failure.
4755  *
4756  * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4757  *	Description
4758  *		Use BTF to write to seq_write a string representation of
4759  *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4760  *		*flags* are identical to those used for bpf_snprintf_btf.
4761  *	Return
4762  *		0 on success or a negative error in case of failure.
4763  *
4764  * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4765  * 	Description
4766  * 		See **bpf_get_cgroup_classid**\ () for the main description.
4767  * 		This helper differs from **bpf_get_cgroup_classid**\ () in that
4768  * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s
4769  * 		associated socket instead of the current process.
4770  * 	Return
4771  * 		The id is returned or 0 in case the id could not be retrieved.
4772  *
4773  * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4774  * 	Description
4775  * 		Redirect the packet to another net device of index *ifindex*
4776  * 		and fill in L2 addresses from neighboring subsystem. This helper
4777  * 		is somewhat similar to **bpf_redirect**\ (), except that it
4778  * 		populates L2 addresses as well, meaning, internally, the helper
4779  * 		relies on the neighbor lookup for the L2 address of the nexthop.
4780  *
4781  * 		The helper will perform a FIB lookup based on the skb's
4782  * 		networking header to get the address of the next hop, unless
4783  * 		this is supplied by the caller in the *params* argument. The
4784  * 		*plen* argument indicates the len of *params* and should be set
4785  * 		to 0 if *params* is NULL.
4786  *
4787  * 		The *flags* argument is reserved and must be 0. The helper is
4788  * 		currently only supported for tc BPF program types, and enabled
4789  * 		for IPv4 and IPv6 protocols.
4790  * 	Return
4791  * 		The helper returns **TC_ACT_REDIRECT** on success or
4792  * 		**TC_ACT_SHOT** on error.
4793  *
4794  * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4795  *     Description
4796  *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4797  *             pointer to the percpu kernel variable on *cpu*. A ksym is an
4798  *             extern variable decorated with '__ksym'. For ksym, there is a
4799  *             global var (either static or global) defined of the same name
4800  *             in the kernel. The ksym is percpu if the global var is percpu.
4801  *             The returned pointer points to the global percpu var on *cpu*.
4802  *
4803  *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4804  *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
4805  *             happens if *cpu* is larger than nr_cpu_ids. The caller of
4806  *             bpf_per_cpu_ptr() must check the returned value.
4807  *     Return
4808  *             A pointer pointing to the kernel percpu variable on *cpu*, or
4809  *             NULL, if *cpu* is invalid.
4810  *
4811  * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4812  *	Description
4813  *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4814  *		pointer to the percpu kernel variable on this cpu. See the
4815  *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4816  *
4817  *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4818  *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4819  *		never return NULL.
4820  *	Return
4821  *		A pointer pointing to the kernel percpu variable on this cpu.
4822  *
4823  * long bpf_redirect_peer(u32 ifindex, u64 flags)
4824  * 	Description
4825  * 		Redirect the packet to another net device of index *ifindex*.
4826  * 		This helper is somewhat similar to **bpf_redirect**\ (), except
4827  * 		that the redirection happens to the *ifindex*' peer device and
4828  * 		the netns switch takes place from ingress to ingress without
4829  * 		going through the CPU's backlog queue.
4830  *
4831  * 		The *flags* argument is reserved and must be 0. The helper is
4832  * 		currently only supported for tc BPF program types at the ingress
4833  * 		hook and for veth device types. The peer device must reside in a
4834  * 		different network namespace.
4835  * 	Return
4836  * 		The helper returns **TC_ACT_REDIRECT** on success or
4837  * 		**TC_ACT_SHOT** on error.
4838  *
4839  * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4840  *	Description
4841  *		Get a bpf_local_storage from the *task*.
4842  *
4843  *		Logically, it could be thought of as getting the value from
4844  *		a *map* with *task* as the **key**.  From this
4845  *		perspective,  the usage is not much different from
4846  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4847  *		helper enforces the key must be a task_struct and the map must also
4848  *		be a **BPF_MAP_TYPE_TASK_STORAGE**.
4849  *
4850  *		Underneath, the value is stored locally at *task* instead of
4851  *		the *map*.  The *map* is used as the bpf-local-storage
4852  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4853  *		searched against all bpf_local_storage residing at *task*.
4854  *
4855  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4856  *		used such that a new bpf_local_storage will be
4857  *		created if one does not exist.  *value* can be used
4858  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4859  *		the initial value of a bpf_local_storage.  If *value* is
4860  *		**NULL**, the new bpf_local_storage will be zero initialized.
4861  *	Return
4862  *		A bpf_local_storage pointer is returned on success.
4863  *
4864  *		**NULL** if not found or there was an error in adding
4865  *		a new bpf_local_storage.
4866  *
4867  * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4868  *	Description
4869  *		Delete a bpf_local_storage from a *task*.
4870  *	Return
4871  *		0 on success.
4872  *
4873  *		**-ENOENT** if the bpf_local_storage cannot be found.
4874  *
4875  * struct task_struct *bpf_get_current_task_btf(void)
4876  *	Description
4877  *		Return a BTF pointer to the "current" task.
4878  *		This pointer can also be used in helpers that accept an
4879  *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.
4880  *	Return
4881  *		Pointer to the current task.
4882  *
4883  * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4884  *	Description
4885  *		Set or clear certain options on *bprm*:
4886  *
4887  *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4888  *		which sets the **AT_SECURE** auxv for glibc. The bit
4889  *		is cleared if the flag is not specified.
4890  *	Return
4891  *		**-EINVAL** if invalid *flags* are passed, zero otherwise.
4892  *
4893  * u64 bpf_ktime_get_coarse_ns(void)
4894  * 	Description
4895  * 		Return a coarse-grained version of the time elapsed since
4896  * 		system boot, in nanoseconds. Does not include time the system
4897  * 		was suspended.
4898  *
4899  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4900  * 	Return
4901  * 		Current *ktime*.
4902  *
4903  * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4904  *	Description
4905  *		Returns the stored IMA hash of the *inode* (if it's available).
4906  *		If the hash is larger than *size*, then only *size*
4907  *		bytes will be copied to *dst*
4908  *	Return
4909  *		The **hash_algo** is returned on success,
4910  *		**-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4911  *		invalid arguments are passed.
4912  *
4913  * struct socket *bpf_sock_from_file(struct file *file)
4914  *	Description
4915  *		If the given file represents a socket, returns the associated
4916  *		socket.
4917  *	Return
4918  *		A pointer to a struct socket on success or NULL if the file is
4919  *		not a socket.
4920  *
4921  * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4922  *	Description
4923  *		Check packet size against exceeding MTU of net device (based
4924  *		on *ifindex*).  This helper will likely be used in combination
4925  *		with helpers that adjust/change the packet size.
4926  *
4927  *		The argument *len_diff* can be used for querying with a planned
4928  *		size change. This allows to check MTU prior to changing packet
4929  *		ctx. Providing a *len_diff* adjustment that is larger than the
4930  *		actual packet size (resulting in negative packet size) will in
4931  *		principle not exceed the MTU, which is why it is not considered
4932  *		a failure.  Other BPF helpers are needed for performing the
4933  *		planned size change; therefore the responsibility for catching
4934  *		a negative packet size belongs in those helpers.
4935  *
4936  *		Specifying *ifindex* zero means the MTU check is performed
4937  *		against the current net device.  This is practical if this isn't
4938  *		used prior to redirect.
4939  *
4940  *		On input *mtu_len* must be a valid pointer, else verifier will
4941  *		reject BPF program.  If the value *mtu_len* is initialized to
4942  *		zero then the ctx packet size is use.  When value *mtu_len* is
4943  *		provided as input this specify the L3 length that the MTU check
4944  *		is done against. Remember XDP and TC length operate at L2, but
4945  *		this value is L3 as this correlate to MTU and IP-header tot_len
4946  *		values which are L3 (similar behavior as bpf_fib_lookup).
4947  *
4948  *		The Linux kernel route table can configure MTUs on a more
4949  *		specific per route level, which is not provided by this helper.
4950  *		For route level MTU checks use the **bpf_fib_lookup**\ ()
4951  *		helper.
4952  *
4953  *		*ctx* is either **struct xdp_md** for XDP programs or
4954  *		**struct sk_buff** for tc cls_act programs.
4955  *
4956  *		The *flags* argument can be a combination of one or more of the
4957  *		following values:
4958  *
4959  *		**BPF_MTU_CHK_SEGS**
4960  *			This flag will only works for *ctx* **struct sk_buff**.
4961  *			If packet context contains extra packet segment buffers
4962  *			(often knows as GSO skb), then MTU check is harder to
4963  *			check at this point, because in transmit path it is
4964  *			possible for the skb packet to get re-segmented
4965  *			(depending on net device features).  This could still be
4966  *			a MTU violation, so this flag enables performing MTU
4967  *			check against segments, with a different violation
4968  *			return code to tell it apart. Check cannot use len_diff.
4969  *
4970  *		On return *mtu_len* pointer contains the MTU value of the net
4971  *		device.  Remember the net device configured MTU is the L3 size,
4972  *		which is returned here and XDP and TC length operate at L2.
4973  *		Helper take this into account for you, but remember when using
4974  *		MTU value in your BPF-code.
4975  *
4976  *	Return
4977  *		* 0 on success, and populate MTU value in *mtu_len* pointer.
4978  *
4979  *		* < 0 if any input argument is invalid (*mtu_len* not updated)
4980  *
4981  *		MTU violations return positive values, but also populate MTU
4982  *		value in *mtu_len* pointer, as this can be needed for
4983  *		implementing PMTU handing:
4984  *
4985  *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**
4986  *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4987  *
4988  * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4989  *	Description
4990  *		For each element in **map**, call **callback_fn** function with
4991  *		**map**, **callback_ctx** and other map-specific parameters.
4992  *		The **callback_fn** should be a static function and
4993  *		the **callback_ctx** should be a pointer to the stack.
4994  *		The **flags** is used to control certain aspects of the helper.
4995  *		Currently, the **flags** must be 0.
4996  *
4997  *		The following are a list of supported map types and their
4998  *		respective expected callback signatures:
4999  *
5000  *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
5001  *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
5002  *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
5003  *
5004  *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
5005  *
5006  *		For per_cpu maps, the map_value is the value on the cpu where the
5007  *		bpf_prog is running.
5008  *
5009  *		If **callback_fn** return 0, the helper will continue to the next
5010  *		element. If return value is 1, the helper will skip the rest of
5011  *		elements and return. Other return values are not used now.
5012  *
5013  *	Return
5014  *		The number of traversed map elements for success, **-EINVAL** for
5015  *		invalid **flags**.
5016  *
5017  * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
5018  *	Description
5019  *		Outputs a string into the **str** buffer of size **str_size**
5020  *		based on a format string stored in a read-only map pointed by
5021  *		**fmt**.
5022  *
5023  *		Each format specifier in **fmt** corresponds to one u64 element
5024  *		in the **data** array. For strings and pointers where pointees
5025  *		are accessed, only the pointer values are stored in the *data*
5026  *		array. The *data_len* is the size of *data* in bytes - must be
5027  *		a multiple of 8.
5028  *
5029  *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel
5030  *		memory. Reading kernel memory may fail due to either invalid
5031  *		address or valid address but requiring a major memory fault. If
5032  *		reading kernel memory fails, the string for **%s** will be an
5033  *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
5034  *		Not returning error to bpf program is consistent with what
5035  *		**bpf_trace_printk**\ () does for now.
5036  *
5037  *	Return
5038  *		The strictly positive length of the formatted string, including
5039  *		the trailing zero character. If the return value is greater than
5040  *		**str_size**, **str** contains a truncated string, guaranteed to
5041  *		be zero-terminated except when **str_size** is 0.
5042  *
5043  *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.
5044  *
5045  * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
5046  * 	Description
5047  * 		Execute bpf syscall with given arguments.
5048  * 	Return
5049  * 		A syscall result.
5050  *
5051  * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
5052  * 	Description
5053  * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
5054  * 	Return
5055  * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.
5056  *
5057  * long bpf_sys_close(u32 fd)
5058  * 	Description
5059  * 		Execute close syscall for given FD.
5060  * 	Return
5061  * 		A syscall result.
5062  *
5063  * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
5064  *	Description
5065  *		Initialize the timer.
5066  *		First 4 bits of *flags* specify clockid.
5067  *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
5068  *		All other bits of *flags* are reserved.
5069  *		The verifier will reject the program if *timer* is not from
5070  *		the same *map*.
5071  *	Return
5072  *		0 on success.
5073  *		**-EBUSY** if *timer* is already initialized.
