xref: /linux-6.15/include/uapi/linux/bpf.h (revision fb3ceec1)
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[];	/* 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 union bpf_iter_link_info {
91 	struct {
92 		__u32	map_fd;
93 	} map;
94 };
95 
96 /* BPF syscall commands, see bpf(2) man-page for more details. */
97 /**
98  * DOC: eBPF Syscall Preamble
99  *
100  * The operation to be performed by the **bpf**\ () system call is determined
101  * by the *cmd* argument. Each operation takes an accompanying argument,
102  * provided via *attr*, which is a pointer to a union of type *bpf_attr* (see
103  * below). The size argument is the size of the union pointed to by *attr*.
104  */
105 /**
106  * DOC: eBPF Syscall Commands
107  *
108  * BPF_MAP_CREATE
109  *	Description
110  *		Create a map and return a file descriptor that refers to the
111  *		map. The close-on-exec file descriptor flag (see **fcntl**\ (2))
112  *		is automatically enabled for the new file descriptor.
113  *
114  *		Applying **close**\ (2) to the file descriptor returned by
115  *		**BPF_MAP_CREATE** will delete the map (but see NOTES).
116  *
117  *	Return
118  *		A new file descriptor (a nonnegative integer), or -1 if an
119  *		error occurred (in which case, *errno* is set appropriately).
120  *
121  * BPF_MAP_LOOKUP_ELEM
122  *	Description
123  *		Look up an element with a given *key* in the map referred to
124  *		by the file descriptor *map_fd*.
125  *
126  *		The *flags* argument may be specified as one of the
127  *		following:
128  *
129  *		**BPF_F_LOCK**
130  *			Look up the value of a spin-locked map without
131  *			returning the lock. This must be specified if the
132  *			elements contain a spinlock.
133  *
134  *	Return
135  *		Returns zero on success. On error, -1 is returned and *errno*
136  *		is set appropriately.
137  *
138  * BPF_MAP_UPDATE_ELEM
139  *	Description
140  *		Create or update an element (key/value pair) in a specified map.
141  *
142  *		The *flags* argument should be specified as one of the
143  *		following:
144  *
145  *		**BPF_ANY**
146  *			Create a new element or update an existing element.
147  *		**BPF_NOEXIST**
148  *			Create a new element only if it did not exist.
149  *		**BPF_EXIST**
150  *			Update an existing element.
151  *		**BPF_F_LOCK**
152  *			Update a spin_lock-ed map element.
153  *
154  *	Return
155  *		Returns zero on success. On error, -1 is returned and *errno*
156  *		is set appropriately.
157  *
158  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**,
159  *		**E2BIG**, **EEXIST**, or **ENOENT**.
160  *
161  *		**E2BIG**
162  *			The number of elements in the map reached the
163  *			*max_entries* limit specified at map creation time.
164  *		**EEXIST**
165  *			If *flags* specifies **BPF_NOEXIST** and the element
166  *			with *key* already exists in the map.
167  *		**ENOENT**
168  *			If *flags* specifies **BPF_EXIST** and the element with
169  *			*key* does not exist in the map.
170  *
171  * BPF_MAP_DELETE_ELEM
172  *	Description
173  *		Look up and delete an element by key in a specified map.
174  *
175  *	Return
176  *		Returns zero on success. On error, -1 is returned and *errno*
177  *		is set appropriately.
178  *
179  * BPF_MAP_GET_NEXT_KEY
180  *	Description
181  *		Look up an element by key in a specified map and return the key
182  *		of the next element. Can be used to iterate over all elements
183  *		in the map.
184  *
185  *	Return
186  *		Returns zero on success. On error, -1 is returned and *errno*
187  *		is set appropriately.
188  *
189  *		The following cases can be used to iterate over all elements of
190  *		the map:
191  *
192  *		* If *key* is not found, the operation returns zero and sets
193  *		  the *next_key* pointer to the key of the first element.
194  *		* If *key* is found, the operation returns zero and sets the
195  *		  *next_key* pointer to the key of the next element.
196  *		* If *key* is the last element, returns -1 and *errno* is set
197  *		  to **ENOENT**.
198  *
199  *		May set *errno* to **ENOMEM**, **EFAULT**, **EPERM**, or
200  *		**EINVAL** on error.
201  *
202  * BPF_PROG_LOAD
203  *	Description
204  *		Verify and load an eBPF program, returning a new file
205  *		descriptor associated with the program.
206  *
207  *		Applying **close**\ (2) to the file descriptor returned by
208  *		**BPF_PROG_LOAD** will unload the eBPF program (but see NOTES).
209  *
210  *		The close-on-exec file descriptor flag (see **fcntl**\ (2)) is
211  *		automatically enabled for the new file descriptor.
212  *
213  *	Return
214  *		A new file descriptor (a nonnegative integer), or -1 if an
215  *		error occurred (in which case, *errno* is set appropriately).
216  *
217  * BPF_OBJ_PIN
218  *	Description
219  *		Pin an eBPF program or map referred by the specified *bpf_fd*
220  *		to the provided *pathname* on the filesystem.
221  *
222  *		The *pathname* argument must not contain a dot (".").
223  *
224  *		On success, *pathname* retains a reference to the eBPF object,
225  *		preventing deallocation of the object when the original
226  *		*bpf_fd* is closed. This allow the eBPF object to live beyond
227  *		**close**\ (\ *bpf_fd*\ ), and hence the lifetime of the parent
228  *		process.
229  *
230  *		Applying **unlink**\ (2) or similar calls to the *pathname*
231  *		unpins the object from the filesystem, removing the reference.
232  *		If no other file descriptors or filesystem nodes refer to the
233  *		same object, it will be deallocated (see NOTES).
234  *
235  *		The filesystem type for the parent directory of *pathname* must
236  *		be **BPF_FS_MAGIC**.
237  *
238  *	Return
239  *		Returns zero on success. On error, -1 is returned and *errno*
240  *		is set appropriately.
241  *
242  * BPF_OBJ_GET
243  *	Description
244  *		Open a file descriptor for the eBPF object pinned to the
245  *		specified *pathname*.
246  *
247  *	Return
248  *		A new file descriptor (a nonnegative integer), or -1 if an
249  *		error occurred (in which case, *errno* is set appropriately).
250  *
251  * BPF_PROG_ATTACH
252  *	Description
253  *		Attach an eBPF program to a *target_fd* at the specified
254  *		*attach_type* hook.
255  *
256  *		The *attach_type* specifies the eBPF attachment point to
257  *		attach the program to, and must be one of *bpf_attach_type*
258  *		(see below).
259  *
260  *		The *attach_bpf_fd* must be a valid file descriptor for a
261  *		loaded eBPF program of a cgroup, flow dissector, LIRC, sockmap
262  *		or sock_ops type corresponding to the specified *attach_type*.
263  *
264  *		The *target_fd* must be a valid file descriptor for a kernel
265  *		object which depends on the attach type of *attach_bpf_fd*:
266  *
267  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
268  *		**BPF_PROG_TYPE_CGROUP_SKB**,
269  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
270  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
271  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
272  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
273  *		**BPF_PROG_TYPE_SOCK_OPS**
274  *
275  *			Control Group v2 hierarchy with the eBPF controller
276  *			enabled. Requires the kernel to be compiled with
277  *			**CONFIG_CGROUP_BPF**.
278  *
279  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
280  *
281  *			Network namespace (eg /proc/self/ns/net).
282  *
283  *		**BPF_PROG_TYPE_LIRC_MODE2**
284  *
285  *			LIRC device path (eg /dev/lircN). Requires the kernel
286  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
287  *
288  *		**BPF_PROG_TYPE_SK_SKB**,
289  *		**BPF_PROG_TYPE_SK_MSG**
290  *
291  *			eBPF map of socket type (eg **BPF_MAP_TYPE_SOCKHASH**).
292  *
293  *	Return
294  *		Returns zero on success. On error, -1 is returned and *errno*
295  *		is set appropriately.
296  *
297  * BPF_PROG_DETACH
298  *	Description
299  *		Detach the eBPF program associated with the *target_fd* at the
300  *		hook specified by *attach_type*. The program must have been
301  *		previously attached using **BPF_PROG_ATTACH**.
302  *
303  *	Return
304  *		Returns zero on success. On error, -1 is returned and *errno*
305  *		is set appropriately.
306  *
307  * BPF_PROG_TEST_RUN
308  *	Description
309  *		Run the eBPF program associated with the *prog_fd* a *repeat*
310  *		number of times against a provided program context *ctx_in* and
311  *		data *data_in*, and return the modified program context
312  *		*ctx_out*, *data_out* (for example, packet data), result of the
313  *		execution *retval*, and *duration* of the test run.
314  *
315  *		The sizes of the buffers provided as input and output
316  *		parameters *ctx_in*, *ctx_out*, *data_in*, and *data_out* must
317  *		be provided in the corresponding variables *ctx_size_in*,
318  *		*ctx_size_out*, *data_size_in*, and/or *data_size_out*. If any
319  *		of these parameters are not provided (ie set to NULL), the
320  *		corresponding size field must be zero.
321  *
322  *		Some program types have particular requirements:
323  *
324  *		**BPF_PROG_TYPE_SK_LOOKUP**
325  *			*data_in* and *data_out* must be NULL.
326  *
327  *		**BPF_PROG_TYPE_RAW_TRACEPOINT**,
328  *		**BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE**
329  *
330  *			*ctx_out*, *data_in* and *data_out* must be NULL.
331  *			*repeat* must be zero.
332  *
333  *		BPF_PROG_RUN is an alias for BPF_PROG_TEST_RUN.
334  *
335  *	Return
336  *		Returns zero on success. On error, -1 is returned and *errno*
337  *		is set appropriately.
338  *
339  *		**ENOSPC**
340  *			Either *data_size_out* or *ctx_size_out* is too small.
341  *		**ENOTSUPP**
342  *			This command is not supported by the program type of
343  *			the program referred to by *prog_fd*.
344  *
345  * BPF_PROG_GET_NEXT_ID
346  *	Description
347  *		Fetch the next eBPF program currently loaded into the kernel.
348  *
349  *		Looks for the eBPF program with an id greater than *start_id*
350  *		and updates *next_id* on success. If no other eBPF programs
351  *		remain with ids higher than *start_id*, returns -1 and sets
352  *		*errno* to **ENOENT**.
353  *
354  *	Return
355  *		Returns zero on success. On error, or when no id remains, -1
356  *		is returned and *errno* is set appropriately.
357  *
358  * BPF_MAP_GET_NEXT_ID
359  *	Description
360  *		Fetch the next eBPF map currently loaded into the kernel.
361  *
362  *		Looks for the eBPF map with an id greater than *start_id*
363  *		and updates *next_id* on success. If no other eBPF maps
364  *		remain with ids higher than *start_id*, returns -1 and sets
365  *		*errno* to **ENOENT**.
366  *
367  *	Return
368  *		Returns zero on success. On error, or when no id remains, -1
369  *		is returned and *errno* is set appropriately.
370  *
371  * BPF_PROG_GET_FD_BY_ID
372  *	Description
373  *		Open a file descriptor for the eBPF program corresponding to
374  *		*prog_id*.
375  *
376  *	Return
377  *		A new file descriptor (a nonnegative integer), or -1 if an
378  *		error occurred (in which case, *errno* is set appropriately).
379  *
380  * BPF_MAP_GET_FD_BY_ID
381  *	Description
382  *		Open a file descriptor for the eBPF map corresponding to
383  *		*map_id*.
384  *
385  *	Return
386  *		A new file descriptor (a nonnegative integer), or -1 if an
387  *		error occurred (in which case, *errno* is set appropriately).
388  *
389  * BPF_OBJ_GET_INFO_BY_FD
390  *	Description
391  *		Obtain information about the eBPF object corresponding to
392  *		*bpf_fd*.
393  *
394  *		Populates up to *info_len* bytes of *info*, which will be in
395  *		one of the following formats depending on the eBPF object type
396  *		of *bpf_fd*:
397  *
398  *		* **struct bpf_prog_info**
399  *		* **struct bpf_map_info**
400  *		* **struct bpf_btf_info**
401  *		* **struct bpf_link_info**
402  *
403  *	Return
404  *		Returns zero on success. On error, -1 is returned and *errno*
405  *		is set appropriately.
406  *
407  * BPF_PROG_QUERY
408  *	Description
409  *		Obtain information about eBPF programs associated with the
410  *		specified *attach_type* hook.
411  *
412  *		The *target_fd* must be a valid file descriptor for a kernel
413  *		object which depends on the attach type of *attach_bpf_fd*:
414  *
415  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
416  *		**BPF_PROG_TYPE_CGROUP_SKB**,
417  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
418  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
419  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
420  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
421  *		**BPF_PROG_TYPE_SOCK_OPS**
422  *
423  *			Control Group v2 hierarchy with the eBPF controller
424  *			enabled. Requires the kernel to be compiled with
425  *			**CONFIG_CGROUP_BPF**.
426  *
427  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
428  *
429  *			Network namespace (eg /proc/self/ns/net).
430  *
431  *		**BPF_PROG_TYPE_LIRC_MODE2**
432  *
433  *			LIRC device path (eg /dev/lircN). Requires the kernel
434  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
435  *
436  *		**BPF_PROG_QUERY** always fetches the number of programs
437  *		attached and the *attach_flags* which were used to attach those
438  *		programs. Additionally, if *prog_ids* is nonzero and the number
439  *		of attached programs is less than *prog_cnt*, populates
440  *		*prog_ids* with the eBPF program ids of the programs attached
441  *		at *target_fd*.
442  *
443  *		The following flags may alter the result:
444  *
445  *		**BPF_F_QUERY_EFFECTIVE**
446  *			Only return information regarding programs which are
447  *			currently effective at the specified *target_fd*.
448  *
449  *	Return
450  *		Returns zero on success. On error, -1 is returned and *errno*
451  *		is set appropriately.
452  *
453  * BPF_RAW_TRACEPOINT_OPEN
454  *	Description
455  *		Attach an eBPF program to a tracepoint *name* to access kernel
456  *		internal arguments of the tracepoint in their raw form.
457  *
458  *		The *prog_fd* must be a valid file descriptor associated with
459  *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
460  *
461  *		No ABI guarantees are made about the content of tracepoint
462  *		arguments exposed to the corresponding eBPF program.
463  *
464  *		Applying **close**\ (2) to the file descriptor returned by
465  *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
466  *
467  *	Return
468  *		A new file descriptor (a nonnegative integer), or -1 if an
469  *		error occurred (in which case, *errno* is set appropriately).
470  *
471  * BPF_BTF_LOAD
472  *	Description
473  *		Verify and load BPF Type Format (BTF) metadata into the kernel,
474  *		returning a new file descriptor associated with the metadata.
475  *		BTF is described in more detail at
476  *		https://www.kernel.org/doc/html/latest/bpf/btf.html.
477  *
478  *		The *btf* parameter must point to valid memory providing
479  *		*btf_size* bytes of BTF binary metadata.
480  *
481  *		The returned file descriptor can be passed to other **bpf**\ ()
482  *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
483  *		associate the BTF with those objects.
484  *
485  *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
486  *		parameters to specify a *btf_log_buf*, *btf_log_size* and
487  *		*btf_log_level* which allow the kernel to return freeform log
488  *		output regarding the BTF verification process.
489  *
490  *	Return
491  *		A new file descriptor (a nonnegative integer), or -1 if an
492  *		error occurred (in which case, *errno* is set appropriately).
493  *
494  * BPF_BTF_GET_FD_BY_ID
495  *	Description
496  *		Open a file descriptor for the BPF Type Format (BTF)
497  *		corresponding to *btf_id*.
498  *
499  *	Return
500  *		A new file descriptor (a nonnegative integer), or -1 if an
501  *		error occurred (in which case, *errno* is set appropriately).
502  *
503  * BPF_TASK_FD_QUERY
504  *	Description
505  *		Obtain information about eBPF programs associated with the
506  *		target process identified by *pid* and *fd*.
507  *
508  *		If the *pid* and *fd* are associated with a tracepoint, kprobe
509  *		or uprobe perf event, then the *prog_id* and *fd_type* will
510  *		be populated with the eBPF program id and file descriptor type
511  *		of type **bpf_task_fd_type**. If associated with a kprobe or
512  *		uprobe, the  *probe_offset* and *probe_addr* will also be
513  *		populated. Optionally, if *buf* is provided, then up to
514  *		*buf_len* bytes of *buf* will be populated with the name of
515  *		the tracepoint, kprobe or uprobe.
516  *
517  *		The resulting *prog_id* may be introspected in deeper detail
518  *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
519  *
520  *	Return
521  *		Returns zero on success. On error, -1 is returned and *errno*
522  *		is set appropriately.
523  *
524  * BPF_MAP_LOOKUP_AND_DELETE_ELEM
525  *	Description
526  *		Look up an element with the given *key* in the map referred to
527  *		by the file descriptor *fd*, and if found, delete the element.
528  *
529  *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
530  *		types, the *flags* argument needs to be set to 0, but for other
531  *		map types, it may be specified as:
532  *
533  *		**BPF_F_LOCK**
534  *			Look up and delete the value of a spin-locked map
535  *			without returning the lock. This must be specified if
536  *			the elements contain a spinlock.
537  *
538  *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
539  *		implement this command as a "pop" operation, deleting the top
540  *		element rather than one corresponding to *key*.
541  *		The *key* and *key_len* parameters should be zeroed when
542  *		issuing this operation for these map types.
543  *
544  *		This command is only valid for the following map types:
545  *		* **BPF_MAP_TYPE_QUEUE**
546  *		* **BPF_MAP_TYPE_STACK**
547  *		* **BPF_MAP_TYPE_HASH**
548  *		* **BPF_MAP_TYPE_PERCPU_HASH**
549  *		* **BPF_MAP_TYPE_LRU_HASH**
550  *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**
551  *
552  *	Return
553  *		Returns zero on success. On error, -1 is returned and *errno*
554  *		is set appropriately.
555  *
556  * BPF_MAP_FREEZE
557  *	Description
558  *		Freeze the permissions of the specified map.
559  *
560  *		Write permissions may be frozen by passing zero *flags*.
561  *		Upon success, no future syscall invocations may alter the
562  *		map state of *map_fd*. Write operations from eBPF programs
563  *		are still possible for a frozen map.
564  *
565  *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
566  *
567  *	Return
568  *		Returns zero on success. On error, -1 is returned and *errno*
569  *		is set appropriately.
570  *
571  * BPF_BTF_GET_NEXT_ID
572  *	Description
573  *		Fetch the next BPF Type Format (BTF) object currently loaded
574  *		into the kernel.
575  *
576  *		Looks for the BTF object with an id greater than *start_id*
577  *		and updates *next_id* on success. If no other BTF objects
578  *		remain with ids higher than *start_id*, returns -1 and sets
579  *		*errno* to **ENOENT**.
580  *
581  *	Return
582  *		Returns zero on success. On error, or when no id remains, -1
583  *		is returned and *errno* is set appropriately.
584  *
585  * BPF_MAP_LOOKUP_BATCH
586  *	Description
587  *		Iterate and fetch multiple elements in a map.
588  *
589  *		Two opaque values are used to manage batch operations,
590  *		*in_batch* and *out_batch*. Initially, *in_batch* must be set
591  *		to NULL to begin the batched operation. After each subsequent
592  *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
593  *		*out_batch* as the *in_batch* for the next operation to
594  *		continue iteration from the current point.
595  *
596  *		The *keys* and *values* are output parameters which must point
597  *		to memory large enough to hold *count* items based on the key
598  *		and value size of the map *map_fd*. The *keys* buffer must be
599  *		of *key_size* * *count*. The *values* buffer must be of
600  *		*value_size* * *count*.
601  *
602  *		The *elem_flags* argument may be specified as one of the
603  *		following:
604  *
605  *		**BPF_F_LOCK**
606  *			Look up the value of a spin-locked map without
607  *			returning the lock. This must be specified if the
608  *			elements contain a spinlock.
609  *
610  *		On success, *count* elements from the map are copied into the
611  *		user buffer, with the keys copied into *keys* and the values
612  *		copied into the corresponding indices in *values*.
613  *
614  *		If an error is returned and *errno* is not **EFAULT**, *count*
615  *		is set to the number of successfully processed elements.
616  *
617  *	Return
618  *		Returns zero on success. On error, -1 is returned and *errno*
619  *		is set appropriately.
620  *
621  *		May set *errno* to **ENOSPC** to indicate that *keys* or
622  *		*values* is too small to dump an entire bucket during
623  *		iteration of a hash-based map type.
624  *
625  * BPF_MAP_LOOKUP_AND_DELETE_BATCH
626  *	Description
627  *		Iterate and delete all elements in a map.
628  *
629  *		This operation has the same behavior as
630  *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:
631  *
632  *		* Every element that is successfully returned is also deleted
633  *		  from the map. This is at least *count* elements. Note that
634  *		  *count* is both an input and an output parameter.
635  *		* Upon returning with *errno* set to **EFAULT**, up to
636  *		  *count* elements may be deleted without returning the keys
637  *		  and values of the deleted elements.
638  *
639  *	Return
640  *		Returns zero on success. On error, -1 is returned and *errno*
641  *		is set appropriately.
642  *
643  * BPF_MAP_UPDATE_BATCH
644  *	Description
645  *		Update multiple elements in a map by *key*.
646  *
647  *		The *keys* and *values* are input parameters which must point
648  *		to memory large enough to hold *count* items based on the key
649  *		and value size of the map *map_fd*. The *keys* buffer must be
650  *		of *key_size* * *count*. The *values* buffer must be of
651  *		*value_size* * *count*.
652  *
653  *		Each element specified in *keys* is sequentially updated to the
654  *		value in the corresponding index in *values*. The *in_batch*
655  *		and *out_batch* parameters are ignored and should be zeroed.
656  *
657  *		The *elem_flags* argument should be specified as one of the
658  *		following:
659  *
660  *		**BPF_ANY**
661  *			Create new elements or update a existing elements.
662  *		**BPF_NOEXIST**
663  *			Create new elements only if they do not exist.
664  *		**BPF_EXIST**
665  *			Update existing elements.
666  *		**BPF_F_LOCK**
667  *			Update spin_lock-ed map elements. This must be
668  *			specified if the map value contains a spinlock.
669  *
670  *		On success, *count* elements from the map are updated.
671  *
672  *		If an error is returned and *errno* is not **EFAULT**, *count*
673  *		is set to the number of successfully processed elements.
674  *
675  *	Return
676  *		Returns zero on success. On error, -1 is returned and *errno*
677  *		is set appropriately.
678  *
679  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
680  *		**E2BIG**. **E2BIG** indicates that the number of elements in
681  *		the map reached the *max_entries* limit specified at map
682  *		creation time.
683  *
684  *		May set *errno* to one of the following error codes under
685  *		specific circumstances:
686  *
687  *		**EEXIST**
688  *			If *flags* specifies **BPF_NOEXIST** and the element
689  *			with *key* already exists in the map.
690  *		**ENOENT**
691  *			If *flags* specifies **BPF_EXIST** and the element with
692  *			*key* does not exist in the map.
693  *
694  * BPF_MAP_DELETE_BATCH
695  *	Description
696  *		Delete multiple elements in a map by *key*.
697  *
698  *		The *keys* parameter is an input parameter which must point
699  *		to memory large enough to hold *count* items based on the key
700  *		size of the map *map_fd*, that is, *key_size* * *count*.
701  *
702  *		Each element specified in *keys* is sequentially deleted. The
703  *		*in_batch*, *out_batch*, and *values* parameters are ignored
704  *		and should be zeroed.
705  *
706  *		The *elem_flags* argument may be specified as one of the
707  *		following:
708  *
709  *		**BPF_F_LOCK**
710  *			Look up the value of a spin-locked map without
711  *			returning the lock. This must be specified if the
712  *			elements contain a spinlock.
713  *
714  *		On success, *count* elements from the map are updated.
715  *
716  *		If an error is returned and *errno* is not **EFAULT**, *count*
717  *		is set to the number of successfully processed elements. If
718  *		*errno* is **EFAULT**, up to *count* elements may be been
719  *		deleted.
720  *
721  *	Return
722  *		Returns zero on success. On error, -1 is returned and *errno*
723  *		is set appropriately.
724  *
725  * BPF_LINK_CREATE
726  *	Description
727  *		Attach an eBPF program to a *target_fd* at the specified
728  *		*attach_type* hook and return a file descriptor handle for
729  *		managing the link.
730  *
731  *	Return
732  *		A new file descriptor (a nonnegative integer), or -1 if an
733  *		error occurred (in which case, *errno* is set appropriately).
734  *
735  * BPF_LINK_UPDATE
736  *	Description
737  *		Update the eBPF program in the specified *link_fd* to
738  *		*new_prog_fd*.
739  *
740  *	Return
741  *		Returns zero on success. On error, -1 is returned and *errno*
742  *		is set appropriately.
743  *
744  * BPF_LINK_GET_FD_BY_ID
745  *	Description
746  *		Open a file descriptor for the eBPF Link corresponding to
747  *		*link_id*.
748  *
749  *	Return
750  *		A new file descriptor (a nonnegative integer), or -1 if an
751  *		error occurred (in which case, *errno* is set appropriately).
752  *
753  * BPF_LINK_GET_NEXT_ID
754  *	Description
755  *		Fetch the next eBPF link currently loaded into the kernel.
756  *
757  *		Looks for the eBPF link with an id greater than *start_id*
758  *		and updates *next_id* on success. If no other eBPF links
759  *		remain with ids higher than *start_id*, returns -1 and sets
760  *		*errno* to **ENOENT**.
761  *
762  *	Return
763  *		Returns zero on success. On error, or when no id remains, -1
764  *		is returned and *errno* is set appropriately.
765  *
766  * BPF_ENABLE_STATS
767  *	Description
768  *		Enable eBPF runtime statistics gathering.
769  *
770  *		Runtime statistics gathering for the eBPF runtime is disabled
771  *		by default to minimize the corresponding performance overhead.
772  *		This command enables statistics globally.
773  *
774  *		Multiple programs may independently enable statistics.
775  *		After gathering the desired statistics, eBPF runtime statistics
776  *		may be disabled again by calling **close**\ (2) for the file
777  *		descriptor returned by this function. Statistics will only be
778  *		disabled system-wide when all outstanding file descriptors
779  *		returned by prior calls for this subcommand are closed.
780  *
781  *	Return
782  *		A new file descriptor (a nonnegative integer), or -1 if an
783  *		error occurred (in which case, *errno* is set appropriately).
784  *
785  * BPF_ITER_CREATE
786  *	Description
787  *		Create an iterator on top of the specified *link_fd* (as
788  *		previously created using **BPF_LINK_CREATE**) and return a
789  *		file descriptor that can be used to trigger the iteration.
790  *
791  *		If the resulting file descriptor is pinned to the filesystem
792  *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
793  *		for that path will trigger the iterator to read kernel state
794  *		using the eBPF program attached to *link_fd*.
795  *
796  *	Return
797  *		A new file descriptor (a nonnegative integer), or -1 if an
798  *		error occurred (in which case, *errno* is set appropriately).
799  *
800  * BPF_LINK_DETACH
801  *	Description
802  *		Forcefully detach the specified *link_fd* from its
803  *		corresponding attachment point.
804  *
805  *	Return
806  *		Returns zero on success. On error, -1 is returned and *errno*
807  *		is set appropriately.
808  *
809  * BPF_PROG_BIND_MAP
810  *	Description
811  *		Bind a map to the lifetime of an eBPF program.
812  *
813  *		The map identified by *map_fd* is bound to the program
814  *		identified by *prog_fd* and only released when *prog_fd* is
815  *		released. This may be used in cases where metadata should be
816  *		associated with a program which otherwise does not contain any
817  *		references to the map (for example, embedded in the eBPF
818  *		program instructions).
819  *
820  *	Return
821  *		Returns zero on success. On error, -1 is returned and *errno*
822  *		is set appropriately.
823  *
824  * NOTES
825  *	eBPF objects (maps and programs) can be shared between processes.
826  *
827  *	* After **fork**\ (2), the child inherits file descriptors
828  *	  referring to the same eBPF objects.
829  *	* File descriptors referring to eBPF objects can be transferred over
830  *	  **unix**\ (7) domain sockets.
831  *	* File descriptors referring to eBPF objects can be duplicated in the
832  *	  usual way, using **dup**\ (2) and similar calls.
833  *	* File descriptors referring to eBPF objects can be pinned to the
834  *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
835  *
836  *	An eBPF object is deallocated only after all file descriptors referring
837  *	to the object have been closed and no references remain pinned to the
838  *	filesystem or attached (for example, bound to a program or device).
839  */
840 enum bpf_cmd {
841 	BPF_MAP_CREATE,
842 	BPF_MAP_LOOKUP_ELEM,
843 	BPF_MAP_UPDATE_ELEM,
844 	BPF_MAP_DELETE_ELEM,
845 	BPF_MAP_GET_NEXT_KEY,
846 	BPF_PROG_LOAD,
847 	BPF_OBJ_PIN,
848 	BPF_OBJ_GET,
849 	BPF_PROG_ATTACH,
850 	BPF_PROG_DETACH,
851 	BPF_PROG_TEST_RUN,
852 	BPF_PROG_RUN = BPF_PROG_TEST_RUN,
853 	BPF_PROG_GET_NEXT_ID,
854 	BPF_MAP_GET_NEXT_ID,
855 	BPF_PROG_GET_FD_BY_ID,
856 	BPF_MAP_GET_FD_BY_ID,
857 	BPF_OBJ_GET_INFO_BY_FD,
858 	BPF_PROG_QUERY,
859 	BPF_RAW_TRACEPOINT_OPEN,
860 	BPF_BTF_LOAD,
861 	BPF_BTF_GET_FD_BY_ID,
862 	BPF_TASK_FD_QUERY,
863 	BPF_MAP_LOOKUP_AND_DELETE_ELEM,
864 	BPF_MAP_FREEZE,
865 	BPF_BTF_GET_NEXT_ID,
866 	BPF_MAP_LOOKUP_BATCH,
867 	BPF_MAP_LOOKUP_AND_DELETE_BATCH,
868 	BPF_MAP_UPDATE_BATCH,
869 	BPF_MAP_DELETE_BATCH,
870 	BPF_LINK_CREATE,
871 	BPF_LINK_UPDATE,
872 	BPF_LINK_GET_FD_BY_ID,
873 	BPF_LINK_GET_NEXT_ID,
874 	BPF_ENABLE_STATS,
875 	BPF_ITER_CREATE,
876 	BPF_LINK_DETACH,
877 	BPF_PROG_BIND_MAP,
878 };
879 
880 enum bpf_map_type {
881 	BPF_MAP_TYPE_UNSPEC,
882 	BPF_MAP_TYPE_HASH,
883 	BPF_MAP_TYPE_ARRAY,
884 	BPF_MAP_TYPE_PROG_ARRAY,
885 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
886 	BPF_MAP_TYPE_PERCPU_HASH,
887 	BPF_MAP_TYPE_PERCPU_ARRAY,
888 	BPF_MAP_TYPE_STACK_TRACE,
889 	BPF_MAP_TYPE_CGROUP_ARRAY,
890 	BPF_MAP_TYPE_LRU_HASH,
891 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
892 	BPF_MAP_TYPE_LPM_TRIE,
893 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
894 	BPF_MAP_TYPE_HASH_OF_MAPS,
895 	BPF_MAP_TYPE_DEVMAP,
896 	BPF_MAP_TYPE_SOCKMAP,
897 	BPF_MAP_TYPE_CPUMAP,
898 	BPF_MAP_TYPE_XSKMAP,
899 	BPF_MAP_TYPE_SOCKHASH,
900 	BPF_MAP_TYPE_CGROUP_STORAGE,
901 	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
902 	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
903 	BPF_MAP_TYPE_QUEUE,
904 	BPF_MAP_TYPE_STACK,
905 	BPF_MAP_TYPE_SK_STORAGE,
906 	BPF_MAP_TYPE_DEVMAP_HASH,
907 	BPF_MAP_TYPE_STRUCT_OPS,
908 	BPF_MAP_TYPE_RINGBUF,
909 	BPF_MAP_TYPE_INODE_STORAGE,
910 	BPF_MAP_TYPE_TASK_STORAGE,
911 	BPF_MAP_TYPE_BLOOM_FILTER,
912 };
913 
914 /* Note that tracing related programs such as
915  * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
916  * are not subject to a stable API since kernel internal data
917  * structures can change from release to release and may
918  * therefore break existing tracing BPF programs. Tracing BPF
919  * programs correspond to /a/ specific kernel which is to be
920  * analyzed, and not /a/ specific kernel /and/ all future ones.
