• 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_ATOMIC	0xc0	/* atomic memory ops - op type in immediate */
23 #define BPF_XADD	0xc0	/* exclusive add - legacy name */
24 
25 /* alu/jmp fields */
26 #define BPF_MOV		0xb0	/* mov reg to reg */
27 #define BPF_ARSH	0xc0	/* sign extending arithmetic shift right */
28 
29 /* change endianness of a register */
30 #define BPF_END		0xd0	/* flags for endianness conversion: */
31 #define BPF_TO_LE	0x00	/* convert to little-endian */
32 #define BPF_TO_BE	0x08	/* convert to big-endian */
33 #define BPF_FROM_LE	BPF_TO_LE
34 #define BPF_FROM_BE	BPF_TO_BE
35 
36 /* jmp encodings */
37 #define BPF_JNE		0x50	/* jump != */
38 #define BPF_JLT		0xa0	/* LT is unsigned, '<' */
39 #define BPF_JLE		0xb0	/* LE is unsigned, '<=' */
40 #define BPF_JSGT	0x60	/* SGT is signed '>', GT in x86 */
41 #define BPF_JSGE	0x70	/* SGE is signed '>=', GE in x86 */
42 #define BPF_JSLT	0xc0	/* SLT is signed, '<' */
43 #define BPF_JSLE	0xd0	/* SLE is signed, '<=' */
44 #define BPF_CALL	0x80	/* function call */
45 #define BPF_EXIT	0x90	/* function return */
46 
47 /* atomic op type fields (stored in immediate) */
48 #define BPF_FETCH	0x01	/* not an opcode on its own, used to build others */
49 #define BPF_XCHG	(0xe0 | BPF_FETCH)	/* atomic exchange */
50 #define BPF_CMPXCHG	(0xf0 | BPF_FETCH)	/* atomic compare-and-write */
51 
52 /* Register numbers */
53 enum {
54 	BPF_REG_0 = 0,
55 	BPF_REG_1,
56 	BPF_REG_2,
57 	BPF_REG_3,
58 	BPF_REG_4,
59 	BPF_REG_5,
60 	BPF_REG_6,
61 	BPF_REG_7,
62 	BPF_REG_8,
63 	BPF_REG_9,
64 	BPF_REG_10,
65 	__MAX_BPF_REG,
66 };
67 
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG	__MAX_BPF_REG
70 
71 struct bpf_insn {
72 	__u8	code;		/* opcode */
73 	__u8	dst_reg:4;	/* dest register */
74 	__u8	src_reg:4;	/* source register */
75 	__s16	off;		/* signed offset */
76 	__s32	imm;		/* signed immediate constant */
77 };
78 
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
82 	__u8	data[0];	/* Arbitrary size */
83 };
84 
85 struct bpf_cgroup_storage_key {
86 	__u64	cgroup_inode_id;	/* cgroup inode id */
87 	__u32	attach_type;		/* program attach type (enum bpf_attach_type) */
88 };
89 
90 enum bpf_cgroup_iter_order {
91 	BPF_CGROUP_ITER_ORDER_UNSPEC = 0,
92 	BPF_CGROUP_ITER_SELF_ONLY,		/* process only a single object. */
93 	BPF_CGROUP_ITER_DESCENDANTS_PRE,	/* walk descendants in pre-order. */
94 	BPF_CGROUP_ITER_DESCENDANTS_POST,	/* walk descendants in post-order. */
95 	BPF_CGROUP_ITER_ANCESTORS_UP,		/* walk ancestors upward. */
96 };
97 
98 union bpf_iter_link_info {
99 	struct {
100 		__u32	map_fd;
101 	} map;
102 	struct {
103 		enum bpf_cgroup_iter_order order;
104 
105 		/* At most one of cgroup_fd and cgroup_id can be non-zero. If
106 		 * both are zero, the walk starts from the default cgroup v2
107 		 * root. For walking v1 hierarchy, one should always explicitly
108 		 * specify cgroup_fd.
109 		 */
110 		__u32	cgroup_fd;
111 		__u64	cgroup_id;
112 	} cgroup;
113 	/* Parameters of task iterators. */
114 	struct {
115 		__u32	tid;
116 		__u32	pid;
117 		__u32	pid_fd;
118 	} task;
119 };
120 
121 /* BPF syscall commands, see bpf(2) man-page for more details. */
122 /**
123  * DOC: eBPF Syscall Preamble
124  *
125  * The operation to be performed by the **bpf**\ () system call is determined
126  * by the *cmd* argument. Each operation takes an accompanying argument,
127  * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
128  * below). The size argument is the size of the union pointed to by *attr*.
129  */
130 /**
131  * DOC: eBPF Syscall Commands
132  *
133  * BPF_MAP_CREATE
134  *	Description
135  *		Create a map and return a file descriptor that refers to the
136  *		map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
137  *		is automatically enabled for the new file descriptor.
138  *
139  *		Applying **close**\ (2) to the file descriptor returned by
140  *		**BPF_MAP_CREATE** will delete the map (but see NOTES).
141  *
142  *	Return
143  *		A new file descriptor (a nonnegative integer), or -1 if an
144  *		error occurred (in which case, *errno* is set appropriately).
145  *
146  * BPF_MAP_LOOKUP_ELEM
147  *	Description
148  *		Look up an element with a given *key* in the map referred to
149  *		by the file descriptor *map_fd*.
150  *
151  *		The *flags* argument may be specified as one of the
152  *		following:
153  *
154  *		**BPF_F_LOCK**
155  *			Look up the value of a spin-locked map without
156  *			returning the lock. This must be specified if the
157  *			elements contain a spinlock.
158  *
159  *	Return
160  *		Returns zero on success. On error, -1 is returned and *errno*
161  *		is set appropriately.
162  *
163  * BPF_MAP_UPDATE_ELEM
164  *	Description
165  *		Create or update an element (key/value pair) in a specified map.
166  *
167  *		The *flags* argument should be specified as one of the
168  *		following:
169  *
170  *		**BPF_ANY**
171  *			Create a new element or update an existing element.
172  *		**BPF_NOEXIST**
173  *			Create a new element only if it did not exist.
174  *		**BPF_EXIST**
175  *			Update an existing element.
176  *		**BPF_F_LOCK**
177  *			Update a spin_lock-ed map element.
178  *
179  *	Return
180  *		Returns zero on success. On error, -1 is returned and *errno*
181  *		is set appropriately.
182  *
183  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
184  *		**E2BIG**, **EEXIST**, or **ENOENT**.
185  *
186  *		**E2BIG**
187  *			The number of elements in the map reached the
188  *			*max_entries* limit specified at map creation time.
189  *		**EEXIST**
190  *			If *flags* specifies **BPF_NOEXIST** and the element
191  *			with *key* already exists in the map.
192  *		**ENOENT**
193  *			If *flags* specifies **BPF_EXIST** and the element with
194  *			*key* does not exist in the map.
195  *
196  * BPF_MAP_DELETE_ELEM
197  *	Description
198  *		Look up and delete an element by key in a specified map.
199  *
200  *	Return
201  *		Returns zero on success. On error, -1 is returned and *errno*
202  *		is set appropriately.
203  *
204  * BPF_MAP_GET_NEXT_KEY
205  *	Description
206  *		Look up an element by key in a specified map and return the key
207  *		of the next element. Can be used to iterate over all elements
208  *		in the map.
209  *
210  *	Return
211  *		Returns zero on success. On error, -1 is returned and *errno*
212  *		is set appropriately.
213  *
214  *		The following cases can be used to iterate over all elements of
215  *		the map:
216  *
217  *		* If *key* is not found, the operation returns zero and sets
218  *		  the *next_key* pointer to the key of the first element.
219  *		* If *key* is found, the operation returns zero and sets the
220  *		  *next_key* pointer to the key of the next element.
221  *		* If *key* is the last element, returns -1 and *errno* is set
222  *		  to **ENOENT**.
223  *
224  *		May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
225  *		**EINVAL** on error.
226  *
227  * BPF_PROG_LOAD
228  *	Description
229  *		Verify and load an eBPF program, returning a new file
230  *		descriptor associated with the program.
231  *
232  *		Applying **close**\ (2) to the file descriptor returned by
233  *		**BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
234  *
235  *		The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
236  *		automatically enabled for the new file descriptor.
237  *
238  *	Return
239  *		A new file descriptor (a nonnegative integer), or -1 if an
240  *		error occurred (in which case, *errno* is set appropriately).
241  *
242  * BPF_OBJ_PIN
243  *	Description
244  *		Pin an eBPF program or map referred by the specified *bpf_fd*
245  *		to the provided *pathname* on the filesystem.
246  *
247  *		The *pathname* argument must not contain a dot (".").
248  *
249  *		On success, *pathname* retains a reference to the eBPF object,
250  *		preventing deallocation of the object when the original
251  *		*bpf_fd* is closed. This allow the eBPF object to live beyond
252  *		**close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
253  *		process.
254  *
255  *		Applying **unlink**\ (2) or similar calls to the *pathname*
256  *		unpins the object from the filesystem, removing the reference.
257  *		If no other file descriptors or filesystem nodes refer to the
258  *		same object, it will be deallocated (see NOTES).
259  *
260  *		The filesystem type for the parent directory of *pathname* must
261  *		be **BPF_FS_MAGIC**.
262  *
263  *	Return
264  *		Returns zero on success. On error, -1 is returned and *errno*
265  *		is set appropriately.
266  *
267  * BPF_OBJ_GET
268  *	Description
269  *		Open a file descriptor for the eBPF object pinned to the
270  *		specified *pathname*.
271  *
272  *	Return
273  *		A new file descriptor (a nonnegative integer), or -1 if an
274  *		error occurred (in which case, *errno* is set appropriately).
275  *
276  * BPF_PROG_ATTACH
277  *	Description
278  *		Attach an eBPF program to a *target_fd* at the specified
279  *		*attach_type* hook.
280  *
281  *		The *attach_type* specifies the eBPF attachment point to
282  *		attach the program to, and must be one of *bpf_attach_type*
283  *		(see below).
284  *
285  *		The *attach_bpf_fd* must be a valid file descriptor for a
286  *		loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
287  *		or sock_ops type corresponding to the specified *attach_type*.
288  *
289  *		The *target_fd* must be a valid file descriptor for a kernel
290  *		object which depends on the attach type of *attach_bpf_fd*:
291  *
292  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
293  *		**BPF_PROG_TYPE_CGROUP_SKB**,
294  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
295  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
296  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
297  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
298  *		**BPF_PROG_TYPE_SOCK_OPS**
299  *
300  *			Control Group v2 hierarchy with the eBPF controller
301  *			enabled. Requires the kernel to be compiled with
302  *			**CONFIG_CGROUP_BPF**.
303  *
304  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
305  *
306  *			Network namespace (eg /proc/self/ns/net).
307  *
308  *		**BPF_PROG_TYPE_LIRC_MODE2**
309  *
310  *			LIRC device path (eg /dev/lircN). Requires the kernel
311  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
312  *
313  *		**BPF_PROG_TYPE_SK_SKB**,
314  *		**BPF_PROG_TYPE_SK_MSG**
315  *
316  *			eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
317  *
318  *	Return
319  *		Returns zero on success. On error, -1 is returned and *errno*
320  *		is set appropriately.
321  *
322  * BPF_PROG_DETACH
323  *	Description
324  *		Detach the eBPF program associated with the *target_fd* at the
325  *		hook specified by *attach_type*. The program must have been
326  *		previously attached using **BPF_PROG_ATTACH**.
327  *
328  *	Return
329  *		Returns zero on success. On error, -1 is returned and *errno*
330  *		is set appropriately.
331  *
332  * BPF_PROG_TEST_RUN
333  *	Description
334  *		Run the eBPF program associated with the *prog_fd* a *repeat*
335  *		number of times against a provided program context *ctx_in* and
336  *		data *data_in*, and return the modified program context
337  *		*ctx_out*, *data_out* (for example, packet data), result of the
338  *		execution *retval*, and *duration* of the test run.
339  *
340  *		The sizes of the buffers provided as input and output
341  *		parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
342  *		be provided in the corresponding variables *ctx_size_in*,
343  *		*ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
344  *		of these parameters are not provided (ie set to NULL), the
345  *		corresponding size field must be zero.
346  *
347  *		Some program types have particular requirements:
348  *
349  *		**BPF_PROG_TYPE_SK_LOOKUP**
350  *			*data_in* and *data_out* must be NULL.
351  *
352  *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,
353  *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
354  *
355  *			*ctx_out*, *data_in* and *data_out* must be NULL.
356  *			*repeat* must be zero.
357  *
358  *		BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
359  *
360  *	Return
361  *		Returns zero on success. On error, -1 is returned and *errno*
362  *		is set appropriately.
363  *
364  *		**ENOSPC**
365  *			Either *data_size_out* or *ctx_size_out* is too small.
366  *		**ENOTSUPP**
367  *			This command is not supported by the program type of
368  *			the program referred to by *prog_fd*.
369  *
370  * BPF_PROG_GET_NEXT_ID
371  *	Description
372  *		Fetch the next eBPF program currently loaded into the kernel.
373  *
374  *		Looks for the eBPF program with an id greater than *start_id*
375  *		and updates *next_id* on success. If no other eBPF programs
376  *		remain with ids higher than *start_id*, returns -1 and sets
377  *		*errno* to **ENOENT**.
378  *
379  *	Return
380  *		Returns zero on success. On error, or when no id remains, -1
381  *		is returned and *errno* is set appropriately.
382  *
383  * BPF_MAP_GET_NEXT_ID
384  *	Description
385  *		Fetch the next eBPF map currently loaded into the kernel.
386  *
387  *		Looks for the eBPF map with an id greater than *start_id*
388  *		and updates *next_id* on success. If no other eBPF maps
389  *		remain with ids higher than *start_id*, returns -1 and sets
390  *		*errno* to **ENOENT**.
391  *
392  *	Return
393  *		Returns zero on success. On error, or when no id remains, -1
394  *		is returned and *errno* is set appropriately.
395  *
396  * BPF_PROG_GET_FD_BY_ID
397  *	Description
398  *		Open a file descriptor for the eBPF program corresponding to
399  *		*prog_id*.
400  *
401  *	Return
402  *		A new file descriptor (a nonnegative integer), or -1 if an
403  *		error occurred (in which case, *errno* is set appropriately).
404  *
405  * BPF_MAP_GET_FD_BY_ID
406  *	Description
407  *		Open a file descriptor for the eBPF map corresponding to
408  *		*map_id*.
409  *
410  *	Return
411  *		A new file descriptor (a nonnegative integer), or -1 if an
412  *		error occurred (in which case, *errno* is set appropriately).
413  *
414  * BPF_OBJ_GET_INFO_BY_FD
415  *	Description
416  *		Obtain information about the eBPF object corresponding to
417  *		*bpf_fd*.
418  *
419  *		Populates up to *info_len* bytes of *info*, which will be in
420  *		one of the following formats depending on the eBPF object type
421  *		of *bpf_fd*:
422  *
423  *		* **struct bpf_prog_info**
424  *		* **struct bpf_map_info**
425  *		* **struct bpf_btf_info**
426  *		* **struct bpf_link_info**
427  *
428  *	Return
429  *		Returns zero on success. On error, -1 is returned and *errno*
430  *		is set appropriately.
431  *
432  * BPF_PROG_QUERY
433  *	Description
434  *		Obtain information about eBPF programs associated with the
435  *		specified *attach_type* hook.
436  *
437  *		The *target_fd* must be a valid file descriptor for a kernel
438  *		object which depends on the attach type of *attach_bpf_fd*:
439  *
440  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
441  *		**BPF_PROG_TYPE_CGROUP_SKB**,
442  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
443  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
444  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
445  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
446  *		**BPF_PROG_TYPE_SOCK_OPS**
447  *
448  *			Control Group v2 hierarchy with the eBPF controller
449  *			enabled. Requires the kernel to be compiled with
450  *			**CONFIG_CGROUP_BPF**.
451  *
452  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
453  *
454  *			Network namespace (eg /proc/self/ns/net).
455  *
456  *		**BPF_PROG_TYPE_LIRC_MODE2**
457  *
458  *			LIRC device path (eg /dev/lircN). Requires the kernel
459  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
460  *
461  *		**BPF_PROG_QUERY** always fetches the number of programs
462  *		attached and the *attach_flags* which were used to attach those
463  *		programs. Additionally, if *prog_ids* is nonzero and the number
464  *		of attached programs is less than *prog_cnt*, populates
465  *		*prog_ids* with the eBPF program ids of the programs attached
466  *		at *target_fd*.
467  *
468  *		The following flags may alter the result:
469  *
470  *		**BPF_F_QUERY_EFFECTIVE**
471  *			Only return information regarding programs which are
472  *			currently effective at the specified *target_fd*.
473  *
474  *	Return
475  *		Returns zero on success. On error, -1 is returned and *errno*
476  *		is set appropriately.
477  *
478  * BPF_RAW_TRACEPOINT_OPEN
479  *	Description
480  *		Attach an eBPF program to a tracepoint *name* to access kernel
481  *		internal arguments of the tracepoint in their raw form.
482  *
483  *		The *prog_fd* must be a valid file descriptor associated with
484  *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
485  *
486  *		No ABI guarantees are made about the content of tracepoint
487  *		arguments exposed to the corresponding eBPF program.
488  *
489  *		Applying **close**\ (2) to the file descriptor returned by
490  *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
491  *
492  *	Return
493  *		A new file descriptor (a nonnegative integer), or -1 if an
494  *		error occurred (in which case, *errno* is set appropriately).
495  *
496  * BPF_BTF_LOAD
497  *	Description
498  *		Verify and load BPF Type Format (BTF) metadata into the kernel,
499  *		returning a new file descriptor associated with the metadata.
500  *		BTF is described in more detail at
501  *		https://www.kernel.org/doc/html/latest/bpf/btf.html.
502  *
503  *		The *btf* parameter must point to valid memory providing
504  *		*btf_size* bytes of BTF binary metadata.
505  *
506  *		The returned file descriptor can be passed to other **bpf**\ ()
507  *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
508  *		associate the BTF with those objects.
509  *
510  *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
511  *		parameters to specify a *btf_log_buf*, *btf_log_size* and
512  *		*btf_log_level* which allow the kernel to return freeform log
513  *		output regarding the BTF verification process.
514  *
515  *	Return
516  *		A new file descriptor (a nonnegative integer), or -1 if an
517  *		error occurred (in which case, *errno* is set appropriately).
518  *
519  * BPF_BTF_GET_FD_BY_ID
520  *	Description
521  *		Open a file descriptor for the BPF Type Format (BTF)
522  *		corresponding to *btf_id*.
523  *
524  *	Return
525  *		A new file descriptor (a nonnegative integer), or -1 if an
526  *		error occurred (in which case, *errno* is set appropriately).
527  *
528  * BPF_TASK_FD_QUERY
529  *	Description
530  *		Obtain information about eBPF programs associated with the
531  *		target process identified by *pid* and *fd*.
532  *
533  *		If the *pid* and *fd* are associated with a tracepoint, kprobe
534  *		or uprobe perf event, then the *prog_id* and *fd_type* will
535  *		be populated with the eBPF program id and file descriptor type
536  *		of type **bpf_task_fd_type**. If associated with a kprobe or
537  *		uprobe, the  *probe_offset* and *probe_addr* will also be
538  *		populated. Optionally, if *buf* is provided, then up to
539  *		*buf_len* bytes of *buf* will be populated with the name of
540  *		the tracepoint, kprobe or uprobe.
541  *
542  *		The resulting *prog_id* may be introspected in deeper detail
543  *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
544  *
545  *	Return
546  *		Returns zero on success. On error, -1 is returned and *errno*
547  *		is set appropriately.
548  *
549  * BPF_MAP_LOOKUP_AND_DELETE_ELEM
550  *	Description
551  *		Look up an element with the given *key* in the map referred to
552  *		by the file descriptor *fd*, and if found, delete the element.
553  *
554  *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
555  *		types, the *flags* argument needs to be set to 0, but for other
556  *		map types, it may be specified as:
557  *
558  *		**BPF_F_LOCK**
559  *			Look up and delete the value of a spin-locked map
560  *			without returning the lock. This must be specified if
561  *			the elements contain a spinlock.
562  *
563  *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
564  *		implement this command as a "pop" operation, deleting the top
565  *		element rather than one corresponding to *key*.
566  *		The *key* and *key_len* parameters should be zeroed when
567  *		issuing this operation for these map types.
568  *
569  *		This command is only valid for the following map types:
570  *		* **BPF_MAP_TYPE_QUEUE**
571  *		* **BPF_MAP_TYPE_STACK**
572  *		* **BPF_MAP_TYPE_HASH**
573  *		* **BPF_MAP_TYPE_PERCPU_HASH**
574  *		* **BPF_MAP_TYPE_LRU_HASH**
575  *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**
576  *
577  *	Return
578  *		Returns zero on success. On error, -1 is returned and *errno*
579  *		is set appropriately.
580  *
581  * BPF_MAP_FREEZE
582  *	Description
583  *		Freeze the permissions of the specified map.
584  *
585  *		Write permissions may be frozen by passing zero *flags*.
586  *		Upon success, no future syscall invocations may alter the
587  *		map state of *map_fd*. Write operations from eBPF programs
588  *		are still possible for a frozen map.
589  *
590  *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
591  *
592  *	Return
593  *		Returns zero on success. On error, -1 is returned and *errno*
594  *		is set appropriately.
595  *
596  * BPF_BTF_GET_NEXT_ID
597  *	Description
598  *		Fetch the next BPF Type Format (BTF) object currently loaded
599  *		into the kernel.
600  *
601  *		Looks for the BTF object with an id greater than *start_id*
602  *		and updates *next_id* on success. If no other BTF objects
603  *		remain with ids higher than *start_id*, returns -1 and sets
604  *		*errno* to **ENOENT**.
605  *
606  *	Return
607  *		Returns zero on success. On error, or when no id remains, -1
608  *		is returned and *errno* is set appropriately.
609  *
610  * BPF_MAP_LOOKUP_BATCH
611  *	Description
612  *		Iterate and fetch multiple elements in a map.
613  *
614  *		Two opaque values are used to manage batch operations,
615  *		*in_batch* and *out_batch*. Initially, *in_batch* must be set
616  *		to NULL to begin the batched operation. After each subsequent
617  *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
618  *		*out_batch* as the *in_batch* for the next operation to
619  *		continue iteration from the current point.
620  *
621  *		The *keys* and *values* are output parameters which must point
622  *		to memory large enough to hold *count* items based on the key
623  *		and value size of the map *map_fd*. The *keys* buffer must be
624  *		of *key_size* * *count*. The *values* buffer must be of
625  *		*value_size* * *count*.
626  *
627  *		The *elem_flags* argument may be specified as one of the
628  *		following:
629  *
630  *		**BPF_F_LOCK**
631  *			Look up the value of a spin-locked map without
632  *			returning the lock. This must be specified if the
633  *			elements contain a spinlock.
634  *
635  *		On success, *count* elements from the map are copied into the
636  *		user buffer, with the keys copied into *keys* and the values
637  *		copied into the corresponding indices in *values*.
638  *
639  *		If an error is returned and *errno* is not **EFAULT**, *count*
640  *		is set to the number of successfully processed elements.
641  *
642  *	Return
643  *		Returns zero on success. On error, -1 is returned and *errno*
644  *		is set appropriately.
645  *
646  *		May set *errno* to **ENOSPC** to indicate that *keys* or
647  *		*values* is too small to dump an entire bucket during
648  *		iteration of a hash-based map type.
649  *
650  * BPF_MAP_LOOKUP_AND_DELETE_BATCH
651  *	Description
652  *		Iterate and delete all elements in a map.
653  *
654  *		This operation has the same behavior as
655  *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:
656  *
657  *		* Every element that is successfully returned is also deleted
658  *		  from the map. This is at least *count* elements. Note that
659  *		  *count* is both an input and an output parameter.
660  *		* Upon returning with *errno* set to **EFAULT**, up to
661  *		  *count* elements may be deleted without returning the keys
662  *		  and values of the deleted elements.
663  *
664  *	Return
665  *		Returns zero on success. On error, -1 is returned and *errno*
666  *		is set appropriately.
667  *
668  * BPF_MAP_UPDATE_BATCH
669  *	Description
670  *		Update multiple elements in a map by *key*.
671  *
672  *		The *keys* and *values* are input parameters which must point
673  *		to memory large enough to hold *count* items based on the key
674  *		and value size of the map *map_fd*. The *keys* buffer must be
675  *		of *key_size* * *count*. The *values* buffer must be of
676  *		*value_size* * *count*.
677  *
678  *		Each element specified in *keys* is sequentially updated to the
679  *		value in the corresponding index in *values*. The *in_batch*
680  *		and *out_batch* parameters are ignored and should be zeroed.
681  *
682  *		The *elem_flags* argument should be specified as one of the
683  *		following:
684  *
685  *		**BPF_ANY**
686  *			Create new elements or update a existing elements.
687  *		**BPF_NOEXIST**
688  *			Create new elements only if they do not exist.
689  *		**BPF_EXIST**
690  *			Update existing elements.
691  *		**BPF_F_LOCK**
692  *			Update spin_lock-ed map elements. This must be
693  *			specified if the map value contains a spinlock.
694  *
695  *		On success, *count* elements from the map are updated.
696  *
697  *		If an error is returned and *errno* is not **EFAULT**, *count*
698  *		is set to the number of successfully processed elements.
699  *
700  *	Return
701  *		Returns zero on success. On error, -1 is returned and *errno*
702  *		is set appropriately.
703  *
704  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
705  *		**E2BIG**. **E2BIG** indicates that the number of elements in
706  *		the map reached the *max_entries* limit specified at map
707  *		creation time.
708  *
709  *		May set *errno* to one of the following error codes under
710  *		specific circumstances:
711  *
712  *		**EEXIST**
713  *			If *flags* specifies **BPF_NOEXIST** and the element
714  *			with *key* already exists in the map.
715  *		**ENOENT**
716  *			If *flags* specifies **BPF_EXIST** and the element with
717  *			*key* does not exist in the map.
718  *
719  * BPF_MAP_DELETE_BATCH
720  *	Description
721  *		Delete multiple elements in a map by *key*.
722  *
723  *		The *keys* parameter is an input parameter which must point
724  *		to memory large enough to hold *count* items based on the key
725  *		size of the map *map_fd*, that is, *key_size* * *count*.
726  *
727  *		Each element specified in *keys* is sequentially deleted. The
728  *		*in_batch*, *out_batch*, and *values* parameters are ignored
729  *		and should be zeroed.
730  *
731  *		The *elem_flags* argument may be specified as one of the
732  *		following:
733  *
734  *		**BPF_F_LOCK**
735  *			Look up the value of a spin-locked map without
736  *			returning the lock. This must be specified if the
737  *			elements contain a spinlock.
738  *
739  *		On success, *count* elements from the map are updated.
740  *
741  *		If an error is returned and *errno* is not **EFAULT**, *count*
742  *		is set to the number of successfully processed elements. If
743  *		*errno* is **EFAULT**, up to *count* elements may be been
744  *		deleted.
745  *
746  *	Return
747  *		Returns zero on success. On error, -1 is returned and *errno*
748  *		is set appropriately.
749  *
750  * BPF_LINK_CREATE
751  *	Description
752  *		Attach an eBPF program to a *target_fd* at the specified
753  *		*attach_type* hook and return a file descriptor handle for
754  *		managing the link.
755  *
756  *	Return
757  *		A new file descriptor (a nonnegative integer), or -1 if an
758  *		error occurred (in which case, *errno* is set appropriately).
759  *
760  * BPF_LINK_UPDATE
761  *	Description
762  *		Update the eBPF program in the specified *link_fd* to
763  *		*new_prog_fd*.
764  *
765  *	Return
766  *		Returns zero on success. On error, -1 is returned and *errno*
767  *		is set appropriately.
768  *
769  * BPF_LINK_GET_FD_BY_ID
770  *	Description
771  *		Open a file descriptor for the eBPF Link corresponding to
772  *		*link_id*.
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_GET_NEXT_ID
779  *	Description
780  *		Fetch the next eBPF link currently loaded into the kernel.
781  *
782  *		Looks for the eBPF link with an id greater than *start_id*
783  *		and updates *next_id* on success. If no other eBPF links
784  *		remain with ids higher than *start_id*, returns -1 and sets
785  *		*errno* to **ENOENT**.
786  *
787  *	Return
788  *		Returns zero on success. On error, or when no id remains, -1
789  *		is returned and *errno* is set appropriately.
790  *
791  * BPF_ENABLE_STATS
792  *	Description
793  *		Enable eBPF runtime statistics gathering.
794  *
795  *		Runtime statistics gathering for the eBPF runtime is disabled
796  *		by default to minimize the corresponding performance overhead.
797  *		This command enables statistics globally.
798  *
799  *		Multiple programs may independently enable statistics.
800  *		After gathering the desired statistics, eBPF runtime statistics
801  *		may be disabled again by calling **close**\ (2) for the file
802  *		descriptor returned by this function. Statistics will only be
803  *		disabled system-wide when all outstanding file descriptors
804  *		returned by prior calls for this subcommand are closed.
805  *
806  *	Return
807  *		A new file descriptor (a nonnegative integer), or -1 if an
808  *		error occurred (in which case, *errno* is set appropriately).
809  *
810  * BPF_ITER_CREATE
811  *	Description
812  *		Create an iterator on top of the specified *link_fd* (as
813  *		previously created using **BPF_LINK_CREATE**) and return a
814  *		file descriptor that can be used to trigger the iteration.
815  *
816  *		If the resulting file descriptor is pinned to the filesystem
817  *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
818  *		for that path will trigger the iterator to read kernel state
819  *		using the eBPF program attached to *link_fd*.
820  *
821  *	Return
822  *		A new file descriptor (a nonnegative integer), or -1 if an
823  *		error occurred (in which case, *errno* is set appropriately).
824  *
825  * BPF_LINK_DETACH
826  *	Description
827  *		Forcefully detach the specified *link_fd* from its
828  *		corresponding attachment point.
829  *
830  *	Return
831  *		Returns zero on success. On error, -1 is returned and *errno*
832  *		is set appropriately.
833  *
834  * BPF_PROG_BIND_MAP
835  *	Description
836  *		Bind a map to the lifetime of an eBPF program.
837  *
838  *		The map identified by *map_fd* is bound to the program
839  *		identified by *prog_fd* and only released when *prog_fd* is
840  *		released. This may be used in cases where metadata should be
841  *		associated with a program which otherwise does not contain any
842  *		references to the map (for example, embedded in the eBPF
843  *		program instructions).
844  *
845  *	Return
846  *		Returns zero on success. On error, -1 is returned and *errno*
847  *		is set appropriately.
848  *
849  * NOTES
850  *	eBPF objects (maps and programs) can be shared between processes.
851  *
852  *	* After **fork**\ (2), the child inherits file descriptors
853  *	  referring to the same eBPF objects.
854  *	* File descriptors referring to eBPF objects can be transferred over
855  *	  **unix**\ (7) domain sockets.
856  *	* File descriptors referring to eBPF objects can be duplicated in the
857  *	  usual way, using **dup**\ (2) and similar calls.
858  *	* File descriptors referring to eBPF objects can be pinned to the
859  *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
860  *
861  *	An eBPF object is deallocated only after all file descriptors referring
862  *	to the object have been closed and no references remain pinned to the
863  *	filesystem or attached (for example, bound to a program or device).
864  */
865 enum bpf_cmd {
866 	BPF_MAP_CREATE,
867 	BPF_MAP_LOOKUP_ELEM,
868 	BPF_MAP_UPDATE_ELEM,
869 	BPF_MAP_DELETE_ELEM,
870 	BPF_MAP_GET_NEXT_KEY,
871 	BPF_PROG_LOAD,
872 	BPF_OBJ_PIN,
873 	BPF_OBJ_GET,
874 	BPF_PROG_ATTACH,
875 	BPF_PROG_DETACH,
876 	BPF_PROG_TEST_RUN,
877 	BPF_PROG_RUN = BPF_PROG_TEST_RUN,
878 	BPF_PROG_GET_NEXT_ID,
879 	BPF_MAP_GET_NEXT_ID,
880 	BPF_PROG_GET_FD_BY_ID,
881 	BPF_MAP_GET_FD_BY_ID,
882 	BPF_OBJ_GET_INFO_BY_FD,
883 	BPF_PROG_QUERY,
884 	BPF_RAW_TRACEPOINT_OPEN,
885 	BPF_BTF_LOAD,
886 	BPF_BTF_GET_FD_BY_ID,
887 	BPF_TASK_FD_QUERY,
888 	BPF_MAP_LOOKUP_AND_DELETE_ELEM,
889 	BPF_MAP_FREEZE,
890 	BPF_BTF_GET_NEXT_ID,
891 	BPF_MAP_LOOKUP_BATCH,
892 	BPF_MAP_LOOKUP_AND_DELETE_BATCH,
893 	BPF_MAP_UPDATE_BATCH,
894 	BPF_MAP_DELETE_BATCH,
895 	BPF_LINK_CREATE,
896 	BPF_LINK_UPDATE,
897 	BPF_LINK_GET_FD_BY_ID,
898 	BPF_LINK_GET_NEXT_ID,
899 	BPF_ENABLE_STATS,
900 	BPF_ITER_CREATE,
901 	BPF_LINK_DETACH,
902 	BPF_PROG_BIND_MAP,
903 };
904 
905 enum bpf_map_type {
906 	BPF_MAP_TYPE_UNSPEC,
907 	BPF_MAP_TYPE_HASH,
908 	BPF_MAP_TYPE_ARRAY,
909 	BPF_MAP_TYPE_PROG_ARRAY,
910 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
911 	BPF_MAP_TYPE_PERCPU_HASH,
912 	BPF_MAP_TYPE_PERCPU_ARRAY,
913 	BPF_MAP_TYPE_STACK_TRACE,
914 	BPF_MAP_TYPE_CGROUP_ARRAY,
915 	BPF_MAP_TYPE_LRU_HASH,
916 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
917 	BPF_MAP_TYPE_LPM_TRIE,
918 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
919 	BPF_MAP_TYPE_HASH_OF_MAPS,
920 	BPF_MAP_TYPE_DEVMAP,
921 	BPF_MAP_TYPE_SOCKMAP,
922 	BPF_MAP_TYPE_CPUMAP,
923 	BPF_MAP_TYPE_XSKMAP,
924 	BPF_MAP_TYPE_SOCKHASH,
925 	BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
926 	/* BPF_MAP_TYPE_CGROUP_STORAGE is available to bpf programs attaching
927 	 * to a cgroup. The newer BPF_MAP_TYPE_CGRP_STORAGE is available to
928 	 * both cgroup-attached and other progs and supports all functionality
929 	 * provided by BPF_MAP_TYPE_CGROUP_STORAGE. So mark
930 	 * BPF_MAP_TYPE_CGROUP_STORAGE deprecated.