5074  *		**-EINVAL** if invalid *flags* are passed.
5075  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
5076  *		The user space should either hold a file descriptor to a map with timers
5077  *		or pin such map in bpffs. When map is unpinned or file descriptor is
5078  *		closed all timers in the map will be cancelled and freed.
5079  *
5080  * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
5081  *	Description
5082  *		Configure the timer to call *callback_fn* static function.
5083  *	Return
5084  *		0 on success.
5085  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5086  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
5087  *		The user space should either hold a file descriptor to a map with timers
5088  *		or pin such map in bpffs. When map is unpinned or file descriptor is
5089  *		closed all timers in the map will be cancelled and freed.
5090  *
5091  * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
5092  *	Description
5093  *		Set timer expiration N nanoseconds from the current time. The
5094  *		configured callback will be invoked in soft irq context on some cpu
5095  *		and will not repeat unless another bpf_timer_start() is made.
5096  *		In such case the next invocation can migrate to a different cpu.
5097  *		Since struct bpf_timer is a field inside map element the map
5098  *		owns the timer. The bpf_timer_set_callback() will increment refcnt
5099  *		of BPF program to make sure that callback_fn code stays valid.
5100  *		When user space reference to a map reaches zero all timers
5101  *		in a map are cancelled and corresponding program's refcnts are
5102  *		decremented. This is done to make sure that Ctrl-C of a user
5103  *		process doesn't leave any timers running. If map is pinned in
5104  *		bpffs the callback_fn can re-arm itself indefinitely.
5105  *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands
5106  *		cancel and free the timer in the given map element.
5107  *		The map can contain timers that invoke callback_fn-s from different
5108  *		programs. The same callback_fn can serve different timers from
5109  *		different maps if key/value layout matches across maps.
5110  *		Every bpf_timer_set_callback() can have different callback_fn.
5111  *
5112  *		*flags* can be one of:
5113  *
5114  *		**BPF_F_TIMER_ABS**
5115  *			Start the timer in absolute expire value instead of the
5116  *			default relative one.
5117  *
5118  *	Return
5119  *		0 on success.
5120  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
5121  *		or invalid *flags* are passed.
5122  *
5123  * long bpf_timer_cancel(struct bpf_timer *timer)
5124  *	Description
5125  *		Cancel the timer and wait for callback_fn to finish if it was running.
5126  *	Return
5127  *		0 if the timer was not active.
5128  *		1 if the timer was active.
5129  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
5130  *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
5131  *		own timer which would have led to a deadlock otherwise.
5132  *
5133  * u64 bpf_get_func_ip(void *ctx)
5134  * 	Description
5135  * 		Get address of the traced function (for tracing and kprobe programs).
5136  *
5137  * 		When called for kprobe program attached as uprobe it returns
5138  * 		probe address for both entry and return uprobe.
5139  *
5140  * 	Return
5141  * 		Address of the traced function for kprobe.
5142  * 		0 for kprobes placed within the function (not at the entry).
5143  * 		Address of the probe for uprobe and return uprobe.
5144  *
5145  * u64 bpf_get_attach_cookie(void *ctx)
5146  * 	Description
5147  * 		Get bpf_cookie value provided (optionally) during the program
5148  * 		attachment. It might be different for each individual
5149  * 		attachment, even if BPF program itself is the same.
5150  * 		Expects BPF program context *ctx* as a first argument.
5151  *
5152  * 		Supported for the following program types:
5153  *			- kprobe/uprobe;
5154  *			- tracepoint;
5155  *			- perf_event.
5156  * 	Return
5157  *		Value specified by user at BPF link creation/attachment time
5158  *		or 0, if it was not specified.
5159  *
5160  * long bpf_task_pt_regs(struct task_struct *task)
5161  *	Description
5162  *		Get the struct pt_regs associated with **task**.
5163  *	Return
5164  *		A pointer to struct pt_regs.
5165  *
5166  * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5167  *	Description
5168  *		Get branch trace from hardware engines like Intel LBR. The
5169  *		hardware engine is stopped shortly after the helper is
5170  *		called. Therefore, the user need to filter branch entries
5171  *		based on the actual use case. To capture branch trace
5172  *		before the trigger point of the BPF program, the helper
5173  *		should be called at the beginning of the BPF program.
5174  *
5175  *		The data is stored as struct perf_branch_entry into output
5176  *		buffer *entries*. *size* is the size of *entries* in bytes.
5177  *		*flags* is reserved for now and must be zero.
5178  *
5179  *	Return
5180  *		On success, number of bytes written to *buf*. On error, a
5181  *		negative value.
5182  *
5183  *		**-EINVAL** if *flags* is not zero.
5184  *
5185  *		**-ENOENT** if architecture does not support branch records.
5186  *
5187  * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5188  *	Description
5189  *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5190  *		to format and can handle more format args as a result.
5191  *
5192  *		Arguments are to be used as in **bpf_seq_printf**\ () helper.
5193  *	Return
5194  *		The number of bytes written to the buffer, or a negative error
5195  *		in case of failure.
5196  *
5197  * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5198  * 	Description
5199  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5200  *	Return
5201  *		*sk* if casting is valid, or **NULL** otherwise.
5202  *
5203  * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5204  *	Description
5205  *		Get the address of a kernel symbol, returned in *res*. *res* is
5206  *		set to 0 if the symbol is not found.
5207  *	Return
5208  *		On success, zero. On error, a negative value.
5209  *
5210  *		**-EINVAL** if *flags* is not zero.
5211  *
5212  *		**-EINVAL** if string *name* is not the same size as *name_sz*.
5213  *
5214  *		**-ENOENT** if symbol is not found.
5215  *
5216  *		**-EPERM** if caller does not have permission to obtain kernel address.
5217  *
5218  * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5219  *	Description
5220  *		Find vma of *task* that contains *addr*, call *callback_fn*
5221  *		function with *task*, *vma*, and *callback_ctx*.
5222  *		The *callback_fn* should be a static function and
5223  *		the *callback_ctx* should be a pointer to the stack.
5224  *		The *flags* is used to control certain aspects of the helper.
5225  *		Currently, the *flags* must be 0.
5226  *
5227  *		The expected callback signature is
5228  *
5229  *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5230  *
5231  *	Return
5232  *		0 on success.
5233  *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5234  *		**-EBUSY** if failed to try lock mmap_lock.
5235  *		**-EINVAL** for invalid **flags**.
5236  *
5237  * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5238  *	Description
5239  *		For **nr_loops**, call **callback_fn** function
5240  *		with **callback_ctx** as the context parameter.
5241  *		The **callback_fn** should be a static function and
5242  *		the **callback_ctx** should be a pointer to the stack.
5243  *		The **flags** is used to control certain aspects of the helper.
5244  *		Currently, the **flags** must be 0. Currently, nr_loops is
5245  *		limited to 1 << 23 (~8 million) loops.
5246  *
5247  *		long (\*callback_fn)(u32 index, void \*ctx);
5248  *
5249  *		where **index** is the current index in the loop. The index
5250  *		is zero-indexed.
5251  *
5252  *		If **callback_fn** returns 0, the helper will continue to the next
5253  *		loop. If return value is 1, the helper will skip the rest of
5254  *		the loops and return. Other return values are not used now,
5255  *		and will be rejected by the verifier.
5256  *
5257  *	Return
5258  *		The number of loops performed, **-EINVAL** for invalid **flags**,
5259  *		**-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5260  *
5261  * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5262  *	Description
5263  *		Do strncmp() between **s1** and **s2**. **s1** doesn't need
5264  *		to be null-terminated and **s1_sz** is the maximum storage
5265  *		size of **s1**. **s2** must be a read-only string.
5266  *	Return
5267  *		An integer less than, equal to, or greater than zero
5268  *		if the first **s1_sz** bytes of **s1** is found to be
5269  *		less than, to match, or be greater than **s2**.
5270  *
5271  * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5272  *	Description
5273  *		Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5274  *		returned in **value**.
5275  *
5276  *	Return
5277  *		0 on success.
5278  *		**-EINVAL** if n >= argument register count of traced function.
5279  *
5280  * long bpf_get_func_ret(void *ctx, u64 *value)
5281  *	Description
5282  *		Get return value of the traced function (for tracing programs)
5283  *		in **value**.
5284  *
5285  *	Return
5286  *		0 on success.
5287  *		**-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5288  *
5289  * long bpf_get_func_arg_cnt(void *ctx)
5290  *	Description
5291  *		Get number of registers of the traced function (for tracing programs) where
5292  *		function arguments are stored in these registers.
5293  *
5294  *	Return
5295  *		The number of argument registers of the traced function.
5296  *
5297  * int bpf_get_retval(void)
5298  *	Description
5299  *		Get the BPF program's return value that will be returned to the upper layers.
5300  *
5301  *		This helper is currently supported by cgroup programs and only by the hooks
5302  *		where BPF program's return value is returned to the userspace via errno.
5303  *	Return
5304  *		The BPF program's return value.
5305  *
5306  * int bpf_set_retval(int retval)
5307  *	Description
5308  *		Set the BPF program's return value that will be returned to the upper layers.
5309  *
5310  *		This helper is currently supported by cgroup programs and only by the hooks
5311  *		where BPF program's return value is returned to the userspace via errno.
5312  *
5313  *		Note that there is the following corner case where the program exports an error
5314  *		via bpf_set_retval but signals success via 'return 1':
5315  *
5316  *			bpf_set_retval(-EPERM);
5317  *			return 1;
5318  *
5319  *		In this case, the BPF program's return value will use helper's -EPERM. This
5320  *		still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5321  *
5322  *	Return
5323  *		0 on success, or a negative error in case of failure.
5324  *
5325  * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5326  *	Description
5327  *		Get the total size of a given xdp buff (linear and paged area)
5328  *	Return
5329  *		The total size of a given xdp buffer.
5330  *
5331  * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5332  *	Description
5333  *		This helper is provided as an easy way to load data from a
5334  *		xdp buffer. It can be used to load *len* bytes from *offset* from
5335  *		the frame associated to *xdp_md*, into the buffer pointed by
5336  *		*buf*.
5337  *	Return
5338  *		0 on success, or a negative error in case of failure.
5339  *
5340  * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5341  *	Description
5342  *		Store *len* bytes from buffer *buf* into the frame
5343  *		associated to *xdp_md*, at *offset*.
5344  *	Return
5345  *		0 on success, or a negative error in case of failure.
5346  *
5347  * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5348  *	Description
5349  *		Read *size* bytes from user space address *user_ptr* in *tsk*'s
5350  *		address space, and stores the data in *dst*. *flags* is not
5351  *		used yet and is provided for future extensibility. This helper
5352  *		can only be used by sleepable programs.
5353  *	Return
5354  *		0 on success, or a negative error in case of failure. On error
5355  *		*dst* buffer is zeroed out.
5356  *
5357  * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5358  *	Description
5359  *		Change the __sk_buff->tstamp_type to *tstamp_type*
5360  *		and set *tstamp* to the __sk_buff->tstamp together.
5361  *
5362  *		If there is no need to change the __sk_buff->tstamp_type,
5363  *		the tstamp value can be directly written to __sk_buff->tstamp
5364  *		instead.
5365  *
5366  *		BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5367  *		will be kept during bpf_redirect_*().  A non zero
5368  *		*tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5369  *		*tstamp_type*.
5370  *
5371  *		A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5372  *		with a zero *tstamp*.
5373  *
5374  *		Only IPv4 and IPv6 skb->protocol are supported.
5375  *
5376  *		This function is most useful when it needs to set a
5377  *		mono delivery time to __sk_buff->tstamp and then
5378  *		bpf_redirect_*() to the egress of an iface.  For example,
5379  *		changing the (rcv) timestamp in __sk_buff->tstamp at
5380  *		ingress to a mono delivery time and then bpf_redirect_*()
5381  *		to sch_fq@phy-dev.
5382  *	Return
5383  *		0 on success.
5384  *		**-EINVAL** for invalid input
5385  *		**-EOPNOTSUPP** for unsupported protocol
5386  *
5387  * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5388  *	Description
5389  *		Returns a calculated IMA hash of the *file*.
5390  *		If the hash is larger than *size*, then only *size*
5391  *		bytes will be copied to *dst*
5392  *	Return
5393  *		The **hash_algo** is returned on success,
5394  *		**-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
5395  *		invalid arguments are passed.
5396  *
5397  * void *bpf_kptr_xchg(void *map_value, void *ptr)
5398  *	Description
5399  *		Exchange kptr at pointer *map_value* with *ptr*, and return the
5400  *		old value. *ptr* can be NULL, otherwise it must be a referenced
5401  *		pointer which will be released when this helper is called.
