1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of version 2 of the GNU General Public
6  * License as published by the Free Software Foundation.
7  */
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
10 
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
13 
14 /* Extended instruction set based on top of classic BPF */
15 
16 /* instruction classes */
17 #define BPF_JMP32	0x06	/* jmp mode in word width */
18 #define BPF_ALU64	0x07	/* alu mode in double word width */
19 
20 /* ld/ldx fields */
21 #define BPF_DW		0x18	/* double word (64-bit) */
22 #define BPF_ATOMIC	0xc0	/* atomic memory ops - op type in immediate */
23 #define BPF_XADD	0xc0	/* exclusive add - legacy name */
24 
25 /* alu/jmp fields */
26 #define BPF_MOV		0xb0	/* mov reg to reg */
27 #define BPF_ARSH	0xc0	/* sign extending arithmetic shift right */
28 
29 /* change endianness of a register */
30 #define BPF_END		0xd0	/* flags for endianness conversion: */
31 #define BPF_TO_LE	0x00	/* convert to little-endian */
32 #define BPF_TO_BE	0x08	/* convert to big-endian */
33 #define BPF_FROM_LE	BPF_TO_LE
34 #define BPF_FROM_BE	BPF_TO_BE
35 
36 /* jmp encodings */
37 #define BPF_JNE		0x50	/* jump != */
38 #define BPF_JLT		0xa0	/* LT is unsigned, '<' */
39 #define BPF_JLE		0xb0	/* LE is unsigned, '<=' */
40 #define BPF_JSGT	0x60	/* SGT is signed '>', GT in x86 */
41 #define BPF_JSGE	0x70	/* SGE is signed '>=', GE in x86 */
42 #define BPF_JSLT	0xc0	/* SLT is signed, '<' */
43 #define BPF_JSLE	0xd0	/* SLE is signed, '<=' */
44 #define BPF_CALL	0x80	/* function call */
45 #define BPF_EXIT	0x90	/* function return */
46 
47 /* atomic op type fields (stored in immediate) */
48 #define BPF_FETCH	0x01	/* not an opcode on its own, used to build others */
49 #define BPF_XCHG	(0xe0 | BPF_FETCH)	/* atomic exchange */
50 #define BPF_CMPXCHG	(0xf0 | BPF_FETCH)	/* atomic compare-and-write */
51 
52 /* Register numbers */
53 enum {
54 	BPF_REG_0 = 0,
55 	BPF_REG_1,
56 	BPF_REG_2,
57 	BPF_REG_3,
58 	BPF_REG_4,
59 	BPF_REG_5,
60 	BPF_REG_6,
61 	BPF_REG_7,
62 	BPF_REG_8,
63 	BPF_REG_9,
64 	BPF_REG_10,
65 	__MAX_BPF_REG,
66 };
67 
68 /* BPF has 10 general purpose 64-bit registers and stack frame. */
69 #define MAX_BPF_REG	__MAX_BPF_REG
70 
71 struct bpf_insn {
72 	__u8	code;		/* opcode */
73 	__u8	dst_reg:4;	/* dest register */
74 	__u8	src_reg:4;	/* source register */
75 	__s16	off;		/* signed offset */
76 	__s32	imm;		/* signed immediate constant */
77 };
78 
79 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
80 struct bpf_lpm_trie_key {
81 	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
82 	__u8	data[0];	/* Arbitrary size */
83 };
84 
85 struct bpf_cgroup_storage_key {
86 	__u64	cgroup_inode_id;	/* cgroup inode id */
87 	__u32	attach_type;		/* program attach type (enum bpf_attach_type) */
88 };
89 
90 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  *	Return
334  *		Returns zero on success. On error, -1 is returned and *errno*
335  *		is set appropriately.
336  *
337  *		**ENOSPC**
338  *			Either *data_size_out* or *ctx_size_out* is too small.
339  *		**ENOTSUPP**
340  *			This command is not supported by the program type of
341  *			the program referred to by *prog_fd*.
342  *
343  * BPF_PROG_GET_NEXT_ID
344  *	Description
345  *		Fetch the next eBPF program currently loaded into the kernel.
346  *
347  *		Looks for the eBPF program with an id greater than *start_id*
348  *		and updates *next_id* on success. If no other eBPF programs
349  *		remain with ids higher than *start_id*, returns -1 and sets
350  *		*errno* to **ENOENT**.
351  *
352  *	Return
353  *		Returns zero on success. On error, or when no id remains, -1
354  *		is returned and *errno* is set appropriately.
355  *
356  * BPF_MAP_GET_NEXT_ID
357  *	Description
358  *		Fetch the next eBPF map currently loaded into the kernel.
359  *
360  *		Looks for the eBPF map with an id greater than *start_id*
361  *		and updates *next_id* on success. If no other eBPF maps
362  *		remain with ids higher than *start_id*, returns -1 and sets
363  *		*errno* to **ENOENT**.
364  *
365  *	Return
366  *		Returns zero on success. On error, or when no id remains, -1
367  *		is returned and *errno* is set appropriately.
368  *
369  * BPF_PROG_GET_FD_BY_ID
370  *	Description
371  *		Open a file descriptor for the eBPF program corresponding to
372  *		*prog_id*.
373  *
374  *	Return
375  *		A new file descriptor (a nonnegative integer), or -1 if an
376  *		error occurred (in which case, *errno* is set appropriately).
377  *
378  * BPF_MAP_GET_FD_BY_ID
379  *	Description
380  *		Open a file descriptor for the eBPF map corresponding to
381  *		*map_id*.
382  *
383  *	Return
384  *		A new file descriptor (a nonnegative integer), or -1 if an
385  *		error occurred (in which case, *errno* is set appropriately).
386  *
387  * BPF_OBJ_GET_INFO_BY_FD
388  *	Description
389  *		Obtain information about the eBPF object corresponding to
390  *		*bpf_fd*.
391  *
392  *		Populates up to *info_len* bytes of *info*, which will be in
393  *		one of the following formats depending on the eBPF object type
394  *		of *bpf_fd*:
395  *
396  *		* **struct bpf_prog_info**
397  *		* **struct bpf_map_info**
398  *		* **struct bpf_btf_info**
399  *		* **struct bpf_link_info**
400  *
401  *	Return
402  *		Returns zero on success. On error, -1 is returned and *errno*
403  *		is set appropriately.
404  *
405  * BPF_PROG_QUERY
406  *	Description
407  *		Obtain information about eBPF programs associated with the
408  *		specified *attach_type* hook.
409  *
410  *		The *target_fd* must be a valid file descriptor for a kernel
411  *		object which depends on the attach type of *attach_bpf_fd*:
412  *
413  *		**BPF_PROG_TYPE_CGROUP_DEVICE**,
414  *		**BPF_PROG_TYPE_CGROUP_SKB**,
415  *		**BPF_PROG_TYPE_CGROUP_SOCK**,
416  *		**BPF_PROG_TYPE_CGROUP_SOCK_ADDR**,
417  *		**BPF_PROG_TYPE_CGROUP_SOCKOPT**,
418  *		**BPF_PROG_TYPE_CGROUP_SYSCTL**,
419  *		**BPF_PROG_TYPE_SOCK_OPS**
420  *
421  *			Control Group v2 hierarchy with the eBPF controller
422  *			enabled. Requires the kernel to be compiled with
423  *			**CONFIG_CGROUP_BPF**.
424  *
425  *		**BPF_PROG_TYPE_FLOW_DISSECTOR**
426  *
427  *			Network namespace (eg /proc/self/ns/net).
428  *
429  *		**BPF_PROG_TYPE_LIRC_MODE2**
430  *
431  *			LIRC device path (eg /dev/lircN). Requires the kernel
432  *			to be compiled with **CONFIG_BPF_LIRC_MODE2**.
433  *
434  *		**BPF_PROG_QUERY** always fetches the number of programs
435  *		attached and the *attach_flags* which were used to attach those
436  *		programs. Additionally, if *prog_ids* is nonzero and the number
437  *		of attached programs is less than *prog_cnt*, populates
438  *		*prog_ids* with the eBPF program ids of the programs attached
439  *		at *target_fd*.
440  *
441  *		The following flags may alter the result:
442  *
443  *		**BPF_F_QUERY_EFFECTIVE**
444  *			Only return information regarding programs which are
445  *			currently effective at the specified *target_fd*.
446  *
447  *	Return
448  *		Returns zero on success. On error, -1 is returned and *errno*
449  *		is set appropriately.
450  *
451  * BPF_RAW_TRACEPOINT_OPEN
452  *	Description
453  *		Attach an eBPF program to a tracepoint *name* to access kernel
454  *		internal arguments of the tracepoint in their raw form.
455  *
456  *		The *prog_fd* must be a valid file descriptor associated with
457  *		a loaded eBPF program of type **BPF_PROG_TYPE_RAW_TRACEPOINT**.
458  *
459  *		No ABI guarantees are made about the content of tracepoint
460  *		arguments exposed to the corresponding eBPF program.
461  *
462  *		Applying **close**\ (2) to the file descriptor returned by
463  *		**BPF_RAW_TRACEPOINT_OPEN** will delete the map (but see NOTES).
464  *
465  *	Return
466  *		A new file descriptor (a nonnegative integer), or -1 if an
467  *		error occurred (in which case, *errno* is set appropriately).
468  *
469  * BPF_BTF_LOAD
470  *	Description
471  *		Verify and load BPF Type Format (BTF) metadata into the kernel,
472  *		returning a new file descriptor associated with the metadata.
473  *		BTF is described in more detail at
474  *		https://www.kernel.org/doc/html/latest/bpf/btf.html.
475  *
476  *		The *btf* parameter must point to valid memory providing
477  *		*btf_size* bytes of BTF binary metadata.
478  *
479  *		The returned file descriptor can be passed to other **bpf**\ ()
480  *		subcommands such as **BPF_PROG_LOAD** or **BPF_MAP_CREATE** to
481  *		associate the BTF with those objects.
482  *
483  *		Similar to **BPF_PROG_LOAD**, **BPF_BTF_LOAD** has optional
484  *		parameters to specify a *btf_log_buf*, *btf_log_size* and
485  *		*btf_log_level* which allow the kernel to return freeform log
486  *		output regarding the BTF verification process.
487  *
488  *	Return
489  *		A new file descriptor (a nonnegative integer), or -1 if an
490  *		error occurred (in which case, *errno* is set appropriately).
491  *
492  * BPF_BTF_GET_FD_BY_ID
493  *	Description
494  *		Open a file descriptor for the BPF Type Format (BTF)
495  *		corresponding to *btf_id*.
496  *
497  *	Return
498  *		A new file descriptor (a nonnegative integer), or -1 if an
499  *		error occurred (in which case, *errno* is set appropriately).
500  *
501  * BPF_TASK_FD_QUERY
502  *	Description
503  *		Obtain information about eBPF programs associated with the
504  *		target process identified by *pid* and *fd*.
505  *
506  *		If the *pid* and *fd* are associated with a tracepoint, kprobe
507  *		or uprobe perf event, then the *prog_id* and *fd_type* will
508  *		be populated with the eBPF program id and file descriptor type
509  *		of type **bpf_task_fd_type**. If associated with a kprobe or
510  *		uprobe, the  *probe_offset* and *probe_addr* will also be
511  *		populated. Optionally, if *buf* is provided, then up to
512  *		*buf_len* bytes of *buf* will be populated with the name of
513  *		the tracepoint, kprobe or uprobe.
514  *
515  *		The resulting *prog_id* may be introspected in deeper detail
516  *		using **BPF_PROG_GET_FD_BY_ID** and **BPF_OBJ_GET_INFO_BY_FD**.
517  *
518  *	Return
519  *		Returns zero on success. On error, -1 is returned and *errno*
520  *		is set appropriately.
521  *
522  * BPF_MAP_LOOKUP_AND_DELETE_ELEM
523  *	Description
524  *		Look up an element with the given *key* in the map referred to
525  *		by the file descriptor *fd*, and if found, delete the element.
526  *
527  *		For **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map
528  *		types, the *flags* argument needs to be set to 0, but for other
529  *		map types, it may be specified as:
530  *
531  *		**BPF_F_LOCK**
532  *			Look up and delete the value of a spin-locked map
533  *			without returning the lock. This must be specified if
534  *			the elements contain a spinlock.
535  *
536  *		The **BPF_MAP_TYPE_QUEUE** and **BPF_MAP_TYPE_STACK** map types
537  *		implement this command as a "pop" operation, deleting the top
538  *		element rather than one corresponding to *key*.
539  *		The *key* and *key_len* parameters should be zeroed when
540  *		issuing this operation for these map types.
541  *
542  *		This command is only valid for the following map types:
543  *		* **BPF_MAP_TYPE_QUEUE**
544  *		* **BPF_MAP_TYPE_STACK**
545  *		* **BPF_MAP_TYPE_HASH**
546  *		* **BPF_MAP_TYPE_PERCPU_HASH**
547  *		* **BPF_MAP_TYPE_LRU_HASH**
548  *		* **BPF_MAP_TYPE_LRU_PERCPU_HASH**
549  *
550  *	Return
551  *		Returns zero on success. On error, -1 is returned and *errno*
552  *		is set appropriately.
553  *
554  * BPF_MAP_FREEZE
555  *	Description
556  *		Freeze the permissions of the specified map.
557  *
558  *		Write permissions may be frozen by passing zero *flags*.
559  *		Upon success, no future syscall invocations may alter the
560  *		map state of *map_fd*. Write operations from eBPF programs
561  *		are still possible for a frozen map.
562  *
563  *		Not supported for maps of type **BPF_MAP_TYPE_STRUCT_OPS**.
564  *
565  *	Return
566  *		Returns zero on success. On error, -1 is returned and *errno*
567  *		is set appropriately.
568  *
569  * BPF_BTF_GET_NEXT_ID
570  *	Description
571  *		Fetch the next BPF Type Format (BTF) object currently loaded
572  *		into the kernel.
573  *
574  *		Looks for the BTF object with an id greater than *start_id*
575  *		and updates *next_id* on success. If no other BTF objects
576  *		remain with ids higher than *start_id*, returns -1 and sets
577  *		*errno* to **ENOENT**.
578  *
579  *	Return
580  *		Returns zero on success. On error, or when no id remains, -1
581  *		is returned and *errno* is set appropriately.
582  *
583  * BPF_MAP_LOOKUP_BATCH
584  *	Description
585  *		Iterate and fetch multiple elements in a map.
586  *
587  *		Two opaque values are used to manage batch operations,
588  *		*in_batch* and *out_batch*. Initially, *in_batch* must be set
589  *		to NULL to begin the batched operation. After each subsequent
590  *		**BPF_MAP_LOOKUP_BATCH**, the caller should pass the resultant
591  *		*out_batch* as the *in_batch* for the next operation to
592  *		continue iteration from the current point.
593  *
594  *		The *keys* and *values* are output parameters which must point
595  *		to memory large enough to hold *count* items based on the key
596  *		and value size of the map *map_fd*. The *keys* buffer must be
597  *		of *key_size* * *count*. The *values* buffer must be of
598  *		*value_size* * *count*.
599  *
600  *		The *elem_flags* argument may be specified as one of the
601  *		following:
602  *
603  *		**BPF_F_LOCK**
604  *			Look up the value of a spin-locked map without
605  *			returning the lock. This must be specified if the
606  *			elements contain a spinlock.
607  *
608  *		On success, *count* elements from the map are copied into the
609  *		user buffer, with the keys copied into *keys* and the values
610  *		copied into the corresponding indices in *values*.
611  *
612  *		If an error is returned and *errno* is not **EFAULT**, *count*
613  *		is set to the number of successfully processed elements.
614  *
615  *	Return
616  *		Returns zero on success. On error, -1 is returned and *errno*
617  *		is set appropriately.
618  *
619  *		May set *errno* to **ENOSPC** to indicate that *keys* or
620  *		*values* is too small to dump an entire bucket during
621  *		iteration of a hash-based map type.
622  *
623  * BPF_MAP_LOOKUP_AND_DELETE_BATCH
624  *	Description
625  *		Iterate and delete all elements in a map.
626  *
627  *		This operation has the same behavior as
628  *		**BPF_MAP_LOOKUP_BATCH** with two exceptions:
629  *
630  *		* Every element that is successfully returned is also deleted
631  *		  from the map. This is at least *count* elements. Note that
632  *		  *count* is both an input and an output parameter.
633  *		* Upon returning with *errno* set to **EFAULT**, up to
634  *		  *count* elements may be deleted without returning the keys
635  *		  and values of the deleted elements.
636  *
637  *	Return
638  *		Returns zero on success. On error, -1 is returned and *errno*
639  *		is set appropriately.
640  *
641  * BPF_MAP_UPDATE_BATCH
642  *	Description
643  *		Update multiple elements in a map by *key*.
644  *
645  *		The *keys* and *values* are input parameters which must point
646  *		to memory large enough to hold *count* items based on the key
647  *		and value size of the map *map_fd*. The *keys* buffer must be
648  *		of *key_size* * *count*. The *values* buffer must be of
649  *		*value_size* * *count*.
650  *
651  *		Each element specified in *keys* is sequentially updated to the
652  *		value in the corresponding index in *values*. The *in_batch*
653  *		and *out_batch* parameters are ignored and should be zeroed.
654  *
655  *		The *elem_flags* argument should be specified as one of the
656  *		following:
657  *
658  *		**BPF_ANY**
659  *			Create new elements or update a existing elements.
660  *		**BPF_NOEXIST**
661  *			Create new elements only if they do not exist.
662  *		**BPF_EXIST**
663  *			Update existing elements.
664  *		**BPF_F_LOCK**
665  *			Update spin_lock-ed map elements. This must be
666  *			specified if the map value contains a spinlock.
667  *
668  *		On success, *count* elements from the map are updated.
669  *
670  *		If an error is returned and *errno* is not **EFAULT**, *count*
671  *		is set to the number of successfully processed elements.
672  *
673  *	Return
674  *		Returns zero on success. On error, -1 is returned and *errno*
675  *		is set appropriately.
676  *
677  *		May set *errno* to **EINVAL**, **EPERM**, **ENOMEM**, or
678  *		**E2BIG**. **E2BIG** indicates that the number of elements in
679  *		the map reached the *max_entries* limit specified at map
680  *		creation time.
681  *
682  *		May set *errno* to one of the following error codes under
683  *		specific circumstances:
684  *
685  *		**EEXIST**
686  *			If *flags* specifies **BPF_NOEXIST** and the element
687  *			with *key* already exists in the map.
688  *		**ENOENT**
689  *			If *flags* specifies **BPF_EXIST** and the element with
690  *			*key* does not exist in the map.
691  *
692  * BPF_MAP_DELETE_BATCH
693  *	Description
694  *		Delete multiple elements in a map by *key*.
695  *
696  *		The *keys* parameter is an input parameter which must point
697  *		to memory large enough to hold *count* items based on the key
698  *		size of the map *map_fd*, that is, *key_size* * *count*.
699  *
700  *		Each element specified in *keys* is sequentially deleted. The
701  *		*in_batch*, *out_batch*, and *values* parameters are ignored
702  *		and should be zeroed.
703  *
704  *		The *elem_flags* argument may be specified as one of the
705  *		following:
706  *
707  *		**BPF_F_LOCK**
708  *			Look up the value of a spin-locked map without
709  *			returning the lock. This must be specified if the
710  *			elements contain a spinlock.
711  *
712  *		On success, *count* elements from the map are updated.
713  *
714  *		If an error is returned and *errno* is not **EFAULT**, *count*
715  *		is set to the number of successfully processed elements. If
716  *		*errno* is **EFAULT**, up to *count* elements may be been
717  *		deleted.
718  *
719  *	Return
720  *		Returns zero on success. On error, -1 is returned and *errno*
721  *		is set appropriately.
722  *
723  * BPF_LINK_CREATE
724  *	Description
725  *		Attach an eBPF program to a *target_fd* at the specified
726  *		*attach_type* hook and return a file descriptor handle for
727  *		managing the link.
728  *
729  *	Return
730  *		A new file descriptor (a nonnegative integer), or -1 if an
731  *		error occurred (in which case, *errno* is set appropriately).
732  *
733  * BPF_LINK_UPDATE
734  *	Description
735  *		Update the eBPF program in the specified *link_fd* to
736  *		*new_prog_fd*.
737  *
738  *	Return
739  *		Returns zero on success. On error, -1 is returned and *errno*
740  *		is set appropriately.
741  *
742  * BPF_LINK_GET_FD_BY_ID
743  *	Description
744  *		Open a file descriptor for the eBPF Link corresponding to
745  *		*link_id*.
746  *
747  *	Return
748  *		A new file descriptor (a nonnegative integer), or -1 if an
749  *		error occurred (in which case, *errno* is set appropriately).
750  *
751  * BPF_LINK_GET_NEXT_ID
752  *	Description
753  *		Fetch the next eBPF link currently loaded into the kernel.
754  *
755  *		Looks for the eBPF link with an id greater than *start_id*
756  *		and updates *next_id* on success. If no other eBPF links
757  *		remain with ids higher than *start_id*, returns -1 and sets
758  *		*errno* to **ENOENT**.
759  *
760  *	Return
761  *		Returns zero on success. On error, or when no id remains, -1
762  *		is returned and *errno* is set appropriately.
763  *
764  * BPF_ENABLE_STATS
765  *	Description
766  *		Enable eBPF runtime statistics gathering.
767  *
768  *		Runtime statistics gathering for the eBPF runtime is disabled
769  *		by default to minimize the corresponding performance overhead.
770  *		This command enables statistics globally.
771  *
772  *		Multiple programs may independently enable statistics.
773  *		After gathering the desired statistics, eBPF runtime statistics
774  *		may be disabled again by calling **close**\ (2) for the file
775  *		descriptor returned by this function. Statistics will only be
776  *		disabled system-wide when all outstanding file descriptors
777  *		returned by prior calls for this subcommand are closed.
778  *
779  *	Return
780  *		A new file descriptor (a nonnegative integer), or -1 if an
781  *		error occurred (in which case, *errno* is set appropriately).
782  *
783  * BPF_ITER_CREATE
784  *	Description
785  *		Create an iterator on top of the specified *link_fd* (as
786  *		previously created using **BPF_LINK_CREATE**) and return a
787  *		file descriptor that can be used to trigger the iteration.
788  *
789  *		If the resulting file descriptor is pinned to the filesystem
790  *		using  **BPF_OBJ_PIN**, then subsequent **read**\ (2) syscalls
791  *		for that path will trigger the iterator to read kernel state
792  *		using the eBPF program attached to *link_fd*.
793  *
794  *	Return
795  *		A new file descriptor (a nonnegative integer), or -1 if an
796  *		error occurred (in which case, *errno* is set appropriately).
797  *
798  * BPF_LINK_DETACH
799  *	Description
800  *		Forcefully detach the specified *link_fd* from its
801  *		corresponding attachment point.
802  *
803  *	Return
804  *		Returns zero on success. On error, -1 is returned and *errno*
805  *		is set appropriately.
806  *
807  * BPF_PROG_BIND_MAP
808  *	Description
809  *		Bind a map to the lifetime of an eBPF program.
810  *
811  *		The map identified by *map_fd* is bound to the program
812  *		identified by *prog_fd* and only released when *prog_fd* is
813  *		released. This may be used in cases where metadata should be
814  *		associated with a program which otherwise does not contain any
815  *		references to the map (for example, embedded in the eBPF
816  *		program instructions).
817  *
818  *	Return
819  *		Returns zero on success. On error, -1 is returned and *errno*
820  *		is set appropriately.
821  *
822  * NOTES
823  *	eBPF objects (maps and programs) can be shared between processes.
824  *
825  *	* After **fork**\ (2), the child inherits file descriptors
826  *	  referring to the same eBPF objects.
827  *	* File descriptors referring to eBPF objects can be transferred over
828  *	  **unix**\ (7) domain sockets.
829  *	* File descriptors referring to eBPF objects can be duplicated in the
830  *	  usual way, using **dup**\ (2) and similar calls.
831  *	* File descriptors referring to eBPF objects can be pinned to the
832  *	  filesystem using the **BPF_OBJ_PIN** command of **bpf**\ (2).
833  *
834  *	An eBPF object is deallocated only after all file descriptors referring
835  *	to the object have been closed and no references remain pinned to the
836  *	filesystem or attached (for example, bound to a program or device).
837  */
838 enum bpf_cmd {
839 	BPF_MAP_CREATE,
840 	BPF_MAP_LOOKUP_ELEM,
841 	BPF_MAP_UPDATE_ELEM,
842 	BPF_MAP_DELETE_ELEM,
843 	BPF_MAP_GET_NEXT_KEY,
844 	BPF_PROG_LOAD,
845 	BPF_OBJ_PIN,
846 	BPF_OBJ_GET,
847 	BPF_PROG_ATTACH,
848 	BPF_PROG_DETACH,
849 	BPF_PROG_TEST_RUN,
850 	BPF_PROG_RUN = BPF_PROG_TEST_RUN,
851 	BPF_PROG_GET_NEXT_ID,
852 	BPF_MAP_GET_NEXT_ID,
853 	BPF_PROG_GET_FD_BY_ID,
854 	BPF_MAP_GET_FD_BY_ID,
855 	BPF_OBJ_GET_INFO_BY_FD,
856 	BPF_PROG_QUERY,
857 	BPF_RAW_TRACEPOINT_OPEN,
858 	BPF_BTF_LOAD,
859 	BPF_BTF_GET_FD_BY_ID,
860 	BPF_TASK_FD_QUERY,
861 	BPF_MAP_LOOKUP_AND_DELETE_ELEM,
862 	BPF_MAP_FREEZE,
863 	BPF_BTF_GET_NEXT_ID,
864 	BPF_MAP_LOOKUP_BATCH,
865 	BPF_MAP_LOOKUP_AND_DELETE_BATCH,
866 	BPF_MAP_UPDATE_BATCH,
867 	BPF_MAP_DELETE_BATCH,
868 	BPF_LINK_CREATE,
869 	BPF_LINK_UPDATE,
870 	BPF_LINK_GET_FD_BY_ID,
871 	BPF_LINK_GET_NEXT_ID,
872 	BPF_ENABLE_STATS,
873 	BPF_ITER_CREATE,
874 	BPF_LINK_DETACH,
875 	BPF_PROG_BIND_MAP,
876 };
877 
878 enum bpf_map_type {
879 	BPF_MAP_TYPE_UNSPEC,
880 	BPF_MAP_TYPE_HASH,
881 	BPF_MAP_TYPE_ARRAY,
882 	BPF_MAP_TYPE_PROG_ARRAY,
883 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
884 	BPF_MAP_TYPE_PERCPU_HASH,
885 	BPF_MAP_TYPE_PERCPU_ARRAY,
886 	BPF_MAP_TYPE_STACK_TRACE,
887 	BPF_MAP_TYPE_CGROUP_ARRAY,
888 	BPF_MAP_TYPE_LRU_HASH,
889 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
890 	BPF_MAP_TYPE_LPM_TRIE,
891 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
892 	BPF_MAP_TYPE_HASH_OF_MAPS,
893 	BPF_MAP_TYPE_DEVMAP,
894 	BPF_MAP_TYPE_SOCKMAP,
895 	BPF_MAP_TYPE_CPUMAP,
896 	BPF_MAP_TYPE_XSKMAP,
897 	BPF_MAP_TYPE_SOCKHASH,
898 	BPF_MAP_TYPE_CGROUP_STORAGE,
899 	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
900 	BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
901 	BPF_MAP_TYPE_QUEUE,
902 	BPF_MAP_TYPE_STACK,
903 	BPF_MAP_TYPE_SK_STORAGE,
904 	BPF_MAP_TYPE_DEVMAP_HASH,
905 	BPF_MAP_TYPE_STRUCT_OPS,
906 	BPF_MAP_TYPE_RINGBUF,
907 	BPF_MAP_TYPE_INODE_STORAGE,
908 	BPF_MAP_TYPE_TASK_STORAGE,
909 	BPF_MAP_TYPE_BLOOM_FILTER,
910 };
911 
912 /* Note that tracing related programs such as
913  * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT}
914  * are not subject to a stable API since kernel internal data
915  * structures can change from release to release and may
916  * therefore break existing tracing BPF programs. Tracing BPF
917  * programs correspond to /a/ specific kernel which is to be
918  * analyzed, and not /a/ specific kernel /and/ all future ones.
919  */
920 enum bpf_prog_type {
921 	BPF_PROG_TYPE_UNSPEC,
922 	BPF_PROG_TYPE_SOCKET_FILTER,
923 	BPF_PROG_TYPE_KPROBE,
924 	BPF_PROG_TYPE_SCHED_CLS,
925 	BPF_PROG_TYPE_SCHED_ACT,
926 	BPF_PROG_TYPE_TRACEPOINT,
927 	BPF_PROG_TYPE_XDP,
928 	BPF_PROG_TYPE_PERF_EVENT,
929 	BPF_PROG_TYPE_CGROUP_SKB,
930 	BPF_PROG_TYPE_CGROUP_SOCK,
931 	BPF_PROG_TYPE_LWT_IN,
932 	BPF_PROG_TYPE_LWT_OUT,
933 	BPF_PROG_TYPE_LWT_XMIT,
934 	BPF_PROG_TYPE_SOCK_OPS,
935 	BPF_PROG_TYPE_SK_SKB,
936 	BPF_PROG_TYPE_CGROUP_DEVICE,
937 	BPF_PROG_TYPE_SK_MSG,
938 	BPF_PROG_TYPE_RAW_TRACEPOINT,
939 	BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
940 	BPF_PROG_TYPE_LWT_SEG6LOCAL,
941 	BPF_PROG_TYPE_LIRC_MODE2,
942 	BPF_PROG_TYPE_SK_REUSEPORT,
943 	BPF_PROG_TYPE_FLOW_DISSECTOR,
944 	BPF_PROG_TYPE_CGROUP_SYSCTL,
945 	BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE,
946 	BPF_PROG_TYPE_CGROUP_SOCKOPT,
947 	BPF_PROG_TYPE_TRACING,
948 	BPF_PROG_TYPE_STRUCT_OPS,
949 	BPF_PROG_TYPE_EXT,
950 	BPF_PROG_TYPE_LSM,
951 	BPF_PROG_TYPE_SK_LOOKUP,
952 	BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */
953 };
954 
955 enum bpf_attach_type {
956 	BPF_CGROUP_INET_INGRESS,
957 	BPF_CGROUP_INET_EGRESS,
958 	BPF_CGROUP_INET_SOCK_CREATE,
959 	BPF_CGROUP_SOCK_OPS,
960 	BPF_SK_SKB_STREAM_PARSER,
961 	BPF_SK_SKB_STREAM_VERDICT,
962 	BPF_CGROUP_DEVICE,
963 	BPF_SK_MSG_VERDICT,
964 	BPF_CGROUP_INET4_BIND,
965 	BPF_CGROUP_INET6_BIND,
966 	BPF_CGROUP_INET4_CONNECT,
967 	BPF_CGROUP_INET6_CONNECT,
968 	BPF_CGROUP_INET4_POST_BIND,
969 	BPF_CGROUP_INET6_POST_BIND,
970 	BPF_CGROUP_UDP4_SENDMSG,
971 	BPF_CGROUP_UDP6_SENDMSG,
972 	BPF_LIRC_MODE2,
973 	BPF_FLOW_DISSECTOR,
974 	BPF_CGROUP_SYSCTL,
975 	BPF_CGROUP_UDP4_RECVMSG,
976 	BPF_CGROUP_UDP6_RECVMSG,
977 	BPF_CGROUP_GETSOCKOPT,
978 	BPF_CGROUP_SETSOCKOPT,
979 	BPF_TRACE_RAW_TP,
980 	BPF_TRACE_FENTRY,
981 	BPF_TRACE_FEXIT,
982 	BPF_MODIFY_RETURN,
983 	BPF_LSM_MAC,
984 	BPF_TRACE_ITER,
985 	BPF_CGROUP_INET4_GETPEERNAME,
986 	BPF_CGROUP_INET6_GETPEERNAME,
987 	BPF_CGROUP_INET4_GETSOCKNAME,
988 	BPF_CGROUP_INET6_GETSOCKNAME,
989 	BPF_XDP_DEVMAP,
990 	BPF_CGROUP_INET_SOCK_RELEASE,
991 	BPF_XDP_CPUMAP,
992 	BPF_SK_LOOKUP,
993 	BPF_XDP,
994 	BPF_SK_SKB_VERDICT,
995 	BPF_SK_REUSEPORT_SELECT,
996 	BPF_SK_REUSEPORT_SELECT_OR_MIGRATE,
997 	BPF_PERF_EVENT,
998 	__MAX_BPF_ATTACH_TYPE
999 };
1000 
1001 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
1002 
1003 enum bpf_link_type {
1004 	BPF_LINK_TYPE_UNSPEC = 0,
1005 	BPF_LINK_TYPE_RAW_TRACEPOINT = 1,
1006 	BPF_LINK_TYPE_TRACING = 2,
1007 	BPF_LINK_TYPE_CGROUP = 3,
1008 	BPF_LINK_TYPE_ITER = 4,
1009 	BPF_LINK_TYPE_NETNS = 5,
1010 	BPF_LINK_TYPE_XDP = 6,
1011 	BPF_LINK_TYPE_PERF_EVENT = 7,
1012 
1013 	MAX_BPF_LINK_TYPE,
1014 };
1015 
1016 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
1017  *
1018  * NONE(default): No further bpf programs allowed in the subtree.