931 	 */
932 	BPF_MAP_TYPE_CGROUP_STORAGE = BPF_MAP_TYPE_CGROUP_STORAGE_DEPRECATED,
933 	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
934 	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
935 	BPF_MAP_TYPE_QUEUE,
936 	BPF_MAP_TYPE_STACK,
937 	BPF_MAP_TYPE_SK_STORAGE,
938 	BPF_MAP_TYPE_DEVMAP_HASH,
939 	BPF_MAP_TYPE_STRUCT_OPS,
940 	BPF_MAP_TYPE_RINGBUF,
941 	BPF_MAP_TYPE_INODE_STORAGE,
942 	BPF_MAP_TYPE_TASK_STORAGE,
943 	BPF_MAP_TYPE_BLOOM_FILTER,
944 	BPF_MAP_TYPE_USER_RINGBUF,
945 	BPF_MAP_TYPE_CGRP_STORAGE,
946 };
947 
948 /* Note that tracing related programs such as
949  * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
950  * are not subject to a stable API since kernel internal data
951  * structures can change from release to release and may
952  * therefore break existing tracing BPF programs. Tracing BPF
953  * programs correspond to /a/ specific kernel which is to be
954  * analyzed, and not /a/ specific kernel /and/ all future ones.
955  */
956 enum bpf_prog_type {
957 	BPF_PROG_TYPE_UNSPEC,
958 	BPF_PROG_TYPE_SOCKET_FILTER,
959 	BPF_PROG_TYPE_KPROBE,
960 	BPF_PROG_TYPE_SCHED_CLS,
961 	BPF_PROG_TYPE_SCHED_ACT,
962 	BPF_PROG_TYPE_TRACEPOINT,
963 	BPF_PROG_TYPE_XDP,
964 	BPF_PROG_TYPE_PERF_EVENT,
965 	BPF_PROG_TYPE_CGROUP_SKB,
966 	BPF_PROG_TYPE_CGROUP_SOCK,
967 	BPF_PROG_TYPE_LWT_IN,
968 	BPF_PROG_TYPE_LWT_OUT,
969 	BPF_PROG_TYPE_LWT_XMIT,
970 	BPF_PROG_TYPE_SOCK_OPS,
971 	BPF_PROG_TYPE_SK_SKB,
972 	BPF_PROG_TYPE_CGROUP_DEVICE,
973 	BPF_PROG_TYPE_SK_MSG,
974 	BPF_PROG_TYPE_RAW_TRACEPOINT,
975 	BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
976 	BPF_PROG_TYPE_LWT_SEG6LOCAL,
977 	BPF_PROG_TYPE_LIRC_MODE2,
978 	BPF_PROG_TYPE_SK_REUSEPORT,
979 	BPF_PROG_TYPE_FLOW_DISSECTOR,
980 	BPF_PROG_TYPE_CGROUP_SYSCTL,
981 	BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
982 	BPF_PROG_TYPE_CGROUP_SOCKOPT,
983 	BPF_PROG_TYPE_TRACING,
984 	BPF_PROG_TYPE_STRUCT_OPS,
985 	BPF_PROG_TYPE_EXT,
986 	BPF_PROG_TYPE_LSM,
987 	BPF_PROG_TYPE_SK_LOOKUP,
988 	BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
989 };
990 
991 enum bpf_attach_type {
992 	BPF_CGROUP_INET_INGRESS,
993 	BPF_CGROUP_INET_EGRESS,
994 	BPF_CGROUP_INET_SOCK_CREATE,
995 	BPF_CGROUP_SOCK_OPS,
996 	BPF_SK_SKB_STREAM_PARSER,
997 	BPF_SK_SKB_STREAM_VERDICT,
998 	BPF_CGROUP_DEVICE,
999 	BPF_SK_MSG_VERDICT,
1000 	BPF_CGROUP_INET4_BIND,
1001 	BPF_CGROUP_INET6_BIND,
1002 	BPF_CGROUP_INET4_CONNECT,
1003 	BPF_CGROUP_INET6_CONNECT,
1004 	BPF_CGROUP_INET4_POST_BIND,
1005 	BPF_CGROUP_INET6_POST_BIND,
1006 	BPF_CGROUP_UDP4_SENDMSG,
1007 	BPF_CGROUP_UDP6_SENDMSG,
1008 	BPF_LIRC_MODE2,
1009 	BPF_FLOW_DISSECTOR,
1010 	BPF_CGROUP_SYSCTL,
1011 	BPF_CGROUP_UDP4_RECVMSG,
1012 	BPF_CGROUP_UDP6_RECVMSG,
1013 	BPF_CGROUP_GETSOCKOPT,
1014 	BPF_CGROUP_SETSOCKOPT,
1015 	BPF_TRACE_RAW_TP,
1016 	BPF_TRACE_FENTRY,
1017 	BPF_TRACE_FEXIT,
1018 	BPF_MODIFY_RETURN,
1019 	BPF_LSM_MAC,
1020 	BPF_TRACE_ITER,
1021 	BPF_CGROUP_INET4_GETPEERNAME,
1022 	BPF_CGROUP_INET6_GETPEERNAME,
1023 	BPF_CGROUP_INET4_GETSOCKNAME,
1024 	BPF_CGROUP_INET6_GETSOCKNAME,
1025 	BPF_XDP_DEVMAP,
1026 	BPF_CGROUP_INET_SOCK_RELEASE,
1027 	BPF_XDP_CPUMAP,
1028 	BPF_SK_LOOKUP,
1029 	BPF_XDP,
1030 	BPF_SK_SKB_VERDICT,
1031 	BPF_SK_REUSEPORT_SELECT,
1032 	BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
1033 	BPF_PERF_EVENT,
1034 	BPF_TRACE_KPROBE_MULTI,
1035 	BPF_LSM_CGROUP,
1036 	__MAX_BPF_ATTACH_TYPE
1037 };
1038 
1039 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1040 
1041 enum bpf_link_type {
1042 	BPF_LINK_TYPE_UNSPEC = 0,
1043 	BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1044 	BPF_LINK_TYPE_TRACING = 2,
1045 	BPF_LINK_TYPE_CGROUP = 3,
1046 	BPF_LINK_TYPE_ITER = 4,
1047 	BPF_LINK_TYPE_NETNS = 5,
1048 	BPF_LINK_TYPE_XDP = 6,
1049 	BPF_LINK_TYPE_PERF_EVENT = 7,
1050 	BPF_LINK_TYPE_KPROBE_MULTI = 8,
1051 	BPF_LINK_TYPE_STRUCT_OPS = 9,
1052 
1053 	MAX_BPF_LINK_TYPE,
1054 };
1055 
1056 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1057  *
1058  * NONE(default): No further bpf programs allowed in the subtree.
1059  *
1060  * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1061  * the program in this cgroup yields to sub-cgroup program.
1062  *
1063  * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1064  * that cgroup program gets run in addition to the program in this cgroup.
1065  *
1066  * Only one program is allowed to be attached to a cgroup with
1067  * NONE or BPF_F_ALLOW_OVERRIDE flag.
1068  * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1069  * release old program and attach the new one. Attach flags has to match.
1070  *
1071  * Multiple programs are allowed to be attached to a cgroup with
1072  * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1073  * (those that were attached first, run first)
1074  * The programs of sub-cgroup are executed first, then programs of
1075  * this cgroup and then programs of parent cgroup.
1076  * When children program makes decision (like picking TCP CA or sock bind)
1077  * parent program has a chance to override it.
1078  *
1079  * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1080  * programs for a cgroup. Though it's possible to replace an old program at
1081  * any position by also specifying BPF_F_REPLACE flag and position itself in
1082  * replace_bpf_fd attribute. Old program at this position will be released.
1083  *
1084  * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1085  * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1086  * Ex1:
1087  * cgrp1 (MULTI progs A, B) ->
1088  *    cgrp2 (OVERRIDE prog C) ->
1089  *      cgrp3 (MULTI prog D) ->
1090  *        cgrp4 (OVERRIDE prog E) ->
1091  *          cgrp5 (NONE prog F)
1092  * the event in cgrp5 triggers execution of F,D,A,B in that order.
1093  * if prog F is detached, the execution is E,D,A,B
1094  * if prog F and D are detached, the execution is E,A,B
1095  * if prog F, E and D are detached, the execution is C,A,B
1096  *
1097  * All eligible programs are executed regardless of return code from
1098  * earlier programs.
1099  */
1100 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
1101 #define BPF_F_ALLOW_MULTI	(1U << 1)
1102 #define BPF_F_REPLACE		(1U << 2)
1103 
1104 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1105  * verifier will perform strict alignment checking as if the kernel
1106  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1107  * and NET_IP_ALIGN defined to 2.
1108  */
1109 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
1110 
1111 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1112  * verifier will allow any alignment whatsoever.  On platforms
1113  * with strict alignment requirements for loads ands stores (such
1114  * as sparc and mips) the verifier validates that all loads and
1115  * stores provably follow this requirement.  This flag turns that
1116  * checking and enforcement off.
1117  *
1118  * It is mostly used for testing when we want to validate the
1119  * context and memory access aspects of the verifier, but because
1120  * of an unaligned access the alignment check would trigger before
1121  * the one we are interested in.
1122  */
1123 #define BPF_F_ANY_ALIGNMENT	(1U << 1)
1124 
1125 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1126  * Verifier does sub-register def/use analysis and identifies instructions whose
1127  * def only matters for low 32-bit, high 32-bit is never referenced later
1128  * through implicit zero extension. Therefore verifier notifies JIT back-ends
1129  * that it is safe to ignore clearing high 32-bit for these instructions. This
1130  * saves some back-ends a lot of code-gen. However such optimization is not
1131  * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1132  * hence hasn't used verifier's analysis result. But, we really want to have a
1133  * way to be able to verify the correctness of the described optimization on
1134  * x86_64 on which testsuites are frequently exercised.
1135  *
1136  * So, this flag is introduced. Once it is set, verifier will randomize high
1137  * 32-bit for those instructions who has been identified as safe to ignore them.
1138  * Then, if verifier is not doing correct analysis, such randomization will
1139  * regress tests to expose bugs.
1140  */
1141 #define BPF_F_TEST_RND_HI32	(1U << 2)
1142 
1143 /* The verifier internal test flag. Behavior is undefined */
1144 #define BPF_F_TEST_STATE_FREQ	(1U << 3)
1145 
1146 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1147  * restrict map and helper usage for such programs. Sleepable BPF programs can
1148  * only be attached to hooks where kernel execution context allows sleeping.
1149  * Such programs are allowed to use helpers that may sleep like
1150  * bpf_copy_from_user().
1151  */
1152 #define BPF_F_SLEEPABLE		(1U << 4)
1153 
1154 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1155  * fully support xdp frags.
1156  */
1157 #define BPF_F_XDP_HAS_FRAGS	(1U << 5)
1158 
1159 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1160  * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1161  */
1162 #define BPF_F_KPROBE_MULTI_RETURN	(1U << 0)
1163 
1164 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1165  * the following extensions:
1166  *
1167  * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
1168  * insn[0].imm:      map fd or fd_idx
1169  * insn[1].imm:      0
1170  * insn[0].off:      0
1171  * insn[1].off:      0
1172  * ldimm64 rewrite:  address of map
1173  * verifier type:    CONST_PTR_TO_MAP
1174  */
1175 #define BPF_PSEUDO_MAP_FD	1
1176 #define BPF_PSEUDO_MAP_IDX	5
1177 
1178 /* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
1179  * insn[0].imm:      map fd or fd_idx
1180  * insn[1].imm:      offset into value
1181  * insn[0].off:      0
1182  * insn[1].off:      0
1183  * ldimm64 rewrite:  address of map[0]+offset
1184  * verifier type:    PTR_TO_MAP_VALUE
1185  */
1186 #define BPF_PSEUDO_MAP_VALUE		2
1187 #define BPF_PSEUDO_MAP_IDX_VALUE	6
1188 
1189 /* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
1190  * insn[0].imm:      kernel btd id of VAR
1191  * insn[1].imm:      0
1192  * insn[0].off:      0
1193  * insn[1].off:      0
1194  * ldimm64 rewrite:  address of the kernel variable
1195  * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1196  *                   is struct/union.
1197  */
1198 #define BPF_PSEUDO_BTF_ID	3
1199 /* insn[0].src_reg:  BPF_PSEUDO_FUNC
1200  * insn[0].imm:      insn offset to the func
1201  * insn[1].imm:      0
1202  * insn[0].off:      0
1203  * insn[1].off:      0
1204  * ldimm64 rewrite:  address of the function
1205  * verifier type:    PTR_TO_FUNC.
1206  */
1207 #define BPF_PSEUDO_FUNC		4
1208 
1209 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1210  * offset to another bpf function
1211  */
1212 #define BPF_PSEUDO_CALL		1
1213 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1214  * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1215  */
1216 #define BPF_PSEUDO_KFUNC_CALL	2
1217 
1218 /* flags for BPF_MAP_UPDATE_ELEM command */
1219 enum {
1220 	BPF_ANY		= 0, /* create new element or update existing */
1221 	BPF_NOEXIST	= 1, /* create new element if it didn't exist */
1222 	BPF_EXIST	= 2, /* update existing element */
1223 	BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */
1224 };
1225 
1226 /* flags for BPF_MAP_CREATE command */
1227 enum {
1228 	BPF_F_NO_PREALLOC	= (1U << 0),
1229 /* Instead of having one common LRU list in the
1230  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1231  * which can scale and perform better.
1232  * Note, the LRU nodes (including free nodes) cannot be moved
1233  * across different LRU lists.
1234  */
1235 	BPF_F_NO_COMMON_LRU	= (1U << 1),
1236 /* Specify numa node during map creation */
1237 	BPF_F_NUMA_NODE		= (1U << 2),
1238 
1239 /* Flags for accessing BPF object from syscall side. */
1240 	BPF_F_RDONLY		= (1U << 3),
1241 	BPF_F_WRONLY		= (1U << 4),
1242 
1243 /* Flag for stack_map, store build_id+offset instead of pointer */
1244 	BPF_F_STACK_BUILD_ID	= (1U << 5),
1245 
1246 /* Zero-initialize hash function seed. This should only be used for testing. */
1247 	BPF_F_ZERO_SEED		= (1U << 6),
1248 
1249 /* Flags for accessing BPF object from program side. */
1250 	BPF_F_RDONLY_PROG	= (1U << 7),
1251 	BPF_F_WRONLY_PROG	= (1U << 8),
1252 
1253 /* Clone map from listener for newly accepted socket */
1254 	BPF_F_CLONE		= (1U << 9),
1255 
1256 /* Enable memory-mapping BPF map */
1257 	BPF_F_MMAPABLE		= (1U << 10),
1258 
1259 /* Share perf_event among processes */
1260 	BPF_F_PRESERVE_ELEMS	= (1U << 11),
1261 
1262 /* Create a map that is suitable to be an inner map with dynamic max entries */
1263 	BPF_F_INNER_MAP		= (1U << 12),
1264 };
1265 
1266 /* Flags for BPF_PROG_QUERY. */
1267 
1268 /* Query effective (directly attached + inherited from ancestor cgroups)
1269  * programs that will be executed for events within a cgroup.
1270  * attach_flags with this flag are always returned 0.
1271  */
1272 #define BPF_F_QUERY_EFFECTIVE	(1U << 0)
1273 
1274 /* Flags for BPF_PROG_TEST_RUN */
1275 
1276 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1277 #define BPF_F_TEST_RUN_ON_CPU	(1U << 0)
1278 /* If set, XDP frames will be transmitted after processing */
1279 #define BPF_F_TEST_XDP_LIVE_FRAMES	(1U << 1)
1280 
1281 /* type for BPF_ENABLE_STATS */
1282 enum bpf_stats_type {
1283 	/* enabled run_time_ns and run_cnt */
1284 	BPF_STATS_RUN_TIME = 0,
1285 };
1286 
1287 enum bpf_stack_build_id_status {
1288 	/* user space need an empty entry to identify end of a trace */
1289 	BPF_STACK_BUILD_ID_EMPTY = 0,
1290 	/* with valid build_id and offset */
1291 	BPF_STACK_BUILD_ID_VALID = 1,
1292 	/* couldn't get build_id, fallback to ip */
1293 	BPF_STACK_BUILD_ID_IP = 2,
1294 };
1295 
1296 #define BPF_BUILD_ID_SIZE 20
1297 struct bpf_stack_build_id {
1298 	__s32		status;
1299 	unsigned char	build_id[BPF_BUILD_ID_SIZE];
1300 	union {
1301 		__u64	offset;
1302 		__u64	ip;
1303 	};
1304 };
1305 
1306 #define BPF_OBJ_NAME_LEN 16U
1307 
1308 union bpf_attr {
1309 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
1310 		__u32	map_type;	/* one of enum bpf_map_type */
1311 		__u32	key_size;	/* size of key in bytes */
1312 		__u32	value_size;	/* size of value in bytes */
1313 		__u32	max_entries;	/* max number of entries in a map */
1314 		__u32	map_flags;	/* BPF_MAP_CREATE related
1315 					 * flags defined above.
1316 					 */
1317 		__u32	inner_map_fd;	/* fd pointing to the inner map */
1318 		__u32	numa_node;	/* numa node (effective only if
1319 					 * BPF_F_NUMA_NODE is set).
1320 					 */
1321 		char	map_name[BPF_OBJ_NAME_LEN];
1322 		__u32	map_ifindex;	/* ifindex of netdev to create on */
1323 		__u32	btf_fd;		/* fd pointing to a BTF type data */
1324 		__u32	btf_key_type_id;	/* BTF type_id of the key */
1325 		__u32	btf_value_type_id;	/* BTF type_id of the value */
1326 		__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1327 						   * struct stored as the
1328 						   * map value
1329 						   */
1330 		/* Any per-map-type extra fields
1331 		 *
1332 		 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1333 		 * number of hash functions (if 0, the bloom filter will default
1334 		 * to using 5 hash functions).
1335 		 */
1336 		__u64	map_extra;
1337 	};
1338 
1339 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1340 		__u32		map_fd;
1341 		__aligned_u64	key;
1342 		union {
1343 			__aligned_u64 value;
1344 			__aligned_u64 next_key;
1345 		};
1346 		__u64		flags;
1347 	};
1348 
1349 	struct { /* struct used by BPF_MAP_*_BATCH commands */
1350 		__aligned_u64	in_batch;	/* start batch,
1351 						 * NULL to start from beginning
1352 						 */
1353 		__aligned_u64	out_batch;	/* output: next start batch */
1354 		__aligned_u64	keys;
1355 		__aligned_u64	values;
1356 		__u32		count;		/* input/output:
1357 						 * input: # of key/value
1358 						 * elements
1359 						 * output: # of filled elements
1360 						 */
1361 		__u32		map_fd;
1362 		__u64		elem_flags;
1363 		__u64		flags;
1364 	} batch;
1365 
1366 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
1367 		__u32		prog_type;	/* one of enum bpf_prog_type */
1368 		__u32		insn_cnt;
1369 		__aligned_u64	insns;
1370 		__aligned_u64	license;
1371 		__u32		log_level;	/* verbosity level of verifier */
1372 		__u32		log_size;	/* size of user buffer */
1373 		__aligned_u64	log_buf;	/* user supplied buffer */
1374 		__u32		kern_version;	/* not used */
1375 		__u32		prog_flags;
1376 		char		prog_name[BPF_OBJ_NAME_LEN];
1377 		__u32		prog_ifindex;	/* ifindex of netdev to prep for */
1378 		/* For some prog types expected attach type must be known at
1379 		 * load time to verify attach type specific parts of prog
1380 		 * (context accesses, allowed helpers, etc).
1381 		 */
1382 		__u32		expected_attach_type;
1383 		__u32		prog_btf_fd;	/* fd pointing to BTF type data */
1384 		__u32		func_info_rec_size;	/* userspace bpf_func_info size */
1385 		__aligned_u64	func_info;	/* func info */
1386 		__u32		func_info_cnt;	/* number of bpf_func_info records */
1387 		__u32		line_info_rec_size;	/* userspace bpf_line_info size */
1388 		__aligned_u64	line_info;	/* line info */
1389 		__u32		line_info_cnt;	/* number of bpf_line_info records */
1390 		__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */
1391 		union {
1392 			/* valid prog_fd to attach to bpf prog */
1393 			__u32		attach_prog_fd;
1394 			/* or valid module BTF object fd or 0 to attach to vmlinux */
1395 			__u32		attach_btf_obj_fd;
1396 		};
1397 		__u32		core_relo_cnt;	/* number of bpf_core_relo */
1398 		__aligned_u64	fd_array;	/* array of FDs */
1399 		__aligned_u64	core_relos;
1400 		__u32		core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1401 	};
1402 
1403 	struct { /* anonymous struct used by BPF_OBJ_* commands */
1404 		__aligned_u64	pathname;
1405 		__u32		bpf_fd;
1406 		__u32		file_flags;
1407 	};
1408 
1409 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1410 		__u32		target_fd;	/* container object to attach to */
1411 		__u32		attach_bpf_fd;	/* eBPF program to attach */
1412 		__u32		attach_type;
1413 		__u32		attach_flags;
1414 		__u32		replace_bpf_fd;	/* previously attached eBPF
1415 						 * program to replace if
1416 						 * BPF_F_REPLACE is used
1417 						 */
1418 	};
1419 
1420 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1421 		__u32		prog_fd;
1422 		__u32		retval;
1423 		__u32		data_size_in;	/* input: len of data_in */
1424 		__u32		data_size_out;	/* input/output: len of data_out
1425 						 *   returns ENOSPC if data_out
1426 						 *   is too small.
1427 						 */
1428 		__aligned_u64	data_in;
1429 		__aligned_u64	data_out;
1430 		__u32		repeat;
1431 		__u32		duration;
1432 		__u32		ctx_size_in;	/* input: len of ctx_in */
1433 		__u32		ctx_size_out;	/* input/output: len of ctx_out
1434 						 *   returns ENOSPC if ctx_out
1435 						 *   is too small.
1436 						 */
1437 		__aligned_u64	ctx_in;
1438 		__aligned_u64	ctx_out;
1439 		__u32		flags;
1440 		__u32		cpu;
1441 		__u32		batch_size;
1442 	} test;
1443 
1444 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
1445 		union {
1446 			__u32		start_id;
1447 			__u32		prog_id;
1448 			__u32		map_id;
1449 			__u32		btf_id;
1450 			__u32		link_id;
1451 		};
1452 		__u32		next_id;
1453 		__u32		open_flags;
1454 	};
1455 
1456 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1457 		__u32		bpf_fd;
1458 		__u32		info_len;
1459 		__aligned_u64	info;
1460 	} info;
1461 
1462 	struct { /* anonymous struct used by BPF_PROG_QUERY command */
1463 		__u32		target_fd;	/* container object to query */
1464 		__u32		attach_type;
1465 		__u32		query_flags;
1466 		__u32		attach_flags;
1467 		__aligned_u64	prog_ids;
1468 		__u32		prog_cnt;
1469 		/* output: per-program attach_flags.
1470 		 * not allowed to be set during effective query.
1471 		 */
1472 		__aligned_u64	prog_attach_flags;
1473 	} query;
1474 
1475 	struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1476 		__u64 name;
1477 		__u32 prog_fd;
1478 	} raw_tracepoint;
1479 
1480 	struct { /* anonymous struct for BPF_BTF_LOAD */
1481 		__aligned_u64	btf;
1482 		__aligned_u64	btf_log_buf;
1483 		__u32		btf_size;
1484 		__u32		btf_log_size;
1485 		__u32		btf_log_level;
1486 	};
1487 
1488 	struct {
1489 		__u32		pid;		/* input: pid */
1490 		__u32		fd;		/* input: fd */
1491 		__u32		flags;		/* input: flags */
1492 		__u32		buf_len;	/* input/output: buf len */
1493 		__aligned_u64	buf;		/* input/output:
1494 						 *   tp_name for tracepoint
1495 						 *   symbol for kprobe
1496 						 *   filename for uprobe
1497 						 */
1498 		__u32		prog_id;	/* output: prod_id */
1499 		__u32		fd_type;	/* output: BPF_FD_TYPE_* */
1500 		__u64		probe_offset;	/* output: probe_offset */
1501 		__u64		probe_addr;	/* output: probe_addr */
1502 	} task_fd_query;
1503 
1504 	struct { /* struct used by BPF_LINK_CREATE command */
1505 		__u32		prog_fd;	/* eBPF program to attach */
1506 		union {
1507 			__u32		target_fd;	/* object to attach to */
1508 			__u32		target_ifindex; /* target ifindex */
1509 		};
1510 		__u32		attach_type;	/* attach type */
1511 		__u32		flags;		/* extra flags */
1512 		union {
1513 			__u32		target_btf_id;	/* btf_id of target to attach to */
1514 			struct {
1515 				__aligned_u64	iter_info;	/* extra bpf_iter_link_info */
1516 				__u32		iter_info_len;	/* iter_info length */
1517 			};
1518 			struct {
1519 				/* black box user-provided value passed through
1520 				 * to BPF program at the execution time and
1521 				 * accessible through bpf_get_attach_cookie() BPF helper
1522 				 */
1523 				__u64		bpf_cookie;
1524 			} perf_event;
1525 			struct {
1526 				__u32		flags;
1527 				__u32		cnt;
1528 				__aligned_u64	syms;
1529 				__aligned_u64	addrs;
1530 				__aligned_u64	cookies;
1531 			} kprobe_multi;
1532 			struct {
1533 				/* this is overlaid with the target_btf_id above. */
1534 				__u32		target_btf_id;
1535 				/* black box user-provided value passed through
1536 				 * to BPF program at the execution time and
1537 				 * accessible through bpf_get_attach_cookie() BPF helper
1538 				 */
1539 				__u64		cookie;
1540 			} tracing;
1541 		};
1542 	} link_create;
1543 
1544 	struct { /* struct used by BPF_LINK_UPDATE command */
1545 		__u32		link_fd;	/* link fd */
1546 		/* new program fd to update link with */
1547 		__u32		new_prog_fd;
1548 		__u32		flags;		/* extra flags */
1549 		/* expected link's program fd; is specified only if
1550 		 * BPF_F_REPLACE flag is set in flags */
1551 		__u32		old_prog_fd;
1552 	} link_update;
1553 
1554 	struct {
1555 		__u32		link_fd;
1556 	} link_detach;
1557 
1558 	struct { /* struct used by BPF_ENABLE_STATS command */
1559 		__u32		type;
1560 	} enable_stats;
1561 
1562 	struct { /* struct used by BPF_ITER_CREATE command */
1563 		__u32		link_fd;
1564 		__u32		flags;
1565 	} iter_create;
1566 
1567 	struct { /* struct used by BPF_PROG_BIND_MAP command */
1568 		__u32		prog_fd;
1569 		__u32		map_fd;
1570 		__u32		flags;		/* extra flags */
1571 	} prog_bind_map;
1572 
1573 } __attribute__((aligned(8)));
1574 
1575 /* The description below is an attempt at providing documentation to eBPF
1576  * developers about the multiple available eBPF helper functions. It can be
1577  * parsed and used to produce a manual page. The workflow is the following,
1578  * and requires the rst2man utility:
1579  *
1580  *     $ ./scripts/bpf_doc.py \
1581  *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1582  *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1583  *     $ man /tmp/bpf-helpers.7
1584  *
1585  * Note that in order to produce this external documentation, some RST
1586  * formatting is used in the descriptions to get "bold" and "italics" in
1587  * manual pages. Also note that the few trailing white spaces are
1588  * intentional, removing them would break paragraphs for rst2man.
1589  *
1590  * Start of BPF helper function descriptions:
1591  *
1592  * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1593  * 	Description
1594  * 		Perform a lookup in *map* for an entry associated to *key*.
1595  * 	Return
1596  * 		Map value associated to *key*, or **NULL** if no entry was
1597  * 		found.
1598  *
1599  * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1600  * 	Description
1601  * 		Add or update the value of the entry associated to *key* in
1602  * 		*map* with *value*. *flags* is one of:
1603  *
1604  * 		**BPF_NOEXIST**
1605  * 			The entry for *key* must not exist in the map.
1606  * 		**BPF_EXIST**
1607  * 			The entry for *key* must already exist in the map.
1608  * 		**BPF_ANY**
1609  * 			No condition on the existence of the entry for *key*.
1610  *
1611  * 		Flag value **BPF_NOEXIST** cannot be used for maps of types
1612  * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
1613  * 		elements always exist), the helper would return an error.
1614  * 	Return
1615  * 		0 on success, or a negative error in case of failure.
1616  *
1617  * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1618  * 	Description
1619  * 		Delete entry with *key* from *map*.
1620  * 	Return
1621  * 		0 on success, or a negative error in case of failure.
1622  *
1623  * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1624  * 	Description
1625  * 		For tracing programs, safely attempt to read *size* bytes from
1626  * 		kernel space address *unsafe_ptr* and store the data in *dst*.
1627  *
1628  * 		Generally, use **bpf_probe_read_user**\ () or
1629  * 		**bpf_probe_read_kernel**\ () instead.
1630  * 	Return
1631  * 		0 on success, or a negative error in case of failure.
1632  *
1633  * u64 bpf_ktime_get_ns(void)
1634  * 	Description
1635  * 		Return the time elapsed since system boot, in nanoseconds.
1636  * 		Does not include time the system was suspended.
1637  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1638  * 	Return
1639  * 		Current *ktime*.
1640  *
1641  * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1642  * 	Description
1643  * 		This helper is a "printk()-like" facility for debugging. It
1644  * 		prints a message defined by format *fmt* (of size *fmt_size*)
1645  * 		to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1646  * 		available. It can take up to three additional **u64**
1647  * 		arguments (as an eBPF helpers, the total number of arguments is
1648  * 		limited to five).
1649  *
1650  * 		Each time the helper is called, it appends a line to the trace.
1651  * 		Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1652  * 		open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1653  * 		The format of the trace is customizable, and the exact output
1654  * 		one will get depends on the options set in
1655  * 		*\/sys/kernel/debug/tracing/trace_options* (see also the
1656  * 		*README* file under the same directory). However, it usually
1657  * 		defaults to something like:
1658  *
1659  * 		::
1660  *
1661  * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1662  *
1663  * 		In the above:
1664  *
1665  * 			* ``telnet`` is the name of the current task.
1666  * 			* ``470`` is the PID of the current task.
1667  * 			* ``001`` is the CPU number on which the task is
1668  * 			  running.
1669  * 			* In ``.N..``, each character refers to a set of
1670  * 			  options (whether irqs are enabled, scheduling
1671  * 			  options, whether hard/softirqs are running, level of
1672  * 			  preempt_disabled respectively). **N** means that
1673  * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1674  * 			  are set.
1675  * 			* ``419421.045894`` is a timestamp.
1676  * 			* ``0x00000001`` is a fake value used by BPF for the
1677  * 			  instruction pointer register.
1678  * 			* ``<formatted msg>`` is the message formatted with
1679  * 			  *fmt*.
1680  *
1681  * 		The conversion specifiers supported by *fmt* are similar, but
1682  * 		more limited than for printk(). They are **%d**, **%i**,
1683  * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1684  * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1685  * 		of field, padding with zeroes, etc.) is available, and the
1686  * 		helper will return **-EINVAL** (but print nothing) if it
1687  * 		encounters an unknown specifier.
1688  *
1689  * 		Also, note that **bpf_trace_printk**\ () is slow, and should
1690  * 		only be used for debugging purposes. For this reason, a notice
1691  * 		block (spanning several lines) is printed to kernel logs and
1692  * 		states that the helper should not be used "for production use"
1693  * 		the first time this helper is used (or more precisely, when
1694  * 		**trace_printk**\ () buffers are allocated). For passing values
1695  * 		to user space, perf events should be preferred.
1696  * 	Return
1697  * 		The number of bytes written to the buffer, or a negative error
1698  * 		in case of failure.
1699  *
1700  * u32 bpf_get_prandom_u32(void)
1701  * 	Description
1702  * 		Get a pseudo-random number.
1703  *
1704  * 		From a security point of view, this helper uses its own
1705  * 		pseudo-random internal state, and cannot be used to infer the
1706  * 		seed of other random functions in the kernel. However, it is
1707  * 		essential to note that the generator used by the helper is not
1708  * 		cryptographically secure.
1709  * 	Return
1710  * 		A random 32-bit unsigned value.
1711  *
1712  * u32 bpf_get_smp_processor_id(void)
1713  * 	Description
1714  * 		Get the SMP (symmetric multiprocessing) processor id. Note that
1715  * 		all programs run with migration disabled, which means that the
1716  * 		SMP processor id is stable during all the execution of the
1717  * 		program.
1718  * 	Return
1719  * 		The SMP id of the processor running the program.
1720  *
1721  * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1722  * 	Description
1723  * 		Store *len* bytes from address *from* into the packet
1724  * 		associated to *skb*, at *offset*. *flags* are a combination of
1725  * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1726  * 		checksum for the packet after storing the bytes) and
1727  * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1728  * 		**->swhash** and *skb*\ **->l4hash** to 0).
1729  *
1730  * 		A call to this helper is susceptible to change the underlying
1731  * 		packet buffer. Therefore, at load time, all checks on pointers
1732  * 		previously done by the verifier are invalidated and must be
1733  * 		performed again, if the helper is used in combination with
1734  * 		direct packet access.
1735  * 	Return
1736  * 		0 on success, or a negative error in case of failure.
1737  *
1738  * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1739  * 	Description
1740  * 		Recompute the layer 3 (e.g. IP) checksum for the packet
1741  * 		associated to *skb*. Computation is incremental, so the helper
1742  * 		must know the former value of the header field that was
1743  * 		modified (*from*), the new value of this field (*to*), and the
1744  * 		number of bytes (2 or 4) for this field, stored in *size*.
1745  * 		Alternatively, it is possible to store the difference between
1746  * 		the previous and the new values of the header field in *to*, by
1747  * 		setting *from* and *size* to 0. For both methods, *offset*
1748  * 		indicates the location of the IP checksum within the packet.
1749  *
1750  * 		This helper works in combination with **bpf_csum_diff**\ (),
1751  * 		which does not update the checksum in-place, but offers more
1752  * 		flexibility and can handle sizes larger than 2 or 4 for the
1753  * 		checksum to update.
1754  *
1755  * 		A call to this helper is susceptible to change the underlying
1756  * 		packet buffer. Therefore, at load time, all checks on pointers
1757  * 		previously done by the verifier are invalidated and must be
1758  * 		performed again, if the helper is used in combination with
1759  * 		direct packet access.
1760  * 	Return
1761  * 		0 on success, or a negative error in case of failure.