5402  *	Return
5403  *		The old value of kptr (which can be NULL). The returned pointer
5404  *		if not NULL, is a reference which must be released using its
5405  *		corresponding release function, or moved into a BPF map before
5406  *		program exit.
5407  *
5408  * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5409  * 	Description
5410  * 		Perform a lookup in *percpu map* for an entry associated to
5411  * 		*key* on *cpu*.
5412  * 	Return
5413  * 		Map value associated to *key* on *cpu*, or **NULL** if no entry
5414  * 		was found or *cpu* is invalid.
5415  *
5416  * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5417  *	Description
5418  *		Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5419  *	Return
5420  *		*sk* if casting is valid, or **NULL** otherwise.
5421  *
5422  * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5423  *	Description
5424  *		Get a dynptr to local memory *data*.
5425  *
5426  *		*data* must be a ptr to a map value.
5427  *		The maximum *size* supported is DYNPTR_MAX_SIZE.
5428  *		*flags* is currently unused.
5429  *	Return
5430  *		0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5431  *		-EINVAL if flags is not 0.
5432  *
5433  * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5434  *	Description
5435  *		Reserve *size* bytes of payload in a ring buffer *ringbuf*
5436  *		through the dynptr interface. *flags* must be 0.
5437  *
5438  *		Please note that a corresponding bpf_ringbuf_submit_dynptr or
5439  *		bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5440  *		reservation fails. This is enforced by the verifier.
5441  *	Return
5442  *		0 on success, or a negative error in case of failure.
5443  *
5444  * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5445  *	Description
5446  *		Submit reserved ring buffer sample, pointed to by *data*,
5447  *		through the dynptr interface. This is a no-op if the dynptr is
5448  *		invalid/null.
5449  *
5450  *		For more information on *flags*, please see
5451  *		'bpf_ringbuf_submit'.
5452  *	Return
5453  *		Nothing. Always succeeds.
5454  *
5455  * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5456  *	Description
5457  *		Discard reserved ring buffer sample through the dynptr
5458  *		interface. This is a no-op if the dynptr is invalid/null.
5459  *
5460  *		For more information on *flags*, please see
5461  *		'bpf_ringbuf_discard'.
5462  *	Return
5463  *		Nothing. Always succeeds.
5464  *
5465  * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
5466  *	Description
5467  *		Read *len* bytes from *src* into *dst*, starting from *offset*
5468  *		into *src*.
5469  *		*flags* is currently unused.
5470  *	Return
5471  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5472  *		of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5473  *		*flags* is not 0.
5474  *
5475  * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5476  *	Description
5477  *		Write *len* bytes from *src* into *dst*, starting from *offset*
5478  *		into *dst*.
5479  *
5480  *		*flags* must be 0 except for skb-type dynptrs.
5481  *
5482  *		For skb-type dynptrs:
5483  *		    *  All data slices of the dynptr are automatically
5484  *		       invalidated after **bpf_dynptr_write**\ (). This is
5485  *		       because writing may pull the skb and change the
5486  *		       underlying packet buffer.
5487  *
5488  *		    *  For *flags*, please see the flags accepted by
5489  *		       **bpf_skb_store_bytes**\ ().
5490  *	Return
5491  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5492  *		of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5493  *		is a read-only dynptr or if *flags* is not correct. For skb-type dynptrs,
5494  *		other errors correspond to errors returned by **bpf_skb_store_bytes**\ ().
5495  *
5496  * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
5497  *	Description
5498  *		Get a pointer to the underlying dynptr data.
5499  *
5500  *		*len* must be a statically known value. The returned data slice
5501  *		is invalidated whenever the dynptr is invalidated.
5502  *
5503  *		skb and xdp type dynptrs may not use bpf_dynptr_data. They should
5504  *		instead use bpf_dynptr_slice and bpf_dynptr_slice_rdwr.
5505  *	Return
5506  *		Pointer to the underlying dynptr data, NULL if the dynptr is
5507  *		read-only, if the dynptr is invalid, or if the offset and length
5508  *		is out of bounds.
5509  *
5510  * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5511  *	Description
5512  *		Try to issue a SYN cookie for the packet with corresponding
5513  *		IPv4/TCP headers, *iph* and *th*, without depending on a
5514  *		listening socket.
5515  *
5516  *		*iph* points to the IPv4 header.
5517  *
5518  *		*th* points to the start of the TCP header, while *th_len*
5519  *		contains the length of the TCP header (at least
5520  *		**sizeof**\ (**struct tcphdr**)).
5521  *	Return
5522  *		On success, lower 32 bits hold the generated SYN cookie in
5523  *		followed by 16 bits which hold the MSS value for that cookie,
5524  *		and the top 16 bits are unused.
5525  *
5526  *		On failure, the returned value is one of the following:
5527  *
5528  *		**-EINVAL** if *th_len* is invalid.
5529  *
5530  * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5531  *	Description
5532  *		Try to issue a SYN cookie for the packet with corresponding
5533  *		IPv6/TCP headers, *iph* and *th*, without depending on a
5534  *		listening socket.
5535  *
5536  *		*iph* points to the IPv6 header.
5537  *
5538  *		*th* points to the start of the TCP header, while *th_len*
5539  *		contains the length of the TCP header (at least
5540  *		**sizeof**\ (**struct tcphdr**)).
5541  *	Return
5542  *		On success, lower 32 bits hold the generated SYN cookie in
5543  *		followed by 16 bits which hold the MSS value for that cookie,
5544  *		and the top 16 bits are unused.
5545  *
5546  *		On failure, the returned value is one of the following:
5547  *
5548  *		**-EINVAL** if *th_len* is invalid.
5549  *
5550  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5551  *
5552  * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5553  *	Description
5554  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5555  *		without depending on a listening socket.
5556  *
5557  *		*iph* points to the IPv4 header.
5558  *
5559  *		*th* points to the TCP header.
5560  *	Return
5561  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5562  *
5563  *		On failure, the returned value is one of the following:
5564  *
5565  *		**-EACCES** if the SYN cookie is not valid.
5566  *
5567  * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5568  *	Description
5569  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5570  *		without depending on a listening socket.
5571  *
5572  *		*iph* points to the IPv6 header.
5573  *
5574  *		*th* points to the TCP header.
5575  *	Return
5576  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5577  *
5578  *		On failure, the returned value is one of the following:
5579  *
5580  *		**-EACCES** if the SYN cookie is not valid.
5581  *
5582  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5583  *
5584  * u64 bpf_ktime_get_tai_ns(void)
5585  *	Description
5586  *		A nonsettable system-wide clock derived from wall-clock time but
5587  *		ignoring leap seconds.  This clock does not experience
5588  *		discontinuities and backwards jumps caused by NTP inserting leap
5589  *		seconds as CLOCK_REALTIME does.
5590  *
5591  *		See: **clock_gettime**\ (**CLOCK_TAI**)
5592  *	Return
5593  *		Current *ktime*.
5594  *
5595  * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5596  *	Description
5597  *		Drain samples from the specified user ring buffer, and invoke
5598  *		the provided callback for each such sample:
5599  *
5600  *		long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5601  *
5602  *		If **callback_fn** returns 0, the helper will continue to try
5603  *		and drain the next sample, up to a maximum of
5604  *		BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5605  *		the helper will skip the rest of the samples and return. Other
5606  *		return values are not used now, and will be rejected by the
5607  *		verifier.
5608  *	Return
5609  *		The number of drained samples if no error was encountered while
5610  *		draining samples, or 0 if no samples were present in the ring
5611  *		buffer. If a user-space producer was epoll-waiting on this map,
5612  *		and at least one sample was drained, they will receive an event
5613  *		notification notifying them of available space in the ring
5614  *		buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5615  *		function, no wakeup notification will be sent. If the
5616  *		BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5617  *		be sent even if no sample was drained.
5618  *
5619  *		On failure, the returned value is one of the following:
5620  *
5621  *		**-EBUSY** if the ring buffer is contended, and another calling
5622  *		context was concurrently draining the ring buffer.
5623  *
5624  *		**-EINVAL** if user-space is not properly tracking the ring
5625  *		buffer due to the producer position not being aligned to 8
5626  *		bytes, a sample not being aligned to 8 bytes, or the producer
5627  *		position not matching the advertised length of a sample.
5628  *
5629  *		**-E2BIG** if user-space has tried to publish a sample which is
5630  *		larger than the size of the ring buffer, or which cannot fit
5631  *		within a struct bpf_dynptr.
5632  *
5633  * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5634  *	Description
5635  *		Get a bpf_local_storage from the *cgroup*.
5636  *
5637  *		Logically, it could be thought of as getting the value from
5638  *		a *map* with *cgroup* as the **key**.  From this
5639  *		perspective,  the usage is not much different from
5640  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5641  *		helper enforces the key must be a cgroup struct and the map must also
5642  *		be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5643  *
5644  *		In reality, the local-storage value is embedded directly inside of the
5645  *		*cgroup* object itself, rather than being located in the
5646  *		**BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5647  *		queried for some *map* on a *cgroup* object, the kernel will perform an
5648  *		O(n) iteration over all of the live local-storage values for that
5649  *		*cgroup* object until the local-storage value for the *map* is found.
5650  *
5651  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5652  *		used such that a new bpf_local_storage will be
5653  *		created if one does not exist.  *value* can be used
5654  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5655  *		the initial value of a bpf_local_storage.  If *value* is
5656  *		**NULL**, the new bpf_local_storage will be zero initialized.
5657  *	Return
5658  *		A bpf_local_storage pointer is returned on success.
5659  *
5660  *		**NULL** if not found or there was an error in adding
5661  *		a new bpf_local_storage.
5662  *
5663  * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5664  *	Description
5665  *		Delete a bpf_local_storage from a *cgroup*.
5666  *	Return
5667  *		0 on success.
5668  *
5669  *		**-ENOENT** if the bpf_local_storage cannot be found.