1019  *
1020  * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
1021  * the program in this cgroup yields to sub-cgroup program.
1022  *
1023  * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
1024  * that cgroup program gets run in addition to the program in this cgroup.
1025  *
1026  * Only one program is allowed to be attached to a cgroup with
1027  * NONE or BPF_F_ALLOW_OVERRIDE flag.
1028  * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
1029  * release old program and attach the new one. Attach flags has to match.
1030  *
1031  * Multiple programs are allowed to be attached to a cgroup with
1032  * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
1033  * (those that were attached first, run first)
1034  * The programs of sub-cgroup are executed first, then programs of
1035  * this cgroup and then programs of parent cgroup.
1036  * When children program makes decision (like picking TCP CA or sock bind)
1037  * parent program has a chance to override it.
1038  *
1039  * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of
1040  * programs for a cgroup. Though it's possible to replace an old program at
1041  * any position by also specifying BPF_F_REPLACE flag and position itself in
1042  * replace_bpf_fd attribute. Old program at this position will be released.
1043  *
1044  * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
1045  * A cgroup with NONE doesn't allow any programs in sub-cgroups.
1046  * Ex1:
1047  * cgrp1 (MULTI progs A, B) ->
1048  *    cgrp2 (OVERRIDE prog C) ->
1049  *      cgrp3 (MULTI prog D) ->
1050  *        cgrp4 (OVERRIDE prog E) ->
1051  *          cgrp5 (NONE prog F)
1052  * the event in cgrp5 triggers execution of F,D,A,B in that order.
1053  * if prog F is detached, the execution is E,D,A,B
1054  * if prog F and D are detached, the execution is E,A,B
1055  * if prog F, E and D are detached, the execution is C,A,B
1056  *
1057  * All eligible programs are executed regardless of return code from
1058  * earlier programs.
1059  */
1060 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
1061 #define BPF_F_ALLOW_MULTI	(1U << 1)
1062 #define BPF_F_REPLACE		(1U << 2)
1063 
1064 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
1065  * verifier will perform strict alignment checking as if the kernel
1066  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
1067  * and NET_IP_ALIGN defined to 2.
1068  */
1069 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
1070 
1071 /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the
1072  * verifier will allow any alignment whatsoever.  On platforms
1073  * with strict alignment requirements for loads ands stores (such
1074  * as sparc and mips) the verifier validates that all loads and
1075  * stores provably follow this requirement.  This flag turns that
1076  * checking and enforcement off.
1077  *
1078  * It is mostly used for testing when we want to validate the
1079  * context and memory access aspects of the verifier, but because
1080  * of an unaligned access the alignment check would trigger before
1081  * the one we are interested in.
1082  */
1083 #define BPF_F_ANY_ALIGNMENT	(1U << 1)
1084 
1085 /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose.
1086  * Verifier does sub-register def/use analysis and identifies instructions whose
1087  * def only matters for low 32-bit, high 32-bit is never referenced later
1088  * through implicit zero extension. Therefore verifier notifies JIT back-ends
1089  * that it is safe to ignore clearing high 32-bit for these instructions. This
1090  * saves some back-ends a lot of code-gen. However such optimization is not
1091  * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends
1092  * hence hasn't used verifier's analysis result. But, we really want to have a
1093  * way to be able to verify the correctness of the described optimization on
1094  * x86_64 on which testsuites are frequently exercised.
1095  *
1096  * So, this flag is introduced. Once it is set, verifier will randomize high
1097  * 32-bit for those instructions who has been identified as safe to ignore them.
1098  * Then, if verifier is not doing correct analysis, such randomization will
1099  * regress tests to expose bugs.
1100  */
1101 #define BPF_F_TEST_RND_HI32	(1U << 2)
1102 
1103 /* The verifier internal test flag. Behavior is undefined */
1104 #define BPF_F_TEST_STATE_FREQ	(1U << 3)
1105 
1106 /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will
1107  * restrict map and helper usage for such programs. Sleepable BPF programs can
1108  * only be attached to hooks where kernel execution context allows sleeping.
1109  * Such programs are allowed to use helpers that may sleep like
1110  * bpf_copy_from_user().
1111  */
1112 #define BPF_F_SLEEPABLE		(1U << 4)
1113 
1114 /* When BPF ldimm64's insn[0].src_reg != 0 then this can have
1115  * the following extensions:
1116  *
1117  * insn[0].src_reg:  BPF_PSEUDO_MAP_[FD|IDX]
1118  * insn[0].imm:      map fd or fd_idx
1119  * insn[1].imm:      0
1120  * insn[0].off:      0
1121  * insn[1].off:      0
1122  * ldimm64 rewrite:  address of map
1123  * verifier type:    CONST_PTR_TO_MAP
1124  */
1125 #define BPF_PSEUDO_MAP_FD	1
1126 #define BPF_PSEUDO_MAP_IDX	5
1127 
1128 /* insn[0].src_reg:  BPF_PSEUDO_MAP_[IDX_]VALUE
1129  * insn[0].imm:      map fd or fd_idx
1130  * insn[1].imm:      offset into value
1131  * insn[0].off:      0
1132  * insn[1].off:      0
1133  * ldimm64 rewrite:  address of map[0]+offset
1134  * verifier type:    PTR_TO_MAP_VALUE
1135  */
1136 #define BPF_PSEUDO_MAP_VALUE		2
1137 #define BPF_PSEUDO_MAP_IDX_VALUE	6
1138 
1139 /* insn[0].src_reg:  BPF_PSEUDO_BTF_ID
1140  * insn[0].imm:      kernel btd id of VAR
1141  * insn[1].imm:      0
1142  * insn[0].off:      0
1143  * insn[1].off:      0
1144  * ldimm64 rewrite:  address of the kernel variable
1145  * verifier type:    PTR_TO_BTF_ID or PTR_TO_MEM, depending on whether the var
1146  *                   is struct/union.
1147  */
1148 #define BPF_PSEUDO_BTF_ID	3
1149 /* insn[0].src_reg:  BPF_PSEUDO_FUNC
1150  * insn[0].imm:      insn offset to the func
1151  * insn[1].imm:      0
1152  * insn[0].off:      0
1153  * insn[1].off:      0
1154  * ldimm64 rewrite:  address of the function
1155  * verifier type:    PTR_TO_FUNC.
1156  */
1157 #define BPF_PSEUDO_FUNC		4
1158 
1159 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
1160  * offset to another bpf function
1161  */
1162 #define BPF_PSEUDO_CALL		1
1163 /* when bpf_call->src_reg == BPF_PSEUDO_KFUNC_CALL,
1164  * bpf_call->imm == btf_id of a BTF_KIND_FUNC in the running kernel
1165  */
1166 #define BPF_PSEUDO_KFUNC_CALL	2
1167 
1168 /* flags for BPF_MAP_UPDATE_ELEM command */
1169 enum {
1170 	BPF_ANY		= 0, /* create new element or update existing */
1171 	BPF_NOEXIST	= 1, /* create new element if it didn't exist */
1172 	BPF_EXIST	= 2, /* update existing element */
1173 	BPF_F_LOCK	= 4, /* spin_lock-ed map_lookup/map_update */
1174 };
1175 
1176 /* flags for BPF_MAP_CREATE command */
1177 enum {
1178 	BPF_F_NO_PREALLOC	= (1U << 0),
1179 /* Instead of having one common LRU list in the
1180  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
1181  * which can scale and perform better.
1182  * Note, the LRU nodes (including free nodes) cannot be moved
1183  * across different LRU lists.
1184  */
1185 	BPF_F_NO_COMMON_LRU	= (1U << 1),
1186 /* Specify numa node during map creation */
1187 	BPF_F_NUMA_NODE		= (1U << 2),
1188 
1189 /* Flags for accessing BPF object from syscall side. */
1190 	BPF_F_RDONLY		= (1U << 3),
1191 	BPF_F_WRONLY		= (1U << 4),
1192 
1193 /* Flag for stack_map, store build_id+offset instead of pointer */
1194 	BPF_F_STACK_BUILD_ID	= (1U << 5),
1195 
1196 /* Zero-initialize hash function seed. This should only be used for testing. */
1197 	BPF_F_ZERO_SEED		= (1U << 6),
1198 
1199 /* Flags for accessing BPF object from program side. */
1200 	BPF_F_RDONLY_PROG	= (1U << 7),
1201 	BPF_F_WRONLY_PROG	= (1U << 8),
1202 
1203 /* Clone map from listener for newly accepted socket */
1204 	BPF_F_CLONE		= (1U << 9),
1205 
1206 /* Enable memory-mapping BPF map */
1207 	BPF_F_MMAPABLE		= (1U << 10),
1208 
1209 /* Share perf_event among processes */
1210 	BPF_F_PRESERVE_ELEMS	= (1U << 11),
1211 
1212 /* Create a map that is suitable to be an inner map with dynamic max entries */
1213 	BPF_F_INNER_MAP		= (1U << 12),
1214 };
1215 
1216 /* Flags for BPF_PROG_QUERY. */
1217 
1218 /* Query effective (directly attached + inherited from ancestor cgroups)
1219  * programs that will be executed for events within a cgroup.
1220  * attach_flags with this flag are returned only for directly attached programs.
1221  */
1222 #define BPF_F_QUERY_EFFECTIVE	(1U << 0)
1223 
1224 /* Flags for BPF_PROG_TEST_RUN */
1225 
1226 /* If set, run the test on the cpu specified by bpf_attr.test.cpu */
1227 #define BPF_F_TEST_RUN_ON_CPU	(1U << 0)
1228 
1229 /* type for BPF_ENABLE_STATS */
1230 enum bpf_stats_type {
1231 	/* enabled run_time_ns and run_cnt */
1232 	BPF_STATS_RUN_TIME = 0,
1233 };
1234 
1235 enum bpf_stack_build_id_status {
1236 	/* user space need an empty entry to identify end of a trace */
1237 	BPF_STACK_BUILD_ID_EMPTY = 0,
1238 	/* with valid build_id and offset */
1239 	BPF_STACK_BUILD_ID_VALID = 1,
1240 	/* couldn't get build_id, fallback to ip */
1241 	BPF_STACK_BUILD_ID_IP = 2,
1242 };
1243 
1244 #define BPF_BUILD_ID_SIZE 20
1245 struct bpf_stack_build_id {
1246 	__s32		status;
1247 	unsigned char	build_id[BPF_BUILD_ID_SIZE];
1248 	union {
1249 		__u64	offset;
1250 		__u64	ip;
1251 	};
1252 };
1253 
1254 #define BPF_OBJ_NAME_LEN 16U
1255 
1256 union bpf_attr {
1257 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
1258 		__u32	map_type;	/* one of enum bpf_map_type */
1259 		__u32	key_size;	/* size of key in bytes */
1260 		__u32	value_size;	/* size of value in bytes */
1261 		__u32	max_entries;	/* max number of entries in a map */
1262 		__u32	map_flags;	/* BPF_MAP_CREATE related
1263 					 * flags defined above.
1264 					 */
1265 		__u32	inner_map_fd;	/* fd pointing to the inner map */
1266 		__u32	numa_node;	/* numa node (effective only if
1267 					 * BPF_F_NUMA_NODE is set).
1268 					 */
1269 		char	map_name[BPF_OBJ_NAME_LEN];
1270 		__u32	map_ifindex;	/* ifindex of netdev to create on */
1271 		__u32	btf_fd;		/* fd pointing to a BTF type data */
1272 		__u32	btf_key_type_id;	/* BTF type_id of the key */
1273 		__u32	btf_value_type_id;	/* BTF type_id of the value */
1274 		__u32	btf_vmlinux_value_type_id;/* BTF type_id of a kernel-
1275 						   * struct stored as the
1276 						   * map value
1277 						   */
1278 		/* Any per-map-type extra fields
1279 		 *
1280 		 * BPF_MAP_TYPE_BLOOM_FILTER - the lowest 4 bits indicate the
1281 		 * number of hash functions (if 0, the bloom filter will default
1282 		 * to using 5 hash functions).
1283 		 */
1284 		__u64	map_extra;
1285 	};
1286 
1287 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
1288 		__u32		map_fd;
1289 		__aligned_u64	key;
1290 		union {
1291 			__aligned_u64 value;
1292 			__aligned_u64 next_key;
1293 		};
1294 		__u64		flags;
1295 	};
1296 
1297 	struct { /* struct used by BPF_MAP_*_BATCH commands */
1298 		__aligned_u64	in_batch;	/* start batch,
1299 						 * NULL to start from beginning
1300 						 */
1301 		__aligned_u64	out_batch;	/* output: next start batch */
1302 		__aligned_u64	keys;
1303 		__aligned_u64	values;
1304 		__u32		count;		/* input/output:
1305 						 * input: # of key/value
1306 						 * elements
1307 						 * output: # of filled elements
1308 						 */
1309 		__u32		map_fd;
1310 		__u64		elem_flags;
1311 		__u64		flags;
1312 	} batch;
1313 
1314 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
1315 		__u32		prog_type;	/* one of enum bpf_prog_type */
1316 		__u32		insn_cnt;
1317 		__aligned_u64	insns;
1318 		__aligned_u64	license;
1319 		__u32		log_level;	/* verbosity level of verifier */
1320 		__u32		log_size;	/* size of user buffer */
1321 		__aligned_u64	log_buf;	/* user supplied buffer */
1322 		__u32		kern_version;	/* not used */
1323 		__u32		prog_flags;
1324 		char		prog_name[BPF_OBJ_NAME_LEN];
1325 		__u32		prog_ifindex;	/* ifindex of netdev to prep for */
1326 		/* For some prog types expected attach type must be known at
1327 		 * load time to verify attach type specific parts of prog
1328 		 * (context accesses, allowed helpers, etc).
1329 		 */
1330 		__u32		expected_attach_type;
1331 		__u32		prog_btf_fd;	/* fd pointing to BTF type data */
1332 		__u32		func_info_rec_size;	/* userspace bpf_func_info size */
1333 		__aligned_u64	func_info;	/* func info */
1334 		__u32		func_info_cnt;	/* number of bpf_func_info records */
1335 		__u32		line_info_rec_size;	/* userspace bpf_line_info size */
1336 		__aligned_u64	line_info;	/* line info */
1337 		__u32		line_info_cnt;	/* number of bpf_line_info records */
1338 		__u32		attach_btf_id;	/* in-kernel BTF type id to attach to */
1339 		union {
1340 			/* valid prog_fd to attach to bpf prog */
1341 			__u32		attach_prog_fd;
1342 			/* or valid module BTF object fd or 0 to attach to vmlinux */
1343 			__u32		attach_btf_obj_fd;
1344 		};
1345 		__u32		:32;		/* pad */
1346 		__aligned_u64	fd_array;	/* array of FDs */
1347 	};
1348 
1349 	struct { /* anonymous struct used by BPF_OBJ_* commands */
1350 		__aligned_u64	pathname;
1351 		__u32		bpf_fd;
1352 		__u32		file_flags;
1353 	};
1354 
1355 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
1356 		__u32		target_fd;	/* container object to attach to */
1357 		__u32		attach_bpf_fd;	/* eBPF program to attach */
1358 		__u32		attach_type;
1359 		__u32		attach_flags;
1360 		__u32		replace_bpf_fd;	/* previously attached eBPF
1361 						 * program to replace if
1362 						 * BPF_F_REPLACE is used
1363 						 */
1364 	};
1365 
1366 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
1367 		__u32		prog_fd;
1368 		__u32		retval;
1369 		__u32		data_size_in;	/* input: len of data_in */
1370 		__u32		data_size_out;	/* input/output: len of data_out
1371 						 *   returns ENOSPC if data_out
1372 						 *   is too small.
1373 						 */
1374 		__aligned_u64	data_in;
1375 		__aligned_u64	data_out;
1376 		__u32		repeat;
1377 		__u32		duration;
1378 		__u32		ctx_size_in;	/* input: len of ctx_in */
1379 		__u32		ctx_size_out;	/* input/output: len of ctx_out
1380 						 *   returns ENOSPC if ctx_out
1381 						 *   is too small.
1382 						 */
1383 		__aligned_u64	ctx_in;
1384 		__aligned_u64	ctx_out;
1385 		__u32		flags;
1386 		__u32		cpu;
1387 	} test;
1388 
1389 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
1390 		union {
1391 			__u32		start_id;
1392 			__u32		prog_id;
1393 			__u32		map_id;
1394 			__u32		btf_id;
1395 			__u32		link_id;
1396 		};
1397 		__u32		next_id;
1398 		__u32		open_flags;
1399 	};
1400 
1401 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
1402 		__u32		bpf_fd;
1403 		__u32		info_len;
1404 		__aligned_u64	info;
1405 	} info;
1406 
1407 	struct { /* anonymous struct used by BPF_PROG_QUERY command */
1408 		__u32		target_fd;	/* container object to query */
1409 		__u32		attach_type;
1410 		__u32		query_flags;
1411 		__u32		attach_flags;
1412 		__aligned_u64	prog_ids;
1413 		__u32		prog_cnt;
1414 	} query;
1415 
1416 	struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */
1417 		__u64 name;
1418 		__u32 prog_fd;
1419 	} raw_tracepoint;
1420 
1421 	struct { /* anonymous struct for BPF_BTF_LOAD */
1422 		__aligned_u64	btf;
1423 		__aligned_u64	btf_log_buf;
1424 		__u32		btf_size;
1425 		__u32		btf_log_size;
1426 		__u32		btf_log_level;
1427 	};
1428 
1429 	struct {
1430 		__u32		pid;		/* input: pid */
1431 		__u32		fd;		/* input: fd */
1432 		__u32		flags;		/* input: flags */
1433 		__u32		buf_len;	/* input/output: buf len */
1434 		__aligned_u64	buf;		/* input/output:
1435 						 *   tp_name for tracepoint
1436 						 *   symbol for kprobe
1437 						 *   filename for uprobe
1438 						 */
1439 		__u32		prog_id;	/* output: prod_id */
1440 		__u32		fd_type;	/* output: BPF_FD_TYPE_* */
1441 		__u64		probe_offset;	/* output: probe_offset */
1442 		__u64		probe_addr;	/* output: probe_addr */
1443 	} task_fd_query;
1444 
1445 	struct { /* struct used by BPF_LINK_CREATE command */
1446 		__u32		prog_fd;	/* eBPF program to attach */
1447 		union {
1448 			__u32		target_fd;	/* object to attach to */
1449 			__u32		target_ifindex; /* target ifindex */
1450 		};
1451 		__u32		attach_type;	/* attach type */
1452 		__u32		flags;		/* extra flags */
1453 		union {
1454 			__u32		target_btf_id;	/* btf_id of target to attach to */
1455 			struct {
1456 				__aligned_u64	iter_info;	/* extra bpf_iter_link_info */
1457 				__u32		iter_info_len;	/* iter_info length */
1458 			};
1459 			struct {
1460 				/* black box user-provided value passed through
1461 				 * to BPF program at the execution time and
1462 				 * accessible through bpf_get_attach_cookie() BPF helper
1463 				 */
1464 				__u64		bpf_cookie;
1465 			} perf_event;
1466 		};
1467 	} link_create;
1468 
1469 	struct { /* struct used by BPF_LINK_UPDATE command */
1470 		__u32		link_fd;	/* link fd */
1471 		/* new program fd to update link with */
1472 		__u32		new_prog_fd;
1473 		__u32		flags;		/* extra flags */
1474 		/* expected link's program fd; is specified only if
1475 		 * BPF_F_REPLACE flag is set in flags */
1476 		__u32		old_prog_fd;
1477 	} link_update;
1478 
1479 	struct {
1480 		__u32		link_fd;
1481 	} link_detach;
1482 
1483 	struct { /* struct used by BPF_ENABLE_STATS command */
1484 		__u32		type;
1485 	} enable_stats;
1486 
1487 	struct { /* struct used by BPF_ITER_CREATE command */
1488 		__u32		link_fd;
1489 		__u32		flags;
1490 	} iter_create;
1491 
1492 	struct { /* struct used by BPF_PROG_BIND_MAP command */
1493 		__u32		prog_fd;
1494 		__u32		map_fd;
1495 		__u32		flags;		/* extra flags */
1496 	} prog_bind_map;
1497 
1498 } __attribute__((aligned(8)));
1499 
1500 /* The description below is an attempt at providing documentation to eBPF
1501  * developers about the multiple available eBPF helper functions. It can be
1502  * parsed and used to produce a manual page. The workflow is the following,
1503  * and requires the rst2man utility:
1504  *
1505  *     $ ./scripts/bpf_doc.py \
1506  *             --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
1507  *     $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
1508  *     $ man /tmp/bpf-helpers.7
1509  *
1510  * Note that in order to produce this external documentation, some RST
1511  * formatting is used in the descriptions to get "bold" and "italics" in
1512  * manual pages. Also note that the few trailing white spaces are
1513  * intentional, removing them would break paragraphs for rst2man.
1514  *
1515  * Start of BPF helper function descriptions:
1516  *
1517  * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
1518  * 	Description
1519  * 		Perform a lookup in *map* for an entry associated to *key*.
1520  * 	Return
1521  * 		Map value associated to *key*, or **NULL** if no entry was
1522  * 		found.
1523  *
1524  * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
1525  * 	Description
1526  * 		Add or update the value of the entry associated to *key* in
1527  * 		*map* with *value*. *flags* is one of:
1528  *
1529  * 		**BPF_NOEXIST**
1530  * 			The entry for *key* must not exist in the map.
1531  * 		**BPF_EXIST**
1532  * 			The entry for *key* must already exist in the map.
1533  * 		**BPF_ANY**
1534  * 			No condition on the existence of the entry for *key*.
1535  *
1536  * 		Flag value **BPF_NOEXIST** cannot be used for maps of types
1537  * 		**BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY**  (all
1538  * 		elements always exist), the helper would return an error.
1539  * 	Return
1540  * 		0 on success, or a negative error in case of failure.
1541  *
1542  * long bpf_map_delete_elem(struct bpf_map *map, const void *key)
1543  * 	Description
1544  * 		Delete entry with *key* from *map*.
1545  * 	Return
1546  * 		0 on success, or a negative error in case of failure.
1547  *
1548  * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr)
1549  * 	Description
1550  * 		For tracing programs, safely attempt to read *size* bytes from
1551  * 		kernel space address *unsafe_ptr* and store the data in *dst*.
1552  *
1553  * 		Generally, use **bpf_probe_read_user**\ () or
1554  * 		**bpf_probe_read_kernel**\ () instead.
1555  * 	Return
1556  * 		0 on success, or a negative error in case of failure.
1557  *
1558  * u64 bpf_ktime_get_ns(void)
1559  * 	Description
1560  * 		Return the time elapsed since system boot, in nanoseconds.
1561  * 		Does not include time the system was suspended.
1562  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC**)
1563  * 	Return
1564  * 		Current *ktime*.
1565  *
1566  * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
1567  * 	Description
1568  * 		This helper is a "printk()-like" facility for debugging. It
1569  * 		prints a message defined by format *fmt* (of size *fmt_size*)
1570  * 		to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
1571  * 		available. It can take up to three additional **u64**
1572  * 		arguments (as an eBPF helpers, the total number of arguments is
1573  * 		limited to five).
1574  *
1575  * 		Each time the helper is called, it appends a line to the trace.
1576  * 		Lines are discarded while *\/sys/kernel/debug/tracing/trace* is
1577  * 		open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this.
1578  * 		The format of the trace is customizable, and the exact output
1579  * 		one will get depends on the options set in
1580  * 		*\/sys/kernel/debug/tracing/trace_options* (see also the
1581  * 		*README* file under the same directory). However, it usually
1582  * 		defaults to something like:
1583  *
1584  * 		::
1585  *
1586  * 			telnet-470   [001] .N.. 419421.045894: 0x00000001: <formatted msg>
1587  *
1588  * 		In the above:
1589  *
1590  * 			* ``telnet`` is the name of the current task.
1591  * 			* ``470`` is the PID of the current task.
1592  * 			* ``001`` is the CPU number on which the task is
1593  * 			  running.
1594  * 			* In ``.N..``, each character refers to a set of
1595  * 			  options (whether irqs are enabled, scheduling
1596  * 			  options, whether hard/softirqs are running, level of
1597  * 			  preempt_disabled respectively). **N** means that
1598  * 			  **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
1599  * 			  are set.
1600  * 			* ``419421.045894`` is a timestamp.
1601  * 			* ``0x00000001`` is a fake value used by BPF for the
1602  * 			  instruction pointer register.
1603  * 			* ``<formatted msg>`` is the message formatted with
1604  * 			  *fmt*.
1605  *
1606  * 		The conversion specifiers supported by *fmt* are similar, but
1607  * 		more limited than for printk(). They are **%d**, **%i**,
1608  * 		**%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
1609  * 		**%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
1610  * 		of field, padding with zeroes, etc.) is available, and the
1611  * 		helper will return **-EINVAL** (but print nothing) if it
1612  * 		encounters an unknown specifier.
1613  *
1614  * 		Also, note that **bpf_trace_printk**\ () is slow, and should
1615  * 		only be used for debugging purposes. For this reason, a notice
1616  * 		block (spanning several lines) is printed to kernel logs and
1617  * 		states that the helper should not be used "for production use"
1618  * 		the first time this helper is used (or more precisely, when
1619  * 		**trace_printk**\ () buffers are allocated). For passing values
1620  * 		to user space, perf events should be preferred.
1621  * 	Return
1622  * 		The number of bytes written to the buffer, or a negative error
1623  * 		in case of failure.
1624  *
1625  * u32 bpf_get_prandom_u32(void)
1626  * 	Description
1627  * 		Get a pseudo-random number.
1628  *
1629  * 		From a security point of view, this helper uses its own
1630  * 		pseudo-random internal state, and cannot be used to infer the
1631  * 		seed of other random functions in the kernel. However, it is
1632  * 		essential to note that the generator used by the helper is not
1633  * 		cryptographically secure.
1634  * 	Return
1635  * 		A random 32-bit unsigned value.
1636  *
1637  * u32 bpf_get_smp_processor_id(void)
1638  * 	Description
1639  * 		Get the SMP (symmetric multiprocessing) processor id. Note that
1640  * 		all programs run with migration disabled, which means that the
1641  * 		SMP processor id is stable during all the execution of the
1642  * 		program.
1643  * 	Return
1644  * 		The SMP id of the processor running the program.
1645  *
1646  * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
1647  * 	Description
1648  * 		Store *len* bytes from address *from* into the packet
1649  * 		associated to *skb*, at *offset*. *flags* are a combination of
1650  * 		**BPF_F_RECOMPUTE_CSUM** (automatically recompute the
1651  * 		checksum for the packet after storing the bytes) and
1652  * 		**BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
1653  * 		**->swhash** and *skb*\ **->l4hash** to 0).
1654  *
1655  * 		A call to this helper is susceptible to change the underlying
1656  * 		packet buffer. Therefore, at load time, all checks on pointers
1657  * 		previously done by the verifier are invalidated and must be
1658  * 		performed again, if the helper is used in combination with
1659  * 		direct packet access.
1660  * 	Return
1661  * 		0 on success, or a negative error in case of failure.
1662  *
1663  * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
1664  * 	Description
1665  * 		Recompute the layer 3 (e.g. IP) checksum for the packet
1666  * 		associated to *skb*. Computation is incremental, so the helper
1667  * 		must know the former value of the header field that was
1668  * 		modified (*from*), the new value of this field (*to*), and the
1669  * 		number of bytes (2 or 4) for this field, stored in *size*.
1670  * 		Alternatively, it is possible to store the difference between
1671  * 		the previous and the new values of the header field in *to*, by
1672  * 		setting *from* and *size* to 0. For both methods, *offset*
1673  * 		indicates the location of the IP checksum within the packet.
1674  *
1675  * 		This helper works in combination with **bpf_csum_diff**\ (),
1676  * 		which does not update the checksum in-place, but offers more
1677  * 		flexibility and can handle sizes larger than 2 or 4 for the
1678  * 		checksum to update.
1679  *
1680  * 		A call to this helper is susceptible to change the underlying
1681  * 		packet buffer. Therefore, at load time, all checks on pointers
1682  * 		previously done by the verifier are invalidated and must be
1683  * 		performed again, if the helper is used in combination with
1684  * 		direct packet access.
1685  * 	Return
1686  * 		0 on success, or a negative error in case of failure.
1687  *
1688  * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
1689  * 	Description
1690  * 		Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
1691  * 		packet associated to *skb*. Computation is incremental, so the
1692  * 		helper must know the former value of the header field that was
1693  * 		modified (*from*), the new value of this field (*to*), and the
1694  * 		number of bytes (2 or 4) for this field, stored on the lowest
1695  * 		four bits of *flags*. Alternatively, it is possible to store
1696  * 		the difference between the previous and the new values of the
1697  * 		header field in *to*, by setting *from* and the four lowest
1698  * 		bits of *flags* to 0. For both methods, *offset* indicates the
1699  * 		location of the IP checksum within the packet. In addition to
1700  * 		the size of the field, *flags* can be added (bitwise OR) actual
1701  * 		flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
1702  * 		untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
1703  * 		for updates resulting in a null checksum the value is set to
1704  * 		**CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
1705  * 		the checksum is to be computed against a pseudo-header.
1706  *
1707  * 		This helper works in combination with **bpf_csum_diff**\ (),
1708  * 		which does not update the checksum in-place, but offers more
1709  * 		flexibility and can handle sizes larger than 2 or 4 for the
1710  * 		checksum to update.
1711  *
1712  * 		A call to this helper is susceptible to change the underlying
1713  * 		packet buffer. Therefore, at load time, all checks on pointers
1714  * 		previously done by the verifier are invalidated and must be
1715  * 		performed again, if the helper is used in combination with
1716  * 		direct packet access.
1717  * 	Return
1718  * 		0 on success, or a negative error in case of failure.
1719  *
1720  * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
1721  * 	Description
1722  * 		This special helper is used to trigger a "tail call", or in
1723  * 		other words, to jump into another eBPF program. The same stack
1724  * 		frame is used (but values on stack and in registers for the
1725  * 		caller are not accessible to the callee). This mechanism allows
1726  * 		for program chaining, either for raising the maximum number of
1727  * 		available eBPF instructions, or to execute given programs in
1728  * 		conditional blocks. For security reasons, there is an upper
1729  * 		limit to the number of successive tail calls that can be
1730  * 		performed.
1731  *
1732  * 		Upon call of this helper, the program attempts to jump into a
1733  * 		program referenced at index *index* in *prog_array_map*, a
1734  * 		special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
1735  * 		*ctx*, a pointer to the context.
1736  *
1737  * 		If the call succeeds, the kernel immediately runs the first
1738  * 		instruction of the new program. This is not a function call,
1739  * 		and it never returns to the previous program. If the call
1740  * 		fails, then the helper has no effect, and the caller continues
1741  * 		to run its subsequent instructions. A call can fail if the
1742  * 		destination program for the jump does not exist (i.e. *index*
1743  * 		is superior to the number of entries in *prog_array_map*), or
1744  * 		if the maximum number of tail calls has been reached for this
1745  * 		chain of programs. This limit is defined in the kernel by the
1746  * 		macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
1747  * 		which is currently set to 32.
1748  * 	Return
1749  * 		0 on success, or a negative error in case of failure.
1750  *
1751  * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
1752  * 	Description
1753  * 		Clone and redirect the packet associated to *skb* to another
1754  * 		net device of index *ifindex*. Both ingress and egress
1755  * 		interfaces can be used for redirection. The **BPF_F_INGRESS**
1756  * 		value in *flags* is used to make the distinction (ingress path
1757  * 		is selected if the flag is present, egress path otherwise).