921  */
922 enum bpf_prog_type {
923 	BPF_PROG_TYPE_UNSPEC,
924 	BPF_PROG_TYPE_SOCKET_FILTER,
925 	BPF_PROG_TYPE_KPROBE,
926 	BPF_PROG_TYPE_SCHED_CLS,
927 	BPF_PROG_TYPE_SCHED_ACT,
928 	BPF_PROG_TYPE_TRACEPOINT,
929 	BPF_PROG_TYPE_XDP,
930 	BPF_PROG_TYPE_PERF_EVENT,
931 	BPF_PROG_TYPE_CGROUP_SKB,
932 	BPF_PROG_TYPE_CGROUP_SOCK,
933 	BPF_PROG_TYPE_LWT_IN,
934 	BPF_PROG_TYPE_LWT_OUT,
935 	BPF_PROG_TYPE_LWT_XMIT,
936 	BPF_PROG_TYPE_SOCK_OPS,
937 	BPF_PROG_TYPE_SK_SKB,
938 	BPF_PROG_TYPE_CGROUP_DEVICE,
939 	BPF_PROG_TYPE_SK_MSG,
940 	BPF_PROG_TYPE_RAW_TRACEPOINT,
941 	BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
942 	BPF_PROG_TYPE_LWT_SEG6LOCAL,
943 	BPF_PROG_TYPE_LIRC_MODE2,
944 	BPF_PROG_TYPE_SK_REUSEPORT,
945 	BPF_PROG_TYPE_FLOW_DISSECTOR,
946 	BPF_PROG_TYPE_CGROUP_SYSCTL,
947 	BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
948 	BPF_PROG_TYPE_CGROUP_SOCKOPT,
949 	BPF_PROG_TYPE_TRACING,
950 	BPF_PROG_TYPE_STRUCT_OPS,
951 	BPF_PROG_TYPE_EXT,
952 	BPF_PROG_TYPE_LSM,
953 	BPF_PROG_TYPE_SK_LOOKUP,
954 	BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
955 };
956 
957 enum bpf_attach_type {
958 	BPF_CGROUP_INET_INGRESS,
959 	BPF_CGROUP_INET_EGRESS,
960 	BPF_CGROUP_INET_SOCK_CREATE,
961 	BPF_CGROUP_SOCK_OPS,
962 	BPF_SK_SKB_STREAM_PARSER,
963 	BPF_SK_SKB_STREAM_VERDICT,
964 	BPF_CGROUP_DEVICE,
965 	BPF_SK_MSG_VERDICT,
966 	BPF_CGROUP_INET4_BIND,
967 	BPF_CGROUP_INET6_BIND,
968 	BPF_CGROUP_INET4_CONNECT,
969 	BPF_CGROUP_INET6_CONNECT,
970 	BPF_CGROUP_INET4_POST_BIND,
971 	BPF_CGROUP_INET6_POST_BIND,
972 	BPF_CGROUP_UDP4_SENDMSG,
973 	BPF_CGROUP_UDP6_SENDMSG,
974 	BPF_LIRC_MODE2,
975 	BPF_FLOW_DISSECTOR,
976 	BPF_CGROUP_SYSCTL,
977 	BPF_CGROUP_UDP4_RECVMSG,
978 	BPF_CGROUP_UDP6_RECVMSG,
979 	BPF_CGROUP_GETSOCKOPT,
980 	BPF_CGROUP_SETSOCKOPT,
981 	BPF_TRACE_RAW_TP,
982 	BPF_TRACE_FENTRY,
983 	BPF_TRACE_FEXIT,
984 	BPF_MODIFY_RETURN,
985 	BPF_LSM_MAC,
986 	BPF_TRACE_ITER,
987 	BPF_CGROUP_INET4_GETPEERNAME,
988 	BPF_CGROUP_INET6_GETPEERNAME,
989 	BPF_CGROUP_INET4_GETSOCKNAME,
990 	BPF_CGROUP_INET6_GETSOCKNAME,
991 	BPF_XDP_DEVMAP,
992 	BPF_CGROUP_INET_SOCK_RELEASE,
993 	BPF_XDP_CPUMAP,
994 	BPF_SK_LOOKUP,
995 	BPF_XDP,
996 	BPF_SK_SKB_VERDICT,
997 	BPF_SK_REUSEPORT_SELECT,
998 	BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
999 	BPF_PERF_EVENT,
1000 	BPF_TRACE_KPROBE_MULTI,
1001 	BPF_LSM_CGROUP,
1002 	__MAX_BPF_ATTACH_TYPE
1003 };
1004 
1005 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1006 
1007 enum bpf_link_type {
1008 	BPF_LINK_TYPE_UNSPEC = 0,
1009 	BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1010 	BPF_LINK_TYPE_TRACING = 2,
1011 	BPF_LINK_TYPE_CGROUP = 3,
1012 	BPF_LINK_TYPE_ITER = 4,
1013 	BPF_LINK_TYPE_NETNS = 5,
1014 	BPF_LINK_TYPE_XDP = 6,
1015 	BPF_LINK_TYPE_PERF_EVENT = 7,
1016 	BPF_LINK_TYPE_KPROBE_MULTI = 8,
1017 	BPF_LINK_TYPE_STRUCT_OPS = 9,
1018 
1019 	MAX_BPF_LINK_TYPE,
1020 };
1021 
1022 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1023  *
1024  * NONE(default): No further bpf programs allowed in the subtree.
1025  *
1026  * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1027  * the program in this cgroup yields to sub-cgroup program.
1028  *
1029  * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1030  * that cgroup program gets run in addition to the program in this cgroup.
1031  *
1032  * Only one program is allowed to be attached to a cgroup with
1033  * NONE or BPF_F_ALLOW_OVERRIDE flag.
1034  * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1035  * release old program and attach the new one. Attach flags has to match.
1036  *
1037  * Multiple programs are allowed to be attached to a cgroup with
1038  * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1039  * (those that were attached first, run first)
1040  * The programs of sub-cgroup are executed first, then programs of
1041  * this cgroup and then programs of parent cgroup.
1042  * When children program makes decision (like picking TCP CA or sock bind)
1043  * parent program has a chance to override it.
1044  *
1045  * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1046  * programs for a cgroup. Though it's possible to replace an old program at
1047  * any position by also specifying BPF_F_REPLACE flag and position itself in
1048  * replace_bpf_fd attribute. Old program at this position will be released.
1049  *
1050  * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1051  * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1052  * Ex1:
1053  * cgrp1 (MULTI progs A, B) ->
1054  *    cgrp2 (OVERRIDE prog C) ->
1055  *      cgrp3 (MULTI prog D) ->
1056  *        cgrp4 (OVERRIDE prog E) ->
1057  *          cgrp5 (NONE prog F)
1058  * the event in cgrp5 triggers execution of F,D,A,B in that order.
1059  * if prog F is detached, the execution is E,D,A,B
1060  * if prog F and D are detached, the execution is E,A,B
1061  * if prog F, E and D are detached, the execution is C,A,B
1062  *
1063  * All eligible programs are executed regardless of return code from
1064  * earlier programs.
1065  */
1066 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
1067 #define BPF_F_ALLOW_MULTI	(1U << 1)
1068 #define BPF_F_REPLACE		(1U << 2)
1069 
1070 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1071  * verifier will perform strict alignment checking as if the kernel
1072  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1073  * and NET_IP_ALIGN defined to 2.
1074  */
1075 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
1076 
1077 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1078  * verifier will allow any alignment whatsoever.  On platforms
1079  * with strict alignment requirements for loads ands stores (such
1080  * as sparc and mips) the verifier validates that all loads and
1081  * stores provably follow this requirement.  This flag turns that
1082  * checking and enforcement off.
1083  *
1084  * It is mostly used for testing when we want to validate the
1085  * context and memory access aspects of the verifier, but because
1086  * of an unaligned access the alignment check would trigger before
1087  * the one we are interested in.
1088  */
1089 #define BPF_F_ANY_ALIGNMENT	(1U << 1)
1090 
1091 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1092  * Verifier does sub-register def/use analysis and identifies instructions whose
1093  * def only matters for low 32-bit, high 32-bit is never referenced later
1094  * through implicit zero extension. Therefore verifier notifies JIT back-ends
1095  * that it is safe to ignore clearing high 32-bit for these instructions. This
1096  * saves some back-ends a lot of code-gen. However such optimization is not
1097  * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1098  * hence hasn't used verifier's analysis result. But, we really want to have a
1099  * way to be able to verify the correctness of the described optimization on
1100  * x86_64 on which testsuites are frequently exercised.
1101  *
1102  * So, this flag is introduced. Once it is set, verifier will randomize high
1103  * 32-bit for those instructions who has been identified as safe to ignore them.
1104  * Then, if verifier is not doing correct analysis, such randomization will
1105  * regress tests to expose bugs.
1106  */
1107 #define BPF_F_TEST_RND_HI32	(1U << 2)
1108 
1109 /* The verifier internal test flag. Behavior is undefined */
1110 #define BPF_F_TEST_STATE_FREQ	(1U << 3)
1111 
1112 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1113  * restrict map and helper usage for such programs. Sleepable BPF programs can
1114  * only be attached to hooks where kernel execution context allows sleeping.
1115  * Such programs are allowed to use helpers that may sleep like
1116  * bpf_copy_from_user().
1117  */
1118 #define BPF_F_SLEEPABLE		(1U << 4)
1119 
1120 /* If BPF_F_XDP_HAS_FRAGS is used in BPF_PROG_LOAD command, the loaded program
1121  * fully support xdp frags.
1122  */
1123 #define BPF_F_XDP_HAS_FRAGS	(1U << 5)
1124 
1125 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
1126  * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
1127  */
1128 #define BPF_F_KPROBE_MULTI_RETURN	(1U << 0)
1129 
1130 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1131  * the following extensions:
1132  *
1133  * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
1134  * insn[0].imm:      map fd or fd_idx
1135  * insn[1].imm:      0
1136  * insn[0].off:      0
1137  * insn[1].off:      0
1138  * ldimm64 rewrite:  address of map
1139  * verifier type:    CONST_PTR_TO_MAP
1140  */
1141 #define BPF_PSEUDO_MAP_FD	1
1142 #define BPF_PSEUDO_MAP_IDX	5
1143 
1144 /* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
1145  * insn[0].imm:      map fd or fd_idx
1146  * insn[1].imm:      offset into value
1147  * insn[0].off:      0
1148  * insn[1].off:      0
1149  * ldimm64 rewrite:  address of map[0]+offset
1150  * verifier type:    PTR_TO_MAP_VALUE
1151  */
1152 #define BPF_PSEUDO_MAP_VALUE		2
1153 #define BPF_PSEUDO_MAP_IDX_VALUE	6
1154 
1155 /* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
1156  * insn[0].imm:      kernel btd id of VAR
1157  * insn[1].imm:      0
1158  * insn[0].off:      0
1159  * insn[1].off:      0
1160  * ldimm64 rewrite:  address of the kernel variable
1161  * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1162  *                   is struct/union.
1163  */
1164 #define BPF_PSEUDO_BTF_ID	3
1165 /* insn[0].src_reg:  BPF_PSEUDO_FUNC
1166  * insn[0].imm:      insn offset to the func
1167  * insn[1].imm:      0
1168  * insn[0].off:      0
1169  * insn[1].off:      0
1170  * ldimm64 rewrite:  address of the function
1171  * verifier type:    PTR_TO_FUNC.
1172  */
1173 #define BPF_PSEUDO_FUNC		4
1174 
1175 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1176  * offset to another bpf function
1177  */
1178 #define BPF_PSEUDO_CALL		1
1179 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1180  * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1181  */
1182 #define BPF_PSEUDO_KFUNC_CALL	2
1183 
1184 /* flags for BPF_MAP_UPDATE_ELEM command */
1185 enum {
1186 	BPF_ANY		= 0, /* create new element or update existing */
1187 	BPF_NOEXIST	= 1, /* create new element if it didn't exist */
1188 	BPF_EXIST	= 2, /* update existing element */
1189 	BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */
1190 };
1191 
1192 /* flags for BPF_MAP_CREATE command */
1193 enum {
1194 	BPF_F_NO_PREALLOC	= (1U << 0),
1195 /* Instead of having one common LRU list in the
1196  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1197  * which can scale and perform better.
1198  * Note, the LRU nodes (including free nodes) cannot be moved
1199  * across different LRU lists.
1200  */
1201 	BPF_F_NO_COMMON_LRU	= (1U << 1),
1202 /* Specify numa node during map creation */
1203 	BPF_F_NUMA_NODE		= (1U << 2),
1204 
1205 /* Flags for accessing BPF object from syscall side. */
1206 	BPF_F_RDONLY		= (1U << 3),
1207 	BPF_F_WRONLY		= (1U << 4),
1208 
1209 /* Flag for stack_map, store build_id+offset instead of pointer */
1210 	BPF_F_STACK_BUILD_ID	= (1U << 5),
1211 
1212 /* Zero-initialize hash function seed. This should only be used for testing. */
1213 	BPF_F_ZERO_SEED		= (1U << 6),
1214 
1215 /* Flags for accessing BPF object from program side. */
1216 	BPF_F_RDONLY_PROG	= (1U << 7),
1217 	BPF_F_WRONLY_PROG	= (1U << 8),
1218 
1219 /* Clone map from listener for newly accepted socket */
1220 	BPF_F_CLONE		= (1U << 9),
1221 
1222 /* Enable memory-mapping BPF map */
1223 	BPF_F_MMAPABLE		= (1U << 10),
1224 
1225 /* Share perf_event among processes */
1226 	BPF_F_PRESERVE_ELEMS	= (1U << 11),
1227 
1228 /* Create a map that is suitable to be an inner map with dynamic max entries */
1229 	BPF_F_INNER_MAP		= (1U << 12),
1230 };
1231 
1232 /* Flags for BPF_PROG_QUERY. */
1233 
1234 /* Query effective (directly attached + inherited from ancestor cgroups)
1235  * programs that will be executed for events within a cgroup.
1236  * attach_flags with this flag are returned only for directly attached programs.
1237  */
1238 #define BPF_F_QUERY_EFFECTIVE	(1U << 0)
1239 
1240 /* Flags for BPF_PROG_TEST_RUN */
1241 
1242 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1243 #define BPF_F_TEST_RUN_ON_CPU	(1U << 0)
1244 /* If set, XDP frames will be transmitted after processing */
1245 #define BPF_F_TEST_XDP_LIVE_FRAMES	(1U << 1)
1246 
1247 /* type for BPF_ENABLE_STATS */
1248 enum bpf_stats_type {
1249 	/* enabled run_time_ns and run_cnt */
1250 	BPF_STATS_RUN_TIME = 0,
1251 };
1252 
1253 enum bpf_stack_build_id_status {
1254 	/* user space need an empty entry to identify end of a trace */
1255 	BPF_STACK_BUILD_ID_EMPTY = 0,
1256 	/* with valid build_id and offset */
1257 	BPF_STACK_BUILD_ID_VALID = 1,
1258 	/* couldn't get build_id, fallback to ip */
1259 	BPF_STACK_BUILD_ID_IP = 2,
1260 };
1261 
1262 #define BPF_BUILD_ID_SIZE 20
1263 struct bpf_stack_build_id {
1264 	__s32		status;
1265 	unsigned char	build_id[BPF_BUILD_ID_SIZE];
1266 	union {
1267 		__u64	offset;
1268 		__u64	ip;
1269 	};
1270 };
1271 
1272 #define BPF_OBJ_NAME_LEN 16U
1273 
1274 union bpf_attr {
1275 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
1276 		__u32	map_type;	/* one of enum bpf_map_type */
1277 		__u32	key_size;	/* size of key in bytes */
1278 		__u32	value_size;	/* size of value in bytes */
1279 		__u32	max_entries;	/* max number of entries in a map */
1280 		__u32	map_flags;	/* BPF_MAP_CREATE related
1281 					 * flags defined above.
1282 					 */
1283 		__u32	inner_map_fd;	/* fd pointing to the inner map */
1284 		__u32	numa_node;	/* numa node (effective only if
1285 					 * BPF_F_NUMA_NODE is set).
1286 					 */
1287 		char	map_name[BPF_OBJ_NAME_LEN];
1288 		__u32	map_ifindex;	/* ifindex of netdev to create on */
1289 		__u32	btf_fd;		/* fd pointing to a BTF type data */
1290 		__u32	btf_key_type_id;	/* BTF type_id of the key */
1291 		__u32	btf_value_type_id;	/* BTF type_id of the value */
1292 		__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1293 						   * struct stored as the
1294 						   * map value
1295 						   */
1296 		/* Any per-map-type extra fields
1297 		 *
1298 		 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1299 		 * number of hash functions (if 0, the bloom filter will default
1300 		 * to using 5 hash functions).
1301 		 */
1302 		__u64	map_extra;
1303 	};
1304 
1305 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1306 		__u32		map_fd;
1307 		__aligned_u64	key;
1308 		union {
1309 			__aligned_u64 value;
1310 			__aligned_u64 next_key;
1311 		};
1312 		__u64		flags;
1313 	};
1314 
1315 	struct { /* struct used by BPF_MAP_*_BATCH commands */
1316 		__aligned_u64	in_batch;	/* start batch,
1317 						 * NULL to start from beginning
1318 						 */
1319 		__aligned_u64	out_batch;	/* output: next start batch */
1320 		__aligned_u64	keys;
1321 		__aligned_u64	values;
1322 		__u32		count;		/* input/output:
1323 						 * input: # of key/value
1324 						 * elements
1325 						 * output: # of filled elements
1326 						 */
1327 		__u32		map_fd;
1328 		__u64		elem_flags;
1329 		__u64		flags;
1330 	} batch;
1331 
1332 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
1333 		__u32		prog_type;	/* one of enum bpf_prog_type */
1334 		__u32		insn_cnt;
1335 		__aligned_u64	insns;
1336 		__aligned_u64	license;
1337 		__u32		log_level;	/* verbosity level of verifier */
1338 		__u32		log_size;	/* size of user buffer */
1339 		__aligned_u64	log_buf;	/* user supplied buffer */
1340 		__u32		kern_version;	/* not used */
1341 		__u32		prog_flags;
1342 		char		prog_name[BPF_OBJ_NAME_LEN];
1343 		__u32		prog_ifindex;	/* ifindex of netdev to prep for */
1344 		/* For some prog types expected attach type must be known at
1345 		 * load time to verify attach type specific parts of prog
1346 		 * (context accesses, allowed helpers, etc).
1347 		 */
1348 		__u32		expected_attach_type;
1349 		__u32		prog_btf_fd;	/* fd pointing to BTF type data */
1350 		__u32		func_info_rec_size;	/* userspace bpf_func_info size */
1351 		__aligned_u64	func_info;	/* func info */
1352 		__u32		func_info_cnt;	/* number of bpf_func_info records */
1353 		__u32		line_info_rec_size;	/* userspace bpf_line_info size */
1354 		__aligned_u64	line_info;	/* line info */
1355 		__u32		line_info_cnt;	/* number of bpf_line_info records */
1356 		__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */
1357 		union {
1358 			/* valid prog_fd to attach to bpf prog */
1359 			__u32		attach_prog_fd;
1360 			/* or valid module BTF object fd or 0 to attach to vmlinux */
1361 			__u32		attach_btf_obj_fd;
1362 		};
1363 		__u32		core_relo_cnt;	/* number of bpf_core_relo */
1364 		__aligned_u64	fd_array;	/* array of FDs */
1365 		__aligned_u64	core_relos;
1366 		__u32		core_relo_rec_size; /* sizeof(struct bpf_core_relo) */
1367 	};
1368 
1369 	struct { /* anonymous struct used by BPF_OBJ_* commands */
1370 		__aligned_u64	pathname;
1371 		__u32		bpf_fd;
1372 		__u32		file_flags;
1373 	};
1374 
1375 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1376 		__u32		target_fd;	/* container object to attach to */
1377 		__u32		attach_bpf_fd;	/* eBPF program to attach */
1378 		__u32		attach_type;
1379 		__u32		attach_flags;
1380 		__u32		replace_bpf_fd;	/* previously attached eBPF
1381 						 * program to replace if
1382 						 * BPF_F_REPLACE is used
1383 						 */
1384 	};
1385 
1386 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1387 		__u32		prog_fd;
1388 		__u32		retval;
1389 		__u32		data_size_in;	/* input: len of data_in */
1390 		__u32		data_size_out;	/* input/output: len of data_out
1391 						 *   returns ENOSPC if data_out
1392 						 *   is too small.
1393 						 */
1394 		__aligned_u64	data_in;
1395 		__aligned_u64	data_out;
1396 		__u32		repeat;
1397 		__u32		duration;
1398 		__u32		ctx_size_in;	/* input: len of ctx_in */
1399 		__u32		ctx_size_out;	/* input/output: len of ctx_out
1400 						 *   returns ENOSPC if ctx_out
1401 						 *   is too small.
1402 						 */
1403 		__aligned_u64	ctx_in;
1404 		__aligned_u64	ctx_out;
1405 		__u32		flags;
1406 		__u32		cpu;
1407 		__u32		batch_size;
1408 	} test;
1409 
1410 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
1411 		union {
1412 			__u32		start_id;
1413 			__u32		prog_id;
1414 			__u32		map_id;
1415 			__u32		btf_id;
1416 			__u32		link_id;
1417 		};
1418 		__u32		next_id;
1419 		__u32		open_flags;
1420 	};
1421 
1422 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1423 		__u32		bpf_fd;
1424 		__u32		info_len;
1425 		__aligned_u64	info;
1426 	} info;
1427 
1428 	struct { /* anonymous struct used by BPF_PROG_QUERY command */
1429 		__u32		target_fd;	/* container object to query */
1430 		__u32		attach_type;
1431 		__u32		query_flags;
1432 		__u32		attach_flags;
1433 		__aligned_u64	prog_ids;
1434 		__u32		prog_cnt;
1435 		__aligned_u64	prog_attach_flags; /* output: per-program attach_flags */
1436 	} query;
1437 
1438 	struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1439 		__u64 name;
1440 		__u32 prog_fd;
1441 	} raw_tracepoint;
1442 
1443 	struct { /* anonymous struct for BPF_BTF_LOAD */
1444 		__aligned_u64	btf;
1445 		__aligned_u64	btf_log_buf;
1446 		__u32		btf_size;
1447 		__u32		btf_log_size;
1448 		__u32		btf_log_level;
1449 	};
1450 
1451 	struct {
1452 		__u32		pid;		/* input: pid */
1453 		__u32		fd;		/* input: fd */
1454 		__u32		flags;		/* input: flags */
1455 		__u32		buf_len;	/* input/output: buf len */
1456 		__aligned_u64	buf;		/* input/output:
1457 						 *   tp_name for tracepoint
1458 						 *   symbol for kprobe
1459 						 *   filename for uprobe
1460 						 */
1461 		__u32		prog_id;	/* output: prod_id */
1462 		__u32		fd_type;	/* output: BPF_FD_TYPE_* */
1463 		__u64		probe_offset;	/* output: probe_offset */
1464 		__u64		probe_addr;	/* output: probe_addr */
1465 	} task_fd_query;
1466 
1467 	struct { /* struct used by BPF_LINK_CREATE command */
1468 		__u32		prog_fd;	/* eBPF program to attach */
1469 		union {
1470 			__u32		target_fd;	/* object to attach to */
1471 			__u32		target_ifindex; /* target ifindex */
1472 		};
1473 		__u32		attach_type;	/* attach type */
1474 		__u32		flags;		/* extra flags */
1475 		union {
1476 			__u32		target_btf_id;	/* btf_id of target to attach to */
1477 			struct {
1478 				__aligned_u64	iter_info;	/* extra bpf_iter_link_info */
1479 				__u32		iter_info_len;	/* iter_info length */
1480 			};
1481 			struct {
1482 				/* black box user-provided value passed through
1483 				 * to BPF program at the execution time and
1484 				 * accessible through bpf_get_attach_cookie() BPF helper
1485 				 */
1486 				__u64		bpf_cookie;
1487 			} perf_event;
1488 			struct {
1489 				__u32		flags;
1490 				__u32		cnt;
1491 				__aligned_u64	syms;
1492 				__aligned_u64	addrs;
1493 				__aligned_u64	cookies;
1494 			} kprobe_multi;
1495 			struct {
1496 				/* this is overlaid with the target_btf_id above. */
1497 				__u32		target_btf_id;
1498 				/* black box user-provided value passed through
1499 				 * to BPF program at the execution time and
1500 				 * accessible through bpf_get_attach_cookie() BPF helper
1501 				 */
1502 				__u64		cookie;
1503 			} tracing;
1504 		};
1505 	} link_create;
1506 
1507 	struct { /* struct used by BPF_LINK_UPDATE command */
1508 		__u32		link_fd;	/* link fd */
1509 		/* new program fd to update link with */
1510 		__u32		new_prog_fd;
1511 		__u32		flags;		/* extra flags */
1512 		/* expected link's program fd; is specified only if
1513 		 * BPF_F_REPLACE flag is set in flags */
1514 		__u32		old_prog_fd;
1515 	} link_update;
1516 
1517 	struct {
1518 		__u32		link_fd;
1519 	} link_detach;
1520 
1521 	struct { /* struct used by BPF_ENABLE_STATS command */
1522 		__u32		type;
1523 	} enable_stats;
1524 
1525 	struct { /* struct used by BPF_ITER_CREATE command */
1526 		__u32		link_fd;
1527 		__u32		flags;
1528 	} iter_create;
1529 
1530 	struct { /* struct used by BPF_PROG_BIND_MAP command */
1531 		__u32		prog_fd;
1532 		__u32		map_fd;
1533 		__u32		flags;		/* extra flags */
1534 	} prog_bind_map;
1535 
1536 } __attribute__((aligned(8)));
1537 
1538 /* The description below is an attempt at providing documentation to eBPF
1539  * developers about the multiple available eBPF helper functions. It can be
1540  * parsed and used to produce a manual page. The workflow is the following,
1541  * and requires the rst2man utility:
1542  *
1543  *     $ ./scripts/bpf_doc.py \
1544  *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1545  *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1546  *     $ man /tmp/bpf-helpers.7
1547  *
1548  * Note that in order to produce this external documentation, some RST
1549  * formatting is used in the descriptions to get "bold" and "italics" in
1550  * manual pages. Also note that the few trailing white spaces are
1551  * intentional, removing them would break paragraphs for rst2man.
1552  *
1553  * Start of BPF helper function descriptions:
1554  *
1555  * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1556  * 	Description
1557  * 		Perform a lookup in *map* for an entry associated to *key*.
1558  * 	Return
1559  * 		Map value associated to *key*, or **NULL** if no entry was
1560  * 		found.
1561  *
1562  * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1563  * 	Description
1564  * 		Add or update the value of the entry associated to *key* in
1565  * 		*map* with *value*. *flags* is one of:
1566  *
1567  * 		**BPF_NOEXIST**
1568  * 			The entry for *key* must not exist in the map.
1569  * 		**BPF_EXIST**
1570  * 			The entry for *key* must already exist in the map.
1571  * 		**BPF_ANY**
1572  * 			No condition on the existence of the entry for *key*.
1573  *
1574  * 		Flag value **BPF_NOEXIST** cannot be used for maps of types
1575  * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
1576  * 		elements always exist), the helper would return an error.
1577  * 	Return
1578  * 		0 on success, or a negative error in case of failure.
1579  *
1580  * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1581  * 	Description
1582  * 		Delete entry with *key* from *map*.
1583  * 	Return
1584  * 		0 on success, or a negative error in case of failure.
1585  *
1586  * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1587  * 	Description
1588  * 		For tracing programs, safely attempt to read *size* bytes from
1589  * 		kernel space address *unsafe_ptr* and store the data in *dst*.
1590  *
1591  * 		Generally, use **bpf_probe_read_user**\ () or
1592  * 		**bpf_probe_read_kernel**\ () instead.
1593  * 	Return
1594  * 		0 on success, or a negative error in case of failure.
1595  *
1596  * u64 bpf_ktime_get_ns(void)
1597  * 	Description
1598  * 		Return the time elapsed since system boot, in nanoseconds.
1599  * 		Does not include time the system was suspended.
1600  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1601  * 	Return
1602  * 		Current *ktime*.
1603  *
1604  * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1605  * 	Description
1606  * 		This helper is a "printk()-like" facility for debugging. It
1607  * 		prints a message defined by format *fmt* (of size *fmt_size*)
1608  * 		to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1609  * 		available. It can take up to three additional **u64**
1610  * 		arguments (as an eBPF helpers, the total number of arguments is
1611  * 		limited to five).
1612  *
1613  * 		Each time the helper is called, it appends a line to the trace.
1614  * 		Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1615  * 		open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1616  * 		The format of the trace is customizable, and the exact output
1617  * 		one will get depends on the options set in
1618  * 		*\/sys/kernel/debug/tracing/trace_options* (see also the
1619  * 		*README* file under the same directory). However, it usually
1620  * 		defaults to something like:
1621  *
1622  * 		::
1623  *
1624  * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1625  *
1626  * 		In the above:
1627  *
1628  * 			* ``telnet`` is the name of the current task.
1629  * 			* ``470`` is the PID of the current task.
1630  * 			* ``001`` is the CPU number on which the task is
1631  * 			  running.
1632  * 			* In ``.N..``, each character refers to a set of
1633  * 			  options (whether irqs are enabled, scheduling
1634  * 			  options, whether hard/softirqs are running, level of
1635  * 			  preempt_disabled respectively). **N** means that
1636  * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1637  * 			  are set.
1638  * 			* ``419421.045894`` is a timestamp.
1639  * 			* ``0x00000001`` is a fake value used by BPF for the
1640  * 			  instruction pointer register.
1641  * 			* ``<formatted msg>`` is the message formatted with
1642  * 			  *fmt*.
1643  *
1644  * 		The conversion specifiers supported by *fmt* are similar, but
1645  * 		more limited than for printk(). They are **%d**, **%i**,
1646  * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1647  * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1648  * 		of field, padding with zeroes, etc.) is available, and the
1649  * 		helper will return **-EINVAL** (but print nothing) if it
1650  * 		encounters an unknown specifier.
1651  *
1652  * 		Also, note that **bpf_trace_printk**\ () is slow, and should
1653  * 		only be used for debugging purposes. For this reason, a notice
1654  * 		block (spanning several lines) is printed to kernel logs and
1655  * 		states that the helper should not be used "for production use"
1656  * 		the first time this helper is used (or more precisely, when
1657  * 		**trace_printk**\ () buffers are allocated). For passing values
1658  * 		to user space, perf events should be preferred.
1659  * 	Return
1660  * 		The number of bytes written to the buffer, or a negative error
1661  * 		in case of failure.
1662  *
1663  * u32 bpf_get_prandom_u32(void)
1664  * 	Description
1665  * 		Get a pseudo-random number.
1666  *
1667  * 		From a security point of view, this helper uses its own
1668  * 		pseudo-random internal state, and cannot be used to infer the
1669  * 		seed of other random functions in the kernel. However, it is
1670  * 		essential to note that the generator used by the helper is not
1671  * 		cryptographically secure.
1672  * 	Return
1673  * 		A random 32-bit unsigned value.
1674  *
1675  * u32 bpf_get_smp_processor_id(void)
1676  * 	Description
1677  * 		Get the SMP (symmetric multiprocessing) processor id. Note that
1678  * 		all programs run with migration disabled, which means that the
1679  * 		SMP processor id is stable during all the execution of the
1680  * 		program.
1681  * 	Return
1682  * 		The SMP id of the processor running the program.
1683  *
1684  * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1685  * 	Description
1686  * 		Store *len* bytes from address *from* into the packet
1687  * 		associated to *skb*, at *offset*. *flags* are a combination of
1688  * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1689  * 		checksum for the packet after storing the bytes) and
1690  * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1691  * 		**->swhash** and *skb*\ **->l4hash** to 0).
1692  *
1693  * 		A call to this helper is susceptible to change the underlying
1694  * 		packet buffer. Therefore, at load time, all checks on pointers
1695  * 		previously done by the verifier are invalidated and must be
1696  * 		performed again, if the helper is used in combination with
1697  * 		direct packet access.
1698  * 	Return
1699  * 		0 on success, or a negative error in case of failure.
1700  *
1701  * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1702  * 	Description
1703  * 		Recompute the layer 3 (e.g. IP) checksum for the packet
1704  * 		associated to *skb*. Computation is incremental, so the helper
1705  * 		must know the former value of the header field that was
1706  * 		modified (*from*), the new value of this field (*to*), and the
1707  * 		number of bytes (2 or 4) for this field, stored in *size*.
1708  * 		Alternatively, it is possible to store the difference between
1709  * 		the previous and the new values of the header field in *to*, by
1710  * 		setting *from* and *size* to 0. For both methods, *offset*
1711  * 		indicates the location of the IP checksum within the packet.
1712  *
1713  * 		This helper works in combination with **bpf_csum_diff**\ (),
1714  * 		which does not update the checksum in-place, but offers more
1715  * 		flexibility and can handle sizes larger than 2 or 4 for the
1716  * 		checksum to update.
1717  *
1718  * 		A call to this helper is susceptible to change the underlying
1719  * 		packet buffer. Therefore, at load time, all checks on pointers
1720  * 		previously done by the verifier are invalidated and must be
1721  * 		performed again, if the helper is used in combination with
1722  * 		direct packet access.
1723  * 	Return
1724  * 		0 on success, or a negative error in case of failure.
1725  *
1726  * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1727  * 	Description
1728  * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1729  * 		packet associated to *skb*. Computation is incremental, so the
1730  * 		helper must know the former value of the header field that was
1731  * 		modified (*from*), the new value of this field (*to*), and the
1732  * 		number of bytes (2 or 4) for this field, stored on the lowest
1733  * 		four bits of *flags*. Alternatively, it is possible to store
1734  * 		the difference between the previous and the new values of the
1735  * 		header field in *to*, by setting *from* and the four lowest
1736  * 		bits of *flags* to 0. For both methods, *offset* indicates the
1737  * 		location of the IP checksum within the packet. In addition to
1738  * 		the size of the field, *flags* can be added (bitwise OR) actual
1739  * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1740  * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1741  * 		for updates resulting in a null checksum the value is set to
1742  * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1743  * 		the checksum is to be computed against a pseudo-header.