5670  */
5671 #define ___BPF_FUNC_MAPPER(FN, ctx...)			\
5672 	FN(unspec, 0, ##ctx)				\
5673 	FN(map_lookup_elem, 1, ##ctx)			\
5674 	FN(map_update_elem, 2, ##ctx)			\
5675 	FN(map_delete_elem, 3, ##ctx)			\
5676 	FN(probe_read, 4, ##ctx)			\
5677 	FN(ktime_get_ns, 5, ##ctx)			\
5678 	FN(trace_printk, 6, ##ctx)			\
5679 	FN(get_prandom_u32, 7, ##ctx)			\
5680 	FN(get_smp_processor_id, 8, ##ctx)		\
5681 	FN(skb_store_bytes, 9, ##ctx)			\
5682 	FN(l3_csum_replace, 10, ##ctx)			\
5683 	FN(l4_csum_replace, 11, ##ctx)			\
5684 	FN(tail_call, 12, ##ctx)			\
5685 	FN(clone_redirect, 13, ##ctx)			\
5686 	FN(get_current_pid_tgid, 14, ##ctx)		\
5687 	FN(get_current_uid_gid, 15, ##ctx)		\
5688 	FN(get_current_comm, 16, ##ctx)			\
5689 	FN(get_cgroup_classid, 17, ##ctx)		\
5690 	FN(skb_vlan_push, 18, ##ctx)			\
5691 	FN(skb_vlan_pop, 19, ##ctx)			\
5692 	FN(skb_get_tunnel_key, 20, ##ctx)		\
5693 	FN(skb_set_tunnel_key, 21, ##ctx)		\
5694 	FN(perf_event_read, 22, ##ctx)			\
5695 	FN(redirect, 23, ##ctx)				\
5696 	FN(get_route_realm, 24, ##ctx)			\
5697 	FN(perf_event_output, 25, ##ctx)		\
5698 	FN(skb_load_bytes, 26, ##ctx)			\
5699 	FN(get_stackid, 27, ##ctx)			\
5700 	FN(csum_diff, 28, ##ctx)			\
5701 	FN(skb_get_tunnel_opt, 29, ##ctx)		\
5702 	FN(skb_set_tunnel_opt, 30, ##ctx)		\
5703 	FN(skb_change_proto, 31, ##ctx)			\
5704 	FN(skb_change_type, 32, ##ctx)			\
5705 	FN(skb_under_cgroup, 33, ##ctx)			\
5706 	FN(get_hash_recalc, 34, ##ctx)			\
5707 	FN(get_current_task, 35, ##ctx)			\
5708 	FN(probe_write_user, 36, ##ctx)			\
5709 	FN(current_task_under_cgroup, 37, ##ctx)	\
5710 	FN(skb_change_tail, 38, ##ctx)			\
5711 	FN(skb_pull_data, 39, ##ctx)			\
5712 	FN(csum_update, 40, ##ctx)			\
5713 	FN(set_hash_invalid, 41, ##ctx)			\
5714 	FN(get_numa_node_id, 42, ##ctx)			\
5715 	FN(skb_change_head, 43, ##ctx)			\
5716 	FN(xdp_adjust_head, 44, ##ctx)			\
5717 	FN(probe_read_str, 45, ##ctx)			\
5718 	FN(get_socket_cookie, 46, ##ctx)		\
5719 	FN(get_socket_uid, 47, ##ctx)			\
5720 	FN(set_hash, 48, ##ctx)				\
5721 	FN(setsockopt, 49, ##ctx)			\
5722 	FN(skb_adjust_room, 50, ##ctx)			\
5723 	FN(redirect_map, 51, ##ctx)			\
5724 	FN(sk_redirect_map, 52, ##ctx)			\
5725 	FN(sock_map_update, 53, ##ctx)			\
5726 	FN(xdp_adjust_meta, 54, ##ctx)			\
5727 	FN(perf_event_read_value, 55, ##ctx)		\
5728 	FN(perf_prog_read_value, 56, ##ctx)		\
5729 	FN(getsockopt, 57, ##ctx)			\
5730 	FN(override_return, 58, ##ctx)			\
5731 	FN(sock_ops_cb_flags_set, 59, ##ctx)		\
5732 	FN(msg_redirect_map, 60, ##ctx)			\
5733 	FN(msg_apply_bytes, 61, ##ctx)			\
5734 	FN(msg_cork_bytes, 62, ##ctx)			\
5735 	FN(msg_pull_data, 63, ##ctx)			\
5736 	FN(bind, 64, ##ctx)				\
5737 	FN(xdp_adjust_tail, 65, ##ctx)			\
5738 	FN(skb_get_xfrm_state, 66, ##ctx)		\
5739 	FN(get_stack, 67, ##ctx)			\
5740 	FN(skb_load_bytes_relative, 68, ##ctx)		\
5741 	FN(fib_lookup, 69, ##ctx)			\
5742 	FN(sock_hash_update, 70, ##ctx)			\
5743 	FN(msg_redirect_hash, 71, ##ctx)		\
5744 	FN(sk_redirect_hash, 72, ##ctx)			\
5745 	FN(lwt_push_encap, 73, ##ctx)			\
5746 	FN(lwt_seg6_store_bytes, 74, ##ctx)		\
5747 	FN(lwt_seg6_adjust_srh, 75, ##ctx)		\
5748 	FN(lwt_seg6_action, 76, ##ctx)			\
5749 	FN(rc_repeat, 77, ##ctx)			\
5750 	FN(rc_keydown, 78, ##ctx)			\
5751 	FN(skb_cgroup_id, 79, ##ctx)			\
5752 	FN(get_current_cgroup_id, 80, ##ctx)		\
5753 	FN(get_local_storage, 81, ##ctx)		\
5754 	FN(sk_select_reuseport, 82, ##ctx)		\
5755 	FN(skb_ancestor_cgroup_id, 83, ##ctx)		\
5756 	FN(sk_lookup_tcp, 84, ##ctx)			\
5757 	FN(sk_lookup_udp, 85, ##ctx)			\
5758 	FN(sk_release, 86, ##ctx)			\
5759 	FN(map_push_elem, 87, ##ctx)			\
5760 	FN(map_pop_elem, 88, ##ctx)			\
5761 	FN(map_peek_elem, 89, ##ctx)			\
5762 	FN(msg_push_data, 90, ##ctx)			\
5763 	FN(msg_pop_data, 91, ##ctx)			\
5764 	FN(rc_pointer_rel, 92, ##ctx)			\
5765 	FN(spin_lock, 93, ##ctx)			\
5766 	FN(spin_unlock, 94, ##ctx)			\
5767 	FN(sk_fullsock, 95, ##ctx)			\
5768 	FN(tcp_sock, 96, ##ctx)				\
5769 	FN(skb_ecn_set_ce, 97, ##ctx)			\
5770 	FN(get_listener_sock, 98, ##ctx)		\
5771 	FN(skc_lookup_tcp, 99, ##ctx)			\
5772 	FN(tcp_check_syncookie, 100, ##ctx)		\
5773 	FN(sysctl_get_name, 101, ##ctx)			\
5774 	FN(sysctl_get_current_value, 102, ##ctx)	\
5775 	FN(sysctl_get_new_value, 103, ##ctx)		\
5776 	FN(sysctl_set_new_value, 104, ##ctx)		\
5777 	FN(strtol, 105, ##ctx)				\
5778 	FN(strtoul, 106, ##ctx)				\
5779 	FN(sk_storage_get, 107, ##ctx)			\
5780 	FN(sk_storage_delete, 108, ##ctx)		\
5781 	FN(send_signal, 109, ##ctx)			\
5782 	FN(tcp_gen_syncookie, 110, ##ctx)		\
5783 	FN(skb_output, 111, ##ctx)			\
5784 	FN(probe_read_user, 112, ##ctx)			\
5785 	FN(probe_read_kernel, 113, ##ctx)		\
5786 	FN(probe_read_user_str, 114, ##ctx)		\
5787 	FN(probe_read_kernel_str, 115, ##ctx)		\
5788 	FN(tcp_send_ack, 116, ##ctx)			\
5789 	FN(send_signal_thread, 117, ##ctx)		\
5790 	FN(jiffies64, 118, ##ctx)			\
5791 	FN(read_branch_records, 119, ##ctx)		\
5792 	FN(get_ns_current_pid_tgid, 120, ##ctx)		\
5793 	FN(xdp_output, 121, ##ctx)			\
5794 	FN(get_netns_cookie, 122, ##ctx)		\
5795 	FN(get_current_ancestor_cgroup_id, 123, ##ctx)	\
5796 	FN(sk_assign, 124, ##ctx)			\
5797 	FN(ktime_get_boot_ns, 125, ##ctx)		\
5798 	FN(seq_printf, 126, ##ctx)			\
5799 	FN(seq_write, 127, ##ctx)			\
5800 	FN(sk_cgroup_id, 128, ##ctx)			\
5801 	FN(sk_ancestor_cgroup_id, 129, ##ctx)		\
5802 	FN(ringbuf_output, 130, ##ctx)			\
5803 	FN(ringbuf_reserve, 131, ##ctx)			\
5804 	FN(ringbuf_submit, 132, ##ctx)			\
5805 	FN(ringbuf_discard, 133, ##ctx)			\
5806 	FN(ringbuf_query, 134, ##ctx)			\
5807 	FN(csum_level, 135, ##ctx)			\
5808 	FN(skc_to_tcp6_sock, 136, ##ctx)		\
5809 	FN(skc_to_tcp_sock, 137, ##ctx)			\
5810 	FN(skc_to_tcp_timewait_sock, 138, ##ctx)	\
5811 	FN(skc_to_tcp_request_sock, 139, ##ctx)		\
5812 	FN(skc_to_udp6_sock, 140, ##ctx)		\
5813 	FN(get_task_stack, 141, ##ctx)			\
5814 	FN(load_hdr_opt, 142, ##ctx)			\
5815 	FN(store_hdr_opt, 143, ##ctx)			\
5816 	FN(reserve_hdr_opt, 144, ##ctx)			\
5817 	FN(inode_storage_get, 145, ##ctx)		\
5818 	FN(inode_storage_delete, 146, ##ctx)		\
5819 	FN(d_path, 147, ##ctx)				\
5820 	FN(copy_from_user, 148, ##ctx)			\
5821 	FN(snprintf_btf, 149, ##ctx)			\
5822 	FN(seq_printf_btf, 150, ##ctx)			\
5823 	FN(skb_cgroup_classid, 151, ##ctx)		\
5824 	FN(redirect_neigh, 152, ##ctx)			\
5825 	FN(per_cpu_ptr, 153, ##ctx)			\
5826 	FN(this_cpu_ptr, 154, ##ctx)			\
5827 	FN(redirect_peer, 155, ##ctx)			\
5828 	FN(task_storage_get, 156, ##ctx)		\
5829 	FN(task_storage_delete, 157, ##ctx)		\
5830 	FN(get_current_task_btf, 158, ##ctx)		\
5831 	FN(bprm_opts_set, 159, ##ctx)			\
5832 	FN(ktime_get_coarse_ns, 160, ##ctx)		\
5833 	FN(ima_inode_hash, 161, ##ctx)			\
5834 	FN(sock_from_file, 162, ##ctx)			\
5835 	FN(check_mtu, 163, ##ctx)			\
5836 	FN(for_each_map_elem, 164, ##ctx)		\
5837 	FN(snprintf, 165, ##ctx)			\
5838 	FN(sys_bpf, 166, ##ctx)				\
5839 	FN(btf_find_by_name_kind, 167, ##ctx)		\
5840 	FN(sys_close, 168, ##ctx)			\
5841 	FN(timer_init, 169, ##ctx)			\
5842 	FN(timer_set_callback, 170, ##ctx)		\
5843 	FN(timer_start, 171, ##ctx)			\
5844 	FN(timer_cancel, 172, ##ctx)			\
5845 	FN(get_func_ip, 173, ##ctx)			\
5846 	FN(get_attach_cookie, 174, ##ctx)		\
5847 	FN(task_pt_regs, 175, ##ctx)			\
5848 	FN(get_branch_snapshot, 176, ##ctx)		\
5849 	FN(trace_vprintk, 177, ##ctx)			\
5850 	FN(skc_to_unix_sock, 178, ##ctx)		\
5851 	FN(kallsyms_lookup_name, 179, ##ctx)		\
5852 	FN(find_vma, 180, ##ctx)			\
5853 	FN(loop, 181, ##ctx)				\
5854 	FN(strncmp, 182, ##ctx)				\
5855 	FN(get_func_arg, 183, ##ctx)			\
5856 	FN(get_func_ret, 184, ##ctx)			\
5857 	FN(get_func_arg_cnt, 185, ##ctx)		\
5858 	FN(get_retval, 186, ##ctx)			\
5859 	FN(set_retval, 187, ##ctx)			\
5860 	FN(xdp_get_buff_len, 188, ##ctx)		\
5861 	FN(xdp_load_bytes, 189, ##ctx)			\
5862 	FN(xdp_store_bytes, 190, ##ctx)			\
5863 	FN(copy_from_user_task, 191, ##ctx)		\
5864 	FN(skb_set_tstamp, 192, ##ctx)			\
5865 	FN(ima_file_hash, 193, ##ctx)			\
5866 	FN(kptr_xchg, 194, ##ctx)			\
5867 	FN(map_lookup_percpu_elem, 195, ##ctx)		\
5868 	FN(skc_to_mptcp_sock, 196, ##ctx)		\
5869 	FN(dynptr_from_mem, 197, ##ctx)			\
5870 	FN(ringbuf_reserve_dynptr, 198, ##ctx)		\
5871 	FN(ringbuf_submit_dynptr, 199, ##ctx)		\
5872 	FN(ringbuf_discard_dynptr, 200, ##ctx)		\
5873 	FN(dynptr_read, 201, ##ctx)			\
5874 	FN(dynptr_write, 202, ##ctx)			\
5875 	FN(dynptr_data, 203, ##ctx)			\
5876 	FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx)	\
5877 	FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx)	\
5878 	FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx)	\
5879 	FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)	\
5880 	FN(ktime_get_tai_ns, 208, ##ctx)		\
5881 	FN(user_ringbuf_drain, 209, ##ctx)		\
5882 	FN(cgrp_storage_get, 210, ##ctx)		\
5883 	FN(cgrp_storage_delete, 211, ##ctx)		\
5884 	/* */
5885 
5886 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
5887  * know or care about integer value that is now passed as second argument
5888  */
5889 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
5890 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
5891 
5892 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5893  * function eBPF program intends to call
5894  */
5895 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
5896 enum bpf_func_id {
5897 	___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5898 	__BPF_FUNC_MAX_ID,
5899 };
5900 #undef __BPF_ENUM_FN
5901 
5902 /* All flags used by eBPF helper functions, placed here. */
5903 
5904 /* BPF_FUNC_skb_store_bytes flags. */
5905 enum {
5906 	BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),
5907 	BPF_F_INVALIDATE_HASH		= (1ULL << 1),
5908 };
5909 
5910 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5911  * First 4 bits are for passing the header field size.