1762  *
1763  * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1764  * 	Description
1765  * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1766  * 		packet associated to *skb*. Computation is incremental, so the
1767  * 		helper must know the former value of the header field that was
1768  * 		modified (*from*), the new value of this field (*to*), and the
1769  * 		number of bytes (2 or 4) for this field, stored on the lowest
1770  * 		four bits of *flags*. Alternatively, it is possible to store
1771  * 		the difference between the previous and the new values of the
1772  * 		header field in *to*, by setting *from* and the four lowest
1773  * 		bits of *flags* to 0. For both methods, *offset* indicates the
1774  * 		location of the IP checksum within the packet. In addition to
1775  * 		the size of the field, *flags* can be added (bitwise OR) actual
1776  * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1777  * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1778  * 		for updates resulting in a null checksum the value is set to
1779  * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1780  * 		the checksum is to be computed against a pseudo-header.
1781  *
1782  * 		This helper works in combination with **bpf_csum_diff**\ (),
1783  * 		which does not update the checksum in-place, but offers more
1784  * 		flexibility and can handle sizes larger than 2 or 4 for the
1785  * 		checksum to update.
1786  *
1787  * 		A call to this helper is susceptible to change the underlying
1788  * 		packet buffer. Therefore, at load time, all checks on pointers
1789  * 		previously done by the verifier are invalidated and must be
1790  * 		performed again, if the helper is used in combination with
1791  * 		direct packet access.
1792  * 	Return
1793  * 		0 on success, or a negative error in case of failure.
1794  *
1795  * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1796  * 	Description
1797  * 		This special helper is used to trigger a "tail call", or in
1798  * 		other words, to jump into another eBPF program. The same stack
1799  * 		frame is used (but values on stack and in registers for the
1800  * 		caller are not accessible to the callee). This mechanism allows
1801  * 		for program chaining, either for raising the maximum number of
1802  * 		available eBPF instructions, or to execute given programs in
1803  * 		conditional blocks. For security reasons, there is an upper
1804  * 		limit to the number of successive tail calls that can be
1805  * 		performed.
1806  *
1807  * 		Upon call of this helper, the program attempts to jump into a
1808  * 		program referenced at index *index* in *prog_array_map*, a
1809  * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1810  * 		*ctx*, a pointer to the context.
1811  *
1812  * 		If the call succeeds, the kernel immediately runs the first
1813  * 		instruction of the new program. This is not a function call,
1814  * 		and it never returns to the previous program. If the call
1815  * 		fails, then the helper has no effect, and the caller continues
1816  * 		to run its subsequent instructions. A call can fail if the
1817  * 		destination program for the jump does not exist (i.e. *index*
1818  * 		is superior to the number of entries in *prog_array_map*), or
1819  * 		if the maximum number of tail calls has been reached for this
1820  * 		chain of programs. This limit is defined in the kernel by the
1821  * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1822  *		which is currently set to 33.
1823  * 	Return
1824  * 		0 on success, or a negative error in case of failure.
1825  *
1826  * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1827  * 	Description
1828  * 		Clone and redirect the packet associated to *skb* to another
1829  * 		net device of index *ifindex*. Both ingress and egress
1830  * 		interfaces can be used for redirection. The **BPF_F_INGRESS**
1831  * 		value in *flags* is used to make the distinction (ingress path
1832  * 		is selected if the flag is present, egress path otherwise).
1833  * 		This is the only flag supported for now.
1834  *
1835  * 		In comparison with **bpf_redirect**\ () helper,
1836  * 		**bpf_clone_redirect**\ () has the associated cost of
1837  * 		duplicating the packet buffer, but this can be executed out of
1838  * 		the eBPF program. Conversely, **bpf_redirect**\ () is more
1839  * 		efficient, but it is handled through an action code where the
1840  * 		redirection happens only after the eBPF program has returned.
1841  *
1842  * 		A call to this helper is susceptible to change the underlying
1843  * 		packet buffer. Therefore, at load time, all checks on pointers
1844  * 		previously done by the verifier are invalidated and must be
1845  * 		performed again, if the helper is used in combination with
1846  * 		direct packet access.
1847  * 	Return
1848  * 		0 on success, or a negative error in case of failure.
1849  *
1850  * u64 bpf_get_current_pid_tgid(void)
1851  * 	Description
1852  * 		Get the current pid and tgid.
1853  * 	Return
1854  * 		A 64-bit integer containing the current tgid and pid, and
1855  * 		created as such:
1856  * 		*current_task*\ **->tgid << 32 \|**
1857  * 		*current_task*\ **->pid**.
1858  *
1859  * u64 bpf_get_current_uid_gid(void)
1860  * 	Description
1861  * 		Get the current uid and gid.
1862  * 	Return
1863  * 		A 64-bit integer containing the current GID and UID, and
1864  * 		created as such: *current_gid* **<< 32 \|** *current_uid*.
1865  *
1866  * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1867  * 	Description
1868  * 		Copy the **comm** attribute of the current task into *buf* of
1869  * 		*size_of_buf*. The **comm** attribute contains the name of
1870  * 		the executable (excluding the path) for the current task. The
1871  * 		*size_of_buf* must be strictly positive. On success, the
1872  * 		helper makes sure that the *buf* is NUL-terminated. On failure,
1873  * 		it is filled with zeroes.
1874  * 	Return
1875  * 		0 on success, or a negative error in case of failure.
1876  *
1877  * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1878  * 	Description
1879  * 		Retrieve the classid for the current task, i.e. for the net_cls
1880  * 		cgroup to which *skb* belongs.
1881  *
1882  * 		This helper can be used on TC egress path, but not on ingress.
1883  *
1884  * 		The net_cls cgroup provides an interface to tag network packets
1885  * 		based on a user-provided identifier for all traffic coming from
1886  * 		the tasks belonging to the related cgroup. See also the related
1887  * 		kernel documentation, available from the Linux sources in file
1888  * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1889  *
1890  * 		The Linux kernel has two versions for cgroups: there are
1891  * 		cgroups v1 and cgroups v2. Both are available to users, who can
1892  * 		use a mixture of them, but note that the net_cls cgroup is for
1893  * 		cgroup v1 only. This makes it incompatible with BPF programs
1894  * 		run on cgroups, which is a cgroup-v2-only feature (a socket can
1895  * 		only hold data for one version of cgroups at a time).
1896  *
1897  * 		This helper is only available is the kernel was compiled with
1898  * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1899  * 		"**y**" or to "**m**".
1900  * 	Return
1901  * 		The classid, or 0 for the default unconfigured classid.
1902  *
1903  * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1904  * 	Description
1905  * 		Push a *vlan_tci* (VLAN tag control information) of protocol
1906  * 		*vlan_proto* to the packet associated to *skb*, then update
1907  * 		the checksum. Note that if *vlan_proto* is different from
1908  * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1909  * 		be **ETH_P_8021Q**.
1910  *
1911  * 		A call to this helper is susceptible to change the underlying
1912  * 		packet buffer. Therefore, at load time, all checks on pointers
1913  * 		previously done by the verifier are invalidated and must be
1914  * 		performed again, if the helper is used in combination with
1915  * 		direct packet access.
1916  * 	Return
1917  * 		0 on success, or a negative error in case of failure.
1918  *
1919  * long bpf_skb_vlan_pop(struct sk_buff *skb)
1920  * 	Description
1921  * 		Pop a VLAN header from the packet associated to *skb*.
1922  *
1923  * 		A call to this helper is susceptible to change the underlying
1924  * 		packet buffer. Therefore, at load time, all checks on pointers
1925  * 		previously done by the verifier are invalidated and must be
1926  * 		performed again, if the helper is used in combination with
1927  * 		direct packet access.
1928  * 	Return
1929  * 		0 on success, or a negative error in case of failure.
1930  *
1931  * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1932  * 	Description
1933  * 		Get tunnel metadata. This helper takes a pointer *key* to an
1934  * 		empty **struct bpf_tunnel_key** of **size**, that will be
1935  * 		filled with tunnel metadata for the packet associated to *skb*.
1936  * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1937  * 		indicates that the tunnel is based on IPv6 protocol instead of
1938  * 		IPv4.
1939  *
1940  * 		The **struct bpf_tunnel_key** is an object that generalizes the
1941  * 		principal parameters used by various tunneling protocols into a
1942  * 		single struct. This way, it can be used to easily make a
1943  * 		decision based on the contents of the encapsulation header,
1944  * 		"summarized" in this struct. In particular, it holds the IP
1945  * 		address of the remote end (IPv4 or IPv6, depending on the case)
1946  * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1947  * 		this struct exposes the *key*\ **->tunnel_id**, which is
1948  * 		generally mapped to a VNI (Virtual Network Identifier), making
1949  * 		it programmable together with the **bpf_skb_set_tunnel_key**\
1950  * 		() helper.
1951  *
1952  * 		Let's imagine that the following code is part of a program
1953  * 		attached to the TC ingress interface, on one end of a GRE
1954  * 		tunnel, and is supposed to filter out all messages coming from
1955  * 		remote ends with IPv4 address other than 10.0.0.1:
1956  *
1957  * 		::
1958  *
1959  * 			int ret;
1960  * 			struct bpf_tunnel_key key = {};
1961  *
1962  * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1963  * 			if (ret < 0)
1964  * 				return TC_ACT_SHOT;	// drop packet
1965  *
1966  * 			if (key.remote_ipv4 != 0x0a000001)
1967  * 				return TC_ACT_SHOT;	// drop packet
1968  *
1969  * 			return TC_ACT_OK;		// accept packet
1970  *
1971  * 		This interface can also be used with all encapsulation devices
1972  * 		that can operate in "collect metadata" mode: instead of having
1973  * 		one network device per specific configuration, the "collect
1974  * 		metadata" mode only requires a single device where the
1975  * 		configuration can be extracted from this helper.
1976  *
1977  * 		This can be used together with various tunnels such as VXLan,
1978  * 		Geneve, GRE or IP in IP (IPIP).
1979  * 	Return
1980  * 		0 on success, or a negative error in case of failure.
1981  *
1982  * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1983  * 	Description
1984  * 		Populate tunnel metadata for packet associated to *skb.* The
1985  * 		tunnel metadata is set to the contents of *key*, of *size*. The
1986  * 		*flags* can be set to a combination of the following values:
1987  *
1988  * 		**BPF_F_TUNINFO_IPV6**
1989  * 			Indicate that the tunnel is based on IPv6 protocol
1990  * 			instead of IPv4.
1991  * 		**BPF_F_ZERO_CSUM_TX**
1992  * 			For IPv4 packets, add a flag to tunnel metadata
1993  * 			indicating that checksum computation should be skipped
1994  * 			and checksum set to zeroes.
1995  * 		**BPF_F_DONT_FRAGMENT**
1996  * 			Add a flag to tunnel metadata indicating that the
1997  * 			packet should not be fragmented.
1998  * 		**BPF_F_SEQ_NUMBER**
1999  * 			Add a flag to tunnel metadata indicating that a
2000  * 			sequence number should be added to tunnel header before
2001  * 			sending the packet. This flag was added for GRE
2002  * 			encapsulation, but might be used with other protocols
2003  * 			as well in the future.
2004  * 		**BPF_F_NO_TUNNEL_KEY**
2005  * 			Add a flag to tunnel metadata indicating that no tunnel
2006  * 			key should be set in the resulting tunnel header.
2007  *
2008  * 		Here is a typical usage on the transmit path:
2009  *
2010  * 		::
2011  *
2012  * 			struct bpf_tunnel_key key;
2013  * 			     populate key ...
2014  * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
2015  * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
2016  *
2017  * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()
2018  * 		helper for additional information.
2019  * 	Return
2020  * 		0 on success, or a negative error in case of failure.
2021  *
2022  * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
2023  * 	Description
2024  * 		Read the value of a perf event counter. This helper relies on a
2025  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
2026  * 		the perf event counter is selected when *map* is updated with
2027  * 		perf event file descriptors. The *map* is an array whose size
2028  * 		is the number of available CPUs, and each cell contains a value
2029  * 		relative to one CPU. The value to retrieve is indicated by
2030  * 		*flags*, that contains the index of the CPU to look up, masked
2031  * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2032  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2033  * 		current CPU should be retrieved.
2034  *
2035  * 		Note that before Linux 4.13, only hardware perf event can be
2036  * 		retrieved.
2037  *
2038  * 		Also, be aware that the newer helper
2039  * 		**bpf_perf_event_read_value**\ () is recommended over
2040  * 		**bpf_perf_event_read**\ () in general. The latter has some ABI
2041  * 		quirks where error and counter value are used as a return code
2042  * 		(which is wrong to do since ranges may overlap). This issue is
2043  * 		fixed with **bpf_perf_event_read_value**\ (), which at the same
2044  * 		time provides more features over the **bpf_perf_event_read**\
2045  * 		() interface. Please refer to the description of
2046  * 		**bpf_perf_event_read_value**\ () for details.
2047  * 	Return
2048  * 		The value of the perf event counter read from the map, or a
2049  * 		negative error code in case of failure.
2050  *
2051  * long bpf_redirect(u32 ifindex, u64 flags)
2052  * 	Description
2053  * 		Redirect the packet to another net device of index *ifindex*.
2054  * 		This helper is somewhat similar to **bpf_clone_redirect**\
2055  * 		(), except that the packet is not cloned, which provides
2056  * 		increased performance.
2057  *
2058  * 		Except for XDP, both ingress and egress interfaces can be used
2059  * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used
2060  * 		to make the distinction (ingress path is selected if the flag
2061  * 		is present, egress path otherwise). Currently, XDP only
2062  * 		supports redirection to the egress interface, and accepts no
2063  * 		flag at all.
2064  *
2065  * 		The same effect can also be attained with the more generic
2066  * 		**bpf_redirect_map**\ (), which uses a BPF map to store the
2067  * 		redirect target instead of providing it directly to the helper.
2068  * 	Return
2069  * 		For XDP, the helper returns **XDP_REDIRECT** on success or
2070  * 		**XDP_ABORTED** on error. For other program types, the values
2071  * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2072  * 		error.
2073  *
2074  * u32 bpf_get_route_realm(struct sk_buff *skb)
2075  * 	Description
2076  * 		Retrieve the realm or the route, that is to say the
2077  * 		**tclassid** field of the destination for the *skb*. The
2078  * 		identifier retrieved is a user-provided tag, similar to the
2079  * 		one used with the net_cls cgroup (see description for
2080  * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is
2081  * 		held by a route (a destination entry), not by a task.
2082  *
2083  * 		Retrieving this identifier works with the clsact TC egress hook
2084  * 		(see also **tc-bpf(8)**), or alternatively on conventional
2085  * 		classful egress qdiscs, but not on TC ingress path. In case of
2086  * 		clsact TC egress hook, this has the advantage that, internally,
2087  * 		the destination entry has not been dropped yet in the transmit
2088  * 		path. Therefore, the destination entry does not need to be
2089  * 		artificially held via **netif_keep_dst**\ () for a classful
2090  * 		qdisc until the *skb* is freed.
2091  *
2092  * 		This helper is available only if the kernel was compiled with
2093  * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.
2094  * 	Return
2095  * 		The realm of the route for the packet associated to *skb*, or 0
2096  * 		if none was found.
2097  *
2098  * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2099  * 	Description
2100  * 		Write raw *data* blob into a special BPF perf event held by
2101  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2102  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
2103  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2104  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2105  *
2106  * 		The *flags* are used to indicate the index in *map* for which
2107  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
2108  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2109  * 		to indicate that the index of the current CPU core should be
2110  * 		used.
2111  *
2112  * 		The value to write, of *size*, is passed through eBPF stack and
2113  * 		pointed by *data*.
2114  *
2115  * 		The context of the program *ctx* needs also be passed to the
2116  * 		helper.
2117  *
2118  * 		On user space, a program willing to read the values needs to
2119  * 		call **perf_event_open**\ () on the perf event (either for
2120  * 		one or for all CPUs) and to store the file descriptor into the
2121  * 		*map*. This must be done before the eBPF program can send data
2122  * 		into it. An example is available in file
2123  * 		*samples/bpf/trace_output_user.c* in the Linux kernel source
2124  * 		tree (the eBPF program counterpart is in
2125  * 		*samples/bpf/trace_output_kern.c*).
2126  *
2127  * 		**bpf_perf_event_output**\ () achieves better performance
2128  * 		than **bpf_trace_printk**\ () for sharing data with user
2129  * 		space, and is much better suitable for streaming data from eBPF
2130  * 		programs.
2131  *
2132  * 		Note that this helper is not restricted to tracing use cases
2133  * 		and can be used with programs attached to TC or XDP as well,
2134  * 		where it allows for passing data to user space listeners. Data
2135  * 		can be:
2136  *
2137  * 		* Only custom structs,
2138  * 		* Only the packet payload, or
2139  * 		* A combination of both.
2140  * 	Return
2141  * 		0 on success, or a negative error in case of failure.
2142  *
2143  * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2144  * 	Description
2145  * 		This helper was provided as an easy way to load data from a
2146  * 		packet. It can be used to load *len* bytes from *offset* from
2147  * 		the packet associated to *skb*, into the buffer pointed by
2148  * 		*to*.
2149  *
2150  * 		Since Linux 4.7, usage of this helper has mostly been replaced
2151  * 		by "direct packet access", enabling packet data to be
2152  * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2153  * 		pointing respectively to the first byte of packet data and to
2154  * 		the byte after the last byte of packet data. However, it
2155  * 		remains useful if one wishes to read large quantities of data
2156  * 		at once from a packet into the eBPF stack.
2157  * 	Return
2158  * 		0 on success, or a negative error in case of failure.
2159  *
2160  * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2161  * 	Description
2162  * 		Walk a user or a kernel stack and return its id. To achieve
2163  * 		this, the helper needs *ctx*, which is a pointer to the context
2164  * 		on which the tracing program is executed, and a pointer to a
2165  * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2166  *
2167  * 		The last argument, *flags*, holds the number of stack frames to
2168  * 		skip (from 0 to 255), masked with
2169  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2170  * 		a combination of the following flags:
2171  *
2172  * 		**BPF_F_USER_STACK**
2173  * 			Collect a user space stack instead of a kernel stack.
2174  * 		**BPF_F_FAST_STACK_CMP**
2175  * 			Compare stacks by hash only.
2176  * 		**BPF_F_REUSE_STACKID**
2177  * 			If two different stacks hash into the same *stackid*,
2178  * 			discard the old one.
2179  *
2180  * 		The stack id retrieved is a 32 bit long integer handle which
2181  * 		can be further combined with other data (including other stack
2182  * 		ids) and used as a key into maps. This can be useful for
2183  * 		generating a variety of graphs (such as flame graphs or off-cpu
2184  * 		graphs).
2185  *
2186  * 		For walking a stack, this helper is an improvement over
2187  * 		**bpf_probe_read**\ (), which can be used with unrolled loops
2188  * 		but is not efficient and consumes a lot of eBPF instructions.
2189  * 		Instead, **bpf_get_stackid**\ () can collect up to
2190  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2191  * 		this limit can be controlled with the **sysctl** program, and
2192  * 		that it should be manually increased in order to profile long
2193  * 		user stacks (such as stacks for Java programs). To do so, use:
2194  *
2195  * 		::
2196  *
2197  * 			# sysctl kernel.perf_event_max_stack=<new value>
2198  * 	Return
2199  * 		The positive or null stack id on success, or a negative error
2200  * 		in case of failure.
2201  *
2202  * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2203  * 	Description
2204  * 		Compute a checksum difference, from the raw buffer pointed by
2205  * 		*from*, of length *from_size* (that must be a multiple of 4),
2206  * 		towards the raw buffer pointed by *to*, of size *to_size*
2207  * 		(same remark). An optional *seed* can be added to the value
2208  * 		(this can be cascaded, the seed may come from a previous call
2209  * 		to the helper).
2210  *
2211  * 		This is flexible enough to be used in several ways:
2212  *
2213  * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to
2214  * 		  checksum, it can be used when pushing new data.
2215  * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to
2216  * 		  checksum, it can be used when removing data from a packet.
2217  * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2218  * 		  can be used to compute a diff. Note that *from_size* and
2219  * 		  *to_size* do not need to be equal.
2220  *
2221  * 		This helper can be used in combination with
2222  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2223  * 		which one can feed in the difference computed with
2224  * 		**bpf_csum_diff**\ ().
2225  * 	Return
2226  * 		The checksum result, or a negative error code in case of
2227  * 		failure.
2228  *
2229  * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2230  * 	Description
2231  * 		Retrieve tunnel options metadata for the packet associated to
2232  * 		*skb*, and store the raw tunnel option data to the buffer *opt*
2233  * 		of *size*.
2234  *
2235  * 		This helper can be used with encapsulation devices that can
2236  * 		operate in "collect metadata" mode (please refer to the related
2237  * 		note in the description of **bpf_skb_get_tunnel_key**\ () for
2238  * 		more details). A particular example where this can be used is
2239  * 		in combination with the Geneve encapsulation protocol, where it
2240  * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2241  * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from
2242  * 		the eBPF program. This allows for full customization of these
2243  * 		headers.
2244  * 	Return
2245  * 		The size of the option data retrieved.
2246  *
2247  * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2248  * 	Description
2249  * 		Set tunnel options metadata for the packet associated to *skb*
2250  * 		to the option data contained in the raw buffer *opt* of *size*.
2251  *
2252  * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2253  * 		helper for additional information.
2254  * 	Return
2255  * 		0 on success, or a negative error in case of failure.
2256  *
2257  * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2258  * 	Description
2259  * 		Change the protocol of the *skb* to *proto*. Currently
2260  * 		supported are transition from IPv4 to IPv6, and from IPv6 to
2261  * 		IPv4. The helper takes care of the groundwork for the
2262  * 		transition, including resizing the socket buffer. The eBPF
2263  * 		program is expected to fill the new headers, if any, via
2264  * 		**skb_store_bytes**\ () and to recompute the checksums with
2265  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2266  * 		(). The main case for this helper is to perform NAT64
2267  * 		operations out of an eBPF program.
2268  *
2269  * 		Internally, the GSO type is marked as dodgy so that headers are
2270  * 		checked and segments are recalculated by the GSO/GRO engine.
2271  * 		The size for GSO target is adapted as well.
2272  *
2273  * 		All values for *flags* are reserved for future usage, and must
2274  * 		be left at zero.
2275  *
2276  * 		A call to this helper is susceptible to change the underlying
2277  * 		packet buffer. Therefore, at load time, all checks on pointers
2278  * 		previously done by the verifier are invalidated and must be
2279  * 		performed again, if the helper is used in combination with
2280  * 		direct packet access.
2281  * 	Return
2282  * 		0 on success, or a negative error in case of failure.
2283  *
2284  * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2285  * 	Description
2286  * 		Change the packet type for the packet associated to *skb*. This
2287  * 		comes down to setting *skb*\ **->pkt_type** to *type*, except
2288  * 		the eBPF program does not have a write access to *skb*\
2289  * 		**->pkt_type** beside this helper. Using a helper here allows
2290  * 		for graceful handling of errors.
2291  *
2292  * 		The major use case is to change incoming *skb*s to
2293  * 		**PACKET_HOST** in a programmatic way instead of having to
2294  * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2295  * 		example.
2296  *
2297  * 		Note that *type* only allows certain values. At this time, they
2298  * 		are:
2299  *
2300  * 		**PACKET_HOST**
2301  * 			Packet is for us.
2302  * 		**PACKET_BROADCAST**
2303  * 			Send packet to all.
2304  * 		**PACKET_MULTICAST**
2305  * 			Send packet to group.
2306  * 		**PACKET_OTHERHOST**
2307  * 			Send packet to someone else.
2308  * 	Return
2309  * 		0 on success, or a negative error in case of failure.
2310  *
2311  * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2312  * 	Description
2313  * 		Check whether *skb* is a descendant of the cgroup2 held by
2314  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2315  * 	Return
2316  * 		The return value depends on the result of the test, and can be:
2317  *
2318  * 		* 0, if the *skb* failed the cgroup2 descendant test.
2319  * 		* 1, if the *skb* succeeded the cgroup2 descendant test.
2320  * 		* A negative error code, if an error occurred.
2321  *
2322  * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2323  * 	Description
2324  * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2325  * 		not set, in particular if the hash was cleared due to mangling,
2326  * 		recompute this hash. Later accesses to the hash can be done
2327  * 		directly with *skb*\ **->hash**.
2328  *
2329  * 		Calling **bpf_set_hash_invalid**\ (), changing a packet
2330  * 		prototype with **bpf_skb_change_proto**\ (), or calling
2331  * 		**bpf_skb_store_bytes**\ () with the
2332  * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2333  * 		the hash and to trigger a new computation for the next call to
2334  * 		**bpf_get_hash_recalc**\ ().
2335  * 	Return
2336  * 		The 32-bit hash.
2337  *
2338  * u64 bpf_get_current_task(void)
2339  * 	Description
2340  * 		Get the current task.
2341  * 	Return
2342  * 		A pointer to the current task struct.
2343  *
2344  * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2345  * 	Description
2346  * 		Attempt in a safe way to write *len* bytes from the buffer
2347  * 		*src* to *dst* in memory. It only works for threads that are in
2348  * 		user context, and *dst* must be a valid user space address.
2349  *
2350  * 		This helper should not be used to implement any kind of
2351  * 		security mechanism because of TOC-TOU attacks, but rather to
2352  * 		debug, divert, and manipulate execution of semi-cooperative
2353  * 		processes.
2354  *
2355  * 		Keep in mind that this feature is meant for experiments, and it
2356  * 		has a risk of crashing the system and running programs.
2357  * 		Therefore, when an eBPF program using this helper is attached,
2358  * 		a warning including PID and process name is printed to kernel
2359  * 		logs.
2360  * 	Return
2361  * 		0 on success, or a negative error in case of failure.
2362  *
2363  * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2364  * 	Description
2365  * 		Check whether the probe is being run is the context of a given
2366  * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2367  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2368  * 	Return
2369  * 		The return value depends on the result of the test, and can be:
2370  *
2371  *		* 1, if current task belongs to the cgroup2.
2372  *		* 0, if current task does not belong to the cgroup2.
2373  * 		* A negative error code, if an error occurred.
2374  *
2375  * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2376  * 	Description
2377  * 		Resize (trim or grow) the packet associated to *skb* to the
2378  * 		new *len*. The *flags* are reserved for future usage, and must
2379  * 		be left at zero.
2380  *
2381  * 		The basic idea is that the helper performs the needed work to
2382  * 		change the size of the packet, then the eBPF program rewrites
2383  * 		the rest via helpers like **bpf_skb_store_bytes**\ (),
2384  * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2385  * 		and others. This helper is a slow path utility intended for
2386  * 		replies with control messages. And because it is targeted for
2387  * 		slow path, the helper itself can afford to be slow: it
2388  * 		implicitly linearizes, unclones and drops offloads from the
2389  * 		*skb*.
2390  *
2391  * 		A call to this helper is susceptible to change the underlying
2392  * 		packet buffer. Therefore, at load time, all checks on pointers
2393  * 		previously done by the verifier are invalidated and must be
2394  * 		performed again, if the helper is used in combination with
2395  * 		direct packet access.
2396  * 	Return
2397  * 		0 on success, or a negative error in case of failure.
2398  *
2399  * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2400  * 	Description
2401  * 		Pull in non-linear data in case the *skb* is non-linear and not
2402  * 		all of *len* are part of the linear section. Make *len* bytes
2403  * 		from *skb* readable and writable. If a zero value is passed for
2404  *		*len*, then all bytes in the linear part of *skb* will be made
2405  *		readable and writable.
2406  *
2407  * 		This helper is only needed for reading and writing with direct
2408  * 		packet access.
2409  *
2410  * 		For direct packet access, testing that offsets to access
2411  * 		are within packet boundaries (test on *skb*\ **->data_end**) is
2412  * 		susceptible to fail if offsets are invalid, or if the requested
2413  * 		data is in non-linear parts of the *skb*. On failure the
2414  * 		program can just bail out, or in the case of a non-linear
2415  * 		buffer, use a helper to make the data available. The
2416  * 		**bpf_skb_load_bytes**\ () helper is a first solution to access
2417  * 		the data. Another one consists in using **bpf_skb_pull_data**
2418  * 		to pull in once the non-linear parts, then retesting and
2419  * 		eventually access the data.
2420  *
2421  * 		At the same time, this also makes sure the *skb* is uncloned,
2422  * 		which is a necessary condition for direct write. As this needs
2423  * 		to be an invariant for the write part only, the verifier
2424  * 		detects writes and adds a prologue that is calling
2425  * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from
2426  * 		the very beginning in case it is indeed cloned.
2427  *
2428  * 		A call to this helper is susceptible to change the underlying
2429  * 		packet buffer. Therefore, at load time, all checks on pointers
2430  * 		previously done by the verifier are invalidated and must be
2431  * 		performed again, if the helper is used in combination with
2432  * 		direct packet access.
2433  * 	Return
2434  * 		0 on success, or a negative error in case of failure.
2435  *
2436  * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2437  * 	Description
2438  * 		Add the checksum *csum* into *skb*\ **->csum** in case the
2439  * 		driver has supplied a checksum for the entire packet into that
2440  * 		field. Return an error otherwise. This helper is intended to be
2441  * 		used in combination with **bpf_csum_diff**\ (), in particular
2442  * 		when the checksum needs to be updated after data has been
2443  * 		written into the packet through direct packet access.
2444  * 	Return
2445  * 		The checksum on success, or a negative error code in case of
2446  * 		failure.
2447  *
2448  * void bpf_set_hash_invalid(struct sk_buff *skb)
2449  * 	Description
2450  * 		Invalidate the current *skb*\ **->hash**. It can be used after
2451  * 		mangling on headers through direct packet access, in order to
2452  * 		indicate that the hash is outdated and to trigger a
2453  * 		recalculation the next time the kernel tries to access this
2454  * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.
2455  * 	Return
2456  * 		void.
2457  *
2458  * long bpf_get_numa_node_id(void)
2459  * 	Description
2460  * 		Return the id of the current NUMA node. The primary use case
2461  * 		for this helper is the selection of sockets for the local NUMA
2462  * 		node, when the program is attached to sockets using the
2463  * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2464  * 		but the helper is also available to other eBPF program types,
2465  * 		similarly to **bpf_get_smp_processor_id**\ ().
2466  * 	Return
2467  * 		The id of current NUMA node.
2468  *
2469  * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2470  * 	Description
2471  * 		Grows headroom of packet associated to *skb* and adjusts the
2472  * 		offset of the MAC header accordingly, adding *len* bytes of
2473  * 		space. It automatically extends and reallocates memory as
2474  * 		required.
2475  *
2476  * 		This helper can be used on a layer 3 *skb* to push a MAC header
2477  * 		for redirection into a layer 2 device.
2478  *
2479  * 		All values for *flags* are reserved for future usage, and must
2480  * 		be left at zero.
2481  *
2482  * 		A call to this helper is susceptible to change the underlying
2483  * 		packet buffer. Therefore, at load time, all checks on pointers
2484  * 		previously done by the verifier are invalidated and must be
2485  * 		performed again, if the helper is used in combination with
2486  * 		direct packet access.
2487  * 	Return
2488  * 		0 on success, or a negative error in case of failure.
2489  *
2490  * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2491  * 	Description
2492  * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2493  * 		it is possible to use a negative value for *delta*. This helper
2494  * 		can be used to prepare the packet for pushing or popping
2495  * 		headers.
2496  *
2497  * 		A call to this helper is susceptible to change the underlying
2498  * 		packet buffer. Therefore, at load time, all checks on pointers
2499  * 		previously done by the verifier are invalidated and must be
2500  * 		performed again, if the helper is used in combination with
2501  * 		direct packet access.
2502  * 	Return
2503  * 		0 on success, or a negative error in case of failure.
2504  *
2505  * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2506  * 	Description
2507  * 		Copy a NUL terminated string from an unsafe kernel address
2508  * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2509  * 		more details.
2510  *
2511  * 		Generally, use **bpf_probe_read_user_str**\ () or
2512  * 		**bpf_probe_read_kernel_str**\ () instead.
2513  * 	Return
2514  * 		On success, the strictly positive length of the string,
2515  * 		including the trailing NUL character. On error, a negative
2516  * 		value.
2517  *
2518  * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2519  * 	Description
2520  * 		If the **struct sk_buff** pointed by *skb* has a known socket,
2521  * 		retrieve the cookie (generated by the kernel) of this socket.
2522  * 		If no cookie has been set yet, generate a new cookie. Once
2523  * 		generated, the socket cookie remains stable for the life of the
2524  * 		socket. This helper can be useful for monitoring per socket
2525  * 		networking traffic statistics as it provides a global socket
2526  * 		identifier that can be assumed unique.
2527  * 	Return
2528  * 		A 8-byte long unique number on success, or 0 if the socket
2529  * 		field is missing inside *skb*.
2530  *
2531  * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2532  * 	Description
2533  * 		Equivalent to bpf_get_socket_cookie() helper that accepts
2534  * 		*skb*, but gets socket from **struct bpf_sock_addr** context.
2535  * 	Return
2536  * 		A 8-byte long unique number.
2537  *
2538  * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2539  * 	Description
2540  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2541  * 		*skb*, but gets socket from **struct bpf_sock_ops** context.
2542  * 	Return
2543  * 		A 8-byte long unique number.
2544  *
2545  * u64 bpf_get_socket_cookie(struct sock *sk)
2546  * 	Description
2547  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2548  * 		*sk*, but gets socket from a BTF **struct sock**. This helper
2549  * 		also works for sleepable programs.
2550  * 	Return
2551  * 		A 8-byte long unique number or 0 if *sk* is NULL.
2552  *
2553  * u32 bpf_get_socket_uid(struct sk_buff *skb)
2554  * 	Description
2555  * 		Get the owner UID of the socked associated to *skb*.
2556  * 	Return
2557  * 		The owner UID of the socket associated to *skb*. If the socket
2558  * 		is **NULL**, or if it is not a full socket (i.e. if it is a
2559  * 		time-wait or a request socket instead), **overflowuid** value
2560  * 		is returned (note that **overflowuid** might also be the actual
2561  * 		UID value for the socket).
2562  *
2563  * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2564  * 	Description
2565  * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)
2566  * 		to value *hash*.
2567  * 	Return
2568  * 		0
2569  *
2570  * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2571  * 	Description
2572  * 		Emulate a call to **setsockopt()** on the socket associated to
2573  * 		*bpf_socket*, which must be a full socket. The *level* at
2574  * 		which the option resides and the name *optname* of the option
2575  * 		must be specified, see **setsockopt(2)** for more information.
2576  * 		The option value of length *optlen* is pointed by *optval*.
2577  *
2578  * 		*bpf_socket* should be one of the following:
2579  *
2580  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2581  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2582  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2583  *
2584  * 		This helper actually implements a subset of **setsockopt()**.