1758  * 		This is the only flag supported for now.
1759  *
1760  * 		In comparison with **bpf_redirect**\ () helper,
1761  * 		**bpf_clone_redirect**\ () has the associated cost of
1762  * 		duplicating the packet buffer, but this can be executed out of
1763  * 		the eBPF program. Conversely, **bpf_redirect**\ () is more
1764  * 		efficient, but it is handled through an action code where the
1765  * 		redirection happens only after the eBPF program has returned.
1766  *
1767  * 		A call to this helper is susceptible to change the underlying
1768  * 		packet buffer. Therefore, at load time, all checks on pointers
1769  * 		previously done by the verifier are invalidated and must be
1770  * 		performed again, if the helper is used in combination with
1771  * 		direct packet access.
1772  * 	Return
1773  * 		0 on success, or a negative error in case of failure.
1774  *
1775  * u64 bpf_get_current_pid_tgid(void)
1776  * 	Return
1777  * 		A 64-bit integer containing the current tgid and pid, and
1778  * 		created as such:
1779  * 		*current_task*\ **->tgid << 32 \|**
1780  * 		*current_task*\ **->pid**.
1781  *
1782  * u64 bpf_get_current_uid_gid(void)
1783  * 	Return
1784  * 		A 64-bit integer containing the current GID and UID, and
1785  * 		created as such: *current_gid* **<< 32 \|** *current_uid*.
1786  *
1787  * long bpf_get_current_comm(void *buf, u32 size_of_buf)
1788  * 	Description
1789  * 		Copy the **comm** attribute of the current task into *buf* of
1790  * 		*size_of_buf*. The **comm** attribute contains the name of
1791  * 		the executable (excluding the path) for the current task. The
1792  * 		*size_of_buf* must be strictly positive. On success, the
1793  * 		helper makes sure that the *buf* is NUL-terminated. On failure,
1794  * 		it is filled with zeroes.
1795  * 	Return
1796  * 		0 on success, or a negative error in case of failure.
1797  *
1798  * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
1799  * 	Description
1800  * 		Retrieve the classid for the current task, i.e. for the net_cls
1801  * 		cgroup to which *skb* belongs.
1802  *
1803  * 		This helper can be used on TC egress path, but not on ingress.
1804  *
1805  * 		The net_cls cgroup provides an interface to tag network packets
1806  * 		based on a user-provided identifier for all traffic coming from
1807  * 		the tasks belonging to the related cgroup. See also the related
1808  * 		kernel documentation, available from the Linux sources in file
1809  * 		*Documentation/admin-guide/cgroup-v1/net_cls.rst*.
1810  *
1811  * 		The Linux kernel has two versions for cgroups: there are
1812  * 		cgroups v1 and cgroups v2. Both are available to users, who can
1813  * 		use a mixture of them, but note that the net_cls cgroup is for
1814  * 		cgroup v1 only. This makes it incompatible with BPF programs
1815  * 		run on cgroups, which is a cgroup-v2-only feature (a socket can
1816  * 		only hold data for one version of cgroups at a time).
1817  *
1818  * 		This helper is only available is the kernel was compiled with
1819  * 		the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
1820  * 		"**y**" or to "**m**".
1821  * 	Return
1822  * 		The classid, or 0 for the default unconfigured classid.
1823  *
1824  * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
1825  * 	Description
1826  * 		Push a *vlan_tci* (VLAN tag control information) of protocol
1827  * 		*vlan_proto* to the packet associated to *skb*, then update
1828  * 		the checksum. Note that if *vlan_proto* is different from
1829  * 		**ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
1830  * 		be **ETH_P_8021Q**.
1831  *
1832  * 		A call to this helper is susceptible to change the underlying
1833  * 		packet buffer. Therefore, at load time, all checks on pointers
1834  * 		previously done by the verifier are invalidated and must be
1835  * 		performed again, if the helper is used in combination with
1836  * 		direct packet access.
1837  * 	Return
1838  * 		0 on success, or a negative error in case of failure.
1839  *
1840  * long bpf_skb_vlan_pop(struct sk_buff *skb)
1841  * 	Description
1842  * 		Pop a VLAN header from the packet associated to *skb*.
1843  *
1844  * 		A call to this helper is susceptible to change the underlying
1845  * 		packet buffer. Therefore, at load time, all checks on pointers
1846  * 		previously done by the verifier are invalidated and must be
1847  * 		performed again, if the helper is used in combination with
1848  * 		direct packet access.
1849  * 	Return
1850  * 		0 on success, or a negative error in case of failure.
1851  *
1852  * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1853  * 	Description
1854  * 		Get tunnel metadata. This helper takes a pointer *key* to an
1855  * 		empty **struct bpf_tunnel_key** of **size**, that will be
1856  * 		filled with tunnel metadata for the packet associated to *skb*.
1857  * 		The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
1858  * 		indicates that the tunnel is based on IPv6 protocol instead of
1859  * 		IPv4.
1860  *
1861  * 		The **struct bpf_tunnel_key** is an object that generalizes the
1862  * 		principal parameters used by various tunneling protocols into a
1863  * 		single struct. This way, it can be used to easily make a
1864  * 		decision based on the contents of the encapsulation header,
1865  * 		"summarized" in this struct. In particular, it holds the IP
1866  * 		address of the remote end (IPv4 or IPv6, depending on the case)
1867  * 		in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
1868  * 		this struct exposes the *key*\ **->tunnel_id**, which is
1869  * 		generally mapped to a VNI (Virtual Network Identifier), making
1870  * 		it programmable together with the **bpf_skb_set_tunnel_key**\
1871  * 		() helper.
1872  *
1873  * 		Let's imagine that the following code is part of a program
1874  * 		attached to the TC ingress interface, on one end of a GRE
1875  * 		tunnel, and is supposed to filter out all messages coming from
1876  * 		remote ends with IPv4 address other than 10.0.0.1:
1877  *
1878  * 		::
1879  *
1880  * 			int ret;
1881  * 			struct bpf_tunnel_key key = {};
1882  *
1883  * 			ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
1884  * 			if (ret < 0)
1885  * 				return TC_ACT_SHOT;	// drop packet
1886  *
1887  * 			if (key.remote_ipv4 != 0x0a000001)
1888  * 				return TC_ACT_SHOT;	// drop packet
1889  *
1890  * 			return TC_ACT_OK;		// accept packet
1891  *
1892  * 		This interface can also be used with all encapsulation devices
1893  * 		that can operate in "collect metadata" mode: instead of having
1894  * 		one network device per specific configuration, the "collect
1895  * 		metadata" mode only requires a single device where the
1896  * 		configuration can be extracted from this helper.
1897  *
1898  * 		This can be used together with various tunnels such as VXLan,
1899  * 		Geneve, GRE or IP in IP (IPIP).
1900  * 	Return
1901  * 		0 on success, or a negative error in case of failure.
1902  *
1903  * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
1904  * 	Description
1905  * 		Populate tunnel metadata for packet associated to *skb.* The
1906  * 		tunnel metadata is set to the contents of *key*, of *size*. The
1907  * 		*flags* can be set to a combination of the following values:
1908  *
1909  * 		**BPF_F_TUNINFO_IPV6**
1910  * 			Indicate that the tunnel is based on IPv6 protocol
1911  * 			instead of IPv4.
1912  * 		**BPF_F_ZERO_CSUM_TX**
1913  * 			For IPv4 packets, add a flag to tunnel metadata
1914  * 			indicating that checksum computation should be skipped
1915  * 			and checksum set to zeroes.
1916  * 		**BPF_F_DONT_FRAGMENT**
1917  * 			Add a flag to tunnel metadata indicating that the
1918  * 			packet should not be fragmented.
1919  * 		**BPF_F_SEQ_NUMBER**
1920  * 			Add a flag to tunnel metadata indicating that a
1921  * 			sequence number should be added to tunnel header before
1922  * 			sending the packet. This flag was added for GRE
1923  * 			encapsulation, but might be used with other protocols
1924  * 			as well in the future.
1925  *
1926  * 		Here is a typical usage on the transmit path:
1927  *
1928  * 		::
1929  *
1930  * 			struct bpf_tunnel_key key;
1931  * 			     populate key ...
1932  * 			bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
1933  * 			bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
1934  *
1935  * 		See also the description of the **bpf_skb_get_tunnel_key**\ ()
1936  * 		helper for additional information.
1937  * 	Return
1938  * 		0 on success, or a negative error in case of failure.
1939  *
1940  * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
1941  * 	Description
1942  * 		Read the value of a perf event counter. This helper relies on a
1943  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
1944  * 		the perf event counter is selected when *map* is updated with
1945  * 		perf event file descriptors. The *map* is an array whose size
1946  * 		is the number of available CPUs, and each cell contains a value
1947  * 		relative to one CPU. The value to retrieve is indicated by
1948  * 		*flags*, that contains the index of the CPU to look up, masked
1949  * 		with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
1950  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
1951  * 		current CPU should be retrieved.
1952  *
1953  * 		Note that before Linux 4.13, only hardware perf event can be
1954  * 		retrieved.
1955  *
1956  * 		Also, be aware that the newer helper
1957  * 		**bpf_perf_event_read_value**\ () is recommended over
1958  * 		**bpf_perf_event_read**\ () in general. The latter has some ABI
1959  * 		quirks where error and counter value are used as a return code
1960  * 		(which is wrong to do since ranges may overlap). This issue is
1961  * 		fixed with **bpf_perf_event_read_value**\ (), which at the same
1962  * 		time provides more features over the **bpf_perf_event_read**\
1963  * 		() interface. Please refer to the description of
1964  * 		**bpf_perf_event_read_value**\ () for details.
1965  * 	Return
1966  * 		The value of the perf event counter read from the map, or a
1967  * 		negative error code in case of failure.
1968  *
1969  * long bpf_redirect(u32 ifindex, u64 flags)
1970  * 	Description
1971  * 		Redirect the packet to another net device of index *ifindex*.
1972  * 		This helper is somewhat similar to **bpf_clone_redirect**\
1973  * 		(), except that the packet is not cloned, which provides
1974  * 		increased performance.
1975  *
1976  * 		Except for XDP, both ingress and egress interfaces can be used
1977  * 		for redirection. The **BPF_F_INGRESS** value in *flags* is used
1978  * 		to make the distinction (ingress path is selected if the flag
1979  * 		is present, egress path otherwise). Currently, XDP only
1980  * 		supports redirection to the egress interface, and accepts no
1981  * 		flag at all.
1982  *
1983  * 		The same effect can also be attained with the more generic
1984  * 		**bpf_redirect_map**\ (), which uses a BPF map to store the
1985  * 		redirect target instead of providing it directly to the helper.
1986  * 	Return
1987  * 		For XDP, the helper returns **XDP_REDIRECT** on success or
1988  * 		**XDP_ABORTED** on error. For other program types, the values
1989  * 		are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
1990  * 		error.
1991  *
1992  * u32 bpf_get_route_realm(struct sk_buff *skb)
1993  * 	Description
1994  * 		Retrieve the realm or the route, that is to say the
1995  * 		**tclassid** field of the destination for the *skb*. The
1996  * 		identifier retrieved is a user-provided tag, similar to the
1997  * 		one used with the net_cls cgroup (see description for
1998  * 		**bpf_get_cgroup_classid**\ () helper), but here this tag is
1999  * 		held by a route (a destination entry), not by a task.
2000  *
2001  * 		Retrieving this identifier works with the clsact TC egress hook
2002  * 		(see also **tc-bpf(8)**), or alternatively on conventional
2003  * 		classful egress qdiscs, but not on TC ingress path. In case of
2004  * 		clsact TC egress hook, this has the advantage that, internally,
2005  * 		the destination entry has not been dropped yet in the transmit
2006  * 		path. Therefore, the destination entry does not need to be
2007  * 		artificially held via **netif_keep_dst**\ () for a classful
2008  * 		qdisc until the *skb* is freed.
2009  *
2010  * 		This helper is available only if the kernel was compiled with
2011  * 		**CONFIG_IP_ROUTE_CLASSID** configuration option.
2012  * 	Return
2013  * 		The realm of the route for the packet associated to *skb*, or 0
2014  * 		if none was found.
2015  *
2016  * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
2017  * 	Description
2018  * 		Write raw *data* blob into a special BPF perf event held by
2019  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
2020  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
2021  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
2022  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
2023  *
2024  * 		The *flags* are used to indicate the index in *map* for which
2025  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
2026  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
2027  * 		to indicate that the index of the current CPU core should be
2028  * 		used.
2029  *
2030  * 		The value to write, of *size*, is passed through eBPF stack and
2031  * 		pointed by *data*.
2032  *
2033  * 		The context of the program *ctx* needs also be passed to the
2034  * 		helper.
2035  *
2036  * 		On user space, a program willing to read the values needs to
2037  * 		call **perf_event_open**\ () on the perf event (either for
2038  * 		one or for all CPUs) and to store the file descriptor into the
2039  * 		*map*. This must be done before the eBPF program can send data
2040  * 		into it. An example is available in file
2041  * 		*samples/bpf/trace_output_user.c* in the Linux kernel source
2042  * 		tree (the eBPF program counterpart is in
2043  * 		*samples/bpf/trace_output_kern.c*).
2044  *
2045  * 		**bpf_perf_event_output**\ () achieves better performance
2046  * 		than **bpf_trace_printk**\ () for sharing data with user
2047  * 		space, and is much better suitable for streaming data from eBPF
2048  * 		programs.
2049  *
2050  * 		Note that this helper is not restricted to tracing use cases
2051  * 		and can be used with programs attached to TC or XDP as well,
2052  * 		where it allows for passing data to user space listeners. Data
2053  * 		can be:
2054  *
2055  * 		* Only custom structs,
2056  * 		* Only the packet payload, or
2057  * 		* A combination of both.
2058  * 	Return
2059  * 		0 on success, or a negative error in case of failure.
2060  *
2061  * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len)
2062  * 	Description
2063  * 		This helper was provided as an easy way to load data from a
2064  * 		packet. It can be used to load *len* bytes from *offset* from
2065  * 		the packet associated to *skb*, into the buffer pointed by
2066  * 		*to*.
2067  *
2068  * 		Since Linux 4.7, usage of this helper has mostly been replaced
2069  * 		by "direct packet access", enabling packet data to be
2070  * 		manipulated with *skb*\ **->data** and *skb*\ **->data_end**
2071  * 		pointing respectively to the first byte of packet data and to
2072  * 		the byte after the last byte of packet data. However, it
2073  * 		remains useful if one wishes to read large quantities of data
2074  * 		at once from a packet into the eBPF stack.
2075  * 	Return
2076  * 		0 on success, or a negative error in case of failure.
2077  *
2078  * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags)
2079  * 	Description
2080  * 		Walk a user or a kernel stack and return its id. To achieve
2081  * 		this, the helper needs *ctx*, which is a pointer to the context
2082  * 		on which the tracing program is executed, and a pointer to a
2083  * 		*map* of type **BPF_MAP_TYPE_STACK_TRACE**.
2084  *
2085  * 		The last argument, *flags*, holds the number of stack frames to
2086  * 		skip (from 0 to 255), masked with
2087  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2088  * 		a combination of the following flags:
2089  *
2090  * 		**BPF_F_USER_STACK**
2091  * 			Collect a user space stack instead of a kernel stack.
2092  * 		**BPF_F_FAST_STACK_CMP**
2093  * 			Compare stacks by hash only.
2094  * 		**BPF_F_REUSE_STACKID**
2095  * 			If two different stacks hash into the same *stackid*,
2096  * 			discard the old one.
2097  *
2098  * 		The stack id retrieved is a 32 bit long integer handle which
2099  * 		can be further combined with other data (including other stack
2100  * 		ids) and used as a key into maps. This can be useful for
2101  * 		generating a variety of graphs (such as flame graphs or off-cpu
2102  * 		graphs).
2103  *
2104  * 		For walking a stack, this helper is an improvement over
2105  * 		**bpf_probe_read**\ (), which can be used with unrolled loops
2106  * 		but is not efficient and consumes a lot of eBPF instructions.
2107  * 		Instead, **bpf_get_stackid**\ () can collect up to
2108  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
2109  * 		this limit can be controlled with the **sysctl** program, and
2110  * 		that it should be manually increased in order to profile long
2111  * 		user stacks (such as stacks for Java programs). To do so, use:
2112  *
2113  * 		::
2114  *
2115  * 			# sysctl kernel.perf_event_max_stack=<new value>
2116  * 	Return
2117  * 		The positive or null stack id on success, or a negative error
2118  * 		in case of failure.
2119  *
2120  * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
2121  * 	Description
2122  * 		Compute a checksum difference, from the raw buffer pointed by
2123  * 		*from*, of length *from_size* (that must be a multiple of 4),
2124  * 		towards the raw buffer pointed by *to*, of size *to_size*
2125  * 		(same remark). An optional *seed* can be added to the value
2126  * 		(this can be cascaded, the seed may come from a previous call
2127  * 		to the helper).
2128  *
2129  * 		This is flexible enough to be used in several ways:
2130  *
2131  * 		* With *from_size* == 0, *to_size* > 0 and *seed* set to
2132  * 		  checksum, it can be used when pushing new data.
2133  * 		* With *from_size* > 0, *to_size* == 0 and *seed* set to
2134  * 		  checksum, it can be used when removing data from a packet.
2135  * 		* With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
2136  * 		  can be used to compute a diff. Note that *from_size* and
2137  * 		  *to_size* do not need to be equal.
2138  *
2139  * 		This helper can be used in combination with
2140  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
2141  * 		which one can feed in the difference computed with
2142  * 		**bpf_csum_diff**\ ().
2143  * 	Return
2144  * 		The checksum result, or a negative error code in case of
2145  * 		failure.
2146  *
2147  * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2148  * 	Description
2149  * 		Retrieve tunnel options metadata for the packet associated to
2150  * 		*skb*, and store the raw tunnel option data to the buffer *opt*
2151  * 		of *size*.
2152  *
2153  * 		This helper can be used with encapsulation devices that can
2154  * 		operate in "collect metadata" mode (please refer to the related
2155  * 		note in the description of **bpf_skb_get_tunnel_key**\ () for
2156  * 		more details). A particular example where this can be used is
2157  * 		in combination with the Geneve encapsulation protocol, where it
2158  * 		allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
2159  * 		and retrieving arbitrary TLVs (Type-Length-Value headers) from
2160  * 		the eBPF program. This allows for full customization of these
2161  * 		headers.
2162  * 	Return
2163  * 		The size of the option data retrieved.
2164  *
2165  * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size)
2166  * 	Description
2167  * 		Set tunnel options metadata for the packet associated to *skb*
2168  * 		to the option data contained in the raw buffer *opt* of *size*.
2169  *
2170  * 		See also the description of the **bpf_skb_get_tunnel_opt**\ ()
2171  * 		helper for additional information.
2172  * 	Return
2173  * 		0 on success, or a negative error in case of failure.
2174  *
2175  * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
2176  * 	Description
2177  * 		Change the protocol of the *skb* to *proto*. Currently
2178  * 		supported are transition from IPv4 to IPv6, and from IPv6 to
2179  * 		IPv4. The helper takes care of the groundwork for the
2180  * 		transition, including resizing the socket buffer. The eBPF
2181  * 		program is expected to fill the new headers, if any, via
2182  * 		**skb_store_bytes**\ () and to recompute the checksums with
2183  * 		**bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
2184  * 		(). The main case for this helper is to perform NAT64
2185  * 		operations out of an eBPF program.
2186  *
2187  * 		Internally, the GSO type is marked as dodgy so that headers are
2188  * 		checked and segments are recalculated by the GSO/GRO engine.
2189  * 		The size for GSO target is adapted as well.
2190  *
2191  * 		All values for *flags* are reserved for future usage, and must
2192  * 		be left at zero.
2193  *
2194  * 		A call to this helper is susceptible to change the underlying
2195  * 		packet buffer. Therefore, at load time, all checks on pointers
2196  * 		previously done by the verifier are invalidated and must be
2197  * 		performed again, if the helper is used in combination with
2198  * 		direct packet access.
2199  * 	Return
2200  * 		0 on success, or a negative error in case of failure.
2201  *
2202  * long bpf_skb_change_type(struct sk_buff *skb, u32 type)
2203  * 	Description
2204  * 		Change the packet type for the packet associated to *skb*. This
2205  * 		comes down to setting *skb*\ **->pkt_type** to *type*, except
2206  * 		the eBPF program does not have a write access to *skb*\
2207  * 		**->pkt_type** beside this helper. Using a helper here allows
2208  * 		for graceful handling of errors.
2209  *
2210  * 		The major use case is to change incoming *skb*s to
2211  * 		**PACKET_HOST** in a programmatic way instead of having to
2212  * 		recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
2213  * 		example.
2214  *
2215  * 		Note that *type* only allows certain values. At this time, they
2216  * 		are:
2217  *
2218  * 		**PACKET_HOST**
2219  * 			Packet is for us.
2220  * 		**PACKET_BROADCAST**
2221  * 			Send packet to all.
2222  * 		**PACKET_MULTICAST**
2223  * 			Send packet to group.
2224  * 		**PACKET_OTHERHOST**
2225  * 			Send packet to someone else.
2226  * 	Return
2227  * 		0 on success, or a negative error in case of failure.
2228  *
2229  * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
2230  * 	Description
2231  * 		Check whether *skb* is a descendant of the cgroup2 held by
2232  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2233  * 	Return
2234  * 		The return value depends on the result of the test, and can be:
2235  *
2236  * 		* 0, if the *skb* failed the cgroup2 descendant test.
2237  * 		* 1, if the *skb* succeeded the cgroup2 descendant test.
2238  * 		* A negative error code, if an error occurred.
2239  *
2240  * u32 bpf_get_hash_recalc(struct sk_buff *skb)
2241  * 	Description
2242  * 		Retrieve the hash of the packet, *skb*\ **->hash**. If it is
2243  * 		not set, in particular if the hash was cleared due to mangling,
2244  * 		recompute this hash. Later accesses to the hash can be done
2245  * 		directly with *skb*\ **->hash**.
2246  *
2247  * 		Calling **bpf_set_hash_invalid**\ (), changing a packet
2248  * 		prototype with **bpf_skb_change_proto**\ (), or calling
2249  * 		**bpf_skb_store_bytes**\ () with the
2250  * 		**BPF_F_INVALIDATE_HASH** are actions susceptible to clear
2251  * 		the hash and to trigger a new computation for the next call to
2252  * 		**bpf_get_hash_recalc**\ ().
2253  * 	Return
2254  * 		The 32-bit hash.
2255  *
2256  * u64 bpf_get_current_task(void)
2257  * 	Return
2258  * 		A pointer to the current task struct.
2259  *
2260  * long bpf_probe_write_user(void *dst, const void *src, u32 len)
2261  * 	Description
2262  * 		Attempt in a safe way to write *len* bytes from the buffer
2263  * 		*src* to *dst* in memory. It only works for threads that are in
2264  * 		user context, and *dst* must be a valid user space address.
2265  *
2266  * 		This helper should not be used to implement any kind of
2267  * 		security mechanism because of TOC-TOU attacks, but rather to
2268  * 		debug, divert, and manipulate execution of semi-cooperative
2269  * 		processes.
2270  *
2271  * 		Keep in mind that this feature is meant for experiments, and it
2272  * 		has a risk of crashing the system and running programs.
2273  * 		Therefore, when an eBPF program using this helper is attached,
2274  * 		a warning including PID and process name is printed to kernel
2275  * 		logs.
2276  * 	Return
2277  * 		0 on success, or a negative error in case of failure.
2278  *
2279  * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
2280  * 	Description
2281  * 		Check whether the probe is being run is the context of a given
2282  * 		subset of the cgroup2 hierarchy. The cgroup2 to test is held by
2283  * 		*map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
2284  * 	Return
2285  * 		The return value depends on the result of the test, and can be:
2286  *
2287  *		* 0, if current task belongs to the cgroup2.
2288  *		* 1, if current task does not belong to the cgroup2.
2289  * 		* A negative error code, if an error occurred.
2290  *
2291  * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
2292  * 	Description
2293  * 		Resize (trim or grow) the packet associated to *skb* to the
2294  * 		new *len*. The *flags* are reserved for future usage, and must
2295  * 		be left at zero.
2296  *
2297  * 		The basic idea is that the helper performs the needed work to
2298  * 		change the size of the packet, then the eBPF program rewrites
2299  * 		the rest via helpers like **bpf_skb_store_bytes**\ (),
2300  * 		**bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
2301  * 		and others. This helper is a slow path utility intended for
2302  * 		replies with control messages. And because it is targeted for
2303  * 		slow path, the helper itself can afford to be slow: it
2304  * 		implicitly linearizes, unclones and drops offloads from the
2305  * 		*skb*.
2306  *
2307  * 		A call to this helper is susceptible to change the underlying
2308  * 		packet buffer. Therefore, at load time, all checks on pointers
2309  * 		previously done by the verifier are invalidated and must be
2310  * 		performed again, if the helper is used in combination with
2311  * 		direct packet access.
2312  * 	Return
2313  * 		0 on success, or a negative error in case of failure.
2314  *
2315  * long bpf_skb_pull_data(struct sk_buff *skb, u32 len)
2316  * 	Description
2317  * 		Pull in non-linear data in case the *skb* is non-linear and not
2318  * 		all of *len* are part of the linear section. Make *len* bytes
2319  * 		from *skb* readable and writable. If a zero value is passed for
2320  * 		*len*, then the whole length of the *skb* is pulled.
2321  *
2322  * 		This helper is only needed for reading and writing with direct
2323  * 		packet access.
2324  *
2325  * 		For direct packet access, testing that offsets to access
2326  * 		are within packet boundaries (test on *skb*\ **->data_end**) is
2327  * 		susceptible to fail if offsets are invalid, or if the requested
2328  * 		data is in non-linear parts of the *skb*. On failure the
2329  * 		program can just bail out, or in the case of a non-linear
2330  * 		buffer, use a helper to make the data available. The
2331  * 		**bpf_skb_load_bytes**\ () helper is a first solution to access
2332  * 		the data. Another one consists in using **bpf_skb_pull_data**
2333  * 		to pull in once the non-linear parts, then retesting and
2334  * 		eventually access the data.
2335  *
2336  * 		At the same time, this also makes sure the *skb* is uncloned,
2337  * 		which is a necessary condition for direct write. As this needs
2338  * 		to be an invariant for the write part only, the verifier
2339  * 		detects writes and adds a prologue that is calling
2340  * 		**bpf_skb_pull_data()** to effectively unclone the *skb* from
2341  * 		the very beginning in case it is indeed cloned.
2342  *
2343  * 		A call to this helper is susceptible to change the underlying
2344  * 		packet buffer. Therefore, at load time, all checks on pointers
2345  * 		previously done by the verifier are invalidated and must be
2346  * 		performed again, if the helper is used in combination with
2347  * 		direct packet access.
2348  * 	Return
2349  * 		0 on success, or a negative error in case of failure.
2350  *
2351  * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
2352  * 	Description
2353  * 		Add the checksum *csum* into *skb*\ **->csum** in case the
2354  * 		driver has supplied a checksum for the entire packet into that
2355  * 		field. Return an error otherwise. This helper is intended to be
2356  * 		used in combination with **bpf_csum_diff**\ (), in particular
2357  * 		when the checksum needs to be updated after data has been
2358  * 		written into the packet through direct packet access.
2359  * 	Return
2360  * 		The checksum on success, or a negative error code in case of
2361  * 		failure.
2362  *
2363  * void bpf_set_hash_invalid(struct sk_buff *skb)
2364  * 	Description
2365  * 		Invalidate the current *skb*\ **->hash**. It can be used after
2366  * 		mangling on headers through direct packet access, in order to
2367  * 		indicate that the hash is outdated and to trigger a
2368  * 		recalculation the next time the kernel tries to access this
2369  * 		hash or when the **bpf_get_hash_recalc**\ () helper is called.
2370  *
2371  * long bpf_get_numa_node_id(void)
2372  * 	Description
2373  * 		Return the id of the current NUMA node. The primary use case
2374  * 		for this helper is the selection of sockets for the local NUMA
2375  * 		node, when the program is attached to sockets using the
2376  * 		**SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
2377  * 		but the helper is also available to other eBPF program types,
2378  * 		similarly to **bpf_get_smp_processor_id**\ ().
2379  * 	Return
2380  * 		The id of current NUMA node.
2381  *
2382  * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
2383  * 	Description
2384  * 		Grows headroom of packet associated to *skb* and adjusts the
2385  * 		offset of the MAC header accordingly, adding *len* bytes of
2386  * 		space. It automatically extends and reallocates memory as
2387  * 		required.
2388  *
2389  * 		This helper can be used on a layer 3 *skb* to push a MAC header
2390  * 		for redirection into a layer 2 device.
2391  *
2392  * 		All values for *flags* are reserved for future usage, and must
2393  * 		be left at zero.
2394  *
2395  * 		A call to this helper is susceptible to change the underlying
2396  * 		packet buffer. Therefore, at load time, all checks on pointers
2397  * 		previously done by the verifier are invalidated and must be
2398  * 		performed again, if the helper is used in combination with
2399  * 		direct packet access.
2400  * 	Return
2401  * 		0 on success, or a negative error in case of failure.
2402  *
2403  * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
2404  * 	Description
2405  * 		Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
2406  * 		it is possible to use a negative value for *delta*. This helper
2407  * 		can be used to prepare the packet for pushing or popping
2408  * 		headers.
2409  *
2410  * 		A call to this helper is susceptible to change the underlying
2411  * 		packet buffer. Therefore, at load time, all checks on pointers
2412  * 		previously done by the verifier are invalidated and must be
2413  * 		performed again, if the helper is used in combination with
2414  * 		direct packet access.
2415  * 	Return
2416  * 		0 on success, or a negative error in case of failure.
2417  *
2418  * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr)
2419  * 	Description
2420  * 		Copy a NUL terminated string from an unsafe kernel address
2421  * 		*unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for
2422  * 		more details.
2423  *
2424  * 		Generally, use **bpf_probe_read_user_str**\ () or
2425  * 		**bpf_probe_read_kernel_str**\ () instead.
2426  * 	Return
2427  * 		On success, the strictly positive length of the string,
2428  * 		including the trailing NUL character. On error, a negative
2429  * 		value.
2430  *
2431  * u64 bpf_get_socket_cookie(struct sk_buff *skb)
2432  * 	Description
2433  * 		If the **struct sk_buff** pointed by *skb* has a known socket,
2434  * 		retrieve the cookie (generated by the kernel) of this socket.
2435  * 		If no cookie has been set yet, generate a new cookie. Once
2436  * 		generated, the socket cookie remains stable for the life of the
2437  * 		socket. This helper can be useful for monitoring per socket
2438  * 		networking traffic statistics as it provides a global socket
2439  * 		identifier that can be assumed unique.
2440  * 	Return
2441  * 		A 8-byte long unique number on success, or 0 if the socket
2442  * 		field is missing inside *skb*.
2443  *
2444  * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
2445  * 	Description
2446  * 		Equivalent to bpf_get_socket_cookie() helper that accepts
2447  * 		*skb*, but gets socket from **struct bpf_sock_addr** context.
2448  * 	Return
2449  * 		A 8-byte long unique number.
2450  *
2451  * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
2452  * 	Description
2453  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2454  * 		*skb*, but gets socket from **struct bpf_sock_ops** context.
2455  * 	Return
2456  * 		A 8-byte long unique number.
2457  *
2458  * u64 bpf_get_socket_cookie(struct sock *sk)
2459  * 	Description
2460  * 		Equivalent to **bpf_get_socket_cookie**\ () helper that accepts
2461  * 		*sk*, but gets socket from a BTF **struct sock**. This helper
2462  * 		also works for sleepable programs.
2463  * 	Return
2464  * 		A 8-byte long unique number or 0 if *sk* is NULL.
2465  *
2466  * u32 bpf_get_socket_uid(struct sk_buff *skb)
2467  * 	Return
2468  * 		The owner UID of the socket associated to *skb*. If the socket
2469  * 		is **NULL**, or if it is not a full socket (i.e. if it is a
2470  * 		time-wait or a request socket instead), **overflowuid** value
2471  * 		is returned (note that **overflowuid** might also be the actual
2472  * 		UID value for the socket).
2473  *
2474  * long bpf_set_hash(struct sk_buff *skb, u32 hash)
2475  * 	Description
2476  * 		Set the full hash for *skb* (set the field *skb*\ **->hash**)
2477  * 		to value *hash*.