5912  */
5913 enum {
5914 	BPF_F_HDR_FIELD_MASK		= 0xfULL,
5915 };
5916 
5917 /* BPF_FUNC_l4_csum_replace flags. */
5918 enum {
5919 	BPF_F_PSEUDO_HDR		= (1ULL << 4),
5920 	BPF_F_MARK_MANGLED_0		= (1ULL << 5),
5921 	BPF_F_MARK_ENFORCE		= (1ULL << 6),
5922 };
5923 
5924 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5925 enum {
5926 	BPF_F_INGRESS			= (1ULL << 0),
5927 };
5928 
5929 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5930 enum {
5931 	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
5932 };
5933 
5934 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5935 enum {
5936 	BPF_F_SKIP_FIELD_MASK		= 0xffULL,
5937 	BPF_F_USER_STACK		= (1ULL << 8),
5938 /* flags used by BPF_FUNC_get_stackid only. */
5939 	BPF_F_FAST_STACK_CMP		= (1ULL << 9),
5940 	BPF_F_REUSE_STACKID		= (1ULL << 10),
5941 /* flags used by BPF_FUNC_get_stack only. */
5942 	BPF_F_USER_BUILD_ID		= (1ULL << 11),
5943 };
5944 
5945 /* BPF_FUNC_skb_set_tunnel_key flags. */
5946 enum {
5947 	BPF_F_ZERO_CSUM_TX		= (1ULL << 1),
5948 	BPF_F_DONT_FRAGMENT		= (1ULL << 2),
5949 	BPF_F_SEQ_NUMBER		= (1ULL << 3),
5950 	BPF_F_NO_TUNNEL_KEY		= (1ULL << 4),
5951 };
5952 
5953 /* BPF_FUNC_skb_get_tunnel_key flags. */
5954 enum {
5955 	BPF_F_TUNINFO_FLAGS		= (1ULL << 4),
5956 };
5957 
5958 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5959  * BPF_FUNC_perf_event_read_value flags.
5960  */
5961 enum {
5962 	BPF_F_INDEX_MASK		= 0xffffffffULL,
5963 	BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,
5964 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5965 	BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),
5966 };
5967 
5968 /* Current network namespace */
5969 enum {
5970 	BPF_F_CURRENT_NETNS		= (-1L),
5971 };
5972 
5973 /* BPF_FUNC_csum_level level values. */
5974 enum {
5975 	BPF_CSUM_LEVEL_QUERY,
5976 	BPF_CSUM_LEVEL_INC,
5977 	BPF_CSUM_LEVEL_DEC,
5978 	BPF_CSUM_LEVEL_RESET,
5979 };
5980 
5981 /* BPF_FUNC_skb_adjust_room flags. */
5982 enum {
5983 	BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),
5984 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),
5985 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),
5986 	BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),
5987 	BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),
5988 	BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),
5989 	BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),
5990 	BPF_F_ADJ_ROOM_DECAP_L3_IPV4	= (1ULL << 7),
5991 	BPF_F_ADJ_ROOM_DECAP_L3_IPV6	= (1ULL << 8),
5992 };
5993 
5994 enum {
5995 	BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,
5996 	BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,
5997 };
5998 
5999 #define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \
6000 					  BPF_ADJ_ROOM_ENCAP_L2_MASK) \
6001 					 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
6002 
6003 /* BPF_FUNC_sysctl_get_name flags. */
6004 enum {
6005 	BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),
6006 };
6007 
6008 /* BPF_FUNC_<kernel_obj>_storage_get flags */
6009 enum {
6010 	BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),
6011 	/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
6012 	 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
6013 	 */
6014 	BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
6015 };
6016 
6017 /* BPF_FUNC_read_branch_records flags. */
6018 enum {
6019 	BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),
6020 };
6021 
6022 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
6023  * BPF_FUNC_bpf_ringbuf_output flags.
6024  */
6025 enum {
6026 	BPF_RB_NO_WAKEUP		= (1ULL << 0),
6027 	BPF_RB_FORCE_WAKEUP		= (1ULL << 1),
6028 };
6029 
6030 /* BPF_FUNC_bpf_ringbuf_query flags */
6031 enum {
6032 	BPF_RB_AVAIL_DATA = 0,
6033 	BPF_RB_RING_SIZE = 1,
6034 	BPF_RB_CONS_POS = 2,
6035 	BPF_RB_PROD_POS = 3,
6036 };
6037 
6038 /* BPF ring buffer constants */
6039 enum {
6040 	BPF_RINGBUF_BUSY_BIT		= (1U << 31),
6041 	BPF_RINGBUF_DISCARD_BIT		= (1U << 30),
6042 	BPF_RINGBUF_HDR_SZ		= 8,
6043 };
6044 
6045 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
6046 enum {
6047 	BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),
6048 	BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),
6049 };
6050 
6051 /* Mode for BPF_FUNC_skb_adjust_room helper. */
6052 enum bpf_adj_room_mode {
6053 	BPF_ADJ_ROOM_NET,
6054 	BPF_ADJ_ROOM_MAC,
6055 };
6056 
6057 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
6058 enum bpf_hdr_start_off {
6059 	BPF_HDR_START_MAC,
6060 	BPF_HDR_START_NET,
6061 };
6062 
6063 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
6064 enum bpf_lwt_encap_mode {
6065 	BPF_LWT_ENCAP_SEG6,
6066 	BPF_LWT_ENCAP_SEG6_INLINE,
6067 	BPF_LWT_ENCAP_IP,
6068 };
6069 
6070 /* Flags for bpf_bprm_opts_set helper */
6071 enum {
6072 	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
6073 };
6074 
6075 /* Flags for bpf_redirect_map helper */
6076 enum {
6077 	BPF_F_BROADCAST		= (1ULL << 3),
6078 	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
6079 };
6080 
6081 #define __bpf_md_ptr(type, name)	\
6082 union {					\
6083 	type name;			\
6084 	__u64 :64;			\
6085 } __attribute__((aligned(8)))
6086 
6087 enum {
6088 	BPF_SKB_TSTAMP_UNSPEC,
6089 	BPF_SKB_TSTAMP_DELIVERY_MONO,	/* tstamp has mono delivery time */
6090 	/* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
6091 	 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
6092 	 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
6093 	 */
6094 };
6095 
6096 /* user accessible mirror of in-kernel sk_buff.
6097  * new fields can only be added to the end of this structure
6098  */
6099 struct __sk_buff {
6100 	__u32 len;
6101 	__u32 pkt_type;
6102 	__u32 mark;
6103 	__u32 queue_mapping;
6104 	__u32 protocol;
6105 	__u32 vlan_present;
6106 	__u32 vlan_tci;
6107 	__u32 vlan_proto;
6108 	__u32 priority;
6109 	__u32 ingress_ifindex;
6110 	__u32 ifindex;
6111 	__u32 tc_index;
6112 	__u32 cb[5];
6113 	__u32 hash;
6114 	__u32 tc_classid;
6115 	__u32 data;
6116 	__u32 data_end;
6117 	__u32 napi_id;
6118 
6119 	/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
6120 	__u32 family;
6121 	__u32 remote_ip4;	/* Stored in network byte order */
6122 	__u32 local_ip4;	/* Stored in network byte order */
6123 	__u32 remote_ip6[4];	/* Stored in network byte order */
6124 	__u32 local_ip6[4];	/* Stored in network byte order */
6125 	__u32 remote_port;	/* Stored in network byte order */
6126 	__u32 local_port;	/* stored in host byte order */
6127 	/* ... here. */
6128 
6129 	__u32 data_meta;
6130 	__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
6131 	__u64 tstamp;
6132 	__u32 wire_len;
6133 	__u32 gso_segs;
6134 	__bpf_md_ptr(struct bpf_sock *, sk);
6135 	__u32 gso_size;
6136 	__u8  tstamp_type;
6137 	__u32 :24;		/* Padding, future use. */
6138 	__u64 hwtstamp;
6139 };
6140 
6141 struct bpf_tunnel_key {
6142 	__u32 tunnel_id;
6143 	union {
6144 		__u32 remote_ipv4;
6145 		__u32 remote_ipv6[4];
6146 	};
6147 	__u8 tunnel_tos;
6148 	__u8 tunnel_ttl;
6149 	union {
6150 		__u16 tunnel_ext;	/* compat */
6151 		__be16 tunnel_flags;
6152 	};
6153 	__u32 tunnel_label;
6154 	union {
6155 		__u32 local_ipv4;
6156 		__u32 local_ipv6[4];
6157 	};
6158 };
6159 
6160 /* user accessible mirror of in-kernel xfrm_state.
6161  * new fields can only be added to the end of this structure
6162  */
6163 struct bpf_xfrm_state {
6164 	__u32 reqid;
6165 	__u32 spi;	/* Stored in network byte order */
6166 	__u16 family;
6167 	__u16 ext;	/* Padding, future use. */
6168 	union {
6169 		__u32 remote_ipv4;	/* Stored in network byte order */
6170 		__u32 remote_ipv6[4];	/* Stored in network byte order */
6171 	};
6172 };
6173 
6174 /* Generic BPF return codes which all BPF program types may support.
6175  * The values are binary compatible with their TC_ACT_* counter-part to
6176  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
6177  * programs.
6178  *
6179  * XDP is handled seprately, see XDP_*.
6180  */
6181 enum bpf_ret_code {
6182 	BPF_OK = 0,
6183 	/* 1 reserved */
6184 	BPF_DROP = 2,
6185 	/* 3-6 reserved */
6186 	BPF_REDIRECT = 7,
6187 	/* >127 are reserved for prog type specific return codes.
6188 	 *
6189 	 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6190 	 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6191 	 *    changed and should be routed based on its new L3 header.
6192 	 *    (This is an L3 redirect, as opposed to L2 redirect
6193 	 *    represented by BPF_REDIRECT above).
6194 	 */
6195 	BPF_LWT_REROUTE = 128,
6196 	/* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6197 	 *   to indicate that no custom dissection was performed, and
6198 	 *   fallback to standard dissector is requested.
6199 	 */
6200 	BPF_FLOW_DISSECTOR_CONTINUE = 129,
6201 };
6202 
6203 struct bpf_sock {
6204 	__u32 bound_dev_if;
6205 	__u32 family;
6206 	__u32 type;
6207 	__u32 protocol;
6208 	__u32 mark;
6209 	__u32 priority;
6210 	/* IP address also allows 1 and 2 bytes access */
6211 	__u32 src_ip4;
6212 	__u32 src_ip6[4];
6213 	__u32 src_port;		/* host byte order */
6214 	__be16 dst_port;	/* network byte order */
6215 	__u16 :16;		/* zero padding */
6216 	__u32 dst_ip4;
6217 	__u32 dst_ip6[4];
6218 	__u32 state;
6219 	__s32 rx_queue_mapping;
6220 };
6221 
6222 struct bpf_tcp_sock {
6223 	__u32 snd_cwnd;		/* Sending congestion window		*/
6224 	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
6225 	__u32 rtt_min;
6226 	__u32 snd_ssthresh;	/* Slow start size threshold		*/
6227 	__u32 rcv_nxt;		/* What we want to receive next		*/
6228 	__u32 snd_nxt;		/* Next sequence we send		*/
6229 	__u32 snd_una;		/* First byte we want an ack for	*/
6230 	__u32 mss_cache;	/* Cached effective mss, not including SACKS */
6231 	__u32 ecn_flags;	/* ECN status bits.			*/
6232 	__u32 rate_delivered;	/* saved rate sample: packets delivered */
6233 	__u32 rate_interval_us;	/* saved rate sample: time elapsed */
6234 	__u32 packets_out;	/* Packets which are "in flight"	*/
6235 	__u32 retrans_out;	/* Retransmitted packets out		*/
6236 	__u32 total_retrans;	/* Total retransmits for entire connection */
6237 	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
6238 				 * total number of segments in.
6239 				 */
6240 	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
6241 				 * total number of data segments in.
6242 				 */
6243 	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
6244 				 * The total number of segments sent.
6245 				 */
6246 	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
6247 				 * total number of data segments sent.
6248 				 */
6249 	__u32 lost_out;		/* Lost packets			*/
6250 	__u32 sacked_out;	/* SACK'd packets			*/
6251 	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
6252 				 * sum(delta(rcv_nxt)), or how many bytes
6253 				 * were acked.
6254 				 */
6255 	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
6256 				 * sum(delta(snd_una)), or how many bytes
6257 				 * were acked.