2585  * 		It supports the following *level*\ s:
2586  *
2587  * 		* **SOL_SOCKET**, which supports the following *optname*\ s:
2588  * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2589  * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2590  * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**, **SO_REUSEADDR**,
2591  * 		  **SO_REUSEPORT**, **SO_BINDTOIFINDEX**, **SO_TXREHASH**.
2592  * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:
2593  * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,
2594  * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2595  * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2596  * 		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**,
2597  * 		  **TCP_NODELAY**, **TCP_MAXSEG**, **TCP_WINDOW_CLAMP**,
2598  * 		  **TCP_THIN_LINEAR_TIMEOUTS**, **TCP_BPF_DELACK_MAX**,
2599  * 		  **TCP_BPF_RTO_MIN**.
2600  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2601  * 		* **IPPROTO_IPV6**, which supports the following *optname*\ s:
2602  * 		  **IPV6_TCLASS**, **IPV6_AUTOFLOWLABEL**.
2603  * 	Return
2604  * 		0 on success, or a negative error in case of failure.
2605  *
2606  * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2607  * 	Description
2608  * 		Grow or shrink the room for data in the packet associated to
2609  * 		*skb* by *len_diff*, and according to the selected *mode*.
2610  *
2611  * 		By default, the helper will reset any offloaded checksum
2612  * 		indicator of the skb to CHECKSUM_NONE. This can be avoided
2613  * 		by the following flag:
2614  *
2615  * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2616  * 		  checksum data of the skb to CHECKSUM_NONE.
2617  *
2618  *		There are two supported modes at this time:
2619  *
2620  *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2621  * 		  (room space is added or removed between the layer 2 and
2622  * 		  layer 3 headers).
2623  *
2624  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2625  * 		  (room space is added or removed between the layer 3 and
2626  * 		  layer 4 headers).
2627  *
2628  *		The following flags are supported at this time:
2629  *
2630  *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2631  *		  Adjusting mss in this way is not allowed for datagrams.
2632  *
2633  *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2634  *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2635  *		  Any new space is reserved to hold a tunnel header.
2636  *		  Configure skb offsets and other fields accordingly.
2637  *
2638  *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2639  *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2640  *		  Use with ENCAP_L3 flags to further specify the tunnel type.
2641  *
2642  *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2643  *		  Use with ENCAP_L3/L4 flags to further specify the tunnel
2644  *		  type; *len* is the length of the inner MAC header.
2645  *
2646  *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2647  *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2648  *		  L2 type as Ethernet.
2649  *
2650  * 		A call to this helper is susceptible to change the underlying
2651  * 		packet buffer. Therefore, at load time, all checks on pointers
2652  * 		previously done by the verifier are invalidated and must be
2653  * 		performed again, if the helper is used in combination with
2654  * 		direct packet access.
2655  * 	Return
2656  * 		0 on success, or a negative error in case of failure.
2657  *
2658  * long bpf_redirect_map(struct bpf_map *map, u64 key, u64 flags)
2659  * 	Description
2660  * 		Redirect the packet to the endpoint referenced by *map* at
2661  * 		index *key*. Depending on its type, this *map* can contain
2662  * 		references to net devices (for forwarding packets through other
2663  * 		ports), or to CPUs (for redirecting XDP frames to another CPU;
2664  * 		but this is only implemented for native XDP (with driver
2665  * 		support) as of this writing).
2666  *
2667  * 		The lower two bits of *flags* are used as the return code if
2668  * 		the map lookup fails. This is so that the return value can be
2669  * 		one of the XDP program return codes up to **XDP_TX**, as chosen
2670  * 		by the caller. The higher bits of *flags* can be set to
2671  * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2672  *
2673  * 		With BPF_F_BROADCAST the packet will be broadcasted to all the
2674  * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2675  * 		interface will be excluded when do broadcasting.
2676  *
2677  * 		See also **bpf_redirect**\ (), which only supports redirecting
2678  * 		to an ifindex, but doesn't require a map to do so.
2679  * 	Return
2680  * 		**XDP_REDIRECT** on success, or the value of the two lower bits
2681  * 		of the *flags* argument on error.
2682  *
2683  * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2684  * 	Description
2685  * 		Redirect the packet to the socket referenced by *map* (of type
2686  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2687  * 		egress interfaces can be used for redirection. The
2688  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2689  * 		distinction (ingress path is selected if the flag is present,
2690  * 		egress path otherwise). This is the only flag supported for now.
2691  * 	Return
2692  * 		**SK_PASS** on success, or **SK_DROP** on error.
2693  *
2694  * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2695  * 	Description
2696  * 		Add an entry to, or update a *map* referencing sockets. The
2697  * 		*skops* is used as a new value for the entry associated to
2698  * 		*key*. *flags* is one of:
2699  *
2700  * 		**BPF_NOEXIST**
2701  * 			The entry for *key* must not exist in the map.
2702  * 		**BPF_EXIST**
2703  * 			The entry for *key* must already exist in the map.
2704  * 		**BPF_ANY**
2705  * 			No condition on the existence of the entry for *key*.
2706  *
2707  * 		If the *map* has eBPF programs (parser and verdict), those will
2708  * 		be inherited by the socket being added. If the socket is
2709  * 		already attached to eBPF programs, this results in an error.
2710  * 	Return
2711  * 		0 on success, or a negative error in case of failure.
2712  *
2713  * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2714  * 	Description
2715  * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by
2716  * 		*delta* (which can be positive or negative). Note that this
2717  * 		operation modifies the address stored in *xdp_md*\ **->data**,
2718  * 		so the latter must be loaded only after the helper has been
2719  * 		called.
2720  *
2721  * 		The use of *xdp_md*\ **->data_meta** is optional and programs
2722  * 		are not required to use it. The rationale is that when the
2723  * 		packet is processed with XDP (e.g. as DoS filter), it is
2724  * 		possible to push further meta data along with it before passing
2725  * 		to the stack, and to give the guarantee that an ingress eBPF
2726  * 		program attached as a TC classifier on the same device can pick
2727  * 		this up for further post-processing. Since TC works with socket
2728  * 		buffers, it remains possible to set from XDP the **mark** or
2729  * 		**priority** pointers, or other pointers for the socket buffer.
2730  * 		Having this scratch space generic and programmable allows for
2731  * 		more flexibility as the user is free to store whatever meta
2732  * 		data they need.
2733  *
2734  * 		A call to this helper is susceptible to change the underlying
2735  * 		packet buffer. Therefore, at load time, all checks on pointers
2736  * 		previously done by the verifier are invalidated and must be
2737  * 		performed again, if the helper is used in combination with
2738  * 		direct packet access.
2739  * 	Return
2740  * 		0 on success, or a negative error in case of failure.
2741  *
2742  * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2743  * 	Description
2744  * 		Read the value of a perf event counter, and store it into *buf*
2745  * 		of size *buf_size*. This helper relies on a *map* of type
2746  * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2747  * 		counter is selected when *map* is updated with perf event file
2748  * 		descriptors. The *map* is an array whose size is the number of
2749  * 		available CPUs, and each cell contains a value relative to one
2750  * 		CPU. The value to retrieve is indicated by *flags*, that
2751  * 		contains the index of the CPU to look up, masked with
2752  * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2753  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2754  * 		current CPU should be retrieved.
2755  *
2756  * 		This helper behaves in a way close to
2757  * 		**bpf_perf_event_read**\ () helper, save that instead of
2758  * 		just returning the value observed, it fills the *buf*
2759  * 		structure. This allows for additional data to be retrieved: in
2760  * 		particular, the enabled and running times (in *buf*\
2761  * 		**->enabled** and *buf*\ **->running**, respectively) are
2762  * 		copied. In general, **bpf_perf_event_read_value**\ () is
2763  * 		recommended over **bpf_perf_event_read**\ (), which has some
2764  * 		ABI issues and provides fewer functionalities.
2765  *
2766  * 		These values are interesting, because hardware PMU (Performance
2767  * 		Monitoring Unit) counters are limited resources. When there are
2768  * 		more PMU based perf events opened than available counters,
2769  * 		kernel will multiplex these events so each event gets certain
2770  * 		percentage (but not all) of the PMU time. In case that
2771  * 		multiplexing happens, the number of samples or counter value
2772  * 		will not reflect the case compared to when no multiplexing
2773  * 		occurs. This makes comparison between different runs difficult.
2774  * 		Typically, the counter value should be normalized before
2775  * 		comparing to other experiments. The usual normalization is done
2776  * 		as follows.
2777  *
2778  * 		::
2779  *
2780  * 			normalized_counter = counter * t_enabled / t_running
2781  *
2782  * 		Where t_enabled is the time enabled for event and t_running is
2783  * 		the time running for event since last normalization. The
2784  * 		enabled and running times are accumulated since the perf event
2785  * 		open. To achieve scaling factor between two invocations of an
2786  * 		eBPF program, users can use CPU id as the key (which is
2787  * 		typical for perf array usage model) to remember the previous
2788  * 		value and do the calculation inside the eBPF program.
2789  * 	Return
2790  * 		0 on success, or a negative error in case of failure.
2791  *
2792  * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2793  * 	Description
2794  * 		For en eBPF program attached to a perf event, retrieve the
2795  * 		value of the event counter associated to *ctx* and store it in
2796  * 		the structure pointed by *buf* and of size *buf_size*. Enabled
2797  * 		and running times are also stored in the structure (see
2798  * 		description of helper **bpf_perf_event_read_value**\ () for
2799  * 		more details).
2800  * 	Return
2801  * 		0 on success, or a negative error in case of failure.
2802  *
2803  * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2804  * 	Description
2805  * 		Emulate a call to **getsockopt()** on the socket associated to
2806  * 		*bpf_socket*, which must be a full socket. The *level* at
2807  * 		which the option resides and the name *optname* of the option
2808  * 		must be specified, see **getsockopt(2)** for more information.
2809  * 		The retrieved value is stored in the structure pointed by
2810  * 		*opval* and of length *optlen*.
2811  *
2812  * 		*bpf_socket* should be one of the following:
2813  *
2814  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2815  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2816  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2817  *
2818  * 		This helper actually implements a subset of **getsockopt()**.
2819  * 		It supports the same set of *optname*\ s that is supported by
2820  * 		the **bpf_setsockopt**\ () helper.  The exceptions are
2821  * 		**TCP_BPF_*** is **bpf_setsockopt**\ () only and
2822  * 		**TCP_SAVED_SYN** is **bpf_getsockopt**\ () only.
2823  * 	Return
2824  * 		0 on success, or a negative error in case of failure.
2825  *
2826  * long bpf_override_return(struct pt_regs *regs, u64 rc)
2827  * 	Description
2828  * 		Used for error injection, this helper uses kprobes to override
2829  * 		the return value of the probed function, and to set it to *rc*.
2830  * 		The first argument is the context *regs* on which the kprobe
2831  * 		works.
2832  *
2833  * 		This helper works by setting the PC (program counter)
2834  * 		to an override function which is run in place of the original
2835  * 		probed function. This means the probed function is not run at
2836  * 		all. The replacement function just returns with the required
2837  * 		value.
2838  *
2839  * 		This helper has security implications, and thus is subject to
2840  * 		restrictions. It is only available if the kernel was compiled
2841  * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2842  * 		option, and in this case it only works on functions tagged with
2843  * 		**ALLOW_ERROR_INJECTION** in the kernel code.
2844  *
2845  * 		Also, the helper is only available for the architectures having
2846  * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2847  * 		x86 architecture is the only one to support this feature.
2848  * 	Return
2849  * 		0
2850  *
2851  * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2852  * 	Description
2853  * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2854  * 		for the full TCP socket associated to *bpf_sock_ops* to
2855  * 		*argval*.
2856  *
2857  * 		The primary use of this field is to determine if there should
2858  * 		be calls to eBPF programs of type
2859  * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2860  * 		code. A program of the same type can change its value, per
2861  * 		connection and as necessary, when the connection is
2862  * 		established. This field is directly accessible for reading, but
2863  * 		this helper must be used for updates in order to return an
2864  * 		error if an eBPF program tries to set a callback that is not
2865  * 		supported in the current kernel.
2866  *
2867  * 		*argval* is a flag array which can combine these flags:
2868  *
2869  * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2870  * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2871  * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2872  * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2873  *
2874  * 		Therefore, this function can be used to clear a callback flag by
2875  * 		setting the appropriate bit to zero. e.g. to disable the RTO
2876  * 		callback:
2877  *
2878  * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**
2879  * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2880  *
2881  * 		Here are some examples of where one could call such eBPF
2882  * 		program:
2883  *
2884  * 		* When RTO fires.
2885  * 		* When a packet is retransmitted.
2886  * 		* When the connection terminates.
2887  * 		* When a packet is sent.
2888  * 		* When a packet is received.
2889  * 	Return
2890  * 		Code **-EINVAL** if the socket is not a full TCP socket;
2891  * 		otherwise, a positive number containing the bits that could not
2892  * 		be set is returned (which comes down to 0 if all bits were set
2893  * 		as required).
2894  *
2895  * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2896  * 	Description
2897  * 		This helper is used in programs implementing policies at the
2898  * 		socket level. If the message *msg* is allowed to pass (i.e. if
2899  * 		the verdict eBPF program returns **SK_PASS**), redirect it to
2900  * 		the socket referenced by *map* (of type
2901  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2902  * 		egress interfaces can be used for redirection. The
2903  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2904  * 		distinction (ingress path is selected if the flag is present,
2905  * 		egress path otherwise). This is the only flag supported for now.
2906  * 	Return
2907  * 		**SK_PASS** on success, or **SK_DROP** on error.
2908  *
2909  * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2910  * 	Description
2911  * 		For socket policies, apply the verdict of the eBPF program to
2912  * 		the next *bytes* (number of bytes) of message *msg*.
2913  *
2914  * 		For example, this helper can be used in the following cases:
2915  *
2916  * 		* A single **sendmsg**\ () or **sendfile**\ () system call
2917  * 		  contains multiple logical messages that the eBPF program is
2918  * 		  supposed to read and for which it should apply a verdict.
2919  * 		* An eBPF program only cares to read the first *bytes* of a
2920  * 		  *msg*. If the message has a large payload, then setting up
2921  * 		  and calling the eBPF program repeatedly for all bytes, even
2922  * 		  though the verdict is already known, would create unnecessary
2923  * 		  overhead.
2924  *
2925  * 		When called from within an eBPF program, the helper sets a
2926  * 		counter internal to the BPF infrastructure, that is used to
2927  * 		apply the last verdict to the next *bytes*. If *bytes* is
2928  * 		smaller than the current data being processed from a
2929  * 		**sendmsg**\ () or **sendfile**\ () system call, the first
2930  * 		*bytes* will be sent and the eBPF program will be re-run with
2931  * 		the pointer for start of data pointing to byte number *bytes*
2932  * 		**+ 1**. If *bytes* is larger than the current data being
2933  * 		processed, then the eBPF verdict will be applied to multiple
2934  * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2935  * 		consumed.
2936  *
2937  * 		Note that if a socket closes with the internal counter holding
2938  * 		a non-zero value, this is not a problem because data is not
2939  * 		being buffered for *bytes* and is sent as it is received.
2940  * 	Return
2941  * 		0
2942  *
2943  * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2944  * 	Description
2945  * 		For socket policies, prevent the execution of the verdict eBPF
2946  * 		program for message *msg* until *bytes* (byte number) have been
2947  * 		accumulated.
2948  *
2949  * 		This can be used when one needs a specific number of bytes
2950  * 		before a verdict can be assigned, even if the data spans
2951  * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2952  * 		case would be a user calling **sendmsg**\ () repeatedly with
2953  * 		1-byte long message segments. Obviously, this is bad for
2954  * 		performance, but it is still valid. If the eBPF program needs
2955  * 		*bytes* bytes to validate a header, this helper can be used to
2956  * 		prevent the eBPF program to be called again until *bytes* have
2957  * 		been accumulated.
2958  * 	Return
2959  * 		0
2960  *
2961  * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2962  * 	Description
2963  * 		For socket policies, pull in non-linear data from user space
2964  * 		for *msg* and set pointers *msg*\ **->data** and *msg*\
2965  * 		**->data_end** to *start* and *end* bytes offsets into *msg*,
2966  * 		respectively.
2967  *
2968  * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2969  * 		*msg* it can only parse data that the (**data**, **data_end**)
2970  * 		pointers have already consumed. For **sendmsg**\ () hooks this
2971  * 		is likely the first scatterlist element. But for calls relying
2972  * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2973  * 		be the range (**0**, **0**) because the data is shared with
2974  * 		user space and by default the objective is to avoid allowing
2975  * 		user space to modify data while (or after) eBPF verdict is
2976  * 		being decided. This helper can be used to pull in data and to
2977  * 		set the start and end pointer to given values. Data will be
2978  * 		copied if necessary (i.e. if data was not linear and if start
2979  * 		and end pointers do not point to the same chunk).
2980  *
2981  * 		A call to this helper is susceptible to change the underlying
2982  * 		packet buffer. Therefore, at load time, all checks on pointers
2983  * 		previously done by the verifier are invalidated and must be
2984  * 		performed again, if the helper is used in combination with
2985  * 		direct packet access.
2986  *
2987  * 		All values for *flags* are reserved for future usage, and must
2988  * 		be left at zero.
2989  * 	Return
2990  * 		0 on success, or a negative error in case of failure.
2991  *
2992  * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2993  * 	Description
2994  * 		Bind the socket associated to *ctx* to the address pointed by
2995  * 		*addr*, of length *addr_len*. This allows for making outgoing
2996  * 		connection from the desired IP address, which can be useful for
2997  * 		example when all processes inside a cgroup should use one
2998  * 		single IP address on a host that has multiple IP configured.
2999  *
3000  * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The
3001  * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or
3002  * 		**AF_INET6**). It's advised to pass zero port (**sin_port**
3003  * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
3004  * 		behavior and lets the kernel efficiently pick up an unused
3005  * 		port as long as 4-tuple is unique. Passing non-zero port might
3006  * 		lead to degraded performance.
3007  * 	Return
3008  * 		0 on success, or a negative error in case of failure.
3009  *
3010  * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
3011  * 	Description
3012  * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
3013  * 		possible to both shrink and grow the packet tail.
3014  * 		Shrink done via *delta* being a negative integer.
3015  *
3016  * 		A call to this helper is susceptible to change the underlying
3017  * 		packet buffer. Therefore, at load time, all checks on pointers
3018  * 		previously done by the verifier are invalidated and must be
3019  * 		performed again, if the helper is used in combination with
3020  * 		direct packet access.
3021  * 	Return
3022  * 		0 on success, or a negative error in case of failure.
3023  *
3024  * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
3025  * 	Description
3026  * 		Retrieve the XFRM state (IP transform framework, see also
3027  * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
3028  *
3029  * 		The retrieved value is stored in the **struct bpf_xfrm_state**
3030  * 		pointed by *xfrm_state* and of length *size*.
3031  *
3032  * 		All values for *flags* are reserved for future usage, and must
3033  * 		be left at zero.
3034  *
3035  * 		This helper is available only if the kernel was compiled with
3036  * 		**CONFIG_XFRM** configuration option.
3037  * 	Return
3038  * 		0 on success, or a negative error in case of failure.
3039  *
3040  * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
3041  * 	Description
3042  * 		Return a user or a kernel stack in bpf program provided buffer.
3043  * 		To achieve this, the helper needs *ctx*, which is a pointer
3044  * 		to the context on which the tracing program is executed.
3045  * 		To store the stacktrace, the bpf program provides *buf* with
3046  * 		a nonnegative *size*.
3047  *
3048  * 		The last argument, *flags*, holds the number of stack frames to
3049  * 		skip (from 0 to 255), masked with
3050  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3051  * 		the following flags:
3052  *
3053  * 		**BPF_F_USER_STACK**
3054  * 			Collect a user space stack instead of a kernel stack.
3055  * 		**BPF_F_USER_BUILD_ID**
3056  * 			Collect (build_id, file_offset) instead of ips for user
3057  * 			stack, only valid if **BPF_F_USER_STACK** is also
3058  * 			specified.
3059  *
3060  * 			*file_offset* is an offset relative to the beginning
3061  * 			of the executable or shared object file backing the vma
3062  * 			which the *ip* falls in. It is *not* an offset relative
3063  * 			to that object's base address. Accordingly, it must be
3064  * 			adjusted by adding (sh_addr - sh_offset), where
3065  * 			sh_{addr,offset} correspond to the executable section
3066  * 			containing *file_offset* in the object, for comparisons
3067  * 			to symbols' st_value to be valid.
3068  *
3069  * 		**bpf_get_stack**\ () can collect up to
3070  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3071  * 		to sufficient large buffer size. Note that
3072  * 		this limit can be controlled with the **sysctl** program, and
3073  * 		that it should be manually increased in order to profile long
3074  * 		user stacks (such as stacks for Java programs). To do so, use:
3075  *
3076  * 		::
3077  *
3078  * 			# sysctl kernel.perf_event_max_stack=<new value>
3079  * 	Return
3080  * 		The non-negative copied *buf* length equal to or less than
3081  * 		*size* on success, or a negative error in case of failure.
3082  *
3083  * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3084  * 	Description
3085  * 		This helper is similar to **bpf_skb_load_bytes**\ () in that
3086  * 		it provides an easy way to load *len* bytes from *offset*
3087  * 		from the packet associated to *skb*, into the buffer pointed
3088  * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3089  * 		a fifth argument *start_header* exists in order to select a
3090  * 		base offset to start from. *start_header* can be one of:
3091  *
3092  * 		**BPF_HDR_START_MAC**
3093  * 			Base offset to load data from is *skb*'s mac header.
3094  * 		**BPF_HDR_START_NET**
3095  * 			Base offset to load data from is *skb*'s network header.
3096  *
3097  * 		In general, "direct packet access" is the preferred method to
3098  * 		access packet data, however, this helper is in particular useful
3099  * 		in socket filters where *skb*\ **->data** does not always point
3100  * 		to the start of the mac header and where "direct packet access"
3101  * 		is not available.
3102  * 	Return
3103  * 		0 on success, or a negative error in case of failure.
3104  *
3105  * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3106  *	Description
3107  *		Do FIB lookup in kernel tables using parameters in *params*.
3108  *		If lookup is successful and result shows packet is to be
3109  *		forwarded, the neighbor tables are searched for the nexthop.
3110  *		If successful (ie., FIB lookup shows forwarding and nexthop
3111  *		is resolved), the nexthop address is returned in ipv4_dst
3112  *		or ipv6_dst based on family, smac is set to mac address of
3113  *		egress device, dmac is set to nexthop mac address, rt_metric
3114  *		is set to metric from route (IPv4/IPv6 only), and ifindex
3115  *		is set to the device index of the nexthop from the FIB lookup.
3116  *
3117  *		*plen* argument is the size of the passed in struct.
3118  *		*flags* argument can be a combination of one or more of the
3119  *		following values:
3120  *
3121  *		**BPF_FIB_LOOKUP_DIRECT**
3122  *			Do a direct table lookup vs full lookup using FIB
3123  *			rules.
3124  *		**BPF_FIB_LOOKUP_OUTPUT**
3125  *			Perform lookup from an egress perspective (default is
3126  *			ingress).
3127  *
3128  *		*ctx* is either **struct xdp_md** for XDP programs or
3129  *		**struct sk_buff** tc cls_act programs.
3130  *	Return
3131  *		* < 0 if any input argument is invalid
3132  *		*   0 on success (packet is forwarded, nexthop neighbor exists)
3133  *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3134  *		  packet is not forwarded or needs assist from full stack
3135  *
3136  *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3137  *		was exceeded and output params->mtu_result contains the MTU.
3138  *
3139  * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3140  *	Description
3141  *		Add an entry to, or update a sockhash *map* referencing sockets.
3142  *		The *skops* is used as a new value for the entry associated to
3143  *		*key*. *flags* is one of:
3144  *
3145  *		**BPF_NOEXIST**
3146  *			The entry for *key* must not exist in the map.
3147  *		**BPF_EXIST**
3148  *			The entry for *key* must already exist in the map.
3149  *		**BPF_ANY**
3150  *			No condition on the existence of the entry for *key*.
3151  *
3152  *		If the *map* has eBPF programs (parser and verdict), those will
3153  *		be inherited by the socket being added. If the socket is
3154  *		already attached to eBPF programs, this results in an error.
3155  *	Return
3156  *		0 on success, or a negative error in case of failure.
3157  *
3158  * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3159  *	Description
3160  *		This helper is used in programs implementing policies at the
3161  *		socket level. If the message *msg* is allowed to pass (i.e. if
3162  *		the verdict eBPF program returns **SK_PASS**), redirect it to
3163  *		the socket referenced by *map* (of type
3164  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3165  *		egress interfaces can be used for redirection. The
3166  *		**BPF_F_INGRESS** value in *flags* is used to make the
3167  *		distinction (ingress path is selected if the flag is present,
3168  *		egress path otherwise). This is the only flag supported for now.
3169  *	Return
3170  *		**SK_PASS** on success, or **SK_DROP** on error.
3171  *
3172  * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3173  *	Description
3174  *		This helper is used in programs implementing policies at the
3175  *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3176  *		if the verdict eBPF program returns **SK_PASS**), redirect it
3177  *		to the socket referenced by *map* (of type
3178  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3179  *		egress interfaces can be used for redirection. The
3180  *		**BPF_F_INGRESS** value in *flags* is used to make the
3181  *		distinction (ingress path is selected if the flag is present,
3182  *		egress otherwise). This is the only flag supported for now.
3183  *	Return
3184  *		**SK_PASS** on success, or **SK_DROP** on error.
3185  *
3186  * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3187  *	Description
3188  *		Encapsulate the packet associated to *skb* within a Layer 3
3189  *		protocol header. This header is provided in the buffer at
3190  *		address *hdr*, with *len* its size in bytes. *type* indicates
3191  *		the protocol of the header and can be one of:
3192  *
3193  *		**BPF_LWT_ENCAP_SEG6**
3194  *			IPv6 encapsulation with Segment Routing Header
3195  *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3196  *			the IPv6 header is computed by the kernel.
3197  *		**BPF_LWT_ENCAP_SEG6_INLINE**
3198  *			Only works if *skb* contains an IPv6 packet. Insert a
3199  *			Segment Routing Header (**struct ipv6_sr_hdr**) inside
3200  *			the IPv6 header.
3201  *		**BPF_LWT_ENCAP_IP**
3202  *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3203  *			must be IPv4 or IPv6, followed by zero or more
3204  *			additional headers, up to **LWT_BPF_MAX_HEADROOM**
3205  *			total bytes in all prepended headers. Please note that
3206  *			if **skb_is_gso**\ (*skb*) is true, no more than two
3207  *			headers can be prepended, and the inner header, if
3208  *			present, should be either GRE or UDP/GUE.
3209  *
3210  *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3211  *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3212  *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3213  *		**BPF_PROG_TYPE_LWT_XMIT**.
3214  *
3215  * 		A call to this helper is susceptible to change the underlying
3216  * 		packet buffer. Therefore, at load time, all checks on pointers
3217  * 		previously done by the verifier are invalidated and must be
3218  * 		performed again, if the helper is used in combination with
3219  * 		direct packet access.
3220  *	Return
3221  * 		0 on success, or a negative error in case of failure.
3222  *
3223  * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3224  *	Description
3225  *		Store *len* bytes from address *from* into the packet
3226  *		associated to *skb*, at *offset*. Only the flags, tag and TLVs
3227  *		inside the outermost IPv6 Segment Routing Header can be
3228  *		modified through this helper.
3229  *
3230  * 		A call to this helper is susceptible to change the underlying
3231  * 		packet buffer. Therefore, at load time, all checks on pointers
3232  * 		previously done by the verifier are invalidated and must be
3233  * 		performed again, if the helper is used in combination with
3234  * 		direct packet access.
3235  *	Return
3236  * 		0 on success, or a negative error in case of failure.
3237  *
3238  * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3239  *	Description
3240  *		Adjust the size allocated to TLVs in the outermost IPv6
3241  *		Segment Routing Header contained in the packet associated to
3242  *		*skb*, at position *offset* by *delta* bytes. Only offsets
3243  *		after the segments are accepted. *delta* can be as well
3244  *		positive (growing) as negative (shrinking).
3245  *
3246  * 		A call to this helper is susceptible to change the underlying
3247  * 		packet buffer. Therefore, at load time, all checks on pointers
3248  * 		previously done by the verifier are invalidated and must be
3249  * 		performed again, if the helper is used in combination with
3250  * 		direct packet access.
3251  *	Return
3252  * 		0 on success, or a negative error in case of failure.
3253  *
3254  * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3255  *	Description
3256  *		Apply an IPv6 Segment Routing action of type *action* to the
3257  *		packet associated to *skb*. Each action takes a parameter
3258  *		contained at address *param*, and of length *param_len* bytes.
3259  *		*action* can be one of:
3260  *
3261  *		**SEG6_LOCAL_ACTION_END_X**
3262  *			End.X action: Endpoint with Layer-3 cross-connect.
3263  *			Type of *param*: **struct in6_addr**.
3264  *		**SEG6_LOCAL_ACTION_END_T**
3265  *			End.T action: Endpoint with specific IPv6 table lookup.
3266  *			Type of *param*: **int**.
3267  *		**SEG6_LOCAL_ACTION_END_B6**
3268  *			End.B6 action: Endpoint bound to an SRv6 policy.
3269  *			Type of *param*: **struct ipv6_sr_hdr**.
3270  *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**
3271  *			End.B6.Encap action: Endpoint bound to an SRv6
3272  *			encapsulation policy.
3273  *			Type of *param*: **struct ipv6_sr_hdr**.
3274  *
3275  * 		A call to this helper is susceptible to change the underlying
3276  * 		packet buffer. Therefore, at load time, all checks on pointers
3277  * 		previously done by the verifier are invalidated and must be
3278  * 		performed again, if the helper is used in combination with
3279  * 		direct packet access.
3280  *	Return
3281  * 		0 on success, or a negative error in case of failure.
3282  *
3283  * long bpf_rc_repeat(void *ctx)
3284  *	Description
3285  *		This helper is used in programs implementing IR decoding, to
3286  *		report a successfully decoded repeat key message. This delays
3287  *		the generation of a key up event for previously generated
3288  *		key down event.
3289  *
3290  *		Some IR protocols like NEC have a special IR message for
3291  *		repeating last button, for when a button is held down.
3292  *
3293  *		The *ctx* should point to the lirc sample as passed into
3294  *		the program.
3295  *
3296  *		This helper is only available is the kernel was compiled with
3297  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3298  *		"**y**".
3299  *	Return
3300  *		0
3301  *
3302  * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3303  *	Description
3304  *		This helper is used in programs implementing IR decoding, to
3305  *		report a successfully decoded key press with *scancode*,
3306  *		*toggle* value in the given *protocol*. The scancode will be
3307  *		translated to a keycode using the rc keymap, and reported as
3308  *		an input key down event. After a period a key up event is
3309  *		generated. This period can be extended by calling either
3310  *		**bpf_rc_keydown**\ () again with the same values, or calling
3311  *		**bpf_rc_repeat**\ ().
3312  *
3313  *		Some protocols include a toggle bit, in case the button was
3314  *		released and pressed again between consecutive scancodes.
3315  *
3316  *		The *ctx* should point to the lirc sample as passed into
3317  *		the program.
3318  *
3319  *		The *protocol* is the decoded protocol number (see
3320  *		**enum rc_proto** for some predefined values).
3321  *
3322  *		This helper is only available is the kernel was compiled with
3323  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3324  *		"**y**".
3325  *	Return
3326  *		0
3327  *
3328  * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3329  * 	Description
3330  * 		Return the cgroup v2 id of the socket associated with the *skb*.
3331  * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3332  * 		helper for cgroup v1 by providing a tag resp. identifier that
3333  * 		can be matched on or used for map lookups e.g. to implement
3334  * 		policy. The cgroup v2 id of a given path in the hierarchy is
3335  * 		exposed in user space through the f_handle API in order to get
3336  * 		to the same 64-bit id.
3337  *
3338  * 		This helper can be used on TC egress path, but not on ingress,
3339  * 		and is available only if the kernel was compiled with the
3340  * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.
3341  * 	Return
3342  * 		The id is returned or 0 in case the id could not be retrieved.
3343  *
3344  * u64 bpf_get_current_cgroup_id(void)
3345  * 	Description
3346  * 		Get the current cgroup id based on the cgroup within which
3347  * 		the current task is running.
3348  * 	Return
3349  * 		A 64-bit integer containing the current cgroup id based
3350  * 		on the cgroup within which the current task is running.
3351  *
3352  * void *bpf_get_local_storage(void *map, u64 flags)
3353  *	Description
3354  *		Get the pointer to the local storage area.
3355  *		The type and the size of the local storage is defined
3356  *		by the *map* argument.
3357  *		The *flags* meaning is specific for each map type,
3358  *		and has to be 0 for cgroup local storage.
3359  *
3360  *		Depending on the BPF program type, a local storage area
3361  *		can be shared between multiple instances of the BPF program,
3362  *		running simultaneously.
3363  *
3364  *		A user should care about the synchronization by himself.
3365  *		For example, by using the **BPF_ATOMIC** instructions to alter
3366  *		the shared data.
3367  *	Return
3368  *		A pointer to the local storage area.
3369  *
3370  * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3371  *	Description
3372  *		Select a **SO_REUSEPORT** socket from a
3373  *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3374  *		It checks the selected socket is matching the incoming
3375  *		request in the socket buffer.
3376  *	Return
3377  *		0 on success, or a negative error in case of failure.
3378  *
3379  * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3380  *	Description
3381  *		Return id of cgroup v2 that is ancestor of cgroup associated
3382  *		with the *skb* at the *ancestor_level*.  The root cgroup is at
3383  *		*ancestor_level* zero and each step down the hierarchy
3384  *		increments the level. If *ancestor_level* == level of cgroup
3385  *		associated with *skb*, then return value will be same as that
3386  *		of **bpf_skb_cgroup_id**\ ().
3387  *
3388  *		The helper is useful to implement policies based on cgroups
3389  *		that are upper in hierarchy than immediate cgroup associated
3390  *		with *skb*.
3391  *
3392  *		The format of returned id and helper limitations are same as in
3393  *		**bpf_skb_cgroup_id**\ ().
3394  *	Return
3395  *		The id is returned or 0 in case the id could not be retrieved.
3396  *
3397  * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3398  *	Description
3399  *		Look for TCP socket matching *tuple*, optionally in a child
3400  *		network namespace *netns*. The return value must be checked,
3401  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3402  *
3403  *		The *ctx* should point to the context of the program, such as
3404  *		the skb or socket (depending on the hook in use). This is used
3405  *		to determine the base network namespace for the lookup.
3406  *
3407  *		*tuple_size* must be one of:
3408  *
3409  *		**sizeof**\ (*tuple*\ **->ipv4**)
3410  *			Look for an IPv4 socket.