1744  *
1745  * 		This helper works in combination with **bpf_csum_diff**\ (),
1746  * 		which does not update the checksum in-place, but offers more
1747  * 		flexibility and can handle sizes larger than 2 or 4 for the
1748  * 		checksum to update.
1749  *
1750  * 		A call to this helper is susceptible to change the underlying
1751  * 		packet buffer. Therefore, at load time, all checks on pointers
1752  * 		previously done by the verifier are invalidated and must be
1753  * 		performed again, if the helper is used in combination with
1754  * 		direct packet access.
1755  * 	Return
1756  * 		0 on success, or a negative error in case of failure.
1757  *
1758  * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1759  * 	Description
1760  * 		This special helper is used to trigger a "tail call", or in
1761  * 		other words, to jump into another eBPF program. The same stack
1762  * 		frame is used (but values on stack and in registers for the
1763  * 		caller are not accessible to the callee). This mechanism allows
1764  * 		for program chaining, either for raising the maximum number of
1765  * 		available eBPF instructions, or to execute given programs in
1766  * 		conditional blocks. For security reasons, there is an upper
1767  * 		limit to the number of successive tail calls that can be
1768  * 		performed.
1769  *
1770  * 		Upon call of this helper, the program attempts to jump into a
1771  * 		program referenced at index *index* in *prog_array_map*, a
1772  * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1773  * 		*ctx*, a pointer to the context.
1774  *
1775  * 		If the call succeeds, the kernel immediately runs the first
1776  * 		instruction of the new program. This is not a function call,
1777  * 		and it never returns to the previous program. If the call
1778  * 		fails, then the helper has no effect, and the caller continues
1779  * 		to run its subsequent instructions. A call can fail if the
1780  * 		destination program for the jump does not exist (i.e. *index*
1781  * 		is superior to the number of entries in *prog_array_map*), or
1782  * 		if the maximum number of tail calls has been reached for this
1783  * 		chain of programs. This limit is defined in the kernel by the
1784  * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1785  *		which is currently set to 33.
1786  * 	Return
1787  * 		0 on success, or a negative error in case of failure.
1788  *
1789  * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1790  * 	Description
1791  * 		Clone and redirect the packet associated to *skb* to another
1792  * 		net device of index *ifindex*. Both ingress and egress
1793  * 		interfaces can be used for redirection. The **BPF_F_INGRESS**
1794  * 		value in *flags* is used to make the distinction (ingress path
1795  * 		is selected if the flag is present, egress path otherwise).
1796  * 		This is the only flag supported for now.
1797  *
1798  * 		In comparison with **bpf_redirect**\ () helper,
1799  * 		**bpf_clone_redirect**\ () has the associated cost of
1800  * 		duplicating the packet buffer, but this can be executed out of
1801  * 		the eBPF program. Conversely, **bpf_redirect**\ () is more
1802  * 		efficient, but it is handled through an action code where the
1803  * 		redirection happens only after the eBPF program has returned.
1804  *
1805  * 		A call to this helper is susceptible to change the underlying
1806  * 		packet buffer. Therefore, at load time, all checks on pointers
1807  * 		previously done by the verifier are invalidated and must be
1808  * 		performed again, if the helper is used in combination with
1809  * 		direct packet access.
1810  * 	Return
1811  * 		0 on success, or a negative error in case of failure.
1812  *
1813  * u64 bpf_get_current_pid_tgid(void)
1814  * 	Description
1815  * 		Get the current pid and tgid.
1816  * 	Return
1817  * 		A 64-bit integer containing the current tgid and pid, and
1818  * 		created as such:
1819  * 		*current_task*\ **->tgid << 32 \|**
1820  * 		*current_task*\ **->pid**.
1821  *
1822  * u64 bpf_get_current_uid_gid(void)
1823  * 	Description
1824  * 		Get the current uid and gid.
1825  * 	Return
1826  * 		A 64-bit integer containing the current GID and UID, and
1827  * 		created as such: *current_gid* **<< 32 \|** *current_uid*.
1828  *
1829  * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1830  * 	Description
1831  * 		Copy the **comm** attribute of the current task into *buf* of
1832  * 		*size_of_buf*. The **comm** attribute contains the name of
1833  * 		the executable (excluding the path) for the current task. The
1834  * 		*size_of_buf* must be strictly positive. On success, the
1835  * 		helper makes sure that the *buf* is NUL-terminated. On failure,
1836  * 		it is filled with zeroes.
1837  * 	Return
1838  * 		0 on success, or a negative error in case of failure.
1839  *
1840  * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1841  * 	Description
1842  * 		Retrieve the classid for the current task, i.e. for the net_cls
1843  * 		cgroup to which *skb* belongs.
1844  *
1845  * 		This helper can be used on TC egress path, but not on ingress.
1846  *
1847  * 		The net_cls cgroup provides an interface to tag network packets
1848  * 		based on a user-provided identifier for all traffic coming from
1849  * 		the tasks belonging to the related cgroup. See also the related
1850  * 		kernel documentation, available from the Linux sources in file
1851  * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1852  *
1853  * 		The Linux kernel has two versions for cgroups: there are
1854  * 		cgroups v1 and cgroups v2. Both are available to users, who can
1855  * 		use a mixture of them, but note that the net_cls cgroup is for
1856  * 		cgroup v1 only. This makes it incompatible with BPF programs
1857  * 		run on cgroups, which is a cgroup-v2-only feature (a socket can
1858  * 		only hold data for one version of cgroups at a time).
1859  *
1860  * 		This helper is only available is the kernel was compiled with
1861  * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1862  * 		"**y**" or to "**m**".
1863  * 	Return
1864  * 		The classid, or 0 for the default unconfigured classid.
1865  *
1866  * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1867  * 	Description
1868  * 		Push a *vlan_tci* (VLAN tag control information) of protocol
1869  * 		*vlan_proto* to the packet associated to *skb*, then update
1870  * 		the checksum. Note that if *vlan_proto* is different from
1871  * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1872  * 		be **ETH_P_8021Q**.
1873  *
1874  * 		A call to this helper is susceptible to change the underlying
1875  * 		packet buffer. Therefore, at load time, all checks on pointers
1876  * 		previously done by the verifier are invalidated and must be
1877  * 		performed again, if the helper is used in combination with
1878  * 		direct packet access.
1879  * 	Return
1880  * 		0 on success, or a negative error in case of failure.
1881  *
1882  * long bpf_skb_vlan_pop(struct sk_buff *skb)
1883  * 	Description
1884  * 		Pop a VLAN header from the packet associated to *skb*.
1885  *
1886  * 		A call to this helper is susceptible to change the underlying
1887  * 		packet buffer. Therefore, at load time, all checks on pointers
1888  * 		previously done by the verifier are invalidated and must be
1889  * 		performed again, if the helper is used in combination with
1890  * 		direct packet access.
1891  * 	Return
1892  * 		0 on success, or a negative error in case of failure.
1893  *
1894  * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1895  * 	Description
1896  * 		Get tunnel metadata. This helper takes a pointer *key* to an
1897  * 		empty **struct bpf_tunnel_key** of **size**, that will be
1898  * 		filled with tunnel metadata for the packet associated to *skb*.
1899  * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1900  * 		indicates that the tunnel is based on IPv6 protocol instead of
1901  * 		IPv4.
1902  *
1903  * 		The **struct bpf_tunnel_key** is an object that generalizes the
1904  * 		principal parameters used by various tunneling protocols into a
1905  * 		single struct. This way, it can be used to easily make a
1906  * 		decision based on the contents of the encapsulation header,
1907  * 		"summarized" in this struct. In particular, it holds the IP
1908  * 		address of the remote end (IPv4 or IPv6, depending on the case)
1909  * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1910  * 		this struct exposes the *key*\ **->tunnel_id**, which is
1911  * 		generally mapped to a VNI (Virtual Network Identifier), making
1912  * 		it programmable together with the **bpf_skb_set_tunnel_key**\
1913  * 		() helper.
1914  *
1915  * 		Let's imagine that the following code is part of a program
1916  * 		attached to the TC ingress interface, on one end of a GRE
1917  * 		tunnel, and is supposed to filter out all messages coming from
1918  * 		remote ends with IPv4 address other than 10.0.0.1:
1919  *
1920  * 		::
1921  *
1922  * 			int ret;
1923  * 			struct bpf_tunnel_key key = {};
1924  *
1925  * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1926  * 			if (ret < 0)
1927  * 				return TC_ACT_SHOT;	// drop packet
1928  *
1929  * 			if (key.remote_ipv4 != 0x0a000001)
1930  * 				return TC_ACT_SHOT;	// drop packet
1931  *
1932  * 			return TC_ACT_OK;		// accept packet
1933  *
1934  * 		This interface can also be used with all encapsulation devices
1935  * 		that can operate in "collect metadata" mode: instead of having
1936  * 		one network device per specific configuration, the "collect
1937  * 		metadata" mode only requires a single device where the
1938  * 		configuration can be extracted from this helper.
1939  *
1940  * 		This can be used together with various tunnels such as VXLan,
1941  * 		Geneve, GRE or IP in IP (IPIP).
1942  * 	Return
1943  * 		0 on success, or a negative error in case of failure.
1944  *
1945  * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1946  * 	Description
1947  * 		Populate tunnel metadata for packet associated to *skb.* The
1948  * 		tunnel metadata is set to the contents of *key*, of *size*. The
1949  * 		*flags* can be set to a combination of the following values:
1950  *
1951  * 		**BPF_F_TUNINFO_IPV6**
1952  * 			Indicate that the tunnel is based on IPv6 protocol
1953  * 			instead of IPv4.
1954  * 		**BPF_F_ZERO_CSUM_TX**
1955  * 			For IPv4 packets, add a flag to tunnel metadata
1956  * 			indicating that checksum computation should be skipped
1957  * 			and checksum set to zeroes.
1958  * 		**BPF_F_DONT_FRAGMENT**
1959  * 			Add a flag to tunnel metadata indicating that the
1960  * 			packet should not be fragmented.
1961  * 		**BPF_F_SEQ_NUMBER**
1962  * 			Add a flag to tunnel metadata indicating that a
1963  * 			sequence number should be added to tunnel header before
1964  * 			sending the packet. This flag was added for GRE
1965  * 			encapsulation, but might be used with other protocols
1966  * 			as well in the future.
1967  *
1968  * 		Here is a typical usage on the transmit path:
1969  *
1970  * 		::
1971  *
1972  * 			struct bpf_tunnel_key key;
1973  * 			     populate key ...
1974  * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1975  * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1976  *
1977  * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()
1978  * 		helper for additional information.
1979  * 	Return
1980  * 		0 on success, or a negative error in case of failure.
1981  *
1982  * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1983  * 	Description
1984  * 		Read the value of a perf event counter. This helper relies on a
1985  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1986  * 		the perf event counter is selected when *map* is updated with
1987  * 		perf event file descriptors. The *map* is an array whose size
1988  * 		is the number of available CPUs, and each cell contains a value
1989  * 		relative to one CPU. The value to retrieve is indicated by
1990  * 		*flags*, that contains the index of the CPU to look up, masked
1991  * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1992  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
1993  * 		current CPU should be retrieved.
1994  *
1995  * 		Note that before Linux 4.13, only hardware perf event can be
1996  * 		retrieved.
1997  *
1998  * 		Also, be aware that the newer helper
1999  * 		**bpf_perf_event_read_value**\ () is recommended over
2000  * 		**bpf_perf_event_read**\ () in general. The latter has some ABI
2001  * 		quirks where error and counter value are used as a return code
2002  * 		(which is wrong to do since ranges may overlap). This issue is
2003  * 		fixed with **bpf_perf_event_read_value**\ (), which at the same
2004  * 		time provides more features over the **bpf_perf_event_read**\
2005  * 		() interface. Please refer to the description of
2006  * 		**bpf_perf_event_read_value**\ () for details.
2007  * 	Return
2008  * 		The value of the perf event counter read from the map, or a
2009  * 		negative error code in case of failure.
2010  *
2011  * long bpf_redirect(u32 ifindex, u64 flags)
2012  * 	Description
2013  * 		Redirect the packet to another net device of index *ifindex*.
2014  * 		This helper is somewhat similar to **bpf_clone_redirect**\
2015  * 		(), except that the packet is not cloned, which provides
2016  * 		increased performance.
2017  *
2018  * 		Except for XDP, both ingress and egress interfaces can be used
2019  * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used
2020  * 		to make the distinction (ingress path is selected if the flag
2021  * 		is present, egress path otherwise). Currently, XDP only
2022  * 		supports redirection to the egress interface, and accepts no
2023  * 		flag at all.
2024  *
2025  * 		The same effect can also be attained with the more generic
2026  * 		**bpf_redirect_map**\ (), which uses a BPF map to store the
2027  * 		redirect target instead of providing it directly to the helper.
2028  * 	Return
2029  * 		For XDP, the helper returns **XDP_REDIRECT** on success or
2030  * 		**XDP_ABORTED** on error. For other program types, the values
2031  * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
2032  * 		error.
2033  *
2034  * u32 bpf_get_route_realm(struct sk_buff *skb)
2035  * 	Description
2036  * 		Retrieve the realm or the route, that is to say the
2037  * 		**tclassid** field of the destination for the *skb*. The
2038  * 		identifier retrieved is a user-provided tag, similar to the
2039  * 		one used with the net_cls cgroup (see description for
2040  * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is
2041  * 		held by a route (a destination entry), not by a task.
2042  *
2043  * 		Retrieving this identifier works with the clsact TC egress hook
2044  * 		(see also **tc-bpf(8)**), or alternatively on conventional
2045  * 		classful egress qdiscs, but not on TC ingress path. In case of
2046  * 		clsact TC egress hook, this has the advantage that, internally,
2047  * 		the destination entry has not been dropped yet in the transmit
2048  * 		path. Therefore, the destination entry does not need to be
2049  * 		artificially held via **netif_keep_dst**\ () for a classful
2050  * 		qdisc until the *skb* is freed.
2051  *
2052  * 		This helper is available only if the kernel was compiled with
2053  * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.
2054  * 	Return
2055  * 		The realm of the route for the packet associated to *skb*, or 0
2056  * 		if none was found.
2057  *
2058  * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2059  * 	Description
2060  * 		Write raw *data* blob into a special BPF perf event held by
2061  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2062  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
2063  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2064  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2065  *
2066  * 		The *flags* are used to indicate the index in *map* for which
2067  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
2068  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2069  * 		to indicate that the index of the current CPU core should be
2070  * 		used.
2071  *
2072  * 		The value to write, of *size*, is passed through eBPF stack and
2073  * 		pointed by *data*.
2074  *
2075  * 		The context of the program *ctx* needs also be passed to the
2076  * 		helper.
2077  *
2078  * 		On user space, a program willing to read the values needs to
2079  * 		call **perf_event_open**\ () on the perf event (either for
2080  * 		one or for all CPUs) and to store the file descriptor into the
2081  * 		*map*. This must be done before the eBPF program can send data
2082  * 		into it. An example is available in file
2083  * 		*samples/bpf/trace_output_user.c* in the Linux kernel source
2084  * 		tree (the eBPF program counterpart is in
2085  * 		*samples/bpf/trace_output_kern.c*).
2086  *
2087  * 		**bpf_perf_event_output**\ () achieves better performance
2088  * 		than **bpf_trace_printk**\ () for sharing data with user
2089  * 		space, and is much better suitable for streaming data from eBPF
2090  * 		programs.
2091  *
2092  * 		Note that this helper is not restricted to tracing use cases
2093  * 		and can be used with programs attached to TC or XDP as well,
2094  * 		where it allows for passing data to user space listeners. Data
2095  * 		can be:
2096  *
2097  * 		* Only custom structs,
2098  * 		* Only the packet payload, or
2099  * 		* A combination of both.
2100  * 	Return
2101  * 		0 on success, or a negative error in case of failure.
2102  *
2103  * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2104  * 	Description
2105  * 		This helper was provided as an easy way to load data from a
2106  * 		packet. It can be used to load *len* bytes from *offset* from
2107  * 		the packet associated to *skb*, into the buffer pointed by
2108  * 		*to*.
2109  *
2110  * 		Since Linux 4.7, usage of this helper has mostly been replaced
2111  * 		by "direct packet access", enabling packet data to be
2112  * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2113  * 		pointing respectively to the first byte of packet data and to
2114  * 		the byte after the last byte of packet data. However, it
2115  * 		remains useful if one wishes to read large quantities of data
2116  * 		at once from a packet into the eBPF stack.
2117  * 	Return
2118  * 		0 on success, or a negative error in case of failure.
2119  *
2120  * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2121  * 	Description
2122  * 		Walk a user or a kernel stack and return its id. To achieve
2123  * 		this, the helper needs *ctx*, which is a pointer to the context
2124  * 		on which the tracing program is executed, and a pointer to a
2125  * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2126  *
2127  * 		The last argument, *flags*, holds the number of stack frames to
2128  * 		skip (from 0 to 255), masked with
2129  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2130  * 		a combination of the following flags:
2131  *
2132  * 		**BPF_F_USER_STACK**
2133  * 			Collect a user space stack instead of a kernel stack.
2134  * 		**BPF_F_FAST_STACK_CMP**
2135  * 			Compare stacks by hash only.
2136  * 		**BPF_F_REUSE_STACKID**
2137  * 			If two different stacks hash into the same *stackid*,
2138  * 			discard the old one.
2139  *
2140  * 		The stack id retrieved is a 32 bit long integer handle which
2141  * 		can be further combined with other data (including other stack
2142  * 		ids) and used as a key into maps. This can be useful for
2143  * 		generating a variety of graphs (such as flame graphs or off-cpu
2144  * 		graphs).
2145  *
2146  * 		For walking a stack, this helper is an improvement over
2147  * 		**bpf_probe_read**\ (), which can be used with unrolled loops
2148  * 		but is not efficient and consumes a lot of eBPF instructions.
2149  * 		Instead, **bpf_get_stackid**\ () can collect up to
2150  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2151  * 		this limit can be controlled with the **sysctl** program, and
2152  * 		that it should be manually increased in order to profile long
2153  * 		user stacks (such as stacks for Java programs). To do so, use:
2154  *
2155  * 		::
2156  *
2157  * 			# sysctl kernel.perf_event_max_stack=<new value>
2158  * 	Return
2159  * 		The positive or null stack id on success, or a negative error
2160  * 		in case of failure.
2161  *
2162  * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2163  * 	Description
2164  * 		Compute a checksum difference, from the raw buffer pointed by
2165  * 		*from*, of length *from_size* (that must be a multiple of 4),
2166  * 		towards the raw buffer pointed by *to*, of size *to_size*
2167  * 		(same remark). An optional *seed* can be added to the value
2168  * 		(this can be cascaded, the seed may come from a previous call
2169  * 		to the helper).
2170  *
2171  * 		This is flexible enough to be used in several ways:
2172  *
2173  * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to
2174  * 		  checksum, it can be used when pushing new data.
2175  * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to
2176  * 		  checksum, it can be used when removing data from a packet.
2177  * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2178  * 		  can be used to compute a diff. Note that *from_size* and
2179  * 		  *to_size* do not need to be equal.
2180  *
2181  * 		This helper can be used in combination with
2182  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2183  * 		which one can feed in the difference computed with
2184  * 		**bpf_csum_diff**\ ().
2185  * 	Return
2186  * 		The checksum result, or a negative error code in case of
2187  * 		failure.
2188  *
2189  * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2190  * 	Description
2191  * 		Retrieve tunnel options metadata for the packet associated to
2192  * 		*skb*, and store the raw tunnel option data to the buffer *opt*
2193  * 		of *size*.
2194  *
2195  * 		This helper can be used with encapsulation devices that can
2196  * 		operate in "collect metadata" mode (please refer to the related
2197  * 		note in the description of **bpf_skb_get_tunnel_key**\ () for
2198  * 		more details). A particular example where this can be used is
2199  * 		in combination with the Geneve encapsulation protocol, where it
2200  * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2201  * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from
2202  * 		the eBPF program. This allows for full customization of these
2203  * 		headers.
2204  * 	Return
2205  * 		The size of the option data retrieved.
2206  *
2207  * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2208  * 	Description
2209  * 		Set tunnel options metadata for the packet associated to *skb*
2210  * 		to the option data contained in the raw buffer *opt* of *size*.
2211  *
2212  * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2213  * 		helper for additional information.
2214  * 	Return
2215  * 		0 on success, or a negative error in case of failure.
2216  *
2217  * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2218  * 	Description
2219  * 		Change the protocol of the *skb* to *proto*. Currently
2220  * 		supported are transition from IPv4 to IPv6, and from IPv6 to
2221  * 		IPv4. The helper takes care of the groundwork for the
2222  * 		transition, including resizing the socket buffer. The eBPF
2223  * 		program is expected to fill the new headers, if any, via
2224  * 		**skb_store_bytes**\ () and to recompute the checksums with
2225  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2226  * 		(). The main case for this helper is to perform NAT64
2227  * 		operations out of an eBPF program.
2228  *
2229  * 		Internally, the GSO type is marked as dodgy so that headers are
2230  * 		checked and segments are recalculated by the GSO/GRO engine.
2231  * 		The size for GSO target is adapted as well.
2232  *
2233  * 		All values for *flags* are reserved for future usage, and must
2234  * 		be left at zero.
2235  *
2236  * 		A call to this helper is susceptible to change the underlying
2237  * 		packet buffer. Therefore, at load time, all checks on pointers
2238  * 		previously done by the verifier are invalidated and must be
2239  * 		performed again, if the helper is used in combination with
2240  * 		direct packet access.
2241  * 	Return
2242  * 		0 on success, or a negative error in case of failure.
2243  *
2244  * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2245  * 	Description
2246  * 		Change the packet type for the packet associated to *skb*. This
2247  * 		comes down to setting *skb*\ **->pkt_type** to *type*, except
2248  * 		the eBPF program does not have a write access to *skb*\
2249  * 		**->pkt_type** beside this helper. Using a helper here allows
2250  * 		for graceful handling of errors.
2251  *
2252  * 		The major use case is to change incoming *skb*s to
2253  * 		**PACKET_HOST** in a programmatic way instead of having to
2254  * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2255  * 		example.
2256  *
2257  * 		Note that *type* only allows certain values. At this time, they
2258  * 		are:
2259  *
2260  * 		**PACKET_HOST**
2261  * 			Packet is for us.
2262  * 		**PACKET_BROADCAST**
2263  * 			Send packet to all.
2264  * 		**PACKET_MULTICAST**
2265  * 			Send packet to group.
2266  * 		**PACKET_OTHERHOST**
2267  * 			Send packet to someone else.
2268  * 	Return
2269  * 		0 on success, or a negative error in case of failure.
2270  *
2271  * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2272  * 	Description
2273  * 		Check whether *skb* is a descendant of the cgroup2 held by
2274  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2275  * 	Return
2276  * 		The return value depends on the result of the test, and can be:
2277  *
2278  * 		* 0, if the *skb* failed the cgroup2 descendant test.
2279  * 		* 1, if the *skb* succeeded the cgroup2 descendant test.
2280  * 		* A negative error code, if an error occurred.
2281  *
2282  * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2283  * 	Description
2284  * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2285  * 		not set, in particular if the hash was cleared due to mangling,
2286  * 		recompute this hash. Later accesses to the hash can be done
2287  * 		directly with *skb*\ **->hash**.
2288  *
2289  * 		Calling **bpf_set_hash_invalid**\ (), changing a packet
2290  * 		prototype with **bpf_skb_change_proto**\ (), or calling
2291  * 		**bpf_skb_store_bytes**\ () with the
2292  * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2293  * 		the hash and to trigger a new computation for the next call to
2294  * 		**bpf_get_hash_recalc**\ ().
2295  * 	Return
2296  * 		The 32-bit hash.
2297  *
2298  * u64 bpf_get_current_task(void)
2299  * 	Description
2300  * 		Get the current task.
2301  * 	Return
2302  * 		A pointer to the current task struct.
2303  *
2304  * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2305  * 	Description
2306  * 		Attempt in a safe way to write *len* bytes from the buffer
2307  * 		*src* to *dst* in memory. It only works for threads that are in
2308  * 		user context, and *dst* must be a valid user space address.
2309  *
2310  * 		This helper should not be used to implement any kind of
2311  * 		security mechanism because of TOC-TOU attacks, but rather to
2312  * 		debug, divert, and manipulate execution of semi-cooperative
2313  * 		processes.
2314  *
2315  * 		Keep in mind that this feature is meant for experiments, and it
2316  * 		has a risk of crashing the system and running programs.
2317  * 		Therefore, when an eBPF program using this helper is attached,
2318  * 		a warning including PID and process name is printed to kernel
2319  * 		logs.
2320  * 	Return
2321  * 		0 on success, or a negative error in case of failure.
2322  *
2323  * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2324  * 	Description
2325  * 		Check whether the probe is being run is the context of a given
2326  * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2327  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2328  * 	Return
2329  * 		The return value depends on the result of the test, and can be:
2330  *
2331  *		* 1, if current task belongs to the cgroup2.
2332  *		* 0, if current task does not belong to the cgroup2.
2333  * 		* A negative error code, if an error occurred.
2334  *
2335  * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2336  * 	Description
2337  * 		Resize (trim or grow) the packet associated to *skb* to the
2338  * 		new *len*. The *flags* are reserved for future usage, and must
2339  * 		be left at zero.
2340  *
2341  * 		The basic idea is that the helper performs the needed work to
2342  * 		change the size of the packet, then the eBPF program rewrites
2343  * 		the rest via helpers like **bpf_skb_store_bytes**\ (),
2344  * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2345  * 		and others. This helper is a slow path utility intended for
2346  * 		replies with control messages. And because it is targeted for
2347  * 		slow path, the helper itself can afford to be slow: it
2348  * 		implicitly linearizes, unclones and drops offloads from the
2349  * 		*skb*.
2350  *
2351  * 		A call to this helper is susceptible to change the underlying
2352  * 		packet buffer. Therefore, at load time, all checks on pointers
2353  * 		previously done by the verifier are invalidated and must be
2354  * 		performed again, if the helper is used in combination with
2355  * 		direct packet access.
2356  * 	Return
2357  * 		0 on success, or a negative error in case of failure.
2358  *
2359  * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2360  * 	Description
2361  * 		Pull in non-linear data in case the *skb* is non-linear and not
2362  * 		all of *len* are part of the linear section. Make *len* bytes
2363  * 		from *skb* readable and writable. If a zero value is passed for
2364  *		*len*, then all bytes in the linear part of *skb* will be made
2365  *		readable and writable.
2366  *
2367  * 		This helper is only needed for reading and writing with direct
2368  * 		packet access.
2369  *
2370  * 		For direct packet access, testing that offsets to access
2371  * 		are within packet boundaries (test on *skb*\ **->data_end**) is
2372  * 		susceptible to fail if offsets are invalid, or if the requested
2373  * 		data is in non-linear parts of the *skb*. On failure the
2374  * 		program can just bail out, or in the case of a non-linear
2375  * 		buffer, use a helper to make the data available. The
2376  * 		**bpf_skb_load_bytes**\ () helper is a first solution to access
2377  * 		the data. Another one consists in using **bpf_skb_pull_data**
2378  * 		to pull in once the non-linear parts, then retesting and
2379  * 		eventually access the data.
2380  *
2381  * 		At the same time, this also makes sure the *skb* is uncloned,
2382  * 		which is a necessary condition for direct write. As this needs
2383  * 		to be an invariant for the write part only, the verifier
2384  * 		detects writes and adds a prologue that is calling
2385  * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from
2386  * 		the very beginning in case it is indeed cloned.
2387  *
2388  * 		A call to this helper is susceptible to change the underlying
2389  * 		packet buffer. Therefore, at load time, all checks on pointers
2390  * 		previously done by the verifier are invalidated and must be
2391  * 		performed again, if the helper is used in combination with
2392  * 		direct packet access.
2393  * 	Return
2394  * 		0 on success, or a negative error in case of failure.
2395  *
2396  * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2397  * 	Description
2398  * 		Add the checksum *csum* into *skb*\ **->csum** in case the
2399  * 		driver has supplied a checksum for the entire packet into that
2400  * 		field. Return an error otherwise. This helper is intended to be
2401  * 		used in combination with **bpf_csum_diff**\ (), in particular
2402  * 		when the checksum needs to be updated after data has been
2403  * 		written into the packet through direct packet access.
2404  * 	Return
2405  * 		The checksum on success, or a negative error code in case of
2406  * 		failure.
2407  *
2408  * void bpf_set_hash_invalid(struct sk_buff *skb)
2409  * 	Description
2410  * 		Invalidate the current *skb*\ **->hash**. It can be used after
2411  * 		mangling on headers through direct packet access, in order to
2412  * 		indicate that the hash is outdated and to trigger a
2413  * 		recalculation the next time the kernel tries to access this
2414  * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.
2415  * 	Return
2416  * 		void.
2417  *
2418  * long bpf_get_numa_node_id(void)
2419  * 	Description
2420  * 		Return the id of the current NUMA node. The primary use case
2421  * 		for this helper is the selection of sockets for the local NUMA
2422  * 		node, when the program is attached to sockets using the
2423  * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2424  * 		but the helper is also available to other eBPF program types,
2425  * 		similarly to **bpf_get_smp_processor_id**\ ().
2426  * 	Return
2427  * 		The id of current NUMA node.
2428  *
2429  * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2430  * 	Description
2431  * 		Grows headroom of packet associated to *skb* and adjusts the
2432  * 		offset of the MAC header accordingly, adding *len* bytes of
2433  * 		space. It automatically extends and reallocates memory as
2434  * 		required.
2435  *
2436  * 		This helper can be used on a layer 3 *skb* to push a MAC header
2437  * 		for redirection into a layer 2 device.
2438  *
2439  * 		All values for *flags* are reserved for future usage, and must
2440  * 		be left at zero.
2441  *
2442  * 		A call to this helper is susceptible to change the underlying
2443  * 		packet buffer. Therefore, at load time, all checks on pointers
2444  * 		previously done by the verifier are invalidated and must be
2445  * 		performed again, if the helper is used in combination with
2446  * 		direct packet access.
2447  * 	Return
2448  * 		0 on success, or a negative error in case of failure.
2449  *
2450  * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2451  * 	Description
2452  * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2453  * 		it is possible to use a negative value for *delta*. This helper
2454  * 		can be used to prepare the packet for pushing or popping
2455  * 		headers.
2456  *
2457  * 		A call to this helper is susceptible to change the underlying
2458  * 		packet buffer. Therefore, at load time, all checks on pointers
2459  * 		previously done by the verifier are invalidated and must be
2460  * 		performed again, if the helper is used in combination with
2461  * 		direct packet access.
2462  * 	Return
2463  * 		0 on success, or a negative error in case of failure.
2464  *
2465  * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2466  * 	Description
2467  * 		Copy a NUL terminated string from an unsafe kernel address
2468  * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2469  * 		more details.
2470  *
2471  * 		Generally, use **bpf_probe_read_user_str**\ () or
2472  * 		**bpf_probe_read_kernel_str**\ () instead.
2473  * 	Return
2474  * 		On success, the strictly positive length of the string,
2475  * 		including the trailing NUL character. On error, a negative
2476  * 		value.
2477  *
2478  * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2479  * 	Description
2480  * 		If the **struct sk_buff** pointed by *skb* has a known socket,
2481  * 		retrieve the cookie (generated by the kernel) of this socket.
2482  * 		If no cookie has been set yet, generate a new cookie. Once
2483  * 		generated, the socket cookie remains stable for the life of the
2484  * 		socket. This helper can be useful for monitoring per socket
2485  * 		networking traffic statistics as it provides a global socket
2486  * 		identifier that can be assumed unique.
2487  * 	Return
2488  * 		A 8-byte long unique number on success, or 0 if the socket
2489  * 		field is missing inside *skb*.
2490  *
2491  * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2492  * 	Description
2493  * 		Equivalent to bpf_get_socket_cookie() helper that accepts
2494  * 		*skb*, but gets socket from **struct bpf_sock_addr** context.
2495  * 	Return
2496  * 		A 8-byte long unique number.
2497  *
2498  * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2499  * 	Description
2500  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2501  * 		*skb*, but gets socket from **struct bpf_sock_ops** context.
2502  * 	Return
2503  * 		A 8-byte long unique number.
2504  *
2505  * u64 bpf_get_socket_cookie(struct sock *sk)
2506  * 	Description
2507  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2508  * 		*sk*, but gets socket from a BTF **struct sock**. This helper
2509  * 		also works for sleepable programs.
2510  * 	Return
2511  * 		A 8-byte long unique number or 0 if *sk* is NULL.
2512  *
2513  * u32 bpf_get_socket_uid(struct sk_buff *skb)
2514  * 	Description
2515  * 		Get the owner UID of the socked associated to *skb*.
2516  * 	Return
2517  * 		The owner UID of the socket associated to *skb*. If the socket
2518  * 		is **NULL**, or if it is not a full socket (i.e. if it is a
2519  * 		time-wait or a request socket instead), **overflowuid** value
2520  * 		is returned (note that **overflowuid** might also be the actual
2521  * 		UID value for the socket).
2522  *
2523  * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2524  * 	Description
2525  * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)
2526  * 		to value *hash*.
2527  * 	Return
2528  * 		0
2529  *
2530  * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2531  * 	Description
2532  * 		Emulate a call to **setsockopt()** on the socket associated to
2533  * 		*bpf_socket*, which must be a full socket. The *level* at
2534  * 		which the option resides and the name *optname* of the option
2535  * 		must be specified, see **setsockopt(2)** for more information.