6258 				 */
6259 	__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups
6260 				 * total number of DSACK blocks received
6261 				 */
6262 	__u32 delivered;	/* Total data packets delivered incl. rexmits */
6263 	__u32 delivered_ce;	/* Like the above but only ECE marked packets */
6264 	__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */
6265 };
6266 
6267 struct bpf_sock_tuple {
6268 	union {
6269 		struct {
6270 			__be32 saddr;
6271 			__be32 daddr;
6272 			__be16 sport;
6273 			__be16 dport;
6274 		} ipv4;
6275 		struct {
6276 			__be32 saddr[4];
6277 			__be32 daddr[4];
6278 			__be16 sport;
6279 			__be16 dport;
6280 		} ipv6;
6281 	};
6282 };
6283 
6284 /* (Simplified) user return codes for tcx prog type.
6285  * A valid tcx program must return one of these defined values. All other
6286  * return codes are reserved for future use. Must remain compatible with
6287  * their TC_ACT_* counter-parts. For compatibility in behavior, unknown
6288  * return codes are mapped to TCX_NEXT.
6289  */
6290 enum tcx_action_base {
6291 	TCX_NEXT	= -1,
6292 	TCX_PASS	= 0,
6293 	TCX_DROP	= 2,
6294 	TCX_REDIRECT	= 7,
6295 };
6296 
6297 struct bpf_xdp_sock {
6298 	__u32 queue_id;
6299 };
6300 
6301 #define XDP_PACKET_HEADROOM 256
6302 
6303 /* User return codes for XDP prog type.
6304  * A valid XDP program must return one of these defined values. All other
6305  * return codes are reserved for future use. Unknown return codes will
6306  * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6307  */
6308 enum xdp_action {
6309 	XDP_ABORTED = 0,
6310 	XDP_DROP,
6311 	XDP_PASS,
6312 	XDP_TX,
6313 	XDP_REDIRECT,
6314 };
6315 
6316 /* user accessible metadata for XDP packet hook
6317  * new fields must be added to the end of this structure
6318  */
6319 struct xdp_md {
6320 	__u32 data;
6321 	__u32 data_end;
6322 	__u32 data_meta;
6323 	/* Below access go through struct xdp_rxq_info */
6324 	__u32 ingress_ifindex; /* rxq->dev->ifindex */
6325 	__u32 rx_queue_index;  /* rxq->queue_index  */
6326 
6327 	__u32 egress_ifindex;  /* txq->dev->ifindex */
6328 };
6329 
6330 /* DEVMAP map-value layout
6331  *
6332  * The struct data-layout of map-value is a configuration interface.
6333  * New members can only be added to the end of this structure.
6334  */
6335 struct bpf_devmap_val {
6336 	__u32 ifindex;   /* device index */
6337 	union {
6338 		int   fd;  /* prog fd on map write */
6339 		__u32 id;  /* prog id on map read */
6340 	} bpf_prog;
6341 };
6342 
6343 /* CPUMAP map-value layout
6344  *
6345  * The struct data-layout of map-value is a configuration interface.
6346  * New members can only be added to the end of this structure.
6347  */
6348 struct bpf_cpumap_val {
6349 	__u32 qsize;	/* queue size to remote target CPU */
6350 	union {
6351 		int   fd;	/* prog fd on map write */
6352 		__u32 id;	/* prog id on map read */
6353 	} bpf_prog;
6354 };
6355 
6356 enum sk_action {
6357 	SK_DROP = 0,
6358 	SK_PASS,
6359 };
6360 
6361 /* user accessible metadata for SK_MSG packet hook, new fields must
6362  * be added to the end of this structure
6363  */
6364 struct sk_msg_md {
6365 	__bpf_md_ptr(void *, data);
6366 	__bpf_md_ptr(void *, data_end);
6367 
6368 	__u32 family;
6369 	__u32 remote_ip4;	/* Stored in network byte order */
6370 	__u32 local_ip4;	/* Stored in network byte order */
6371 	__u32 remote_ip6[4];	/* Stored in network byte order */
6372 	__u32 local_ip6[4];	/* Stored in network byte order */
6373 	__u32 remote_port;	/* Stored in network byte order */
6374 	__u32 local_port;	/* stored in host byte order */
6375 	__u32 size;		/* Total size of sk_msg */
6376 
6377 	__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6378 };
6379 
6380 struct sk_reuseport_md {
6381 	/*
6382 	 * Start of directly accessible data. It begins from
6383 	 * the tcp/udp header.
6384 	 */
6385 	__bpf_md_ptr(void *, data);
6386 	/* End of directly accessible data */
6387 	__bpf_md_ptr(void *, data_end);
6388 	/*
6389 	 * Total length of packet (starting from the tcp/udp header).
6390 	 * Note that the directly accessible bytes (data_end - data)
6391 	 * could be less than this "len".  Those bytes could be
6392 	 * indirectly read by a helper "bpf_skb_load_bytes()".
6393 	 */
6394 	__u32 len;
6395 	/*
6396 	 * Eth protocol in the mac header (network byte order). e.g.
6397 	 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6398 	 */
6399 	__u32 eth_protocol;
6400 	__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6401 	__u32 bind_inany;	/* Is sock bound to an INANY address? */
6402 	__u32 hash;		/* A hash of the packet 4 tuples */
6403 	/* When reuse->migrating_sk is NULL, it is selecting a sk for the
6404 	 * new incoming connection request (e.g. selecting a listen sk for
6405 	 * the received SYN in the TCP case).  reuse->sk is one of the sk
6406 	 * in the reuseport group. The bpf prog can use reuse->sk to learn
6407 	 * the local listening ip/port without looking into the skb.
6408 	 *
6409 	 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6410 	 * reuse->migrating_sk is the socket that needs to be migrated
6411 	 * to another listening socket.  migrating_sk could be a fullsock
6412 	 * sk that is fully established or a reqsk that is in-the-middle
6413 	 * of 3-way handshake.
6414 	 */
6415 	__bpf_md_ptr(struct bpf_sock *, sk);
6416 	__bpf_md_ptr(struct bpf_sock *, migrating_sk);
6417 };
6418 
6419 #define BPF_TAG_SIZE	8
6420 
6421 struct bpf_prog_info {
6422 	__u32 type;
6423 	__u32 id;
6424 	__u8  tag[BPF_TAG_SIZE];
6425 	__u32 jited_prog_len;
6426 	__u32 xlated_prog_len;
6427 	__aligned_u64 jited_prog_insns;
6428 	__aligned_u64 xlated_prog_insns;
6429 	__u64 load_time;	/* ns since boottime */
6430 	__u32 created_by_uid;
6431 	__u32 nr_map_ids;
6432 	__aligned_u64 map_ids;
6433 	char name[BPF_OBJ_NAME_LEN];
6434 	__u32 ifindex;
6435 	__u32 gpl_compatible:1;
6436 	__u32 :31; /* alignment pad */
6437 	__u64 netns_dev;
6438 	__u64 netns_ino;
6439 	__u32 nr_jited_ksyms;
6440 	__u32 nr_jited_func_lens;
6441 	__aligned_u64 jited_ksyms;
6442 	__aligned_u64 jited_func_lens;
6443 	__u32 btf_id;
6444 	__u32 func_info_rec_size;
6445 	__aligned_u64 func_info;
6446 	__u32 nr_func_info;
6447 	__u32 nr_line_info;
6448 	__aligned_u64 line_info;
6449 	__aligned_u64 jited_line_info;
6450 	__u32 nr_jited_line_info;
6451 	__u32 line_info_rec_size;
6452 	__u32 jited_line_info_rec_size;
6453 	__u32 nr_prog_tags;
6454 	__aligned_u64 prog_tags;
6455 	__u64 run_time_ns;
6456 	__u64 run_cnt;
6457 	__u64 recursion_misses;
6458 	__u32 verified_insns;
6459 	__u32 attach_btf_obj_id;
6460 	__u32 attach_btf_id;
6461 } __attribute__((aligned(8)));
6462 
6463 struct bpf_map_info {
6464 	__u32 type;
6465 	__u32 id;
6466 	__u32 key_size;
6467 	__u32 value_size;
6468 	__u32 max_entries;
6469 	__u32 map_flags;
6470 	char  name[BPF_OBJ_NAME_LEN];
6471 	__u32 ifindex;
6472 	__u32 btf_vmlinux_value_type_id;
6473 	__u64 netns_dev;
6474 	__u64 netns_ino;
6475 	__u32 btf_id;
6476 	__u32 btf_key_type_id;
6477 	__u32 btf_value_type_id;
6478 	__u32 :32;	/* alignment pad */
6479 	__u64 map_extra;
6480 } __attribute__((aligned(8)));
6481 
6482 struct bpf_btf_info {
6483 	__aligned_u64 btf;
6484 	__u32 btf_size;
6485 	__u32 id;
6486 	__aligned_u64 name;
6487 	__u32 name_len;
6488 	__u32 kernel_btf;
6489 } __attribute__((aligned(8)));
6490 
6491 struct bpf_link_info {
6492 	__u32 type;
6493 	__u32 id;
6494 	__u32 prog_id;
6495 	union {
6496 		struct {
6497 			__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6498 			__u32 tp_name_len;     /* in/out: tp_name buffer len */
6499 		} raw_tracepoint;
6500 		struct {
6501 			__u32 attach_type;
6502 			__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6503 			__u32 target_btf_id; /* BTF type id inside the object */
6504 		} tracing;
6505 		struct {
6506 			__u64 cgroup_id;
6507 			__u32 attach_type;
6508 		} cgroup;
6509 		struct {
6510 			__aligned_u64 target_name; /* in/out: target_name buffer ptr */
6511 			__u32 target_name_len;	   /* in/out: target_name buffer len */
6512 
6513 			/* If the iter specific field is 32 bits, it can be put
6514 			 * in the first or second union. Otherwise it should be
6515 			 * put in the second union.
6516 			 */
6517 			union {
6518 				struct {
6519 					__u32 map_id;
6520 				} map;
6521 			};
6522 			union {
6523 				struct {
6524 					__u64 cgroup_id;
6525 					__u32 order;
6526 				} cgroup;
6527 				struct {
6528 					__u32 tid;
6529 					__u32 pid;
6530 				} task;
6531 			};
6532 		} iter;
6533 		struct  {
6534 			__u32 netns_ino;
6535 			__u32 attach_type;
6536 		} netns;
6537 		struct {
6538 			__u32 ifindex;
6539 		} xdp;
6540 		struct {
6541 			__u32 map_id;
6542 		} struct_ops;
6543 		struct {
6544 			__u32 pf;
6545 			__u32 hooknum;
6546 			__s32 priority;
6547 			__u32 flags;
6548 		} netfilter;
6549 		struct {
6550 			__aligned_u64 addrs;
6551 			__u32 count; /* in/out: kprobe_multi function count */
6552 			__u32 flags;
6553 		} kprobe_multi;
6554 		struct {
6555 			__u32 type; /* enum bpf_perf_event_type */
6556 			__u32 :32;
6557 			union {
6558 				struct {
6559 					__aligned_u64 file_name; /* in/out */
6560 					__u32 name_len;
6561 					__u32 offset; /* offset from file_name */
6562 				} uprobe; /* BPF_PERF_EVENT_UPROBE, BPF_PERF_EVENT_URETPROBE */
6563 				struct {
6564 					__aligned_u64 func_name; /* in/out */
6565 					__u32 name_len;
6566 					__u32 offset; /* offset from func_name */
6567 					__u64 addr;
6568 				} kprobe; /* BPF_PERF_EVENT_KPROBE, BPF_PERF_EVENT_KRETPROBE */
6569 				struct {
6570 					__aligned_u64 tp_name;   /* in/out */
6571 					__u32 name_len;
6572 				} tracepoint; /* BPF_PERF_EVENT_TRACEPOINT */
6573 				struct {
6574 					__u64 config;
6575 					__u32 type;
6576 				} event; /* BPF_PERF_EVENT_EVENT */
6577 			};
6578 		} perf_event;
6579 		struct {
6580 			__u32 ifindex;
6581 			__u32 attach_type;
6582 		} tcx;
6583 	};
6584 } __attribute__((aligned(8)));
6585 
6586 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6587  * by user and intended to be used by socket (e.g. to bind to, depends on
6588  * attach type).
6589  */
6590 struct bpf_sock_addr {
6591 	__u32 user_family;	/* Allows 4-byte read, but no write. */
6592 	__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.
6593 				 * Stored in network byte order.
6594 				 */
6595 	__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6596 				 * Stored in network byte order.
6597 				 */
6598 	__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.
6599 				 * Stored in network byte order
6600 				 */
6601 	__u32 family;		/* Allows 4-byte read, but no write */
6602 	__u32 type;		/* Allows 4-byte read, but no write */
6603 	__u32 protocol;		/* Allows 4-byte read, but no write */
6604 	__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.
6605 				 * Stored in network byte order.