2478  * 	Return
2479  * 		0
2480  *
2481  * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2482  * 	Description
2483  * 		Emulate a call to **setsockopt()** on the socket associated to
2484  * 		*bpf_socket*, which must be a full socket. The *level* at
2485  * 		which the option resides and the name *optname* of the option
2486  * 		must be specified, see **setsockopt(2)** for more information.
2487  * 		The option value of length *optlen* is pointed by *optval*.
2488  *
2489  * 		*bpf_socket* should be one of the following:
2490  *
2491  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2492  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2493  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2494  *
2495  * 		This helper actually implements a subset of **setsockopt()**.
2496  * 		It supports the following *level*\ s:
2497  *
2498  * 		* **SOL_SOCKET**, which supports the following *optname*\ s:
2499  * 		  **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
2500  * 		  **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**,
2501  * 		  **SO_BINDTODEVICE**, **SO_KEEPALIVE**.
2502  * 		* **IPPROTO_TCP**, which supports the following *optname*\ s:
2503  * 		  **TCP_CONGESTION**, **TCP_BPF_IW**,
2504  * 		  **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**,
2505  * 		  **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**,
2506  *		  **TCP_SYNCNT**, **TCP_USER_TIMEOUT**, **TCP_NOTSENT_LOWAT**.
2507  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2508  * 		* **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2509  * 	Return
2510  * 		0 on success, or a negative error in case of failure.
2511  *
2512  * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
2513  * 	Description
2514  * 		Grow or shrink the room for data in the packet associated to
2515  * 		*skb* by *len_diff*, and according to the selected *mode*.
2516  *
2517  * 		By default, the helper will reset any offloaded checksum
2518  * 		indicator of the skb to CHECKSUM_NONE. This can be avoided
2519  * 		by the following flag:
2520  *
2521  * 		* **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded
2522  * 		  checksum data of the skb to CHECKSUM_NONE.
2523  *
2524  *		There are two supported modes at this time:
2525  *
2526  *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
2527  *		  (room space is added or removed below the layer 2 header).
2528  *
2529  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
2530  * 		  (room space is added or removed below the layer 3 header).
2531  *
2532  *		The following flags are supported at this time:
2533  *
2534  *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
2535  *		  Adjusting mss in this way is not allowed for datagrams.
2536  *
2537  *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**,
2538  *		  **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**:
2539  *		  Any new space is reserved to hold a tunnel header.
2540  *		  Configure skb offsets and other fields accordingly.
2541  *
2542  *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**,
2543  *		  **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**:
2544  *		  Use with ENCAP_L3 flags to further specify the tunnel type.
2545  *
2546  *		* **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*):
2547  *		  Use with ENCAP_L3/L4 flags to further specify the tunnel
2548  *		  type; *len* is the length of the inner MAC header.
2549  *
2550  *		* **BPF_F_ADJ_ROOM_ENCAP_L2_ETH**:
2551  *		  Use with BPF_F_ADJ_ROOM_ENCAP_L2 flag to further specify the
2552  *		  L2 type as Ethernet.
2553  *
2554  * 		A call to this helper is susceptible to change the underlying
2555  * 		packet buffer. Therefore, at load time, all checks on pointers
2556  * 		previously done by the verifier are invalidated and must be
2557  * 		performed again, if the helper is used in combination with
2558  * 		direct packet access.
2559  * 	Return
2560  * 		0 on success, or a negative error in case of failure.
2561  *
2562  * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
2563  * 	Description
2564  * 		Redirect the packet to the endpoint referenced by *map* at
2565  * 		index *key*. Depending on its type, this *map* can contain
2566  * 		references to net devices (for forwarding packets through other
2567  * 		ports), or to CPUs (for redirecting XDP frames to another CPU;
2568  * 		but this is only implemented for native XDP (with driver
2569  * 		support) as of this writing).
2570  *
2571  * 		The lower two bits of *flags* are used as the return code if
2572  * 		the map lookup fails. This is so that the return value can be
2573  * 		one of the XDP program return codes up to **XDP_TX**, as chosen
2574  * 		by the caller. The higher bits of *flags* can be set to
2575  * 		BPF_F_BROADCAST or BPF_F_EXCLUDE_INGRESS as defined below.
2576  *
2577  * 		With BPF_F_BROADCAST the packet will be broadcasted to all the
2578  * 		interfaces in the map, with BPF_F_EXCLUDE_INGRESS the ingress
2579  * 		interface will be excluded when do broadcasting.
2580  *
2581  * 		See also **bpf_redirect**\ (), which only supports redirecting
2582  * 		to an ifindex, but doesn't require a map to do so.
2583  * 	Return
2584  * 		**XDP_REDIRECT** on success, or the value of the two lower bits
2585  * 		of the *flags* argument on error.
2586  *
2587  * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags)
2588  * 	Description
2589  * 		Redirect the packet to the socket referenced by *map* (of type
2590  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2591  * 		egress interfaces can be used for redirection. The
2592  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2593  * 		distinction (ingress path is selected if the flag is present,
2594  * 		egress path otherwise). This is the only flag supported for now.
2595  * 	Return
2596  * 		**SK_PASS** on success, or **SK_DROP** on error.
2597  *
2598  * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
2599  * 	Description
2600  * 		Add an entry to, or update a *map* referencing sockets. The
2601  * 		*skops* is used as a new value for the entry associated to
2602  * 		*key*. *flags* is one of:
2603  *
2604  * 		**BPF_NOEXIST**
2605  * 			The entry for *key* must not exist in the map.
2606  * 		**BPF_EXIST**
2607  * 			The entry for *key* must already exist in the map.
2608  * 		**BPF_ANY**
2609  * 			No condition on the existence of the entry for *key*.
2610  *
2611  * 		If the *map* has eBPF programs (parser and verdict), those will
2612  * 		be inherited by the socket being added. If the socket is
2613  * 		already attached to eBPF programs, this results in an error.
2614  * 	Return
2615  * 		0 on success, or a negative error in case of failure.
2616  *
2617  * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
2618  * 	Description
2619  * 		Adjust the address pointed by *xdp_md*\ **->data_meta** by
2620  * 		*delta* (which can be positive or negative). Note that this
2621  * 		operation modifies the address stored in *xdp_md*\ **->data**,
2622  * 		so the latter must be loaded only after the helper has been
2623  * 		called.
2624  *
2625  * 		The use of *xdp_md*\ **->data_meta** is optional and programs
2626  * 		are not required to use it. The rationale is that when the
2627  * 		packet is processed with XDP (e.g. as DoS filter), it is
2628  * 		possible to push further meta data along with it before passing
2629  * 		to the stack, and to give the guarantee that an ingress eBPF
2630  * 		program attached as a TC classifier on the same device can pick
2631  * 		this up for further post-processing. Since TC works with socket
2632  * 		buffers, it remains possible to set from XDP the **mark** or
2633  * 		**priority** pointers, or other pointers for the socket buffer.
2634  * 		Having this scratch space generic and programmable allows for
2635  * 		more flexibility as the user is free to store whatever meta
2636  * 		data they need.
2637  *
2638  * 		A call to this helper is susceptible to change the underlying
2639  * 		packet buffer. Therefore, at load time, all checks on pointers
2640  * 		previously done by the verifier are invalidated and must be
2641  * 		performed again, if the helper is used in combination with
2642  * 		direct packet access.
2643  * 	Return
2644  * 		0 on success, or a negative error in case of failure.
2645  *
2646  * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
2647  * 	Description
2648  * 		Read the value of a perf event counter, and store it into *buf*
2649  * 		of size *buf_size*. This helper relies on a *map* of type
2650  * 		**BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
2651  * 		counter is selected when *map* is updated with perf event file
2652  * 		descriptors. The *map* is an array whose size is the number of
2653  * 		available CPUs, and each cell contains a value relative to one
2654  * 		CPU. The value to retrieve is indicated by *flags*, that
2655  * 		contains the index of the CPU to look up, masked with
2656  * 		**BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
2657  * 		**BPF_F_CURRENT_CPU** to indicate that the value for the
2658  * 		current CPU should be retrieved.
2659  *
2660  * 		This helper behaves in a way close to
2661  * 		**bpf_perf_event_read**\ () helper, save that instead of
2662  * 		just returning the value observed, it fills the *buf*
2663  * 		structure. This allows for additional data to be retrieved: in
2664  * 		particular, the enabled and running times (in *buf*\
2665  * 		**->enabled** and *buf*\ **->running**, respectively) are
2666  * 		copied. In general, **bpf_perf_event_read_value**\ () is
2667  * 		recommended over **bpf_perf_event_read**\ (), which has some
2668  * 		ABI issues and provides fewer functionalities.
2669  *
2670  * 		These values are interesting, because hardware PMU (Performance
2671  * 		Monitoring Unit) counters are limited resources. When there are
2672  * 		more PMU based perf events opened than available counters,
2673  * 		kernel will multiplex these events so each event gets certain
2674  * 		percentage (but not all) of the PMU time. In case that
2675  * 		multiplexing happens, the number of samples or counter value
2676  * 		will not reflect the case compared to when no multiplexing
2677  * 		occurs. This makes comparison between different runs difficult.
2678  * 		Typically, the counter value should be normalized before
2679  * 		comparing to other experiments. The usual normalization is done
2680  * 		as follows.
2681  *
2682  * 		::
2683  *
2684  * 			normalized_counter = counter * t_enabled / t_running
2685  *
2686  * 		Where t_enabled is the time enabled for event and t_running is
2687  * 		the time running for event since last normalization. The
2688  * 		enabled and running times are accumulated since the perf event
2689  * 		open. To achieve scaling factor between two invocations of an
2690  * 		eBPF program, users can use CPU id as the key (which is
2691  * 		typical for perf array usage model) to remember the previous
2692  * 		value and do the calculation inside the eBPF program.
2693  * 	Return
2694  * 		0 on success, or a negative error in case of failure.
2695  *
2696  * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
2697  * 	Description
2698  * 		For en eBPF program attached to a perf event, retrieve the
2699  * 		value of the event counter associated to *ctx* and store it in
2700  * 		the structure pointed by *buf* and of size *buf_size*. Enabled
2701  * 		and running times are also stored in the structure (see
2702  * 		description of helper **bpf_perf_event_read_value**\ () for
2703  * 		more details).
2704  * 	Return
2705  * 		0 on success, or a negative error in case of failure.
2706  *
2707  * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen)
2708  * 	Description
2709  * 		Emulate a call to **getsockopt()** on the socket associated to
2710  * 		*bpf_socket*, which must be a full socket. The *level* at
2711  * 		which the option resides and the name *optname* of the option
2712  * 		must be specified, see **getsockopt(2)** for more information.
2713  * 		The retrieved value is stored in the structure pointed by
2714  * 		*opval* and of length *optlen*.
2715  *
2716  * 		*bpf_socket* should be one of the following:
2717  *
2718  * 		* **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**.
2719  * 		* **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT**
2720  * 		  and **BPF_CGROUP_INET6_CONNECT**.
2721  *
2722  * 		This helper actually implements a subset of **getsockopt()**.
2723  * 		It supports the following *level*\ s:
2724  *
2725  * 		* **IPPROTO_TCP**, which supports *optname*
2726  * 		  **TCP_CONGESTION**.
2727  * 		* **IPPROTO_IP**, which supports *optname* **IP_TOS**.
2728  * 		* **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
2729  * 	Return
2730  * 		0 on success, or a negative error in case of failure.
2731  *
2732  * long bpf_override_return(struct pt_regs *regs, u64 rc)
2733  * 	Description
2734  * 		Used for error injection, this helper uses kprobes to override
2735  * 		the return value of the probed function, and to set it to *rc*.
2736  * 		The first argument is the context *regs* on which the kprobe
2737  * 		works.
2738  *
2739  * 		This helper works by setting the PC (program counter)
2740  * 		to an override function which is run in place of the original
2741  * 		probed function. This means the probed function is not run at
2742  * 		all. The replacement function just returns with the required
2743  * 		value.
2744  *
2745  * 		This helper has security implications, and thus is subject to
2746  * 		restrictions. It is only available if the kernel was compiled
2747  * 		with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
2748  * 		option, and in this case it only works on functions tagged with
2749  * 		**ALLOW_ERROR_INJECTION** in the kernel code.
2750  *
2751  * 		Also, the helper is only available for the architectures having
2752  * 		the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
2753  * 		x86 architecture is the only one to support this feature.
2754  * 	Return
2755  * 		0
2756  *
2757  * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
2758  * 	Description
2759  * 		Attempt to set the value of the **bpf_sock_ops_cb_flags** field
2760  * 		for the full TCP socket associated to *bpf_sock_ops* to
2761  * 		*argval*.
2762  *
2763  * 		The primary use of this field is to determine if there should
2764  * 		be calls to eBPF programs of type
2765  * 		**BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
2766  * 		code. A program of the same type can change its value, per
2767  * 		connection and as necessary, when the connection is
2768  * 		established. This field is directly accessible for reading, but
2769  * 		this helper must be used for updates in order to return an
2770  * 		error if an eBPF program tries to set a callback that is not
2771  * 		supported in the current kernel.
2772  *
2773  * 		*argval* is a flag array which can combine these flags:
2774  *
2775  * 		* **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
2776  * 		* **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
2777  * 		* **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
2778  * 		* **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT)
2779  *
2780  * 		Therefore, this function can be used to clear a callback flag by
2781  * 		setting the appropriate bit to zero. e.g. to disable the RTO
2782  * 		callback:
2783  *
2784  * 		**bpf_sock_ops_cb_flags_set(bpf_sock,**
2785  * 			**bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)**
2786  *
2787  * 		Here are some examples of where one could call such eBPF
2788  * 		program:
2789  *
2790  * 		* When RTO fires.
2791  * 		* When a packet is retransmitted.
2792  * 		* When the connection terminates.
2793  * 		* When a packet is sent.
2794  * 		* When a packet is received.
2795  * 	Return
2796  * 		Code **-EINVAL** if the socket is not a full TCP socket;
2797  * 		otherwise, a positive number containing the bits that could not
2798  * 		be set is returned (which comes down to 0 if all bits were set
2799  * 		as required).
2800  *
2801  * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
2802  * 	Description
2803  * 		This helper is used in programs implementing policies at the
2804  * 		socket level. If the message *msg* is allowed to pass (i.e. if
2805  * 		the verdict eBPF program returns **SK_PASS**), redirect it to
2806  * 		the socket referenced by *map* (of type
2807  * 		**BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
2808  * 		egress interfaces can be used for redirection. The
2809  * 		**BPF_F_INGRESS** value in *flags* is used to make the
2810  * 		distinction (ingress path is selected if the flag is present,
2811  * 		egress path otherwise). This is the only flag supported for now.
2812  * 	Return
2813  * 		**SK_PASS** on success, or **SK_DROP** on error.
2814  *
2815  * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
2816  * 	Description
2817  * 		For socket policies, apply the verdict of the eBPF program to
2818  * 		the next *bytes* (number of bytes) of message *msg*.
2819  *
2820  * 		For example, this helper can be used in the following cases:
2821  *
2822  * 		* A single **sendmsg**\ () or **sendfile**\ () system call
2823  * 		  contains multiple logical messages that the eBPF program is
2824  * 		  supposed to read and for which it should apply a verdict.
2825  * 		* An eBPF program only cares to read the first *bytes* of a
2826  * 		  *msg*. If the message has a large payload, then setting up
2827  * 		  and calling the eBPF program repeatedly for all bytes, even
2828  * 		  though the verdict is already known, would create unnecessary
2829  * 		  overhead.
2830  *
2831  * 		When called from within an eBPF program, the helper sets a
2832  * 		counter internal to the BPF infrastructure, that is used to
2833  * 		apply the last verdict to the next *bytes*. If *bytes* is
2834  * 		smaller than the current data being processed from a
2835  * 		**sendmsg**\ () or **sendfile**\ () system call, the first
2836  * 		*bytes* will be sent and the eBPF program will be re-run with
2837  * 		the pointer for start of data pointing to byte number *bytes*
2838  * 		**+ 1**. If *bytes* is larger than the current data being
2839  * 		processed, then the eBPF verdict will be applied to multiple
2840  * 		**sendmsg**\ () or **sendfile**\ () calls until *bytes* are
2841  * 		consumed.
2842  *
2843  * 		Note that if a socket closes with the internal counter holding
2844  * 		a non-zero value, this is not a problem because data is not
2845  * 		being buffered for *bytes* and is sent as it is received.
2846  * 	Return
2847  * 		0
2848  *
2849  * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
2850  * 	Description
2851  * 		For socket policies, prevent the execution of the verdict eBPF
2852  * 		program for message *msg* until *bytes* (byte number) have been
2853  * 		accumulated.
2854  *
2855  * 		This can be used when one needs a specific number of bytes
2856  * 		before a verdict can be assigned, even if the data spans
2857  * 		multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
2858  * 		case would be a user calling **sendmsg**\ () repeatedly with
2859  * 		1-byte long message segments. Obviously, this is bad for
2860  * 		performance, but it is still valid. If the eBPF program needs
2861  * 		*bytes* bytes to validate a header, this helper can be used to
2862  * 		prevent the eBPF program to be called again until *bytes* have
2863  * 		been accumulated.
2864  * 	Return
2865  * 		0
2866  *
2867  * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
2868  * 	Description
2869  * 		For socket policies, pull in non-linear data from user space
2870  * 		for *msg* and set pointers *msg*\ **->data** and *msg*\
2871  * 		**->data_end** to *start* and *end* bytes offsets into *msg*,
2872  * 		respectively.
2873  *
2874  * 		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
2875  * 		*msg* it can only parse data that the (**data**, **data_end**)
2876  * 		pointers have already consumed. For **sendmsg**\ () hooks this
2877  * 		is likely the first scatterlist element. But for calls relying
2878  * 		on the **sendpage** handler (e.g. **sendfile**\ ()) this will
2879  * 		be the range (**0**, **0**) because the data is shared with
2880  * 		user space and by default the objective is to avoid allowing
2881  * 		user space to modify data while (or after) eBPF verdict is
2882  * 		being decided. This helper can be used to pull in data and to
2883  * 		set the start and end pointer to given values. Data will be
2884  * 		copied if necessary (i.e. if data was not linear and if start
2885  * 		and end pointers do not point to the same chunk).
2886  *
2887  * 		A call to this helper is susceptible to change the underlying
2888  * 		packet buffer. Therefore, at load time, all checks on pointers
2889  * 		previously done by the verifier are invalidated and must be
2890  * 		performed again, if the helper is used in combination with
2891  * 		direct packet access.
2892  *
2893  * 		All values for *flags* are reserved for future usage, and must
2894  * 		be left at zero.
2895  * 	Return
2896  * 		0 on success, or a negative error in case of failure.
2897  *
2898  * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
2899  * 	Description
2900  * 		Bind the socket associated to *ctx* to the address pointed by
2901  * 		*addr*, of length *addr_len*. This allows for making outgoing
2902  * 		connection from the desired IP address, which can be useful for
2903  * 		example when all processes inside a cgroup should use one
2904  * 		single IP address on a host that has multiple IP configured.
2905  *
2906  * 		This helper works for IPv4 and IPv6, TCP and UDP sockets. The
2907  * 		domain (*addr*\ **->sa_family**) must be **AF_INET** (or
2908  * 		**AF_INET6**). It's advised to pass zero port (**sin_port**
2909  * 		or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like
2910  * 		behavior and lets the kernel efficiently pick up an unused
2911  * 		port as long as 4-tuple is unique. Passing non-zero port might
2912  * 		lead to degraded performance.
2913  * 	Return
2914  * 		0 on success, or a negative error in case of failure.
2915  *
2916  * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
2917  * 	Description
2918  * 		Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
2919  * 		possible to both shrink and grow the packet tail.
2920  * 		Shrink done via *delta* being a negative integer.
2921  *
2922  * 		A call to this helper is susceptible to change the underlying
2923  * 		packet buffer. Therefore, at load time, all checks on pointers
2924  * 		previously done by the verifier are invalidated and must be
2925  * 		performed again, if the helper is used in combination with
2926  * 		direct packet access.
2927  * 	Return
2928  * 		0 on success, or a negative error in case of failure.
2929  *
2930  * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
2931  * 	Description
2932  * 		Retrieve the XFRM state (IP transform framework, see also
2933  * 		**ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
2934  *
2935  * 		The retrieved value is stored in the **struct bpf_xfrm_state**
2936  * 		pointed by *xfrm_state* and of length *size*.
2937  *
2938  * 		All values for *flags* are reserved for future usage, and must
2939  * 		be left at zero.
2940  *
2941  * 		This helper is available only if the kernel was compiled with
2942  * 		**CONFIG_XFRM** configuration option.
2943  * 	Return
2944  * 		0 on success, or a negative error in case of failure.
2945  *
2946  * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags)
2947  * 	Description
2948  * 		Return a user or a kernel stack in bpf program provided buffer.
2949  * 		To achieve this, the helper needs *ctx*, which is a pointer
2950  * 		to the context on which the tracing program is executed.
2951  * 		To store the stacktrace, the bpf program provides *buf* with
2952  * 		a nonnegative *size*.
2953  *
2954  * 		The last argument, *flags*, holds the number of stack frames to
2955  * 		skip (from 0 to 255), masked with
2956  * 		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
2957  * 		the following flags:
2958  *
2959  * 		**BPF_F_USER_STACK**
2960  * 			Collect a user space stack instead of a kernel stack.
2961  * 		**BPF_F_USER_BUILD_ID**
2962  * 			Collect buildid+offset instead of ips for user stack,
2963  * 			only valid if **BPF_F_USER_STACK** is also specified.
2964  *
2965  * 		**bpf_get_stack**\ () can collect up to
2966  * 		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
2967  * 		to sufficient large buffer size. Note that
2968  * 		this limit can be controlled with the **sysctl** program, and
2969  * 		that it should be manually increased in order to profile long
2970  * 		user stacks (such as stacks for Java programs). To do so, use:
2971  *
2972  * 		::
2973  *
2974  * 			# sysctl kernel.perf_event_max_stack=<new value>
2975  * 	Return
2976  * 		A non-negative value equal to or less than *size* on success,
2977  * 		or a negative error in case of failure.
2978  *
2979  * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header)
2980  * 	Description
2981  * 		This helper is similar to **bpf_skb_load_bytes**\ () in that
2982  * 		it provides an easy way to load *len* bytes from *offset*
2983  * 		from the packet associated to *skb*, into the buffer pointed
2984  * 		by *to*. The difference to **bpf_skb_load_bytes**\ () is that
2985  * 		a fifth argument *start_header* exists in order to select a
2986  * 		base offset to start from. *start_header* can be one of:
2987  *
2988  * 		**BPF_HDR_START_MAC**
2989  * 			Base offset to load data from is *skb*'s mac header.
2990  * 		**BPF_HDR_START_NET**
2991  * 			Base offset to load data from is *skb*'s network header.
2992  *
2993  * 		In general, "direct packet access" is the preferred method to
2994  * 		access packet data, however, this helper is in particular useful
2995  * 		in socket filters where *skb*\ **->data** does not always point
2996  * 		to the start of the mac header and where "direct packet access"
2997  * 		is not available.
2998  * 	Return
2999  * 		0 on success, or a negative error in case of failure.
3000  *
3001  * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
3002  *	Description
3003  *		Do FIB lookup in kernel tables using parameters in *params*.
3004  *		If lookup is successful and result shows packet is to be
3005  *		forwarded, the neighbor tables are searched for the nexthop.
3006  *		If successful (ie., FIB lookup shows forwarding and nexthop
3007  *		is resolved), the nexthop address is returned in ipv4_dst
3008  *		or ipv6_dst based on family, smac is set to mac address of
3009  *		egress device, dmac is set to nexthop mac address, rt_metric
3010  *		is set to metric from route (IPv4/IPv6 only), and ifindex
3011  *		is set to the device index of the nexthop from the FIB lookup.
3012  *
3013  *		*plen* argument is the size of the passed in struct.
3014  *		*flags* argument can be a combination of one or more of the
3015  *		following values:
3016  *
3017  *		**BPF_FIB_LOOKUP_DIRECT**
3018  *			Do a direct table lookup vs full lookup using FIB
3019  *			rules.
3020  *		**BPF_FIB_LOOKUP_OUTPUT**
3021  *			Perform lookup from an egress perspective (default is
3022  *			ingress).
3023  *
3024  *		*ctx* is either **struct xdp_md** for XDP programs or
3025  *		**struct sk_buff** tc cls_act programs.
3026  *	Return
3027  *		* < 0 if any input argument is invalid
3028  *		*   0 on success (packet is forwarded, nexthop neighbor exists)
3029  *		* > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
3030  *		  packet is not forwarded or needs assist from full stack
3031  *
3032  *		If lookup fails with BPF_FIB_LKUP_RET_FRAG_NEEDED, then the MTU
3033  *		was exceeded and output params->mtu_result contains the MTU.
3034  *
3035  * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
3036  *	Description
3037  *		Add an entry to, or update a sockhash *map* referencing sockets.
3038  *		The *skops* is used as a new value for the entry associated to
3039  *		*key*. *flags* is one of:
3040  *
3041  *		**BPF_NOEXIST**
3042  *			The entry for *key* must not exist in the map.
3043  *		**BPF_EXIST**
3044  *			The entry for *key* must already exist in the map.
3045  *		**BPF_ANY**
3046  *			No condition on the existence of the entry for *key*.
3047  *
3048  *		If the *map* has eBPF programs (parser and verdict), those will
3049  *		be inherited by the socket being added. If the socket is
3050  *		already attached to eBPF programs, this results in an error.
3051  *	Return
3052  *		0 on success, or a negative error in case of failure.
3053  *
3054  * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
3055  *	Description
3056  *		This helper is used in programs implementing policies at the
3057  *		socket level. If the message *msg* is allowed to pass (i.e. if
3058  *		the verdict eBPF program returns **SK_PASS**), redirect it to
3059  *		the socket referenced by *map* (of type
3060  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3061  *		egress interfaces can be used for redirection. The
3062  *		**BPF_F_INGRESS** value in *flags* is used to make the
3063  *		distinction (ingress path is selected if the flag is present,
3064  *		egress path otherwise). This is the only flag supported for now.
3065  *	Return
3066  *		**SK_PASS** on success, or **SK_DROP** on error.
3067  *
3068  * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
3069  *	Description
3070  *		This helper is used in programs implementing policies at the
3071  *		skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
3072  *		if the verdict eBPF program returns **SK_PASS**), redirect it
3073  *		to the socket referenced by *map* (of type
3074  *		**BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
3075  *		egress interfaces can be used for redirection. The
3076  *		**BPF_F_INGRESS** value in *flags* is used to make the
3077  *		distinction (ingress path is selected if the flag is present,
3078  *		egress otherwise). This is the only flag supported for now.
3079  *	Return
3080  *		**SK_PASS** on success, or **SK_DROP** on error.
3081  *
3082  * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
3083  *	Description
3084  *		Encapsulate the packet associated to *skb* within a Layer 3
3085  *		protocol header. This header is provided in the buffer at
3086  *		address *hdr*, with *len* its size in bytes. *type* indicates
3087  *		the protocol of the header and can be one of:
3088  *
3089  *		**BPF_LWT_ENCAP_SEG6**
3090  *			IPv6 encapsulation with Segment Routing Header
3091  *			(**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
3092  *			the IPv6 header is computed by the kernel.
3093  *		**BPF_LWT_ENCAP_SEG6_INLINE**
3094  *			Only works if *skb* contains an IPv6 packet. Insert a
3095  *			Segment Routing Header (**struct ipv6_sr_hdr**) inside
3096  *			the IPv6 header.
3097  *		**BPF_LWT_ENCAP_IP**
3098  *			IP encapsulation (GRE/GUE/IPIP/etc). The outer header
3099  *			must be IPv4 or IPv6, followed by zero or more
3100  *			additional headers, up to **LWT_BPF_MAX_HEADROOM**
3101  *			total bytes in all prepended headers. Please note that
3102  *			if **skb_is_gso**\ (*skb*) is true, no more than two
3103  *			headers can be prepended, and the inner header, if
3104  *			present, should be either GRE or UDP/GUE.
3105  *
3106  *		**BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs
3107  *		of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can
3108  *		be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and
3109  *		**BPF_PROG_TYPE_LWT_XMIT**.
3110  *
3111  * 		A call to this helper is susceptible to change the underlying
3112  * 		packet buffer. Therefore, at load time, all checks on pointers
3113  * 		previously done by the verifier are invalidated and must be
3114  * 		performed again, if the helper is used in combination with
3115  * 		direct packet access.
3116  *	Return
3117  * 		0 on success, or a negative error in case of failure.
3118  *
3119  * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
3120  *	Description
3121  *		Store *len* bytes from address *from* into the packet
3122  *		associated to *skb*, at *offset*. Only the flags, tag and TLVs
3123  *		inside the outermost IPv6 Segment Routing Header can be
3124  *		modified through this helper.
3125  *
3126  * 		A call to this helper is susceptible to change the underlying
3127  * 		packet buffer. Therefore, at load time, all checks on pointers
3128  * 		previously done by the verifier are invalidated and must be
3129  * 		performed again, if the helper is used in combination with
3130  * 		direct packet access.
3131  *	Return
3132  * 		0 on success, or a negative error in case of failure.
3133  *
3134  * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
3135  *	Description
3136  *		Adjust the size allocated to TLVs in the outermost IPv6
3137  *		Segment Routing Header contained in the packet associated to
3138  *		*skb*, at position *offset* by *delta* bytes. Only offsets
3139  *		after the segments are accepted. *delta* can be as well
3140  *		positive (growing) as negative (shrinking).
3141  *
3142  * 		A call to this helper is susceptible to change the underlying
3143  * 		packet buffer. Therefore, at load time, all checks on pointers
3144  * 		previously done by the verifier are invalidated and must be
3145  * 		performed again, if the helper is used in combination with
3146  * 		direct packet access.
3147  *	Return
3148  * 		0 on success, or a negative error in case of failure.
3149  *
3150  * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
3151  *	Description
3152  *		Apply an IPv6 Segment Routing action of type *action* to the
3153  *		packet associated to *skb*. Each action takes a parameter
3154  *		contained at address *param*, and of length *param_len* bytes.
3155  *		*action* can be one of:
3156  *
3157  *		**SEG6_LOCAL_ACTION_END_X**
3158  *			End.X action: Endpoint with Layer-3 cross-connect.
3159  *			Type of *param*: **struct in6_addr**.
3160  *		**SEG6_LOCAL_ACTION_END_T**
3161  *			End.T action: Endpoint with specific IPv6 table lookup.
3162  *			Type of *param*: **int**.
3163  *		**SEG6_LOCAL_ACTION_END_B6**
3164  *			End.B6 action: Endpoint bound to an SRv6 policy.
3165  *			Type of *param*: **struct ipv6_sr_hdr**.
3166  *		**SEG6_LOCAL_ACTION_END_B6_ENCAP**
3167  *			End.B6.Encap action: Endpoint bound to an SRv6
3168  *			encapsulation policy.
3169  *			Type of *param*: **struct ipv6_sr_hdr**.
3170  *
3171  * 		A call to this helper is susceptible to change the underlying
3172  * 		packet buffer. Therefore, at load time, all checks on pointers
3173  * 		previously done by the verifier are invalidated and must be
3174  * 		performed again, if the helper is used in combination with
3175  * 		direct packet access.
3176  *	Return
3177  * 		0 on success, or a negative error in case of failure.
3178  *
3179  * long bpf_rc_repeat(void *ctx)
3180  *	Description
3181  *		This helper is used in programs implementing IR decoding, to
3182  *		report a successfully decoded repeat key message. This delays
3183  *		the generation of a key up event for previously generated
3184  *		key down event.
3185  *
3186  *		Some IR protocols like NEC have a special IR message for
3187  *		repeating last button, for when a button is held down.
3188  *
3189  *		The *ctx* should point to the lirc sample as passed into
3190  *		the program.
3191  *
3192  *		This helper is only available is the kernel was compiled with
3193  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3194  *		"**y**".
3195  *	Return
3196  *		0
3197  *
3198  * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
3199  *	Description
3200  *		This helper is used in programs implementing IR decoding, to
3201  *		report a successfully decoded key press with *scancode*,
3202  *		*toggle* value in the given *protocol*. The scancode will be
3203  *		translated to a keycode using the rc keymap, and reported as
3204  *		an input key down event. After a period a key up event is
3205  *		generated. This period can be extended by calling either
3206  *		**bpf_rc_keydown**\ () again with the same values, or calling
3207  *		**bpf_rc_repeat**\ ().