2536  * 		The option value of length *optlen* is pointed by *optval*.
2537  *
2538  * 		*bpf_socket* should be one of the following:
2539  *
2540  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2541  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2542  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2543  *
2544  * 		This helper actually implements a subset of **setsockopt()**.
2545  * 		It supports the following *level*\ s:
2546  *
2547  * 		* **SOL_SOCKET**, which supports the following *optname*\ s:
2548  * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2549  * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2550  * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2551  * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:
2552  * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,
2553  * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2554  * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2555  *		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2556  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2557  * 		* **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2558  * 	Return
2559  * 		0 on success, or a negative error in case of failure.
2560  *
2561  * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2562  * 	Description
2563  * 		Grow or shrink the room for data in the packet associated to
2564  * 		*skb* by *len_diff*, and according to the selected *mode*.
2565  *
2566  * 		By default, the helper will reset any offloaded checksum
2567  * 		indicator of the skb to CHECKSUM_NONE. This can be avoided
2568  * 		by the following flag:
2569  *
2570  * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2571  * 		  checksum data of the skb to CHECKSUM_NONE.
2572  *
2573  *		There are two supported modes at this time:
2574  *
2575  *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2576  * 		  (room space is added or removed between the layer 2 and
2577  * 		  layer 3 headers).
2578  *
2579  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2580  * 		  (room space is added or removed between the layer 3 and
2581  * 		  layer 4 headers).
2582  *
2583  *		The following flags are supported at this time:
2584  *
2585  *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2586  *		  Adjusting mss in this way is not allowed for datagrams.
2587  *
2588  *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2589  *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2590  *		  Any new space is reserved to hold a tunnel header.
2591  *		  Configure skb offsets and other fields accordingly.
2592  *
2593  *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2594  *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2595  *		  Use with ENCAP_L3 flags to further specify the tunnel type.
2596  *
2597  *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2598  *		  Use with ENCAP_L3/L4 flags to further specify the tunnel
2599  *		  type; *len* is the length of the inner MAC header.
2600  *
2601  *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2602  *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2603  *		  L2 type as Ethernet.
2604  *
2605  * 		A call to this helper is susceptible to change the underlying
2606  * 		packet buffer. Therefore, at load time, all checks on pointers
2607  * 		previously done by the verifier are invalidated and must be
2608  * 		performed again, if the helper is used in combination with
2609  * 		direct packet access.
2610  * 	Return
2611  * 		0 on success, or a negative error in case of failure.
2612  *
2613  * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2614  * 	Description
2615  * 		Redirect the packet to the endpoint referenced by *map* at
2616  * 		index *key*. Depending on its type, this *map* can contain
2617  * 		references to net devices (for forwarding packets through other
2618  * 		ports), or to CPUs (for redirecting XDP frames to another CPU;
2619  * 		but this is only implemented for native XDP (with driver
2620  * 		support) as of this writing).
2621  *
2622  * 		The lower two bits of *flags* are used as the return code if
2623  * 		the map lookup fails. This is so that the return value can be
2624  * 		one of the XDP program return codes up to **XDP_TX**, as chosen
2625  * 		by the caller. The higher bits of *flags* can be set to
2626  * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2627  *
2628  * 		With BPF_F_BROADCAST the packet will be broadcasted to all the
2629  * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2630  * 		interface will be excluded when do broadcasting.
2631  *
2632  * 		See also **bpf_redirect**\ (), which only supports redirecting
2633  * 		to an ifindex, but doesn't require a map to do so.
2634  * 	Return
2635  * 		**XDP_REDIRECT** on success, or the value of the two lower bits
2636  * 		of the *flags* argument on error.
2637  *
2638  * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2639  * 	Description
2640  * 		Redirect the packet to the socket referenced by *map* (of type
2641  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2642  * 		egress interfaces can be used for redirection. The
2643  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2644  * 		distinction (ingress path is selected if the flag is present,
2645  * 		egress path otherwise). This is the only flag supported for now.
2646  * 	Return
2647  * 		**SK_PASS** on success, or **SK_DROP** on error.
2648  *
2649  * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2650  * 	Description
2651  * 		Add an entry to, or update a *map* referencing sockets. The
2652  * 		*skops* is used as a new value for the entry associated to
2653  * 		*key*. *flags* is one of:
2654  *
2655  * 		**BPF_NOEXIST**
2656  * 			The entry for *key* must not exist in the map.
2657  * 		**BPF_EXIST**
2658  * 			The entry for *key* must already exist in the map.
2659  * 		**BPF_ANY**
2660  * 			No condition on the existence of the entry for *key*.
2661  *
2662  * 		If the *map* has eBPF programs (parser and verdict), those will
2663  * 		be inherited by the socket being added. If the socket is
2664  * 		already attached to eBPF programs, this results in an error.
2665  * 	Return
2666  * 		0 on success, or a negative error in case of failure.
2667  *
2668  * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2669  * 	Description
2670  * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by
2671  * 		*delta* (which can be positive or negative). Note that this
2672  * 		operation modifies the address stored in *xdp_md*\ **->data**,
2673  * 		so the latter must be loaded only after the helper has been
2674  * 		called.
2675  *
2676  * 		The use of *xdp_md*\ **->data_meta** is optional and programs
2677  * 		are not required to use it. The rationale is that when the
2678  * 		packet is processed with XDP (e.g. as DoS filter), it is
2679  * 		possible to push further meta data along with it before passing
2680  * 		to the stack, and to give the guarantee that an ingress eBPF
2681  * 		program attached as a TC classifier on the same device can pick
2682  * 		this up for further post-processing. Since TC works with socket
2683  * 		buffers, it remains possible to set from XDP the **mark** or
2684  * 		**priority** pointers, or other pointers for the socket buffer.
2685  * 		Having this scratch space generic and programmable allows for
2686  * 		more flexibility as the user is free to store whatever meta
2687  * 		data they need.
2688  *
2689  * 		A call to this helper is susceptible to change the underlying
2690  * 		packet buffer. Therefore, at load time, all checks on pointers
2691  * 		previously done by the verifier are invalidated and must be
2692  * 		performed again, if the helper is used in combination with
2693  * 		direct packet access.
2694  * 	Return
2695  * 		0 on success, or a negative error in case of failure.
2696  *
2697  * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2698  * 	Description
2699  * 		Read the value of a perf event counter, and store it into *buf*
2700  * 		of size *buf_size*. This helper relies on a *map* of type
2701  * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2702  * 		counter is selected when *map* is updated with perf event file
2703  * 		descriptors. The *map* is an array whose size is the number of
2704  * 		available CPUs, and each cell contains a value relative to one
2705  * 		CPU. The value to retrieve is indicated by *flags*, that
2706  * 		contains the index of the CPU to look up, masked with
2707  * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2708  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2709  * 		current CPU should be retrieved.
2710  *
2711  * 		This helper behaves in a way close to
2712  * 		**bpf_perf_event_read**\ () helper, save that instead of
2713  * 		just returning the value observed, it fills the *buf*
2714  * 		structure. This allows for additional data to be retrieved: in
2715  * 		particular, the enabled and running times (in *buf*\
2716  * 		**->enabled** and *buf*\ **->running**, respectively) are
2717  * 		copied. In general, **bpf_perf_event_read_value**\ () is
2718  * 		recommended over **bpf_perf_event_read**\ (), which has some
2719  * 		ABI issues and provides fewer functionalities.
2720  *
2721  * 		These values are interesting, because hardware PMU (Performance
2722  * 		Monitoring Unit) counters are limited resources. When there are
2723  * 		more PMU based perf events opened than available counters,
2724  * 		kernel will multiplex these events so each event gets certain
2725  * 		percentage (but not all) of the PMU time. In case that
2726  * 		multiplexing happens, the number of samples or counter value
2727  * 		will not reflect the case compared to when no multiplexing
2728  * 		occurs. This makes comparison between different runs difficult.
2729  * 		Typically, the counter value should be normalized before
2730  * 		comparing to other experiments. The usual normalization is done
2731  * 		as follows.
2732  *
2733  * 		::
2734  *
2735  * 			normalized_counter = counter * t_enabled / t_running
2736  *
2737  * 		Where t_enabled is the time enabled for event and t_running is
2738  * 		the time running for event since last normalization. The
2739  * 		enabled and running times are accumulated since the perf event
2740  * 		open. To achieve scaling factor between two invocations of an
2741  * 		eBPF program, users can use CPU id as the key (which is
2742  * 		typical for perf array usage model) to remember the previous
2743  * 		value and do the calculation inside the eBPF program.
2744  * 	Return
2745  * 		0 on success, or a negative error in case of failure.
2746  *
2747  * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2748  * 	Description
2749  * 		For en eBPF program attached to a perf event, retrieve the
2750  * 		value of the event counter associated to *ctx* and store it in
2751  * 		the structure pointed by *buf* and of size *buf_size*. Enabled
2752  * 		and running times are also stored in the structure (see
2753  * 		description of helper **bpf_perf_event_read_value**\ () for
2754  * 		more details).
2755  * 	Return
2756  * 		0 on success, or a negative error in case of failure.
2757  *
2758  * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2759  * 	Description
2760  * 		Emulate a call to **getsockopt()** on the socket associated to
2761  * 		*bpf_socket*, which must be a full socket. The *level* at
2762  * 		which the option resides and the name *optname* of the option
2763  * 		must be specified, see **getsockopt(2)** for more information.
2764  * 		The retrieved value is stored in the structure pointed by
2765  * 		*opval* and of length *optlen*.
2766  *
2767  * 		*bpf_socket* should be one of the following:
2768  *
2769  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2770  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2771  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2772  *
2773  * 		This helper actually implements a subset of **getsockopt()**.
2774  * 		It supports the following *level*\ s:
2775  *
2776  * 		* **IPPROTO_TCP**, which supports *optname*
2777  * 		  **TCP_CONGESTION**.
2778  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2779  * 		* **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2780  * 	Return
2781  * 		0 on success, or a negative error in case of failure.
2782  *
2783  * long bpf_override_return(struct pt_regs *regs, u64 rc)
2784  * 	Description
2785  * 		Used for error injection, this helper uses kprobes to override
2786  * 		the return value of the probed function, and to set it to *rc*.
2787  * 		The first argument is the context *regs* on which the kprobe
2788  * 		works.
2789  *
2790  * 		This helper works by setting the PC (program counter)
2791  * 		to an override function which is run in place of the original
2792  * 		probed function. This means the probed function is not run at
2793  * 		all. The replacement function just returns with the required
2794  * 		value.
2795  *
2796  * 		This helper has security implications, and thus is subject to
2797  * 		restrictions. It is only available if the kernel was compiled
2798  * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2799  * 		option, and in this case it only works on functions tagged with
2800  * 		**ALLOW_ERROR_INJECTION** in the kernel code.
2801  *
2802  * 		Also, the helper is only available for the architectures having
2803  * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2804  * 		x86 architecture is the only one to support this feature.
2805  * 	Return
2806  * 		0
2807  *
2808  * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2809  * 	Description
2810  * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2811  * 		for the full TCP socket associated to *bpf_sock_ops* to
2812  * 		*argval*.
2813  *
2814  * 		The primary use of this field is to determine if there should
2815  * 		be calls to eBPF programs of type
2816  * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2817  * 		code. A program of the same type can change its value, per
2818  * 		connection and as necessary, when the connection is
2819  * 		established. This field is directly accessible for reading, but
2820  * 		this helper must be used for updates in order to return an
2821  * 		error if an eBPF program tries to set a callback that is not
2822  * 		supported in the current kernel.
2823  *
2824  * 		*argval* is a flag array which can combine these flags:
2825  *
2826  * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2827  * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2828  * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2829  * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2830  *
2831  * 		Therefore, this function can be used to clear a callback flag by
2832  * 		setting the appropriate bit to zero. e.g. to disable the RTO
2833  * 		callback:
2834  *
2835  * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**
2836  * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2837  *
2838  * 		Here are some examples of where one could call such eBPF
2839  * 		program:
2840  *
2841  * 		* When RTO fires.
2842  * 		* When a packet is retransmitted.
2843  * 		* When the connection terminates.
2844  * 		* When a packet is sent.
2845  * 		* When a packet is received.
2846  * 	Return
2847  * 		Code **-EINVAL** if the socket is not a full TCP socket;
2848  * 		otherwise, a positive number containing the bits that could not
2849  * 		be set is returned (which comes down to 0 if all bits were set
2850  * 		as required).
2851  *
2852  * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2853  * 	Description
2854  * 		This helper is used in programs implementing policies at the
2855  * 		socket level. If the message *msg* is allowed to pass (i.e. if
2856  * 		the verdict eBPF program returns **SK_PASS**), redirect it to
2857  * 		the socket referenced by *map* (of type
2858  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2859  * 		egress interfaces can be used for redirection. The
2860  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2861  * 		distinction (ingress path is selected if the flag is present,
2862  * 		egress path otherwise). This is the only flag supported for now.
2863  * 	Return
2864  * 		**SK_PASS** on success, or **SK_DROP** on error.
2865  *
2866  * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2867  * 	Description
2868  * 		For socket policies, apply the verdict of the eBPF program to
2869  * 		the next *bytes* (number of bytes) of message *msg*.
2870  *
2871  * 		For example, this helper can be used in the following cases:
2872  *
2873  * 		* A single **sendmsg**\ () or **sendfile**\ () system call
2874  * 		  contains multiple logical messages that the eBPF program is
2875  * 		  supposed to read and for which it should apply a verdict.
2876  * 		* An eBPF program only cares to read the first *bytes* of a
2877  * 		  *msg*. If the message has a large payload, then setting up
2878  * 		  and calling the eBPF program repeatedly for all bytes, even
2879  * 		  though the verdict is already known, would create unnecessary
2880  * 		  overhead.
2881  *
2882  * 		When called from within an eBPF program, the helper sets a
2883  * 		counter internal to the BPF infrastructure, that is used to
2884  * 		apply the last verdict to the next *bytes*. If *bytes* is
2885  * 		smaller than the current data being processed from a
2886  * 		**sendmsg**\ () or **sendfile**\ () system call, the first
2887  * 		*bytes* will be sent and the eBPF program will be re-run with
2888  * 		the pointer for start of data pointing to byte number *bytes*
2889  * 		**+ 1**. If *bytes* is larger than the current data being
2890  * 		processed, then the eBPF verdict will be applied to multiple
2891  * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2892  * 		consumed.
2893  *
2894  * 		Note that if a socket closes with the internal counter holding
2895  * 		a non-zero value, this is not a problem because data is not
2896  * 		being buffered for *bytes* and is sent as it is received.
2897  * 	Return
2898  * 		0
2899  *
2900  * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2901  * 	Description
2902  * 		For socket policies, prevent the execution of the verdict eBPF
2903  * 		program for message *msg* until *bytes* (byte number) have been
2904  * 		accumulated.
2905  *
2906  * 		This can be used when one needs a specific number of bytes
2907  * 		before a verdict can be assigned, even if the data spans
2908  * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2909  * 		case would be a user calling **sendmsg**\ () repeatedly with
2910  * 		1-byte long message segments. Obviously, this is bad for
2911  * 		performance, but it is still valid. If the eBPF program needs
2912  * 		*bytes* bytes to validate a header, this helper can be used to
2913  * 		prevent the eBPF program to be called again until *bytes* have
2914  * 		been accumulated.
2915  * 	Return
2916  * 		0
2917  *
2918  * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2919  * 	Description
2920  * 		For socket policies, pull in non-linear data from user space
2921  * 		for *msg* and set pointers *msg*\ **->data** and *msg*\
2922  * 		**->data_end** to *start* and *end* bytes offsets into *msg*,
2923  * 		respectively.
2924  *
2925  * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2926  * 		*msg* it can only parse data that the (**data**, **data_end**)
2927  * 		pointers have already consumed. For **sendmsg**\ () hooks this
2928  * 		is likely the first scatterlist element. But for calls relying
2929  * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2930  * 		be the range (**0**, **0**) because the data is shared with
2931  * 		user space and by default the objective is to avoid allowing
2932  * 		user space to modify data while (or after) eBPF verdict is
2933  * 		being decided. This helper can be used to pull in data and to
2934  * 		set the start and end pointer to given values. Data will be
2935  * 		copied if necessary (i.e. if data was not linear and if start
2936  * 		and end pointers do not point to the same chunk).
2937  *
2938  * 		A call to this helper is susceptible to change the underlying
2939  * 		packet buffer. Therefore, at load time, all checks on pointers
2940  * 		previously done by the verifier are invalidated and must be
2941  * 		performed again, if the helper is used in combination with
2942  * 		direct packet access.
2943  *
2944  * 		All values for *flags* are reserved for future usage, and must
2945  * 		be left at zero.
2946  * 	Return
2947  * 		0 on success, or a negative error in case of failure.
2948  *
2949  * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2950  * 	Description
2951  * 		Bind the socket associated to *ctx* to the address pointed by
2952  * 		*addr*, of length *addr_len*. This allows for making outgoing
2953  * 		connection from the desired IP address, which can be useful for
2954  * 		example when all processes inside a cgroup should use one
2955  * 		single IP address on a host that has multiple IP configured.
2956  *
2957  * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2958  * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2959  * 		**AF_INET6**). It's advised to pass zero port (**sin_port**
2960  * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2961  * 		behavior and lets the kernel efficiently pick up an unused
2962  * 		port as long as 4-tuple is unique. Passing non-zero port might
2963  * 		lead to degraded performance.
2964  * 	Return
2965  * 		0 on success, or a negative error in case of failure.
2966  *
2967  * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2968  * 	Description
2969  * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2970  * 		possible to both shrink and grow the packet tail.
2971  * 		Shrink done via *delta* being a negative integer.
2972  *
2973  * 		A call to this helper is susceptible to change the underlying
2974  * 		packet buffer. Therefore, at load time, all checks on pointers
2975  * 		previously done by the verifier are invalidated and must be
2976  * 		performed again, if the helper is used in combination with
2977  * 		direct packet access.
2978  * 	Return
2979  * 		0 on success, or a negative error in case of failure.
2980  *
2981  * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2982  * 	Description
2983  * 		Retrieve the XFRM state (IP transform framework, see also
2984  * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2985  *
2986  * 		The retrieved value is stored in the **struct bpf_xfrm_state**
2987  * 		pointed by *xfrm_state* and of length *size*.
2988  *
2989  * 		All values for *flags* are reserved for future usage, and must
2990  * 		be left at zero.
2991  *
2992  * 		This helper is available only if the kernel was compiled with
2993  * 		**CONFIG_XFRM** configuration option.
2994  * 	Return
2995  * 		0 on success, or a negative error in case of failure.
2996  *
2997  * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2998  * 	Description
2999  * 		Return a user or a kernel stack in bpf program provided buffer.
3000  * 		To achieve this, the helper needs *ctx*, which is a pointer
3001  * 		to the context on which the tracing program is executed.
3002  * 		To store the stacktrace, the bpf program provides *buf* with
3003  * 		a nonnegative *size*.
3004  *
3005  * 		The last argument, *flags*, holds the number of stack frames to
3006  * 		skip (from 0 to 255), masked with
3007  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
3008  * 		the following flags:
3009  *
3010  * 		**BPF_F_USER_STACK**
3011  * 			Collect a user space stack instead of a kernel stack.
3012  * 		**BPF_F_USER_BUILD_ID**
3013  * 			Collect (build_id, file_offset) instead of ips for user
3014  * 			stack, only valid if **BPF_F_USER_STACK** is also
3015  * 			specified.
3016  *
3017  * 			*file_offset* is an offset relative to the beginning
3018  * 			of the executable or shared object file backing the vma
3019  * 			which the *ip* falls in. It is *not* an offset relative
3020  * 			to that object's base address. Accordingly, it must be
3021  * 			adjusted by adding (sh_addr - sh_offset), where
3022  * 			sh_{addr,offset} correspond to the executable section
3023  * 			containing *file_offset* in the object, for comparisons
3024  * 			to symbols' st_value to be valid.
3025  *
3026  * 		**bpf_get_stack**\ () can collect up to
3027  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
3028  * 		to sufficient large buffer size. Note that
3029  * 		this limit can be controlled with the **sysctl** program, and
3030  * 		that it should be manually increased in order to profile long
3031  * 		user stacks (such as stacks for Java programs). To do so, use:
3032  *
3033  * 		::
3034  *
3035  * 			# sysctl kernel.perf_event_max_stack=<new value>
3036  * 	Return
3037  * 		The non-negative copied *buf* length equal to or less than
3038  * 		*size* on success, or a negative error in case of failure.
3039  *
3040  * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
3041  * 	Description
3042  * 		This helper is similar to **bpf_skb_load_bytes**\ () in that
3043  * 		it provides an easy way to load *len* bytes from *offset*
3044  * 		from the packet associated to *skb*, into the buffer pointed
3045  * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that
3046  * 		a fifth argument *start_header* exists in order to select a
3047  * 		base offset to start from. *start_header* can be one of:
3048  *
3049  * 		**BPF_HDR_START_MAC**
3050  * 			Base offset to load data from is *skb*'s mac header.
3051  * 		**BPF_HDR_START_NET**
3052  * 			Base offset to load data from is *skb*'s network header.
3053  *
3054  * 		In general, "direct packet access" is the preferred method to
3055  * 		access packet data, however, this helper is in particular useful
3056  * 		in socket filters where *skb*\ **->data** does not always point
3057  * 		to the start of the mac header and where "direct packet access"
3058  * 		is not available.
3059  * 	Return
3060  * 		0 on success, or a negative error in case of failure.
3061  *
3062  * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3063  *	Description
3064  *		Do FIB lookup in kernel tables using parameters in *params*.
3065  *		If lookup is successful and result shows packet is to be
3066  *		forwarded, the neighbor tables are searched for the nexthop.
3067  *		If successful (ie., FIB lookup shows forwarding and nexthop
3068  *		is resolved), the nexthop address is returned in ipv4_dst
3069  *		or ipv6_dst based on family, smac is set to mac address of
3070  *		egress device, dmac is set to nexthop mac address, rt_metric
3071  *		is set to metric from route (IPv4/IPv6 only), and ifindex
3072  *		is set to the device index of the nexthop from the FIB lookup.
3073  *
3074  *		*plen* argument is the size of the passed in struct.
3075  *		*flags* argument can be a combination of one or more of the
3076  *		following values:
3077  *
3078  *		**BPF_FIB_LOOKUP_DIRECT**
3079  *			Do a direct table lookup vs full lookup using FIB
3080  *			rules.
3081  *		**BPF_FIB_LOOKUP_OUTPUT**
3082  *			Perform lookup from an egress perspective (default is
3083  *			ingress).
3084  *
3085  *		*ctx* is either **struct xdp_md** for XDP programs or
3086  *		**struct sk_buff** tc cls_act programs.
3087  *	Return
3088  *		* < 0 if any input argument is invalid
3089  *		*   0 on success (packet is forwarded, nexthop neighbor exists)
3090  *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3091  *		  packet is not forwarded or needs assist from full stack
3092  *
3093  *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3094  *		was exceeded and output params->mtu_result contains the MTU.
3095  *
3096  * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3097  *	Description
3098  *		Add an entry to, or update a sockhash *map* referencing sockets.
3099  *		The *skops* is used as a new value for the entry associated to
3100  *		*key*. *flags* is one of:
3101  *
3102  *		**BPF_NOEXIST**
3103  *			The entry for *key* must not exist in the map.
3104  *		**BPF_EXIST**
3105  *			The entry for *key* must already exist in the map.
3106  *		**BPF_ANY**
3107  *			No condition on the existence of the entry for *key*.
3108  *
3109  *		If the *map* has eBPF programs (parser and verdict), those will
3110  *		be inherited by the socket being added. If the socket is
3111  *		already attached to eBPF programs, this results in an error.
3112  *	Return
3113  *		0 on success, or a negative error in case of failure.
3114  *
3115  * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3116  *	Description
3117  *		This helper is used in programs implementing policies at the
3118  *		socket level. If the message *msg* is allowed to pass (i.e. if
3119  *		the verdict eBPF program returns **SK_PASS**), redirect it to
3120  *		the socket referenced by *map* (of type
3121  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3122  *		egress interfaces can be used for redirection. The
3123  *		**BPF_F_INGRESS** value in *flags* is used to make the
3124  *		distinction (ingress path is selected if the flag is present,
3125  *		egress path otherwise). This is the only flag supported for now.
3126  *	Return
3127  *		**SK_PASS** on success, or **SK_DROP** on error.
3128  *
3129  * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3130  *	Description
3131  *		This helper is used in programs implementing policies at the
3132  *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3133  *		if the verdict eBPF program returns **SK_PASS**), redirect it
3134  *		to the socket referenced by *map* (of type
3135  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3136  *		egress interfaces can be used for redirection. The
3137  *		**BPF_F_INGRESS** value in *flags* is used to make the
3138  *		distinction (ingress path is selected if the flag is present,
3139  *		egress otherwise). This is the only flag supported for now.
3140  *	Return
3141  *		**SK_PASS** on success, or **SK_DROP** on error.
3142  *
3143  * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3144  *	Description
3145  *		Encapsulate the packet associated to *skb* within a Layer 3
3146  *		protocol header. This header is provided in the buffer at
3147  *		address *hdr*, with *len* its size in bytes. *type* indicates
3148  *		the protocol of the header and can be one of:
3149  *
3150  *		**BPF_LWT_ENCAP_SEG6**
3151  *			IPv6 encapsulation with Segment Routing Header
3152  *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3153  *			the IPv6 header is computed by the kernel.
3154  *		**BPF_LWT_ENCAP_SEG6_INLINE**
3155  *			Only works if *skb* contains an IPv6 packet. Insert a
3156  *			Segment Routing Header (**struct ipv6_sr_hdr**) inside
3157  *			the IPv6 header.
3158  *		**BPF_LWT_ENCAP_IP**
3159  *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3160  *			must be IPv4 or IPv6, followed by zero or more
3161  *			additional headers, up to **LWT_BPF_MAX_HEADROOM**
3162  *			total bytes in all prepended headers. Please note that
3163  *			if **skb_is_gso**\ (*skb*) is true, no more than two
3164  *			headers can be prepended, and the inner header, if
3165  *			present, should be either GRE or UDP/GUE.
3166  *
3167  *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3168  *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3169  *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3170  *		**BPF_PROG_TYPE_LWT_XMIT**.
3171  *
3172  * 		A call to this helper is susceptible to change the underlying
3173  * 		packet buffer. Therefore, at load time, all checks on pointers
3174  * 		previously done by the verifier are invalidated and must be
3175  * 		performed again, if the helper is used in combination with
3176  * 		direct packet access.
3177  *	Return
3178  * 		0 on success, or a negative error in case of failure.
3179  *
3180  * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3181  *	Description
3182  *		Store *len* bytes from address *from* into the packet
3183  *		associated to *skb*, at *offset*. Only the flags, tag and TLVs
3184  *		inside the outermost IPv6 Segment Routing Header can be
3185  *		modified through this helper.
3186  *
3187  * 		A call to this helper is susceptible to change the underlying
3188  * 		packet buffer. Therefore, at load time, all checks on pointers
3189  * 		previously done by the verifier are invalidated and must be
3190  * 		performed again, if the helper is used in combination with
3191  * 		direct packet access.
3192  *	Return
3193  * 		0 on success, or a negative error in case of failure.
3194  *
3195  * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3196  *	Description
3197  *		Adjust the size allocated to TLVs in the outermost IPv6
3198  *		Segment Routing Header contained in the packet associated to
3199  *		*skb*, at position *offset* by *delta* bytes. Only offsets
3200  *		after the segments are accepted. *delta* can be as well
3201  *		positive (growing) as negative (shrinking).
3202  *
3203  * 		A call to this helper is susceptible to change the underlying
3204  * 		packet buffer. Therefore, at load time, all checks on pointers
3205  * 		previously done by the verifier are invalidated and must be
3206  * 		performed again, if the helper is used in combination with
3207  * 		direct packet access.
3208  *	Return
3209  * 		0 on success, or a negative error in case of failure.
3210  *
3211  * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3212  *	Description
3213  *		Apply an IPv6 Segment Routing action of type *action* to the
3214  *		packet associated to *skb*. Each action takes a parameter
3215  *		contained at address *param*, and of length *param_len* bytes.
3216  *		*action* can be one of:
3217  *
3218  *		**SEG6_LOCAL_ACTION_END_X**
3219  *			End.X action: Endpoint with Layer-3 cross-connect.
3220  *			Type of *param*: **struct in6_addr**.
3221  *		**SEG6_LOCAL_ACTION_END_T**
3222  *			End.T action: Endpoint with specific IPv6 table lookup.
3223  *			Type of *param*: **int**.
3224  *		**SEG6_LOCAL_ACTION_END_B6**
3225  *			End.B6 action: Endpoint bound to an SRv6 policy.
3226  *			Type of *param*: **struct ipv6_sr_hdr**.
3227  *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**
3228  *			End.B6.Encap action: Endpoint bound to an SRv6
3229  *			encapsulation policy.
3230  *			Type of *param*: **struct ipv6_sr_hdr**.
3231  *
3232  * 		A call to this helper is susceptible to change the underlying
3233  * 		packet buffer. Therefore, at load time, all checks on pointers
3234  * 		previously done by the verifier are invalidated and must be
3235  * 		performed again, if the helper is used in combination with
3236  * 		direct packet access.
3237  *	Return
3238  * 		0 on success, or a negative error in case of failure.
3239  *
3240  * long bpf_rc_repeat(void *ctx)
3241  *	Description
3242  *		This helper is used in programs implementing IR decoding, to
3243  *		report a successfully decoded repeat key message. This delays
3244  *		the generation of a key up event for previously generated
3245  *		key down event.
3246  *
3247  *		Some IR protocols like NEC have a special IR message for
3248  *		repeating last button, for when a button is held down.
3249  *
3250  *		The *ctx* should point to the lirc sample as passed into
3251  *		the program.
3252  *
3253  *		This helper is only available is the kernel was compiled with
3254  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3255  *		"**y**".
3256  *	Return
3257  *		0
3258  *
3259  * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3260  *	Description
3261  *		This helper is used in programs implementing IR decoding, to
3262  *		report a successfully decoded key press with *scancode*,
3263  *		*toggle* value in the given *protocol*. The scancode will be
3264  *		translated to a keycode using the rc keymap, and reported as
3265  *		an input key down event. After a period a key up event is
3266  *		generated. This period can be extended by calling either
3267  *		**bpf_rc_keydown**\ () again with the same values, or calling
3268  *		**bpf_rc_repeat**\ ().
3269  *
3270  *		Some protocols include a toggle bit, in case the button was
3271  *		released and pressed again between consecutive scancodes.
3272  *
3273  *		The *ctx* should point to the lirc sample as passed into
3274  *		the program.
3275  *
3276  *		The *protocol* is the decoded protocol number (see
3277  *		**enum rc_proto** for some predefined values).
3278  *
3279  *		This helper is only available is the kernel was compiled with
3280  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3281  *		"**y**".
3282  *	Return
3283  *		0
3284  *
3285  * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3286  * 	Description
3287  * 		Return the cgroup v2 id of the socket associated with the *skb*.
3288  * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3289  * 		helper for cgroup v1 by providing a tag resp. identifier that
3290  * 		can be matched on or used for map lookups e.g. to implement
3291  * 		policy. The cgroup v2 id of a given path in the hierarchy is
3292  * 		exposed in user space through the f_handle API in order to get
3293  * 		to the same 64-bit id.
3294  *
3295  * 		This helper can be used on TC egress path, but not on ingress,
3296  * 		and is available only if the kernel was compiled with the
3297  * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.
3298  * 	Return
3299  * 		The id is returned or 0 in case the id could not be retrieved.
3300  *
3301  * u64 bpf_get_current_cgroup_id(void)
3302  * 	Description
3303  * 		Get the current cgroup id based on the cgroup within which
3304  * 		the current task is running.
3305  * 	Return
3306  * 		A 64-bit integer containing the current cgroup id based
3307  * 		on the cgroup within which the current task is running.
3308  *
3309  * void *bpf_get_local_storage(void *map, u64 flags)
3310  *	Description
3311  *		Get the pointer to the local storage area.
3312  *		The type and the size of the local storage is defined
3313  *		by the *map* argument.
3314  *		The *flags* meaning is specific for each map type,
3315  *		and has to be 0 for cgroup local storage.
3316  *
3317  *		Depending on the BPF program type, a local storage area
3318  *		can be shared between multiple instances of the BPF program,
3319  *		running simultaneously.
3320  *
3321  *		A user should care about the synchronization by himself.
3322  *		For example, by using the **BPF_ATOMIC** instructions to alter
3323  *		the shared data.
3324  *	Return
3325  *		A pointer to the local storage area.
3326  *
3327  * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3328  *	Description
3329  *		Select a **SO_REUSEPORT** socket from a
3330  *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3331  *		It checks the selected socket is matching the incoming
3332  *		request in the socket buffer.
3333  *	Return
3334  *		0 on success, or a negative error in case of failure.
3335  *
3336  * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3337  *	Description
3338  *		Return id of cgroup v2 that is ancestor of cgroup associated
3339  *		with the *skb* at the *ancestor_level*.  The root cgroup is at
3340  *		*ancestor_level* zero and each step down the hierarchy
3341  *		increments the level. If *ancestor_level* == level of cgroup
3342  *		associated with *skb*, then return value will be same as that
3343  *		of **bpf_skb_cgroup_id**\ ().