3411  *		**sizeof**\ (*tuple*\ **->ipv6**)
3412  *			Look for an IPv6 socket.
3413  *
3414  *		If the *netns* is a negative signed 32-bit integer, then the
3415  *		socket lookup table in the netns associated with the *ctx*
3416  *		will be used. For the TC hooks, this is the netns of the device
3417  *		in the skb. For socket hooks, this is the netns of the socket.
3418  *		If *netns* is any other signed 32-bit value greater than or
3419  *		equal to zero then it specifies the ID of the netns relative to
3420  *		the netns associated with the *ctx*. *netns* values beyond the
3421  *		range of 32-bit integers are reserved for future use.
3422  *
3423  *		All values for *flags* are reserved for future usage, and must
3424  *		be left at zero.
3425  *
3426  *		This helper is available only if the kernel was compiled with
3427  *		**CONFIG_NET** configuration option.
3428  *	Return
3429  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3430  *		For sockets with reuseport option, the **struct bpf_sock**
3431  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3432  *		tuple.
3433  *
3434  * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3435  *	Description
3436  *		Look for UDP socket matching *tuple*, optionally in a child
3437  *		network namespace *netns*. The return value must be checked,
3438  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3439  *
3440  *		The *ctx* should point to the context of the program, such as
3441  *		the skb or socket (depending on the hook in use). This is used
3442  *		to determine the base network namespace for the lookup.
3443  *
3444  *		*tuple_size* must be one of:
3445  *
3446  *		**sizeof**\ (*tuple*\ **->ipv4**)
3447  *			Look for an IPv4 socket.
3448  *		**sizeof**\ (*tuple*\ **->ipv6**)
3449  *			Look for an IPv6 socket.
3450  *
3451  *		If the *netns* is a negative signed 32-bit integer, then the
3452  *		socket lookup table in the netns associated with the *ctx*
3453  *		will be used. For the TC hooks, this is the netns of the device
3454  *		in the skb. For socket hooks, this is the netns of the socket.
3455  *		If *netns* is any other signed 32-bit value greater than or
3456  *		equal to zero then it specifies the ID of the netns relative to
3457  *		the netns associated with the *ctx*. *netns* values beyond the
3458  *		range of 32-bit integers are reserved for future use.
3459  *
3460  *		All values for *flags* are reserved for future usage, and must
3461  *		be left at zero.
3462  *
3463  *		This helper is available only if the kernel was compiled with
3464  *		**CONFIG_NET** configuration option.
3465  *	Return
3466  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3467  *		For sockets with reuseport option, the **struct bpf_sock**
3468  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3469  *		tuple.
3470  *
3471  * long bpf_sk_release(void *sock)
3472  *	Description
3473  *		Release the reference held by *sock*. *sock* must be a
3474  *		non-**NULL** pointer that was returned from
3475  *		**bpf_sk_lookup_xxx**\ ().
3476  *	Return
3477  *		0 on success, or a negative error in case of failure.
3478  *
3479  * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3480  * 	Description
3481  * 		Push an element *value* in *map*. *flags* is one of:
3482  *
3483  * 		**BPF_EXIST**
3484  * 			If the queue/stack is full, the oldest element is
3485  * 			removed to make room for this.
3486  * 	Return
3487  * 		0 on success, or a negative error in case of failure.
3488  *
3489  * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3490  * 	Description
3491  * 		Pop an element from *map*.
3492  * 	Return
3493  * 		0 on success, or a negative error in case of failure.
3494  *
3495  * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3496  * 	Description
3497  * 		Get an element from *map* without removing it.
3498  * 	Return
3499  * 		0 on success, or a negative error in case of failure.
3500  *
3501  * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3502  *	Description
3503  *		For socket policies, insert *len* bytes into *msg* at offset
3504  *		*start*.
3505  *
3506  *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3507  *		*msg* it may want to insert metadata or options into the *msg*.
3508  *		This can later be read and used by any of the lower layer BPF
3509  *		hooks.
3510  *
3511  *		This helper may fail if under memory pressure (a malloc
3512  *		fails) in these cases BPF programs will get an appropriate
3513  *		error and BPF programs will need to handle them.
3514  *	Return
3515  *		0 on success, or a negative error in case of failure.
3516  *
3517  * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3518  *	Description
3519  *		Will remove *len* bytes from a *msg* starting at byte *start*.
3520  *		This may result in **ENOMEM** errors under certain situations if
3521  *		an allocation and copy are required due to a full ring buffer.
3522  *		However, the helper will try to avoid doing the allocation
3523  *		if possible. Other errors can occur if input parameters are
3524  *		invalid either due to *start* byte not being valid part of *msg*
3525  *		payload and/or *pop* value being to large.
3526  *	Return
3527  *		0 on success, or a negative error in case of failure.
3528  *
3529  * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3530  *	Description
3531  *		This helper is used in programs implementing IR decoding, to
3532  *		report a successfully decoded pointer movement.
3533  *
3534  *		The *ctx* should point to the lirc sample as passed into
3535  *		the program.
3536  *
3537  *		This helper is only available is the kernel was compiled with
3538  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3539  *		"**y**".
3540  *	Return
3541  *		0
3542  *
3543  * long bpf_spin_lock(struct bpf_spin_lock *lock)
3544  *	Description
3545  *		Acquire a spinlock represented by the pointer *lock*, which is
3546  *		stored as part of a value of a map. Taking the lock allows to
3547  *		safely update the rest of the fields in that value. The
3548  *		spinlock can (and must) later be released with a call to
3549  *		**bpf_spin_unlock**\ (\ *lock*\ ).
3550  *
3551  *		Spinlocks in BPF programs come with a number of restrictions
3552  *		and constraints:
3553  *
3554  *		* **bpf_spin_lock** objects are only allowed inside maps of
3555  *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3556  *		  list could be extended in the future).
3557  *		* BTF description of the map is mandatory.
3558  *		* The BPF program can take ONE lock at a time, since taking two
3559  *		  or more could cause dead locks.
3560  *		* Only one **struct bpf_spin_lock** is allowed per map element.
3561  *		* When the lock is taken, calls (either BPF to BPF or helpers)
3562  *		  are not allowed.
3563  *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3564  *		  allowed inside a spinlock-ed region.
3565  *		* The BPF program MUST call **bpf_spin_unlock**\ () to release
3566  *		  the lock, on all execution paths, before it returns.
3567  *		* The BPF program can access **struct bpf_spin_lock** only via
3568  *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3569  *		  helpers. Loading or storing data into the **struct
3570  *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3571  *		* To use the **bpf_spin_lock**\ () helper, the BTF description
3572  *		  of the map value must be a struct and have **struct
3573  *		  bpf_spin_lock** *anyname*\ **;** field at the top level.
3574  *		  Nested lock inside another struct is not allowed.
3575  *		* The **struct bpf_spin_lock** *lock* field in a map value must
3576  *		  be aligned on a multiple of 4 bytes in that value.
3577  *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3578  *		  the **bpf_spin_lock** field to user space.
3579  *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3580  *		  a BPF program, do not update the **bpf_spin_lock** field.
3581  *		* **bpf_spin_lock** cannot be on the stack or inside a
3582  *		  networking packet (it can only be inside of a map values).
3583  *		* **bpf_spin_lock** is available to root only.
3584  *		* Tracing programs and socket filter programs cannot use
3585  *		  **bpf_spin_lock**\ () due to insufficient preemption checks
3586  *		  (but this may change in the future).
3587  *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3588  *	Return
3589  *		0
3590  *
3591  * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3592  *	Description
3593  *		Release the *lock* previously locked by a call to
3594  *		**bpf_spin_lock**\ (\ *lock*\ ).
3595  *	Return
3596  *		0
3597  *
3598  * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3599  *	Description
3600  *		This helper gets a **struct bpf_sock** pointer such
3601  *		that all the fields in this **bpf_sock** can be accessed.
3602  *	Return
3603  *		A **struct bpf_sock** pointer on success, or **NULL** in
3604  *		case of failure.
3605  *
3606  * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3607  *	Description
3608  *		This helper gets a **struct bpf_tcp_sock** pointer from a
3609  *		**struct bpf_sock** pointer.
3610  *	Return
3611  *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3612  *		case of failure.
3613  *
3614  * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3615  *	Description
3616  *		Set ECN (Explicit Congestion Notification) field of IP header
3617  *		to **CE** (Congestion Encountered) if current value is **ECT**
3618  *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3619  *		and IPv4.
3620  *	Return
3621  *		1 if the **CE** flag is set (either by the current helper call
3622  *		or because it was already present), 0 if it is not set.
3623  *
3624  * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3625  *	Description
3626  *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3627  *		**bpf_sk_release**\ () is unnecessary and not allowed.
3628  *	Return
3629  *		A **struct bpf_sock** pointer on success, or **NULL** in
3630  *		case of failure.
3631  *
3632  * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3633  *	Description
3634  *		Look for TCP socket matching *tuple*, optionally in a child
3635  *		network namespace *netns*. The return value must be checked,
3636  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3637  *
3638  *		This function is identical to **bpf_sk_lookup_tcp**\ (), except
3639  *		that it also returns timewait or request sockets. Use
3640  *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3641  *		full structure.
3642  *
3643  *		This helper is available only if the kernel was compiled with
3644  *		**CONFIG_NET** configuration option.
3645  *	Return
3646  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3647  *		For sockets with reuseport option, the **struct bpf_sock**
3648  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3649  *		tuple.
3650  *
3651  * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3652  * 	Description
3653  * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for
3654  * 		the listening socket in *sk*.
3655  *
3656  * 		*iph* points to the start of the IPv4 or IPv6 header, while
3657  * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3658  * 		**sizeof**\ (**struct ipv6hdr**).
3659  *
3660  * 		*th* points to the start of the TCP header, while *th_len*
3661  *		contains the length of the TCP header (at least
3662  *		**sizeof**\ (**struct tcphdr**)).
3663  * 	Return
3664  * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3665  * 		error otherwise.
3666  *
3667  * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3668  *	Description
3669  *		Get name of sysctl in /proc/sys/ and copy it into provided by
3670  *		program buffer *buf* of size *buf_len*.
3671  *
3672  *		The buffer is always NUL terminated, unless it's zero-sized.
3673  *
3674  *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3675  *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3676  *		only (e.g. "tcp_mem").
3677  *	Return
3678  *		Number of character copied (not including the trailing NUL).
3679  *
3680  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3681  *		truncated name in this case).
3682  *
3683  * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3684  *	Description
3685  *		Get current value of sysctl as it is presented in /proc/sys
3686  *		(incl. newline, etc), and copy it as a string into provided
3687  *		by program buffer *buf* of size *buf_len*.
3688  *
3689  *		The whole value is copied, no matter what file position user
3690  *		space issued e.g. sys_read at.
3691  *
3692  *		The buffer is always NUL terminated, unless it's zero-sized.
3693  *	Return
3694  *		Number of character copied (not including the trailing NUL).
3695  *
3696  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3697  *		truncated name in this case).
3698  *
3699  *		**-EINVAL** if current value was unavailable, e.g. because
3700  *		sysctl is uninitialized and read returns -EIO for it.
3701  *
3702  * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3703  *	Description
3704  *		Get new value being written by user space to sysctl (before
3705  *		the actual write happens) and copy it as a string into
3706  *		provided by program buffer *buf* of size *buf_len*.
3707  *
3708  *		User space may write new value at file position > 0.
3709  *
3710  *		The buffer is always NUL terminated, unless it's zero-sized.
3711  *	Return
3712  *		Number of character copied (not including the trailing NUL).
3713  *
3714  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3715  *		truncated name in this case).
3716  *
3717  *		**-EINVAL** if sysctl is being read.
3718  *
3719  * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3720  *	Description
3721  *		Override new value being written by user space to sysctl with
3722  *		value provided by program in buffer *buf* of size *buf_len*.
3723  *
3724  *		*buf* should contain a string in same form as provided by user
3725  *		space on sysctl write.
3726  *
3727  *		User space may write new value at file position > 0. To override
3728  *		the whole sysctl value file position should be set to zero.
3729  *	Return
3730  *		0 on success.
3731  *
3732  *		**-E2BIG** if the *buf_len* is too big.
3733  *
3734  *		**-EINVAL** if sysctl is being read.
3735  *
3736  * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3737  *	Description
3738  *		Convert the initial part of the string from buffer *buf* of
3739  *		size *buf_len* to a long integer according to the given base
3740  *		and save the result in *res*.
3741  *
3742  *		The string may begin with an arbitrary amount of white space
3743  *		(as determined by **isspace**\ (3)) followed by a single
3744  *		optional '**-**' sign.
3745  *
3746  *		Five least significant bits of *flags* encode base, other bits
3747  *		are currently unused.
3748  *
3749  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3750  *		similar to user space **strtol**\ (3).
3751  *	Return
3752  *		Number of characters consumed on success. Must be positive but
3753  *		no more than *buf_len*.
3754  *
3755  *		**-EINVAL** if no valid digits were found or unsupported base
3756  *		was provided.
3757  *
3758  *		**-ERANGE** if resulting value was out of range.
3759  *
3760  * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3761  *	Description
3762  *		Convert the initial part of the string from buffer *buf* of
3763  *		size *buf_len* to an unsigned long integer according to the
3764  *		given base and save the result in *res*.
3765  *
3766  *		The string may begin with an arbitrary amount of white space
3767  *		(as determined by **isspace**\ (3)).
3768  *
3769  *		Five least significant bits of *flags* encode base, other bits
3770  *		are currently unused.
3771  *
3772  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3773  *		similar to user space **strtoul**\ (3).
3774  *	Return
3775  *		Number of characters consumed on success. Must be positive but
3776  *		no more than *buf_len*.
3777  *
3778  *		**-EINVAL** if no valid digits were found or unsupported base
3779  *		was provided.
3780  *
3781  *		**-ERANGE** if resulting value was out of range.
3782  *
3783  * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3784  *	Description
3785  *		Get a bpf-local-storage from a *sk*.
3786  *
3787  *		Logically, it could be thought of getting the value from
3788  *		a *map* with *sk* as the **key**.  From this
3789  *		perspective,  the usage is not much different from
3790  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3791  *		helper enforces the key must be a full socket and the map must
3792  *		be a **BPF_MAP_TYPE_SK_STORAGE** also.
3793  *
3794  *		Underneath, the value is stored locally at *sk* instead of
3795  *		the *map*.  The *map* is used as the bpf-local-storage
3796  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
3797  *		searched against all bpf-local-storages residing at *sk*.
3798  *
3799  *		*sk* is a kernel **struct sock** pointer for LSM program.
3800  *		*sk* is a **struct bpf_sock** pointer for other program types.
3801  *
3802  *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3803  *		used such that a new bpf-local-storage will be
3804  *		created if one does not exist.  *value* can be used
3805  *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3806  *		the initial value of a bpf-local-storage.  If *value* is
3807  *		**NULL**, the new bpf-local-storage will be zero initialized.
3808  *	Return
3809  *		A bpf-local-storage pointer is returned on success.
3810  *
3811  *		**NULL** if not found or there was an error in adding
3812  *		a new bpf-local-storage.
3813  *
3814  * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3815  *	Description
3816  *		Delete a bpf-local-storage from a *sk*.
3817  *	Return
3818  *		0 on success.
3819  *
3820  *		**-ENOENT** if the bpf-local-storage cannot be found.
3821  *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3822  *
3823  * long bpf_send_signal(u32 sig)
3824  *	Description
3825  *		Send signal *sig* to the process of the current task.
3826  *		The signal may be delivered to any of this process's threads.
3827  *	Return
3828  *		0 on success or successfully queued.
3829  *
3830  *		**-EBUSY** if work queue under nmi is full.
3831  *
3832  *		**-EINVAL** if *sig* is invalid.
3833  *
3834  *		**-EPERM** if no permission to send the *sig*.
3835  *
3836  *		**-EAGAIN** if bpf program can try again.
3837  *
3838  * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3839  *	Description
3840  *		Try to issue a SYN cookie for the packet with corresponding
3841  *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3842  *
3843  *		*iph* points to the start of the IPv4 or IPv6 header, while
3844  *		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3845  *		**sizeof**\ (**struct ipv6hdr**).
3846  *
3847  *		*th* points to the start of the TCP header, while *th_len*
3848  *		contains the length of the TCP header with options (at least
3849  *		**sizeof**\ (**struct tcphdr**)).
3850  *	Return
3851  *		On success, lower 32 bits hold the generated SYN cookie in
3852  *		followed by 16 bits which hold the MSS value for that cookie,
3853  *		and the top 16 bits are unused.
3854  *
3855  *		On failure, the returned value is one of the following:
3856  *
3857  *		**-EINVAL** SYN cookie cannot be issued due to error
3858  *
3859  *		**-ENOENT** SYN cookie should not be issued (no SYN flood)
3860  *
3861  *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3862  *
3863  *		**-EPROTONOSUPPORT** IP packet version is not 4 or 6
3864  *
3865  * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3866  * 	Description
3867  * 		Write raw *data* blob into a special BPF perf event held by
3868  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3869  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
3870  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3871  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3872  *
3873  * 		The *flags* are used to indicate the index in *map* for which
3874  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
3875  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3876  * 		to indicate that the index of the current CPU core should be
3877  * 		used.
3878  *
3879  * 		The value to write, of *size*, is passed through eBPF stack and
3880  * 		pointed by *data*.
3881  *
3882  * 		*ctx* is a pointer to in-kernel struct sk_buff.
3883  *
3884  * 		This helper is similar to **bpf_perf_event_output**\ () but
3885  * 		restricted to raw_tracepoint bpf programs.
3886  * 	Return
3887  * 		0 on success, or a negative error in case of failure.
3888  *
3889  * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3890  * 	Description
3891  * 		Safely attempt to read *size* bytes from user space address
3892  * 		*unsafe_ptr* and store the data in *dst*.
3893  * 	Return
3894  * 		0 on success, or a negative error in case of failure.
3895  *
3896  * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3897  * 	Description
3898  * 		Safely attempt to read *size* bytes from kernel space address
3899  * 		*unsafe_ptr* and store the data in *dst*.
3900  * 	Return
3901  * 		0 on success, or a negative error in case of failure.
3902  *
3903  * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3904  * 	Description
3905  * 		Copy a NUL terminated string from an unsafe user address
3906  * 		*unsafe_ptr* to *dst*. The *size* should include the
3907  * 		terminating NUL byte. In case the string length is smaller than
3908  * 		*size*, the target is not padded with further NUL bytes. If the
3909  * 		string length is larger than *size*, just *size*-1 bytes are
3910  * 		copied and the last byte is set to NUL.
3911  *
3912  * 		On success, returns the number of bytes that were written,
3913  * 		including the terminal NUL. This makes this helper useful in
3914  * 		tracing programs for reading strings, and more importantly to
3915  * 		get its length at runtime. See the following snippet:
3916  *
3917  * 		::
3918  *
3919  * 			SEC("kprobe/sys_open")
3920  * 			void bpf_sys_open(struct pt_regs *ctx)
3921  * 			{
3922  * 			        char buf[PATHLEN]; // PATHLEN is defined to 256
3923  * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),
3924  * 				                                  ctx->di);
3925  *
3926  * 				// Consume buf, for example push it to
3927  * 				// userspace via bpf_perf_event_output(); we
3928  * 				// can use res (the string length) as event
3929  * 				// size, after checking its boundaries.
3930  * 			}
3931  *
3932  * 		In comparison, using **bpf_probe_read_user**\ () helper here
3933  * 		instead to read the string would require to estimate the length
3934  * 		at compile time, and would often result in copying more memory
3935  * 		than necessary.
3936  *
3937  * 		Another useful use case is when parsing individual process
3938  * 		arguments or individual environment variables navigating
3939  * 		*current*\ **->mm->arg_start** and *current*\
3940  * 		**->mm->env_start**: using this helper and the return value,
3941  * 		one can quickly iterate at the right offset of the memory area.
3942  * 	Return
3943  * 		On success, the strictly positive length of the output string,
3944  * 		including the trailing NUL character. On error, a negative
3945  * 		value.
3946  *
3947  * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3948  * 	Description
3949  * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3950  * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3951  * 	Return
3952  * 		On success, the strictly positive length of the string, including
3953  * 		the trailing NUL character. On error, a negative value.
3954  *
3955  * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3956  *	Description
3957  *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3958  *		*rcv_nxt* is the ack_seq to be sent out.
3959  *	Return
3960  *		0 on success, or a negative error in case of failure.
3961  *
3962  * long bpf_send_signal_thread(u32 sig)
3963  *	Description
3964  *		Send signal *sig* to the thread corresponding to the current task.
3965  *	Return
3966  *		0 on success or successfully queued.
3967  *
3968  *		**-EBUSY** if work queue under nmi is full.
3969  *
3970  *		**-EINVAL** if *sig* is invalid.
3971  *
3972  *		**-EPERM** if no permission to send the *sig*.
3973  *
3974  *		**-EAGAIN** if bpf program can try again.
3975  *
3976  * u64 bpf_jiffies64(void)
3977  *	Description
3978  *		Obtain the 64bit jiffies
3979  *	Return
3980  *		The 64 bit jiffies
3981  *
3982  * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3983  *	Description
3984  *		For an eBPF program attached to a perf event, retrieve the
3985  *		branch records (**struct perf_branch_entry**) associated to *ctx*
3986  *		and store it in the buffer pointed by *buf* up to size
3987  *		*size* bytes.
3988  *	Return
3989  *		On success, number of bytes written to *buf*. On error, a
3990  *		negative value.
3991  *
3992  *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3993  *		instead return the number of bytes required to store all the
3994  *		branch entries. If this flag is set, *buf* may be NULL.
3995  *
3996  *		**-EINVAL** if arguments invalid or **size** not a multiple
3997  *		of **sizeof**\ (**struct perf_branch_entry**\ ).
3998  *
3999  *		**-ENOENT** if architecture does not support branch records.
4000  *
4001  * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
4002  *	Description
4003  *		Returns 0 on success, values for *pid* and *tgid* as seen from the current
4004  *		*namespace* will be returned in *nsdata*.
4005  *	Return
4006  *		0 on success, or one of the following in case of failure:
4007  *
4008  *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number
4009  *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
4010  *
4011  *		**-ENOENT** if pidns does not exists for the current task.
4012  *
4013  * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
4014  *	Description
4015  *		Write raw *data* blob into a special BPF perf event held by
4016  *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
4017  *		event must have the following attributes: **PERF_SAMPLE_RAW**
4018  *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
4019  *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
4020  *
4021  *		The *flags* are used to indicate the index in *map* for which
4022  *		the value must be put, masked with **BPF_F_INDEX_MASK**.
4023  *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
4024  *		to indicate that the index of the current CPU core should be
4025  *		used.
4026  *
4027  *		The value to write, of *size*, is passed through eBPF stack and
4028  *		pointed by *data*.
4029  *
4030  *		*ctx* is a pointer to in-kernel struct xdp_buff.
4031  *
4032  *		This helper is similar to **bpf_perf_eventoutput**\ () but
4033  *		restricted to raw_tracepoint bpf programs.
4034  *	Return
4035  *		0 on success, or a negative error in case of failure.
4036  *
4037  * u64 bpf_get_netns_cookie(void *ctx)
4038  * 	Description
4039  * 		Retrieve the cookie (generated by the kernel) of the network
4040  * 		namespace the input *ctx* is associated with. The network
4041  * 		namespace cookie remains stable for its lifetime and provides
4042  * 		a global identifier that can be assumed unique. If *ctx* is
4043  * 		NULL, then the helper returns the cookie for the initial
4044  * 		network namespace. The cookie itself is very similar to that
4045  * 		of **bpf_get_socket_cookie**\ () helper, but for network
4046  * 		namespaces instead of sockets.
4047  * 	Return
4048  * 		A 8-byte long opaque number.
4049  *
4050  * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4051  * 	Description
4052  * 		Return id of cgroup v2 that is ancestor of the cgroup associated
4053  * 		with the current task at the *ancestor_level*. The root cgroup
4054  * 		is at *ancestor_level* zero and each step down the hierarchy
4055  * 		increments the level. If *ancestor_level* == level of cgroup
4056  * 		associated with the current task, then return value will be the
4057  * 		same as that of **bpf_get_current_cgroup_id**\ ().
4058  *
4059  * 		The helper is useful to implement policies based on cgroups
4060  * 		that are upper in hierarchy than immediate cgroup associated
4061  * 		with the current task.
4062  *
4063  * 		The format of returned id and helper limitations are same as in
4064  * 		**bpf_get_current_cgroup_id**\ ().
4065  * 	Return
4066  * 		The id is returned or 0 in case the id could not be retrieved.
4067  *
4068  * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4069  *	Description
4070  *		Helper is overloaded depending on BPF program type. This
4071  *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4072  *		**BPF_PROG_TYPE_SCHED_ACT** programs.
4073  *
4074  *		Assign the *sk* to the *skb*. When combined with appropriate
4075  *		routing configuration to receive the packet towards the socket,
4076  *		will cause *skb* to be delivered to the specified socket.
4077  *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),
4078  *		**bpf_clone_redirect**\ () or other methods outside of BPF may
4079  *		interfere with successful delivery to the socket.
4080  *
4081  *		This operation is only valid from TC ingress path.
4082  *
4083  *		The *flags* argument must be zero.
4084  *	Return
4085  *		0 on success, or a negative error in case of failure:
4086  *
4087  *		**-EINVAL** if specified *flags* are not supported.
4088  *
4089  *		**-ENOENT** if the socket is unavailable for assignment.
4090  *
4091  *		**-ENETUNREACH** if the socket is unreachable (wrong netns).
4092  *
4093  *		**-EOPNOTSUPP** if the operation is not supported, for example
4094  *		a call from outside of TC ingress.
4095  *
4096  *		**-ESOCKTNOSUPPORT** if the socket type is not supported
4097  *		(reuseport).
4098  *
4099  * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4100  *	Description
4101  *		Helper is overloaded depending on BPF program type. This
4102  *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4103  *
4104  *		Select the *sk* as a result of a socket lookup.
4105  *
4106  *		For the operation to succeed passed socket must be compatible
4107  *		with the packet description provided by the *ctx* object.
4108  *
4109  *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4110  *		be an exact match. While IP family (**AF_INET** or
4111  *		**AF_INET6**) must be compatible, that is IPv6 sockets
4112  *		that are not v6-only can be selected for IPv4 packets.
4113  *
4114  *		Only TCP listeners and UDP unconnected sockets can be
4115  *		selected. *sk* can also be NULL to reset any previous
4116  *		selection.
4117  *
4118  *		*flags* argument can combination of following values:
4119  *
4120  *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4121  *		  socket selection, potentially done by a BPF program
4122  *		  that ran before us.
4123  *
4124  *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4125  *		  load-balancing within reuseport group for the socket
4126  *		  being selected.
4127  *
4128  *		On success *ctx->sk* will point to the selected socket.
4129  *
4130  *	Return
4131  *		0 on success, or a negative errno in case of failure.
4132  *
4133  *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is
4134  *		  not compatible with packet family (*ctx->family*).
4135  *
4136  *		* **-EEXIST** if socket has been already selected,
4137  *		  potentially by another program, and
4138  *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4139  *
4140  *		* **-EINVAL** if unsupported flags were specified.
4141  *
4142  *		* **-EPROTOTYPE** if socket L4 protocol
4143  *		  (*sk->protocol*) doesn't match packet protocol
4144  *		  (*ctx->protocol*).
4145  *
4146  *		* **-ESOCKTNOSUPPORT** if socket is not in allowed
4147  *		  state (TCP listening or UDP unconnected).
4148  *
4149  * u64 bpf_ktime_get_boot_ns(void)
4150  * 	Description
4151  * 		Return the time elapsed since system boot, in nanoseconds.
4152  * 		Does include the time the system was suspended.
4153  * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4154  * 	Return
4155  * 		Current *ktime*.
4156  *
4157  * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4158  * 	Description
4159  * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4160  * 		out the format string.
4161  * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4162  * 		the format string itself. The *data* and *data_len* are format string
4163  * 		arguments. The *data* are a **u64** array and corresponding format string
4164  * 		values are stored in the array. For strings and pointers where pointees
4165  * 		are accessed, only the pointer values are stored in the *data* array.
4166  * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4167  *
4168  *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4169  *		Reading kernel memory may fail due to either invalid address or
4170  *		valid address but requiring a major memory fault. If reading kernel memory
4171  *		fails, the string for **%s** will be an empty string, and the ip
4172  *		address for **%p{i,I}{4,6}** will be 0. Not returning error to
4173  *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4174  * 	Return
4175  * 		0 on success, or a negative error in case of failure:
4176  *
4177  *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again
4178  *		by returning 1 from bpf program.
4179  *
4180  *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4181  *
4182  *		**-E2BIG** if *fmt* contains too many format specifiers.
4183  *
4184  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4185  *
4186  * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4187  * 	Description
4188  * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4189  * 		The *m* represents the seq_file. The *data* and *len* represent the
4190  * 		data to write in bytes.
4191  * 	Return
4192  * 		0 on success, or a negative error in case of failure:
4193  *
4194  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4195  *
4196  * u64 bpf_sk_cgroup_id(void *sk)
4197  *	Description
4198  *		Return the cgroup v2 id of the socket *sk*.
4199  *
4200  *		*sk* must be a non-**NULL** pointer to a socket, e.g. one
4201  *		returned from **bpf_sk_lookup_xxx**\ (),
4202  *		**bpf_sk_fullsock**\ (), etc. The format of returned id is
4203  *		same as in **bpf_skb_cgroup_id**\ ().
4204  *
4205  *		This helper is available only if the kernel was compiled with
4206  *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4207  *	Return
4208  *		The id is returned or 0 in case the id could not be retrieved.
4209  *
4210  * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4211  *	Description
4212  *		Return id of cgroup v2 that is ancestor of cgroup associated
4213  *		with the *sk* at the *ancestor_level*.  The root cgroup is at
4214  *		*ancestor_level* zero and each step down the hierarchy
4215  *		increments the level. If *ancestor_level* == level of cgroup
4216  *		associated with *sk*, then return value will be same as that
4217  *		of **bpf_sk_cgroup_id**\ ().
4218  *
4219  *		The helper is useful to implement policies based on cgroups
4220  *		that are upper in hierarchy than immediate cgroup associated
4221  *		with *sk*.
4222  *
4223  *		The format of returned id and helper limitations are same as in
4224  *		**bpf_sk_cgroup_id**\ ().
4225  *	Return
4226  *		The id is returned or 0 in case the id could not be retrieved.
4227  *
4228  * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4229  * 	Description
4230  * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4231  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4232  * 		of new data availability is sent.
4233  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4234  * 		of new data availability is sent unconditionally.
4235  * 		If **0** is specified in *flags*, an adaptive notification
4236  * 		of new data availability is sent.
4237  *
4238  * 		An adaptive notification is a notification sent whenever the user-space
4239  * 		process has caught up and consumed all available payloads. In case the user-space
4240  * 		process is still processing a previous payload, then no notification is needed
4241  * 		as it will process the newly added payload automatically.
4242  * 	Return
4243  * 		0 on success, or a negative error in case of failure.
4244  *
4245  * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4246  * 	Description
4247  * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4248  * 		*flags* must be 0.
4249  * 	Return
4250  * 		Valid pointer with *size* bytes of memory available; NULL,
4251  * 		otherwise.
4252  *
4253  * void bpf_ringbuf_submit(void *data, u64 flags)
4254  * 	Description
4255  * 		Submit reserved ring buffer sample, pointed to by *data*.
4256  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4257  * 		of new data availability is sent.
4258  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4259  * 		of new data availability is sent unconditionally.
4260  * 		If **0** is specified in *flags*, an adaptive notification
4261  * 		of new data availability is sent.
4262  *
4263  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4264  * 	Return
4265  * 		Nothing. Always succeeds.
4266  *
4267  * void bpf_ringbuf_discard(void *data, u64 flags)
4268  * 	Description
4269  * 		Discard reserved ring buffer sample, pointed to by *data*.
4270  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4271  * 		of new data availability is sent.
4272  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4273  * 		of new data availability is sent unconditionally.
4274  * 		If **0** is specified in *flags*, an adaptive notification
4275  * 		of new data availability is sent.
4276  *
4277  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4278  * 	Return
4279  * 		Nothing. Always succeeds.
4280  *
4281  * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4282  *	Description
4283  *		Query various characteristics of provided ring buffer. What
4284  *		exactly is queries is determined by *flags*:
4285  *
4286  *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4287  *		* **BPF_RB_RING_SIZE**: The size of ring buffer.
4288  *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4289  *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4290  *
4291  *		Data returned is just a momentary snapshot of actual values
4292  *		and could be inaccurate, so this facility should be used to
4293  *		power heuristics and for reporting, not to make 100% correct
4294  *		calculation.
4295  *	Return
4296  *		Requested value, or 0, if *flags* are not recognized.
4297  *
4298  * long bpf_csum_level(struct sk_buff *skb, u64 level)
4299  * 	Description
4300  * 		Change the skbs checksum level by one layer up or down, or
4301  * 		reset it entirely to none in order to have the stack perform
4302  * 		checksum validation. The level is applicable to the following
4303  * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4304  * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4305  * 		through **bpf_skb_adjust_room**\ () helper with passing in
4306  * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call
4307  * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4308  * 		the UDP header is removed. Similarly, an encap of the latter
4309  * 		into the former could be accompanied by a helper call to
4310  * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4311  * 		skb is still intended to be processed in higher layers of the
4312  * 		stack instead of just egressing at tc.
4313  *
4314  * 		There are three supported level settings at this time:
4315  *
4316  * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4317  * 		  with CHECKSUM_UNNECESSARY.
4318  * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4319  * 		  with CHECKSUM_UNNECESSARY.
4320  * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4321  * 		  sets CHECKSUM_NONE to force checksum validation by the stack.
4322  * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4323  * 		  skb->csum_level.
4324  * 	Return
4325  * 		0 on success, or a negative error in case of failure. In the
4326  * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4327  * 		is returned or the error code -EACCES in case the skb is not
4328  * 		subject to CHECKSUM_UNNECESSARY.
4329  *
4330  * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4331  *	Description
4332  *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4333  *	Return
4334  *		*sk* if casting is valid, or **NULL** otherwise.
4335  *
4336  * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4337  *	Description
4338  *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4339  *	Return
4340  *		*sk* if casting is valid, or **NULL** otherwise.
4341  *
4342  * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4343  * 	Description
4344  *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4345  *	Return
4346  *		*sk* if casting is valid, or **NULL** otherwise.
4347  *
4348  * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4349  * 	Description
4350  *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4351  *	Return
4352  *		*sk* if casting is valid, or **NULL** otherwise.
4353  *
4354  * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4355  * 	Description
4356  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4357  *	Return
4358  *		*sk* if casting is valid, or **NULL** otherwise.