3208  *
3209  *		Some protocols include a toggle bit, in case the button was
3210  *		released and pressed again between consecutive scancodes.
3211  *
3212  *		The *ctx* should point to the lirc sample as passed into
3213  *		the program.
3214  *
3215  *		The *protocol* is the decoded protocol number (see
3216  *		**enum rc_proto** for some predefined values).
3217  *
3218  *		This helper is only available is the kernel was compiled with
3219  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3220  *		"**y**".
3221  *	Return
3222  *		0
3223  *
3224  * u64 bpf_skb_cgroup_id(struct sk_buff *skb)
3225  * 	Description
3226  * 		Return the cgroup v2 id of the socket associated with the *skb*.
3227  * 		This is roughly similar to the **bpf_get_cgroup_classid**\ ()
3228  * 		helper for cgroup v1 by providing a tag resp. identifier that
3229  * 		can be matched on or used for map lookups e.g. to implement
3230  * 		policy. The cgroup v2 id of a given path in the hierarchy is
3231  * 		exposed in user space through the f_handle API in order to get
3232  * 		to the same 64-bit id.
3233  *
3234  * 		This helper can be used on TC egress path, but not on ingress,
3235  * 		and is available only if the kernel was compiled with the
3236  * 		**CONFIG_SOCK_CGROUP_DATA** configuration option.
3237  * 	Return
3238  * 		The id is returned or 0 in case the id could not be retrieved.
3239  *
3240  * u64 bpf_get_current_cgroup_id(void)
3241  * 	Return
3242  * 		A 64-bit integer containing the current cgroup id based
3243  * 		on the cgroup within which the current task is running.
3244  *
3245  * void *bpf_get_local_storage(void *map, u64 flags)
3246  *	Description
3247  *		Get the pointer to the local storage area.
3248  *		The type and the size of the local storage is defined
3249  *		by the *map* argument.
3250  *		The *flags* meaning is specific for each map type,
3251  *		and has to be 0 for cgroup local storage.
3252  *
3253  *		Depending on the BPF program type, a local storage area
3254  *		can be shared between multiple instances of the BPF program,
3255  *		running simultaneously.
3256  *
3257  *		A user should care about the synchronization by himself.
3258  *		For example, by using the **BPF_ATOMIC** instructions to alter
3259  *		the shared data.
3260  *	Return
3261  *		A pointer to the local storage area.
3262  *
3263  * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
3264  *	Description
3265  *		Select a **SO_REUSEPORT** socket from a
3266  *		**BPF_MAP_TYPE_REUSEPORT_SOCKARRAY** *map*.
3267  *		It checks the selected socket is matching the incoming
3268  *		request in the socket buffer.
3269  *	Return
3270  *		0 on success, or a negative error in case of failure.
3271  *
3272  * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
3273  *	Description
3274  *		Return id of cgroup v2 that is ancestor of cgroup associated
3275  *		with the *skb* at the *ancestor_level*.  The root cgroup is at
3276  *		*ancestor_level* zero and each step down the hierarchy
3277  *		increments the level. If *ancestor_level* == level of cgroup
3278  *		associated with *skb*, then return value will be same as that
3279  *		of **bpf_skb_cgroup_id**\ ().
3280  *
3281  *		The helper is useful to implement policies based on cgroups
3282  *		that are upper in hierarchy than immediate cgroup associated
3283  *		with *skb*.
3284  *
3285  *		The format of returned id and helper limitations are same as in
3286  *		**bpf_skb_cgroup_id**\ ().
3287  *	Return
3288  *		The id is returned or 0 in case the id could not be retrieved.
3289  *
3290  * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3291  *	Description
3292  *		Look for TCP socket matching *tuple*, optionally in a child
3293  *		network namespace *netns*. The return value must be checked,
3294  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3295  *
3296  *		The *ctx* should point to the context of the program, such as
3297  *		the skb or socket (depending on the hook in use). This is used
3298  *		to determine the base network namespace for the lookup.
3299  *
3300  *		*tuple_size* must be one of:
3301  *
3302  *		**sizeof**\ (*tuple*\ **->ipv4**)
3303  *			Look for an IPv4 socket.
3304  *		**sizeof**\ (*tuple*\ **->ipv6**)
3305  *			Look for an IPv6 socket.
3306  *
3307  *		If the *netns* is a negative signed 32-bit integer, then the
3308  *		socket lookup table in the netns associated with the *ctx*
3309  *		will be used. For the TC hooks, this is the netns of the device
3310  *		in the skb. For socket hooks, this is the netns of the socket.
3311  *		If *netns* is any other signed 32-bit value greater than or
3312  *		equal to zero then it specifies the ID of the netns relative to
3313  *		the netns associated with the *ctx*. *netns* values beyond the
3314  *		range of 32-bit integers are reserved for future use.
3315  *
3316  *		All values for *flags* are reserved for future usage, and must
3317  *		be left at zero.
3318  *
3319  *		This helper is available only if the kernel was compiled with
3320  *		**CONFIG_NET** configuration option.
3321  *	Return
3322  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3323  *		For sockets with reuseport option, the **struct bpf_sock**
3324  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3325  *		tuple.
3326  *
3327  * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3328  *	Description
3329  *		Look for UDP socket matching *tuple*, optionally in a child
3330  *		network namespace *netns*. The return value must be checked,
3331  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3332  *
3333  *		The *ctx* should point to the context of the program, such as
3334  *		the skb or socket (depending on the hook in use). This is used
3335  *		to determine the base network namespace for the lookup.
3336  *
3337  *		*tuple_size* must be one of:
3338  *
3339  *		**sizeof**\ (*tuple*\ **->ipv4**)
3340  *			Look for an IPv4 socket.
3341  *		**sizeof**\ (*tuple*\ **->ipv6**)
3342  *			Look for an IPv6 socket.
3343  *
3344  *		If the *netns* is a negative signed 32-bit integer, then the
3345  *		socket lookup table in the netns associated with the *ctx*
3346  *		will be used. For the TC hooks, this is the netns of the device
3347  *		in the skb. For socket hooks, this is the netns of the socket.
3348  *		If *netns* is any other signed 32-bit value greater than or
3349  *		equal to zero then it specifies the ID of the netns relative to
3350  *		the netns associated with the *ctx*. *netns* values beyond the
3351  *		range of 32-bit integers are reserved for future use.
3352  *
3353  *		All values for *flags* are reserved for future usage, and must
3354  *		be left at zero.
3355  *
3356  *		This helper is available only if the kernel was compiled with
3357  *		**CONFIG_NET** configuration option.
3358  *	Return
3359  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3360  *		For sockets with reuseport option, the **struct bpf_sock**
3361  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3362  *		tuple.
3363  *
3364  * long bpf_sk_release(void *sock)
3365  *	Description
3366  *		Release the reference held by *sock*. *sock* must be a
3367  *		non-**NULL** pointer that was returned from
3368  *		**bpf_sk_lookup_xxx**\ ().
3369  *	Return
3370  *		0 on success, or a negative error in case of failure.
3371  *
3372  * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
3373  * 	Description
3374  * 		Push an element *value* in *map*. *flags* is one of:
3375  *
3376  * 		**BPF_EXIST**
3377  * 			If the queue/stack is full, the oldest element is
3378  * 			removed to make room for this.
3379  * 	Return
3380  * 		0 on success, or a negative error in case of failure.
3381  *
3382  * long bpf_map_pop_elem(struct bpf_map *map, void *value)
3383  * 	Description
3384  * 		Pop an element from *map*.
3385  * 	Return
3386  * 		0 on success, or a negative error in case of failure.
3387  *
3388  * long bpf_map_peek_elem(struct bpf_map *map, void *value)
3389  * 	Description
3390  * 		Get an element from *map* without removing it.
3391  * 	Return
3392  * 		0 on success, or a negative error in case of failure.
3393  *
3394  * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3395  *	Description
3396  *		For socket policies, insert *len* bytes into *msg* at offset
3397  *		*start*.
3398  *
3399  *		If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
3400  *		*msg* it may want to insert metadata or options into the *msg*.
3401  *		This can later be read and used by any of the lower layer BPF
3402  *		hooks.
3403  *
3404  *		This helper may fail if under memory pressure (a malloc
3405  *		fails) in these cases BPF programs will get an appropriate
3406  *		error and BPF programs will need to handle them.
3407  *	Return
3408  *		0 on success, or a negative error in case of failure.
3409  *
3410  * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags)
3411  *	Description
3412  *		Will remove *len* bytes from a *msg* starting at byte *start*.
3413  *		This may result in **ENOMEM** errors under certain situations if
3414  *		an allocation and copy are required due to a full ring buffer.
3415  *		However, the helper will try to avoid doing the allocation
3416  *		if possible. Other errors can occur if input parameters are
3417  *		invalid either due to *start* byte not being valid part of *msg*
3418  *		payload and/or *pop* value being to large.
3419  *	Return
3420  *		0 on success, or a negative error in case of failure.
3421  *
3422  * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y)
3423  *	Description
3424  *		This helper is used in programs implementing IR decoding, to
3425  *		report a successfully decoded pointer movement.
3426  *
3427  *		The *ctx* should point to the lirc sample as passed into
3428  *		the program.
3429  *
3430  *		This helper is only available is the kernel was compiled with
3431  *		the **CONFIG_BPF_LIRC_MODE2** configuration option set to
3432  *		"**y**".
3433  *	Return
3434  *		0
3435  *
3436  * long bpf_spin_lock(struct bpf_spin_lock *lock)
3437  *	Description
3438  *		Acquire a spinlock represented by the pointer *lock*, which is
3439  *		stored as part of a value of a map. Taking the lock allows to
3440  *		safely update the rest of the fields in that value. The
3441  *		spinlock can (and must) later be released with a call to
3442  *		**bpf_spin_unlock**\ (\ *lock*\ ).
3443  *
3444  *		Spinlocks in BPF programs come with a number of restrictions
3445  *		and constraints:
3446  *
3447  *		* **bpf_spin_lock** objects are only allowed inside maps of
3448  *		  types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this
3449  *		  list could be extended in the future).
3450  *		* BTF description of the map is mandatory.
3451  *		* The BPF program can take ONE lock at a time, since taking two
3452  *		  or more could cause dead locks.
3453  *		* Only one **struct bpf_spin_lock** is allowed per map element.
3454  *		* When the lock is taken, calls (either BPF to BPF or helpers)
3455  *		  are not allowed.
3456  *		* The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not
3457  *		  allowed inside a spinlock-ed region.
3458  *		* The BPF program MUST call **bpf_spin_unlock**\ () to release
3459  *		  the lock, on all execution paths, before it returns.
3460  *		* The BPF program can access **struct bpf_spin_lock** only via
3461  *		  the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ ()
3462  *		  helpers. Loading or storing data into the **struct
3463  *		  bpf_spin_lock** *lock*\ **;** field of a map is not allowed.
3464  *		* To use the **bpf_spin_lock**\ () helper, the BTF description
3465  *		  of the map value must be a struct and have **struct
3466  *		  bpf_spin_lock** *anyname*\ **;** field at the top level.
3467  *		  Nested lock inside another struct is not allowed.
3468  *		* The **struct bpf_spin_lock** *lock* field in a map value must
3469  *		  be aligned on a multiple of 4 bytes in that value.
3470  *		* Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy
3471  *		  the **bpf_spin_lock** field to user space.
3472  *		* Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from
3473  *		  a BPF program, do not update the **bpf_spin_lock** field.
3474  *		* **bpf_spin_lock** cannot be on the stack or inside a
3475  *		  networking packet (it can only be inside of a map values).
3476  *		* **bpf_spin_lock** is available to root only.
3477  *		* Tracing programs and socket filter programs cannot use
3478  *		  **bpf_spin_lock**\ () due to insufficient preemption checks
3479  *		  (but this may change in the future).
3480  *		* **bpf_spin_lock** is not allowed in inner maps of map-in-map.
3481  *	Return
3482  *		0
3483  *
3484  * long bpf_spin_unlock(struct bpf_spin_lock *lock)
3485  *	Description
3486  *		Release the *lock* previously locked by a call to
3487  *		**bpf_spin_lock**\ (\ *lock*\ ).
3488  *	Return
3489  *		0
3490  *
3491  * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk)
3492  *	Description
3493  *		This helper gets a **struct bpf_sock** pointer such
3494  *		that all the fields in this **bpf_sock** can be accessed.
3495  *	Return
3496  *		A **struct bpf_sock** pointer on success, or **NULL** in
3497  *		case of failure.
3498  *
3499  * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
3500  *	Description
3501  *		This helper gets a **struct bpf_tcp_sock** pointer from a
3502  *		**struct bpf_sock** pointer.
3503  *	Return
3504  *		A **struct bpf_tcp_sock** pointer on success, or **NULL** in
3505  *		case of failure.
3506  *
3507  * long bpf_skb_ecn_set_ce(struct sk_buff *skb)
3508  *	Description
3509  *		Set ECN (Explicit Congestion Notification) field of IP header
3510  *		to **CE** (Congestion Encountered) if current value is **ECT**
3511  *		(ECN Capable Transport). Otherwise, do nothing. Works with IPv6
3512  *		and IPv4.
3513  *	Return
3514  *		1 if the **CE** flag is set (either by the current helper call
3515  *		or because it was already present), 0 if it is not set.
3516  *
3517  * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk)
3518  *	Description
3519  *		Return a **struct bpf_sock** pointer in **TCP_LISTEN** state.
3520  *		**bpf_sk_release**\ () is unnecessary and not allowed.
3521  *	Return
3522  *		A **struct bpf_sock** pointer on success, or **NULL** in
3523  *		case of failure.
3524  *
3525  * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
3526  *	Description
3527  *		Look for TCP socket matching *tuple*, optionally in a child
3528  *		network namespace *netns*. The return value must be checked,
3529  *		and if non-**NULL**, released via **bpf_sk_release**\ ().
3530  *
3531  *		This function is identical to **bpf_sk_lookup_tcp**\ (), except
3532  *		that it also returns timewait or request sockets. Use
3533  *		**bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the
3534  *		full structure.
3535  *
3536  *		This helper is available only if the kernel was compiled with
3537  *		**CONFIG_NET** configuration option.
3538  *	Return
3539  *		Pointer to **struct bpf_sock**, or **NULL** in case of failure.
3540  *		For sockets with reuseport option, the **struct bpf_sock**
3541  *		result is from *reuse*\ **->socks**\ [] using the hash of the
3542  *		tuple.
3543  *
3544  * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3545  * 	Description
3546  * 		Check whether *iph* and *th* contain a valid SYN cookie ACK for
3547  * 		the listening socket in *sk*.
3548  *
3549  * 		*iph* points to the start of the IPv4 or IPv6 header, while
3550  * 		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3551  * 		**sizeof**\ (**struct ip6hdr**).
3552  *
3553  * 		*th* points to the start of the TCP header, while *th_len*
3554  * 		contains **sizeof**\ (**struct tcphdr**).
3555  * 	Return
3556  * 		0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
3557  * 		error otherwise.
3558  *
3559  * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
3560  *	Description
3561  *		Get name of sysctl in /proc/sys/ and copy it into provided by
3562  *		program buffer *buf* of size *buf_len*.
3563  *
3564  *		The buffer is always NUL terminated, unless it's zero-sized.
3565  *
3566  *		If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is
3567  *		copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name
3568  *		only (e.g. "tcp_mem").
3569  *	Return
3570  *		Number of character copied (not including the trailing NUL).
3571  *
3572  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3573  *		truncated name in this case).
3574  *
3575  * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3576  *	Description
3577  *		Get current value of sysctl as it is presented in /proc/sys
3578  *		(incl. newline, etc), and copy it as a string into provided
3579  *		by program buffer *buf* of size *buf_len*.
3580  *
3581  *		The whole value is copied, no matter what file position user
3582  *		space issued e.g. sys_read at.
3583  *
3584  *		The buffer is always NUL terminated, unless it's zero-sized.
3585  *	Return
3586  *		Number of character copied (not including the trailing NUL).
3587  *
3588  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3589  *		truncated name in this case).
3590  *
3591  *		**-EINVAL** if current value was unavailable, e.g. because
3592  *		sysctl is uninitialized and read returns -EIO for it.
3593  *
3594  * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len)
3595  *	Description
3596  *		Get new value being written by user space to sysctl (before
3597  *		the actual write happens) and copy it as a string into
3598  *		provided by program buffer *buf* of size *buf_len*.
3599  *
3600  *		User space may write new value at file position > 0.
3601  *
3602  *		The buffer is always NUL terminated, unless it's zero-sized.
3603  *	Return
3604  *		Number of character copied (not including the trailing NUL).
3605  *
3606  *		**-E2BIG** if the buffer wasn't big enough (*buf* will contain
3607  *		truncated name in this case).
3608  *
3609  *		**-EINVAL** if sysctl is being read.
3610  *
3611  * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len)
3612  *	Description
3613  *		Override new value being written by user space to sysctl with
3614  *		value provided by program in buffer *buf* of size *buf_len*.
3615  *
3616  *		*buf* should contain a string in same form as provided by user
3617  *		space on sysctl write.
3618  *
3619  *		User space may write new value at file position > 0. To override
3620  *		the whole sysctl value file position should be set to zero.
3621  *	Return
3622  *		0 on success.
3623  *
3624  *		**-E2BIG** if the *buf_len* is too big.
3625  *
3626  *		**-EINVAL** if sysctl is being read.
3627  *
3628  * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res)
3629  *	Description
3630  *		Convert the initial part of the string from buffer *buf* of
3631  *		size *buf_len* to a long integer according to the given base
3632  *		and save the result in *res*.
3633  *
3634  *		The string may begin with an arbitrary amount of white space
3635  *		(as determined by **isspace**\ (3)) followed by a single
3636  *		optional '**-**' sign.
3637  *
3638  *		Five least significant bits of *flags* encode base, other bits
3639  *		are currently unused.
3640  *
3641  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3642  *		similar to user space **strtol**\ (3).
3643  *	Return
3644  *		Number of characters consumed on success. Must be positive but
3645  *		no more than *buf_len*.
3646  *
3647  *		**-EINVAL** if no valid digits were found or unsupported base
3648  *		was provided.
3649  *
3650  *		**-ERANGE** if resulting value was out of range.
3651  *
3652  * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res)
3653  *	Description
3654  *		Convert the initial part of the string from buffer *buf* of
3655  *		size *buf_len* to an unsigned long integer according to the
3656  *		given base and save the result in *res*.
3657  *
3658  *		The string may begin with an arbitrary amount of white space
3659  *		(as determined by **isspace**\ (3)).
3660  *
3661  *		Five least significant bits of *flags* encode base, other bits
3662  *		are currently unused.
3663  *
3664  *		Base must be either 8, 10, 16 or 0 to detect it automatically
3665  *		similar to user space **strtoul**\ (3).
3666  *	Return
3667  *		Number of characters consumed on success. Must be positive but
3668  *		no more than *buf_len*.
3669  *
3670  *		**-EINVAL** if no valid digits were found or unsupported base
3671  *		was provided.
3672  *
3673  *		**-ERANGE** if resulting value was out of range.
3674  *
3675  * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags)
3676  *	Description
3677  *		Get a bpf-local-storage from a *sk*.
3678  *
3679  *		Logically, it could be thought of getting the value from
3680  *		a *map* with *sk* as the **key**.  From this
3681  *		perspective,  the usage is not much different from
3682  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this
3683  *		helper enforces the key must be a full socket and the map must
3684  *		be a **BPF_MAP_TYPE_SK_STORAGE** also.
3685  *
3686  *		Underneath, the value is stored locally at *sk* instead of
3687  *		the *map*.  The *map* is used as the bpf-local-storage
3688  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
3689  *		searched against all bpf-local-storages residing at *sk*.
3690  *
3691  *		*sk* is a kernel **struct sock** pointer for LSM program.
3692  *		*sk* is a **struct bpf_sock** pointer for other program types.
3693  *
3694  *		An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be
3695  *		used such that a new bpf-local-storage will be
3696  *		created if one does not exist.  *value* can be used
3697  *		together with **BPF_SK_STORAGE_GET_F_CREATE** to specify
3698  *		the initial value of a bpf-local-storage.  If *value* is
3699  *		**NULL**, the new bpf-local-storage will be zero initialized.
3700  *	Return
3701  *		A bpf-local-storage pointer is returned on success.
3702  *
3703  *		**NULL** if not found or there was an error in adding
3704  *		a new bpf-local-storage.
3705  *
3706  * long bpf_sk_storage_delete(struct bpf_map *map, void *sk)
3707  *	Description
3708  *		Delete a bpf-local-storage from a *sk*.
3709  *	Return
3710  *		0 on success.
3711  *
3712  *		**-ENOENT** if the bpf-local-storage cannot be found.
3713  *		**-EINVAL** if sk is not a fullsock (e.g. a request_sock).
3714  *
3715  * long bpf_send_signal(u32 sig)
3716  *	Description
3717  *		Send signal *sig* to the process of the current task.
3718  *		The signal may be delivered to any of this process's threads.
3719  *	Return
3720  *		0 on success or successfully queued.
3721  *
3722  *		**-EBUSY** if work queue under nmi is full.
3723  *
3724  *		**-EINVAL** if *sig* is invalid.
3725  *
3726  *		**-EPERM** if no permission to send the *sig*.
3727  *
3728  *		**-EAGAIN** if bpf program can try again.
3729  *
3730  * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len)
3731  *	Description
3732  *		Try to issue a SYN cookie for the packet with corresponding
3733  *		IP/TCP headers, *iph* and *th*, on the listening socket in *sk*.
3734  *
3735  *		*iph* points to the start of the IPv4 or IPv6 header, while
3736  *		*iph_len* contains **sizeof**\ (**struct iphdr**) or
3737  *		**sizeof**\ (**struct ip6hdr**).
3738  *
3739  *		*th* points to the start of the TCP header, while *th_len*
3740  *		contains the length of the TCP header.
3741  *	Return
3742  *		On success, lower 32 bits hold the generated SYN cookie in
3743  *		followed by 16 bits which hold the MSS value for that cookie,
3744  *		and the top 16 bits are unused.
3745  *
3746  *		On failure, the returned value is one of the following:
3747  *
3748  *		**-EINVAL** SYN cookie cannot be issued due to error
3749  *
3750  *		**-ENOENT** SYN cookie should not be issued (no SYN flood)
3751  *
3752  *		**-EOPNOTSUPP** kernel configuration does not enable SYN cookies
3753  *
3754  *		**-EPROTONOSUPPORT** IP packet version is not 4 or 6
3755  *
3756  * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3757  * 	Description
3758  * 		Write raw *data* blob into a special BPF perf event held by
3759  * 		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3760  * 		event must have the following attributes: **PERF_SAMPLE_RAW**
3761  * 		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3762  * 		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3763  *
3764  * 		The *flags* are used to indicate the index in *map* for which
3765  * 		the value must be put, masked with **BPF_F_INDEX_MASK**.
3766  * 		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3767  * 		to indicate that the index of the current CPU core should be
3768  * 		used.
3769  *
3770  * 		The value to write, of *size*, is passed through eBPF stack and
3771  * 		pointed by *data*.
3772  *
3773  * 		*ctx* is a pointer to in-kernel struct sk_buff.
3774  *
3775  * 		This helper is similar to **bpf_perf_event_output**\ () but
3776  * 		restricted to raw_tracepoint bpf programs.
3777  * 	Return
3778  * 		0 on success, or a negative error in case of failure.
3779  *
3780  * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr)
3781  * 	Description
3782  * 		Safely attempt to read *size* bytes from user space address
3783  * 		*unsafe_ptr* and store the data in *dst*.
3784  * 	Return
3785  * 		0 on success, or a negative error in case of failure.
3786  *
3787  * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr)
3788  * 	Description
3789  * 		Safely attempt to read *size* bytes from kernel space address
3790  * 		*unsafe_ptr* and store the data in *dst*.
3791  * 	Return
3792  * 		0 on success, or a negative error in case of failure.
3793  *
3794  * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr)
3795  * 	Description
3796  * 		Copy a NUL terminated string from an unsafe user address
3797  * 		*unsafe_ptr* to *dst*. The *size* should include the
3798  * 		terminating NUL byte. In case the string length is smaller than
3799  * 		*size*, the target is not padded with further NUL bytes. If the
3800  * 		string length is larger than *size*, just *size*-1 bytes are
3801  * 		copied and the last byte is set to NUL.
3802  *
3803  * 		On success, returns the number of bytes that were written,
3804  * 		including the terminal NUL. This makes this helper useful in
3805  * 		tracing programs for reading strings, and more importantly to
3806  * 		get its length at runtime. See the following snippet:
3807  *
3808  * 		::
3809  *
3810  * 			SEC("kprobe/sys_open")
3811  * 			void bpf_sys_open(struct pt_regs *ctx)
3812  * 			{
3813  * 			        char buf[PATHLEN]; // PATHLEN is defined to 256
3814  * 			        int res = bpf_probe_read_user_str(buf, sizeof(buf),
3815  * 				                                  ctx->di);
3816  *
3817  * 				// Consume buf, for example push it to
3818  * 				// userspace via bpf_perf_event_output(); we
3819  * 				// can use res (the string length) as event
3820  * 				// size, after checking its boundaries.
3821  * 			}
3822  *
3823  * 		In comparison, using **bpf_probe_read_user**\ () helper here
3824  * 		instead to read the string would require to estimate the length
3825  * 		at compile time, and would often result in copying more memory
3826  * 		than necessary.
3827  *
3828  * 		Another useful use case is when parsing individual process
3829  * 		arguments or individual environment variables navigating
3830  * 		*current*\ **->mm->arg_start** and *current*\
3831  * 		**->mm->env_start**: using this helper and the return value,
3832  * 		one can quickly iterate at the right offset of the memory area.
3833  * 	Return
3834  * 		On success, the strictly positive length of the output string,
3835  * 		including the trailing NUL character. On error, a negative
3836  * 		value.
3837  *
3838  * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr)
3839  * 	Description
3840  * 		Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr*
3841  * 		to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply.
3842  * 	Return
3843  * 		On success, the strictly positive length of the string, including
3844  * 		the trailing NUL character. On error, a negative value.
3845  *
3846  * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt)
3847  *	Description
3848  *		Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**.
3849  *		*rcv_nxt* is the ack_seq to be sent out.
3850  *	Return
3851  *		0 on success, or a negative error in case of failure.
3852  *
3853  * long bpf_send_signal_thread(u32 sig)
3854  *	Description
3855  *		Send signal *sig* to the thread corresponding to the current task.
3856  *	Return
3857  *		0 on success or successfully queued.
3858  *
3859  *		**-EBUSY** if work queue under nmi is full.
3860  *
3861  *		**-EINVAL** if *sig* is invalid.
3862  *
3863  *		**-EPERM** if no permission to send the *sig*.
3864  *
3865  *		**-EAGAIN** if bpf program can try again.
3866  *
3867  * u64 bpf_jiffies64(void)
3868  *	Description
3869  *		Obtain the 64bit jiffies
3870  *	Return
3871  *		The 64 bit jiffies
3872  *
3873  * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags)
3874  *	Description
3875  *		For an eBPF program attached to a perf event, retrieve the
3876  *		branch records (**struct perf_branch_entry**) associated to *ctx*
3877  *		and store it in the buffer pointed by *buf* up to size
3878  *		*size* bytes.
3879  *	Return
3880  *		On success, number of bytes written to *buf*. On error, a
3881  *		negative value.
3882  *
3883  *		The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to
3884  *		instead return the number of bytes required to store all the
3885  *		branch entries. If this flag is set, *buf* may be NULL.
3886  *
3887  *		**-EINVAL** if arguments invalid or **size** not a multiple
3888  *		of **sizeof**\ (**struct perf_branch_entry**\ ).
3889  *
3890  *		**-ENOENT** if architecture does not support branch records.
3891  *
3892  * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size)
3893  *	Description
3894  *		Returns 0 on success, values for *pid* and *tgid* as seen from the current
3895  *		*namespace* will be returned in *nsdata*.
3896  *	Return
3897  *		0 on success, or one of the following in case of failure:
3898  *
3899  *		**-EINVAL** if dev and inum supplied don't match dev_t and inode number
3900  *              with nsfs of current task, or if dev conversion to dev_t lost high bits.
3901  *
3902  *		**-ENOENT** if pidns does not exists for the current task.
3903  *
3904  * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
3905  *	Description
3906  *		Write raw *data* blob into a special BPF perf event held by
3907  *		*map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
3908  *		event must have the following attributes: **PERF_SAMPLE_RAW**
3909  *		as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
3910  *		**PERF_COUNT_SW_BPF_OUTPUT** as **config**.
3911  *
3912  *		The *flags* are used to indicate the index in *map* for which
3913  *		the value must be put, masked with **BPF_F_INDEX_MASK**.
3914  *		Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
3915  *		to indicate that the index of the current CPU core should be
3916  *		used.
3917  *
3918  *		The value to write, of *size*, is passed through eBPF stack and
3919  *		pointed by *data*.
3920  *
3921  *		*ctx* is a pointer to in-kernel struct xdp_buff.
3922  *
3923  *		This helper is similar to **bpf_perf_eventoutput**\ () but
3924  *		restricted to raw_tracepoint bpf programs.
3925  *	Return
3926  *		0 on success, or a negative error in case of failure.
3927  *
3928  * u64 bpf_get_netns_cookie(void *ctx)
3929  * 	Description
3930  * 		Retrieve the cookie (generated by the kernel) of the network
3931  * 		namespace the input *ctx* is associated with. The network
3932  * 		namespace cookie remains stable for its lifetime and provides
3933  * 		a global identifier that can be assumed unique. If *ctx* is
3934  * 		NULL, then the helper returns the cookie for the initial
3935  * 		network namespace. The cookie itself is very similar to that
3936  * 		of **bpf_get_socket_cookie**\ () helper, but for network
3937  * 		namespaces instead of sockets.
3938  * 	Return
3939  * 		A 8-byte long opaque number.
3940  *
3941  * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level)
3942  * 	Description
3943  * 		Return id of cgroup v2 that is ancestor of the cgroup associated
3944  * 		with the current task at the *ancestor_level*. The root cgroup
3945  * 		is at *ancestor_level* zero and each step down the hierarchy
3946  * 		increments the level. If *ancestor_level* == level of cgroup
3947  * 		associated with the current task, then return value will be the
3948  * 		same as that of **bpf_get_current_cgroup_id**\ ().
3949  *
3950  * 		The helper is useful to implement policies based on cgroups
3951  * 		that are upper in hierarchy than immediate cgroup associated
3952  * 		with the current task.
3953  *
3954  * 		The format of returned id and helper limitations are same as in
3955  * 		**bpf_get_current_cgroup_id**\ ().
3956  * 	Return
3957  * 		The id is returned or 0 in case the id could not be retrieved.
3958  *
3959  * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags)
3960  *	Description
3961  *		Helper is overloaded depending on BPF program type. This
3962  *		description applies to **BPF_PROG_TYPE_SCHED_CLS** and
3963  *		**BPF_PROG_TYPE_SCHED_ACT** programs.
3964  *
3965  *		Assign the *sk* to the *skb*. When combined with appropriate
3966  *		routing configuration to receive the packet towards the socket,
3967  *		will cause *skb* to be delivered to the specified socket.
3968  *		Subsequent redirection of *skb* via  **bpf_redirect**\ (),
3969  *		**bpf_clone_redirect**\ () or other methods outside of BPF may
3970  *		interfere with successful delivery to the socket.
3971  *
3972  *		This operation is only valid from TC ingress path.
3973  *
3974  *		The *flags* argument must be zero.
3975  *	Return
3976  *		0 on success, or a negative error in case of failure:
3977  *
3978  *		**-EINVAL** if specified *flags* are not supported.
3979  *
3980  *		**-ENOENT** if the socket is unavailable for assignment.
3981  *
3982  *		**-ENETUNREACH** if the socket is unreachable (wrong netns).
3983  *
3984  *		**-EOPNOTSUPP** if the operation is not supported, for example
3985  *		a call from outside of TC ingress.
3986  *
3987  *		**-ESOCKTNOSUPPORT** if the socket type is not supported
3988  *		(reuseport).
3989  *
3990  * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags)
3991  *	Description
3992  *		Helper is overloaded depending on BPF program type. This
3993  *		description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs.
3994  *
3995  *		Select the *sk* as a result of a socket lookup.
3996  *
3997  *		For the operation to succeed passed socket must be compatible
3998  *		with the packet description provided by the *ctx* object.