3344  *
3345  *		The helper is useful to implement policies based on cgroups
3346  *		that are upper in hierarchy than immediate cgroup associated
3347  *		with *skb*.
3348  *
3349  *		The format of returned id and helper limitations are same as in
3350  *		**bpf_skb_cgroup_id**\ ().
3351  *	Return
3352  *		The id is returned or 0 in case the id could not be retrieved.
3353  *
3354  * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3355  *	Description
3356  *		Look for TCP socket matching *tuple*, optionally in a child
3357  *		network namespace *netns*. The return value must be checked,
3358  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3359  *
3360  *		The *ctx* should point to the context of the program, such as
3361  *		the skb or socket (depending on the hook in use). This is used
3362  *		to determine the base network namespace for the lookup.
3363  *
3364  *		*tuple_size* must be one of:
3365  *
3366  *		**sizeof**\ (*tuple*\ **->ipv4**)
3367  *			Look for an IPv4 socket.
3368  *		**sizeof**\ (*tuple*\ **->ipv6**)
3369  *			Look for an IPv6 socket.
3370  *
3371  *		If the *netns* is a negative signed 32-bit integer, then the
3372  *		socket lookup table in the netns associated with the *ctx*
3373  *		will be used. For the TC hooks, this is the netns of the device
3374  *		in the skb. For socket hooks, this is the netns of the socket.
3375  *		If *netns* is any other signed 32-bit value greater than or
3376  *		equal to zero then it specifies the ID of the netns relative to
3377  *		the netns associated with the *ctx*. *netns* values beyond the
3378  *		range of 32-bit integers are reserved for future use.
3379  *
3380  *		All values for *flags* are reserved for future usage, and must
3381  *		be left at zero.
3382  *
3383  *		This helper is available only if the kernel was compiled with
3384  *		**CONFIG_NET** configuration option.
3385  *	Return
3386  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3387  *		For sockets with reuseport option, the **struct bpf_sock**
3388  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3389  *		tuple.
3390  *
3391  * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3392  *	Description
3393  *		Look for UDP socket matching *tuple*, optionally in a child
3394  *		network namespace *netns*. The return value must be checked,
3395  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3396  *
3397  *		The *ctx* should point to the context of the program, such as
3398  *		the skb or socket (depending on the hook in use). This is used
3399  *		to determine the base network namespace for the lookup.
3400  *
3401  *		*tuple_size* must be one of:
3402  *
3403  *		**sizeof**\ (*tuple*\ **->ipv4**)
3404  *			Look for an IPv4 socket.
3405  *		**sizeof**\ (*tuple*\ **->ipv6**)
3406  *			Look for an IPv6 socket.
3407  *
3408  *		If the *netns* is a negative signed 32-bit integer, then the
3409  *		socket lookup table in the netns associated with the *ctx*
3410  *		will be used. For the TC hooks, this is the netns of the device
3411  *		in the skb. For socket hooks, this is the netns of the socket.
3412  *		If *netns* is any other signed 32-bit value greater than or
3413  *		equal to zero then it specifies the ID of the netns relative to
3414  *		the netns associated with the *ctx*. *netns* values beyond the
3415  *		range of 32-bit integers are reserved for future use.
3416  *
3417  *		All values for *flags* are reserved for future usage, and must
3418  *		be left at zero.
3419  *
3420  *		This helper is available only if the kernel was compiled with
3421  *		**CONFIG_NET** configuration option.
3422  *	Return
3423  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3424  *		For sockets with reuseport option, the **struct bpf_sock**
3425  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3426  *		tuple.
3427  *
3428  * long bpf_sk_release(void *sock)
3429  *	Description
3430  *		Release the reference held by *sock*. *sock* must be a
3431  *		non-**NULL** pointer that was returned from
3432  *		**bpf_sk_lookup_xxx**\ ().
3433  *	Return
3434  *		0 on success, or a negative error in case of failure.
3435  *
3436  * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3437  * 	Description
3438  * 		Push an element *value* in *map*. *flags* is one of:
3439  *
3440  * 		**BPF_EXIST**
3441  * 			If the queue/stack is full, the oldest element is
3442  * 			removed to make room for this.
3443  * 	Return
3444  * 		0 on success, or a negative error in case of failure.
3445  *
3446  * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3447  * 	Description
3448  * 		Pop an element from *map*.
3449  * 	Return
3450  * 		0 on success, or a negative error in case of failure.
3451  *
3452  * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3453  * 	Description
3454  * 		Get an element from *map* without removing it.
3455  * 	Return
3456  * 		0 on success, or a negative error in case of failure.
3457  *
3458  * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3459  *	Description
3460  *		For socket policies, insert *len* bytes into *msg* at offset
3461  *		*start*.
3462  *
3463  *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3464  *		*msg* it may want to insert metadata or options into the *msg*.
3465  *		This can later be read and used by any of the lower layer BPF
3466  *		hooks.
3467  *
3468  *		This helper may fail if under memory pressure (a malloc
3469  *		fails) in these cases BPF programs will get an appropriate
3470  *		error and BPF programs will need to handle them.
3471  *	Return
3472  *		0 on success, or a negative error in case of failure.
3473  *
3474  * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3475  *	Description
3476  *		Will remove *len* bytes from a *msg* starting at byte *start*.
3477  *		This may result in **ENOMEM** errors under certain situations if
3478  *		an allocation and copy are required due to a full ring buffer.
3479  *		However, the helper will try to avoid doing the allocation
3480  *		if possible. Other errors can occur if input parameters are
3481  *		invalid either due to *start* byte not being valid part of *msg*
3482  *		payload and/or *pop* value being to large.
3483  *	Return
3484  *		0 on success, or a negative error in case of failure.
3485  *
3486  * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3487  *	Description
3488  *		This helper is used in programs implementing IR decoding, to
3489  *		report a successfully decoded pointer movement.
3490  *
3491  *		The *ctx* should point to the lirc sample as passed into
3492  *		the program.
3493  *
3494  *		This helper is only available is the kernel was compiled with
3495  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3496  *		"**y**".
3497  *	Return
3498  *		0
3499  *
3500  * long bpf_spin_lock(struct bpf_spin_lock *lock)
3501  *	Description
3502  *		Acquire a spinlock represented by the pointer *lock*, which is
3503  *		stored as part of a value of a map. Taking the lock allows to
3504  *		safely update the rest of the fields in that value. The
3505  *		spinlock can (and must) later be released with a call to
3506  *		**bpf_spin_unlock**\ (\ *lock*\ ).
3507  *
3508  *		Spinlocks in BPF programs come with a number of restrictions
3509  *		and constraints:
3510  *
3511  *		* **bpf_spin_lock** objects are only allowed inside maps of
3512  *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3513  *		  list could be extended in the future).
3514  *		* BTF description of the map is mandatory.
3515  *		* The BPF program can take ONE lock at a time, since taking two
3516  *		  or more could cause dead locks.
3517  *		* Only one **struct bpf_spin_lock** is allowed per map element.
3518  *		* When the lock is taken, calls (either BPF to BPF or helpers)
3519  *		  are not allowed.
3520  *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3521  *		  allowed inside a spinlock-ed region.
3522  *		* The BPF program MUST call **bpf_spin_unlock**\ () to release
3523  *		  the lock, on all execution paths, before it returns.
3524  *		* The BPF program can access **struct bpf_spin_lock** only via
3525  *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3526  *		  helpers. Loading or storing data into the **struct
3527  *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3528  *		* To use the **bpf_spin_lock**\ () helper, the BTF description
3529  *		  of the map value must be a struct and have **struct
3530  *		  bpf_spin_lock** *anyname*\ **;** field at the top level.
3531  *		  Nested lock inside another struct is not allowed.
3532  *		* The **struct bpf_spin_lock** *lock* field in a map value must
3533  *		  be aligned on a multiple of 4 bytes in that value.
3534  *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3535  *		  the **bpf_spin_lock** field to user space.
3536  *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3537  *		  a BPF program, do not update the **bpf_spin_lock** field.
3538  *		* **bpf_spin_lock** cannot be on the stack or inside a
3539  *		  networking packet (it can only be inside of a map values).
3540  *		* **bpf_spin_lock** is available to root only.
3541  *		* Tracing programs and socket filter programs cannot use
3542  *		  **bpf_spin_lock**\ () due to insufficient preemption checks
3543  *		  (but this may change in the future).
3544  *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3545  *	Return
3546  *		0
3547  *
3548  * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3549  *	Description
3550  *		Release the *lock* previously locked by a call to
3551  *		**bpf_spin_lock**\ (\ *lock*\ ).
3552  *	Return
3553  *		0
3554  *
3555  * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3556  *	Description
3557  *		This helper gets a **struct bpf_sock** pointer such
3558  *		that all the fields in this **bpf_sock** can be accessed.
3559  *	Return
3560  *		A **struct bpf_sock** pointer on success, or **NULL** in
3561  *		case of failure.
3562  *
3563  * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3564  *	Description
3565  *		This helper gets a **struct bpf_tcp_sock** pointer from a
3566  *		**struct bpf_sock** pointer.
3567  *	Return
3568  *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3569  *		case of failure.
3570  *
3571  * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3572  *	Description
3573  *		Set ECN (Explicit Congestion Notification) field of IP header
3574  *		to **CE** (Congestion Encountered) if current value is **ECT**
3575  *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3576  *		and IPv4.
3577  *	Return
3578  *		1 if the **CE** flag is set (either by the current helper call
3579  *		or because it was already present), 0 if it is not set.
3580  *
3581  * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3582  *	Description
3583  *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3584  *		**bpf_sk_release**\ () is unnecessary and not allowed.
3585  *	Return
3586  *		A **struct bpf_sock** pointer on success, or **NULL** in
3587  *		case of failure.
3588  *
3589  * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3590  *	Description
3591  *		Look for TCP socket matching *tuple*, optionally in a child
3592  *		network namespace *netns*. The return value must be checked,
3593  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3594  *
3595  *		This function is identical to **bpf_sk_lookup_tcp**\ (), except
3596  *		that it also returns timewait or request sockets. Use
3597  *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3598  *		full structure.
3599  *
3600  *		This helper is available only if the kernel was compiled with
3601  *		**CONFIG_NET** configuration option.
3602  *	Return
3603  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3604  *		For sockets with reuseport option, the **struct bpf_sock**
3605  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3606  *		tuple.
3607  *
3608  * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3609  * 	Description
3610  * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for
3611  * 		the listening socket in *sk*.
3612  *
3613  * 		*iph* points to the start of the IPv4 or IPv6 header, while
3614  * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3615  * 		**sizeof**\ (**struct ipv6hdr**).
3616  *
3617  * 		*th* points to the start of the TCP header, while *th_len*
3618  *		contains the length of the TCP header (at least
3619  *		**sizeof**\ (**struct tcphdr**)).
3620  * 	Return
3621  * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3622  * 		error otherwise.
3623  *
3624  * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3625  *	Description
3626  *		Get name of sysctl in /proc/sys/ and copy it into provided by
3627  *		program buffer *buf* of size *buf_len*.
3628  *
3629  *		The buffer is always NUL terminated, unless it's zero-sized.
3630  *
3631  *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3632  *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3633  *		only (e.g. "tcp_mem").
3634  *	Return
3635  *		Number of character copied (not including the trailing NUL).
3636  *
3637  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3638  *		truncated name in this case).
3639  *
3640  * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3641  *	Description
3642  *		Get current value of sysctl as it is presented in /proc/sys
3643  *		(incl. newline, etc), and copy it as a string into provided
3644  *		by program buffer *buf* of size *buf_len*.
3645  *
3646  *		The whole value is copied, no matter what file position user
3647  *		space issued e.g. sys_read at.
3648  *
3649  *		The buffer is always NUL terminated, unless it's zero-sized.
3650  *	Return
3651  *		Number of character copied (not including the trailing NUL).
3652  *
3653  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3654  *		truncated name in this case).
3655  *
3656  *		**-EINVAL** if current value was unavailable, e.g. because
3657  *		sysctl is uninitialized and read returns -EIO for it.
3658  *
3659  * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3660  *	Description
3661  *		Get new value being written by user space to sysctl (before
3662  *		the actual write happens) and copy it as a string into
3663  *		provided by program buffer *buf* of size *buf_len*.
3664  *
3665  *		User space may write new value at file position > 0.
3666  *
3667  *		The buffer is always NUL terminated, unless it's zero-sized.
3668  *	Return
3669  *		Number of character copied (not including the trailing NUL).
3670  *
3671  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3672  *		truncated name in this case).
3673  *
3674  *		**-EINVAL** if sysctl is being read.
3675  *
3676  * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3677  *	Description
3678  *		Override new value being written by user space to sysctl with
3679  *		value provided by program in buffer *buf* of size *buf_len*.
3680  *
3681  *		*buf* should contain a string in same form as provided by user
3682  *		space on sysctl write.
3683  *
3684  *		User space may write new value at file position > 0. To override
3685  *		the whole sysctl value file position should be set to zero.
3686  *	Return
3687  *		0 on success.
3688  *
3689  *		**-E2BIG** if the *buf_len* is too big.
3690  *
3691  *		**-EINVAL** if sysctl is being read.
3692  *
3693  * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3694  *	Description
3695  *		Convert the initial part of the string from buffer *buf* of
3696  *		size *buf_len* to a long integer according to the given base
3697  *		and save the result in *res*.
3698  *
3699  *		The string may begin with an arbitrary amount of white space
3700  *		(as determined by **isspace**\ (3)) followed by a single
3701  *		optional '**-**' sign.
3702  *
3703  *		Five least significant bits of *flags* encode base, other bits
3704  *		are currently unused.
3705  *
3706  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3707  *		similar to user space **strtol**\ (3).
3708  *	Return
3709  *		Number of characters consumed on success. Must be positive but
3710  *		no more than *buf_len*.
3711  *
3712  *		**-EINVAL** if no valid digits were found or unsupported base
3713  *		was provided.
3714  *
3715  *		**-ERANGE** if resulting value was out of range.
3716  *
3717  * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3718  *	Description
3719  *		Convert the initial part of the string from buffer *buf* of
3720  *		size *buf_len* to an unsigned long integer according to the
3721  *		given base and save the result in *res*.
3722  *
3723  *		The string may begin with an arbitrary amount of white space
3724  *		(as determined by **isspace**\ (3)).
3725  *
3726  *		Five least significant bits of *flags* encode base, other bits
3727  *		are currently unused.
3728  *
3729  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3730  *		similar to user space **strtoul**\ (3).
3731  *	Return
3732  *		Number of characters consumed on success. Must be positive but
3733  *		no more than *buf_len*.
3734  *
3735  *		**-EINVAL** if no valid digits were found or unsupported base
3736  *		was provided.
3737  *
3738  *		**-ERANGE** if resulting value was out of range.
3739  *
3740  * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3741  *	Description
3742  *		Get a bpf-local-storage from a *sk*.
3743  *
3744  *		Logically, it could be thought of getting the value from
3745  *		a *map* with *sk* as the **key**.  From this
3746  *		perspective,  the usage is not much different from
3747  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3748  *		helper enforces the key must be a full socket and the map must
3749  *		be a **BPF_MAP_TYPE_SK_STORAGE** also.
3750  *
3751  *		Underneath, the value is stored locally at *sk* instead of
3752  *		the *map*.  The *map* is used as the bpf-local-storage
3753  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
3754  *		searched against all bpf-local-storages residing at *sk*.
3755  *
3756  *		*sk* is a kernel **struct sock** pointer for LSM program.
3757  *		*sk* is a **struct bpf_sock** pointer for other program types.
3758  *
3759  *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3760  *		used such that a new bpf-local-storage will be
3761  *		created if one does not exist.  *value* can be used
3762  *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3763  *		the initial value of a bpf-local-storage.  If *value* is
3764  *		**NULL**, the new bpf-local-storage will be zero initialized.
3765  *	Return
3766  *		A bpf-local-storage pointer is returned on success.
3767  *
3768  *		**NULL** if not found or there was an error in adding
3769  *		a new bpf-local-storage.
3770  *
3771  * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3772  *	Description
3773  *		Delete a bpf-local-storage from a *sk*.
3774  *	Return
3775  *		0 on success.
3776  *
3777  *		**-ENOENT** if the bpf-local-storage cannot be found.
3778  *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3779  *
3780  * long bpf_send_signal(u32 sig)
3781  *	Description
3782  *		Send signal *sig* to the process of the current task.
3783  *		The signal may be delivered to any of this process's threads.
3784  *	Return
3785  *		0 on success or successfully queued.
3786  *
3787  *		**-EBUSY** if work queue under nmi is full.
3788  *
3789  *		**-EINVAL** if *sig* is invalid.
3790  *
3791  *		**-EPERM** if no permission to send the *sig*.
3792  *
3793  *		**-EAGAIN** if bpf program can try again.
3794  *
3795  * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3796  *	Description
3797  *		Try to issue a SYN cookie for the packet with corresponding
3798  *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3799  *
3800  *		*iph* points to the start of the IPv4 or IPv6 header, while
3801  *		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3802  *		**sizeof**\ (**struct ipv6hdr**).
3803  *
3804  *		*th* points to the start of the TCP header, while *th_len*
3805  *		contains the length of the TCP header with options (at least
3806  *		**sizeof**\ (**struct tcphdr**)).
3807  *	Return
3808  *		On success, lower 32 bits hold the generated SYN cookie in
3809  *		followed by 16 bits which hold the MSS value for that cookie,
3810  *		and the top 16 bits are unused.
3811  *
3812  *		On failure, the returned value is one of the following:
3813  *
3814  *		**-EINVAL** SYN cookie cannot be issued due to error
3815  *
3816  *		**-ENOENT** SYN cookie should not be issued (no SYN flood)
3817  *
3818  *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3819  *
3820  *		**-EPROTONOSUPPORT** IP packet version is not 4 or 6
3821  *
3822  * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3823  * 	Description
3824  * 		Write raw *data* blob into a special BPF perf event held by
3825  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3826  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
3827  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3828  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3829  *
3830  * 		The *flags* are used to indicate the index in *map* for which
3831  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
3832  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3833  * 		to indicate that the index of the current CPU core should be
3834  * 		used.
3835  *
3836  * 		The value to write, of *size*, is passed through eBPF stack and
3837  * 		pointed by *data*.
3838  *
3839  * 		*ctx* is a pointer to in-kernel struct sk_buff.
3840  *
3841  * 		This helper is similar to **bpf_perf_event_output**\ () but
3842  * 		restricted to raw_tracepoint bpf programs.
3843  * 	Return
3844  * 		0 on success, or a negative error in case of failure.
3845  *
3846  * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3847  * 	Description
3848  * 		Safely attempt to read *size* bytes from user space address
3849  * 		*unsafe_ptr* and store the data in *dst*.
3850  * 	Return
3851  * 		0 on success, or a negative error in case of failure.
3852  *
3853  * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3854  * 	Description
3855  * 		Safely attempt to read *size* bytes from kernel space address
3856  * 		*unsafe_ptr* and store the data in *dst*.
3857  * 	Return
3858  * 		0 on success, or a negative error in case of failure.
3859  *
3860  * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3861  * 	Description
3862  * 		Copy a NUL terminated string from an unsafe user address
3863  * 		*unsafe_ptr* to *dst*. The *size* should include the
3864  * 		terminating NUL byte. In case the string length is smaller than
3865  * 		*size*, the target is not padded with further NUL bytes. If the
3866  * 		string length is larger than *size*, just *size*-1 bytes are
3867  * 		copied and the last byte is set to NUL.
3868  *
3869  * 		On success, returns the number of bytes that were written,
3870  * 		including the terminal NUL. This makes this helper useful in
3871  * 		tracing programs for reading strings, and more importantly to
3872  * 		get its length at runtime. See the following snippet:
3873  *
3874  * 		::
3875  *
3876  * 			SEC("kprobe/sys_open")
3877  * 			void bpf_sys_open(struct pt_regs *ctx)
3878  * 			{
3879  * 			        char buf[PATHLEN]; // PATHLEN is defined to 256
3880  * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),
3881  * 				                                  ctx->di);
3882  *
3883  * 				// Consume buf, for example push it to
3884  * 				// userspace via bpf_perf_event_output(); we
3885  * 				// can use res (the string length) as event
3886  * 				// size, after checking its boundaries.
3887  * 			}
3888  *
3889  * 		In comparison, using **bpf_probe_read_user**\ () helper here
3890  * 		instead to read the string would require to estimate the length
3891  * 		at compile time, and would often result in copying more memory
3892  * 		than necessary.
3893  *
3894  * 		Another useful use case is when parsing individual process
3895  * 		arguments or individual environment variables navigating
3896  * 		*current*\ **->mm->arg_start** and *current*\
3897  * 		**->mm->env_start**: using this helper and the return value,
3898  * 		one can quickly iterate at the right offset of the memory area.
3899  * 	Return
3900  * 		On success, the strictly positive length of the output string,
3901  * 		including the trailing NUL character. On error, a negative
3902  * 		value.
3903  *
3904  * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3905  * 	Description
3906  * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3907  * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3908  * 	Return
3909  * 		On success, the strictly positive length of the string, including
3910  * 		the trailing NUL character. On error, a negative value.
3911  *
3912  * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3913  *	Description
3914  *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3915  *		*rcv_nxt* is the ack_seq to be sent out.
3916  *	Return
3917  *		0 on success, or a negative error in case of failure.
3918  *
3919  * long bpf_send_signal_thread(u32 sig)
3920  *	Description
3921  *		Send signal *sig* to the thread corresponding to the current task.
3922  *	Return
3923  *		0 on success or successfully queued.
3924  *
3925  *		**-EBUSY** if work queue under nmi is full.
3926  *
3927  *		**-EINVAL** if *sig* is invalid.
3928  *
3929  *		**-EPERM** if no permission to send the *sig*.
3930  *
3931  *		**-EAGAIN** if bpf program can try again.
3932  *
3933  * u64 bpf_jiffies64(void)
3934  *	Description
3935  *		Obtain the 64bit jiffies
3936  *	Return
3937  *		The 64 bit jiffies
3938  *
3939  * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3940  *	Description
3941  *		For an eBPF program attached to a perf event, retrieve the
3942  *		branch records (**struct perf_branch_entry**) associated to *ctx*
3943  *		and store it in the buffer pointed by *buf* up to size
3944  *		*size* bytes.
3945  *	Return
3946  *		On success, number of bytes written to *buf*. On error, a
3947  *		negative value.
3948  *
3949  *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3950  *		instead return the number of bytes required to store all the
3951  *		branch entries. If this flag is set, *buf* may be NULL.
3952  *
3953  *		**-EINVAL** if arguments invalid or **size** not a multiple
3954  *		of **sizeof**\ (**struct perf_branch_entry**\ ).
3955  *
3956  *		**-ENOENT** if architecture does not support branch records.
3957  *
3958  * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3959  *	Description
3960  *		Returns 0 on success, values for *pid* and *tgid* as seen from the current
3961  *		*namespace* will be returned in *nsdata*.
3962  *	Return
3963  *		0 on success, or one of the following in case of failure:
3964  *
3965  *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number
3966  *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
3967  *
3968  *		**-ENOENT** if pidns does not exists for the current task.
3969  *
3970  * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3971  *	Description
3972  *		Write raw *data* blob into a special BPF perf event held by
3973  *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3974  *		event must have the following attributes: **PERF_SAMPLE_RAW**
3975  *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3976  *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3977  *
3978  *		The *flags* are used to indicate the index in *map* for which
3979  *		the value must be put, masked with **BPF_F_INDEX_MASK**.
3980  *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3981  *		to indicate that the index of the current CPU core should be
3982  *		used.
3983  *
3984  *		The value to write, of *size*, is passed through eBPF stack and
3985  *		pointed by *data*.
3986  *
3987  *		*ctx* is a pointer to in-kernel struct xdp_buff.
3988  *
3989  *		This helper is similar to **bpf_perf_eventoutput**\ () but
3990  *		restricted to raw_tracepoint bpf programs.
3991  *	Return
3992  *		0 on success, or a negative error in case of failure.
3993  *
3994  * u64 bpf_get_netns_cookie(void *ctx)
3995  * 	Description
3996  * 		Retrieve the cookie (generated by the kernel) of the network
3997  * 		namespace the input *ctx* is associated with. The network
3998  * 		namespace cookie remains stable for its lifetime and provides
3999  * 		a global identifier that can be assumed unique. If *ctx* is
4000  * 		NULL, then the helper returns the cookie for the initial
4001  * 		network namespace. The cookie itself is very similar to that
4002  * 		of **bpf_get_socket_cookie**\ () helper, but for network
4003  * 		namespaces instead of sockets.
4004  * 	Return
4005  * 		A 8-byte long opaque number.
4006  *
4007  * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
4008  * 	Description
4009  * 		Return id of cgroup v2 that is ancestor of the cgroup associated
4010  * 		with the current task at the *ancestor_level*. The root cgroup
4011  * 		is at *ancestor_level* zero and each step down the hierarchy
4012  * 		increments the level. If *ancestor_level* == level of cgroup
4013  * 		associated with the current task, then return value will be the
4014  * 		same as that of **bpf_get_current_cgroup_id**\ ().
4015  *
4016  * 		The helper is useful to implement policies based on cgroups
4017  * 		that are upper in hierarchy than immediate cgroup associated
4018  * 		with the current task.
4019  *
4020  * 		The format of returned id and helper limitations are same as in
4021  * 		**bpf_get_current_cgroup_id**\ ().
4022  * 	Return
4023  * 		The id is returned or 0 in case the id could not be retrieved.
4024  *
4025  * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
4026  *	Description
4027  *		Helper is overloaded depending on BPF program type. This
4028  *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and
4029  *		**BPF_PROG_TYPE_SCHED_ACT** programs.
4030  *
4031  *		Assign the *sk* to the *skb*. When combined with appropriate
4032  *		routing configuration to receive the packet towards the socket,
4033  *		will cause *skb* to be delivered to the specified socket.
4034  *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),
4035  *		**bpf_clone_redirect**\ () or other methods outside of BPF may
4036  *		interfere with successful delivery to the socket.
4037  *
4038  *		This operation is only valid from TC ingress path.
4039  *
4040  *		The *flags* argument must be zero.
4041  *	Return
4042  *		0 on success, or a negative error in case of failure:
4043  *
4044  *		**-EINVAL** if specified *flags* are not supported.
4045  *
4046  *		**-ENOENT** if the socket is unavailable for assignment.
4047  *
4048  *		**-ENETUNREACH** if the socket is unreachable (wrong netns).
4049  *
4050  *		**-EOPNOTSUPP** if the operation is not supported, for example
4051  *		a call from outside of TC ingress.
4052  *
4053  *		**-ESOCKTNOSUPPORT** if the socket type is not supported
4054  *		(reuseport).
4055  *
4056  * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
4057  *	Description
4058  *		Helper is overloaded depending on BPF program type. This
4059  *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
4060  *
4061  *		Select the *sk* as a result of a socket lookup.
4062  *
4063  *		For the operation to succeed passed socket must be compatible
4064  *		with the packet description provided by the *ctx* object.
4065  *
4066  *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4067  *		be an exact match. While IP family (**AF_INET** or
4068  *		**AF_INET6**) must be compatible, that is IPv6 sockets
4069  *		that are not v6-only can be selected for IPv4 packets.
4070  *
4071  *		Only TCP listeners and UDP unconnected sockets can be
4072  *		selected. *sk* can also be NULL to reset any previous
4073  *		selection.
4074  *
4075  *		*flags* argument can combination of following values:
4076  *
4077  *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4078  *		  socket selection, potentially done by a BPF program
4079  *		  that ran before us.
4080  *
4081  *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4082  *		  load-balancing within reuseport group for the socket
4083  *		  being selected.
4084  *
4085  *		On success *ctx->sk* will point to the selected socket.
4086  *
4087  *	Return
4088  *		0 on success, or a negative errno in case of failure.
4089  *
4090  *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is
4091  *		  not compatible with packet family (*ctx->family*).
4092  *
4093  *		* **-EEXIST** if socket has been already selected,
4094  *		  potentially by another program, and
4095  *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4096  *
4097  *		* **-EINVAL** if unsupported flags were specified.
4098  *
4099  *		* **-EPROTOTYPE** if socket L4 protocol
4100  *		  (*sk->protocol*) doesn't match packet protocol
4101  *		  (*ctx->protocol*).
4102  *
4103  *		* **-ESOCKTNOSUPPORT** if socket is not in allowed
4104  *		  state (TCP listening or UDP unconnected).
4105  *
4106  * u64 bpf_ktime_get_boot_ns(void)
4107  * 	Description
4108  * 		Return the time elapsed since system boot, in nanoseconds.
4109  * 		Does include the time the system was suspended.
4110  * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4111  * 	Return
4112  * 		Current *ktime*.
4113  *
4114  * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4115  * 	Description
4116  * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4117  * 		out the format string.
4118  * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4119  * 		the format string itself. The *data* and *data_len* are format string
4120  * 		arguments. The *data* are a **u64** array and corresponding format string
4121  * 		values are stored in the array. For strings and pointers where pointees
4122  * 		are accessed, only the pointer values are stored in the *data* array.
4123  * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4124  *
4125  *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4126  *		Reading kernel memory may fail due to either invalid address or
4127  *		valid address but requiring a major memory fault. If reading kernel memory
4128  *		fails, the string for **%s** will be an empty string, and the ip
4129  *		address for **%p{i,I}{4,6}** will be 0. Not returning error to
4130  *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4131  * 	Return
4132  * 		0 on success, or a negative error in case of failure:
4133  *
4134  *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again
4135  *		by returning 1 from bpf program.
4136  *
4137  *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4138  *
4139  *		**-E2BIG** if *fmt* contains too many format specifiers.
4140  *
4141  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4142  *
4143  * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4144  * 	Description
4145  * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4146  * 		The *m* represents the seq_file. The *data* and *len* represent the
4147  * 		data to write in bytes.
4148  * 	Return
4149  * 		0 on success, or a negative error in case of failure:
4150  *
4151  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4152  *
4153  * u64 bpf_sk_cgroup_id(void *sk)
4154  *	Description
4155  *		Return the cgroup v2 id of the socket *sk*.
4156  *
4157  *		*sk* must be a non-**NULL** pointer to a socket, e.g. one
4158  *		returned from **bpf_sk_lookup_xxx**\ (),
4159  *		**bpf_sk_fullsock**\ (), etc. The format of returned id is
4160  *		same as in **bpf_skb_cgroup_id**\ ().
4161  *
4162  *		This helper is available only if the kernel was compiled with
4163  *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4164  *	Return
4165  *		The id is returned or 0 in case the id could not be retrieved.
4166  *
4167  * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4168  *	Description
4169  *		Return id of cgroup v2 that is ancestor of cgroup associated
4170  *		with the *sk* at the *ancestor_level*.  The root cgroup is at
4171  *		*ancestor_level* zero and each step down the hierarchy
4172  *		increments the level. If *ancestor_level* == level of cgroup
4173  *		associated with *sk*, then return value will be same as that
4174  *		of **bpf_sk_cgroup_id**\ ().
4175  *
4176  *		The helper is useful to implement policies based on cgroups
4177  *		that are upper in hierarchy than immediate cgroup associated
4178  *		with *sk*.
4179  *
4180  *		The format of returned id and helper limitations are same as in
4181  *		**bpf_sk_cgroup_id**\ ().
4182  *	Return
4183  *		The id is returned or 0 in case the id could not be retrieved.
4184  *
4185  * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4186  * 	Description
4187  * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4188  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4189  * 		of new data availability is sent.
4190  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4191  * 		of new data availability is sent unconditionally.
4192  * 		If **0** is specified in *flags*, an adaptive notification
4193  * 		of new data availability is sent.
4194  *
4195  * 		An adaptive notification is a notification sent whenever the user-space
4196  * 		process has caught up and consumed all available payloads. In case the user-space
4197  * 		process is still processing a previous payload, then no notification is needed
4198  * 		as it will process the newly added payload automatically.
4199  * 	Return
4200  * 		0 on success, or a negative error in case of failure.
4201  *
4202  * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4203  * 	Description
4204  * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4205  * 		*flags* must be 0.
4206  * 	Return
4207  * 		Valid pointer with *size* bytes of memory available; NULL,
4208  * 		otherwise.
4209  *
4210  * void bpf_ringbuf_submit(void *data, u64 flags)
4211  * 	Description
4212  * 		Submit reserved ring buffer sample, pointed to by *data*.
4213  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4214  * 		of new data availability is sent.
4215  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4216  * 		of new data availability is sent unconditionally.
4217  * 		If **0** is specified in *flags*, an adaptive notification
4218  * 		of new data availability is sent.
4219  *
4220  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4221  * 	Return
4222  * 		Nothing. Always succeeds.
4223  *
4224  * void bpf_ringbuf_discard(void *data, u64 flags)
4225  * 	Description
4226  * 		Discard reserved ring buffer sample, pointed to by *data*.
4227  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4228  * 		of new data availability is sent.
4229  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4230  * 		of new data availability is sent unconditionally.
4231  * 		If **0** is specified in *flags*, an adaptive notification
4232  * 		of new data availability is sent.
4233  *
4234  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4235  * 	Return
4236  * 		Nothing. Always succeeds.
4237  *
4238  * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4239  *	Description
4240  *		Query various characteristics of provided ring buffer. What
4241  *		exactly is queries is determined by *flags*:
4242  *
4243  *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4244  *		* **BPF_RB_RING_SIZE**: The size of ring buffer.