6606 				 */
6607 	__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6608 				 * Stored in network byte order.
6609 				 */
6610 	__bpf_md_ptr(struct bpf_sock *, sk);
6611 };
6612 
6613 /* User bpf_sock_ops struct to access socket values and specify request ops
6614  * and their replies.
6615  * Some of this fields are in network (bigendian) byte order and may need
6616  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6617  * New fields can only be added at the end of this structure
6618  */
6619 struct bpf_sock_ops {
6620 	__u32 op;
6621 	union {
6622 		__u32 args[4];		/* Optionally passed to bpf program */
6623 		__u32 reply;		/* Returned by bpf program	    */
6624 		__u32 replylong[4];	/* Optionally returned by bpf prog  */
6625 	};
6626 	__u32 family;
6627 	__u32 remote_ip4;	/* Stored in network byte order */
6628 	__u32 local_ip4;	/* Stored in network byte order */
6629 	__u32 remote_ip6[4];	/* Stored in network byte order */
6630 	__u32 local_ip6[4];	/* Stored in network byte order */
6631 	__u32 remote_port;	/* Stored in network byte order */
6632 	__u32 local_port;	/* stored in host byte order */
6633 	__u32 is_fullsock;	/* Some TCP fields are only valid if
6634 				 * there is a full socket. If not, the
6635 				 * fields read as zero.
6636 				 */
6637 	__u32 snd_cwnd;
6638 	__u32 srtt_us;		/* Averaged RTT << 3 in usecs */
6639 	__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6640 	__u32 state;
6641 	__u32 rtt_min;
6642 	__u32 snd_ssthresh;
6643 	__u32 rcv_nxt;
6644 	__u32 snd_nxt;
6645 	__u32 snd_una;
6646 	__u32 mss_cache;
6647 	__u32 ecn_flags;
6648 	__u32 rate_delivered;
6649 	__u32 rate_interval_us;
6650 	__u32 packets_out;
6651 	__u32 retrans_out;
6652 	__u32 total_retrans;
6653 	__u32 segs_in;
6654 	__u32 data_segs_in;
6655 	__u32 segs_out;
6656 	__u32 data_segs_out;
6657 	__u32 lost_out;
6658 	__u32 sacked_out;
6659 	__u32 sk_txhash;
6660 	__u64 bytes_received;
6661 	__u64 bytes_acked;
6662 	__bpf_md_ptr(struct bpf_sock *, sk);
6663 	/* [skb_data, skb_data_end) covers the whole TCP header.
6664 	 *
6665 	 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6666 	 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
6667 	 *                                header has not been written.
6668 	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6669 	 *				  been written so far.
6670 	 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
6671 	 *					the 3WHS.
6672 	 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6673 	 *					the 3WHS.
6674 	 *
6675 	 * bpf_load_hdr_opt() can also be used to read a particular option.
6676 	 */
6677 	__bpf_md_ptr(void *, skb_data);
6678 	__bpf_md_ptr(void *, skb_data_end);
6679 	__u32 skb_len;		/* The total length of a packet.
6680 				 * It includes the header, options,
6681 				 * and payload.
6682 				 */
6683 	__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides
6684 				 * an easy way to check for tcp_flags
6685 				 * without parsing skb_data.
6686 				 *
6687 				 * In particular, the skb_tcp_flags
6688 				 * will still be available in
6689 				 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6690 				 * the outgoing header has not
6691 				 * been written yet.
6692 				 */
6693 	__u64 skb_hwtstamp;
6694 };
6695 
6696 /* Definitions for bpf_sock_ops_cb_flags */
6697 enum {
6698 	BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),
6699 	BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),
6700 	BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),
6701 	BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),
6702 	/* Call bpf for all received TCP headers.  The bpf prog will be
6703 	 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6704 	 *
6705 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6706 	 * for the header option related helpers that will be useful
6707 	 * to the bpf programs.
6708 	 *
6709 	 * It could be used at the client/active side (i.e. connect() side)
6710 	 * when the server told it that the server was in syncookie
6711 	 * mode and required the active side to resend the bpf-written
6712 	 * options.  The active side can keep writing the bpf-options until
6713 	 * it received a valid packet from the server side to confirm
6714 	 * the earlier packet (and options) has been received.  The later
6715 	 * example patch is using it like this at the active side when the
6716 	 * server is in syncookie mode.
6717 	 *
6718 	 * The bpf prog will usually turn this off in the common cases.
6719 	 */
6720 	BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),
6721 	/* Call bpf when kernel has received a header option that
6722 	 * the kernel cannot handle.  The bpf prog will be called under
6723 	 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6724 	 *
6725 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6726 	 * for the header option related helpers that will be useful
6727 	 * to the bpf programs.
6728 	 */
6729 	BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6730 	/* Call bpf when the kernel is writing header options for the
6731 	 * outgoing packet.  The bpf prog will first be called
6732 	 * to reserve space in a skb under
6733 	 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
6734 	 * the bpf prog will be called to write the header option(s)
6735 	 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6736 	 *
6737 	 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6738 	 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6739 	 * related helpers that will be useful to the bpf programs.
6740 	 *
6741 	 * The kernel gets its chance to reserve space and write
6742 	 * options first before the BPF program does.
6743 	 */
6744 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6745 /* Mask of all currently supported cb flags */
6746 	BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
6747 };
6748 
6749 /* List of known BPF sock_ops operators.
6750  * New entries can only be added at the end
6751  */
6752 enum {
6753 	BPF_SOCK_OPS_VOID,
6754 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
6755 					 * -1 if default value should be used
6756 					 */
6757 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
6758 					 * window (in packets) or -1 if default
6759 					 * value should be used
6760 					 */
6761 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
6762 					 * active connection is initialized
6763 					 */
6764 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
6765 						 * active connection is
6766 						 * established
6767 						 */
6768 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
6769 						 * passive connection is
6770 						 * established
6771 						 */
6772 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
6773 					 * needs ECN
6774 					 */
6775 	BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is
6776 					 * based on the path and may be
6777 					 * dependent on the congestion control
6778 					 * algorithm. In general it indicates
6779 					 * a congestion threshold. RTTs above
6780 					 * this indicate congestion
6781 					 */
6782 	BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.
6783 					 * Arg1: value of icsk_retransmits
6784 					 * Arg2: value of icsk_rto
6785 					 * Arg3: whether RTO has expired
6786 					 */
6787 	BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.
6788 					 * Arg1: sequence number of 1st byte
6789 					 * Arg2: # segments
6790 					 * Arg3: return value of
6791 					 *       tcp_transmit_skb (0 => success)
6792 					 */
6793 	BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.
6794 					 * Arg1: old_state
6795 					 * Arg2: new_state
6796 					 */
6797 	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
6798 					 * socket transition to LISTEN state.
6799 					 */
6800 	BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.
6801 					 */
6802 	BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.
6803 					 * It will be called to handle
6804 					 * the packets received at
6805 					 * an already established
6806 					 * connection.
6807 					 *
6808 					 * sock_ops->skb_data:
6809 					 * Referring to the received skb.
6810 					 * It covers the TCP header only.
6811 					 *
6812 					 * bpf_load_hdr_opt() can also
6813 					 * be used to search for a
6814 					 * particular option.
6815 					 */
6816 	BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the
6817 					 * header option later in
6818 					 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6819 					 * Arg1: bool want_cookie. (in
6820 					 *       writing SYNACK only)
6821 					 *
6822 					 * sock_ops->skb_data:
6823 					 * Not available because no header has
6824 					 * been	written yet.
6825 					 *
6826 					 * sock_ops->skb_tcp_flags:
6827 					 * The tcp_flags of the
6828 					 * outgoing skb. (e.g. SYN, ACK, FIN).
6829 					 *
6830 					 * bpf_reserve_hdr_opt() should
6831 					 * be used to reserve space.
6832 					 */
6833 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options
6834 					 * Arg1: bool want_cookie. (in
6835 					 *       writing SYNACK only)
6836 					 *
6837 					 * sock_ops->skb_data:
6838 					 * Referring to the outgoing skb.
6839 					 * It covers the TCP header
6840 					 * that has already been written
6841 					 * by the kernel and the
6842 					 * earlier bpf-progs.
6843 					 *
6844 					 * sock_ops->skb_tcp_flags:
6845 					 * The tcp_flags of the outgoing
6846 					 * skb. (e.g. SYN, ACK, FIN).
6847 					 *
6848 					 * bpf_store_hdr_opt() should
6849 					 * be used to write the
6850 					 * option.
6851 					 *
6852 					 * bpf_load_hdr_opt() can also
6853 					 * be used to search for a
6854 					 * particular option that
6855 					 * has already been written
6856 					 * by the kernel or the
6857 					 * earlier bpf-progs.
6858 					 */
6859 };
6860 
6861 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
6862  * changes between the TCP and BPF versions. Ideally this should never happen.
6863  * If it does, we need to add code to convert them before calling
6864  * the BPF sock_ops function.
6865  */
6866 enum {
6867 	BPF_TCP_ESTABLISHED = 1,
6868 	BPF_TCP_SYN_SENT,
6869 	BPF_TCP_SYN_RECV,
6870 	BPF_TCP_FIN_WAIT1,
6871 	BPF_TCP_FIN_WAIT2,
6872 	BPF_TCP_TIME_WAIT,
6873 	BPF_TCP_CLOSE,
6874 	BPF_TCP_CLOSE_WAIT,
6875 	BPF_TCP_LAST_ACK,
6876 	BPF_TCP_LISTEN,
6877 	BPF_TCP_CLOSING,	/* Now a valid state */
6878 	BPF_TCP_NEW_SYN_RECV,
6879 
6880 	BPF_TCP_MAX_STATES	/* Leave at the end! */
6881 };
6882 
6883 enum {
6884 	TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */
6885 	TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */
6886 	TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */
6887 	TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */
6888 	/* Copy the SYN pkt to optval
6889 	 *
6890 	 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
6891 	 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6892 	 * to only getting from the saved_syn.  It can either get the
6893 	 * syn packet from:
6894 	 *
6895 	 * 1. the just-received SYN packet (only available when writing the
6896 	 *    SYNACK).  It will be useful when it is not necessary to
6897 	 *    save the SYN packet for latter use.  It is also the only way
6898 	 *    to get the SYN during syncookie mode because the syn
6899 	 *    packet cannot be saved during syncookie.
6900 	 *
6901 	 * OR
6902 	 *
6903 	 * 2. the earlier saved syn which was done by
6904 	 *    bpf_setsockopt(TCP_SAVE_SYN).
6905 	 *
6906 	 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6907 	 * SYN packet is obtained.
6908 	 *
6909 	 * If the bpf-prog does not need the IP[46] header,  the
6910 	 * bpf-prog can avoid parsing the IP header by using
6911 	 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
6912 	 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6913 	 *
6914 	 *      >0: Total number of bytes copied
6915 	 * -ENOSPC: Not enough space in optval. Only optlen number of
6916 	 *          bytes is copied.
6917 	 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6918 	 *	    is not saved by setsockopt(TCP_SAVE_SYN).
6919 	 */
6920 	TCP_BPF_SYN		= 1005, /* Copy the TCP header */
6921 	TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */
6922 	TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
6923 };
6924 
6925 enum {
6926 	BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6927 };
6928 
6929 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6930  * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6931  */
6932 enum {
6933 	BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the
6934 						 * total option spaces
6935 						 * required for an established
6936 						 * sk in order to calculate the
6937 						 * MSS.  No skb is actually
6938 						 * sent.
6939 						 */
6940 	BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode
6941 						 * when sending a SYN.