4359  *
4360  * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4361  *	Description
4362  *		Return a user or a kernel stack in bpf program provided buffer.
4363  *		To achieve this, the helper needs *task*, which is a valid
4364  *		pointer to **struct task_struct**. To store the stacktrace, the
4365  *		bpf program provides *buf* with a nonnegative *size*.
4366  *
4367  *		The last argument, *flags*, holds the number of stack frames to
4368  *		skip (from 0 to 255), masked with
4369  *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4370  *		the following flags:
4371  *
4372  *		**BPF_F_USER_STACK**
4373  *			Collect a user space stack instead of a kernel stack.
4374  *		**BPF_F_USER_BUILD_ID**
4375  *			Collect buildid+offset instead of ips for user stack,
4376  *			only valid if **BPF_F_USER_STACK** is also specified.
4377  *
4378  *		**bpf_get_task_stack**\ () can collect up to
4379  *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4380  *		to sufficient large buffer size. Note that
4381  *		this limit can be controlled with the **sysctl** program, and
4382  *		that it should be manually increased in order to profile long
4383  *		user stacks (such as stacks for Java programs). To do so, use:
4384  *
4385  *		::
4386  *
4387  *			# sysctl kernel.perf_event_max_stack=<new value>
4388  *	Return
4389  * 		The non-negative copied *buf* length equal to or less than
4390  * 		*size* on success, or a negative error in case of failure.
4391  *
4392  * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4393  *	Description
4394  *		Load header option.  Support reading a particular TCP header
4395  *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4396  *
4397  *		If *flags* is 0, it will search the option from the
4398  *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
4399  *		has details on what skb_data contains under different
4400  *		*skops*\ **->op**.
4401  *
4402  *		The first byte of the *searchby_res* specifies the
4403  *		kind that it wants to search.
4404  *
4405  *		If the searching kind is an experimental kind
4406  *		(i.e. 253 or 254 according to RFC6994).  It also
4407  *		needs to specify the "magic" which is either
4408  *		2 bytes or 4 bytes.  It then also needs to
4409  *		specify the size of the magic by using
4410  *		the 2nd byte which is "kind-length" of a TCP
4411  *		header option and the "kind-length" also
4412  *		includes the first 2 bytes "kind" and "kind-length"
4413  *		itself as a normal TCP header option also does.
4414  *
4415  *		For example, to search experimental kind 254 with
4416  *		2 byte magic 0xeB9F, the searchby_res should be
4417  *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4418  *
4419  *		To search for the standard window scale option (3),
4420  *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4421  *		Note, kind-length must be 0 for regular option.
4422  *
4423  *		Searching for No-Op (0) and End-of-Option-List (1) are
4424  *		not supported.
4425  *
4426  *		*len* must be at least 2 bytes which is the minimal size
4427  *		of a header option.
4428  *
4429  *		Supported flags:
4430  *
4431  *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4432  *		  saved_syn packet or the just-received syn packet.
4433  *
4434  *	Return
4435  *		> 0 when found, the header option is copied to *searchby_res*.
4436  *		The return value is the total length copied. On failure, a
4437  *		negative error code is returned:
4438  *
4439  *		**-EINVAL** if a parameter is invalid.
4440  *
4441  *		**-ENOMSG** if the option is not found.
4442  *
4443  *		**-ENOENT** if no syn packet is available when
4444  *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4445  *
4446  *		**-ENOSPC** if there is not enough space.  Only *len* number of
4447  *		bytes are copied.
4448  *
4449  *		**-EFAULT** on failure to parse the header options in the
4450  *		packet.
4451  *
4452  *		**-EPERM** if the helper cannot be used under the current
4453  *		*skops*\ **->op**.
4454  *
4455  * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4456  *	Description
4457  *		Store header option.  The data will be copied
4458  *		from buffer *from* with length *len* to the TCP header.
4459  *
4460  *		The buffer *from* should have the whole option that
4461  *		includes the kind, kind-length, and the actual
4462  *		option data.  The *len* must be at least kind-length
4463  *		long.  The kind-length does not have to be 4 byte
4464  *		aligned.  The kernel will take care of the padding
4465  *		and setting the 4 bytes aligned value to th->doff.
4466  *
4467  *		This helper will check for duplicated option
4468  *		by searching the same option in the outgoing skb.
4469  *
4470  *		This helper can only be called during
4471  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4472  *
4473  *	Return
4474  *		0 on success, or negative error in case of failure:
4475  *
4476  *		**-EINVAL** If param is invalid.
4477  *
4478  *		**-ENOSPC** if there is not enough space in the header.
4479  *		Nothing has been written
4480  *
4481  *		**-EEXIST** if the option already exists.
4482  *
4483  *		**-EFAULT** on failure to parse the existing header options.
4484  *
4485  *		**-EPERM** if the helper cannot be used under the current
4486  *		*skops*\ **->op**.
4487  *
4488  * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4489  *	Description
4490  *		Reserve *len* bytes for the bpf header option.  The
4491  *		space will be used by **bpf_store_hdr_opt**\ () later in
4492  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4493  *
4494  *		If **bpf_reserve_hdr_opt**\ () is called multiple times,
4495  *		the total number of bytes will be reserved.
4496  *
4497  *		This helper can only be called during
4498  *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4499  *
4500  *	Return
4501  *		0 on success, or negative error in case of failure:
4502  *
4503  *		**-EINVAL** if a parameter is invalid.
4504  *
4505  *		**-ENOSPC** if there is not enough space in the header.
4506  *
4507  *		**-EPERM** if the helper cannot be used under the current
4508  *		*skops*\ **->op**.
4509  *
4510  * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4511  *	Description
4512  *		Get a bpf_local_storage from an *inode*.
4513  *
4514  *		Logically, it could be thought of as getting the value from
4515  *		a *map* with *inode* as the **key**.  From this
4516  *		perspective,  the usage is not much different from
4517  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4518  *		helper enforces the key must be an inode and the map must also
4519  *		be a **BPF_MAP_TYPE_INODE_STORAGE**.
4520  *
4521  *		Underneath, the value is stored locally at *inode* instead of
4522  *		the *map*.  The *map* is used as the bpf-local-storage
4523  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4524  *		searched against all bpf_local_storage residing at *inode*.
4525  *
4526  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4527  *		used such that a new bpf_local_storage will be
4528  *		created if one does not exist.  *value* can be used
4529  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4530  *		the initial value of a bpf_local_storage.  If *value* is
4531  *		**NULL**, the new bpf_local_storage will be zero initialized.
4532  *	Return
4533  *		A bpf_local_storage pointer is returned on success.
4534  *
4535  *		**NULL** if not found or there was an error in adding
4536  *		a new bpf_local_storage.
4537  *
4538  * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4539  *	Description
4540  *		Delete a bpf_local_storage from an *inode*.
4541  *	Return
4542  *		0 on success.
4543  *
4544  *		**-ENOENT** if the bpf_local_storage cannot be found.
4545  *
4546  * long bpf_d_path(struct path *path, char *buf, u32 sz)
4547  *	Description
4548  *		Return full path for given **struct path** object, which
4549  *		needs to be the kernel BTF *path* object. The path is
4550  *		returned in the provided buffer *buf* of size *sz* and
4551  *		is zero terminated.
4552  *
4553  *	Return
4554  *		On success, the strictly positive length of the string,
4555  *		including the trailing NUL character. On error, a negative
4556  *		value.
4557  *
4558  * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4559  * 	Description
4560  * 		Read *size* bytes from user space address *user_ptr* and store
4561  * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4562  * 	Return
4563  * 		0 on success, or a negative error in case of failure.
4564  *
4565  * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4566  *	Description
4567  *		Use BTF to store a string representation of *ptr*->ptr in *str*,
4568  *		using *ptr*->type_id.  This value should specify the type
4569  *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4570  *		can be used to look up vmlinux BTF type ids. Traversing the
4571  *		data structure using BTF, the type information and values are
4572  *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
4573  *		the pointer data is carried out to avoid kernel crashes during
4574  *		operation.  Smaller types can use string space on the stack;
4575  *		larger programs can use map data to store the string
4576  *		representation.
4577  *
4578  *		The string can be subsequently shared with userspace via
4579  *		bpf_perf_event_output() or ring buffer interfaces.
4580  *		bpf_trace_printk() is to be avoided as it places too small
4581  *		a limit on string size to be useful.
4582  *
4583  *		*flags* is a combination of
4584  *
4585  *		**BTF_F_COMPACT**
4586  *			no formatting around type information
4587  *		**BTF_F_NONAME**
4588  *			no struct/union member names/types
4589  *		**BTF_F_PTR_RAW**
4590  *			show raw (unobfuscated) pointer values;
4591  *			equivalent to printk specifier %px.
4592  *		**BTF_F_ZERO**
4593  *			show zero-valued struct/union members; they
4594  *			are not displayed by default
4595  *
4596  *	Return
4597  *		The number of bytes that were written (or would have been
4598  *		written if output had to be truncated due to string size),
4599  *		or a negative error in cases of failure.
4600  *
4601  * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4602  *	Description
4603  *		Use BTF to write to seq_write a string representation of
4604  *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4605  *		*flags* are identical to those used for bpf_snprintf_btf.
4606  *	Return
4607  *		0 on success or a negative error in case of failure.
4608  *
4609  * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4610  * 	Description
4611  * 		See **bpf_get_cgroup_classid**\ () for the main description.
4612  * 		This helper differs from **bpf_get_cgroup_classid**\ () in that
4613  * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s
4614  * 		associated socket instead of the current process.
4615  * 	Return
4616  * 		The id is returned or 0 in case the id could not be retrieved.
4617  *
4618  * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4619  * 	Description
4620  * 		Redirect the packet to another net device of index *ifindex*
4621  * 		and fill in L2 addresses from neighboring subsystem. This helper
4622  * 		is somewhat similar to **bpf_redirect**\ (), except that it
4623  * 		populates L2 addresses as well, meaning, internally, the helper
4624  * 		relies on the neighbor lookup for the L2 address of the nexthop.
4625  *
4626  * 		The helper will perform a FIB lookup based on the skb's
4627  * 		networking header to get the address of the next hop, unless
4628  * 		this is supplied by the caller in the *params* argument. The
4629  * 		*plen* argument indicates the len of *params* and should be set
4630  * 		to 0 if *params* is NULL.
4631  *
4632  * 		The *flags* argument is reserved and must be 0. The helper is
4633  * 		currently only supported for tc BPF program types, and enabled
4634  * 		for IPv4 and IPv6 protocols.
4635  * 	Return
4636  * 		The helper returns **TC_ACT_REDIRECT** on success or
4637  * 		**TC_ACT_SHOT** on error.
4638  *
4639  * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4640  *     Description
4641  *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4642  *             pointer to the percpu kernel variable on *cpu*. A ksym is an
4643  *             extern variable decorated with '__ksym'. For ksym, there is a
4644  *             global var (either static or global) defined of the same name
4645  *             in the kernel. The ksym is percpu if the global var is percpu.
4646  *             The returned pointer points to the global percpu var on *cpu*.
4647  *
4648  *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4649  *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
4650  *             happens if *cpu* is larger than nr_cpu_ids. The caller of
4651  *             bpf_per_cpu_ptr() must check the returned value.
4652  *     Return
4653  *             A pointer pointing to the kernel percpu variable on *cpu*, or
4654  *             NULL, if *cpu* is invalid.
4655  *
4656  * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4657  *	Description
4658  *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4659  *		pointer to the percpu kernel variable on this cpu. See the
4660  *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4661  *
4662  *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4663  *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4664  *		never return NULL.
4665  *	Return
4666  *		A pointer pointing to the kernel percpu variable on this cpu.
4667  *
4668  * long bpf_redirect_peer(u32 ifindex, u64 flags)
4669  * 	Description
4670  * 		Redirect the packet to another net device of index *ifindex*.
4671  * 		This helper is somewhat similar to **bpf_redirect**\ (), except
4672  * 		that the redirection happens to the *ifindex*' peer device and
4673  * 		the netns switch takes place from ingress to ingress without
4674  * 		going through the CPU's backlog queue.
4675  *
4676  * 		The *flags* argument is reserved and must be 0. The helper is
4677  * 		currently only supported for tc BPF program types at the ingress
4678  * 		hook and for veth device types. The peer device must reside in a
4679  * 		different network namespace.
4680  * 	Return
4681  * 		The helper returns **TC_ACT_REDIRECT** on success or
4682  * 		**TC_ACT_SHOT** on error.
4683  *
4684  * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4685  *	Description
4686  *		Get a bpf_local_storage from the *task*.
4687  *
4688  *		Logically, it could be thought of as getting the value from
4689  *		a *map* with *task* as the **key**.  From this
4690  *		perspective,  the usage is not much different from
4691  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4692  *		helper enforces the key must be a task_struct and the map must also
4693  *		be a **BPF_MAP_TYPE_TASK_STORAGE**.
4694  *
4695  *		Underneath, the value is stored locally at *task* instead of
4696  *		the *map*.  The *map* is used as the bpf-local-storage
4697  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4698  *		searched against all bpf_local_storage residing at *task*.
4699  *
4700  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4701  *		used such that a new bpf_local_storage will be
4702  *		created if one does not exist.  *value* can be used
4703  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4704  *		the initial value of a bpf_local_storage.  If *value* is
4705  *		**NULL**, the new bpf_local_storage will be zero initialized.
4706  *	Return
4707  *		A bpf_local_storage pointer is returned on success.
4708  *
4709  *		**NULL** if not found or there was an error in adding
4710  *		a new bpf_local_storage.
4711  *
4712  * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4713  *	Description
4714  *		Delete a bpf_local_storage from a *task*.
4715  *	Return
4716  *		0 on success.
4717  *
4718  *		**-ENOENT** if the bpf_local_storage cannot be found.
4719  *
4720  * struct task_struct *bpf_get_current_task_btf(void)
4721  *	Description
4722  *		Return a BTF pointer to the "current" task.
4723  *		This pointer can also be used in helpers that accept an
4724  *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.
4725  *	Return
4726  *		Pointer to the current task.
4727  *
4728  * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4729  *	Description
4730  *		Set or clear certain options on *bprm*:
4731  *
4732  *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4733  *		which sets the **AT_SECURE** auxv for glibc. The bit
4734  *		is cleared if the flag is not specified.
4735  *	Return
4736  *		**-EINVAL** if invalid *flags* are passed, zero otherwise.
4737  *
4738  * u64 bpf_ktime_get_coarse_ns(void)
4739  * 	Description
4740  * 		Return a coarse-grained version of the time elapsed since
4741  * 		system boot, in nanoseconds. Does not include time the system
4742  * 		was suspended.
4743  *
4744  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4745  * 	Return
4746  * 		Current *ktime*.
4747  *
4748  * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4749  *	Description
4750  *		Returns the stored IMA hash of the *inode* (if it's available).
4751  *		If the hash is larger than *size*, then only *size*
4752  *		bytes will be copied to *dst*
4753  *	Return
4754  *		The **hash_algo** is returned on success,
4755  *		**-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4756  *		invalid arguments are passed.
4757  *
4758  * struct socket *bpf_sock_from_file(struct file *file)
4759  *	Description
4760  *		If the given file represents a socket, returns the associated
4761  *		socket.
4762  *	Return
4763  *		A pointer to a struct socket on success or NULL if the file is
4764  *		not a socket.
4765  *
4766  * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4767  *	Description
4768  *		Check packet size against exceeding MTU of net device (based
4769  *		on *ifindex*).  This helper will likely be used in combination
4770  *		with helpers that adjust/change the packet size.
4771  *
4772  *		The argument *len_diff* can be used for querying with a planned
4773  *		size change. This allows to check MTU prior to changing packet
4774  *		ctx. Providing a *len_diff* adjustment that is larger than the
4775  *		actual packet size (resulting in negative packet size) will in
4776  *		principle not exceed the MTU, which is why it is not considered
4777  *		a failure.  Other BPF helpers are needed for performing the
4778  *		planned size change; therefore the responsibility for catching
4779  *		a negative packet size belongs in those helpers.
4780  *
4781  *		Specifying *ifindex* zero means the MTU check is performed
4782  *		against the current net device.  This is practical if this isn't
4783  *		used prior to redirect.
4784  *
4785  *		On input *mtu_len* must be a valid pointer, else verifier will
4786  *		reject BPF program.  If the value *mtu_len* is initialized to
4787  *		zero then the ctx packet size is use.  When value *mtu_len* is
4788  *		provided as input this specify the L3 length that the MTU check
4789  *		is done against. Remember XDP and TC length operate at L2, but
4790  *		this value is L3 as this correlate to MTU and IP-header tot_len
4791  *		values which are L3 (similar behavior as bpf_fib_lookup).
4792  *
4793  *		The Linux kernel route table can configure MTUs on a more
4794  *		specific per route level, which is not provided by this helper.
4795  *		For route level MTU checks use the **bpf_fib_lookup**\ ()
4796  *		helper.
4797  *
4798  *		*ctx* is either **struct xdp_md** for XDP programs or
4799  *		**struct sk_buff** for tc cls_act programs.
4800  *
4801  *		The *flags* argument can be a combination of one or more of the
4802  *		following values:
4803  *
4804  *		**BPF_MTU_CHK_SEGS**
4805  *			This flag will only works for *ctx* **struct sk_buff**.
4806  *			If packet context contains extra packet segment buffers
4807  *			(often knows as GSO skb), then MTU check is harder to
4808  *			check at this point, because in transmit path it is
4809  *			possible for the skb packet to get re-segmented
4810  *			(depending on net device features).  This could still be
4811  *			a MTU violation, so this flag enables performing MTU
4812  *			check against segments, with a different violation
4813  *			return code to tell it apart. Check cannot use len_diff.
4814  *
4815  *		On return *mtu_len* pointer contains the MTU value of the net
4816  *		device.  Remember the net device configured MTU is the L3 size,
4817  *		which is returned here and XDP and TC length operate at L2.
4818  *		Helper take this into account for you, but remember when using
4819  *		MTU value in your BPF-code.
4820  *
4821  *	Return
4822  *		* 0 on success, and populate MTU value in *mtu_len* pointer.
4823  *
4824  *		* < 0 if any input argument is invalid (*mtu_len* not updated)
4825  *
4826  *		MTU violations return positive values, but also populate MTU
4827  *		value in *mtu_len* pointer, as this can be needed for
4828  *		implementing PMTU handing:
4829  *
4830  *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**
4831  *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4832  *
4833  * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4834  *	Description
4835  *		For each element in **map**, call **callback_fn** function with
4836  *		**map**, **callback_ctx** and other map-specific parameters.
4837  *		The **callback_fn** should be a static function and
4838  *		the **callback_ctx** should be a pointer to the stack.
4839  *		The **flags** is used to control certain aspects of the helper.
4840  *		Currently, the **flags** must be 0.
4841  *
4842  *		The following are a list of supported map types and their
4843  *		respective expected callback signatures:
4844  *
4845  *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4846  *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4847  *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4848  *
4849  *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4850  *
4851  *		For per_cpu maps, the map_value is the value on the cpu where the
4852  *		bpf_prog is running.
4853  *
4854  *		If **callback_fn** return 0, the helper will continue to the next
4855  *		element. If return value is 1, the helper will skip the rest of
4856  *		elements and return. Other return values are not used now.
4857  *
4858  *	Return
4859  *		The number of traversed map elements for success, **-EINVAL** for
4860  *		invalid **flags**.
4861  *
4862  * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4863  *	Description
4864  *		Outputs a string into the **str** buffer of size **str_size**
4865  *		based on a format string stored in a read-only map pointed by
4866  *		**fmt**.
4867  *
4868  *		Each format specifier in **fmt** corresponds to one u64 element
4869  *		in the **data** array. For strings and pointers where pointees
4870  *		are accessed, only the pointer values are stored in the *data*
4871  *		array. The *data_len* is the size of *data* in bytes - must be
4872  *		a multiple of 8.
4873  *
4874  *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4875  *		memory. Reading kernel memory may fail due to either invalid
4876  *		address or valid address but requiring a major memory fault. If
4877  *		reading kernel memory fails, the string for **%s** will be an
4878  *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4879  *		Not returning error to bpf program is consistent with what
4880  *		**bpf_trace_printk**\ () does for now.
4881  *
4882  *	Return
4883  *		The strictly positive length of the formatted string, including
4884  *		the trailing zero character. If the return value is greater than
4885  *		**str_size**, **str** contains a truncated string, guaranteed to
4886  *		be zero-terminated except when **str_size** is 0.
4887  *
4888  *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4889  *
4890  * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
4891  * 	Description
4892  * 		Execute bpf syscall with given arguments.
4893  * 	Return
4894  * 		A syscall result.
4895  *
4896  * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
4897  * 	Description
4898  * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
4899  * 	Return
4900  * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.
4901  *
4902  * long bpf_sys_close(u32 fd)
4903  * 	Description
4904  * 		Execute close syscall for given FD.
4905  * 	Return
4906  * 		A syscall result.
4907  *
4908  * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
4909  *	Description
4910  *		Initialize the timer.
4911  *		First 4 bits of *flags* specify clockid.
4912  *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
4913  *		All other bits of *flags* are reserved.
4914  *		The verifier will reject the program if *timer* is not from
4915  *		the same *map*.
4916  *	Return
4917  *		0 on success.
4918  *		**-EBUSY** if *timer* is already initialized.
4919  *		**-EINVAL** if invalid *flags* are passed.
4920  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4921  *		The user space should either hold a file descriptor to a map with timers
4922  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4923  *		closed all timers in the map will be cancelled and freed.
4924  *
4925  * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
4926  *	Description
4927  *		Configure the timer to call *callback_fn* static function.
4928  *	Return
4929  *		0 on success.
4930  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4931  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4932  *		The user space should either hold a file descriptor to a map with timers
4933  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4934  *		closed all timers in the map will be cancelled and freed.
4935  *
4936  * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
4937  *	Description
4938  *		Set timer expiration N nanoseconds from the current time. The
4939  *		configured callback will be invoked in soft irq context on some cpu
4940  *		and will not repeat unless another bpf_timer_start() is made.
4941  *		In such case the next invocation can migrate to a different cpu.
4942  *		Since struct bpf_timer is a field inside map element the map
4943  *		owns the timer. The bpf_timer_set_callback() will increment refcnt
4944  *		of BPF program to make sure that callback_fn code stays valid.
4945  *		When user space reference to a map reaches zero all timers
4946  *		in a map are cancelled and corresponding program's refcnts are
4947  *		decremented. This is done to make sure that Ctrl-C of a user
4948  *		process doesn't leave any timers running. If map is pinned in
4949  *		bpffs the callback_fn can re-arm itself indefinitely.
4950  *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands
4951  *		cancel and free the timer in the given map element.
4952  *		The map can contain timers that invoke callback_fn-s from different
4953  *		programs. The same callback_fn can serve different timers from
4954  *		different maps if key/value layout matches across maps.
4955  *		Every bpf_timer_set_callback() can have different callback_fn.
4956  *
4957  *	Return
4958  *		0 on success.
4959  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
4960  *		or invalid *flags* are passed.
4961  *
4962  * long bpf_timer_cancel(struct bpf_timer *timer)
4963  *	Description
4964  *		Cancel the timer and wait for callback_fn to finish if it was running.
4965  *	Return
4966  *		0 if the timer was not active.
4967  *		1 if the timer was active.
4968  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4969  *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
4970  *		own timer which would have led to a deadlock otherwise.
4971  *
4972  * u64 bpf_get_func_ip(void *ctx)
4973  * 	Description
4974  * 		Get address of the traced function (for tracing and kprobe programs).
4975  * 	Return
4976  * 		Address of the traced function.
4977  * 		0 for kprobes placed within the function (not at the entry).
4978  *
4979  * u64 bpf_get_attach_cookie(void *ctx)
4980  * 	Description
4981  * 		Get bpf_cookie value provided (optionally) during the program
4982  * 		attachment. It might be different for each individual
4983  * 		attachment, even if BPF program itself is the same.
4984  * 		Expects BPF program context *ctx* as a first argument.
4985  *
4986  * 		Supported for the following program types:
4987  *			- kprobe/uprobe;
4988  *			- tracepoint;
4989  *			- perf_event.
4990  * 	Return
4991  *		Value specified by user at BPF link creation/attachment time
4992  *		or 0, if it was not specified.
4993  *
4994  * long bpf_task_pt_regs(struct task_struct *task)
4995  *	Description
4996  *		Get the struct pt_regs associated with **task**.
4997  *	Return
4998  *		A pointer to struct pt_regs.
4999  *
5000  * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
5001  *	Description
5002  *		Get branch trace from hardware engines like Intel LBR. The
5003  *		hardware engine is stopped shortly after the helper is
5004  *		called. Therefore, the user need to filter branch entries
5005  *		based on the actual use case. To capture branch trace
5006  *		before the trigger point of the BPF program, the helper
5007  *		should be called at the beginning of the BPF program.
5008  *
5009  *		The data is stored as struct perf_branch_entry into output
5010  *		buffer *entries*. *size* is the size of *entries* in bytes.
5011  *		*flags* is reserved for now and must be zero.
5012  *
5013  *	Return
5014  *		On success, number of bytes written to *buf*. On error, a
5015  *		negative value.
5016  *
5017  *		**-EINVAL** if *flags* is not zero.
5018  *
5019  *		**-ENOENT** if architecture does not support branch records.
5020  *
5021  * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
5022  *	Description
5023  *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
5024  *		to format and can handle more format args as a result.
5025  *
5026  *		Arguments are to be used as in **bpf_seq_printf**\ () helper.
5027  *	Return
5028  *		The number of bytes written to the buffer, or a negative error
5029  *		in case of failure.
5030  *
5031  * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
5032  * 	Description
5033  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
5034  *	Return
5035  *		*sk* if casting is valid, or **NULL** otherwise.
5036  *
5037  * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
5038  *	Description
5039  *		Get the address of a kernel symbol, returned in *res*. *res* is
5040  *		set to 0 if the symbol is not found.
5041  *	Return
5042  *		On success, zero. On error, a negative value.
5043  *
5044  *		**-EINVAL** if *flags* is not zero.
5045  *
5046  *		**-EINVAL** if string *name* is not the same size as *name_sz*.
5047  *
5048  *		**-ENOENT** if symbol is not found.
5049  *
5050  *		**-EPERM** if caller does not have permission to obtain kernel address.
5051  *
5052  * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5053  *	Description
5054  *		Find vma of *task* that contains *addr*, call *callback_fn*
5055  *		function with *task*, *vma*, and *callback_ctx*.
5056  *		The *callback_fn* should be a static function and
5057  *		the *callback_ctx* should be a pointer to the stack.
5058  *		The *flags* is used to control certain aspects of the helper.
5059  *		Currently, the *flags* must be 0.
5060  *
5061  *		The expected callback signature is
5062  *
5063  *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5064  *
5065  *	Return
5066  *		0 on success.
5067  *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5068  *		**-EBUSY** if failed to try lock mmap_lock.
5069  *		**-EINVAL** for invalid **flags**.
5070  *
5071  * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5072  *	Description
5073  *		For **nr_loops**, call **callback_fn** function
5074  *		with **callback_ctx** as the context parameter.
5075  *		The **callback_fn** should be a static function and
5076  *		the **callback_ctx** should be a pointer to the stack.
5077  *		The **flags** is used to control certain aspects of the helper.
5078  *		Currently, the **flags** must be 0. Currently, nr_loops is
5079  *		limited to 1 << 23 (~8 million) loops.
5080  *
5081  *		long (\*callback_fn)(u32 index, void \*ctx);
5082  *
5083  *		where **index** is the current index in the loop. The index
5084  *		is zero-indexed.
5085  *
5086  *		If **callback_fn** returns 0, the helper will continue to the next
5087  *		loop. If return value is 1, the helper will skip the rest of
5088  *		the loops and return. Other return values are not used now,
5089  *		and will be rejected by the verifier.
5090  *
5091  *	Return
5092  *		The number of loops performed, **-EINVAL** for invalid **flags**,
5093  *		**-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5094  *
5095  * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5096  *	Description
5097  *		Do strncmp() between **s1** and **s2**. **s1** doesn't need
5098  *		to be null-terminated and **s1_sz** is the maximum storage
5099  *		size of **s1**. **s2** must be a read-only string.
5100  *	Return
5101  *		An integer less than, equal to, or greater than zero
5102  *		if the first **s1_sz** bytes of **s1** is found to be
5103  *		less than, to match, or be greater than **s2**.
5104  *
5105  * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5106  *	Description
5107  *		Get **n**-th argument register (zero based) of the traced function (for tracing programs)
5108  *		returned in **value**.
5109  *
5110  *	Return
5111  *		0 on success.
5112  *		**-EINVAL** if n >= argument register count of traced function.
5113  *
5114  * long bpf_get_func_ret(void *ctx, u64 *value)
5115  *	Description
5116  *		Get return value of the traced function (for tracing programs)
5117  *		in **value**.
5118  *
5119  *	Return
5120  *		0 on success.
5121  *		**-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5122  *
5123  * long bpf_get_func_arg_cnt(void *ctx)
5124  *	Description
5125  *		Get number of registers of the traced function (for tracing programs) where
5126  *		function arguments are stored in these registers.
5127  *
5128  *	Return
5129  *		The number of argument registers of the traced function.
5130  *
5131  * int bpf_get_retval(void)
5132  *	Description
5133  *		Get the BPF program's return value that will be returned to the upper layers.
5134  *
5135  *		This helper is currently supported by cgroup programs and only by the hooks
5136  *		where BPF program's return value is returned to the userspace via errno.
5137  *	Return
5138  *		The BPF program's return value.
5139  *
5140  * int bpf_set_retval(int retval)
5141  *	Description
5142  *		Set the BPF program's return value that will be returned to the upper layers.
5143  *
5144  *		This helper is currently supported by cgroup programs and only by the hooks
5145  *		where BPF program's return value is returned to the userspace via errno.
5146  *
5147  *		Note that there is the following corner case where the program exports an error
5148  *		via bpf_set_retval but signals success via 'return 1':
5149  *
5150  *			bpf_set_retval(-EPERM);
5151  *			return 1;
5152  *
5153  *		In this case, the BPF program's return value will use helper's -EPERM. This
5154  *		still holds true for cgroup/bind{4,6} which supports extra 'return 3' success case.
5155  *
5156  *	Return
5157  *		0 on success, or a negative error in case of failure.
5158  *
5159  * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5160  *	Description
5161  *		Get the total size of a given xdp buff (linear and paged area)
5162  *	Return
5163  *		The total size of a given xdp buffer.
5164  *
5165  * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5166  *	Description
5167  *		This helper is provided as an easy way to load data from a
5168  *		xdp buffer. It can be used to load *len* bytes from *offset* from
5169  *		the frame associated to *xdp_md*, into the buffer pointed by
5170  *		*buf*.
5171  *	Return
5172  *		0 on success, or a negative error in case of failure.
5173  *
5174  * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5175  *	Description
5176  *		Store *len* bytes from buffer *buf* into the frame
5177  *		associated to *xdp_md*, at *offset*.
5178  *	Return
5179  *		0 on success, or a negative error in case of failure.
5180  *
5181  * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5182  *	Description
5183  *		Read *size* bytes from user space address *user_ptr* in *tsk*'s
5184  *		address space, and stores the data in *dst*. *flags* is not
5185  *		used yet and is provided for future extensibility. This helper
5186  *		can only be used by sleepable programs.
5187  *	Return
5188  *		0 on success, or a negative error in case of failure. On error
5189  *		*dst* buffer is zeroed out.
5190  *
5191  * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5192  *	Description
5193  *		Change the __sk_buff->tstamp_type to *tstamp_type*
5194  *		and set *tstamp* to the __sk_buff->tstamp together.
5195  *
5196  *		If there is no need to change the __sk_buff->tstamp_type,
5197  *		the tstamp value can be directly written to __sk_buff->tstamp
5198  *		instead.
5199  *
5200  *		BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5201  *		will be kept during bpf_redirect_*().  A non zero
5202  *		*tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5203  *		*tstamp_type*.
5204  *
5205  *		A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5206  *		with a zero *tstamp*.
5207  *
5208  *		Only IPv4 and IPv6 skb->protocol are supported.
5209  *
5210  *		This function is most useful when it needs to set a
5211  *		mono delivery time to __sk_buff->tstamp and then
5212  *		bpf_redirect_*() to the egress of an iface.  For example,
5213  *		changing the (rcv) timestamp in __sk_buff->tstamp at
5214  *		ingress to a mono delivery time and then bpf_redirect_*()
5215  *		to sch_fq@phy-dev.
5216  *	Return
5217  *		0 on success.
5218  *		**-EINVAL** for invalid input
5219  *		**-EOPNOTSUPP** for unsupported protocol
5220  *
5221  * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5222  *	Description
5223  *		Returns a calculated IMA hash of the *file*.
5224  *		If the hash is larger than *size*, then only *size*
5225  *		bytes will be copied to *dst*
5226  *	Return
5227  *		The **hash_algo** is returned on success,
5228  *		**-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
5229  *		invalid arguments are passed.
5230  *
5231  * void *bpf_kptr_xchg(void *map_value, void *ptr)
5232  *	Description
5233  *		Exchange kptr at pointer *map_value* with *ptr*, and return the
5234  *		old value. *ptr* can be NULL, otherwise it must be a referenced
5235  *		pointer which will be released when this helper is called.
5236  *	Return
5237  *		The old value of kptr (which can be NULL). The returned pointer
5238  *		if not NULL, is a reference which must be released using its
5239  *		corresponding release function, or moved into a BPF map before
5240  *		program exit.
5241  *
5242  * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5243  * 	Description
5244  * 		Perform a lookup in *percpu map* for an entry associated to
5245  * 		*key* on *cpu*.
5246  * 	Return
5247  * 		Map value associated to *key* on *cpu*, or **NULL** if no entry
5248  * 		was found or *cpu* is invalid.
5249  *
5250  * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5251  *	Description
5252  *		Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5253  *	Return
5254  *		*sk* if casting is valid, or **NULL** otherwise.
5255  *
5256  * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5257  *	Description
5258  *		Get a dynptr to local memory *data*.
5259  *
5260  *		*data* must be a ptr to a map value.
5261  *		The maximum *size* supported is DYNPTR_MAX_SIZE.
5262  *		*flags* is currently unused.
5263  *	Return
5264  *		0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5265  *		-EINVAL if flags is not 0.
5266  *
5267  * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5268  *	Description
5269  *		Reserve *size* bytes of payload in a ring buffer *ringbuf*
5270  *		through the dynptr interface. *flags* must be 0.
5271  *
5272  *		Please note that a corresponding bpf_ringbuf_submit_dynptr or
5273  *		bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5274  *		reservation fails. This is enforced by the verifier.
5275  *	Return
5276  *		0 on success, or a negative error in case of failure.