3999  *
4000  *		L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must
4001  *		be an exact match. While IP family (**AF_INET** or
4002  *		**AF_INET6**) must be compatible, that is IPv6 sockets
4003  *		that are not v6-only can be selected for IPv4 packets.
4004  *
4005  *		Only TCP listeners and UDP unconnected sockets can be
4006  *		selected. *sk* can also be NULL to reset any previous
4007  *		selection.
4008  *
4009  *		*flags* argument can combination of following values:
4010  *
4011  *		* **BPF_SK_LOOKUP_F_REPLACE** to override the previous
4012  *		  socket selection, potentially done by a BPF program
4013  *		  that ran before us.
4014  *
4015  *		* **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip
4016  *		  load-balancing within reuseport group for the socket
4017  *		  being selected.
4018  *
4019  *		On success *ctx->sk* will point to the selected socket.
4020  *
4021  *	Return
4022  *		0 on success, or a negative errno in case of failure.
4023  *
4024  *		* **-EAFNOSUPPORT** if socket family (*sk->family*) is
4025  *		  not compatible with packet family (*ctx->family*).
4026  *
4027  *		* **-EEXIST** if socket has been already selected,
4028  *		  potentially by another program, and
4029  *		  **BPF_SK_LOOKUP_F_REPLACE** flag was not specified.
4030  *
4031  *		* **-EINVAL** if unsupported flags were specified.
4032  *
4033  *		* **-EPROTOTYPE** if socket L4 protocol
4034  *		  (*sk->protocol*) doesn't match packet protocol
4035  *		  (*ctx->protocol*).
4036  *
4037  *		* **-ESOCKTNOSUPPORT** if socket is not in allowed
4038  *		  state (TCP listening or UDP unconnected).
4039  *
4040  * u64 bpf_ktime_get_boot_ns(void)
4041  * 	Description
4042  * 		Return the time elapsed since system boot, in nanoseconds.
4043  * 		Does include the time the system was suspended.
4044  * 		See: **clock_gettime**\ (**CLOCK_BOOTTIME**)
4045  * 	Return
4046  * 		Current *ktime*.
4047  *
4048  * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4049  * 	Description
4050  * 		**bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print
4051  * 		out the format string.
4052  * 		The *m* represents the seq_file. The *fmt* and *fmt_size* are for
4053  * 		the format string itself. The *data* and *data_len* are format string
4054  * 		arguments. The *data* are a **u64** array and corresponding format string
4055  * 		values are stored in the array. For strings and pointers where pointees
4056  * 		are accessed, only the pointer values are stored in the *data* array.
4057  * 		The *data_len* is the size of *data* in bytes - must be a multiple of 8.
4058  *
4059  *		Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory.
4060  *		Reading kernel memory may fail due to either invalid address or
4061  *		valid address but requiring a major memory fault. If reading kernel memory
4062  *		fails, the string for **%s** will be an empty string, and the ip
4063  *		address for **%p{i,I}{4,6}** will be 0. Not returning error to
4064  *		bpf program is consistent with what **bpf_trace_printk**\ () does for now.
4065  * 	Return
4066  * 		0 on success, or a negative error in case of failure:
4067  *
4068  *		**-EBUSY** if per-CPU memory copy buffer is busy, can try again
4069  *		by returning 1 from bpf program.
4070  *
4071  *		**-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported.
4072  *
4073  *		**-E2BIG** if *fmt* contains too many format specifiers.
4074  *
4075  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4076  *
4077  * long bpf_seq_write(struct seq_file *m, const void *data, u32 len)
4078  * 	Description
4079  * 		**bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data.
4080  * 		The *m* represents the seq_file. The *data* and *len* represent the
4081  * 		data to write in bytes.
4082  * 	Return
4083  * 		0 on success, or a negative error in case of failure:
4084  *
4085  *		**-EOVERFLOW** if an overflow happened: The same object will be tried again.
4086  *
4087  * u64 bpf_sk_cgroup_id(void *sk)
4088  *	Description
4089  *		Return the cgroup v2 id of the socket *sk*.
4090  *
4091  *		*sk* must be a non-**NULL** pointer to a socket, e.g. one
4092  *		returned from **bpf_sk_lookup_xxx**\ (),
4093  *		**bpf_sk_fullsock**\ (), etc. The format of returned id is
4094  *		same as in **bpf_skb_cgroup_id**\ ().
4095  *
4096  *		This helper is available only if the kernel was compiled with
4097  *		the **CONFIG_SOCK_CGROUP_DATA** configuration option.
4098  *	Return
4099  *		The id is returned or 0 in case the id could not be retrieved.
4100  *
4101  * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level)
4102  *	Description
4103  *		Return id of cgroup v2 that is ancestor of cgroup associated
4104  *		with the *sk* at the *ancestor_level*.  The root cgroup is at
4105  *		*ancestor_level* zero and each step down the hierarchy
4106  *		increments the level. If *ancestor_level* == level of cgroup
4107  *		associated with *sk*, then return value will be same as that
4108  *		of **bpf_sk_cgroup_id**\ ().
4109  *
4110  *		The helper is useful to implement policies based on cgroups
4111  *		that are upper in hierarchy than immediate cgroup associated
4112  *		with *sk*.
4113  *
4114  *		The format of returned id and helper limitations are same as in
4115  *		**bpf_sk_cgroup_id**\ ().
4116  *	Return
4117  *		The id is returned or 0 in case the id could not be retrieved.
4118  *
4119  * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags)
4120  * 	Description
4121  * 		Copy *size* bytes from *data* into a ring buffer *ringbuf*.
4122  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4123  * 		of new data availability is sent.
4124  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4125  * 		of new data availability is sent unconditionally.
4126  * 		If **0** is specified in *flags*, an adaptive notification
4127  * 		of new data availability is sent.
4128  *
4129  * 		An adaptive notification is a notification sent whenever the user-space
4130  * 		process has caught up and consumed all available payloads. In case the user-space
4131  * 		process is still processing a previous payload, then no notification is needed
4132  * 		as it will process the newly added payload automatically.
4133  * 	Return
4134  * 		0 on success, or a negative error in case of failure.
4135  *
4136  * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags)
4137  * 	Description
4138  * 		Reserve *size* bytes of payload in a ring buffer *ringbuf*.
4139  * 		*flags* must be 0.
4140  * 	Return
4141  * 		Valid pointer with *size* bytes of memory available; NULL,
4142  * 		otherwise.
4143  *
4144  * void bpf_ringbuf_submit(void *data, u64 flags)
4145  * 	Description
4146  * 		Submit reserved ring buffer sample, pointed to by *data*.
4147  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4148  * 		of new data availability is sent.
4149  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4150  * 		of new data availability is sent unconditionally.
4151  * 		If **0** is specified in *flags*, an adaptive notification
4152  * 		of new data availability is sent.
4153  *
4154  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4155  * 	Return
4156  * 		Nothing. Always succeeds.
4157  *
4158  * void bpf_ringbuf_discard(void *data, u64 flags)
4159  * 	Description
4160  * 		Discard reserved ring buffer sample, pointed to by *data*.
4161  * 		If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification
4162  * 		of new data availability is sent.
4163  * 		If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification
4164  * 		of new data availability is sent unconditionally.
4165  * 		If **0** is specified in *flags*, an adaptive notification
4166  * 		of new data availability is sent.
4167  *
4168  * 		See 'bpf_ringbuf_output()' for the definition of adaptive notification.
4169  * 	Return
4170  * 		Nothing. Always succeeds.
4171  *
4172  * u64 bpf_ringbuf_query(void *ringbuf, u64 flags)
4173  *	Description
4174  *		Query various characteristics of provided ring buffer. What
4175  *		exactly is queries is determined by *flags*:
4176  *
4177  *		* **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed.
4178  *		* **BPF_RB_RING_SIZE**: The size of ring buffer.
4179  *		* **BPF_RB_CONS_POS**: Consumer position (can wrap around).
4180  *		* **BPF_RB_PROD_POS**: Producer(s) position (can wrap around).
4181  *
4182  *		Data returned is just a momentary snapshot of actual values
4183  *		and could be inaccurate, so this facility should be used to
4184  *		power heuristics and for reporting, not to make 100% correct
4185  *		calculation.
4186  *	Return
4187  *		Requested value, or 0, if *flags* are not recognized.
4188  *
4189  * long bpf_csum_level(struct sk_buff *skb, u64 level)
4190  * 	Description
4191  * 		Change the skbs checksum level by one layer up or down, or
4192  * 		reset it entirely to none in order to have the stack perform
4193  * 		checksum validation. The level is applicable to the following
4194  * 		protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of
4195  * 		| ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP |
4196  * 		through **bpf_skb_adjust_room**\ () helper with passing in
4197  * 		**BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one	call
4198  * 		to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since
4199  * 		the UDP header is removed. Similarly, an encap of the latter
4200  * 		into the former could be accompanied by a helper call to
4201  * 		**bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the
4202  * 		skb is still intended to be processed in higher layers of the
4203  * 		stack instead of just egressing at tc.
4204  *
4205  * 		There are three supported level settings at this time:
4206  *
4207  * 		* **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs
4208  * 		  with CHECKSUM_UNNECESSARY.
4209  * 		* **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs
4210  * 		  with CHECKSUM_UNNECESSARY.
4211  * 		* **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and
4212  * 		  sets CHECKSUM_NONE to force checksum validation by the stack.
4213  * 		* **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current
4214  * 		  skb->csum_level.
4215  * 	Return
4216  * 		0 on success, or a negative error in case of failure. In the
4217  * 		case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level
4218  * 		is returned or the error code -EACCES in case the skb is not
4219  * 		subject to CHECKSUM_UNNECESSARY.
4220  *
4221  * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk)
4222  *	Description
4223  *		Dynamically cast a *sk* pointer to a *tcp6_sock* pointer.
4224  *	Return
4225  *		*sk* if casting is valid, or **NULL** otherwise.
4226  *
4227  * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk)
4228  *	Description
4229  *		Dynamically cast a *sk* pointer to a *tcp_sock* pointer.
4230  *	Return
4231  *		*sk* if casting is valid, or **NULL** otherwise.
4232  *
4233  * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk)
4234  * 	Description
4235  *		Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer.
4236  *	Return
4237  *		*sk* if casting is valid, or **NULL** otherwise.
4238  *
4239  * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk)
4240  * 	Description
4241  *		Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer.
4242  *	Return
4243  *		*sk* if casting is valid, or **NULL** otherwise.
4244  *
4245  * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk)
4246  * 	Description
4247  *		Dynamically cast a *sk* pointer to a *udp6_sock* pointer.
4248  *	Return
4249  *		*sk* if casting is valid, or **NULL** otherwise.
4250  *
4251  * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags)
4252  *	Description
4253  *		Return a user or a kernel stack in bpf program provided buffer.
4254  *		To achieve this, the helper needs *task*, which is a valid
4255  *		pointer to **struct task_struct**. To store the stacktrace, the
4256  *		bpf program provides *buf* with a nonnegative *size*.
4257  *
4258  *		The last argument, *flags*, holds the number of stack frames to
4259  *		skip (from 0 to 255), masked with
4260  *		**BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
4261  *		the following flags:
4262  *
4263  *		**BPF_F_USER_STACK**
4264  *			Collect a user space stack instead of a kernel stack.
4265  *		**BPF_F_USER_BUILD_ID**
4266  *			Collect buildid+offset instead of ips for user stack,
4267  *			only valid if **BPF_F_USER_STACK** is also specified.
4268  *
4269  *		**bpf_get_task_stack**\ () can collect up to
4270  *		**PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
4271  *		to sufficient large buffer size. Note that
4272  *		this limit can be controlled with the **sysctl** program, and
4273  *		that it should be manually increased in order to profile long
4274  *		user stacks (such as stacks for Java programs). To do so, use:
4275  *
4276  *		::
4277  *
4278  *			# sysctl kernel.perf_event_max_stack=<new value>
4279  *	Return
4280  *		A non-negative value equal to or less than *size* on success,
4281  *		or a negative error in case of failure.
4282  *
4283  * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags)
4284  *	Description
4285  *		Load header option.  Support reading a particular TCP header
4286  *		option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**).
4287  *
4288  *		If *flags* is 0, it will search the option from the
4289  *		*skops*\ **->skb_data**.  The comment in **struct bpf_sock_ops**
4290  *		has details on what skb_data contains under different
4291  *		*skops*\ **->op**.
4292  *
4293  *		The first byte of the *searchby_res* specifies the
4294  *		kind that it wants to search.
4295  *
4296  *		If the searching kind is an experimental kind
4297  *		(i.e. 253 or 254 according to RFC6994).  It also
4298  *		needs to specify the "magic" which is either
4299  *		2 bytes or 4 bytes.  It then also needs to
4300  *		specify the size of the magic by using
4301  *		the 2nd byte which is "kind-length" of a TCP
4302  *		header option and the "kind-length" also
4303  *		includes the first 2 bytes "kind" and "kind-length"
4304  *		itself as a normal TCP header option also does.
4305  *
4306  *		For example, to search experimental kind 254 with
4307  *		2 byte magic 0xeB9F, the searchby_res should be
4308  *		[ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ].
4309  *
4310  *		To search for the standard window scale option (3),
4311  *		the *searchby_res* should be [ 3, 0, 0, .... 0 ].
4312  *		Note, kind-length must be 0 for regular option.
4313  *
4314  *		Searching for No-Op (0) and End-of-Option-List (1) are
4315  *		not supported.
4316  *
4317  *		*len* must be at least 2 bytes which is the minimal size
4318  *		of a header option.
4319  *
4320  *		Supported flags:
4321  *
4322  *		* **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the
4323  *		  saved_syn packet or the just-received syn packet.
4324  *
4325  *	Return
4326  *		> 0 when found, the header option is copied to *searchby_res*.
4327  *		The return value is the total length copied. On failure, a
4328  *		negative error code is returned:
4329  *
4330  *		**-EINVAL** if a parameter is invalid.
4331  *
4332  *		**-ENOMSG** if the option is not found.
4333  *
4334  *		**-ENOENT** if no syn packet is available when
4335  *		**BPF_LOAD_HDR_OPT_TCP_SYN** is used.
4336  *
4337  *		**-ENOSPC** if there is not enough space.  Only *len* number of
4338  *		bytes are copied.
4339  *
4340  *		**-EFAULT** on failure to parse the header options in the
4341  *		packet.
4342  *
4343  *		**-EPERM** if the helper cannot be used under the current
4344  *		*skops*\ **->op**.
4345  *
4346  * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags)
4347  *	Description
4348  *		Store header option.  The data will be copied
4349  *		from buffer *from* with length *len* to the TCP header.
4350  *
4351  *		The buffer *from* should have the whole option that
4352  *		includes the kind, kind-length, and the actual
4353  *		option data.  The *len* must be at least kind-length
4354  *		long.  The kind-length does not have to be 4 byte
4355  *		aligned.  The kernel will take care of the padding
4356  *		and setting the 4 bytes aligned value to th->doff.
4357  *
4358  *		This helper will check for duplicated option
4359  *		by searching the same option in the outgoing skb.
4360  *
4361  *		This helper can only be called during
4362  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4363  *
4364  *	Return
4365  *		0 on success, or negative error in case of failure:
4366  *
4367  *		**-EINVAL** If param is invalid.
4368  *
4369  *		**-ENOSPC** if there is not enough space in the header.
4370  *		Nothing has been written
4371  *
4372  *		**-EEXIST** if the option already exists.
4373  *
4374  *		**-EFAULT** on failrue to parse the existing header options.
4375  *
4376  *		**-EPERM** if the helper cannot be used under the current
4377  *		*skops*\ **->op**.
4378  *
4379  * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags)
4380  *	Description
4381  *		Reserve *len* bytes for the bpf header option.  The
4382  *		space will be used by **bpf_store_hdr_opt**\ () later in
4383  *		**BPF_SOCK_OPS_WRITE_HDR_OPT_CB**.
4384  *
4385  *		If **bpf_reserve_hdr_opt**\ () is called multiple times,
4386  *		the total number of bytes will be reserved.
4387  *
4388  *		This helper can only be called during
4389  *		**BPF_SOCK_OPS_HDR_OPT_LEN_CB**.
4390  *
4391  *	Return
4392  *		0 on success, or negative error in case of failure:
4393  *
4394  *		**-EINVAL** if a parameter is invalid.
4395  *
4396  *		**-ENOSPC** if there is not enough space in the header.
4397  *
4398  *		**-EPERM** if the helper cannot be used under the current
4399  *		*skops*\ **->op**.
4400  *
4401  * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags)
4402  *	Description
4403  *		Get a bpf_local_storage from an *inode*.
4404  *
4405  *		Logically, it could be thought of as getting the value from
4406  *		a *map* with *inode* as the **key**.  From this
4407  *		perspective,  the usage is not much different from
4408  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this
4409  *		helper enforces the key must be an inode and the map must also
4410  *		be a **BPF_MAP_TYPE_INODE_STORAGE**.
4411  *
4412  *		Underneath, the value is stored locally at *inode* instead of
4413  *		the *map*.  The *map* is used as the bpf-local-storage
4414  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4415  *		searched against all bpf_local_storage residing at *inode*.
4416  *
4417  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4418  *		used such that a new bpf_local_storage will be
4419  *		created if one does not exist.  *value* can be used
4420  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4421  *		the initial value of a bpf_local_storage.  If *value* is
4422  *		**NULL**, the new bpf_local_storage will be zero initialized.
4423  *	Return
4424  *		A bpf_local_storage pointer is returned on success.
4425  *
4426  *		**NULL** if not found or there was an error in adding
4427  *		a new bpf_local_storage.
4428  *
4429  * int bpf_inode_storage_delete(struct bpf_map *map, void *inode)
4430  *	Description
4431  *		Delete a bpf_local_storage from an *inode*.
4432  *	Return
4433  *		0 on success.
4434  *
4435  *		**-ENOENT** if the bpf_local_storage cannot be found.
4436  *
4437  * long bpf_d_path(struct path *path, char *buf, u32 sz)
4438  *	Description
4439  *		Return full path for given **struct path** object, which
4440  *		needs to be the kernel BTF *path* object. The path is
4441  *		returned in the provided buffer *buf* of size *sz* and
4442  *		is zero terminated.
4443  *
4444  *	Return
4445  *		On success, the strictly positive length of the string,
4446  *		including the trailing NUL character. On error, a negative
4447  *		value.
4448  *
4449  * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr)
4450  * 	Description
4451  * 		Read *size* bytes from user space address *user_ptr* and store
4452  * 		the data in *dst*. This is a wrapper of **copy_from_user**\ ().
4453  * 	Return
4454  * 		0 on success, or a negative error in case of failure.
4455  *
4456  * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags)
4457  *	Description
4458  *		Use BTF to store a string representation of *ptr*->ptr in *str*,
4459  *		using *ptr*->type_id.  This value should specify the type
4460  *		that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1)
4461  *		can be used to look up vmlinux BTF type ids. Traversing the
4462  *		data structure using BTF, the type information and values are
4463  *		stored in the first *str_size* - 1 bytes of *str*.  Safe copy of
4464  *		the pointer data is carried out to avoid kernel crashes during
4465  *		operation.  Smaller types can use string space on the stack;
4466  *		larger programs can use map data to store the string
4467  *		representation.
4468  *
4469  *		The string can be subsequently shared with userspace via
4470  *		bpf_perf_event_output() or ring buffer interfaces.
4471  *		bpf_trace_printk() is to be avoided as it places too small
4472  *		a limit on string size to be useful.
4473  *
4474  *		*flags* is a combination of
4475  *
4476  *		**BTF_F_COMPACT**
4477  *			no formatting around type information
4478  *		**BTF_F_NONAME**
4479  *			no struct/union member names/types
4480  *		**BTF_F_PTR_RAW**
4481  *			show raw (unobfuscated) pointer values;
4482  *			equivalent to printk specifier %px.
4483  *		**BTF_F_ZERO**
4484  *			show zero-valued struct/union members; they
4485  *			are not displayed by default
4486  *
4487  *	Return
4488  *		The number of bytes that were written (or would have been
4489  *		written if output had to be truncated due to string size),
4490  *		or a negative error in cases of failure.
4491  *
4492  * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags)
4493  *	Description
4494  *		Use BTF to write to seq_write a string representation of
4495  *		*ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf().
4496  *		*flags* are identical to those used for bpf_snprintf_btf.
4497  *	Return
4498  *		0 on success or a negative error in case of failure.
4499  *
4500  * u64 bpf_skb_cgroup_classid(struct sk_buff *skb)
4501  * 	Description
4502  * 		See **bpf_get_cgroup_classid**\ () for the main description.
4503  * 		This helper differs from **bpf_get_cgroup_classid**\ () in that
4504  * 		the cgroup v1 net_cls class is retrieved only from the *skb*'s
4505  * 		associated socket instead of the current process.
4506  * 	Return
4507  * 		The id is returned or 0 in case the id could not be retrieved.
4508  *
4509  * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
4510  * 	Description
4511  * 		Redirect the packet to another net device of index *ifindex*
4512  * 		and fill in L2 addresses from neighboring subsystem. This helper
4513  * 		is somewhat similar to **bpf_redirect**\ (), except that it
4514  * 		populates L2 addresses as well, meaning, internally, the helper
4515  * 		relies on the neighbor lookup for the L2 address of the nexthop.
4516  *
4517  * 		The helper will perform a FIB lookup based on the skb's
4518  * 		networking header to get the address of the next hop, unless
4519  * 		this is supplied by the caller in the *params* argument. The
4520  * 		*plen* argument indicates the len of *params* and should be set
4521  * 		to 0 if *params* is NULL.
4522  *
4523  * 		The *flags* argument is reserved and must be 0. The helper is
4524  * 		currently only supported for tc BPF program types, and enabled
4525  * 		for IPv4 and IPv6 protocols.
4526  * 	Return
4527  * 		The helper returns **TC_ACT_REDIRECT** on success or
4528  * 		**TC_ACT_SHOT** on error.
4529  *
4530  * void *bpf_per_cpu_ptr(const void *percpu_ptr, u32 cpu)
4531  *     Description
4532  *             Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4533  *             pointer to the percpu kernel variable on *cpu*. A ksym is an
4534  *             extern variable decorated with '__ksym'. For ksym, there is a
4535  *             global var (either static or global) defined of the same name
4536  *             in the kernel. The ksym is percpu if the global var is percpu.
4537  *             The returned pointer points to the global percpu var on *cpu*.
4538  *
4539  *             bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the
4540  *             kernel, except that bpf_per_cpu_ptr() may return NULL. This
4541  *             happens if *cpu* is larger than nr_cpu_ids. The caller of
4542  *             bpf_per_cpu_ptr() must check the returned value.
4543  *     Return
4544  *             A pointer pointing to the kernel percpu variable on *cpu*, or
4545  *             NULL, if *cpu* is invalid.
4546  *
4547  * void *bpf_this_cpu_ptr(const void *percpu_ptr)
4548  *	Description
4549  *		Take a pointer to a percpu ksym, *percpu_ptr*, and return a
4550  *		pointer to the percpu kernel variable on this cpu. See the
4551  *		description of 'ksym' in **bpf_per_cpu_ptr**\ ().
4552  *
4553  *		bpf_this_cpu_ptr() has the same semantic as this_cpu_ptr() in
4554  *		the kernel. Different from **bpf_per_cpu_ptr**\ (), it would
4555  *		never return NULL.
4556  *	Return
4557  *		A pointer pointing to the kernel percpu variable on this cpu.
4558  *
4559  * long bpf_redirect_peer(u32 ifindex, u64 flags)
4560  * 	Description
4561  * 		Redirect the packet to another net device of index *ifindex*.
4562  * 		This helper is somewhat similar to **bpf_redirect**\ (), except
4563  * 		that the redirection happens to the *ifindex*' peer device and
4564  * 		the netns switch takes place from ingress to ingress without
4565  * 		going through the CPU's backlog queue.
4566  *
4567  * 		The *flags* argument is reserved and must be 0. The helper is
4568  * 		currently only supported for tc BPF program types at the ingress
4569  * 		hook and for veth device types. The peer device must reside in a
4570  * 		different network namespace.
4571  * 	Return
4572  * 		The helper returns **TC_ACT_REDIRECT** on success or
4573  * 		**TC_ACT_SHOT** on error.
4574  *
4575  * void *bpf_task_storage_get(struct bpf_map *map, struct task_struct *task, void *value, u64 flags)
4576  *	Description
4577  *		Get a bpf_local_storage from the *task*.
4578  *
4579  *		Logically, it could be thought of as getting the value from
4580  *		a *map* with *task* as the **key**.  From this
4581  *		perspective,  the usage is not much different from
4582  *		**bpf_map_lookup_elem**\ (*map*, **&**\ *task*) except this
4583  *		helper enforces the key must be an task_struct and the map must also
4584  *		be a **BPF_MAP_TYPE_TASK_STORAGE**.
4585  *
4586  *		Underneath, the value is stored locally at *task* instead of
4587  *		the *map*.  The *map* is used as the bpf-local-storage
4588  *		"type". The bpf-local-storage "type" (i.e. the *map*) is
4589  *		searched against all bpf_local_storage residing at *task*.
4590  *
4591  *		An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be
4592  *		used such that a new bpf_local_storage will be
4593  *		created if one does not exist.  *value* can be used
4594  *		together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify
4595  *		the initial value of a bpf_local_storage.  If *value* is
4596  *		**NULL**, the new bpf_local_storage will be zero initialized.
4597  *	Return
4598  *		A bpf_local_storage pointer is returned on success.
4599  *
4600  *		**NULL** if not found or there was an error in adding
4601  *		a new bpf_local_storage.
4602  *
4603  * long bpf_task_storage_delete(struct bpf_map *map, struct task_struct *task)
4604  *	Description
4605  *		Delete a bpf_local_storage from a *task*.
4606  *	Return
4607  *		0 on success.
4608  *
4609  *		**-ENOENT** if the bpf_local_storage cannot be found.
4610  *
4611  * struct task_struct *bpf_get_current_task_btf(void)
4612  *	Description
4613  *		Return a BTF pointer to the "current" task.
4614  *		This pointer can also be used in helpers that accept an
4615  *		*ARG_PTR_TO_BTF_ID* of type *task_struct*.
4616  *	Return
4617  *		Pointer to the current task.
4618  *
4619  * long bpf_bprm_opts_set(struct linux_binprm *bprm, u64 flags)
4620  *	Description
4621  *		Set or clear certain options on *bprm*:
4622  *
4623  *		**BPF_F_BPRM_SECUREEXEC** Set the secureexec bit
4624  *		which sets the **AT_SECURE** auxv for glibc. The bit
4625  *		is cleared if the flag is not specified.
4626  *	Return
4627  *		**-EINVAL** if invalid *flags* are passed, zero otherwise.
4628  *
4629  * u64 bpf_ktime_get_coarse_ns(void)
4630  * 	Description
4631  * 		Return a coarse-grained version of the time elapsed since
4632  * 		system boot, in nanoseconds. Does not include time the system
4633  * 		was suspended.
4634  *
4635  * 		See: **clock_gettime**\ (**CLOCK_MONOTONIC_COARSE**)
4636  * 	Return
4637  * 		Current *ktime*.
4638  *
4639  * long bpf_ima_inode_hash(struct inode *inode, void *dst, u32 size)
4640  *	Description
4641  *		Returns the stored IMA hash of the *inode* (if it's avaialable).
4642  *		If the hash is larger than *size*, then only *size*
4643  *		bytes will be copied to *dst*
4644  *	Return
4645  *		The **hash_algo** is returned on success,
4646  *		**-EOPNOTSUP** if IMA is disabled or **-EINVAL** if
4647  *		invalid arguments are passed.
4648  *
4649  * struct socket *bpf_sock_from_file(struct file *file)
4650  *	Description
4651  *		If the given file represents a socket, returns the associated
4652  *		socket.
4653  *	Return
4654  *		A pointer to a struct socket on success or NULL if the file is
4655  *		not a socket.
4656  *
4657  * long bpf_check_mtu(void *ctx, u32 ifindex, u32 *mtu_len, s32 len_diff, u64 flags)
4658  *	Description
4659  *		Check packet size against exceeding MTU of net device (based
4660  *		on *ifindex*).  This helper will likely be used in combination
4661  *		with helpers that adjust/change the packet size.
4662  *
4663  *		The argument *len_diff* can be used for querying with a planned
4664  *		size change. This allows to check MTU prior to changing packet
4665  *		ctx. Providing an *len_diff* adjustment that is larger than the
4666  *		actual packet size (resulting in negative packet size) will in
4667  *		principle not exceed the MTU, why it is not considered a
4668  *		failure.  Other BPF-helpers are needed for performing the
4669  *		planned size change, why the responsability for catch a negative
4670  *		packet size belong in those helpers.
4671  *
4672  *		Specifying *ifindex* zero means the MTU check is performed
4673  *		against the current net device.  This is practical if this isn't
4674  *		used prior to redirect.
4675  *
4676  *		On input *mtu_len* must be a valid pointer, else verifier will
4677  *		reject BPF program.  If the value *mtu_len* is initialized to
4678  *		zero then the ctx packet size is use.  When value *mtu_len* is
4679  *		provided as input this specify the L3 length that the MTU check
4680  *		is done against. Remember XDP and TC length operate at L2, but
4681  *		this value is L3 as this correlate to MTU and IP-header tot_len
4682  *		values which are L3 (similar behavior as bpf_fib_lookup).
4683  *
4684  *		The Linux kernel route table can configure MTUs on a more
4685  *		specific per route level, which is not provided by this helper.
4686  *		For route level MTU checks use the **bpf_fib_lookup**\ ()
4687  *		helper.
4688  *
4689  *		*ctx* is either **struct xdp_md** for XDP programs or
4690  *		**struct sk_buff** for tc cls_act programs.
4691  *
4692  *		The *flags* argument can be a combination of one or more of the
4693  *		following values:
4694  *
4695  *		**BPF_MTU_CHK_SEGS**
4696  *			This flag will only works for *ctx* **struct sk_buff**.
4697  *			If packet context contains extra packet segment buffers
4698  *			(often knows as GSO skb), then MTU check is harder to
4699  *			check at this point, because in transmit path it is
4700  *			possible for the skb packet to get re-segmented
4701  *			(depending on net device features).  This could still be
4702  *			a MTU violation, so this flag enables performing MTU
4703  *			check against segments, with a different violation
4704  *			return code to tell it apart. Check cannot use len_diff.
4705  *
4706  *		On return *mtu_len* pointer contains the MTU value of the net
4707  *		device.  Remember the net device configured MTU is the L3 size,
4708  *		which is returned here and XDP and TC length operate at L2.
4709  *		Helper take this into account for you, but remember when using
4710  *		MTU value in your BPF-code.
4711  *
4712  *	Return
4713  *		* 0 on success, and populate MTU value in *mtu_len* pointer.
4714  *
4715  *		* < 0 if any input argument is invalid (*mtu_len* not updated)
4716  *
4717  *		MTU violations return positive values, but also populate MTU
4718  *		value in *mtu_len* pointer, as this can be needed for
4719  *		implementing PMTU handing:
4720  *
4721  *		* **BPF_MTU_CHK_RET_FRAG_NEEDED**
4722  *		* **BPF_MTU_CHK_RET_SEGS_TOOBIG**
4723  *
4724  * long bpf_for_each_map_elem(struct bpf_map *map, void *callback_fn, void *callback_ctx, u64 flags)
4725  *	Description
4726  *		For each element in **map**, call **callback_fn** function with
4727  *		**map**, **callback_ctx** and other map-specific parameters.
4728  *		The **callback_fn** should be a static function and
4729  *		the **callback_ctx** should be a pointer to the stack.
4730  *		The **flags** is used to control certain aspects of the helper.
4731  *		Currently, the **flags** must be 0.
4732  *
4733  *		The following are a list of supported map types and their
4734  *		respective expected callback signatures:
4735  *
4736  *		BPF_MAP_TYPE_HASH, BPF_MAP_TYPE_PERCPU_HASH,
4737  *		BPF_MAP_TYPE_LRU_HASH, BPF_MAP_TYPE_LRU_PERCPU_HASH,
4738  *		BPF_MAP_TYPE_ARRAY, BPF_MAP_TYPE_PERCPU_ARRAY
4739  *
4740  *		long (\*callback_fn)(struct bpf_map \*map, const void \*key, void \*value, void \*ctx);
4741  *
4742  *		For per_cpu maps, the map_value is the value on the cpu where the
4743  *		bpf_prog is running.