6942 						 */
6943 };
6944 
6945 struct bpf_perf_event_value {
6946 	__u64 counter;
6947 	__u64 enabled;
6948 	__u64 running;
6949 };
6950 
6951 enum {
6952 	BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),
6953 	BPF_DEVCG_ACC_READ	= (1ULL << 1),
6954 	BPF_DEVCG_ACC_WRITE	= (1ULL << 2),
6955 };
6956 
6957 enum {
6958 	BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),
6959 	BPF_DEVCG_DEV_CHAR	= (1ULL << 1),
6960 };
6961 
6962 struct bpf_cgroup_dev_ctx {
6963 	/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6964 	__u32 access_type;
6965 	__u32 major;
6966 	__u32 minor;
6967 };
6968 
6969 struct bpf_raw_tracepoint_args {
6970 	__u64 args[0];
6971 };
6972 
6973 /* DIRECT:  Skip the FIB rules and go to FIB table associated with device
6974  * OUTPUT:  Do lookup from egress perspective; default is ingress
6975  */
6976 enum {
6977 	BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
6978 	BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
6979 	BPF_FIB_LOOKUP_SKIP_NEIGH = (1U << 2),
6980 	BPF_FIB_LOOKUP_TBID    = (1U << 3),
6981 	BPF_FIB_LOOKUP_SRC     = (1U << 4),
6982 };
6983 
6984 enum {
6985 	BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
6986 	BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
6987 	BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
6988 	BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
6989 	BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
6990 	BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6991 	BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
6992 	BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
6993 	BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6994 	BPF_FIB_LKUP_RET_NO_SRC_ADDR,  /* failed to derive IP src addr */
6995 };
6996 
6997 struct bpf_fib_lookup {
6998 	/* input:  network family for lookup (AF_INET, AF_INET6)
6999 	 * output: network family of egress nexthop
7000 	 */
7001 	__u8	family;
7002 
7003 	/* set if lookup is to consider L4 data - e.g., FIB rules */
7004 	__u8	l4_protocol;
7005 	__be16	sport;
7006 	__be16	dport;
7007 
7008 	union {	/* used for MTU check */
7009 		/* input to lookup */
7010 		__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */
7011 
7012 		/* output: MTU value */
7013 		__u16	mtu_result;
7014 	} __attribute__((packed, aligned(2)));
7015 	/* input: L3 device index for lookup
7016 	 * output: device index from FIB lookup
7017 	 */
7018 	__u32	ifindex;
7019 
7020 	union {
7021 		/* inputs to lookup */
7022 		__u8	tos;		/* AF_INET  */
7023 		__be32	flowinfo;	/* AF_INET6, flow_label + priority */
7024 
7025 		/* output: metric of fib result (IPv4/IPv6 only) */
7026 		__u32	rt_metric;
7027 	};
7028 
7029 	/* input: source address to consider for lookup
7030 	 * output: source address result from lookup
7031 	 */
7032 	union {
7033 		__be32		ipv4_src;
7034 		__u32		ipv6_src[4];  /* in6_addr; network order */
7035 	};
7036 
7037 	/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
7038 	 * network header. output: bpf_fib_lookup sets to gateway address
7039 	 * if FIB lookup returns gateway route
7040 	 */
7041 	union {
7042 		__be32		ipv4_dst;
7043 		__u32		ipv6_dst[4];  /* in6_addr; network order */
7044 	};
7045 
7046 	union {
7047 		struct {
7048 			/* output */
7049 			__be16	h_vlan_proto;
7050 			__be16	h_vlan_TCI;
7051 		};
7052 		/* input: when accompanied with the
7053 		 * 'BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_TBID` flags, a
7054 		 * specific routing table to use for the fib lookup.
7055 		 */
7056 		__u32	tbid;
7057 	};
7058 
7059 	__u8	smac[6];     /* ETH_ALEN */
7060 	__u8	dmac[6];     /* ETH_ALEN */
7061 };
7062 
7063 struct bpf_redir_neigh {
7064 	/* network family for lookup (AF_INET, AF_INET6) */
7065 	__u32 nh_family;
7066 	/* network address of nexthop; skips fib lookup to find gateway */
7067 	union {
7068 		__be32		ipv4_nh;
7069 		__u32		ipv6_nh[4];  /* in6_addr; network order */
7070 	};
7071 };
7072 
7073 /* bpf_check_mtu flags*/
7074 enum  bpf_check_mtu_flags {
7075 	BPF_MTU_CHK_SEGS  = (1U << 0),
7076 };
7077 
7078 enum bpf_check_mtu_ret {
7079 	BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
7080 	BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
7081 	BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
7082 };
7083 
7084 enum bpf_task_fd_type {
7085 	BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */
7086 	BPF_FD_TYPE_TRACEPOINT,		/* tp name */
7087 	BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */
7088 	BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */
7089 	BPF_FD_TYPE_UPROBE,		/* filename + offset */
7090 	BPF_FD_TYPE_URETPROBE,		/* filename + offset */
7091 };
7092 
7093 enum {
7094 	BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),
7095 	BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),
7096 	BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),
7097 };
7098 
7099 struct bpf_flow_keys {
7100 	__u16	nhoff;
7101 	__u16	thoff;
7102 	__u16	addr_proto;			/* ETH_P_* of valid addrs */
7103 	__u8	is_frag;
7104 	__u8	is_first_frag;
7105 	__u8	is_encap;
7106 	__u8	ip_proto;
7107 	__be16	n_proto;
7108 	__be16	sport;
7109 	__be16	dport;
7110 	union {
7111 		struct {
7112 			__be32	ipv4_src;
7113 			__be32	ipv4_dst;
7114 		};
7115 		struct {
7116 			__u32	ipv6_src[4];	/* in6_addr; network order */
7117 			__u32	ipv6_dst[4];	/* in6_addr; network order */
7118 		};
7119 	};
7120 	__u32	flags;
7121 	__be32	flow_label;
7122 };
7123 
7124 struct bpf_func_info {
7125 	__u32	insn_off;
7126 	__u32	type_id;
7127 };
7128 
7129 #define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)
7130 #define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)
7131 
7132 struct bpf_line_info {
7133 	__u32	insn_off;
7134 	__u32	file_name_off;
7135 	__u32	line_off;
7136 	__u32	line_col;
7137 };
7138 
7139 struct bpf_spin_lock {
7140 	__u32	val;
7141 };
7142 
7143 struct bpf_timer {
7144 	__u64 :64;
7145 	__u64 :64;
7146 } __attribute__((aligned(8)));
7147 
7148 struct bpf_dynptr {
7149 	__u64 :64;
7150 	__u64 :64;
7151 } __attribute__((aligned(8)));
7152 
7153 struct bpf_list_head {
7154 	__u64 :64;
7155 	__u64 :64;
7156 } __attribute__((aligned(8)));
7157 
7158 struct bpf_list_node {
7159 	__u64 :64;
7160 	__u64 :64;
7161 	__u64 :64;
7162 } __attribute__((aligned(8)));
7163 
7164 struct bpf_rb_root {
7165 	__u64 :64;
7166 	__u64 :64;
7167 } __attribute__((aligned(8)));
7168 
7169 struct bpf_rb_node {
7170 	__u64 :64;
7171 	__u64 :64;
7172 	__u64 :64;
7173 	__u64 :64;
7174 } __attribute__((aligned(8)));
7175 
7176 struct bpf_refcount {
7177 	__u32 :32;
7178 } __attribute__((aligned(4)));
7179 
7180 struct bpf_sysctl {
7181 	__u32	write;		/* Sysctl is being read (= 0) or written (= 1).
7182 				 * Allows 1,2,4-byte read, but no write.
7183 				 */
7184 	__u32	file_pos;	/* Sysctl file position to read from, write to.
7185 				 * Allows 1,2,4-byte read an 4-byte write.
7186 				 */
7187 };
7188 
7189 struct bpf_sockopt {
7190 	__bpf_md_ptr(struct bpf_sock *, sk);
7191 	__bpf_md_ptr(void *, optval);
7192 	__bpf_md_ptr(void *, optval_end);
7193 
7194 	__s32	level;
7195 	__s32	optname;
7196 	__s32	optlen;
7197 	__s32	retval;
7198 };
7199 
7200 struct bpf_pidns_info {
7201 	__u32 pid;
7202 	__u32 tgid;
7203 };
7204 
7205 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
7206 struct bpf_sk_lookup {
7207 	union {
7208 		__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
7209 		__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
7210 	};
7211 
7212 	__u32 family;		/* Protocol family (AF_INET, AF_INET6) */
7213 	__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
7214 	__u32 remote_ip4;	/* Network byte order */
7215 	__u32 remote_ip6[4];	/* Network byte order */
7216 	__be16 remote_port;	/* Network byte order */
7217 	__u16 :16;		/* Zero padding */
7218 	__u32 local_ip4;	/* Network byte order */
7219 	__u32 local_ip6[4];	/* Network byte order */
7220 	__u32 local_port;	/* Host byte order */
7221 	__u32 ingress_ifindex;		/* The arriving interface. Determined by inet_iif. */
7222 };
7223 
7224 /*
7225  * struct btf_ptr is used for typed pointer representation; the
7226  * type id is used to render the pointer data as the appropriate type
7227  * via the bpf_snprintf_btf() helper described above.  A flags field -
7228  * potentially to specify additional details about the BTF pointer
7229  * (rather than its mode of display) - is included for future use.
7230  * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
7231  */
7232 struct btf_ptr {
7233 	void *ptr;
7234 	__u32 type_id;
7235 	__u32 flags;		/* BTF ptr flags; unused at present. */
7236 };
7237 
7238 /*
7239  * Flags to control bpf_snprintf_btf() behaviour.
7240  *     - BTF_F_COMPACT: no formatting around type information
7241  *     - BTF_F_NONAME: no struct/union member names/types
7242  *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
7243  *       equivalent to %px.
7244  *     - BTF_F_ZERO: show zero-valued struct/union members; they
7245  *       are not displayed by default
7246  */
7247 enum {
7248 	BTF_F_COMPACT	=	(1ULL << 0),
7249 	BTF_F_NONAME	=	(1ULL << 1),
7250 	BTF_F_PTR_RAW	=	(1ULL << 2),
7251 	BTF_F_ZERO	=	(1ULL << 3),
7252 };
7253 
7254 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
7255  * has to be adjusted by relocations. It is emitted by llvm and passed to
7256  * libbpf and later to the kernel.
7257  */
7258 enum bpf_core_relo_kind {
7259 	BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */
7260 	BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */
7261 	BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */
7262 	BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */
7263 	BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */
7264 	BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */
7265 	BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */
7266 	BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */
7267 	BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */
7268 	BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */
7269 	BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */
7270 	BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */
7271 	BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */
7272 };
7273 
7274 /*
7275  * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7276  * and from libbpf to the kernel.
7277  *
7278  * CO-RE relocation captures the following data:
7279  * - insn_off - instruction offset (in bytes) within a BPF program that needs
7280  *   its insn->imm field to be relocated with actual field info;
7281  * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7282  *   type or field;
7283  * - access_str_off - offset into corresponding .BTF string section. String
7284  *   interpretation depends on specific relocation kind:
7285  *     - for field-based relocations, string encodes an accessed field using
7286  *       a sequence of field and array indices, separated by colon (:). It's
7287  *       conceptually very close to LLVM's getelementptr ([0]) instruction's
7288  *       arguments for identifying offset to a field.
7289  *     - for type-based relocations, strings is expected to be just "0";
7290  *     - for enum value-based relocations, string contains an index of enum
7291  *       value within its enum type;
7292  * - kind - one of enum bpf_core_relo_kind;
7293  *
7294  * Example:
7295  *   struct sample {
7296  *       int a;
7297  *       struct {
7298  *           int b[10];
7299  *       };
7300  *   };
7301  *
7302  *   struct sample *s = ...;
7303  *   int *x = &s->a;     // encoded as "0:0" (a is field #0)
7304  *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,
7305  *                       // b is field #0 inside anon struct, accessing elem #5)
7306  *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7307  *
7308  * type_id for all relocs in this example will capture BTF type id of
7309  * `struct sample`.
7310  *
7311  * Such relocation is emitted when using __builtin_preserve_access_index()
7312  * Clang built-in, passing expression that captures field address, e.g.:
7313  *
7314  * bpf_probe_read(&dst, sizeof(dst),
7315  *		  __builtin_preserve_access_index(&src->a.b.c));
7316  *
7317  * In this case Clang will emit field relocation recording necessary data to
7318  * be able to find offset of embedded `a.b.c` field within `src` struct.
7319  *
7320  * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7321  */
7322 struct bpf_core_relo {
7323 	__u32 insn_off;
7324 	__u32 type_id;
7325 	__u32 access_str_off;
7326 	enum bpf_core_relo_kind kind;
7327 };
7328 
7329 /*
7330  * Flags to control bpf_timer_start() behaviour.
7331  *     - BPF_F_TIMER_ABS: Timeout passed is absolute time, by default it is
7332  *       relative to current time.
7333  */
7334 enum {
7335 	BPF_F_TIMER_ABS = (1ULL << 0),
7336 };
7337 
7338 /* BPF numbers iterator state */
7339 struct bpf_iter_num {
7340 	/* opaque iterator state; having __u64 here allows to preserve correct
7341 	 * alignment requirements in vmlinux.h, generated from BTF
7342 	 */
7343 	__u64 __opaque[1];
7344 } __attribute__((aligned(8)));
7345 
7346 #endif /* _UAPI__LINUX_BPF_H__ */
7347