5277  *
5278  * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5279  *	Description
5280  *		Submit reserved ring buffer sample, pointed to by *data*,
5281  *		through the dynptr interface. This is a no-op if the dynptr is
5282  *		invalid/null.
5283  *
5284  *		For more information on *flags*, please see
5285  *		'bpf_ringbuf_submit'.
5286  *	Return
5287  *		Nothing. Always succeeds.
5288  *
5289  * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5290  *	Description
5291  *		Discard reserved ring buffer sample through the dynptr
5292  *		interface. This is a no-op if the dynptr is invalid/null.
5293  *
5294  *		For more information on *flags*, please see
5295  *		'bpf_ringbuf_discard'.
5296  *	Return
5297  *		Nothing. Always succeeds.
5298  *
5299  * long bpf_dynptr_read(void *dst, u32 len, const struct bpf_dynptr *src, u32 offset, u64 flags)
5300  *	Description
5301  *		Read *len* bytes from *src* into *dst*, starting from *offset*
5302  *		into *src*.
5303  *		*flags* is currently unused.
5304  *	Return
5305  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5306  *		of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5307  *		*flags* is not 0.
5308  *
5309  * long bpf_dynptr_write(const struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5310  *	Description
5311  *		Write *len* bytes from *src* into *dst*, starting from *offset*
5312  *		into *dst*.
5313  *		*flags* is currently unused.
5314  *	Return
5315  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5316  *		of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5317  *		is a read-only dynptr or if *flags* is not 0.
5318  *
5319  * void *bpf_dynptr_data(const struct bpf_dynptr *ptr, u32 offset, u32 len)
5320  *	Description
5321  *		Get a pointer to the underlying dynptr data.
5322  *
5323  *		*len* must be a statically known value. The returned data slice
5324  *		is invalidated whenever the dynptr is invalidated.
5325  *	Return
5326  *		Pointer to the underlying dynptr data, NULL if the dynptr is
5327  *		read-only, if the dynptr is invalid, or if the offset and length
5328  *		is out of bounds.
5329  *
5330  * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5331  *	Description
5332  *		Try to issue a SYN cookie for the packet with corresponding
5333  *		IPv4/TCP headers, *iph* and *th*, without depending on a
5334  *		listening socket.
5335  *
5336  *		*iph* points to the IPv4 header.
5337  *
5338  *		*th* points to the start of the TCP header, while *th_len*
5339  *		contains the length of the TCP header (at least
5340  *		**sizeof**\ (**struct tcphdr**)).
5341  *	Return
5342  *		On success, lower 32 bits hold the generated SYN cookie in
5343  *		followed by 16 bits which hold the MSS value for that cookie,
5344  *		and the top 16 bits are unused.
5345  *
5346  *		On failure, the returned value is one of the following:
5347  *
5348  *		**-EINVAL** if *th_len* is invalid.
5349  *
5350  * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5351  *	Description
5352  *		Try to issue a SYN cookie for the packet with corresponding
5353  *		IPv6/TCP headers, *iph* and *th*, without depending on a
5354  *		listening socket.
5355  *
5356  *		*iph* points to the IPv6 header.
5357  *
5358  *		*th* points to the start of the TCP header, while *th_len*
5359  *		contains the length of the TCP header (at least
5360  *		**sizeof**\ (**struct tcphdr**)).
5361  *	Return
5362  *		On success, lower 32 bits hold the generated SYN cookie in
5363  *		followed by 16 bits which hold the MSS value for that cookie,
5364  *		and the top 16 bits are unused.
5365  *
5366  *		On failure, the returned value is one of the following:
5367  *
5368  *		**-EINVAL** if *th_len* is invalid.
5369  *
5370  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5371  *
5372  * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5373  *	Description
5374  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5375  *		without depending on a listening socket.
5376  *
5377  *		*iph* points to the IPv4 header.
5378  *
5379  *		*th* points to the TCP header.
5380  *	Return
5381  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5382  *
5383  *		On failure, the returned value is one of the following:
5384  *
5385  *		**-EACCES** if the SYN cookie is not valid.
5386  *
5387  * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5388  *	Description
5389  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5390  *		without depending on a listening socket.
5391  *
5392  *		*iph* points to the IPv6 header.
5393  *
5394  *		*th* points to the TCP header.
5395  *	Return
5396  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5397  *
5398  *		On failure, the returned value is one of the following:
5399  *
5400  *		**-EACCES** if the SYN cookie is not valid.
5401  *
5402  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5403  *
5404  * u64 bpf_ktime_get_tai_ns(void)
5405  *	Description
5406  *		A nonsettable system-wide clock derived from wall-clock time but
5407  *		ignoring leap seconds.  This clock does not experience
5408  *		discontinuities and backwards jumps caused by NTP inserting leap
5409  *		seconds as CLOCK_REALTIME does.
5410  *
5411  *		See: **clock_gettime**\ (**CLOCK_TAI**)
5412  *	Return
5413  *		Current *ktime*.
5414  *
5415  * long bpf_user_ringbuf_drain(struct bpf_map *map, void *callback_fn, void *ctx, u64 flags)
5416  *	Description
5417  *		Drain samples from the specified user ring buffer, and invoke
5418  *		the provided callback for each such sample:
5419  *
5420  *		long (\*callback_fn)(const struct bpf_dynptr \*dynptr, void \*ctx);
5421  *
5422  *		If **callback_fn** returns 0, the helper will continue to try
5423  *		and drain the next sample, up to a maximum of
5424  *		BPF_MAX_USER_RINGBUF_SAMPLES samples. If the return value is 1,
5425  *		the helper will skip the rest of the samples and return. Other
5426  *		return values are not used now, and will be rejected by the
5427  *		verifier.
5428  *	Return
5429  *		The number of drained samples if no error was encountered while
5430  *		draining samples, or 0 if no samples were present in the ring
5431  *		buffer. If a user-space producer was epoll-waiting on this map,
5432  *		and at least one sample was drained, they will receive an event
5433  *		notification notifying them of available space in the ring
5434  *		buffer. If the BPF_RB_NO_WAKEUP flag is passed to this
5435  *		function, no wakeup notification will be sent. If the
5436  *		BPF_RB_FORCE_WAKEUP flag is passed, a wakeup notification will
5437  *		be sent even if no sample was drained.
5438  *
5439  *		On failure, the returned value is one of the following:
5440  *
5441  *		**-EBUSY** if the ring buffer is contended, and another calling
5442  *		context was concurrently draining the ring buffer.
5443  *
5444  *		**-EINVAL** if user-space is not properly tracking the ring
5445  *		buffer due to the producer position not being aligned to 8
5446  *		bytes, a sample not being aligned to 8 bytes, or the producer
5447  *		position not matching the advertised length of a sample.
5448  *
5449  *		**-E2BIG** if user-space has tried to publish a sample which is
5450  *		larger than the size of the ring buffer, or which cannot fit
5451  *		within a struct bpf_dynptr.
5452  *
5453  * void *bpf_cgrp_storage_get(struct bpf_map *map, struct cgroup *cgroup, void *value, u64 flags)
5454  *	Description
5455  *		Get a bpf_local_storage from the *cgroup*.
5456  *
5457  *		Logically, it could be thought of as getting the value from
5458  *		a *map* with *cgroup* as the **key**.  From this
5459  *		perspective,  the usage is not much different from
5460  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *cgroup*) except this
5461  *		helper enforces the key must be a cgroup struct and the map must also
5462  *		be a **BPF_MAP_TYPE_CGRP_STORAGE**.
5463  *
5464  *		In reality, the local-storage value is embedded directly inside of the
5465  *		*cgroup* object itself, rather than being located in the
5466  *		**BPF_MAP_TYPE_CGRP_STORAGE** map. When the local-storage value is
5467  *		queried for some *map* on a *cgroup* object, the kernel will perform an
5468  *		O(n) iteration over all of the live local-storage values for that
5469  *		*cgroup* object until the local-storage value for the *map* is found.
5470  *
5471  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
5472  *		used such that a new bpf_local_storage will be
5473  *		created if one does not exist.  *value* can be used
5474  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
5475  *		the initial value of a bpf_local_storage.  If *value* is
5476  *		**NULL**, the new bpf_local_storage will be zero initialized.
5477  *	Return
5478  *		A bpf_local_storage pointer is returned on success.
5479  *
5480  *		**NULL** if not found or there was an error in adding
5481  *		a new bpf_local_storage.
5482  *
5483  * long bpf_cgrp_storage_delete(struct bpf_map *map, struct cgroup *cgroup)
5484  *	Description
5485  *		Delete a bpf_local_storage from a *cgroup*.
5486  *	Return
5487  *		0 on success.
5488  *
5489  *		**-ENOENT** if the bpf_local_storage cannot be found.
5490  */
5491 #define ___BPF_FUNC_MAPPER(FN, ctx...)			\
5492 	FN(unspec, 0, ##ctx)				\
5493 	FN(map_lookup_elem, 1, ##ctx)			\
5494 	FN(map_update_elem, 2, ##ctx)			\
5495 	FN(map_delete_elem, 3, ##ctx)			\
5496 	FN(probe_read, 4, ##ctx)			\
5497 	FN(ktime_get_ns, 5, ##ctx)			\
5498 	FN(trace_printk, 6, ##ctx)			\
5499 	FN(get_prandom_u32, 7, ##ctx)			\
5500 	FN(get_smp_processor_id, 8, ##ctx)		\
5501 	FN(skb_store_bytes, 9, ##ctx)			\
5502 	FN(l3_csum_replace, 10, ##ctx)			\
5503 	FN(l4_csum_replace, 11, ##ctx)			\
5504 	FN(tail_call, 12, ##ctx)			\
5505 	FN(clone_redirect, 13, ##ctx)			\
5506 	FN(get_current_pid_tgid, 14, ##ctx)		\
5507 	FN(get_current_uid_gid, 15, ##ctx)		\
5508 	FN(get_current_comm, 16, ##ctx)			\
5509 	FN(get_cgroup_classid, 17, ##ctx)		\
5510 	FN(skb_vlan_push, 18, ##ctx)			\
5511 	FN(skb_vlan_pop, 19, ##ctx)			\
5512 	FN(skb_get_tunnel_key, 20, ##ctx)		\
5513 	FN(skb_set_tunnel_key, 21, ##ctx)		\
5514 	FN(perf_event_read, 22, ##ctx)			\
5515 	FN(redirect, 23, ##ctx)				\
5516 	FN(get_route_realm, 24, ##ctx)			\
5517 	FN(perf_event_output, 25, ##ctx)		\
5518 	FN(skb_load_bytes, 26, ##ctx)			\
5519 	FN(get_stackid, 27, ##ctx)			\
5520 	FN(csum_diff, 28, ##ctx)			\
5521 	FN(skb_get_tunnel_opt, 29, ##ctx)		\
5522 	FN(skb_set_tunnel_opt, 30, ##ctx)		\
5523 	FN(skb_change_proto, 31, ##ctx)			\
5524 	FN(skb_change_type, 32, ##ctx)			\
5525 	FN(skb_under_cgroup, 33, ##ctx)			\
5526 	FN(get_hash_recalc, 34, ##ctx)			\
5527 	FN(get_current_task, 35, ##ctx)			\
5528 	FN(probe_write_user, 36, ##ctx)			\
5529 	FN(current_task_under_cgroup, 37, ##ctx)	\
5530 	FN(skb_change_tail, 38, ##ctx)			\
5531 	FN(skb_pull_data, 39, ##ctx)			\
5532 	FN(csum_update, 40, ##ctx)			\
5533 	FN(set_hash_invalid, 41, ##ctx)			\
5534 	FN(get_numa_node_id, 42, ##ctx)			\
5535 	FN(skb_change_head, 43, ##ctx)			\
5536 	FN(xdp_adjust_head, 44, ##ctx)			\
5537 	FN(probe_read_str, 45, ##ctx)			\
5538 	FN(get_socket_cookie, 46, ##ctx)		\
5539 	FN(get_socket_uid, 47, ##ctx)			\
5540 	FN(set_hash, 48, ##ctx)				\
5541 	FN(setsockopt, 49, ##ctx)			\
5542 	FN(skb_adjust_room, 50, ##ctx)			\
5543 	FN(redirect_map, 51, ##ctx)			\
5544 	FN(sk_redirect_map, 52, ##ctx)			\
5545 	FN(sock_map_update, 53, ##ctx)			\
5546 	FN(xdp_adjust_meta, 54, ##ctx)			\
5547 	FN(perf_event_read_value, 55, ##ctx)		\
5548 	FN(perf_prog_read_value, 56, ##ctx)		\
5549 	FN(getsockopt, 57, ##ctx)			\
5550 	FN(override_return, 58, ##ctx)			\
5551 	FN(sock_ops_cb_flags_set, 59, ##ctx)		\
5552 	FN(msg_redirect_map, 60, ##ctx)			\
5553 	FN(msg_apply_bytes, 61, ##ctx)			\
5554 	FN(msg_cork_bytes, 62, ##ctx)			\
5555 	FN(msg_pull_data, 63, ##ctx)			\
5556 	FN(bind, 64, ##ctx)				\
5557 	FN(xdp_adjust_tail, 65, ##ctx)			\
5558 	FN(skb_get_xfrm_state, 66, ##ctx)		\
5559 	FN(get_stack, 67, ##ctx)			\
5560 	FN(skb_load_bytes_relative, 68, ##ctx)		\
5561 	FN(fib_lookup, 69, ##ctx)			\
5562 	FN(sock_hash_update, 70, ##ctx)			\
5563 	FN(msg_redirect_hash, 71, ##ctx)		\
5564 	FN(sk_redirect_hash, 72, ##ctx)			\
5565 	FN(lwt_push_encap, 73, ##ctx)			\
5566 	FN(lwt_seg6_store_bytes, 74, ##ctx)		\
5567 	FN(lwt_seg6_adjust_srh, 75, ##ctx)		\
5568 	FN(lwt_seg6_action, 76, ##ctx)			\
5569 	FN(rc_repeat, 77, ##ctx)			\
5570 	FN(rc_keydown, 78, ##ctx)			\
5571 	FN(skb_cgroup_id, 79, ##ctx)			\
5572 	FN(get_current_cgroup_id, 80, ##ctx)		\
5573 	FN(get_local_storage, 81, ##ctx)		\
5574 	FN(sk_select_reuseport, 82, ##ctx)		\
5575 	FN(skb_ancestor_cgroup_id, 83, ##ctx)		\
5576 	FN(sk_lookup_tcp, 84, ##ctx)			\
5577 	FN(sk_lookup_udp, 85, ##ctx)			\
5578 	FN(sk_release, 86, ##ctx)			\
5579 	FN(map_push_elem, 87, ##ctx)			\
5580 	FN(map_pop_elem, 88, ##ctx)			\
5581 	FN(map_peek_elem, 89, ##ctx)			\
5582 	FN(msg_push_data, 90, ##ctx)			\
5583 	FN(msg_pop_data, 91, ##ctx)			\
5584 	FN(rc_pointer_rel, 92, ##ctx)			\
5585 	FN(spin_lock, 93, ##ctx)			\
5586 	FN(spin_unlock, 94, ##ctx)			\
5587 	FN(sk_fullsock, 95, ##ctx)			\
5588 	FN(tcp_sock, 96, ##ctx)				\
5589 	FN(skb_ecn_set_ce, 97, ##ctx)			\
5590 	FN(get_listener_sock, 98, ##ctx)		\
5591 	FN(skc_lookup_tcp, 99, ##ctx)			\
5592 	FN(tcp_check_syncookie, 100, ##ctx)		\
5593 	FN(sysctl_get_name, 101, ##ctx)			\
5594 	FN(sysctl_get_current_value, 102, ##ctx)	\
5595 	FN(sysctl_get_new_value, 103, ##ctx)		\
5596 	FN(sysctl_set_new_value, 104, ##ctx)		\
5597 	FN(strtol, 105, ##ctx)				\
5598 	FN(strtoul, 106, ##ctx)				\
5599 	FN(sk_storage_get, 107, ##ctx)			\
5600 	FN(sk_storage_delete, 108, ##ctx)		\
5601 	FN(send_signal, 109, ##ctx)			\
5602 	FN(tcp_gen_syncookie, 110, ##ctx)		\
5603 	FN(skb_output, 111, ##ctx)			\
5604 	FN(probe_read_user, 112, ##ctx)			\
5605 	FN(probe_read_kernel, 113, ##ctx)		\
5606 	FN(probe_read_user_str, 114, ##ctx)		\
5607 	FN(probe_read_kernel_str, 115, ##ctx)		\
5608 	FN(tcp_send_ack, 116, ##ctx)			\
5609 	FN(send_signal_thread, 117, ##ctx)		\
5610 	FN(jiffies64, 118, ##ctx)			\
5611 	FN(read_branch_records, 119, ##ctx)		\
5612 	FN(get_ns_current_pid_tgid, 120, ##ctx)		\
5613 	FN(xdp_output, 121, ##ctx)			\
5614 	FN(get_netns_cookie, 122, ##ctx)		\
5615 	FN(get_current_ancestor_cgroup_id, 123, ##ctx)	\
5616 	FN(sk_assign, 124, ##ctx)			\
5617 	FN(ktime_get_boot_ns, 125, ##ctx)		\
5618 	FN(seq_printf, 126, ##ctx)			\
5619 	FN(seq_write, 127, ##ctx)			\
5620 	FN(sk_cgroup_id, 128, ##ctx)			\
5621 	FN(sk_ancestor_cgroup_id, 129, ##ctx)		\
5622 	FN(ringbuf_output, 130, ##ctx)			\
5623 	FN(ringbuf_reserve, 131, ##ctx)			\
5624 	FN(ringbuf_submit, 132, ##ctx)			\
5625 	FN(ringbuf_discard, 133, ##ctx)			\
5626 	FN(ringbuf_query, 134, ##ctx)			\
5627 	FN(csum_level, 135, ##ctx)			\
5628 	FN(skc_to_tcp6_sock, 136, ##ctx)		\
5629 	FN(skc_to_tcp_sock, 137, ##ctx)			\
5630 	FN(skc_to_tcp_timewait_sock, 138, ##ctx)	\
5631 	FN(skc_to_tcp_request_sock, 139, ##ctx)		\
5632 	FN(skc_to_udp6_sock, 140, ##ctx)		\
5633 	FN(get_task_stack, 141, ##ctx)			\
5634 	FN(load_hdr_opt, 142, ##ctx)			\
5635 	FN(store_hdr_opt, 143, ##ctx)			\
5636 	FN(reserve_hdr_opt, 144, ##ctx)			\
5637 	FN(inode_storage_get, 145, ##ctx)		\
5638 	FN(inode_storage_delete, 146, ##ctx)		\
5639 	FN(d_path, 147, ##ctx)				\
5640 	FN(copy_from_user, 148, ##ctx)			\
5641 	FN(snprintf_btf, 149, ##ctx)			\
5642 	FN(seq_printf_btf, 150, ##ctx)			\
5643 	FN(skb_cgroup_classid, 151, ##ctx)		\
5644 	FN(redirect_neigh, 152, ##ctx)			\
5645 	FN(per_cpu_ptr, 153, ##ctx)			\
5646 	FN(this_cpu_ptr, 154, ##ctx)			\
5647 	FN(redirect_peer, 155, ##ctx)			\
5648 	FN(task_storage_get, 156, ##ctx)		\
5649 	FN(task_storage_delete, 157, ##ctx)		\
5650 	FN(get_current_task_btf, 158, ##ctx)		\
5651 	FN(bprm_opts_set, 159, ##ctx)			\
5652 	FN(ktime_get_coarse_ns, 160, ##ctx)		\
5653 	FN(ima_inode_hash, 161, ##ctx)			\
5654 	FN(sock_from_file, 162, ##ctx)			\
5655 	FN(check_mtu, 163, ##ctx)			\
5656 	FN(for_each_map_elem, 164, ##ctx)		\
5657 	FN(snprintf, 165, ##ctx)			\
5658 	FN(sys_bpf, 166, ##ctx)				\
5659 	FN(btf_find_by_name_kind, 167, ##ctx)		\
5660 	FN(sys_close, 168, ##ctx)			\
5661 	FN(timer_init, 169, ##ctx)			\
5662 	FN(timer_set_callback, 170, ##ctx)		\
5663 	FN(timer_start, 171, ##ctx)			\
5664 	FN(timer_cancel, 172, ##ctx)			\
5665 	FN(get_func_ip, 173, ##ctx)			\
5666 	FN(get_attach_cookie, 174, ##ctx)		\
5667 	FN(task_pt_regs, 175, ##ctx)			\
5668 	FN(get_branch_snapshot, 176, ##ctx)		\
5669 	FN(trace_vprintk, 177, ##ctx)			\
5670 	FN(skc_to_unix_sock, 178, ##ctx)		\
5671 	FN(kallsyms_lookup_name, 179, ##ctx)		\
5672 	FN(find_vma, 180, ##ctx)			\
5673 	FN(loop, 181, ##ctx)				\
5674 	FN(strncmp, 182, ##ctx)				\
5675 	FN(get_func_arg, 183, ##ctx)			\
5676 	FN(get_func_ret, 184, ##ctx)			\
5677 	FN(get_func_arg_cnt, 185, ##ctx)		\
5678 	FN(get_retval, 186, ##ctx)			\
5679 	FN(set_retval, 187, ##ctx)			\
5680 	FN(xdp_get_buff_len, 188, ##ctx)		\
5681 	FN(xdp_load_bytes, 189, ##ctx)			\
5682 	FN(xdp_store_bytes, 190, ##ctx)			\
5683 	FN(copy_from_user_task, 191, ##ctx)		\
5684 	FN(skb_set_tstamp, 192, ##ctx)			\
5685 	FN(ima_file_hash, 193, ##ctx)			\
5686 	FN(kptr_xchg, 194, ##ctx)			\
5687 	FN(map_lookup_percpu_elem, 195, ##ctx)		\
5688 	FN(skc_to_mptcp_sock, 196, ##ctx)		\
5689 	FN(dynptr_from_mem, 197, ##ctx)			\
5690 	FN(ringbuf_reserve_dynptr, 198, ##ctx)		\
5691 	FN(ringbuf_submit_dynptr, 199, ##ctx)		\
5692 	FN(ringbuf_discard_dynptr, 200, ##ctx)		\
5693 	FN(dynptr_read, 201, ##ctx)			\
5694 	FN(dynptr_write, 202, ##ctx)			\
5695 	FN(dynptr_data, 203, ##ctx)			\
5696 	FN(tcp_raw_gen_syncookie_ipv4, 204, ##ctx)	\
5697 	FN(tcp_raw_gen_syncookie_ipv6, 205, ##ctx)	\
5698 	FN(tcp_raw_check_syncookie_ipv4, 206, ##ctx)	\
5699 	FN(tcp_raw_check_syncookie_ipv6, 207, ##ctx)	\
5700 	FN(ktime_get_tai_ns, 208, ##ctx)		\
5701 	FN(user_ringbuf_drain, 209, ##ctx)		\
5702 	FN(cgrp_storage_get, 210, ##ctx)		\
5703 	FN(cgrp_storage_delete, 211, ##ctx)		\
5704 	/* */
5705 
5706 /* backwards-compatibility macros for users of __BPF_FUNC_MAPPER that don't
5707  * know or care about integer value that is now passed as second argument
5708  */
5709 #define __BPF_FUNC_MAPPER_APPLY(name, value, FN) FN(name),
5710 #define __BPF_FUNC_MAPPER(FN) ___BPF_FUNC_MAPPER(__BPF_FUNC_MAPPER_APPLY, FN)
5711 
5712 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5713  * function eBPF program intends to call
5714  */
5715 #define __BPF_ENUM_FN(x, y) BPF_FUNC_ ## x = y,
5716 enum bpf_func_id {
5717 	___BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5718 	__BPF_FUNC_MAX_ID,
5719 };
5720 #undef __BPF_ENUM_FN
5721 
5722 /* All flags used by eBPF helper functions, placed here. */
5723 
5724 /* BPF_FUNC_skb_store_bytes flags. */
5725 enum {
5726 	BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),
5727 	BPF_F_INVALIDATE_HASH		= (1ULL << 1),
5728 };
5729 
5730 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5731  * First 4 bits are for passing the header field size.
5732  */
5733 enum {
5734 	BPF_F_HDR_FIELD_MASK		= 0xfULL,
5735 };
5736 
5737 /* BPF_FUNC_l4_csum_replace flags. */
5738 enum {
5739 	BPF_F_PSEUDO_HDR		= (1ULL << 4),
5740 	BPF_F_MARK_MANGLED_0		= (1ULL << 5),
5741 	BPF_F_MARK_ENFORCE		= (1ULL << 6),
5742 };
5743 
5744 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5745 enum {
5746 	BPF_F_INGRESS			= (1ULL << 0),
5747 };
5748 
5749 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5750 enum {
5751 	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
5752 };
5753 
5754 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5755 enum {
5756 	BPF_F_SKIP_FIELD_MASK		= 0xffULL,
5757 	BPF_F_USER_STACK		= (1ULL << 8),
5758 /* flags used by BPF_FUNC_get_stackid only. */
5759 	BPF_F_FAST_STACK_CMP		= (1ULL << 9),
5760 	BPF_F_REUSE_STACKID		= (1ULL << 10),
5761 /* flags used by BPF_FUNC_get_stack only. */
5762 	BPF_F_USER_BUILD_ID		= (1ULL << 11),
5763 };
5764 
5765 /* BPF_FUNC_skb_set_tunnel_key flags. */
5766 enum {
5767 	BPF_F_ZERO_CSUM_TX		= (1ULL << 1),
5768 	BPF_F_DONT_FRAGMENT		= (1ULL << 2),
5769 	BPF_F_SEQ_NUMBER		= (1ULL << 3),
5770 	BPF_F_NO_TUNNEL_KEY		= (1ULL << 4),
5771 };
5772 
5773 /* BPF_FUNC_skb_get_tunnel_key flags. */
5774 enum {
5775 	BPF_F_TUNINFO_FLAGS		= (1ULL << 4),
5776 };
5777 
5778 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5779  * BPF_FUNC_perf_event_read_value flags.
5780  */
5781 enum {
5782 	BPF_F_INDEX_MASK		= 0xffffffffULL,
5783 	BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,
5784 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5785 	BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),
5786 };
5787 
5788 /* Current network namespace */
5789 enum {
5790 	BPF_F_CURRENT_NETNS		= (-1L),
5791 };
5792 
5793 /* BPF_FUNC_csum_level level values. */
5794 enum {
5795 	BPF_CSUM_LEVEL_QUERY,
5796 	BPF_CSUM_LEVEL_INC,
5797 	BPF_CSUM_LEVEL_DEC,
5798 	BPF_CSUM_LEVEL_RESET,
5799 };
5800 
5801 /* BPF_FUNC_skb_adjust_room flags. */
5802 enum {
5803 	BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),
5804 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),
5805 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),
5806 	BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),
5807 	BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),
5808 	BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),
5809 	BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),
5810 };
5811 
5812 enum {
5813 	BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,
5814 	BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,
5815 };
5816 
5817 #define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \
5818 					  BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5819 					 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5820 
5821 /* BPF_FUNC_sysctl_get_name flags. */
5822 enum {
5823 	BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),
5824 };
5825 
5826 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5827 enum {
5828 	BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),
5829 	/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5830 	 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5831 	 */
5832 	BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
5833 };
5834 
5835 /* BPF_FUNC_read_branch_records flags. */
5836 enum {
5837 	BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),
5838 };
5839 
5840 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5841  * BPF_FUNC_bpf_ringbuf_output flags.
5842  */
5843 enum {
5844 	BPF_RB_NO_WAKEUP		= (1ULL << 0),
5845 	BPF_RB_FORCE_WAKEUP		= (1ULL << 1),
5846 };
5847 
5848 /* BPF_FUNC_bpf_ringbuf_query flags */
5849 enum {
5850 	BPF_RB_AVAIL_DATA = 0,
5851 	BPF_RB_RING_SIZE = 1,
5852 	BPF_RB_CONS_POS = 2,
5853 	BPF_RB_PROD_POS = 3,
5854 };
5855 
5856 /* BPF ring buffer constants */
5857 enum {
5858 	BPF_RINGBUF_BUSY_BIT		= (1U << 31),
5859 	BPF_RINGBUF_DISCARD_BIT		= (1U << 30),
5860 	BPF_RINGBUF_HDR_SZ		= 8,
5861 };
5862 
5863 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5864 enum {
5865 	BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),
5866 	BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),
5867 };
5868 
5869 /* Mode for BPF_FUNC_skb_adjust_room helper. */
5870 enum bpf_adj_room_mode {
5871 	BPF_ADJ_ROOM_NET,
5872 	BPF_ADJ_ROOM_MAC,
5873 };
5874 
5875 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5876 enum bpf_hdr_start_off {
5877 	BPF_HDR_START_MAC,
5878 	BPF_HDR_START_NET,
5879 };
5880 
5881 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5882 enum bpf_lwt_encap_mode {
5883 	BPF_LWT_ENCAP_SEG6,
5884 	BPF_LWT_ENCAP_SEG6_INLINE,
5885 	BPF_LWT_ENCAP_IP,
5886 };
5887 
5888 /* Flags for bpf_bprm_opts_set helper */
5889 enum {
5890 	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
5891 };
5892 
5893 /* Flags for bpf_redirect_map helper */
5894 enum {
5895 	BPF_F_BROADCAST		= (1ULL << 3),
5896 	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
5897 };
5898 
5899 #define __bpf_md_ptr(type, name)	\
5900 union {					\
5901 	type name;			\
5902 	__u64 :64;			\
5903 } __attribute__((aligned(8)))
5904 
5905 enum {
5906 	BPF_SKB_TSTAMP_UNSPEC,
5907 	BPF_SKB_TSTAMP_DELIVERY_MONO,	/* tstamp has mono delivery time */
5908 	/* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
5909 	 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
5910 	 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
5911 	 */
5912 };
5913 
5914 /* user accessible mirror of in-kernel sk_buff.
5915  * new fields can only be added to the end of this structure
5916  */
5917 struct __sk_buff {
5918 	__u32 len;
5919 	__u32 pkt_type;
5920 	__u32 mark;
5921 	__u32 queue_mapping;
5922 	__u32 protocol;
5923 	__u32 vlan_present;
5924 	__u32 vlan_tci;
5925 	__u32 vlan_proto;
5926 	__u32 priority;
5927 	__u32 ingress_ifindex;
5928 	__u32 ifindex;
5929 	__u32 tc_index;
5930 	__u32 cb[5];
5931 	__u32 hash;
5932 	__u32 tc_classid;
5933 	__u32 data;
5934 	__u32 data_end;
5935 	__u32 napi_id;
5936 
5937 	/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5938 	__u32 family;
5939 	__u32 remote_ip4;	/* Stored in network byte order */
5940 	__u32 local_ip4;	/* Stored in network byte order */
5941 	__u32 remote_ip6[4];	/* Stored in network byte order */
5942 	__u32 local_ip6[4];	/* Stored in network byte order */
5943 	__u32 remote_port;	/* Stored in network byte order */
5944 	__u32 local_port;	/* stored in host byte order */
5945 	/* ... here. */
5946 
5947 	__u32 data_meta;
5948 	__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5949 	__u64 tstamp;
5950 	__u32 wire_len;
5951 	__u32 gso_segs;
5952 	__bpf_md_ptr(struct bpf_sock *, sk);
5953 	__u32 gso_size;
5954 	__u8  tstamp_type;
5955 	__u32 :24;		/* Padding, future use. */
5956 	__u64 hwtstamp;
5957 };
5958 
5959 struct bpf_tunnel_key {
5960 	__u32 tunnel_id;
5961 	union {
5962 		__u32 remote_ipv4;
5963 		__u32 remote_ipv6[4];
5964 	};
5965 	__u8 tunnel_tos;
5966 	__u8 tunnel_ttl;
5967 	union {
5968 		__u16 tunnel_ext;	/* compat */
5969 		__be16 tunnel_flags;
5970 	};
5971 	__u32 tunnel_label;
5972 	union {
5973 		__u32 local_ipv4;
5974 		__u32 local_ipv6[4];
5975 	};
5976 };
5977 
5978 /* user accessible mirror of in-kernel xfrm_state.
5979  * new fields can only be added to the end of this structure
5980  */
5981 struct bpf_xfrm_state {
5982 	__u32 reqid;
5983 	__u32 spi;	/* Stored in network byte order */
5984 	__u16 family;
5985 	__u16 ext;	/* Padding, future use. */
5986 	union {
5987 		__u32 remote_ipv4;	/* Stored in network byte order */
5988 		__u32 remote_ipv6[4];	/* Stored in network byte order */
5989 	};
5990 };
5991 
5992 /* Generic BPF return codes which all BPF program types may support.
5993  * The values are binary compatible with their TC_ACT_* counter-part to
5994  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5995  * programs.
5996  *
5997  * XDP is handled seprately, see XDP_*.
5998  */
5999 enum bpf_ret_code {
6000 	BPF_OK = 0,
6001 	/* 1 reserved */
6002 	BPF_DROP = 2,
6003 	/* 3-6 reserved */
6004 	BPF_REDIRECT = 7,
6005 	/* >127 are reserved for prog type specific return codes.
6006 	 *
6007 	 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
6008 	 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
6009 	 *    changed and should be routed based on its new L3 header.
6010 	 *    (This is an L3 redirect, as opposed to L2 redirect
6011 	 *    represented by BPF_REDIRECT above).
6012 	 */
6013 	BPF_LWT_REROUTE = 128,
6014 	/* BPF_FLOW_DISSECTOR_CONTINUE: used by BPF_PROG_TYPE_FLOW_DISSECTOR
6015 	 *   to indicate that no custom dissection was performed, and
6016 	 *   fallback to standard dissector is requested.
6017 	 */
6018 	BPF_FLOW_DISSECTOR_CONTINUE = 129,
6019 };
6020 
6021 struct bpf_sock {
6022 	__u32 bound_dev_if;
6023 	__u32 family;
6024 	__u32 type;
6025 	__u32 protocol;
6026 	__u32 mark;
6027 	__u32 priority;
6028 	/* IP address also allows 1 and 2 bytes access */
6029 	__u32 src_ip4;
6030 	__u32 src_ip6[4];
6031 	__u32 src_port;		/* host byte order */
6032 	__be16 dst_port;	/* network byte order */
6033 	__u16 :16;		/* zero padding */
6034 	__u32 dst_ip4;
6035 	__u32 dst_ip6[4];
6036 	__u32 state;
6037 	__s32 rx_queue_mapping;
6038 };
6039 
6040 struct bpf_tcp_sock {
6041 	__u32 snd_cwnd;		/* Sending congestion window		*/
6042 	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
6043 	__u32 rtt_min;
6044 	__u32 snd_ssthresh;	/* Slow start size threshold		*/
6045 	__u32 rcv_nxt;		/* What we want to receive next		*/
6046 	__u32 snd_nxt;		/* Next sequence we send		*/
6047 	__u32 snd_una;		/* First byte we want an ack for	*/
6048 	__u32 mss_cache;	/* Cached effective mss, not including SACKS */
6049 	__u32 ecn_flags;	/* ECN status bits.			*/
6050 	__u32 rate_delivered;	/* saved rate sample: packets delivered */
6051 	__u32 rate_interval_us;	/* saved rate sample: time elapsed */
6052 	__u32 packets_out;	/* Packets which are "in flight"	*/
6053 	__u32 retrans_out;	/* Retransmitted packets out		*/
6054 	__u32 total_retrans;	/* Total retransmits for entire connection */
6055 	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
6056 				 * total number of segments in.