4744  *
4745  *		If **callback_fn** return 0, the helper will continue to the next
4746  *		element. If return value is 1, the helper will skip the rest of
4747  *		elements and return. Other return values are not used now.
4748  *
4749  *	Return
4750  *		The number of traversed map elements for success, **-EINVAL** for
4751  *		invalid **flags**.
4752  *
4753  * long bpf_snprintf(char *str, u32 str_size, const char *fmt, u64 *data, u32 data_len)
4754  *	Description
4755  *		Outputs a string into the **str** buffer of size **str_size**
4756  *		based on a format string stored in a read-only map pointed by
4757  *		**fmt**.
4758  *
4759  *		Each format specifier in **fmt** corresponds to one u64 element
4760  *		in the **data** array. For strings and pointers where pointees
4761  *		are accessed, only the pointer values are stored in the *data*
4762  *		array. The *data_len* is the size of *data* in bytes - must be
4763  *		a multiple of 8.
4764  *
4765  *		Formats **%s** and **%p{i,I}{4,6}** require to read kernel
4766  *		memory. Reading kernel memory may fail due to either invalid
4767  *		address or valid address but requiring a major memory fault. If
4768  *		reading kernel memory fails, the string for **%s** will be an
4769  *		empty string, and the ip address for **%p{i,I}{4,6}** will be 0.
4770  *		Not returning error to bpf program is consistent with what
4771  *		**bpf_trace_printk**\ () does for now.
4772  *
4773  *	Return
4774  *		The strictly positive length of the formatted string, including
4775  *		the trailing zero character. If the return value is greater than
4776  *		**str_size**, **str** contains a truncated string, guaranteed to
4777  *		be zero-terminated except when **str_size** is 0.
4778  *
4779  *		Or **-EBUSY** if the per-CPU memory copy buffer is busy.
4780  *
4781  * long bpf_sys_bpf(u32 cmd, void *attr, u32 attr_size)
4782  * 	Description
4783  * 		Execute bpf syscall with given arguments.
4784  * 	Return
4785  * 		A syscall result.
4786  *
4787  * long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
4788  * 	Description
4789  * 		Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
4790  * 	Return
4791  * 		Returns btf_id and btf_obj_fd in lower and upper 32 bits.
4792  *
4793  * long bpf_sys_close(u32 fd)
4794  * 	Description
4795  * 		Execute close syscall for given FD.
4796  * 	Return
4797  * 		A syscall result.
4798  *
4799  * long bpf_timer_init(struct bpf_timer *timer, struct bpf_map *map, u64 flags)
4800  *	Description
4801  *		Initialize the timer.
4802  *		First 4 bits of *flags* specify clockid.
4803  *		Only CLOCK_MONOTONIC, CLOCK_REALTIME, CLOCK_BOOTTIME are allowed.
4804  *		All other bits of *flags* are reserved.
4805  *		The verifier will reject the program if *timer* is not from
4806  *		the same *map*.
4807  *	Return
4808  *		0 on success.
4809  *		**-EBUSY** if *timer* is already initialized.
4810  *		**-EINVAL** if invalid *flags* are passed.
4811  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4812  *		The user space should either hold a file descriptor to a map with timers
4813  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4814  *		closed all timers in the map will be cancelled and freed.
4815  *
4816  * long bpf_timer_set_callback(struct bpf_timer *timer, void *callback_fn)
4817  *	Description
4818  *		Configure the timer to call *callback_fn* static function.
4819  *	Return
4820  *		0 on success.
4821  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4822  *		**-EPERM** if *timer* is in a map that doesn't have any user references.
4823  *		The user space should either hold a file descriptor to a map with timers
4824  *		or pin such map in bpffs. When map is unpinned or file descriptor is
4825  *		closed all timers in the map will be cancelled and freed.
4826  *
4827  * long bpf_timer_start(struct bpf_timer *timer, u64 nsecs, u64 flags)
4828  *	Description
4829  *		Set timer expiration N nanoseconds from the current time. The
4830  *		configured callback will be invoked in soft irq context on some cpu
4831  *		and will not repeat unless another bpf_timer_start() is made.
4832  *		In such case the next invocation can migrate to a different cpu.
4833  *		Since struct bpf_timer is a field inside map element the map
4834  *		owns the timer. The bpf_timer_set_callback() will increment refcnt
4835  *		of BPF program to make sure that callback_fn code stays valid.
4836  *		When user space reference to a map reaches zero all timers
4837  *		in a map are cancelled and corresponding program's refcnts are
4838  *		decremented. This is done to make sure that Ctrl-C of a user
4839  *		process doesn't leave any timers running. If map is pinned in
4840  *		bpffs the callback_fn can re-arm itself indefinitely.
4841  *		bpf_map_update/delete_elem() helpers and user space sys_bpf commands
4842  *		cancel and free the timer in the given map element.
4843  *		The map can contain timers that invoke callback_fn-s from different
4844  *		programs. The same callback_fn can serve different timers from
4845  *		different maps if key/value layout matches across maps.
4846  *		Every bpf_timer_set_callback() can have different callback_fn.
4847  *
4848  *	Return
4849  *		0 on success.
4850  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier
4851  *		or invalid *flags* are passed.
4852  *
4853  * long bpf_timer_cancel(struct bpf_timer *timer)
4854  *	Description
4855  *		Cancel the timer and wait for callback_fn to finish if it was running.
4856  *	Return
4857  *		0 if the timer was not active.
4858  *		1 if the timer was active.
4859  *		**-EINVAL** if *timer* was not initialized with bpf_timer_init() earlier.
4860  *		**-EDEADLK** if callback_fn tried to call bpf_timer_cancel() on its
4861  *		own timer which would have led to a deadlock otherwise.
4862  *
4863  * u64 bpf_get_func_ip(void *ctx)
4864  * 	Description
4865  * 		Get address of the traced function (for tracing and kprobe programs).
4866  * 	Return
4867  * 		Address of the traced function.
4868  *
4869  * u64 bpf_get_attach_cookie(void *ctx)
4870  * 	Description
4871  * 		Get bpf_cookie value provided (optionally) during the program
4872  * 		attachment. It might be different for each individual
4873  * 		attachment, even if BPF program itself is the same.
4874  * 		Expects BPF program context *ctx* as a first argument.
4875  *
4876  * 		Supported for the following program types:
4877  *			- kprobe/uprobe;
4878  *			- tracepoint;
4879  *			- perf_event.
4880  * 	Return
4881  *		Value specified by user at BPF link creation/attachment time
4882  *		or 0, if it was not specified.
4883  *
4884  * long bpf_task_pt_regs(struct task_struct *task)
4885  *	Description
4886  *		Get the struct pt_regs associated with **task**.
4887  *	Return
4888  *		A pointer to struct pt_regs.
4889  *
4890  * long bpf_get_branch_snapshot(void *entries, u32 size, u64 flags)
4891  *	Description
4892  *		Get branch trace from hardware engines like Intel LBR. The
4893  *		hardware engine is stopped shortly after the helper is
4894  *		called. Therefore, the user need to filter branch entries
4895  *		based on the actual use case. To capture branch trace
4896  *		before the trigger point of the BPF program, the helper
4897  *		should be called at the beginning of the BPF program.
4898  *
4899  *		The data is stored as struct perf_branch_entry into output
4900  *		buffer *entries*. *size* is the size of *entries* in bytes.
4901  *		*flags* is reserved for now and must be zero.
4902  *
4903  *	Return
4904  *		On success, number of bytes written to *buf*. On error, a
4905  *		negative value.
4906  *
4907  *		**-EINVAL** if *flags* is not zero.
4908  *
4909  *		**-ENOENT** if architecture does not support branch records.
4910  *
4911  * long bpf_trace_vprintk(const char *fmt, u32 fmt_size, const void *data, u32 data_len)
4912  *	Description
4913  *		Behaves like **bpf_trace_printk**\ () helper, but takes an array of u64
4914  *		to format and can handle more format args as a result.
4915  *
4916  *		Arguments are to be used as in **bpf_seq_printf**\ () helper.
4917  *	Return
4918  *		The number of bytes written to the buffer, or a negative error
4919  *		in case of failure.
4920  *
4921  * struct unix_sock *bpf_skc_to_unix_sock(void *sk)
4922  * 	Description
4923  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
4924  *	Return
4925  *		*sk* if casting is valid, or **NULL** otherwise.
4926  *
4927  * long bpf_kallsyms_lookup_name(const char *name, int name_sz, int flags, u64 *res)
4928  *	Description
4929  *		Get the address of a kernel symbol, returned in *res*. *res* is
4930  *		set to 0 if the symbol is not found.
4931  *	Return
4932  *		On success, zero. On error, a negative value.
4933  *
4934  *		**-EINVAL** if *flags* is not zero.
4935  *
4936  *		**-EINVAL** if string *name* is not the same size as *name_sz*.
4937  *
4938  *		**-ENOENT** if symbol is not found.
4939  *
4940  *		**-EPERM** if caller does not have permission to obtain kernel address.
4941  */
4942 #define __BPF_FUNC_MAPPER(FN)		\
4943 	FN(unspec),			\
4944 	FN(map_lookup_elem),		\
4945 	FN(map_update_elem),		\
4946 	FN(map_delete_elem),		\
4947 	FN(probe_read),			\
4948 	FN(ktime_get_ns),		\
4949 	FN(trace_printk),		\
4950 	FN(get_prandom_u32),		\
4951 	FN(get_smp_processor_id),	\
4952 	FN(skb_store_bytes),		\
4953 	FN(l3_csum_replace),		\
4954 	FN(l4_csum_replace),		\
4955 	FN(tail_call),			\
4956 	FN(clone_redirect),		\
4957 	FN(get_current_pid_tgid),	\
4958 	FN(get_current_uid_gid),	\
4959 	FN(get_current_comm),		\
4960 	FN(get_cgroup_classid),		\
4961 	FN(skb_vlan_push),		\
4962 	FN(skb_vlan_pop),		\
4963 	FN(skb_get_tunnel_key),		\
4964 	FN(skb_set_tunnel_key),		\
4965 	FN(perf_event_read),		\
4966 	FN(redirect),			\
4967 	FN(get_route_realm),		\
4968 	FN(perf_event_output),		\
4969 	FN(skb_load_bytes),		\
4970 	FN(get_stackid),		\
4971 	FN(csum_diff),			\
4972 	FN(skb_get_tunnel_opt),		\
4973 	FN(skb_set_tunnel_opt),		\
4974 	FN(skb_change_proto),		\
4975 	FN(skb_change_type),		\
4976 	FN(skb_under_cgroup),		\
4977 	FN(get_hash_recalc),		\
4978 	FN(get_current_task),		\
4979 	FN(probe_write_user),		\
4980 	FN(current_task_under_cgroup),	\
4981 	FN(skb_change_tail),		\
4982 	FN(skb_pull_data),		\
4983 	FN(csum_update),		\
4984 	FN(set_hash_invalid),		\
4985 	FN(get_numa_node_id),		\
4986 	FN(skb_change_head),		\
4987 	FN(xdp_adjust_head),		\
4988 	FN(probe_read_str),		\
4989 	FN(get_socket_cookie),		\
4990 	FN(get_socket_uid),		\
4991 	FN(set_hash),			\
4992 	FN(setsockopt),			\
4993 	FN(skb_adjust_room),		\
4994 	FN(redirect_map),		\
4995 	FN(sk_redirect_map),		\
4996 	FN(sock_map_update),		\
4997 	FN(xdp_adjust_meta),		\
4998 	FN(perf_event_read_value),	\
4999 	FN(perf_prog_read_value),	\
5000 	FN(getsockopt),			\
5001 	FN(override_return),		\
5002 	FN(sock_ops_cb_flags_set),	\
5003 	FN(msg_redirect_map),		\
5004 	FN(msg_apply_bytes),		\
5005 	FN(msg_cork_bytes),		\
5006 	FN(msg_pull_data),		\
5007 	FN(bind),			\
5008 	FN(xdp_adjust_tail),		\
5009 	FN(skb_get_xfrm_state),		\
5010 	FN(get_stack),			\
5011 	FN(skb_load_bytes_relative),	\
5012 	FN(fib_lookup),			\
5013 	FN(sock_hash_update),		\
5014 	FN(msg_redirect_hash),		\
5015 	FN(sk_redirect_hash),		\
5016 	FN(lwt_push_encap),		\
5017 	FN(lwt_seg6_store_bytes),	\
5018 	FN(lwt_seg6_adjust_srh),	\
5019 	FN(lwt_seg6_action),		\
5020 	FN(rc_repeat),			\
5021 	FN(rc_keydown),			\
5022 	FN(skb_cgroup_id),		\
5023 	FN(get_current_cgroup_id),	\
5024 	FN(get_local_storage),		\
5025 	FN(sk_select_reuseport),	\
5026 	FN(skb_ancestor_cgroup_id),	\
5027 	FN(sk_lookup_tcp),		\
5028 	FN(sk_lookup_udp),		\
5029 	FN(sk_release),			\
5030 	FN(map_push_elem),		\
5031 	FN(map_pop_elem),		\
5032 	FN(map_peek_elem),		\
5033 	FN(msg_push_data),		\
5034 	FN(msg_pop_data),		\
5035 	FN(rc_pointer_rel),		\
5036 	FN(spin_lock),			\
5037 	FN(spin_unlock),		\
5038 	FN(sk_fullsock),		\
5039 	FN(tcp_sock),			\
5040 	FN(skb_ecn_set_ce),		\
5041 	FN(get_listener_sock),		\
5042 	FN(skc_lookup_tcp),		\
5043 	FN(tcp_check_syncookie),	\
5044 	FN(sysctl_get_name),		\
5045 	FN(sysctl_get_current_value),	\
5046 	FN(sysctl_get_new_value),	\
5047 	FN(sysctl_set_new_value),	\
5048 	FN(strtol),			\
5049 	FN(strtoul),			\
5050 	FN(sk_storage_get),		\
5051 	FN(sk_storage_delete),		\
5052 	FN(send_signal),		\
5053 	FN(tcp_gen_syncookie),		\
5054 	FN(skb_output),			\
5055 	FN(probe_read_user),		\
5056 	FN(probe_read_kernel),		\
5057 	FN(probe_read_user_str),	\
5058 	FN(probe_read_kernel_str),	\
5059 	FN(tcp_send_ack),		\
5060 	FN(send_signal_thread),		\
5061 	FN(jiffies64),			\
5062 	FN(read_branch_records),	\
5063 	FN(get_ns_current_pid_tgid),	\
5064 	FN(xdp_output),			\
5065 	FN(get_netns_cookie),		\
5066 	FN(get_current_ancestor_cgroup_id),	\
5067 	FN(sk_assign),			\
5068 	FN(ktime_get_boot_ns),		\
5069 	FN(seq_printf),			\
5070 	FN(seq_write),			\
5071 	FN(sk_cgroup_id),		\
5072 	FN(sk_ancestor_cgroup_id),	\
5073 	FN(ringbuf_output),		\
5074 	FN(ringbuf_reserve),		\
5075 	FN(ringbuf_submit),		\
5076 	FN(ringbuf_discard),		\
5077 	FN(ringbuf_query),		\
5078 	FN(csum_level),			\
5079 	FN(skc_to_tcp6_sock),		\
5080 	FN(skc_to_tcp_sock),		\
5081 	FN(skc_to_tcp_timewait_sock),	\
5082 	FN(skc_to_tcp_request_sock),	\
5083 	FN(skc_to_udp6_sock),		\
5084 	FN(get_task_stack),		\
5085 	FN(load_hdr_opt),		\
5086 	FN(store_hdr_opt),		\
5087 	FN(reserve_hdr_opt),		\
5088 	FN(inode_storage_get),		\
5089 	FN(inode_storage_delete),	\
5090 	FN(d_path),			\
5091 	FN(copy_from_user),		\
5092 	FN(snprintf_btf),		\
5093 	FN(seq_printf_btf),		\
5094 	FN(skb_cgroup_classid),		\
5095 	FN(redirect_neigh),		\
5096 	FN(per_cpu_ptr),		\
5097 	FN(this_cpu_ptr),		\
5098 	FN(redirect_peer),		\
5099 	FN(task_storage_get),		\
5100 	FN(task_storage_delete),	\
5101 	FN(get_current_task_btf),	\
5102 	FN(bprm_opts_set),		\
5103 	FN(ktime_get_coarse_ns),	\
5104 	FN(ima_inode_hash),		\
5105 	FN(sock_from_file),		\
5106 	FN(check_mtu),			\
5107 	FN(for_each_map_elem),		\
5108 	FN(snprintf),			\
5109 	FN(sys_bpf),			\
5110 	FN(btf_find_by_name_kind),	\
5111 	FN(sys_close),			\
5112 	FN(timer_init),			\
5113 	FN(timer_set_callback),		\
5114 	FN(timer_start),		\
5115 	FN(timer_cancel),		\
5116 	FN(get_func_ip),		\
5117 	FN(get_attach_cookie),		\
5118 	FN(task_pt_regs),		\
5119 	FN(get_branch_snapshot),	\
5120 	FN(trace_vprintk),		\
5121 	FN(skc_to_unix_sock),		\
5122 	FN(kallsyms_lookup_name),	\
5123 	/* */
5124 
5125 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
5126  * function eBPF program intends to call
5127  */
5128 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
5129 enum bpf_func_id {
5130 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
5131 	__BPF_FUNC_MAX_ID,
5132 };
5133 #undef __BPF_ENUM_FN
5134 
5135 /* All flags used by eBPF helper functions, placed here. */
5136 
5137 /* BPF_FUNC_skb_store_bytes flags. */
5138 enum {
5139 	BPF_F_RECOMPUTE_CSUM		= (1ULL << 0),
5140 	BPF_F_INVALIDATE_HASH		= (1ULL << 1),
5141 };
5142 
5143 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
5144  * First 4 bits are for passing the header field size.
5145  */
5146 enum {
5147 	BPF_F_HDR_FIELD_MASK		= 0xfULL,
5148 };
5149 
5150 /* BPF_FUNC_l4_csum_replace flags. */
5151 enum {
5152 	BPF_F_PSEUDO_HDR		= (1ULL << 4),
5153 	BPF_F_MARK_MANGLED_0		= (1ULL << 5),
5154 	BPF_F_MARK_ENFORCE		= (1ULL << 6),
5155 };
5156 
5157 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
5158 enum {
5159 	BPF_F_INGRESS			= (1ULL << 0),
5160 };
5161 
5162 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
5163 enum {
5164 	BPF_F_TUNINFO_IPV6		= (1ULL << 0),
5165 };
5166 
5167 /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
5168 enum {
5169 	BPF_F_SKIP_FIELD_MASK		= 0xffULL,
5170 	BPF_F_USER_STACK		= (1ULL << 8),
5171 /* flags used by BPF_FUNC_get_stackid only. */
5172 	BPF_F_FAST_STACK_CMP		= (1ULL << 9),
5173 	BPF_F_REUSE_STACKID		= (1ULL << 10),
5174 /* flags used by BPF_FUNC_get_stack only. */
5175 	BPF_F_USER_BUILD_ID		= (1ULL << 11),
5176 };
5177 
5178 /* BPF_FUNC_skb_set_tunnel_key flags. */
5179 enum {
5180 	BPF_F_ZERO_CSUM_TX		= (1ULL << 1),
5181 	BPF_F_DONT_FRAGMENT		= (1ULL << 2),
5182 	BPF_F_SEQ_NUMBER		= (1ULL << 3),
5183 };
5184 
5185 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
5186  * BPF_FUNC_perf_event_read_value flags.
5187  */
5188 enum {
5189 	BPF_F_INDEX_MASK		= 0xffffffffULL,
5190 	BPF_F_CURRENT_CPU		= BPF_F_INDEX_MASK,
5191 /* BPF_FUNC_perf_event_output for sk_buff input context. */
5192 	BPF_F_CTXLEN_MASK		= (0xfffffULL << 32),
5193 };
5194 
5195 /* Current network namespace */
5196 enum {
5197 	BPF_F_CURRENT_NETNS		= (-1L),
5198 };
5199 
5200 /* BPF_FUNC_csum_level level values. */
5201 enum {
5202 	BPF_CSUM_LEVEL_QUERY,
5203 	BPF_CSUM_LEVEL_INC,
5204 	BPF_CSUM_LEVEL_DEC,
5205 	BPF_CSUM_LEVEL_RESET,
5206 };
5207 
5208 /* BPF_FUNC_skb_adjust_room flags. */
5209 enum {
5210 	BPF_F_ADJ_ROOM_FIXED_GSO	= (1ULL << 0),
5211 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	= (1ULL << 1),
5212 	BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	= (1ULL << 2),
5213 	BPF_F_ADJ_ROOM_ENCAP_L4_GRE	= (1ULL << 3),
5214 	BPF_F_ADJ_ROOM_ENCAP_L4_UDP	= (1ULL << 4),
5215 	BPF_F_ADJ_ROOM_NO_CSUM_RESET	= (1ULL << 5),
5216 	BPF_F_ADJ_ROOM_ENCAP_L2_ETH	= (1ULL << 6),
5217 };
5218 
5219 enum {
5220 	BPF_ADJ_ROOM_ENCAP_L2_MASK	= 0xff,
5221 	BPF_ADJ_ROOM_ENCAP_L2_SHIFT	= 56,
5222 };
5223 
5224 #define BPF_F_ADJ_ROOM_ENCAP_L2(len)	(((__u64)len & \
5225 					  BPF_ADJ_ROOM_ENCAP_L2_MASK) \
5226 					 << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
5227 
5228 /* BPF_FUNC_sysctl_get_name flags. */
5229 enum {
5230 	BPF_F_SYSCTL_BASE_NAME		= (1ULL << 0),
5231 };
5232 
5233 /* BPF_FUNC_<kernel_obj>_storage_get flags */
5234 enum {
5235 	BPF_LOCAL_STORAGE_GET_F_CREATE	= (1ULL << 0),
5236 	/* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility
5237 	 * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead.
5238 	 */
5239 	BPF_SK_STORAGE_GET_F_CREATE  = BPF_LOCAL_STORAGE_GET_F_CREATE,
5240 };
5241 
5242 /* BPF_FUNC_read_branch_records flags. */
5243 enum {
5244 	BPF_F_GET_BRANCH_RECORDS_SIZE	= (1ULL << 0),
5245 };
5246 
5247 /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and
5248  * BPF_FUNC_bpf_ringbuf_output flags.
5249  */
5250 enum {
5251 	BPF_RB_NO_WAKEUP		= (1ULL << 0),
5252 	BPF_RB_FORCE_WAKEUP		= (1ULL << 1),
5253 };
5254 
5255 /* BPF_FUNC_bpf_ringbuf_query flags */
5256 enum {
5257 	BPF_RB_AVAIL_DATA = 0,
5258 	BPF_RB_RING_SIZE = 1,
5259 	BPF_RB_CONS_POS = 2,
5260 	BPF_RB_PROD_POS = 3,
5261 };
5262 
5263 /* BPF ring buffer constants */
5264 enum {
5265 	BPF_RINGBUF_BUSY_BIT		= (1U << 31),
5266 	BPF_RINGBUF_DISCARD_BIT		= (1U << 30),
5267 	BPF_RINGBUF_HDR_SZ		= 8,
5268 };
5269 
5270 /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */
5271 enum {
5272 	BPF_SK_LOOKUP_F_REPLACE		= (1ULL << 0),
5273 	BPF_SK_LOOKUP_F_NO_REUSEPORT	= (1ULL << 1),
5274 };
5275 
5276 /* Mode for BPF_FUNC_skb_adjust_room helper. */
5277 enum bpf_adj_room_mode {
5278 	BPF_ADJ_ROOM_NET,
5279 	BPF_ADJ_ROOM_MAC,
5280 };
5281 
5282 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
5283 enum bpf_hdr_start_off {
5284 	BPF_HDR_START_MAC,
5285 	BPF_HDR_START_NET,
5286 };
5287 
5288 /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
5289 enum bpf_lwt_encap_mode {
5290 	BPF_LWT_ENCAP_SEG6,
5291 	BPF_LWT_ENCAP_SEG6_INLINE,
5292 	BPF_LWT_ENCAP_IP,
5293 };
5294 
5295 /* Flags for bpf_bprm_opts_set helper */
5296 enum {
5297 	BPF_F_BPRM_SECUREEXEC	= (1ULL << 0),
5298 };
5299 
5300 /* Flags for bpf_redirect_map helper */
5301 enum {
5302 	BPF_F_BROADCAST		= (1ULL << 3),
5303 	BPF_F_EXCLUDE_INGRESS	= (1ULL << 4),
5304 };
5305 
5306 #define __bpf_md_ptr(type, name)	\
5307 union {					\
5308 	type name;			\
5309 	__u64 :64;			\
5310 } __attribute__((aligned(8)))
5311 
5312 /* user accessible mirror of in-kernel sk_buff.
5313  * new fields can only be added to the end of this structure
5314  */
5315 struct __sk_buff {
5316 	__u32 len;
5317 	__u32 pkt_type;
5318 	__u32 mark;
5319 	__u32 queue_mapping;
5320 	__u32 protocol;
5321 	__u32 vlan_present;
5322 	__u32 vlan_tci;
5323 	__u32 vlan_proto;
5324 	__u32 priority;
5325 	__u32 ingress_ifindex;
5326 	__u32 ifindex;
5327 	__u32 tc_index;
5328 	__u32 cb[5];
5329 	__u32 hash;
5330 	__u32 tc_classid;
5331 	__u32 data;
5332 	__u32 data_end;
5333 	__u32 napi_id;
5334 
5335 	/* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
5336 	__u32 family;
5337 	__u32 remote_ip4;	/* Stored in network byte order */
5338 	__u32 local_ip4;	/* Stored in network byte order */
5339 	__u32 remote_ip6[4];	/* Stored in network byte order */
5340 	__u32 local_ip6[4];	/* Stored in network byte order */
5341 	__u32 remote_port;	/* Stored in network byte order */
5342 	__u32 local_port;	/* stored in host byte order */
5343 	/* ... here. */
5344 
5345 	__u32 data_meta;
5346 	__bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
5347 	__u64 tstamp;
5348 	__u32 wire_len;
5349 	__u32 gso_segs;
5350 	__bpf_md_ptr(struct bpf_sock *, sk);
5351 	__u32 gso_size;
5352 	__u32 :32;		/* Padding, future use. */
5353 	__u64 hwtstamp;
5354 };
5355 
5356 struct bpf_tunnel_key {
5357 	__u32 tunnel_id;
5358 	union {
5359 		__u32 remote_ipv4;
5360 		__u32 remote_ipv6[4];
5361 	};
5362 	__u8 tunnel_tos;
5363 	__u8 tunnel_ttl;
5364 	__u16 tunnel_ext;	/* Padding, future use. */
5365 	__u32 tunnel_label;
5366 };
5367 
5368 /* user accessible mirror of in-kernel xfrm_state.
5369  * new fields can only be added to the end of this structure
5370  */
5371 struct bpf_xfrm_state {
5372 	__u32 reqid;
5373 	__u32 spi;	/* Stored in network byte order */
5374 	__u16 family;
5375 	__u16 ext;	/* Padding, future use. */
5376 	union {
5377 		__u32 remote_ipv4;	/* Stored in network byte order */
5378 		__u32 remote_ipv6[4];	/* Stored in network byte order */
5379 	};
5380 };
5381 
5382 /* Generic BPF return codes which all BPF program types may support.
5383  * The values are binary compatible with their TC_ACT_* counter-part to
5384  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
5385  * programs.
5386  *
5387  * XDP is handled seprately, see XDP_*.
5388  */
5389 enum bpf_ret_code {
5390 	BPF_OK = 0,
5391 	/* 1 reserved */
5392 	BPF_DROP = 2,
5393 	/* 3-6 reserved */
5394 	BPF_REDIRECT = 7,
5395 	/* >127 are reserved for prog type specific return codes.
5396 	 *
5397 	 * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and
5398 	 *    BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been
5399 	 *    changed and should be routed based on its new L3 header.
5400 	 *    (This is an L3 redirect, as opposed to L2 redirect
5401 	 *    represented by BPF_REDIRECT above).
5402 	 */
5403 	BPF_LWT_REROUTE = 128,
5404 };
5405 
5406 struct bpf_sock {
5407 	__u32 bound_dev_if;
5408 	__u32 family;
5409 	__u32 type;
5410 	__u32 protocol;
5411 	__u32 mark;
5412 	__u32 priority;
5413 	/* IP address also allows 1 and 2 bytes access */
5414 	__u32 src_ip4;
5415 	__u32 src_ip6[4];
5416 	__u32 src_port;		/* host byte order */
5417 	__u32 dst_port;		/* network byte order */
5418 	__u32 dst_ip4;
5419 	__u32 dst_ip6[4];
5420 	__u32 state;
5421 	__s32 rx_queue_mapping;
5422 };
5423 
5424 struct bpf_tcp_sock {
5425 	__u32 snd_cwnd;		/* Sending congestion window		*/
5426 	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
5427 	__u32 rtt_min;
5428 	__u32 snd_ssthresh;	/* Slow start size threshold		*/
5429 	__u32 rcv_nxt;		/* What we want to receive next		*/
5430 	__u32 snd_nxt;		/* Next sequence we send		*/
5431 	__u32 snd_una;		/* First byte we want an ack for	*/
5432 	__u32 mss_cache;	/* Cached effective mss, not including SACKS */
5433 	__u32 ecn_flags;	/* ECN status bits.			*/
5434 	__u32 rate_delivered;	/* saved rate sample: packets delivered */
5435 	__u32 rate_interval_us;	/* saved rate sample: time elapsed */
5436 	__u32 packets_out;	/* Packets which are "in flight"	*/
5437 	__u32 retrans_out;	/* Retransmitted packets out		*/
5438 	__u32 total_retrans;	/* Total retransmits for entire connection */
5439 	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
5440 				 * total number of segments in.
5441 				 */
5442 	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
5443 				 * total number of data segments in.
5444 				 */
5445 	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
5446 				 * The total number of segments sent.
5447 				 */
5448 	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
5449 				 * total number of data segments sent.
5450 				 */
5451 	__u32 lost_out;		/* Lost packets			*/
5452 	__u32 sacked_out;	/* SACK'd packets			*/
5453 	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
5454 				 * sum(delta(rcv_nxt)), or how many bytes
5455 				 * were acked.
5456 				 */
5457 	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
5458 				 * sum(delta(snd_una)), or how many bytes
5459 				 * were acked.
5460 				 */
5461 	__u32 dsack_dups;	/* RFC4898 tcpEStatsStackDSACKDups
5462 				 * total number of DSACK blocks received
5463 				 */
5464 	__u32 delivered;	/* Total data packets delivered incl. rexmits */
5465 	__u32 delivered_ce;	/* Like the above but only ECE marked packets */
5466 	__u32 icsk_retransmits;	/* Number of unrecovered [RTO] timeouts */
5467 };
5468 
5469 struct bpf_sock_tuple {
5470 	union {
5471 		struct {
5472 			__be32 saddr;
5473 			__be32 daddr;
5474 			__be16 sport;
5475 			__be16 dport;
5476 		} ipv4;
5477 		struct {
5478 			__be32 saddr[4];
5479 			__be32 daddr[4];
5480 			__be16 sport;
5481 			__be16 dport;
5482 		} ipv6;
5483 	};
5484 };
5485 
5486 struct bpf_xdp_sock {
5487 	__u32 queue_id;
5488 };
5489 
5490 #define XDP_PACKET_HEADROOM 256
5491 
5492 /* User return codes for XDP prog type.
5493  * A valid XDP program must return one of these defined values. All other
5494  * return codes are reserved for future use. Unknown return codes will
5495  * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
5496  */
5497 enum xdp_action {
5498 	XDP_ABORTED = 0,
5499 	XDP_DROP,
5500 	XDP_PASS,
5501 	XDP_TX,
5502 	XDP_REDIRECT,
5503 };
5504 
5505 /* user accessible metadata for XDP packet hook
5506  * new fields must be added to the end of this structure
5507  */
5508 struct xdp_md {
5509 	__u32 data;
5510 	__u32 data_end;
5511 	__u32 data_meta;
5512 	/* Below access go through struct xdp_rxq_info */
5513 	__u32 ingress_ifindex; /* rxq->dev->ifindex */
5514 	__u32 rx_queue_index;  /* rxq->queue_index  */
5515 
5516 	__u32 egress_ifindex;  /* txq->dev->ifindex */
5517 };
5518 
5519 /* DEVMAP map-value layout
5520  *
5521  * The struct data-layout of map-value is a configuration interface.