4245  *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4246  *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4247  *
4248  *		Data returned is just a momentary snapshot of actual values
4249  *		and could be inaccurate, so this facility should be used to
4250  *		power heuristics and for reporting, not to make 100% correct
4251  *		calculation.
4252  *	Return
4253  *		Requested value, or 0, if *flags* are not recognized.
4254  *
4255  * long bpf_csum_level(struct sk_buff *skb, u64 level)
4256  * 	Description
4257  * 		Change the skbs checksum level by one layer up or down, or
4258  * 		reset it entirely to none in order to have the stack perform
4259  * 		checksum validation. The level is applicable to the following
4260  * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4261  * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4262  * 		through **bpf_skb_adjust_room**\ () helper with passing in
4263  * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call
4264  * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4265  * 		the UDP header is removed. Similarly, an encap of the latter
4266  * 		into the former could be accompanied by a helper call to
4267  * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4268  * 		skb is still intended to be processed in higher layers of the
4269  * 		stack instead of just egressing at tc.
4270  *
4271  * 		There are three supported level settings at this time:
4272  *
4273  * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4274  * 		  with CHECKSUM_UNNECESSARY.
4275  * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4276  * 		  with CHECKSUM_UNNECESSARY.
4277  * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4278  * 		  sets CHECKSUM_NONE to force checksum validation by the stack.
4279  * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4280  * 		  skb->csum_level.
4281  * 	Return
4282  * 		0 on success, or a negative error in case of failure. In the
4283  * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4284  * 		is returned or the error code -EACCES in case the skb is not
4285  * 		subject to CHECKSUM_UNNECESSARY.
4286  *
4287  * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4288  *	Description
4289  *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4290  *	Return
4291  *		*sk* if casting is valid, or **NULL** otherwise.
4292  *
4293  * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4294  *	Description
4295  *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4296  *	Return
4297  *		*sk* if casting is valid, or **NULL** otherwise.
4298  *
4299  * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4300  * 	Description
4301  *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4302  *	Return
4303  *		*sk* if casting is valid, or **NULL** otherwise.
4304  *
4305  * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4306  * 	Description
4307  *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4308  *	Return
4309  *		*sk* if casting is valid, or **NULL** otherwise.
4310  *
4311  * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4312  * 	Description
4313  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4314  *	Return
4315  *		*sk* if casting is valid, or **NULL** otherwise.
4316  *
4317  * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4318  *	Description
4319  *		Return a user or a kernel stack in bpf program provided buffer.
4320  *		To achieve this, the helper needs *task*, which is a valid
4321  *		pointer to **struct task_struct**. To store the stacktrace, the
4322  *		bpf program provides *buf* with a nonnegative *size*.
4323  *
4324  *		The last argument, *flags*, holds the number of stack frames to
4325  *		skip (from 0 to 255), masked with
4326  *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4327  *		the following flags:
4328  *
4329  *		**BPF_F_USER_STACK**
4330  *			Collect a user space stack instead of a kernel stack.
4331  *		**BPF_F_USER_BUILD_ID**
4332  *			Collect buildid+offset instead of ips for user stack,
4333  *			only valid if **BPF_F_USER_STACK** is also specified.
4334  *
4335  *		**bpf_get_task_stack**\ () can collect up to
4336  *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4337  *		to sufficient large buffer size. Note that
4338  *		this limit can be controlled with the **sysctl** program, and
4339  *		that it should be manually increased in order to profile long
4340  *		user stacks (such as stacks for Java programs). To do so, use:
4341  *
4342  *		::
4343  *
4344  *			# sysctl kernel.perf_event_max_stack=<new value>
4345  *	Return
4346  * 		The non-negative copied *buf* length equal to or less than
4347  * 		*size* on success, or a negative error in case of failure.
4348  *
4349  * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4350  *	Description
4351  *		Load header option.  Support reading a particular TCP header
4352  *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4353  *
4354  *		If *flags* is 0, it will search the option from the
4355  *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
4356  *		has details on what skb_data contains under different
4357  *		*skops*\ **->op**.
4358  *
4359  *		The first byte of the *searchby_res* specifies the
4360  *		kind that it wants to search.
4361  *
4362  *		If the searching kind is an experimental kind
4363  *		(i.e. 253 or 254 according to RFC6994).  It also
4364  *		needs to specify the "magic" which is either
4365  *		2 bytes or 4 bytes.  It then also needs to
4366  *		specify the size of the magic by using
4367  *		the 2nd byte which is "kind-length" of a TCP
4368  *		header option and the "kind-length" also
4369  *		includes the first 2 bytes "kind" and "kind-length"
4370  *		itself as a normal TCP header option also does.
4371  *
4372  *		For example, to search experimental kind 254 with
4373  *		2 byte magic 0xeB9F, the searchby_res should be
4374  *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4375  *
4376  *		To search for the standard window scale option (3),
4377  *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4378  *		Note, kind-length must be 0 for regular option.
4379  *
4380  *		Searching for No-Op (0) and End-of-Option-List (1) are
4381  *		not supported.
4382  *
4383  *		*len* must be at least 2 bytes which is the minimal size
4384  *		of a header option.
4385  *
4386  *		Supported flags:
4387  *
4388  *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4389  *		  saved_syn packet or the just-received syn packet.
4390  *
4391  *	Return
4392  *		> 0 when found, the header option is copied to *searchby_res*.
4393  *		The return value is the total length copied. On failure, a
4394  *		negative error code is returned:
4395  *
4396  *		**-EINVAL** if a parameter is invalid.
4397  *
4398  *		**-ENOMSG** if the option is not found.
4399  *
4400  *		**-ENOENT** if no syn packet is available when
4401  *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4402  *
4403  *		**-ENOSPC** if there is not enough space.  Only *len* number of
4404  *		bytes are copied.
4405  *
4406  *		**-EFAULT** on failure to parse the header options in the
4407  *		packet.
4408  *
4409  *		**-EPERM** if the helper cannot be used under the current
4410  *		*skops*\ **->op**.
4411  *
4412  * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4413  *	Description
4414  *		Store header option.  The data will be copied
4415  *		from buffer *from* with length *len* to the TCP header.
4416  *
4417  *		The buffer *from* should have the whole option that
4418  *		includes the kind, kind-length, and the actual
4419  *		option data.  The *len* must be at least kind-length
4420  *		long.  The kind-length does not have to be 4 byte
4421  *		aligned.  The kernel will take care of the padding
4422  *		and setting the 4 bytes aligned value to th->doff.
4423  *
4424  *		This helper will check for duplicated option
4425  *		by searching the same option in the outgoing skb.
4426  *
4427  *		This helper can only be called during
4428  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4429  *
4430  *	Return
4431  *		0 on success, or negative error in case of failure:
4432  *
4433  *		**-EINVAL** If param is invalid.
4434  *
4435  *		**-ENOSPC** if there is not enough space in the header.
4436  *		Nothing has been written
4437  *
4438  *		**-EEXIST** if the option already exists.
4439  *
4440  *		**-EFAULT** on failrue to parse the existing header options.
4441  *
4442  *		**-EPERM** if the helper cannot be used under the current
4443  *		*skops*\ **->op**.
4444  *
4445  * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4446  *	Description
4447  *		Reserve *len* bytes for the bpf header option.  The
4448  *		space will be used by **bpf_store_hdr_opt**\ () later in
4449  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4450  *
4451  *		If **bpf_reserve_hdr_opt**\ () is called multiple times,
4452  *		the total number of bytes will be reserved.
4453  *
4454  *		This helper can only be called during
4455  *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4456  *
4457  *	Return
4458  *		0 on success, or negative error in case of failure:
4459  *
4460  *		**-EINVAL** if a parameter is invalid.
4461  *
4462  *		**-ENOSPC** if there is not enough space in the header.
4463  *
4464  *		**-EPERM** if the helper cannot be used under the current
4465  *		*skops*\ **->op**.
4466  *
4467  * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4468  *	Description
4469  *		Get a bpf_local_storage from an *inode*.
4470  *
4471  *		Logically, it could be thought of as getting the value from
4472  *		a *map* with *inode* as the **key**.  From this
4473  *		perspective,  the usage is not much different from
4474  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4475  *		helper enforces the key must be an inode and the map must also
4476  *		be a **BPF_MAP_TYPE_INODE_STORAGE**.
4477  *
4478  *		Underneath, the value is stored locally at *inode* instead of
4479  *		the *map*.  The *map* is used as the bpf-local-storage
4480  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4481  *		searched against all bpf_local_storage residing at *inode*.
4482  *
4483  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4484  *		used such that a new bpf_local_storage will be
4485  *		created if one does not exist.  *value* can be used
4486  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4487  *		the initial value of a bpf_local_storage.  If *value* is
4488  *		**NULL**, the new bpf_local_storage will be zero initialized.
4489  *	Return
4490  *		A bpf_local_storage pointer is returned on success.
4491  *
4492  *		**NULL** if not found or there was an error in adding
4493  *		a new bpf_local_storage.
4494  *
4495  * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4496  *	Description
4497  *		Delete a bpf_local_storage from an *inode*.
4498  *	Return
4499  *		0 on success.
4500  *
4501  *		**-ENOENT** if the bpf_local_storage cannot be found.
4502  *
4503  * long bpf_d_path(struct path *path, char *buf, u32 sz)
4504  *	Description
4505  *		Return full path for given **struct path** object, which
4506  *		needs to be the kernel BTF *path* object. The path is
4507  *		returned in the provided buffer *buf* of size *sz* and
4508  *		is zero terminated.
4509  *
4510  *	Return
4511  *		On success, the strictly positive length of the string,
4512  *		including the trailing NUL character. On error, a negative
4513  *		value.
4514  *
4515  * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4516  * 	Description
4517  * 		Read *size* bytes from user space address *user_ptr* and store
4518  * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4519  * 	Return
4520  * 		0 on success, or a negative error in case of failure.
4521  *
4522  * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4523  *	Description
4524  *		Use BTF to store a string representation of *ptr*->ptr in *str*,
4525  *		using *ptr*->type_id.  This value should specify the type
4526  *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4527  *		can be used to look up vmlinux BTF type ids. Traversing the
4528  *		data structure using BTF, the type information and values are
4529  *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
4530  *		the pointer data is carried out to avoid kernel crashes during
4531  *		operation.  Smaller types can use string space on the stack;
4532  *		larger programs can use map data to store the string
4533  *		representation.
4534  *
4535  *		The string can be subsequently shared with userspace via
4536  *		bpf_perf_event_output() or ring buffer interfaces.
4537  *		bpf_trace_printk() is to be avoided as it places too small
4538  *		a limit on string size to be useful.
4539  *
4540  *		*flags* is a combination of
4541  *
4542  *		**BTF_F_COMPACT**
4543  *			no formatting around type information
4544  *		**BTF_F_NONAME**
4545  *			no struct/union member names/types
4546  *		**BTF_F_PTR_RAW**
4547  *			show raw (unobfuscated) pointer values;
4548  *			equivalent to printk specifier %px.
4549  *		**BTF_F_ZERO**
4550  *			show zero-valued struct/union members; they
4551  *			are not displayed by default
4552  *
4553  *	Return
4554  *		The number of bytes that were written (or would have been
4555  *		written if output had to be truncated due to string size),
4556  *		or a negative error in cases of failure.
4557  *
4558  * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4559  *	Description
4560  *		Use BTF to write to seq_write a string representation of
4561  *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4562  *		*flags* are identical to those used for bpf_snprintf_btf.
4563  *	Return
4564  *		0 on success or a negative error in case of failure.
4565  *
4566  * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4567  * 	Description
4568  * 		See **bpf_get_cgroup_classid**\ () for the main description.
4569  * 		This helper differs from **bpf_get_cgroup_classid**\ () in that
4570  * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s
4571  * 		associated socket instead of the current process.
4572  * 	Return
4573  * 		The id is returned or 0 in case the id could not be retrieved.
4574  *
4575  * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4576  * 	Description
4577  * 		Redirect the packet to another net device of index *ifindex*
4578  * 		and fill in L2 addresses from neighboring subsystem. This helper
4579  * 		is somewhat similar to **bpf_redirect**\ (), except that it
4580  * 		populates L2 addresses as well, meaning, internally, the helper
4581  * 		relies on the neighbor lookup for the L2 address of the nexthop.
4582  *
4583  * 		The helper will perform a FIB lookup based on the skb's
4584  * 		networking header to get the address of the next hop, unless
4585  * 		this is supplied by the caller in the *params* argument. The
4586  * 		*plen* argument indicates the len of *params* and should be set
4587  * 		to 0 if *params* is NULL.
4588  *
4589  * 		The *flags* argument is reserved and must be 0. The helper is
4590  * 		currently only supported for tc BPF program types, and enabled
4591  * 		for IPv4 and IPv6 protocols.
4592  * 	Return
4593  * 		The helper returns **TC_ACT_REDIRECT** on success or
4594  * 		**TC_ACT_SHOT** on error.
4595  *
4596  * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4597  *     Description
4598  *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4599  *             pointer to the percpu kernel variable on *cpu*. A ksym is an
4600  *             extern variable decorated with '__ksym'. For ksym, there is a
4601  *             global var (either static or global) defined of the same name
4602  *             in the kernel. The ksym is percpu if the global var is percpu.
4603  *             The returned pointer points to the global percpu var on *cpu*.
4604  *
4605  *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4606  *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
4607  *             happens if *cpu* is larger than nr_cpu_ids. The caller of
4608  *             bpf_per_cpu_ptr() must check the returned value.
4609  *     Return
4610  *             A pointer pointing to the kernel percpu variable on *cpu*, or
4611  *             NULL, if *cpu* is invalid.
4612  *
4613  * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4614  *	Description
4615  *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4616  *		pointer to the percpu kernel variable on this cpu. See the
4617  *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4618  *
4619  *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4620  *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4621  *		never return NULL.
4622  *	Return
4623  *		A pointer pointing to the kernel percpu variable on this cpu.
4624  *
4625  * long bpf_redirect_peer(u32 ifindex, u64 flags)
4626  * 	Description
4627  * 		Redirect the packet to another net device of index *ifindex*.
4628  * 		This helper is somewhat similar to **bpf_redirect**\ (), except
4629  * 		that the redirection happens to the *ifindex*' peer device and
4630  * 		the netns switch takes place from ingress to ingress without
4631  * 		going through the CPU's backlog queue.
4632  *
4633  * 		The *flags* argument is reserved and must be 0. The helper is
4634  * 		currently only supported for tc BPF program types at the ingress
4635  * 		hook and for veth device types. The peer device must reside in a
4636  * 		different network namespace.
4637  * 	Return
4638  * 		The helper returns **TC_ACT_REDIRECT** on success or
4639  * 		**TC_ACT_SHOT** on error.
4640  *
4641  * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4642  *	Description
4643  *		Get a bpf_local_storage from the *task*.
4644  *
4645  *		Logically, it could be thought of as getting the value from
4646  *		a *map* with *task* as the **key**.  From this
4647  *		perspective,  the usage is not much different from
4648  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4649  *		helper enforces the key must be an task_struct and the map must also
4650  *		be a **BPF_MAP_TYPE_TASK_STORAGE**.
4651  *
4652  *		Underneath, the value is stored locally at *task* instead of
4653  *		the *map*.  The *map* is used as the bpf-local-storage
4654  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4655  *		searched against all bpf_local_storage residing at *task*.
4656  *
4657  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4658  *		used such that a new bpf_local_storage will be
4659  *		created if one does not exist.  *value* can be used
4660  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4661  *		the initial value of a bpf_local_storage.  If *value* is
4662  *		**NULL**, the new bpf_local_storage will be zero initialized.
4663  *	Return
4664  *		A bpf_local_storage pointer is returned on success.
4665  *
4666  *		**NULL** if not found or there was an error in adding
4667  *		a new bpf_local_storage.
4668  *
4669  * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4670  *	Description
4671  *		Delete a bpf_local_storage from a *task*.
4672  *	Return
4673  *		0 on success.
4674  *
4675  *		**-ENOENT** if the bpf_local_storage cannot be found.
4676  *
4677  * struct task_struct *bpf_get_current_task_btf(void)
4678  *	Description
4679  *		Return a BTF pointer to the "current" task.
4680  *		This pointer can also be used in helpers that accept an
4681  *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.
4682  *	Return
4683  *		Pointer to the current task.
4684  *
4685  * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4686  *	Description
4687  *		Set or clear certain options on *bprm*:
4688  *
4689  *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4690  *		which sets the **AT_SECURE** auxv for glibc. The bit
4691  *		is cleared if the flag is not specified.
4692  *	Return
4693  *		**-EINVAL** if invalid *flags* are passed, zero otherwise.
4694  *
4695  * u64 bpf_ktime_get_coarse_ns(void)
4696  * 	Description
4697  * 		Return a coarse-grained version of the time elapsed since
4698  * 		system boot, in nanoseconds. Does not include time the system
4699  * 		was suspended.
4700  *
4701  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4702  * 	Return
4703  * 		Current *ktime*.
4704  *
4705  * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4706  *	Description
4707  *		Returns the stored IMA hash of the *inode* (if it's avaialable).
4708  *		If the hash is larger than *size*, then only *size*
4709  *		bytes will be copied to *dst*
4710  *	Return
4711  *		The **hash_algo** is returned on success,
4712  *		**-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4713  *		invalid arguments are passed.
4714  *
4715  * struct socket *bpf_sock_from_file(struct file *file)
4716  *	Description
4717  *		If the given file represents a socket, returns the associated
4718  *		socket.
4719  *	Return
4720  *		A pointer to a struct socket on success or NULL if the file is
4721  *		not a socket.
4722  *
4723  * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4724  *	Description
4725  *		Check packet size against exceeding MTU of net device (based
4726  *		on *ifindex*).  This helper will likely be used in combination
4727  *		with helpers that adjust/change the packet size.
4728  *
4729  *		The argument *len_diff* can be used for querying with a planned
4730  *		size change. This allows to check MTU prior to changing packet
4731  *		ctx. Providing an *len_diff* adjustment that is larger than the
4732  *		actual packet size (resulting in negative packet size) will in
4733  *		principle not exceed the MTU, why it is not considered a
4734  *		failure.  Other BPF-helpers are needed for performing the
4735  *		planned size change, why the responsability for catch a negative
4736  *		packet size belong in those helpers.
4737  *
4738  *		Specifying *ifindex* zero means the MTU check is performed
4739  *		against the current net device.  This is practical if this isn't
4740  *		used prior to redirect.
4741  *
4742  *		On input *mtu_len* must be a valid pointer, else verifier will
4743  *		reject BPF program.  If the value *mtu_len* is initialized to
4744  *		zero then the ctx packet size is use.  When value *mtu_len* is
4745  *		provided as input this specify the L3 length that the MTU check
4746  *		is done against. Remember XDP and TC length operate at L2, but
4747  *		this value is L3 as this correlate to MTU and IP-header tot_len
4748  *		values which are L3 (similar behavior as bpf_fib_lookup).
4749  *
4750  *		The Linux kernel route table can configure MTUs on a more
4751  *		specific per route level, which is not provided by this helper.
4752  *		For route level MTU checks use the **bpf_fib_lookup**\ ()
4753  *		helper.
4754  *
4755  *		*ctx* is either **struct xdp_md** for XDP programs or
4756  *		**struct sk_buff** for tc cls_act programs.
4757  *
4758  *		The *flags* argument can be a combination of one or more of the
4759  *		following values:
4760  *
4761  *		**BPF_MTU_CHK_SEGS**
4762  *			This flag will only works for *ctx* **struct sk_buff**.
4763  *			If packet context contains extra packet segment buffers
4764  *			(often knows as GSO skb), then MTU check is harder to
4765  *			check at this point, because in transmit path it is
4766  *			possible for the skb packet to get re-segmented
4767  *			(depending on net device features).  This could still be
4768  *			a MTU violation, so this flag enables performing MTU
4769  *			check against segments, with a different violation
4770  *			return code to tell it apart. Check cannot use len_diff.
4771  *
4772  *		On return *mtu_len* pointer contains the MTU value of the net
4773  *		device.  Remember the net device configured MTU is the L3 size,
4774  *		which is returned here and XDP and TC length operate at L2.
4775  *		Helper take this into account for you, but remember when using
4776  *		MTU value in your BPF-code.
4777  *
4778  *	Return
4779  *		* 0 on success, and populate MTU value in *mtu_len* pointer.
4780  *
4781  *		* < 0 if any input argument is invalid (*mtu_len* not updated)
4782  *
4783  *		MTU violations return positive values, but also populate MTU
4784  *		value in *mtu_len* pointer, as this can be needed for
4785  *		implementing PMTU handing:
4786  *
4787  *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**
4788  *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4789  *
4790  * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4791  *	Description
4792  *		For each element in **map**, call **callback_fn** function with
4793  *		**map**, **callback_ctx** and other map-specific parameters.
4794  *		The **callback_fn** should be a static function and
4795  *		the **callback_ctx** should be a pointer to the stack.
4796  *		The **flags** is used to control certain aspects of the helper.
4797  *		Currently, the **flags** must be 0.
4798  *
4799  *		The following are a list of supported map types and their
4800  *		respective expected callback signatures:
4801  *
4802  *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4803  *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4804  *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4805  *
4806  *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4807  *
4808  *		For per_cpu maps, the map_value is the value on the cpu where the
4809  *		bpf_prog is running.
4810  *
4811  *		If **callback_fn** return 0, the helper will continue to the next
4812  *		element. If return value is 1, the helper will skip the rest of
4813  *		elements and return. Other return values are not used now.
4814  *
4815  *	Return
4816  *		The number of traversed map elements for success, **-EINVAL** for
4817  *		invalid **flags**.
4818  *
4819  * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4820  *	Description
4821  *		Outputs a string into the **str** buffer of size **str_size**
4822  *		based on a format string stored in a read-only map pointed by
4823  *		**fmt**.
4824  *
4825  *		Each format specifier in **fmt** corresponds to one u64 element
4826  *		in the **data** array. For strings and pointers where pointees
4827  *		are accessed, only the pointer values are stored in the *data*
4828  *		array. The *data_len* is the size of *data* in bytes - must be
4829  *		a multiple of 8.
4830  *
4831  *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4832  *		memory. Reading kernel memory may fail due to either invalid
4833  *		address or valid address but requiring a major memory fault. If
4834  *		reading kernel memory fails, the string for **%s** will be an
4835  *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4836  *		Not returning error to bpf program is consistent with what
4837  *		**bpf_trace_printk**\ () does for now.
4838  *
4839  *	Return
4840  *		The strictly positive length of the formatted string, including
4841  *		the trailing zero character. If the return value is greater than
4842  *		**str_size**, **str** contains a truncated string, guaranteed to
4843  *		be zero-terminated except when **str_size** is 0.
4844  *
4845  *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4846  *
4847  * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
4848  * 	Description
4849  * 		Execute bpf syscall with given arguments.
4850  * 	Return
4851  * 		A syscall result.
4852  *
4853  * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
4854  * 	Description
4855  * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
4856  * 	Return
4857  * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.
4858  *
4859  * long bpf_sys_close(u32 fd)
4860  * 	Description
4861  * 		Execute close syscall for given FD.
4862  * 	Return
4863  * 		A syscall result.
4864  *
4865  * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
4866  *	Description
4867  *		Initialize the timer.
4868  *		First 4 bits of *flags* specify clockid.
4869  *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
4870  *		All other bits of *flags* are reserved.
4871  *		The verifier will reject the program if *timer* is not from
4872  *		the same *map*.
4873  *	Return
4874  *		0 on success.
4875  *		**-EBUSY** if *timer* is already initialized.
4876  *		**-EINVAL** if invalid *flags* are passed.
4877  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4878  *		The user space should either hold a file descriptor to a map with timers
4879  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4880  *		closed all timers in the map will be cancelled and freed.
4881  *
4882  * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
4883  *	Description
4884  *		Configure the timer to call *callback_fn* static function.
4885  *	Return
4886  *		0 on success.
4887  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4888  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4889  *		The user space should either hold a file descriptor to a map with timers
4890  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4891  *		closed all timers in the map will be cancelled and freed.
4892  *
4893  * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
4894  *	Description
4895  *		Set timer expiration N nanoseconds from the current time. The
4896  *		configured callback will be invoked in soft irq context on some cpu
4897  *		and will not repeat unless another bpf_timer_start() is made.
4898  *		In such case the next invocation can migrate to a different cpu.
4899  *		Since struct bpf_timer is a field inside map element the map
4900  *		owns the timer. The bpf_timer_set_callback() will increment refcnt
4901  *		of BPF program to make sure that callback_fn code stays valid.
4902  *		When user space reference to a map reaches zero all timers
4903  *		in a map are cancelled and corresponding program's refcnts are
4904  *		decremented. This is done to make sure that Ctrl-C of a user
4905  *		process doesn't leave any timers running. If map is pinned in
4906  *		bpffs the callback_fn can re-arm itself indefinitely.
4907  *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands
4908  *		cancel and free the timer in the given map element.
4909  *		The map can contain timers that invoke callback_fn-s from different
4910  *		programs. The same callback_fn can serve different timers from
4911  *		different maps if key/value layout matches across maps.
4912  *		Every bpf_timer_set_callback() can have different callback_fn.
4913  *
4914  *	Return
4915  *		0 on success.
4916  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
4917  *		or invalid *flags* are passed.
4918  *
4919  * long bpf_timer_cancel(struct bpf_timer *timer)
4920  *	Description
4921  *		Cancel the timer and wait for callback_fn to finish if it was running.
4922  *	Return
4923  *		0 if the timer was not active.
4924  *		1 if the timer was active.
4925  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4926  *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
4927  *		own timer which would have led to a deadlock otherwise.
4928  *
4929  * u64 bpf_get_func_ip(void *ctx)
4930  * 	Description
4931  * 		Get address of the traced function (for tracing and kprobe programs).
4932  * 	Return
4933  * 		Address of the traced function.
4934  *
4935  * u64 bpf_get_attach_cookie(void *ctx)
4936  * 	Description
4937  * 		Get bpf_cookie value provided (optionally) during the program
4938  * 		attachment. It might be different for each individual
4939  * 		attachment, even if BPF program itself is the same.
4940  * 		Expects BPF program context *ctx* as a first argument.
4941  *
4942  * 		Supported for the following program types:
4943  *			- kprobe/uprobe;
4944  *			- tracepoint;
4945  *			- perf_event.
4946  * 	Return
4947  *		Value specified by user at BPF link creation/attachment time
4948  *		or 0, if it was not specified.
4949  *
4950  * long bpf_task_pt_regs(struct task_struct *task)
4951  *	Description
4952  *		Get the struct pt_regs associated with **task**.
4953  *	Return
4954  *		A pointer to struct pt_regs.
4955  *
4956  * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
4957  *	Description
4958  *		Get branch trace from hardware engines like Intel LBR. The
4959  *		hardware engine is stopped shortly after the helper is
4960  *		called. Therefore, the user need to filter branch entries
4961  *		based on the actual use case. To capture branch trace
4962  *		before the trigger point of the BPF program, the helper
4963  *		should be called at the beginning of the BPF program.
4964  *
4965  *		The data is stored as struct perf_branch_entry into output
4966  *		buffer *entries*. *size* is the size of *entries* in bytes.
4967  *		*flags* is reserved for now and must be zero.
4968  *
4969  *	Return
4970  *		On success, number of bytes written to *buf*. On error, a
4971  *		negative value.
4972  *
4973  *		**-EINVAL** if *flags* is not zero.
4974  *
4975  *		**-ENOENT** if architecture does not support branch records.
4976  *
4977  * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4978  *	Description
4979  *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
4980  *		to format and can handle more format args as a result.
4981  *
4982  *		Arguments are to be used as in **bpf_seq_printf**\ () helper.
4983  *	Return
4984  *		The number of bytes written to the buffer, or a negative error
4985  *		in case of failure.
4986  *
4987  * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
4988  * 	Description
4989  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
4990  *	Return
4991  *		*sk* if casting is valid, or **NULL** otherwise.
4992  *
4993  * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
4994  *	Description
4995  *		Get the address of a kernel symbol, returned in *res*. *res* is
4996  *		set to 0 if the symbol is not found.
4997  *	Return
4998  *		On success, zero. On error, a negative value.
4999  *
5000  *		**-EINVAL** if *flags* is not zero.
5001  *
5002  *		**-EINVAL** if string *name* is not the same size as *name_sz*.
5003  *
5004  *		**-ENOENT** if symbol is not found.
5005  *
5006  *		**-EPERM** if caller does not have permission to obtain kernel address.
5007  *
5008  * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
5009  *	Description
5010  *		Find vma of *task* that contains *addr*, call *callback_fn*
5011  *		function with *task*, *vma*, and *callback_ctx*.
5012  *		The *callback_fn* should be a static function and
5013  *		the *callback_ctx* should be a pointer to the stack.
5014  *		The *flags* is used to control certain aspects of the helper.
5015  *		Currently, the *flags* must be 0.
5016  *
5017  *		The expected callback signature is
5018  *
5019  *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
5020  *
5021  *	Return
5022  *		0 on success.
5023  *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
5024  *		**-EBUSY** if failed to try lock mmap_lock.
5025  *		**-EINVAL** for invalid **flags**.
5026  *
5027  * long bpf_loop(u32 nr_loops, void *callback_fn, void *callback_ctx, u64 flags)
5028  *	Description
5029  *		For **nr_loops**, call **callback_fn** function
5030  *		with **callback_ctx** as the context parameter.
5031  *		The **callback_fn** should be a static function and
5032  *		the **callback_ctx** should be a pointer to the stack.
5033  *		The **flags** is used to control certain aspects of the helper.
5034  *		Currently, the **flags** must be 0. Currently, nr_loops is
5035  *		limited to 1 << 23 (~8 million) loops.
5036  *
5037  *		long (\*callback_fn)(u32 index, void \*ctx);
5038  *
5039  *		where **index** is the current index in the loop. The index
5040  *		is zero-indexed.
5041  *
5042  *		If **callback_fn** returns 0, the helper will continue to the next
5043  *		loop. If return value is 1, the helper will skip the rest of
5044  *		the loops and return. Other return values are not used now,
5045  *		and will be rejected by the verifier.
5046  *
5047  *	Return
5048  *		The number of loops performed, **-EINVAL** for invalid **flags**,
5049  *		**-E2BIG** if **nr_loops** exceeds the maximum number of loops.
5050  *
5051  * long bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
5052  *	Description
5053  *		Do strncmp() between **s1** and **s2**. **s1** doesn't need
5054  *		to be null-terminated and **s1_sz** is the maximum storage
5055  *		size of **s1**. **s2** must be a read-only string.
5056  *	Return
5057  *		An integer less than, equal to, or greater than zero
5058  *		if the first **s1_sz** bytes of **s1** is found to be
5059  *		less than, to match, or be greater than **s2**.
5060  *
5061  * long bpf_get_func_arg(void *ctx, u32 n, u64 *value)
5062  *	Description
5063  *		Get **n**-th argument (zero based) of the traced function (for tracing programs)
5064  *		returned in **value**.
5065  *
5066  *	Return
5067  *		0 on success.
5068  *		**-EINVAL** if n >= arguments count of traced function.
5069  *
5070  * long bpf_get_func_ret(void *ctx, u64 *value)
5071  *	Description
5072  *		Get return value of the traced function (for tracing programs)
5073  *		in **value**.
5074  *
5075  *	Return
5076  *		0 on success.
5077  *		**-EOPNOTSUPP** for tracing programs other than BPF_TRACE_FEXIT or BPF_MODIFY_RETURN.
5078  *
5079  * long bpf_get_func_arg_cnt(void *ctx)
5080  *	Description
5081  *		Get number of arguments of the traced function (for tracing programs).
5082  *
5083  *	Return
5084  *		The number of arguments of the traced function.
5085  *
5086  * int bpf_get_retval(void)
5087  *	Description
5088  *		Get the syscall's return value that will be returned to userspace.
5089  *
5090  *		This helper is currently supported by cgroup programs only.
5091  *	Return
5092  *		The syscall's return value.
5093  *
5094  * int bpf_set_retval(int retval)
5095  *	Description
5096  *		Set the syscall's return value that will be returned to userspace.
5097  *
5098  *		This helper is currently supported by cgroup programs only.
5099  *	Return
5100  *		0 on success, or a negative error in case of failure.
5101  *
5102  * u64 bpf_xdp_get_buff_len(struct xdp_buff *xdp_md)
5103  *	Description
5104  *		Get the total size of a given xdp buff (linear and paged area)
5105  *	Return
5106  *		The total size of a given xdp buffer.
5107  *
5108  * long bpf_xdp_load_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5109  *	Description
5110  *		This helper is provided as an easy way to load data from a
5111  *		xdp buffer. It can be used to load *len* bytes from *offset* from
5112  *		the frame associated to *xdp_md*, into the buffer pointed by
5113  *		*buf*.
5114  *	Return
5115  *		0 on success, or a negative error in case of failure.
5116  *
5117  * long bpf_xdp_store_bytes(struct xdp_buff *xdp_md, u32 offset, void *buf, u32 len)
5118  *	Description
5119  *		Store *len* bytes from buffer *buf* into the frame
5120  *		associated to *xdp_md*, at *offset*.
5121  *	Return
5122  *		0 on success, or a negative error in case of failure.