6057 				 */
6058 	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
6059 				 * total number of data segments in.
6060 				 */
6061 	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
6062 				 * The total number of segments sent.
6063 				 */
6064 	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
6065 				 * total number of data segments sent.
6066 				 */
6067 	__u32 lost_out;		/* Lost packets			*/
6068 	__u32 sacked_out;	/* SACK'd packets			*/
6069 	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
6070 				 * sum(delta(rcv_nxt)), or how many bytes
6071 				 * were acked.
6072 				 */
6073 	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
6074 				 * sum(delta(snd_una)), or how many bytes
6075 				 * were acked.
6076 				 */
6077 	__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups
6078 				 * total number of DSACK blocks received
6079 				 */
6080 	__u32 delivered;	/* Total data packets delivered incl. rexmits */
6081 	__u32 delivered_ce;	/* Like the above but only ECE marked packets */
6082 	__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */
6083 };
6084 
6085 struct bpf_sock_tuple {
6086 	union {
6087 		struct {
6088 			__be32 saddr;
6089 			__be32 daddr;
6090 			__be16 sport;
6091 			__be16 dport;
6092 		} ipv4;
6093 		struct {
6094 			__be32 saddr[4];
6095 			__be32 daddr[4];
6096 			__be16 sport;
6097 			__be16 dport;
6098 		} ipv6;
6099 	};
6100 };
6101 
6102 struct bpf_xdp_sock {
6103 	__u32 queue_id;
6104 };
6105 
6106 #define XDP_PACKET_HEADROOM 256
6107 
6108 /* User return codes for XDP prog type.
6109  * A valid XDP program must return one of these defined values. All other
6110  * return codes are reserved for future use. Unknown return codes will
6111  * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
6112  */
6113 enum xdp_action {
6114 	XDP_ABORTED = 0,
6115 	XDP_DROP,
6116 	XDP_PASS,
6117 	XDP_TX,
6118 	XDP_REDIRECT,
6119 };
6120 
6121 /* user accessible metadata for XDP packet hook
6122  * new fields must be added to the end of this structure
6123  */
6124 struct xdp_md {
6125 	__u32 data;
6126 	__u32 data_end;
6127 	__u32 data_meta;
6128 	/* Below access go through struct xdp_rxq_info */
6129 	__u32 ingress_ifindex; /* rxq->dev->ifindex */
6130 	__u32 rx_queue_index;  /* rxq->queue_index  */
6131 
6132 	__u32 egress_ifindex;  /* txq->dev->ifindex */
6133 };
6134 
6135 /* DEVMAP map-value layout
6136  *
6137  * The struct data-layout of map-value is a configuration interface.
6138  * New members can only be added to the end of this structure.
6139  */
6140 struct bpf_devmap_val {
6141 	__u32 ifindex;   /* device index */
6142 	union {
6143 		int   fd;  /* prog fd on map write */
6144 		__u32 id;  /* prog id on map read */
6145 	} bpf_prog;
6146 };
6147 
6148 /* CPUMAP map-value layout
6149  *
6150  * The struct data-layout of map-value is a configuration interface.
6151  * New members can only be added to the end of this structure.
6152  */
6153 struct bpf_cpumap_val {
6154 	__u32 qsize;	/* queue size to remote target CPU */
6155 	union {
6156 		int   fd;	/* prog fd on map write */
6157 		__u32 id;	/* prog id on map read */
6158 	} bpf_prog;
6159 };
6160 
6161 enum sk_action {
6162 	SK_DROP = 0,
6163 	SK_PASS,
6164 };
6165 
6166 /* user accessible metadata for SK_MSG packet hook, new fields must
6167  * be added to the end of this structure
6168  */
6169 struct sk_msg_md {
6170 	__bpf_md_ptr(void *, data);
6171 	__bpf_md_ptr(void *, data_end);
6172 
6173 	__u32 family;
6174 	__u32 remote_ip4;	/* Stored in network byte order */
6175 	__u32 local_ip4;	/* Stored in network byte order */
6176 	__u32 remote_ip6[4];	/* Stored in network byte order */
6177 	__u32 local_ip6[4];	/* Stored in network byte order */
6178 	__u32 remote_port;	/* Stored in network byte order */
6179 	__u32 local_port;	/* stored in host byte order */
6180 	__u32 size;		/* Total size of sk_msg */
6181 
6182 	__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6183 };
6184 
6185 struct sk_reuseport_md {
6186 	/*
6187 	 * Start of directly accessible data. It begins from
6188 	 * the tcp/udp header.
6189 	 */
6190 	__bpf_md_ptr(void *, data);
6191 	/* End of directly accessible data */
6192 	__bpf_md_ptr(void *, data_end);
6193 	/*
6194 	 * Total length of packet (starting from the tcp/udp header).
6195 	 * Note that the directly accessible bytes (data_end - data)
6196 	 * could be less than this "len".  Those bytes could be
6197 	 * indirectly read by a helper "bpf_skb_load_bytes()".
6198 	 */
6199 	__u32 len;
6200 	/*
6201 	 * Eth protocol in the mac header (network byte order). e.g.
6202 	 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6203 	 */
6204 	__u32 eth_protocol;
6205 	__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6206 	__u32 bind_inany;	/* Is sock bound to an INANY address? */
6207 	__u32 hash;		/* A hash of the packet 4 tuples */
6208 	/* When reuse->migrating_sk is NULL, it is selecting a sk for the
6209 	 * new incoming connection request (e.g. selecting a listen sk for
6210 	 * the received SYN in the TCP case).  reuse->sk is one of the sk
6211 	 * in the reuseport group. The bpf prog can use reuse->sk to learn
6212 	 * the local listening ip/port without looking into the skb.
6213 	 *
6214 	 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6215 	 * reuse->migrating_sk is the socket that needs to be migrated
6216 	 * to another listening socket.  migrating_sk could be a fullsock
6217 	 * sk that is fully established or a reqsk that is in-the-middle
6218 	 * of 3-way handshake.
6219 	 */
6220 	__bpf_md_ptr(struct bpf_sock *, sk);
6221 	__bpf_md_ptr(struct bpf_sock *, migrating_sk);
6222 };
6223 
6224 #define BPF_TAG_SIZE	8
6225 
6226 struct bpf_prog_info {
6227 	__u32 type;
6228 	__u32 id;
6229 	__u8  tag[BPF_TAG_SIZE];
6230 	__u32 jited_prog_len;
6231 	__u32 xlated_prog_len;
6232 	__aligned_u64 jited_prog_insns;
6233 	__aligned_u64 xlated_prog_insns;
6234 	__u64 load_time;	/* ns since boottime */
6235 	__u32 created_by_uid;
6236 	__u32 nr_map_ids;
6237 	__aligned_u64 map_ids;
6238 	char name[BPF_OBJ_NAME_LEN];
6239 	__u32 ifindex;
6240 	__u32 gpl_compatible:1;
6241 	__u32 :31; /* alignment pad */
6242 	__u64 netns_dev;
6243 	__u64 netns_ino;
6244 	__u32 nr_jited_ksyms;
6245 	__u32 nr_jited_func_lens;
6246 	__aligned_u64 jited_ksyms;
6247 	__aligned_u64 jited_func_lens;
6248 	__u32 btf_id;
6249 	__u32 func_info_rec_size;
6250 	__aligned_u64 func_info;
6251 	__u32 nr_func_info;
6252 	__u32 nr_line_info;
6253 	__aligned_u64 line_info;
6254 	__aligned_u64 jited_line_info;
6255 	__u32 nr_jited_line_info;
6256 	__u32 line_info_rec_size;
6257 	__u32 jited_line_info_rec_size;
6258 	__u32 nr_prog_tags;
6259 	__aligned_u64 prog_tags;
6260 	__u64 run_time_ns;
6261 	__u64 run_cnt;
6262 	__u64 recursion_misses;
6263 	__u32 verified_insns;
6264 	__u32 attach_btf_obj_id;
6265 	__u32 attach_btf_id;
6266 } __attribute__((aligned(8)));
6267 
6268 struct bpf_map_info {
6269 	__u32 type;
6270 	__u32 id;
6271 	__u32 key_size;
6272 	__u32 value_size;
6273 	__u32 max_entries;
6274 	__u32 map_flags;
6275 	char  name[BPF_OBJ_NAME_LEN];
6276 	__u32 ifindex;
6277 	__u32 btf_vmlinux_value_type_id;
6278 	__u64 netns_dev;
6279 	__u64 netns_ino;
6280 	__u32 btf_id;
6281 	__u32 btf_key_type_id;
6282 	__u32 btf_value_type_id;
6283 	__u32 :32;	/* alignment pad */
6284 	__u64 map_extra;
6285 } __attribute__((aligned(8)));
6286 
6287 struct bpf_btf_info {
6288 	__aligned_u64 btf;
6289 	__u32 btf_size;
6290 	__u32 id;
6291 	__aligned_u64 name;
6292 	__u32 name_len;
6293 	__u32 kernel_btf;
6294 } __attribute__((aligned(8)));
6295 
6296 struct bpf_link_info {
6297 	__u32 type;
6298 	__u32 id;
6299 	__u32 prog_id;
6300 	union {
6301 		struct {
6302 			__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6303 			__u32 tp_name_len;     /* in/out: tp_name buffer len */
6304 		} raw_tracepoint;
6305 		struct {
6306 			__u32 attach_type;
6307 			__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6308 			__u32 target_btf_id; /* BTF type id inside the object */
6309 		} tracing;
6310 		struct {
6311 			__u64 cgroup_id;
6312 			__u32 attach_type;
6313 		} cgroup;
6314 		struct {
6315 			__aligned_u64 target_name; /* in/out: target_name buffer ptr */
6316 			__u32 target_name_len;	   /* in/out: target_name buffer len */
6317 
6318 			/* If the iter specific field is 32 bits, it can be put
6319 			 * in the first or second union. Otherwise it should be
6320 			 * put in the second union.
6321 			 */
6322 			union {
6323 				struct {
6324 					__u32 map_id;
6325 				} map;
6326 			};
6327 			union {
6328 				struct {
6329 					__u64 cgroup_id;
6330 					__u32 order;
6331 				} cgroup;
6332 				struct {
6333 					__u32 tid;
6334 					__u32 pid;
6335 				} task;
6336 			};
6337 		} iter;
6338 		struct  {
6339 			__u32 netns_ino;
6340 			__u32 attach_type;
6341 		} netns;
6342 		struct {
6343 			__u32 ifindex;
6344 		} xdp;
6345 	};
6346 } __attribute__((aligned(8)));
6347 
6348 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6349  * by user and intended to be used by socket (e.g. to bind to, depends on
6350  * attach type).
6351  */
6352 struct bpf_sock_addr {
6353 	__u32 user_family;	/* Allows 4-byte read, but no write. */
6354 	__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.
6355 				 * Stored in network byte order.
6356 				 */
6357 	__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6358 				 * Stored in network byte order.
6359 				 */
6360 	__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.
6361 				 * Stored in network byte order
6362 				 */
6363 	__u32 family;		/* Allows 4-byte read, but no write */
6364 	__u32 type;		/* Allows 4-byte read, but no write */
6365 	__u32 protocol;		/* Allows 4-byte read, but no write */
6366 	__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.
6367 				 * Stored in network byte order.
6368 				 */
6369 	__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6370 				 * Stored in network byte order.
6371 				 */
6372 	__bpf_md_ptr(struct bpf_sock *, sk);
6373 };
6374 
6375 /* User bpf_sock_ops struct to access socket values and specify request ops
6376  * and their replies.
6377  * Some of this fields are in network (bigendian) byte order and may need
6378  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6379  * New fields can only be added at the end of this structure
6380  */
6381 struct bpf_sock_ops {
6382 	__u32 op;
6383 	union {
6384 		__u32 args[4];		/* Optionally passed to bpf program */
6385 		__u32 reply;		/* Returned by bpf program	    */
6386 		__u32 replylong[4];	/* Optionally returned by bpf prog  */
6387 	};
6388 	__u32 family;
6389 	__u32 remote_ip4;	/* Stored in network byte order */
6390 	__u32 local_ip4;	/* Stored in network byte order */
6391 	__u32 remote_ip6[4];	/* Stored in network byte order */
6392 	__u32 local_ip6[4];	/* Stored in network byte order */
6393 	__u32 remote_port;	/* Stored in network byte order */
6394 	__u32 local_port;	/* stored in host byte order */
6395 	__u32 is_fullsock;	/* Some TCP fields are only valid if
6396 				 * there is a full socket. If not, the
6397 				 * fields read as zero.
6398 				 */
6399 	__u32 snd_cwnd;
6400 	__u32 srtt_us;		/* Averaged RTT << 3 in usecs */
6401 	__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6402 	__u32 state;
6403 	__u32 rtt_min;
6404 	__u32 snd_ssthresh;
6405 	__u32 rcv_nxt;
6406 	__u32 snd_nxt;
6407 	__u32 snd_una;
6408 	__u32 mss_cache;
6409 	__u32 ecn_flags;
6410 	__u32 rate_delivered;
6411 	__u32 rate_interval_us;
6412 	__u32 packets_out;
6413 	__u32 retrans_out;
6414 	__u32 total_retrans;
6415 	__u32 segs_in;
6416 	__u32 data_segs_in;
6417 	__u32 segs_out;
6418 	__u32 data_segs_out;
6419 	__u32 lost_out;
6420 	__u32 sacked_out;
6421 	__u32 sk_txhash;
6422 	__u64 bytes_received;
6423 	__u64 bytes_acked;
6424 	__bpf_md_ptr(struct bpf_sock *, sk);
6425 	/* [skb_data, skb_data_end) covers the whole TCP header.
6426 	 *
6427 	 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6428 	 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
6429 	 *                                header has not been written.
6430 	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6431 	 *				  been written so far.
6432 	 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
6433 	 *					the 3WHS.
6434 	 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6435 	 *					the 3WHS.
6436 	 *
6437 	 * bpf_load_hdr_opt() can also be used to read a particular option.
6438 	 */
6439 	__bpf_md_ptr(void *, skb_data);
6440 	__bpf_md_ptr(void *, skb_data_end);
6441 	__u32 skb_len;		/* The total length of a packet.
6442 				 * It includes the header, options,
6443 				 * and payload.
6444 				 */
6445 	__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides
6446 				 * an easy way to check for tcp_flags
6447 				 * without parsing skb_data.
6448 				 *
6449 				 * In particular, the skb_tcp_flags
6450 				 * will still be available in
6451 				 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6452 				 * the outgoing header has not
6453 				 * been written yet.
6454 				 */
6455 	__u64 skb_hwtstamp;
6456 };
6457 
6458 /* Definitions for bpf_sock_ops_cb_flags */
6459 enum {
6460 	BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),
6461 	BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),
6462 	BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),
6463 	BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),
6464 	/* Call bpf for all received TCP headers.  The bpf prog will be
6465 	 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6466 	 *
6467 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6468 	 * for the header option related helpers that will be useful
6469 	 * to the bpf programs.
6470 	 *
6471 	 * It could be used at the client/active side (i.e. connect() side)
6472 	 * when the server told it that the server was in syncookie
6473 	 * mode and required the active side to resend the bpf-written
6474 	 * options.  The active side can keep writing the bpf-options until
6475 	 * it received a valid packet from the server side to confirm
6476 	 * the earlier packet (and options) has been received.  The later
6477 	 * example patch is using it like this at the active side when the
6478 	 * server is in syncookie mode.
6479 	 *
6480 	 * The bpf prog will usually turn this off in the common cases.
6481 	 */
6482 	BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),
6483 	/* Call bpf when kernel has received a header option that
6484 	 * the kernel cannot handle.  The bpf prog will be called under
6485 	 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6486 	 *
6487 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6488 	 * for the header option related helpers that will be useful
6489 	 * to the bpf programs.
6490 	 */
6491 	BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6492 	/* Call bpf when the kernel is writing header options for the
6493 	 * outgoing packet.  The bpf prog will first be called
6494 	 * to reserve space in a skb under
6495 	 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
6496 	 * the bpf prog will be called to write the header option(s)
6497 	 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6498 	 *
6499 	 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6500 	 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6501 	 * related helpers that will be useful to the bpf programs.
6502 	 *
6503 	 * The kernel gets its chance to reserve space and write
6504 	 * options first before the BPF program does.
6505 	 */
6506 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6507 /* Mask of all currently supported cb flags */
6508 	BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
6509 };
6510 
6511 /* List of known BPF sock_ops operators.
6512  * New entries can only be added at the end
6513  */
6514 enum {
6515 	BPF_SOCK_OPS_VOID,
6516 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
6517 					 * -1 if default value should be used
6518 					 */
6519 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
6520 					 * window (in packets) or -1 if default
6521 					 * value should be used
6522 					 */
6523 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
6524 					 * active connection is initialized
6525 					 */
6526 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
6527 						 * active connection is
6528 						 * established
6529 						 */
6530 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
6531 						 * passive connection is
6532 						 * established
6533 						 */
6534 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
6535 					 * needs ECN
6536 					 */
6537 	BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is
6538 					 * based on the path and may be
6539 					 * dependent on the congestion control
6540 					 * algorithm. In general it indicates
6541 					 * a congestion threshold. RTTs above
6542 					 * this indicate congestion
6543 					 */
6544 	BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.
6545 					 * Arg1: value of icsk_retransmits
6546 					 * Arg2: value of icsk_rto
6547 					 * Arg3: whether RTO has expired
6548 					 */
6549 	BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.
6550 					 * Arg1: sequence number of 1st byte
6551 					 * Arg2: # segments
6552 					 * Arg3: return value of
6553 					 *       tcp_transmit_skb (0 => success)
6554 					 */
6555 	BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.
6556 					 * Arg1: old_state
6557 					 * Arg2: new_state
6558 					 */
6559 	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
6560 					 * socket transition to LISTEN state.
6561 					 */
6562 	BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.
6563 					 */
6564 	BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.
6565 					 * It will be called to handle
6566 					 * the packets received at
6567 					 * an already established
6568 					 * connection.
6569 					 *
6570 					 * sock_ops->skb_data:
6571 					 * Referring to the received skb.
6572 					 * It covers the TCP header only.
6573 					 *
6574 					 * bpf_load_hdr_opt() can also
6575 					 * be used to search for a
6576 					 * particular option.
6577 					 */
6578 	BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the
6579 					 * header option later in
6580 					 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6581 					 * Arg1: bool want_cookie. (in
6582 					 *       writing SYNACK only)
6583 					 *
6584 					 * sock_ops->skb_data:
6585 					 * Not available because no header has
6586 					 * been	written yet.
6587 					 *
6588 					 * sock_ops->skb_tcp_flags:
6589 					 * The tcp_flags of the
6590 					 * outgoing skb. (e.g. SYN, ACK, FIN).
6591 					 *
6592 					 * bpf_reserve_hdr_opt() should
6593 					 * be used to reserve space.
6594 					 */
6595 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options
6596 					 * Arg1: bool want_cookie. (in
6597 					 *       writing SYNACK only)
6598 					 *
6599 					 * sock_ops->skb_data:
6600 					 * Referring to the outgoing skb.
6601 					 * It covers the TCP header
6602 					 * that has already been written
6603 					 * by the kernel and the
6604 					 * earlier bpf-progs.
6605 					 *
6606 					 * sock_ops->skb_tcp_flags:
6607 					 * The tcp_flags of the outgoing
6608 					 * skb. (e.g. SYN, ACK, FIN).
6609 					 *
6610 					 * bpf_store_hdr_opt() should
6611 					 * be used to write the
6612 					 * option.
6613 					 *
6614 					 * bpf_load_hdr_opt() can also
6615 					 * be used to search for a
6616 					 * particular option that
6617 					 * has already been written
6618 					 * by the kernel or the
6619 					 * earlier bpf-progs.
6620 					 */
6621 };
6622 
6623 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
6624  * changes between the TCP and BPF versions. Ideally this should never happen.
6625  * If it does, we need to add code to convert them before calling
6626  * the BPF sock_ops function.
6627  */
6628 enum {
6629 	BPF_TCP_ESTABLISHED = 1,
6630 	BPF_TCP_SYN_SENT,
6631 	BPF_TCP_SYN_RECV,
6632 	BPF_TCP_FIN_WAIT1,
6633 	BPF_TCP_FIN_WAIT2,
6634 	BPF_TCP_TIME_WAIT,
6635 	BPF_TCP_CLOSE,
6636 	BPF_TCP_CLOSE_WAIT,
6637 	BPF_TCP_LAST_ACK,
6638 	BPF_TCP_LISTEN,
6639 	BPF_TCP_CLOSING,	/* Now a valid state */
6640 	BPF_TCP_NEW_SYN_RECV,
6641 
6642 	BPF_TCP_MAX_STATES	/* Leave at the end! */
6643 };
6644 
6645 enum {
6646 	TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */
6647 	TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */
6648 	TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */
6649 	TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */
6650 	/* Copy the SYN pkt to optval
6651 	 *
6652 	 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
6653 	 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6654 	 * to only getting from the saved_syn.  It can either get the
6655 	 * syn packet from:
6656 	 *
6657 	 * 1. the just-received SYN packet (only available when writing the
6658 	 *    SYNACK).  It will be useful when it is not necessary to
6659 	 *    save the SYN packet for latter use.  It is also the only way
6660 	 *    to get the SYN during syncookie mode because the syn
6661 	 *    packet cannot be saved during syncookie.
6662 	 *
6663 	 * OR
6664 	 *
6665 	 * 2. the earlier saved syn which was done by
6666 	 *    bpf_setsockopt(TCP_SAVE_SYN).
6667 	 *
6668 	 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6669 	 * SYN packet is obtained.
6670 	 *
6671 	 * If the bpf-prog does not need the IP[46] header,  the
6672 	 * bpf-prog can avoid parsing the IP header by using
6673 	 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
6674 	 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6675 	 *
6676 	 *      >0: Total number of bytes copied
6677 	 * -ENOSPC: Not enough space in optval. Only optlen number of
6678 	 *          bytes is copied.
6679 	 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6680 	 *	    is not saved by setsockopt(TCP_SAVE_SYN).
6681 	 */
6682 	TCP_BPF_SYN		= 1005, /* Copy the TCP header */
6683 	TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */
6684 	TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
6685 };
6686 
6687 enum {
6688 	BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6689 };
6690 
6691 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6692  * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6693  */
6694 enum {
6695 	BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the
6696 						 * total option spaces
6697 						 * required for an established
6698 						 * sk in order to calculate the
6699 						 * MSS.  No skb is actually
6700 						 * sent.
6701 						 */
6702 	BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode
6703 						 * when sending a SYN.
6704 						 */
6705 };
6706 
6707 struct bpf_perf_event_value {
6708 	__u64 counter;
6709 	__u64 enabled;
6710 	__u64 running;
6711 };
6712 
6713 enum {
6714 	BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),
6715 	BPF_DEVCG_ACC_READ	= (1ULL << 1),
6716 	BPF_DEVCG_ACC_WRITE	= (1ULL << 2),
6717 };
6718 
6719 enum {
6720 	BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),
6721 	BPF_DEVCG_DEV_CHAR	= (1ULL << 1),
6722 };
6723 
6724 struct bpf_cgroup_dev_ctx {
6725 	/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6726 	__u32 access_type;
6727 	__u32 major;
6728 	__u32 minor;
6729 };
6730 
6731 struct bpf_raw_tracepoint_args {
6732 	__u64 args[0];
6733 };
6734 
6735 /* DIRECT:  Skip the FIB rules and go to FIB table associated with device
6736  * OUTPUT:  Do lookup from egress perspective; default is ingress
6737  */
6738 enum {
6739 	BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
6740 	BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
6741 };
6742 
6743 enum {
6744 	BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
6745 	BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
6746 	BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
6747 	BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
6748 	BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
6749 	BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6750 	BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
6751 	BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
6752 	BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6753 };
6754 
6755 struct bpf_fib_lookup {
6756 	/* input:  network family for lookup (AF_INET, AF_INET6)
6757 	 * output: network family of egress nexthop
6758 	 */
6759 	__u8	family;
6760 
6761 	/* set if lookup is to consider L4 data - e.g., FIB rules */
6762 	__u8	l4_protocol;
6763 	__be16	sport;
6764 	__be16	dport;
6765 
6766 	union {	/* used for MTU check */
6767 		/* input to lookup */
6768 		__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */
6769 
6770 		/* output: MTU value */
6771 		__u16	mtu_result;
6772 	};
6773 	/* input: L3 device index for lookup
6774 	 * output: device index from FIB lookup
6775 	 */
6776 	__u32	ifindex;
6777 
6778 	union {
6779 		/* inputs to lookup */
6780 		__u8	tos;		/* AF_INET  */
6781 		__be32	flowinfo;	/* AF_INET6, flow_label + priority */
6782 
6783 		/* output: metric of fib result (IPv4/IPv6 only) */
6784 		__u32	rt_metric;
6785 	};
6786 
6787 	union {
6788 		__be32		ipv4_src;
6789 		__u32		ipv6_src[4];  /* in6_addr; network order */
6790 	};
6791 
6792 	/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
6793 	 * network header. output: bpf_fib_lookup sets to gateway address
6794 	 * if FIB lookup returns gateway route
6795 	 */
6796 	union {
6797 		__be32		ipv4_dst;
6798 		__u32		ipv6_dst[4];  /* in6_addr; network order */
6799 	};
6800 
6801 	/* output */
6802 	__be16	h_vlan_proto;
6803 	__be16	h_vlan_TCI;
6804 	__u8	smac[6];     /* ETH_ALEN */
6805 	__u8	dmac[6];     /* ETH_ALEN */
6806 };
6807 
6808 struct bpf_redir_neigh {
6809 	/* network family for lookup (AF_INET, AF_INET6) */
6810 	__u32 nh_family;
6811 	/* network address of nexthop; skips fib lookup to find gateway */
6812 	union {
6813 		__be32		ipv4_nh;
6814 		__u32		ipv6_nh[4];  /* in6_addr; network order */
6815 	};
6816 };
6817 
6818 /* bpf_check_mtu flags*/
6819 enum  bpf_check_mtu_flags {
6820 	BPF_MTU_CHK_SEGS  = (1U << 0),
6821 };
6822 
6823 enum bpf_check_mtu_ret {
6824 	BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
6825 	BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6826 	BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
6827 };
6828 
6829 enum bpf_task_fd_type {
6830 	BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */
6831 	BPF_FD_TYPE_TRACEPOINT,		/* tp name */
6832 	BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */
6833 	BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */
6834 	BPF_FD_TYPE_UPROBE,		/* filename + offset */
6835 	BPF_FD_TYPE_URETPROBE,		/* filename + offset */
6836 };
6837 
6838 enum {
6839 	BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),
6840 	BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),
6841 	BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),
6842 };
6843 
6844 struct bpf_flow_keys {
6845 	__u16	nhoff;
6846 	__u16	thoff;
6847 	__u16	addr_proto;			/* ETH_P_* of valid addrs */
6848 	__u8	is_frag;
6849 	__u8	is_first_frag;
6850 	__u8	is_encap;
6851 	__u8	ip_proto;
6852 	__be16	n_proto;
6853 	__be16	sport;
6854 	__be16	dport;
6855 	union {
6856 		struct {
6857 			__be32	ipv4_src;
6858 			__be32	ipv4_dst;
6859 		};
6860 		struct {
6861 			__u32	ipv6_src[4];	/* in6_addr; network order */
6862 			__u32	ipv6_dst[4];	/* in6_addr; network order */
6863 		};
6864 	};
6865 	__u32	flags;
6866 	__be32	flow_label;
6867 };
6868 
6869 struct bpf_func_info {
6870 	__u32	insn_off;
6871 	__u32	type_id;
6872 };
6873 
6874 #define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)
6875 #define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)
6876 
6877 struct bpf_line_info {
6878 	__u32	insn_off;
6879 	__u32	file_name_off;
6880 	__u32	line_off;
6881 	__u32	line_col;
6882 };
6883 
6884 struct bpf_spin_lock {
6885 	__u32	val;
6886 };
6887 
6888 struct bpf_timer {
6889 	__u64 :64;
6890 	__u64 :64;
6891 } __attribute__((aligned(8)));
6892 
6893 struct bpf_dynptr {
6894 	__u64 :64;
6895 	__u64 :64;
6896 } __attribute__((aligned(8)));
6897 
6898 struct bpf_list_head {
6899 	__u64 :64;
6900 	__u64 :64;
6901 } __attribute__((aligned(8)));
6902 
6903 struct bpf_list_node {
6904 	__u64 :64;
6905 	__u64 :64;
6906 } __attribute__((aligned(8)));
6907 
6908 struct bpf_sysctl {
6909 	__u32	write;		/* Sysctl is being read (= 0) or written (= 1).
6910 				 * Allows 1,2,4-byte read, but no write.
6911 				 */
6912 	__u32	file_pos;	/* Sysctl file position to read from, write to.
6913 				 * Allows 1,2,4-byte read an 4-byte write.
6914 				 */
6915 };
6916 
6917 struct bpf_sockopt {
6918 	__bpf_md_ptr(struct bpf_sock *, sk);
6919 	__bpf_md_ptr(void *, optval);
6920 	__bpf_md_ptr(void *, optval_end);
6921 
6922 	__s32	level;
6923 	__s32	optname;
6924 	__s32	optlen;
6925 	__s32	retval;
6926 };
6927 
6928 struct bpf_pidns_info {
6929 	__u32 pid;
6930 	__u32 tgid;
6931 };
6932 
6933 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6934 struct bpf_sk_lookup {
6935 	union {
6936 		__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6937 		__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6938 	};
6939 
6940 	__u32 family;		/* Protocol family (AF_INET, AF_INET6) */
6941 	__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6942 	__u32 remote_ip4;	/* Network byte order */
6943 	__u32 remote_ip6[4];	/* Network byte order */
6944 	__be16 remote_port;	/* Network byte order */
6945 	__u16 :16;		/* Zero padding */
6946 	__u32 local_ip4;	/* Network byte order */
6947 	__u32 local_ip6[4];	/* Network byte order */
6948 	__u32 local_port;	/* Host byte order */
6949 	__u32 ingress_ifindex;		/* The arriving interface. Determined by inet_iif. */
6950 };
6951 
6952 /*
6953  * struct btf_ptr is used for typed pointer representation; the
6954  * type id is used to render the pointer data as the appropriate type
6955  * via the bpf_snprintf_btf() helper described above.  A flags field -
6956  * potentially to specify additional details about the BTF pointer
6957  * (rather than its mode of display) - is included for future use.
6958  * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6959  */
6960 struct btf_ptr {
6961 	void *ptr;
6962 	__u32 type_id;
6963 	__u32 flags;		/* BTF ptr flags; unused at present. */
6964 };
6965 
6966 /*
6967  * Flags to control bpf_snprintf_btf() behaviour.
6968  *     - BTF_F_COMPACT: no formatting around type information
6969  *     - BTF_F_NONAME: no struct/union member names/types
6970  *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6971  *       equivalent to %px.
6972  *     - BTF_F_ZERO: show zero-valued struct/union members; they
6973  *       are not displayed by default
6974  */
6975 enum {
6976 	BTF_F_COMPACT	=	(1ULL << 0),
6977 	BTF_F_NONAME	=	(1ULL << 1),
6978 	BTF_F_PTR_RAW	=	(1ULL << 2),
6979 	BTF_F_ZERO	=	(1ULL << 3),
6980 };
6981 
6982 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
6983  * has to be adjusted by relocations. It is emitted by llvm and passed to
6984  * libbpf and later to the kernel.
6985  */
6986 enum bpf_core_relo_kind {
6987 	BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */
6988 	BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */
6989 	BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */
6990 	BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */
6991 	BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */
6992 	BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */
6993 	BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */
6994 	BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */
6995 	BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */
6996 	BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */
6997 	BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */
6998 	BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */
6999 	BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */
7000 };
7001 
7002 /*
7003  * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
7004  * and from libbpf to the kernel.
7005  *
7006  * CO-RE relocation captures the following data:
7007  * - insn_off - instruction offset (in bytes) within a BPF program that needs
7008  *   its insn->imm field to be relocated with actual field info;
7009  * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
7010  *   type or field;
7011  * - access_str_off - offset into corresponding .BTF string section. String
7012  *   interpretation depends on specific relocation kind:
7013  *     - for field-based relocations, string encodes an accessed field using
7014  *       a sequence of field and array indices, separated by colon (:). It's
7015  *       conceptually very close to LLVM's getelementptr ([0]) instruction's
7016  *       arguments for identifying offset to a field.
7017  *     - for type-based relocations, strings is expected to be just "0";
7018  *     - for enum value-based relocations, string contains an index of enum
7019  *       value within its enum type;
7020  * - kind - one of enum bpf_core_relo_kind;
7021  *
7022  * Example:
7023  *   struct sample {
7024  *       int a;
7025  *       struct {
7026  *           int b[10];
7027  *       };
7028  *   };
7029  *
7030  *   struct sample *s = ...;
7031  *   int *x = &s->a;     // encoded as "0:0" (a is field #0)
7032  *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,
7033  *                       // b is field #0 inside anon struct, accessing elem #5)
7034  *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
7035  *
7036  * type_id for all relocs in this example will capture BTF type id of
7037  * `struct sample`.
7038  *
7039  * Such relocation is emitted when using __builtin_preserve_access_index()
7040  * Clang built-in, passing expression that captures field address, e.g.:
7041  *
7042  * bpf_probe_read(&dst, sizeof(dst),
7043  *		  __builtin_preserve_access_index(&src->a.b.c));
7044  *
7045  * In this case Clang will emit field relocation recording necessary data to
7046  * be able to find offset of embedded `a.b.c` field within `src` struct.
7047  *
7048  * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
7049  */
7050 struct bpf_core_relo {
7051 	__u32 insn_off;
7052 	__u32 type_id;
7053 	__u32 access_str_off;
7054 	enum bpf_core_relo_kind kind;
7055 };
7056 
7057 #endif /* _UAPI__LINUX_BPF_H__ */
7058