5123  *
5124  * long bpf_copy_from_user_task(void *dst, u32 size, const void *user_ptr, struct task_struct *tsk, u64 flags)
5125  *	Description
5126  *		Read *size* bytes from user space address *user_ptr* in *tsk*'s
5127  *		address space, and stores the data in *dst*. *flags* is not
5128  *		used yet and is provided for future extensibility. This helper
5129  *		can only be used by sleepable programs.
5130  *	Return
5131  *		0 on success, or a negative error in case of failure. On error
5132  *		*dst* buffer is zeroed out.
5133  *
5134  * long bpf_skb_set_tstamp(struct sk_buff *skb, u64 tstamp, u32 tstamp_type)
5135  *	Description
5136  *		Change the __sk_buff->tstamp_type to *tstamp_type*
5137  *		and set *tstamp* to the __sk_buff->tstamp together.
5138  *
5139  *		If there is no need to change the __sk_buff->tstamp_type,
5140  *		the tstamp value can be directly written to __sk_buff->tstamp
5141  *		instead.
5142  *
5143  *		BPF_SKB_TSTAMP_DELIVERY_MONO is the only tstamp that
5144  *		will be kept during bpf_redirect_*().  A non zero
5145  *		*tstamp* must be used with the BPF_SKB_TSTAMP_DELIVERY_MONO
5146  *		*tstamp_type*.
5147  *
5148  *		A BPF_SKB_TSTAMP_UNSPEC *tstamp_type* can only be used
5149  *		with a zero *tstamp*.
5150  *
5151  *		Only IPv4 and IPv6 skb->protocol are supported.
5152  *
5153  *		This function is most useful when it needs to set a
5154  *		mono delivery time to __sk_buff->tstamp and then
5155  *		bpf_redirect_*() to the egress of an iface.  For example,
5156  *		changing the (rcv) timestamp in __sk_buff->tstamp at
5157  *		ingress to a mono delivery time and then bpf_redirect_*()
5158  *		to sch_fq@phy-dev.
5159  *	Return
5160  *		0 on success.
5161  *		**-EINVAL** for invalid input
5162  *		**-EOPNOTSUPP** for unsupported protocol
5163  *
5164  * long bpf_ima_file_hash(struct file *file, void *dst, u32 size)
5165  *	Description
5166  *		Returns a calculated IMA hash of the *file*.
5167  *		If the hash is larger than *size*, then only *size*
5168  *		bytes will be copied to *dst*
5169  *	Return
5170  *		The **hash_algo** is returned on success,
5171  *		**-EOPNOTSUP** if the hash calculation failed or **-EINVAL** if
5172  *		invalid arguments are passed.
5173  *
5174  * void *bpf_kptr_xchg(void *map_value, void *ptr)
5175  *	Description
5176  *		Exchange kptr at pointer *map_value* with *ptr*, and return the
5177  *		old value. *ptr* can be NULL, otherwise it must be a referenced
5178  *		pointer which will be released when this helper is called.
5179  *	Return
5180  *		The old value of kptr (which can be NULL). The returned pointer
5181  *		if not NULL, is a reference which must be released using its
5182  *		corresponding release function, or moved into a BPF map before
5183  *		program exit.
5184  *
5185  * void *bpf_map_lookup_percpu_elem(struct bpf_map *map, const void *key, u32 cpu)
5186  * 	Description
5187  * 		Perform a lookup in *percpu map* for an entry associated to
5188  * 		*key* on *cpu*.
5189  * 	Return
5190  * 		Map value associated to *key* on *cpu*, or **NULL** if no entry
5191  * 		was found or *cpu* is invalid.
5192  *
5193  * struct mptcp_sock *bpf_skc_to_mptcp_sock(void *sk)
5194  *	Description
5195  *		Dynamically cast a *sk* pointer to a *mptcp_sock* pointer.
5196  *	Return
5197  *		*sk* if casting is valid, or **NULL** otherwise.
5198  *
5199  * long bpf_dynptr_from_mem(void *data, u32 size, u64 flags, struct bpf_dynptr *ptr)
5200  *	Description
5201  *		Get a dynptr to local memory *data*.
5202  *
5203  *		*data* must be a ptr to a map value.
5204  *		The maximum *size* supported is DYNPTR_MAX_SIZE.
5205  *		*flags* is currently unused.
5206  *	Return
5207  *		0 on success, -E2BIG if the size exceeds DYNPTR_MAX_SIZE,
5208  *		-EINVAL if flags is not 0.
5209  *
5210  * long bpf_ringbuf_reserve_dynptr(void *ringbuf, u32 size, u64 flags, struct bpf_dynptr *ptr)
5211  *	Description
5212  *		Reserve *size* bytes of payload in a ring buffer *ringbuf*
5213  *		through the dynptr interface. *flags* must be 0.
5214  *
5215  *		Please note that a corresponding bpf_ringbuf_submit_dynptr or
5216  *		bpf_ringbuf_discard_dynptr must be called on *ptr*, even if the
5217  *		reservation fails. This is enforced by the verifier.
5218  *	Return
5219  *		0 on success, or a negative error in case of failure.
5220  *
5221  * void bpf_ringbuf_submit_dynptr(struct bpf_dynptr *ptr, u64 flags)
5222  *	Description
5223  *		Submit reserved ring buffer sample, pointed to by *data*,
5224  *		through the dynptr interface. This is a no-op if the dynptr is
5225  *		invalid/null.
5226  *
5227  *		For more information on *flags*, please see
5228  *		'bpf_ringbuf_submit'.
5229  *	Return
5230  *		Nothing. Always succeeds.
5231  *
5232  * void bpf_ringbuf_discard_dynptr(struct bpf_dynptr *ptr, u64 flags)
5233  *	Description
5234  *		Discard reserved ring buffer sample through the dynptr
5235  *		interface. This is a no-op if the dynptr is invalid/null.
5236  *
5237  *		For more information on *flags*, please see
5238  *		'bpf_ringbuf_discard'.
5239  *	Return
5240  *		Nothing. Always succeeds.
5241  *
5242  * long bpf_dynptr_read(void *dst, u32 len, struct bpf_dynptr *src, u32 offset, u64 flags)
5243  *	Description
5244  *		Read *len* bytes from *src* into *dst*, starting from *offset*
5245  *		into *src*.
5246  *		*flags* is currently unused.
5247  *	Return
5248  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5249  *		of *src*'s data, -EINVAL if *src* is an invalid dynptr or if
5250  *		*flags* is not 0.
5251  *
5252  * long bpf_dynptr_write(struct bpf_dynptr *dst, u32 offset, void *src, u32 len, u64 flags)
5253  *	Description
5254  *		Write *len* bytes from *src* into *dst*, starting from *offset*
5255  *		into *dst*.
5256  *		*flags* is currently unused.
5257  *	Return
5258  *		0 on success, -E2BIG if *offset* + *len* exceeds the length
5259  *		of *dst*'s data, -EINVAL if *dst* is an invalid dynptr or if *dst*
5260  *		is a read-only dynptr or if *flags* is not 0.
5261  *
5262  * void *bpf_dynptr_data(struct bpf_dynptr *ptr, u32 offset, u32 len)
5263  *	Description
5264  *		Get a pointer to the underlying dynptr data.
5265  *
5266  *		*len* must be a statically known value. The returned data slice
5267  *		is invalidated whenever the dynptr is invalidated.
5268  *	Return
5269  *		Pointer to the underlying dynptr data, NULL if the dynptr is
5270  *		read-only, if the dynptr is invalid, or if the offset and length
5271  *		is out of bounds.
5272  *
5273  * s64 bpf_tcp_raw_gen_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th, u32 th_len)
5274  *	Description
5275  *		Try to issue a SYN cookie for the packet with corresponding
5276  *		IPv4/TCP headers, *iph* and *th*, without depending on a
5277  *		listening socket.
5278  *
5279  *		*iph* points to the IPv4 header.
5280  *
5281  *		*th* points to the start of the TCP header, while *th_len*
5282  *		contains the length of the TCP header (at least
5283  *		**sizeof**\ (**struct tcphdr**)).
5284  *	Return
5285  *		On success, lower 32 bits hold the generated SYN cookie in
5286  *		followed by 16 bits which hold the MSS value for that cookie,
5287  *		and the top 16 bits are unused.
5288  *
5289  *		On failure, the returned value is one of the following:
5290  *
5291  *		**-EINVAL** if *th_len* is invalid.
5292  *
5293  * s64 bpf_tcp_raw_gen_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th, u32 th_len)
5294  *	Description
5295  *		Try to issue a SYN cookie for the packet with corresponding
5296  *		IPv6/TCP headers, *iph* and *th*, without depending on a
5297  *		listening socket.
5298  *
5299  *		*iph* points to the IPv6 header.
5300  *
5301  *		*th* points to the start of the TCP header, while *th_len*
5302  *		contains the length of the TCP header (at least
5303  *		**sizeof**\ (**struct tcphdr**)).
5304  *	Return
5305  *		On success, lower 32 bits hold the generated SYN cookie in
5306  *		followed by 16 bits which hold the MSS value for that cookie,
5307  *		and the top 16 bits are unused.
5308  *
5309  *		On failure, the returned value is one of the following:
5310  *
5311  *		**-EINVAL** if *th_len* is invalid.
5312  *
5313  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5314  *
5315  * long bpf_tcp_raw_check_syncookie_ipv4(struct iphdr *iph, struct tcphdr *th)
5316  *	Description
5317  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5318  *		without depending on a listening socket.
5319  *
5320  *		*iph* points to the IPv4 header.
5321  *
5322  *		*th* points to the TCP header.
5323  *	Return
5324  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5325  *
5326  *		On failure, the returned value is one of the following:
5327  *
5328  *		**-EACCES** if the SYN cookie is not valid.
5329  *
5330  * long bpf_tcp_raw_check_syncookie_ipv6(struct ipv6hdr *iph, struct tcphdr *th)
5331  *	Description
5332  *		Check whether *iph* and *th* contain a valid SYN cookie ACK
5333  *		without depending on a listening socket.
5334  *
5335  *		*iph* points to the IPv6 header.
5336  *
5337  *		*th* points to the TCP header.
5338  *	Return
5339  *		0 if *iph* and *th* are a valid SYN cookie ACK.
5340  *
5341  *		On failure, the returned value is one of the following:
5342  *
5343  *		**-EACCES** if the SYN cookie is not valid.
5344  *
5345  *		**-EPROTONOSUPPORT** if CONFIG_IPV6 is not builtin.
5346  *
5347  * u64 bpf_ktime_get_tai_ns(void)
5348  *	Description
5349  *		A nonsettable system-wide clock derived from wall-clock time but
5350  *		ignoring leap seconds.  This clock does not experience
5351  *		discontinuities and backwards jumps caused by NTP inserting leap
5352  *		seconds as CLOCK_REALTIME does.
5353  *
5354  *		See: **clock_gettime**\ (**CLOCK_TAI**)
5355  *	Return
5356  *		Current *ktime*.
5357  *
5358  */
5359 #define __BPF_FUNC_MAPPER(FN)		\
5360 	FN(unspec),			\
5361 	FN(map_lookup_elem),		\
5362 	FN(map_update_elem),		\
5363 	FN(map_delete_elem),		\
5364 	FN(probe_read),			\
5365 	FN(ktime_get_ns),		\
5366 	FN(trace_printk),		\
5367 	FN(get_prandom_u32),		\
5368 	FN(get_smp_processor_id),	\
5369 	FN(skb_store_bytes),		\
5370 	FN(l3_csum_replace),		\
5371 	FN(l4_csum_replace),		\
5372 	FN(tail_call),			\
5373 	FN(clone_redirect),		\
5374 	FN(get_current_pid_tgid),	\
5375 	FN(get_current_uid_gid),	\
5376 	FN(get_current_comm),		\
5377 	FN(get_cgroup_classid),		\
5378 	FN(skb_vlan_push),		\
5379 	FN(skb_vlan_pop),		\
5380 	FN(skb_get_tunnel_key),		\
5381 	FN(skb_set_tunnel_key),		\
5382 	FN(perf_event_read),		\
5383 	FN(redirect),			\
5384 	FN(get_route_realm),		\
5385 	FN(perf_event_output),		\
5386 	FN(skb_load_bytes),		\
5387 	FN(get_stackid),		\
5388 	FN(csum_diff),			\
5389 	FN(skb_get_tunnel_opt),		\
5390 	FN(skb_set_tunnel_opt),		\
5391 	FN(skb_change_proto),		\
5392 	FN(skb_change_type),		\
5393 	FN(skb_under_cgroup),		\
5394 	FN(get_hash_recalc),		\
5395 	FN(get_current_task),		\
5396 	FN(probe_write_user),		\
5397 	FN(current_task_under_cgroup),	\
5398 	FN(skb_change_tail),		\
5399 	FN(skb_pull_data),		\
5400 	FN(csum_update),		\
5401 	FN(set_hash_invalid),		\
5402 	FN(get_numa_node_id),		\
5403 	FN(skb_change_head),		\
5404 	FN(xdp_adjust_head),		\
5405 	FN(probe_read_str),		\
5406 	FN(get_socket_cookie),		\
5407 	FN(get_socket_uid),		\
5408 	FN(set_hash),			\
5409 	FN(setsockopt),			\
5410 	FN(skb_adjust_room),		\
5411 	FN(redirect_map),		\
5412 	FN(sk_redirect_map),		\
5413 	FN(sock_map_update),		\
5414 	FN(xdp_adjust_meta),		\
5415 	FN(perf_event_read_value),	\
5416 	FN(perf_prog_read_value),	\
5417 	FN(getsockopt),			\
5418 	FN(override_return),		\
5419 	FN(sock_ops_cb_flags_set),	\
5420 	FN(msg_redirect_map),		\
5421 	FN(msg_apply_bytes),		\
5422 	FN(msg_cork_bytes),		\
5423 	FN(msg_pull_data),		\
5424 	FN(bind),			\
5425 	FN(xdp_adjust_tail),		\
5426 	FN(skb_get_xfrm_state),		\
5427 	FN(get_stack),			\
5428 	FN(skb_load_bytes_relative),	\
5429 	FN(fib_lookup),			\
5430 	FN(sock_hash_update),		\
5431 	FN(msg_redirect_hash),		\
5432 	FN(sk_redirect_hash),		\
5433 	FN(lwt_push_encap),		\
5434 	FN(lwt_seg6_store_bytes),	\
5435 	FN(lwt_seg6_adjust_srh),	\
5436 	FN(lwt_seg6_action),		\
5437 	FN(rc_repeat),			\
5438 	FN(rc_keydown),			\
5439 	FN(skb_cgroup_id),		\
5440 	FN(get_current_cgroup_id),	\
5441 	FN(get_local_storage),		\
5442 	FN(sk_select_reuseport),	\
5443 	FN(skb_ancestor_cgroup_id),	\
5444 	FN(sk_lookup_tcp),		\
5445 	FN(sk_lookup_udp),		\
5446 	FN(sk_release),			\
5447 	FN(map_push_elem),		\
5448 	FN(map_pop_elem),		\
5449 	FN(map_peek_elem),		\
5450 	FN(msg_push_data),		\
5451 	FN(msg_pop_data),		\
5452 	FN(rc_pointer_rel),		\
5453 	FN(spin_lock),			\
5454 	FN(spin_unlock),		\
5455 	FN(sk_fullsock),		\
5456 	FN(tcp_sock),			\
5457 	FN(skb_ecn_set_ce),		\
5458 	FN(get_listener_sock),		\
5459 	FN(skc_lookup_tcp),		\
5460 	FN(tcp_check_syncookie),	\
5461 	FN(sysctl_get_name),		\
5462 	FN(sysctl_get_current_value),	\
5463 	FN(sysctl_get_new_value),	\
5464 	FN(sysctl_set_new_value),	\
5465 	FN(strtol),			\
5466 	FN(strtoul),			\
5467 	FN(sk_storage_get),		\
5468 	FN(sk_storage_delete),		\
5469 	FN(send_signal),		\
5470 	FN(tcp_gen_syncookie),		\
5471 	FN(skb_output),			\
5472 	FN(probe_read_user),		\
5473 	FN(probe_read_kernel),		\
5474 	FN(probe_read_user_str),	\
5475 	FN(probe_read_kernel_str),	\
5476 	FN(tcp_send_ack),		\
5477 	FN(send_signal_thread),		\
5478 	FN(jiffies64),			\
5479 	FN(read_branch_records),	\
5480 	FN(get_ns_current_pid_tgid),	\
5481 	FN(xdp_output),			\
5482 	FN(get_netns_cookie),		\
5483 	FN(get_current_ancestor_cgroup_id),	\
5484 	FN(sk_assign),			\
5485 	FN(ktime_get_boot_ns),		\
5486 	FN(seq_printf),			\
5487 	FN(seq_write),			\
5488 	FN(sk_cgroup_id),		\
5489 	FN(sk_ancestor_cgroup_id),	\
5490 	FN(ringbuf_output),		\
5491 	FN(ringbuf_reserve),		\
5492 	FN(ringbuf_submit),		\
5493 	FN(ringbuf_discard),		\
5494 	FN(ringbuf_query),		\
5495 	FN(csum_level),			\
5496 	FN(skc_to_tcp6_sock),		\
5497 	FN(skc_to_tcp_sock),		\
5498 	FN(skc_to_tcp_timewait_sock),	\
5499 	FN(skc_to_tcp_request_sock),	\
5500 	FN(skc_to_udp6_sock),		\
5501 	FN(get_task_stack),		\
5502 	FN(load_hdr_opt),		\
5503 	FN(store_hdr_opt),		\
5504 	FN(reserve_hdr_opt),		\
5505 	FN(inode_storage_get),		\
5506 	FN(inode_storage_delete),	\
5507 	FN(d_path),			\
5508 	FN(copy_from_user),		\
5509 	FN(snprintf_btf),		\
5510 	FN(seq_printf_btf),		\
5511 	FN(skb_cgroup_classid),		\
5512 	FN(redirect_neigh),		\
5513 	FN(per_cpu_ptr),		\
5514 	FN(this_cpu_ptr),		\
5515 	FN(redirect_peer),		\
5516 	FN(task_storage_get),		\
5517 	FN(task_storage_delete),	\
5518 	FN(get_current_task_btf),	\
5519 	FN(bprm_opts_set),		\
5520 	FN(ktime_get_coarse_ns),	\
5521 	FN(ima_inode_hash),		\
5522 	FN(sock_from_file),		\
5523 	FN(check_mtu),			\
5524 	FN(for_each_map_elem),		\
5525 	FN(snprintf),			\
5526 	FN(sys_bpf),			\
5527 	FN(btf_find_by_name_kind),	\
5528 	FN(sys_close),			\
5529 	FN(timer_init),			\
5530 	FN(timer_set_callback),		\
5531 	FN(timer_start),		\
5532 	FN(timer_cancel),		\
5533 	FN(get_func_ip),		\
5534 	FN(get_attach_cookie),		\
5535 	FN(task_pt_regs),		\
5536 	FN(get_branch_snapshot),	\
5537 	FN(trace_vprintk),		\
5538 	FN(skc_to_unix_sock),		\
5539 	FN(kallsyms_lookup_name),	\
5540 	FN(find_vma),			\
5541 	FN(loop),			\
5542 	FN(strncmp),			\
5543 	FN(get_func_arg),		\
5544 	FN(get_func_ret),		\
5545 	FN(get_func_arg_cnt),		\
5546 	FN(get_retval),			\
5547 	FN(set_retval),			\
5548 	FN(xdp_get_buff_len),		\
5549 	FN(xdp_load_bytes),		\
5550 	FN(xdp_store_bytes),		\
5551 	FN(copy_from_user_task),	\
5552 	FN(skb_set_tstamp),		\
5553 	FN(ima_file_hash),		\
5554 	FN(kptr_xchg),			\
5555 	FN(map_lookup_percpu_elem),     \
5556 	FN(skc_to_mptcp_sock),		\
5557 	FN(dynptr_from_mem),		\
5558 	FN(ringbuf_reserve_dynptr),	\
5559 	FN(ringbuf_submit_dynptr),	\
5560 	FN(ringbuf_discard_dynptr),	\
5561 	FN(dynptr_read),		\
5562 	FN(dynptr_write),		\
5563 	FN(dynptr_data),		\
5564 	FN(tcp_raw_gen_syncookie_ipv4),	\
5565 	FN(tcp_raw_gen_syncookie_ipv6),	\
5566 	FN(tcp_raw_check_syncookie_ipv4),	\
5567 	FN(tcp_raw_check_syncookie_ipv6),	\
5568 	FN(ktime_get_tai_ns),		\
5569 	/* */
5570 
5571 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5572  * function eBPF program intends to call
5573  */
5574 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
5575 enum bpf_func_id {
5576 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5577 	__BPF_FUNC_MAX_ID,
5578 };
5579 #undef __BPF_ENUM_FN
5580 
5581 /* All flags used by eBPF helper functions, placed here. */
5582 
5583 /* BPF_FUNC_skb_store_bytes flags. */
5584 enum {
5585 	BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),
5586 	BPF_F_INVALIDATE_HASH		= (1ULL << 1),
5587 };
5588 
5589 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5590  * First 4 bits are for passing the header field size.
5591  */
5592 enum {
5593 	BPF_F_HDR_FIELD_MASK		= 0xfULL,
5594 };
5595 
5596 /* BPF_FUNC_l4_csum_replace flags. */
5597 enum {
5598 	BPF_F_PSEUDO_HDR		= (1ULL << 4),
5599 	BPF_F_MARK_MANGLED_0		= (1ULL << 5),
5600 	BPF_F_MARK_ENFORCE		= (1ULL << 6),
5601 };
5602 
5603 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5604 enum {
5605 	BPF_F_INGRESS			= (1ULL << 0),
5606 };
5607 
5608 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5609 enum {
5610 	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
5611 };
5612 
5613 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5614 enum {
5615 	BPF_F_SKIP_FIELD_MASK		= 0xffULL,
5616 	BPF_F_USER_STACK		= (1ULL << 8),
5617 /* flags used by BPF_FUNC_get_stackid only. */
5618 	BPF_F_FAST_STACK_CMP		= (1ULL << 9),
5619 	BPF_F_REUSE_STACKID		= (1ULL << 10),
5620 /* flags used by BPF_FUNC_get_stack only. */
5621 	BPF_F_USER_BUILD_ID		= (1ULL << 11),
5622 };
5623 
5624 /* BPF_FUNC_skb_set_tunnel_key flags. */
5625 enum {
5626 	BPF_F_ZERO_CSUM_TX		= (1ULL << 1),
5627 	BPF_F_DONT_FRAGMENT		= (1ULL << 2),
5628 	BPF_F_SEQ_NUMBER		= (1ULL << 3),
5629 };
5630 
5631 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5632  * BPF_FUNC_perf_event_read_value flags.
5633  */
5634 enum {
5635 	BPF_F_INDEX_MASK		= 0xffffffffULL,
5636 	BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,
5637 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5638 	BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),
5639 };
5640 
5641 /* Current network namespace */
5642 enum {
5643 	BPF_F_CURRENT_NETNS		= (-1L),
5644 };
5645 
5646 /* BPF_FUNC_csum_level level values. */
5647 enum {
5648 	BPF_CSUM_LEVEL_QUERY,
5649 	BPF_CSUM_LEVEL_INC,
5650 	BPF_CSUM_LEVEL_DEC,
5651 	BPF_CSUM_LEVEL_RESET,
5652 };
5653 
5654 /* BPF_FUNC_skb_adjust_room flags. */
5655 enum {
5656 	BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),
5657 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),
5658 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),
5659 	BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),
5660 	BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),
5661 	BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),
5662 	BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),
5663 };
5664 
5665 enum {
5666 	BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,
5667 	BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,
5668 };
5669 
5670 #define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \
5671 					  BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5672 					 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5673 
5674 /* BPF_FUNC_sysctl_get_name flags. */
5675 enum {
5676 	BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),
5677 };
5678 
5679 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5680 enum {
5681 	BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),
5682 	/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5683 	 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5684 	 */
5685 	BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
5686 };
5687 
5688 /* BPF_FUNC_read_branch_records flags. */
5689 enum {
5690 	BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),
5691 };
5692 
5693 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5694  * BPF_FUNC_bpf_ringbuf_output flags.
5695  */
5696 enum {
5697 	BPF_RB_NO_WAKEUP		= (1ULL << 0),
5698 	BPF_RB_FORCE_WAKEUP		= (1ULL << 1),
5699 };
5700 
5701 /* BPF_FUNC_bpf_ringbuf_query flags */
5702 enum {
5703 	BPF_RB_AVAIL_DATA = 0,
5704 	BPF_RB_RING_SIZE = 1,
5705 	BPF_RB_CONS_POS = 2,
5706 	BPF_RB_PROD_POS = 3,
5707 };
5708 
5709 /* BPF ring buffer constants */
5710 enum {
5711 	BPF_RINGBUF_BUSY_BIT		= (1U << 31),
5712 	BPF_RINGBUF_DISCARD_BIT		= (1U << 30),
5713 	BPF_RINGBUF_HDR_SZ		= 8,
5714 };
5715 
5716 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5717 enum {
5718 	BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),
5719 	BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),
5720 };
5721 
5722 /* Mode for BPF_FUNC_skb_adjust_room helper. */
5723 enum bpf_adj_room_mode {
5724 	BPF_ADJ_ROOM_NET,
5725 	BPF_ADJ_ROOM_MAC,
5726 };
5727 
5728 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5729 enum bpf_hdr_start_off {
5730 	BPF_HDR_START_MAC,
5731 	BPF_HDR_START_NET,
5732 };
5733 
5734 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5735 enum bpf_lwt_encap_mode {
5736 	BPF_LWT_ENCAP_SEG6,
5737 	BPF_LWT_ENCAP_SEG6_INLINE,
5738 	BPF_LWT_ENCAP_IP,
5739 };
5740 
5741 /* Flags for bpf_bprm_opts_set helper */
5742 enum {
5743 	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
5744 };
5745 
5746 /* Flags for bpf_redirect_map helper */
5747 enum {
5748 	BPF_F_BROADCAST		= (1ULL << 3),
5749 	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
5750 };
5751 
5752 #define __bpf_md_ptr(type, name)	\
5753 union {					\
5754 	type name;			\
5755 	__u64 :64;			\
5756 } __attribute__((aligned(8)))
5757 
5758 enum {
5759 	BPF_SKB_TSTAMP_UNSPEC,
5760 	BPF_SKB_TSTAMP_DELIVERY_MONO,	/* tstamp has mono delivery time */
5761 	/* For any BPF_SKB_TSTAMP_* that the bpf prog cannot handle,
5762 	 * the bpf prog should handle it like BPF_SKB_TSTAMP_UNSPEC
5763 	 * and try to deduce it by ingress, egress or skb->sk->sk_clockid.
5764 	 */
5765 };
5766 
5767 /* user accessible mirror of in-kernel sk_buff.
5768  * new fields can only be added to the end of this structure
5769  */
5770 struct __sk_buff {
5771 	__u32 len;
5772 	__u32 pkt_type;
5773 	__u32 mark;
5774 	__u32 queue_mapping;
5775 	__u32 protocol;
5776 	__u32 vlan_present;
5777 	__u32 vlan_tci;
5778 	__u32 vlan_proto;
5779 	__u32 priority;
5780 	__u32 ingress_ifindex;
5781 	__u32 ifindex;
5782 	__u32 tc_index;
5783 	__u32 cb[5];
5784 	__u32 hash;
5785 	__u32 tc_classid;
5786 	__u32 data;
5787 	__u32 data_end;
5788 	__u32 napi_id;
5789 
5790 	/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5791 	__u32 family;
5792 	__u32 remote_ip4;	/* Stored in network byte order */
5793 	__u32 local_ip4;	/* Stored in network byte order */
5794 	__u32 remote_ip6[4];	/* Stored in network byte order */
5795 	__u32 local_ip6[4];	/* Stored in network byte order */
5796 	__u32 remote_port;	/* Stored in network byte order */
5797 	__u32 local_port;	/* stored in host byte order */
5798 	/* ... here. */
5799 
5800 	__u32 data_meta;
5801 	__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5802 	__u64 tstamp;
5803 	__u32 wire_len;
5804 	__u32 gso_segs;
5805 	__bpf_md_ptr(struct bpf_sock *, sk);
5806 	__u32 gso_size;
5807 	__u8  tstamp_type;
5808 	__u32 :24;		/* Padding, future use. */
5809 	__u64 hwtstamp;
5810 };
5811 
5812 struct bpf_tunnel_key {
5813 	__u32 tunnel_id;
5814 	union {
5815 		__u32 remote_ipv4;
5816 		__u32 remote_ipv6[4];
5817 	};
5818 	__u8 tunnel_tos;
5819 	__u8 tunnel_ttl;
5820 	__u16 tunnel_ext;	/* Padding, future use. */
5821 	__u32 tunnel_label;
5822 	union {
5823 		__u32 local_ipv4;
5824 		__u32 local_ipv6[4];
5825 	};
5826 };
5827 
5828 /* user accessible mirror of in-kernel xfrm_state.
5829  * new fields can only be added to the end of this structure
5830  */
5831 struct bpf_xfrm_state {
5832 	__u32 reqid;
5833 	__u32 spi;	/* Stored in network byte order */
5834 	__u16 family;
5835 	__u16 ext;	/* Padding, future use. */
5836 	union {
5837 		__u32 remote_ipv4;	/* Stored in network byte order */
5838 		__u32 remote_ipv6[4];	/* Stored in network byte order */
5839 	};
5840 };
5841 
5842 /* Generic BPF return codes which all BPF program types may support.
5843  * The values are binary compatible with their TC_ACT_* counter-part to
5844  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5845  * programs.
5846  *
5847  * XDP is handled seprately, see XDP_*.
5848  */
5849 enum bpf_ret_code {
5850 	BPF_OK = 0,
5851 	/* 1 reserved */
5852 	BPF_DROP = 2,
5853 	/* 3-6 reserved */
5854 	BPF_REDIRECT = 7,
5855 	/* >127 are reserved for prog type specific return codes.
5856 	 *
5857 	 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5858 	 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5859 	 *    changed and should be routed based on its new L3 header.
5860 	 *    (This is an L3 redirect, as opposed to L2 redirect
5861 	 *    represented by BPF_REDIRECT above).
5862 	 */
5863 	BPF_LWT_REROUTE = 128,
5864 };
5865 
5866 struct bpf_sock {
5867 	__u32 bound_dev_if;
5868 	__u32 family;
5869 	__u32 type;
5870 	__u32 protocol;
5871 	__u32 mark;
5872 	__u32 priority;
5873 	/* IP address also allows 1 and 2 bytes access */
5874 	__u32 src_ip4;
5875 	__u32 src_ip6[4];
5876 	__u32 src_port;		/* host byte order */
5877 	__be16 dst_port;	/* network byte order */
5878 	__u16 :16;		/* zero padding */
5879 	__u32 dst_ip4;
5880 	__u32 dst_ip6[4];
5881 	__u32 state;
5882 	__s32 rx_queue_mapping;
5883 };
5884 
5885 struct bpf_tcp_sock {
5886 	__u32 snd_cwnd;		/* Sending congestion window		*/
5887 	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
5888 	__u32 rtt_min;
5889 	__u32 snd_ssthresh;	/* Slow start size threshold		*/
5890 	__u32 rcv_nxt;		/* What we want to receive next		*/
5891 	__u32 snd_nxt;		/* Next sequence we send		*/
5892 	__u32 snd_una;		/* First byte we want an ack for	*/
5893 	__u32 mss_cache;	/* Cached effective mss, not including SACKS */
5894 	__u32 ecn_flags;	/* ECN status bits.			*/
5895 	__u32 rate_delivered;	/* saved rate sample: packets delivered */
5896 	__u32 rate_interval_us;	/* saved rate sample: time elapsed */
5897 	__u32 packets_out;	/* Packets which are "in flight"	*/
5898 	__u32 retrans_out;	/* Retransmitted packets out		*/
5899 	__u32 total_retrans;	/* Total retransmits for entire connection */
5900 	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
5901 				 * total number of segments in.
5902 				 */
5903 	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
5904 				 * total number of data segments in.
5905 				 */
5906 	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
5907 				 * The total number of segments sent.
5908 				 */
5909 	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
5910 				 * total number of data segments sent.
5911 				 */
5912 	__u32 lost_out;		/* Lost packets			*/
5913 	__u32 sacked_out;	/* SACK'd packets			*/
5914 	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
5915 				 * sum(delta(rcv_nxt)), or how many bytes
5916 				 * were acked.
5917 				 */
5918 	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
5919 				 * sum(delta(snd_una)), or how many bytes
5920 				 * were acked.
5921 				 */
5922 	__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups
5923 				 * total number of DSACK blocks received
5924 				 */
5925 	__u32 delivered;	/* Total data packets delivered incl. rexmits */
5926 	__u32 delivered_ce;	/* Like the above but only ECE marked packets */
5927 	__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */
5928 };
5929 
5930 struct bpf_sock_tuple {
5931 	union {
5932 		struct {
5933 			__be32 saddr;
5934 			__be32 daddr;
5935 			__be16 sport;
5936 			__be16 dport;
5937 		} ipv4;
5938 		struct {
5939 			__be32 saddr[4];
5940 			__be32 daddr[4];
5941 			__be16 sport;
5942 			__be16 dport;
5943 		} ipv6;
5944 	};
5945 };
5946 
5947 struct bpf_xdp_sock {
5948 	__u32 queue_id;
5949 };
5950 
5951 #define XDP_PACKET_HEADROOM 256
5952 
5953 /* User return codes for XDP prog type.
5954  * A valid XDP program must return one of these defined values. All other
5955  * return codes are reserved for future use. Unknown return codes will
5956  * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5957  */
5958 enum xdp_action {
5959 	XDP_ABORTED = 0,
5960 	XDP_DROP,
5961 	XDP_PASS,
5962 	XDP_TX,
5963 	XDP_REDIRECT,
5964 };
5965 
5966 /* user accessible metadata for XDP packet hook
5967  * new fields must be added to the end of this structure
5968  */
5969 struct xdp_md {
5970 	__u32 data;
5971 	__u32 data_end;
5972 	__u32 data_meta;
5973 	/* Below access go through struct xdp_rxq_info */
5974 	__u32 ingress_ifindex; /* rxq->dev->ifindex */
5975 	__u32 rx_queue_index;  /* rxq->queue_index  */
5976 
5977 	__u32 egress_ifindex;  /* txq->dev->ifindex */
5978 };
5979 
5980 /* DEVMAP map-value layout
5981  *
5982  * The struct data-layout of map-value is a configuration interface.
5983  * New members can only be added to the end of this structure.
5984  */
5985 struct bpf_devmap_val {
5986 	__u32 ifindex;   /* device index */
5987 	union {
5988 		int   fd;  /* prog fd on map write */
5989 		__u32 id;  /* prog id on map read */
5990 	} bpf_prog;
5991 };
5992 
5993 /* CPUMAP map-value layout
5994  *
5995  * The struct data-layout of map-value is a configuration interface.
5996  * New members can only be added to the end of this structure.
5997  */
5998 struct bpf_cpumap_val {
5999 	__u32 qsize;	/* queue size to remote target CPU */
6000 	union {
6001 		int   fd;	/* prog fd on map write */
6002 		__u32 id;	/* prog id on map read */
6003 	} bpf_prog;
6004 };
6005 
6006 enum sk_action {
6007 	SK_DROP = 0,
6008 	SK_PASS,
6009 };
6010 
6011 /* user accessible metadata for SK_MSG packet hook, new fields must
6012  * be added to the end of this structure
6013  */
6014 struct sk_msg_md {
6015 	__bpf_md_ptr(void *, data);
6016 	__bpf_md_ptr(void *, data_end);
6017 
6018 	__u32 family;
6019 	__u32 remote_ip4;	/* Stored in network byte order */
6020 	__u32 local_ip4;	/* Stored in network byte order */
6021 	__u32 remote_ip6[4];	/* Stored in network byte order */
6022 	__u32 local_ip6[4];	/* Stored in network byte order */
6023 	__u32 remote_port;	/* Stored in network byte order */
6024 	__u32 local_port;	/* stored in host byte order */
6025 	__u32 size;		/* Total size of sk_msg */
6026 
6027 	__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
6028 };
6029 
6030 struct sk_reuseport_md {
6031 	/*
6032 	 * Start of directly accessible data. It begins from
6033 	 * the tcp/udp header.
6034 	 */
6035 	__bpf_md_ptr(void *, data);
6036 	/* End of directly accessible data */
6037 	__bpf_md_ptr(void *, data_end);
6038 	/*
6039 	 * Total length of packet (starting from the tcp/udp header).
6040 	 * Note that the directly accessible bytes (data_end - data)
6041 	 * could be less than this "len".  Those bytes could be
6042 	 * indirectly read by a helper "bpf_skb_load_bytes()".
6043 	 */
6044 	__u32 len;
6045 	/*
6046 	 * Eth protocol in the mac header (network byte order). e.g.
6047 	 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
6048 	 */
6049 	__u32 eth_protocol;
6050 	__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
6051 	__u32 bind_inany;	/* Is sock bound to an INANY address? */
6052 	__u32 hash;		/* A hash of the packet 4 tuples */
6053 	/* When reuse->migrating_sk is NULL, it is selecting a sk for the
6054 	 * new incoming connection request (e.g. selecting a listen sk for
6055 	 * the received SYN in the TCP case).  reuse->sk is one of the sk
6056 	 * in the reuseport group. The bpf prog can use reuse->sk to learn
6057 	 * the local listening ip/port without looking into the skb.
6058 	 *
6059 	 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
6060 	 * reuse->migrating_sk is the socket that needs to be migrated
6061 	 * to another listening socket.  migrating_sk could be a fullsock
6062 	 * sk that is fully established or a reqsk that is in-the-middle
6063 	 * of 3-way handshake.
6064 	 */
6065 	__bpf_md_ptr(struct bpf_sock *, sk);
6066 	__bpf_md_ptr(struct bpf_sock *, migrating_sk);
6067 };
6068 
6069 #define BPF_TAG_SIZE	8
6070 
6071 struct bpf_prog_info {
6072 	__u32 type;
6073 	__u32 id;
6074 	__u8  tag[BPF_TAG_SIZE];
6075 	__u32 jited_prog_len;
6076 	__u32 xlated_prog_len;
6077 	__aligned_u64 jited_prog_insns;
6078 	__aligned_u64 xlated_prog_insns;
6079 	__u64 load_time;	/* ns since boottime */
6080 	__u32 created_by_uid;
6081 	__u32 nr_map_ids;
6082 	__aligned_u64 map_ids;
6083 	char name[BPF_OBJ_NAME_LEN];
6084 	__u32 ifindex;
6085 	__u32 gpl_compatible:1;
6086 	__u32 :31; /* alignment pad */
6087 	__u64 netns_dev;
6088 	__u64 netns_ino;
6089 	__u32 nr_jited_ksyms;
6090 	__u32 nr_jited_func_lens;
6091 	__aligned_u64 jited_ksyms;
6092 	__aligned_u64 jited_func_lens;
6093 	__u32 btf_id;
6094 	__u32 func_info_rec_size;
6095 	__aligned_u64 func_info;
6096 	__u32 nr_func_info;
6097 	__u32 nr_line_info;
6098 	__aligned_u64 line_info;
6099 	__aligned_u64 jited_line_info;
6100 	__u32 nr_jited_line_info;
6101 	__u32 line_info_rec_size;
6102 	__u32 jited_line_info_rec_size;
6103 	__u32 nr_prog_tags;
6104 	__aligned_u64 prog_tags;
6105 	__u64 run_time_ns;
6106 	__u64 run_cnt;
6107 	__u64 recursion_misses;
6108 	__u32 verified_insns;
6109 	__u32 attach_btf_obj_id;
6110 	__u32 attach_btf_id;
6111 } __attribute__((aligned(8)));
6112 
6113 struct bpf_map_info {
6114 	__u32 type;
6115 	__u32 id;
6116 	__u32 key_size;
6117 	__u32 value_size;
6118 	__u32 max_entries;
6119 	__u32 map_flags;
6120 	char  name[BPF_OBJ_NAME_LEN];
6121 	__u32 ifindex;
6122 	__u32 btf_vmlinux_value_type_id;
6123 	__u64 netns_dev;
6124 	__u64 netns_ino;
6125 	__u32 btf_id;
6126 	__u32 btf_key_type_id;
6127 	__u32 btf_value_type_id;
6128 	__u32 :32;	/* alignment pad */
6129 	__u64 map_extra;
6130 } __attribute__((aligned(8)));
6131 
6132 struct bpf_btf_info {
6133 	__aligned_u64 btf;
6134 	__u32 btf_size;
6135 	__u32 id;
6136 	__aligned_u64 name;
6137 	__u32 name_len;
6138 	__u32 kernel_btf;
6139 } __attribute__((aligned(8)));
6140 
6141 struct bpf_link_info {
6142 	__u32 type;
6143 	__u32 id;
6144 	__u32 prog_id;
6145 	union {
6146 		struct {
6147 			__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
6148 			__u32 tp_name_len;     /* in/out: tp_name buffer len */
6149 		} raw_tracepoint;
6150 		struct {
6151 			__u32 attach_type;
6152 			__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
6153 			__u32 target_btf_id; /* BTF type id inside the object */
6154 		} tracing;
6155 		struct {
6156 			__u64 cgroup_id;
6157 			__u32 attach_type;
6158 		} cgroup;
6159 		struct {
6160 			__aligned_u64 target_name; /* in/out: target_name buffer ptr */
6161 			__u32 target_name_len;	   /* in/out: target_name buffer len */
6162 			union {
6163 				struct {
6164 					__u32 map_id;
6165 				} map;
6166 			};
6167 		} iter;
6168 		struct  {
6169 			__u32 netns_ino;
6170 			__u32 attach_type;
6171 		} netns;
6172 		struct {
6173 			__u32 ifindex;
6174 		} xdp;
6175 	};
6176 } __attribute__((aligned(8)));
6177 
6178 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
6179  * by user and intended to be used by socket (e.g. to bind to, depends on
6180  * attach type).
6181  */
6182 struct bpf_sock_addr {
6183 	__u32 user_family;	/* Allows 4-byte read, but no write. */
6184 	__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.
6185 				 * Stored in network byte order.
6186 				 */
6187 	__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6188 				 * Stored in network byte order.
6189 				 */
6190 	__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.
6191 				 * Stored in network byte order
6192 				 */
6193 	__u32 family;		/* Allows 4-byte read, but no write */
6194 	__u32 type;		/* Allows 4-byte read, but no write */
6195 	__u32 protocol;		/* Allows 4-byte read, but no write */
6196 	__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.
6197 				 * Stored in network byte order.
6198 				 */
6199 	__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
6200 				 * Stored in network byte order.
6201 				 */
6202 	__bpf_md_ptr(struct bpf_sock *, sk);
6203 };
6204 
6205 /* User bpf_sock_ops struct to access socket values and specify request ops
6206  * and their replies.
6207  * Some of this fields are in network (bigendian) byte order and may need
6208  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
6209  * New fields can only be added at the end of this structure
6210  */
6211 struct bpf_sock_ops {
6212 	__u32 op;
6213 	union {
6214 		__u32 args[4];		/* Optionally passed to bpf program */
6215 		__u32 reply;		/* Returned by bpf program	    */
6216 		__u32 replylong[4];	/* Optionally returned by bpf prog  */
6217 	};
6218 	__u32 family;
6219 	__u32 remote_ip4;	/* Stored in network byte order */
6220 	__u32 local_ip4;	/* Stored in network byte order */
6221 	__u32 remote_ip6[4];	/* Stored in network byte order */
6222 	__u32 local_ip6[4];	/* Stored in network byte order */
6223 	__u32 remote_port;	/* Stored in network byte order */
6224 	__u32 local_port;	/* stored in host byte order */
6225 	__u32 is_fullsock;	/* Some TCP fields are only valid if
6226 				 * there is a full socket. If not, the
6227 				 * fields read as zero.
6228 				 */
6229 	__u32 snd_cwnd;
6230 	__u32 srtt_us;		/* Averaged RTT << 3 in usecs */
6231 	__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
6232 	__u32 state;
6233 	__u32 rtt_min;
6234 	__u32 snd_ssthresh;
6235 	__u32 rcv_nxt;
6236 	__u32 snd_nxt;
6237 	__u32 snd_una;
6238 	__u32 mss_cache;
6239 	__u32 ecn_flags;
6240 	__u32 rate_delivered;
6241 	__u32 rate_interval_us;
6242 	__u32 packets_out;
6243 	__u32 retrans_out;
6244 	__u32 total_retrans;
6245 	__u32 segs_in;
6246 	__u32 data_segs_in;
6247 	__u32 segs_out;
6248 	__u32 data_segs_out;
6249 	__u32 lost_out;
6250 	__u32 sacked_out;
6251 	__u32 sk_txhash;
6252 	__u64 bytes_received;
6253 	__u64 bytes_acked;
6254 	__bpf_md_ptr(struct bpf_sock *, sk);
6255 	/* [skb_data, skb_data_end) covers the whole TCP header.
6256 	 *
6257 	 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
6258 	 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
6259 	 *                                header has not been written.
6260 	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
6261 	 *				  been written so far.
6262 	 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
6263 	 *					the 3WHS.
6264 	 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
6265 	 *					the 3WHS.
6266 	 *
6267 	 * bpf_load_hdr_opt() can also be used to read a particular option.
6268 	 */
6269 	__bpf_md_ptr(void *, skb_data);
6270 	__bpf_md_ptr(void *, skb_data_end);
6271 	__u32 skb_len;		/* The total length of a packet.
6272 				 * It includes the header, options,
6273 				 * and payload.
6274 				 */
6275 	__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides
6276 				 * an easy way to check for tcp_flags
6277 				 * without parsing skb_data.
6278 				 *
6279 				 * In particular, the skb_tcp_flags
6280 				 * will still be available in
6281 				 * BPF_SOCK_OPS_HDR_OPT_LEN even though
6282 				 * the outgoing header has not
6283 				 * been written yet.
6284 				 */
6285 };
6286 
6287 /* Definitions for bpf_sock_ops_cb_flags */
6288 enum {
6289 	BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),
6290 	BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),
6291 	BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),
6292 	BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),
6293 	/* Call bpf for all received TCP headers.  The bpf prog will be
6294 	 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6295 	 *
6296 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6297 	 * for the header option related helpers that will be useful
6298 	 * to the bpf programs.
6299 	 *
6300 	 * It could be used at the client/active side (i.e. connect() side)
6301 	 * when the server told it that the server was in syncookie
6302 	 * mode and required the active side to resend the bpf-written
6303 	 * options.  The active side can keep writing the bpf-options until
6304 	 * it received a valid packet from the server side to confirm
6305 	 * the earlier packet (and options) has been received.  The later
6306 	 * example patch is using it like this at the active side when the
6307 	 * server is in syncookie mode.
6308 	 *
6309 	 * The bpf prog will usually turn this off in the common cases.
6310 	 */
6311 	BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),
6312 	/* Call bpf when kernel has received a header option that
6313 	 * the kernel cannot handle.  The bpf prog will be called under
6314 	 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
6315 	 *
6316 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
6317 	 * for the header option related helpers that will be useful
6318 	 * to the bpf programs.
6319 	 */
6320 	BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
6321 	/* Call bpf when the kernel is writing header options for the
6322 	 * outgoing packet.  The bpf prog will first be called
6323 	 * to reserve space in a skb under
6324 	 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
6325 	 * the bpf prog will be called to write the header option(s)
6326 	 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6327 	 *
6328 	 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
6329 	 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
6330 	 * related helpers that will be useful to the bpf programs.
6331 	 *
6332 	 * The kernel gets its chance to reserve space and write
6333 	 * options first before the BPF program does.
6334 	 */
6335 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
6336 /* Mask of all currently supported cb flags */
6337 	BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
6338 };
6339 
6340 /* List of known BPF sock_ops operators.
6341  * New entries can only be added at the end
6342  */
6343 enum {
6344 	BPF_SOCK_OPS_VOID,
6345 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
6346 					 * -1 if default value should be used
6347 					 */
6348 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
6349 					 * window (in packets) or -1 if default
6350 					 * value should be used
6351 					 */
6352 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
6353 					 * active connection is initialized
6354 					 */
6355 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
6356 						 * active connection is
6357 						 * established
6358 						 */
6359 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
6360 						 * passive connection is
6361 						 * established
6362 						 */
6363 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
6364 					 * needs ECN
6365 					 */
6366 	BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is
6367 					 * based on the path and may be
6368 					 * dependent on the congestion control
6369 					 * algorithm. In general it indicates
6370 					 * a congestion threshold. RTTs above
6371 					 * this indicate congestion
6372 					 */
6373 	BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.
6374 					 * Arg1: value of icsk_retransmits
6375 					 * Arg2: value of icsk_rto
6376 					 * Arg3: whether RTO has expired
6377 					 */
6378 	BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.
6379 					 * Arg1: sequence number of 1st byte
6380 					 * Arg2: # segments
6381 					 * Arg3: return value of
6382 					 *       tcp_transmit_skb (0 => success)
6383 					 */
6384 	BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.
6385 					 * Arg1: old_state
6386 					 * Arg2: new_state
6387 					 */
6388 	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
6389 					 * socket transition to LISTEN state.
6390 					 */
6391 	BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.
6392 					 */
6393 	BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.
6394 					 * It will be called to handle
6395 					 * the packets received at
6396 					 * an already established
6397 					 * connection.
6398 					 *
6399 					 * sock_ops->skb_data:
6400 					 * Referring to the received skb.
6401 					 * It covers the TCP header only.
6402 					 *
6403 					 * bpf_load_hdr_opt() can also
6404 					 * be used to search for a
6405 					 * particular option.
6406 					 */
6407 	BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the
6408 					 * header option later in
6409 					 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6410 					 * Arg1: bool want_cookie. (in
6411 					 *       writing SYNACK only)
6412 					 *
6413 					 * sock_ops->skb_data:
6414 					 * Not available because no header has
6415 					 * been	written yet.
6416 					 *
6417 					 * sock_ops->skb_tcp_flags:
6418 					 * The tcp_flags of the
6419 					 * outgoing skb. (e.g. SYN, ACK, FIN).
6420 					 *
6421 					 * bpf_reserve_hdr_opt() should
6422 					 * be used to reserve space.
6423 					 */
6424 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options
6425 					 * Arg1: bool want_cookie. (in
6426 					 *       writing SYNACK only)
6427 					 *
6428 					 * sock_ops->skb_data:
6429 					 * Referring to the outgoing skb.
6430 					 * It covers the TCP header
6431 					 * that has already been written
6432 					 * by the kernel and the
6433 					 * earlier bpf-progs.
6434 					 *
6435 					 * sock_ops->skb_tcp_flags:
6436 					 * The tcp_flags of the outgoing
6437 					 * skb. (e.g. SYN, ACK, FIN).
6438 					 *
6439 					 * bpf_store_hdr_opt() should
6440 					 * be used to write the
6441 					 * option.
6442 					 *
6443 					 * bpf_load_hdr_opt() can also
6444 					 * be used to search for a
6445 					 * particular option that
6446 					 * has already been written
6447 					 * by the kernel or the
6448 					 * earlier bpf-progs.
6449 					 */
6450 };
6451 
6452 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
6453  * changes between the TCP and BPF versions. Ideally this should never happen.
6454  * If it does, we need to add code to convert them before calling
6455  * the BPF sock_ops function.
6456  */
6457 enum {
6458 	BPF_TCP_ESTABLISHED = 1,
6459 	BPF_TCP_SYN_SENT,
6460 	BPF_TCP_SYN_RECV,
6461 	BPF_TCP_FIN_WAIT1,
6462 	BPF_TCP_FIN_WAIT2,
6463 	BPF_TCP_TIME_WAIT,
6464 	BPF_TCP_CLOSE,
6465 	BPF_TCP_CLOSE_WAIT,
6466 	BPF_TCP_LAST_ACK,
6467 	BPF_TCP_LISTEN,
6468 	BPF_TCP_CLOSING,	/* Now a valid state */
6469 	BPF_TCP_NEW_SYN_RECV,
6470 
6471 	BPF_TCP_MAX_STATES	/* Leave at the end! */
6472 };
6473 
6474 enum {
6475 	TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */
6476 	TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */
6477 	TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */
6478 	TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */
6479 	/* Copy the SYN pkt to optval
6480 	 *
6481 	 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
6482 	 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6483 	 * to only getting from the saved_syn.  It can either get the
6484 	 * syn packet from:
6485 	 *
6486 	 * 1. the just-received SYN packet (only available when writing the
6487 	 *    SYNACK).  It will be useful when it is not necessary to
6488 	 *    save the SYN packet for latter use.  It is also the only way
6489 	 *    to get the SYN during syncookie mode because the syn
6490 	 *    packet cannot be saved during syncookie.
6491 	 *
6492 	 * OR
6493 	 *
6494 	 * 2. the earlier saved syn which was done by
6495 	 *    bpf_setsockopt(TCP_SAVE_SYN).
6496 	 *
6497 	 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6498 	 * SYN packet is obtained.
6499 	 *
6500 	 * If the bpf-prog does not need the IP[46] header,  the
6501 	 * bpf-prog can avoid parsing the IP header by using
6502 	 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
6503 	 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6504 	 *
6505 	 *      >0: Total number of bytes copied
6506 	 * -ENOSPC: Not enough space in optval. Only optlen number of
6507 	 *          bytes is copied.
6508 	 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6509 	 *	    is not saved by setsockopt(TCP_SAVE_SYN).
6510 	 */
6511 	TCP_BPF_SYN		= 1005, /* Copy the TCP header */
6512 	TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */
6513 	TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
6514 };
6515 
6516 enum {
6517 	BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6518 };
6519 
6520 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6521  * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6522  */
6523 enum {
6524 	BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the
6525 						 * total option spaces
6526 						 * required for an established
6527 						 * sk in order to calculate the
6528 						 * MSS.  No skb is actually
6529 						 * sent.
6530 						 */
6531 	BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode
6532 						 * when sending a SYN.
6533 						 */
6534 };
6535 
6536 struct bpf_perf_event_value {
6537 	__u64 counter;
6538 	__u64 enabled;
6539 	__u64 running;
6540 };
6541 
6542 enum {
6543 	BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),
6544 	BPF_DEVCG_ACC_READ	= (1ULL << 1),
6545 	BPF_DEVCG_ACC_WRITE	= (1ULL << 2),
6546 };
6547 
6548 enum {
6549 	BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),
6550 	BPF_DEVCG_DEV_CHAR	= (1ULL << 1),
6551 };
6552 
6553 struct bpf_cgroup_dev_ctx {
6554 	/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6555 	__u32 access_type;
6556 	__u32 major;
6557 	__u32 minor;
6558 };
6559 
6560 struct bpf_raw_tracepoint_args {
6561 	__u64 args[0];
6562 };
6563 
6564 /* DIRECT:  Skip the FIB rules and go to FIB table associated with device
6565  * OUTPUT:  Do lookup from egress perspective; default is ingress
6566  */
6567 enum {
6568 	BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
6569 	BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
6570 };
6571 
6572 enum {
6573 	BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
6574 	BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
6575 	BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
6576 	BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
6577 	BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
6578 	BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6579 	BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
6580 	BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
6581 	BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6582 };
6583 
6584 struct bpf_fib_lookup {
6585 	/* input:  network family for lookup (AF_INET, AF_INET6)
6586 	 * output: network family of egress nexthop
6587 	 */
6588 	__u8	family;
6589 
6590 	/* set if lookup is to consider L4 data - e.g., FIB rules */
6591 	__u8	l4_protocol;
6592 	__be16	sport;
6593 	__be16	dport;
6594 
6595 	union {	/* used for MTU check */
6596 		/* input to lookup */
6597 		__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */
6598 
6599 		/* output: MTU value */
6600 		__u16	mtu_result;
6601 	};
6602 	/* input: L3 device index for lookup
6603 	 * output: device index from FIB lookup
6604 	 */
6605 	__u32	ifindex;
6606 
6607 	union {
6608 		/* inputs to lookup */
6609 		__u8	tos;		/* AF_INET  */
6610 		__be32	flowinfo;	/* AF_INET6, flow_label + priority */
6611 
6612 		/* output: metric of fib result (IPv4/IPv6 only) */
6613 		__u32	rt_metric;
6614 	};
6615 
6616 	union {
6617 		__be32		ipv4_src;
6618 		__u32		ipv6_src[4];  /* in6_addr; network order */
6619 	};
6620 
6621 	/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
6622 	 * network header. output: bpf_fib_lookup sets to gateway address
6623 	 * if FIB lookup returns gateway route
6624 	 */
6625 	union {
6626 		__be32		ipv4_dst;
6627 		__u32		ipv6_dst[4];  /* in6_addr; network order */
6628 	};
6629 
6630 	/* output */
6631 	__be16	h_vlan_proto;
6632 	__be16	h_vlan_TCI;
6633 	__u8	smac[6];     /* ETH_ALEN */
6634 	__u8	dmac[6];     /* ETH_ALEN */
6635 };
6636 
6637 struct bpf_redir_neigh {
6638 	/* network family for lookup (AF_INET, AF_INET6) */
6639 	__u32 nh_family;
6640 	/* network address of nexthop; skips fib lookup to find gateway */
6641 	union {
6642 		__be32		ipv4_nh;
6643 		__u32		ipv6_nh[4];  /* in6_addr; network order */
6644 	};
6645 };
6646 
6647 /* bpf_check_mtu flags*/
6648 enum  bpf_check_mtu_flags {
6649 	BPF_MTU_CHK_SEGS  = (1U << 0),
6650 };
6651 
6652 enum bpf_check_mtu_ret {
6653 	BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
6654 	BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6655 	BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
6656 };
6657 
6658 enum bpf_task_fd_type {
6659 	BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */
6660 	BPF_FD_TYPE_TRACEPOINT,		/* tp name */
6661 	BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */
6662 	BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */
6663 	BPF_FD_TYPE_UPROBE,		/* filename + offset */
6664 	BPF_FD_TYPE_URETPROBE,		/* filename + offset */
6665 };
6666 
6667 enum {
6668 	BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),
6669 	BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),
6670 	BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),
6671 };
6672 
6673 struct bpf_flow_keys {
6674 	__u16	nhoff;
6675 	__u16	thoff;
6676 	__u16	addr_proto;			/* ETH_P_* of valid addrs */
6677 	__u8	is_frag;
6678 	__u8	is_first_frag;
6679 	__u8	is_encap;
6680 	__u8	ip_proto;
6681 	__be16	n_proto;
6682 	__be16	sport;
6683 	__be16	dport;
6684 	union {
6685 		struct {
6686 			__be32	ipv4_src;
6687 			__be32	ipv4_dst;
6688 		};
6689 		struct {
6690 			__u32	ipv6_src[4];	/* in6_addr; network order */
6691 			__u32	ipv6_dst[4];	/* in6_addr; network order */
6692 		};
6693 	};
6694 	__u32	flags;
6695 	__be32	flow_label;
6696 };
6697 
6698 struct bpf_func_info {
6699 	__u32	insn_off;
6700 	__u32	type_id;
6701 };
6702 
6703 #define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)
6704 #define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)
6705 
6706 struct bpf_line_info {
6707 	__u32	insn_off;
6708 	__u32	file_name_off;
6709 	__u32	line_off;
6710 	__u32	line_col;
6711 };
6712 
6713 struct bpf_spin_lock {
6714 	__u32	val;
6715 };
6716 
6717 struct bpf_timer {
6718 	__u64 :64;
6719 	__u64 :64;
6720 } __attribute__((aligned(8)));
6721 
6722 struct bpf_dynptr {
6723 	__u64 :64;
6724 	__u64 :64;
6725 } __attribute__((aligned(8)));
6726 
6727 struct bpf_sysctl {
6728 	__u32	write;		/* Sysctl is being read (= 0) or written (= 1).
6729 				 * Allows 1,2,4-byte read, but no write.
6730 				 */
6731 	__u32	file_pos;	/* Sysctl file position to read from, write to.
6732 				 * Allows 1,2,4-byte read an 4-byte write.
6733 				 */
6734 };
6735 
6736 struct bpf_sockopt {
6737 	__bpf_md_ptr(struct bpf_sock *, sk);
6738 	__bpf_md_ptr(void *, optval);
6739 	__bpf_md_ptr(void *, optval_end);
6740 
6741 	__s32	level;
6742 	__s32	optname;
6743 	__s32	optlen;
6744 	__s32	retval;
6745 };
6746 
6747 struct bpf_pidns_info {
6748 	__u32 pid;
6749 	__u32 tgid;
6750 };
6751 
6752 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6753 struct bpf_sk_lookup {
6754 	union {
6755 		__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6756 		__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6757 	};
6758 
6759 	__u32 family;		/* Protocol family (AF_INET, AF_INET6) */
6760 	__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6761 	__u32 remote_ip4;	/* Network byte order */
6762 	__u32 remote_ip6[4];	/* Network byte order */
6763 	__be16 remote_port;	/* Network byte order */
6764 	__u16 :16;		/* Zero padding */
6765 	__u32 local_ip4;	/* Network byte order */
6766 	__u32 local_ip6[4];	/* Network byte order */
6767 	__u32 local_port;	/* Host byte order */
6768 	__u32 ingress_ifindex;		/* The arriving interface. Determined by inet_iif. */
6769 };
6770 
6771 /*
6772  * struct btf_ptr is used for typed pointer representation; the
6773  * type id is used to render the pointer data as the appropriate type
6774  * via the bpf_snprintf_btf() helper described above.  A flags field -
6775  * potentially to specify additional details about the BTF pointer
6776  * (rather than its mode of display) - is included for future use.
6777  * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6778  */
6779 struct btf_ptr {
6780 	void *ptr;
6781 	__u32 type_id;
6782 	__u32 flags;		/* BTF ptr flags; unused at present. */
6783 };
6784 
6785 /*
6786  * Flags to control bpf_snprintf_btf() behaviour.
6787  *     - BTF_F_COMPACT: no formatting around type information
6788  *     - BTF_F_NONAME: no struct/union member names/types
6789  *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6790  *       equivalent to %px.
6791  *     - BTF_F_ZERO: show zero-valued struct/union members; they
6792  *       are not displayed by default
6793  */
6794 enum {
6795 	BTF_F_COMPACT	=	(1ULL << 0),
6796 	BTF_F_NONAME	=	(1ULL << 1),
6797 	BTF_F_PTR_RAW	=	(1ULL << 2),
6798 	BTF_F_ZERO	=	(1ULL << 3),
6799 };
6800 
6801 /* bpf_core_relo_kind encodes which aspect of captured field/type/enum value
6802  * has to be adjusted by relocations. It is emitted by llvm and passed to
6803  * libbpf and later to the kernel.
6804  */
6805 enum bpf_core_relo_kind {
6806 	BPF_CORE_FIELD_BYTE_OFFSET = 0,      /* field byte offset */
6807 	BPF_CORE_FIELD_BYTE_SIZE = 1,        /* field size in bytes */
6808 	BPF_CORE_FIELD_EXISTS = 2,           /* field existence in target kernel */
6809 	BPF_CORE_FIELD_SIGNED = 3,           /* field signedness (0 - unsigned, 1 - signed) */
6810 	BPF_CORE_FIELD_LSHIFT_U64 = 4,       /* bitfield-specific left bitshift */
6811 	BPF_CORE_FIELD_RSHIFT_U64 = 5,       /* bitfield-specific right bitshift */
6812 	BPF_CORE_TYPE_ID_LOCAL = 6,          /* type ID in local BPF object */
6813 	BPF_CORE_TYPE_ID_TARGET = 7,         /* type ID in target kernel */
6814 	BPF_CORE_TYPE_EXISTS = 8,            /* type existence in target kernel */
6815 	BPF_CORE_TYPE_SIZE = 9,              /* type size in bytes */
6816 	BPF_CORE_ENUMVAL_EXISTS = 10,        /* enum value existence in target kernel */
6817 	BPF_CORE_ENUMVAL_VALUE = 11,         /* enum value integer value */
6818 	BPF_CORE_TYPE_MATCHES = 12,          /* type match in target kernel */
6819 };
6820 
6821 /*
6822  * "struct bpf_core_relo" is used to pass relocation data form LLVM to libbpf
6823  * and from libbpf to the kernel.
6824  *
6825  * CO-RE relocation captures the following data:
6826  * - insn_off - instruction offset (in bytes) within a BPF program that needs
6827  *   its insn->imm field to be relocated with actual field info;
6828  * - type_id - BTF type ID of the "root" (containing) entity of a relocatable
6829  *   type or field;
6830  * - access_str_off - offset into corresponding .BTF string section. String
6831  *   interpretation depends on specific relocation kind:
6832  *     - for field-based relocations, string encodes an accessed field using
6833  *       a sequence of field and array indices, separated by colon (:). It's
6834  *       conceptually very close to LLVM's getelementptr ([0]) instruction's
6835  *       arguments for identifying offset to a field.
6836  *     - for type-based relocations, strings is expected to be just "0";
6837  *     - for enum value-based relocations, string contains an index of enum
6838  *       value within its enum type;
6839  * - kind - one of enum bpf_core_relo_kind;
6840  *
6841  * Example:
6842  *   struct sample {
6843  *       int a;
6844  *       struct {
6845  *           int b[10];
6846  *       };
6847  *   };
6848  *
6849  *   struct sample *s = ...;
6850  *   int *x = &s->a;     // encoded as "0:0" (a is field #0)
6851  *   int *y = &s->b[5];  // encoded as "0:1:0:5" (anon struct is field #1,
6852  *                       // b is field #0 inside anon struct, accessing elem #5)
6853  *   int *z = &s[10]->b; // encoded as "10:1" (ptr is used as an array)
6854  *
6855  * type_id for all relocs in this example will capture BTF type id of
6856  * `struct sample`.
6857  *
6858  * Such relocation is emitted when using __builtin_preserve_access_index()
6859  * Clang built-in, passing expression that captures field address, e.g.:
6860  *
6861  * bpf_probe_read(&dst, sizeof(dst),
6862  *		  __builtin_preserve_access_index(&src->a.b.c));
6863  *
6864  * In this case Clang will emit field relocation recording necessary data to
6865  * be able to find offset of embedded `a.b.c` field within `src` struct.
6866  *
6867  * [0] https://llvm.org/docs/LangRef.html#getelementptr-instruction
6868  */
6869 struct bpf_core_relo {
6870 	__u32 insn_off;
6871 	__u32 type_id;
6872 	__u32 access_str_off;
6873 	enum bpf_core_relo_kind kind;
6874 };
6875 
6876 #endif /* _UAPI__LINUX_BPF_H__ */
6877