5522  * New members can only be added to the end of this structure.
5523  */
5524 struct bpf_devmap_val {
5525 	__u32 ifindex;   /* device index */
5526 	union {
5527 		int   fd;  /* prog fd on map write */
5528 		__u32 id;  /* prog id on map read */
5529 	} bpf_prog;
5530 };
5531 
5532 /* CPUMAP map-value layout
5533  *
5534  * The struct data-layout of map-value is a configuration interface.
5535  * New members can only be added to the end of this structure.
5536  */
5537 struct bpf_cpumap_val {
5538 	__u32 qsize;	/* queue size to remote target CPU */
5539 	union {
5540 		int   fd;	/* prog fd on map write */
5541 		__u32 id;	/* prog id on map read */
5542 	} bpf_prog;
5543 };
5544 
5545 enum sk_action {
5546 	SK_DROP = 0,
5547 	SK_PASS,
5548 };
5549 
5550 /* user accessible metadata for SK_MSG packet hook, new fields must
5551  * be added to the end of this structure
5552  */
5553 struct sk_msg_md {
5554 	__bpf_md_ptr(void *, data);
5555 	__bpf_md_ptr(void *, data_end);
5556 
5557 	__u32 family;
5558 	__u32 remote_ip4;	/* Stored in network byte order */
5559 	__u32 local_ip4;	/* Stored in network byte order */
5560 	__u32 remote_ip6[4];	/* Stored in network byte order */
5561 	__u32 local_ip6[4];	/* Stored in network byte order */
5562 	__u32 remote_port;	/* Stored in network byte order */
5563 	__u32 local_port;	/* stored in host byte order */
5564 	__u32 size;		/* Total size of sk_msg */
5565 
5566 	__bpf_md_ptr(struct bpf_sock *, sk); /* current socket */
5567 };
5568 
5569 struct sk_reuseport_md {
5570 	/*
5571 	 * Start of directly accessible data. It begins from
5572 	 * the tcp/udp header.
5573 	 */
5574 	__bpf_md_ptr(void *, data);
5575 	/* End of directly accessible data */
5576 	__bpf_md_ptr(void *, data_end);
5577 	/*
5578 	 * Total length of packet (starting from the tcp/udp header).
5579 	 * Note that the directly accessible bytes (data_end - data)
5580 	 * could be less than this "len".  Those bytes could be
5581 	 * indirectly read by a helper "bpf_skb_load_bytes()".
5582 	 */
5583 	__u32 len;
5584 	/*
5585 	 * Eth protocol in the mac header (network byte order). e.g.
5586 	 * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
5587 	 */
5588 	__u32 eth_protocol;
5589 	__u32 ip_protocol;	/* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
5590 	__u32 bind_inany;	/* Is sock bound to an INANY address? */
5591 	__u32 hash;		/* A hash of the packet 4 tuples */
5592 	/* When reuse->migrating_sk is NULL, it is selecting a sk for the
5593 	 * new incoming connection request (e.g. selecting a listen sk for
5594 	 * the received SYN in the TCP case).  reuse->sk is one of the sk
5595 	 * in the reuseport group. The bpf prog can use reuse->sk to learn
5596 	 * the local listening ip/port without looking into the skb.
5597 	 *
5598 	 * When reuse->migrating_sk is not NULL, reuse->sk is closed and
5599 	 * reuse->migrating_sk is the socket that needs to be migrated
5600 	 * to another listening socket.  migrating_sk could be a fullsock
5601 	 * sk that is fully established or a reqsk that is in-the-middle
5602 	 * of 3-way handshake.
5603 	 */
5604 	__bpf_md_ptr(struct bpf_sock *, sk);
5605 	__bpf_md_ptr(struct bpf_sock *, migrating_sk);
5606 };
5607 
5608 #define BPF_TAG_SIZE	8
5609 
5610 struct bpf_prog_info {
5611 	__u32 type;
5612 	__u32 id;
5613 	__u8  tag[BPF_TAG_SIZE];
5614 	__u32 jited_prog_len;
5615 	__u32 xlated_prog_len;
5616 	__aligned_u64 jited_prog_insns;
5617 	__aligned_u64 xlated_prog_insns;
5618 	__u64 load_time;	/* ns since boottime */
5619 	__u32 created_by_uid;
5620 	__u32 nr_map_ids;
5621 	__aligned_u64 map_ids;
5622 	char name[BPF_OBJ_NAME_LEN];
5623 	__u32 ifindex;
5624 	__u32 gpl_compatible:1;
5625 	__u32 :31; /* alignment pad */
5626 	__u64 netns_dev;
5627 	__u64 netns_ino;
5628 	__u32 nr_jited_ksyms;
5629 	__u32 nr_jited_func_lens;
5630 	__aligned_u64 jited_ksyms;
5631 	__aligned_u64 jited_func_lens;
5632 	__u32 btf_id;
5633 	__u32 func_info_rec_size;
5634 	__aligned_u64 func_info;
5635 	__u32 nr_func_info;
5636 	__u32 nr_line_info;
5637 	__aligned_u64 line_info;
5638 	__aligned_u64 jited_line_info;
5639 	__u32 nr_jited_line_info;
5640 	__u32 line_info_rec_size;
5641 	__u32 jited_line_info_rec_size;
5642 	__u32 nr_prog_tags;
5643 	__aligned_u64 prog_tags;
5644 	__u64 run_time_ns;
5645 	__u64 run_cnt;
5646 	__u64 recursion_misses;
5647 	__u32 verified_insns;
5648 } __attribute__((aligned(8)));
5649 
5650 struct bpf_map_info {
5651 	__u32 type;
5652 	__u32 id;
5653 	__u32 key_size;
5654 	__u32 value_size;
5655 	__u32 max_entries;
5656 	__u32 map_flags;
5657 	char  name[BPF_OBJ_NAME_LEN];
5658 	__u32 ifindex;
5659 	__u32 btf_vmlinux_value_type_id;
5660 	__u64 netns_dev;
5661 	__u64 netns_ino;
5662 	__u32 btf_id;
5663 	__u32 btf_key_type_id;
5664 	__u32 btf_value_type_id;
5665 	__u32 :32;	/* alignment pad */
5666 	__u64 map_extra;
5667 } __attribute__((aligned(8)));
5668 
5669 struct bpf_btf_info {
5670 	__aligned_u64 btf;
5671 	__u32 btf_size;
5672 	__u32 id;
5673 	__aligned_u64 name;
5674 	__u32 name_len;
5675 	__u32 kernel_btf;
5676 } __attribute__((aligned(8)));
5677 
5678 struct bpf_link_info {
5679 	__u32 type;
5680 	__u32 id;
5681 	__u32 prog_id;
5682 	union {
5683 		struct {
5684 			__aligned_u64 tp_name; /* in/out: tp_name buffer ptr */
5685 			__u32 tp_name_len;     /* in/out: tp_name buffer len */
5686 		} raw_tracepoint;
5687 		struct {
5688 			__u32 attach_type;
5689 			__u32 target_obj_id; /* prog_id for PROG_EXT, otherwise btf object id */
5690 			__u32 target_btf_id; /* BTF type id inside the object */
5691 		} tracing;
5692 		struct {
5693 			__u64 cgroup_id;
5694 			__u32 attach_type;
5695 		} cgroup;
5696 		struct {
5697 			__aligned_u64 target_name; /* in/out: target_name buffer ptr */
5698 			__u32 target_name_len;	   /* in/out: target_name buffer len */
5699 			union {
5700 				struct {
5701 					__u32 map_id;
5702 				} map;
5703 			};
5704 		} iter;
5705 		struct  {
5706 			__u32 netns_ino;
5707 			__u32 attach_type;
5708 		} netns;
5709 		struct {
5710 			__u32 ifindex;
5711 		} xdp;
5712 	};
5713 } __attribute__((aligned(8)));
5714 
5715 /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
5716  * by user and intended to be used by socket (e.g. to bind to, depends on
5717  * attach type).
5718  */
5719 struct bpf_sock_addr {
5720 	__u32 user_family;	/* Allows 4-byte read, but no write. */
5721 	__u32 user_ip4;		/* Allows 1,2,4-byte read and 4-byte write.
5722 				 * Stored in network byte order.
5723 				 */
5724 	__u32 user_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
5725 				 * Stored in network byte order.
5726 				 */
5727 	__u32 user_port;	/* Allows 1,2,4-byte read and 4-byte write.
5728 				 * Stored in network byte order
5729 				 */
5730 	__u32 family;		/* Allows 4-byte read, but no write */
5731 	__u32 type;		/* Allows 4-byte read, but no write */
5732 	__u32 protocol;		/* Allows 4-byte read, but no write */
5733 	__u32 msg_src_ip4;	/* Allows 1,2,4-byte read and 4-byte write.
5734 				 * Stored in network byte order.
5735 				 */
5736 	__u32 msg_src_ip6[4];	/* Allows 1,2,4,8-byte read and 4,8-byte write.
5737 				 * Stored in network byte order.
5738 				 */
5739 	__bpf_md_ptr(struct bpf_sock *, sk);
5740 };
5741 
5742 /* User bpf_sock_ops struct to access socket values and specify request ops
5743  * and their replies.
5744  * Some of this fields are in network (bigendian) byte order and may need
5745  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
5746  * New fields can only be added at the end of this structure
5747  */
5748 struct bpf_sock_ops {
5749 	__u32 op;
5750 	union {
5751 		__u32 args[4];		/* Optionally passed to bpf program */
5752 		__u32 reply;		/* Returned by bpf program	    */
5753 		__u32 replylong[4];	/* Optionally returned by bpf prog  */
5754 	};
5755 	__u32 family;
5756 	__u32 remote_ip4;	/* Stored in network byte order */
5757 	__u32 local_ip4;	/* Stored in network byte order */
5758 	__u32 remote_ip6[4];	/* Stored in network byte order */
5759 	__u32 local_ip6[4];	/* Stored in network byte order */
5760 	__u32 remote_port;	/* Stored in network byte order */
5761 	__u32 local_port;	/* stored in host byte order */
5762 	__u32 is_fullsock;	/* Some TCP fields are only valid if
5763 				 * there is a full socket. If not, the
5764 				 * fields read as zero.
5765 				 */
5766 	__u32 snd_cwnd;
5767 	__u32 srtt_us;		/* Averaged RTT << 3 in usecs */
5768 	__u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
5769 	__u32 state;
5770 	__u32 rtt_min;
5771 	__u32 snd_ssthresh;
5772 	__u32 rcv_nxt;
5773 	__u32 snd_nxt;
5774 	__u32 snd_una;
5775 	__u32 mss_cache;
5776 	__u32 ecn_flags;
5777 	__u32 rate_delivered;
5778 	__u32 rate_interval_us;
5779 	__u32 packets_out;
5780 	__u32 retrans_out;
5781 	__u32 total_retrans;
5782 	__u32 segs_in;
5783 	__u32 data_segs_in;
5784 	__u32 segs_out;
5785 	__u32 data_segs_out;
5786 	__u32 lost_out;
5787 	__u32 sacked_out;
5788 	__u32 sk_txhash;
5789 	__u64 bytes_received;
5790 	__u64 bytes_acked;
5791 	__bpf_md_ptr(struct bpf_sock *, sk);
5792 	/* [skb_data, skb_data_end) covers the whole TCP header.
5793 	 *
5794 	 * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received
5795 	 * BPF_SOCK_OPS_HDR_OPT_LEN_CB:   Not useful because the
5796 	 *                                header has not been written.
5797 	 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have
5798 	 *				  been written so far.
5799 	 * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:  The SYNACK that concludes
5800 	 *					the 3WHS.
5801 	 * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes
5802 	 *					the 3WHS.
5803 	 *
5804 	 * bpf_load_hdr_opt() can also be used to read a particular option.
5805 	 */
5806 	__bpf_md_ptr(void *, skb_data);
5807 	__bpf_md_ptr(void *, skb_data_end);
5808 	__u32 skb_len;		/* The total length of a packet.
5809 				 * It includes the header, options,
5810 				 * and payload.
5811 				 */
5812 	__u32 skb_tcp_flags;	/* tcp_flags of the header.  It provides
5813 				 * an easy way to check for tcp_flags
5814 				 * without parsing skb_data.
5815 				 *
5816 				 * In particular, the skb_tcp_flags
5817 				 * will still be available in
5818 				 * BPF_SOCK_OPS_HDR_OPT_LEN even though
5819 				 * the outgoing header has not
5820 				 * been written yet.
5821 				 */
5822 };
5823 
5824 /* Definitions for bpf_sock_ops_cb_flags */
5825 enum {
5826 	BPF_SOCK_OPS_RTO_CB_FLAG	= (1<<0),
5827 	BPF_SOCK_OPS_RETRANS_CB_FLAG	= (1<<1),
5828 	BPF_SOCK_OPS_STATE_CB_FLAG	= (1<<2),
5829 	BPF_SOCK_OPS_RTT_CB_FLAG	= (1<<3),
5830 	/* Call bpf for all received TCP headers.  The bpf prog will be
5831 	 * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5832 	 *
5833 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5834 	 * for the header option related helpers that will be useful
5835 	 * to the bpf programs.
5836 	 *
5837 	 * It could be used at the client/active side (i.e. connect() side)
5838 	 * when the server told it that the server was in syncookie
5839 	 * mode and required the active side to resend the bpf-written
5840 	 * options.  The active side can keep writing the bpf-options until
5841 	 * it received a valid packet from the server side to confirm
5842 	 * the earlier packet (and options) has been received.  The later
5843 	 * example patch is using it like this at the active side when the
5844 	 * server is in syncookie mode.
5845 	 *
5846 	 * The bpf prog will usually turn this off in the common cases.
5847 	 */
5848 	BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG	= (1<<4),
5849 	/* Call bpf when kernel has received a header option that
5850 	 * the kernel cannot handle.  The bpf prog will be called under
5851 	 * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB.
5852 	 *
5853 	 * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB
5854 	 * for the header option related helpers that will be useful
5855 	 * to the bpf programs.
5856 	 */
5857 	BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5),
5858 	/* Call bpf when the kernel is writing header options for the
5859 	 * outgoing packet.  The bpf prog will first be called
5860 	 * to reserve space in a skb under
5861 	 * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB.  Then
5862 	 * the bpf prog will be called to write the header option(s)
5863 	 * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5864 	 *
5865 	 * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB
5866 	 * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option
5867 	 * related helpers that will be useful to the bpf programs.
5868 	 *
5869 	 * The kernel gets its chance to reserve space and write
5870 	 * options first before the BPF program does.
5871 	 */
5872 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6),
5873 /* Mask of all currently supported cb flags */
5874 	BPF_SOCK_OPS_ALL_CB_FLAGS       = 0x7F,
5875 };
5876 
5877 /* List of known BPF sock_ops operators.
5878  * New entries can only be added at the end
5879  */
5880 enum {
5881 	BPF_SOCK_OPS_VOID,
5882 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
5883 					 * -1 if default value should be used
5884 					 */
5885 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
5886 					 * window (in packets) or -1 if default
5887 					 * value should be used
5888 					 */
5889 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
5890 					 * active connection is initialized
5891 					 */
5892 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
5893 						 * active connection is
5894 						 * established
5895 						 */
5896 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
5897 						 * passive connection is
5898 						 * established
5899 						 */
5900 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
5901 					 * needs ECN
5902 					 */
5903 	BPF_SOCK_OPS_BASE_RTT,		/* Get base RTT. The correct value is
5904 					 * based on the path and may be
5905 					 * dependent on the congestion control
5906 					 * algorithm. In general it indicates
5907 					 * a congestion threshold. RTTs above
5908 					 * this indicate congestion
5909 					 */
5910 	BPF_SOCK_OPS_RTO_CB,		/* Called when an RTO has triggered.
5911 					 * Arg1: value of icsk_retransmits
5912 					 * Arg2: value of icsk_rto
5913 					 * Arg3: whether RTO has expired
5914 					 */
5915 	BPF_SOCK_OPS_RETRANS_CB,	/* Called when skb is retransmitted.
5916 					 * Arg1: sequence number of 1st byte
5917 					 * Arg2: # segments
5918 					 * Arg3: return value of
5919 					 *       tcp_transmit_skb (0 => success)
5920 					 */
5921 	BPF_SOCK_OPS_STATE_CB,		/* Called when TCP changes state.
5922 					 * Arg1: old_state
5923 					 * Arg2: new_state
5924 					 */
5925 	BPF_SOCK_OPS_TCP_LISTEN_CB,	/* Called on listen(2), right after
5926 					 * socket transition to LISTEN state.
5927 					 */
5928 	BPF_SOCK_OPS_RTT_CB,		/* Called on every RTT.
5929 					 */
5930 	BPF_SOCK_OPS_PARSE_HDR_OPT_CB,	/* Parse the header option.
5931 					 * It will be called to handle
5932 					 * the packets received at
5933 					 * an already established
5934 					 * connection.
5935 					 *
5936 					 * sock_ops->skb_data:
5937 					 * Referring to the received skb.
5938 					 * It covers the TCP header only.
5939 					 *
5940 					 * bpf_load_hdr_opt() can also
5941 					 * be used to search for a
5942 					 * particular option.
5943 					 */
5944 	BPF_SOCK_OPS_HDR_OPT_LEN_CB,	/* Reserve space for writing the
5945 					 * header option later in
5946 					 * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
5947 					 * Arg1: bool want_cookie. (in
5948 					 *       writing SYNACK only)
5949 					 *
5950 					 * sock_ops->skb_data:
5951 					 * Not available because no header has
5952 					 * been	written yet.
5953 					 *
5954 					 * sock_ops->skb_tcp_flags:
5955 					 * The tcp_flags of the
5956 					 * outgoing skb. (e.g. SYN, ACK, FIN).
5957 					 *
5958 					 * bpf_reserve_hdr_opt() should
5959 					 * be used to reserve space.
5960 					 */
5961 	BPF_SOCK_OPS_WRITE_HDR_OPT_CB,	/* Write the header options
5962 					 * Arg1: bool want_cookie. (in
5963 					 *       writing SYNACK only)
5964 					 *
5965 					 * sock_ops->skb_data:
5966 					 * Referring to the outgoing skb.
5967 					 * It covers the TCP header
5968 					 * that has already been written
5969 					 * by the kernel and the
5970 					 * earlier bpf-progs.
5971 					 *
5972 					 * sock_ops->skb_tcp_flags:
5973 					 * The tcp_flags of the outgoing
5974 					 * skb. (e.g. SYN, ACK, FIN).
5975 					 *
5976 					 * bpf_store_hdr_opt() should
5977 					 * be used to write the
5978 					 * option.
5979 					 *
5980 					 * bpf_load_hdr_opt() can also
5981 					 * be used to search for a
5982 					 * particular option that
5983 					 * has already been written
5984 					 * by the kernel or the
5985 					 * earlier bpf-progs.
5986 					 */
5987 };
5988 
5989 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
5990  * changes between the TCP and BPF versions. Ideally this should never happen.
5991  * If it does, we need to add code to convert them before calling
5992  * the BPF sock_ops function.
5993  */
5994 enum {
5995 	BPF_TCP_ESTABLISHED = 1,
5996 	BPF_TCP_SYN_SENT,
5997 	BPF_TCP_SYN_RECV,
5998 	BPF_TCP_FIN_WAIT1,
5999 	BPF_TCP_FIN_WAIT2,
6000 	BPF_TCP_TIME_WAIT,
6001 	BPF_TCP_CLOSE,
6002 	BPF_TCP_CLOSE_WAIT,
6003 	BPF_TCP_LAST_ACK,
6004 	BPF_TCP_LISTEN,
6005 	BPF_TCP_CLOSING,	/* Now a valid state */
6006 	BPF_TCP_NEW_SYN_RECV,
6007 
6008 	BPF_TCP_MAX_STATES	/* Leave at the end! */
6009 };
6010 
6011 enum {
6012 	TCP_BPF_IW		= 1001,	/* Set TCP initial congestion window */
6013 	TCP_BPF_SNDCWND_CLAMP	= 1002,	/* Set sndcwnd_clamp */
6014 	TCP_BPF_DELACK_MAX	= 1003, /* Max delay ack in usecs */
6015 	TCP_BPF_RTO_MIN		= 1004, /* Min delay ack in usecs */
6016 	/* Copy the SYN pkt to optval
6017 	 *
6018 	 * BPF_PROG_TYPE_SOCK_OPS only.  It is similar to the
6019 	 * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit
6020 	 * to only getting from the saved_syn.  It can either get the
6021 	 * syn packet from:
6022 	 *
6023 	 * 1. the just-received SYN packet (only available when writing the
6024 	 *    SYNACK).  It will be useful when it is not necessary to
6025 	 *    save the SYN packet for latter use.  It is also the only way
6026 	 *    to get the SYN during syncookie mode because the syn
6027 	 *    packet cannot be saved during syncookie.
6028 	 *
6029 	 * OR
6030 	 *
6031 	 * 2. the earlier saved syn which was done by
6032 	 *    bpf_setsockopt(TCP_SAVE_SYN).
6033 	 *
6034 	 * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the
6035 	 * SYN packet is obtained.
6036 	 *
6037 	 * If the bpf-prog does not need the IP[46] header,  the
6038 	 * bpf-prog can avoid parsing the IP header by using
6039 	 * TCP_BPF_SYN.  Otherwise, the bpf-prog can get both
6040 	 * IP[46] and TCP header by using TCP_BPF_SYN_IP.
6041 	 *
6042 	 *      >0: Total number of bytes copied
6043 	 * -ENOSPC: Not enough space in optval. Only optlen number of
6044 	 *          bytes is copied.
6045 	 * -ENOENT: The SYN skb is not available now and the earlier SYN pkt
6046 	 *	    is not saved by setsockopt(TCP_SAVE_SYN).
6047 	 */
6048 	TCP_BPF_SYN		= 1005, /* Copy the TCP header */
6049 	TCP_BPF_SYN_IP		= 1006, /* Copy the IP[46] and TCP header */
6050 	TCP_BPF_SYN_MAC         = 1007, /* Copy the MAC, IP[46], and TCP header */
6051 };
6052 
6053 enum {
6054 	BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0),
6055 };
6056 
6057 /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and
6058  * BPF_SOCK_OPS_WRITE_HDR_OPT_CB.
6059  */
6060 enum {
6061 	BPF_WRITE_HDR_TCP_CURRENT_MSS = 1,	/* Kernel is finding the
6062 						 * total option spaces
6063 						 * required for an established
6064 						 * sk in order to calculate the
6065 						 * MSS.  No skb is actually
6066 						 * sent.
6067 						 */
6068 	BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2,	/* Kernel is in syncookie mode
6069 						 * when sending a SYN.
6070 						 */
6071 };
6072 
6073 struct bpf_perf_event_value {
6074 	__u64 counter;
6075 	__u64 enabled;
6076 	__u64 running;
6077 };
6078 
6079 enum {
6080 	BPF_DEVCG_ACC_MKNOD	= (1ULL << 0),
6081 	BPF_DEVCG_ACC_READ	= (1ULL << 1),
6082 	BPF_DEVCG_ACC_WRITE	= (1ULL << 2),
6083 };
6084 
6085 enum {
6086 	BPF_DEVCG_DEV_BLOCK	= (1ULL << 0),
6087 	BPF_DEVCG_DEV_CHAR	= (1ULL << 1),
6088 };
6089 
6090 struct bpf_cgroup_dev_ctx {
6091 	/* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
6092 	__u32 access_type;
6093 	__u32 major;
6094 	__u32 minor;
6095 };
6096 
6097 struct bpf_raw_tracepoint_args {
6098 	__u64 args[0];
6099 };
6100 
6101 /* DIRECT:  Skip the FIB rules and go to FIB table associated with device
6102  * OUTPUT:  Do lookup from egress perspective; default is ingress
6103  */
6104 enum {
6105 	BPF_FIB_LOOKUP_DIRECT  = (1U << 0),
6106 	BPF_FIB_LOOKUP_OUTPUT  = (1U << 1),
6107 };
6108 
6109 enum {
6110 	BPF_FIB_LKUP_RET_SUCCESS,      /* lookup successful */
6111 	BPF_FIB_LKUP_RET_BLACKHOLE,    /* dest is blackholed; can be dropped */
6112 	BPF_FIB_LKUP_RET_UNREACHABLE,  /* dest is unreachable; can be dropped */
6113 	BPF_FIB_LKUP_RET_PROHIBIT,     /* dest not allowed; can be dropped */
6114 	BPF_FIB_LKUP_RET_NOT_FWDED,    /* packet is not forwarded */
6115 	BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
6116 	BPF_FIB_LKUP_RET_UNSUPP_LWT,   /* fwd requires encapsulation */
6117 	BPF_FIB_LKUP_RET_NO_NEIGH,     /* no neighbor entry for nh */
6118 	BPF_FIB_LKUP_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6119 };
6120 
6121 struct bpf_fib_lookup {
6122 	/* input:  network family for lookup (AF_INET, AF_INET6)
6123 	 * output: network family of egress nexthop
6124 	 */
6125 	__u8	family;
6126 
6127 	/* set if lookup is to consider L4 data - e.g., FIB rules */
6128 	__u8	l4_protocol;
6129 	__be16	sport;
6130 	__be16	dport;
6131 
6132 	union {	/* used for MTU check */
6133 		/* input to lookup */
6134 		__u16	tot_len; /* L3 length from network hdr (iph->tot_len) */
6135 
6136 		/* output: MTU value */
6137 		__u16	mtu_result;
6138 	};
6139 	/* input: L3 device index for lookup
6140 	 * output: device index from FIB lookup
6141 	 */
6142 	__u32	ifindex;
6143 
6144 	union {
6145 		/* inputs to lookup */
6146 		__u8	tos;		/* AF_INET  */
6147 		__be32	flowinfo;	/* AF_INET6, flow_label + priority */
6148 
6149 		/* output: metric of fib result (IPv4/IPv6 only) */
6150 		__u32	rt_metric;
6151 	};
6152 
6153 	union {
6154 		__be32		ipv4_src;
6155 		__u32		ipv6_src[4];  /* in6_addr; network order */
6156 	};
6157 
6158 	/* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
6159 	 * network header. output: bpf_fib_lookup sets to gateway address
6160 	 * if FIB lookup returns gateway route
6161 	 */
6162 	union {
6163 		__be32		ipv4_dst;
6164 		__u32		ipv6_dst[4];  /* in6_addr; network order */
6165 	};
6166 
6167 	/* output */
6168 	__be16	h_vlan_proto;
6169 	__be16	h_vlan_TCI;
6170 	__u8	smac[6];     /* ETH_ALEN */
6171 	__u8	dmac[6];     /* ETH_ALEN */
6172 };
6173 
6174 struct bpf_redir_neigh {
6175 	/* network family for lookup (AF_INET, AF_INET6) */
6176 	__u32 nh_family;
6177 	/* network address of nexthop; skips fib lookup to find gateway */
6178 	union {
6179 		__be32		ipv4_nh;
6180 		__u32		ipv6_nh[4];  /* in6_addr; network order */
6181 	};
6182 };
6183 
6184 /* bpf_check_mtu flags*/
6185 enum  bpf_check_mtu_flags {
6186 	BPF_MTU_CHK_SEGS  = (1U << 0),
6187 };
6188 
6189 enum bpf_check_mtu_ret {
6190 	BPF_MTU_CHK_RET_SUCCESS,      /* check and lookup successful */
6191 	BPF_MTU_CHK_RET_FRAG_NEEDED,  /* fragmentation required to fwd */
6192 	BPF_MTU_CHK_RET_SEGS_TOOBIG,  /* GSO re-segmentation needed to fwd */
6193 };
6194 
6195 enum bpf_task_fd_type {
6196 	BPF_FD_TYPE_RAW_TRACEPOINT,	/* tp name */
6197 	BPF_FD_TYPE_TRACEPOINT,		/* tp name */
6198 	BPF_FD_TYPE_KPROBE,		/* (symbol + offset) or addr */
6199 	BPF_FD_TYPE_KRETPROBE,		/* (symbol + offset) or addr */
6200 	BPF_FD_TYPE_UPROBE,		/* filename + offset */
6201 	BPF_FD_TYPE_URETPROBE,		/* filename + offset */
6202 };
6203 
6204 enum {
6205 	BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG		= (1U << 0),
6206 	BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL		= (1U << 1),
6207 	BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP		= (1U << 2),
6208 };
6209 
6210 struct bpf_flow_keys {
6211 	__u16	nhoff;
6212 	__u16	thoff;
6213 	__u16	addr_proto;			/* ETH_P_* of valid addrs */
6214 	__u8	is_frag;
6215 	__u8	is_first_frag;
6216 	__u8	is_encap;
6217 	__u8	ip_proto;
6218 	__be16	n_proto;
6219 	__be16	sport;
6220 	__be16	dport;
6221 	union {
6222 		struct {
6223 			__be32	ipv4_src;
6224 			__be32	ipv4_dst;
6225 		};
6226 		struct {
6227 			__u32	ipv6_src[4];	/* in6_addr; network order */
6228 			__u32	ipv6_dst[4];	/* in6_addr; network order */
6229 		};
6230 	};
6231 	__u32	flags;
6232 	__be32	flow_label;
6233 };
6234 
6235 struct bpf_func_info {
6236 	__u32	insn_off;
6237 	__u32	type_id;
6238 };
6239 
6240 #define BPF_LINE_INFO_LINE_NUM(line_col)	((line_col) >> 10)
6241 #define BPF_LINE_INFO_LINE_COL(line_col)	((line_col) & 0x3ff)
6242 
6243 struct bpf_line_info {
6244 	__u32	insn_off;
6245 	__u32	file_name_off;
6246 	__u32	line_off;
6247 	__u32	line_col;
6248 };
6249 
6250 struct bpf_spin_lock {
6251 	__u32	val;
6252 };
6253 
6254 struct bpf_timer {
6255 	__u64 :64;
6256 	__u64 :64;
6257 } __attribute__((aligned(8)));
6258 
6259 struct bpf_sysctl {
6260 	__u32	write;		/* Sysctl is being read (= 0) or written (= 1).
6261 				 * Allows 1,2,4-byte read, but no write.
6262 				 */
6263 	__u32	file_pos;	/* Sysctl file position to read from, write to.
6264 				 * Allows 1,2,4-byte read an 4-byte write.
6265 				 */
6266 };
6267 
6268 struct bpf_sockopt {
6269 	__bpf_md_ptr(struct bpf_sock *, sk);
6270 	__bpf_md_ptr(void *, optval);
6271 	__bpf_md_ptr(void *, optval_end);
6272 
6273 	__s32	level;
6274 	__s32	optname;
6275 	__s32	optlen;
6276 	__s32	retval;
6277 };
6278 
6279 struct bpf_pidns_info {
6280 	__u32 pid;
6281 	__u32 tgid;
6282 };
6283 
6284 /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */
6285 struct bpf_sk_lookup {
6286 	union {
6287 		__bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */
6288 		__u64 cookie; /* Non-zero if socket was selected in PROG_TEST_RUN */
6289 	};
6290 
6291 	__u32 family;		/* Protocol family (AF_INET, AF_INET6) */
6292 	__u32 protocol;		/* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */
6293 	__u32 remote_ip4;	/* Network byte order */
6294 	__u32 remote_ip6[4];	/* Network byte order */
6295 	__u32 remote_port;	/* Network byte order */
6296 	__u32 local_ip4;	/* Network byte order */
6297 	__u32 local_ip6[4];	/* Network byte order */
6298 	__u32 local_port;	/* Host byte order */
6299 };
6300 
6301 /*
6302  * struct btf_ptr is used for typed pointer representation; the
6303  * type id is used to render the pointer data as the appropriate type
6304  * via the bpf_snprintf_btf() helper described above.  A flags field -
6305  * potentially to specify additional details about the BTF pointer
6306  * (rather than its mode of display) - is included for future use.
6307  * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately.
6308  */
6309 struct btf_ptr {
6310 	void *ptr;
6311 	__u32 type_id;
6312 	__u32 flags;		/* BTF ptr flags; unused at present. */
6313 };
6314 
6315 /*
6316  * Flags to control bpf_snprintf_btf() behaviour.
6317  *     - BTF_F_COMPACT: no formatting around type information
6318  *     - BTF_F_NONAME: no struct/union member names/types
6319  *     - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values;
6320  *       equivalent to %px.
6321  *     - BTF_F_ZERO: show zero-valued struct/union members; they
6322  *       are not displayed by default
6323  */
6324 enum {
6325 	BTF_F_COMPACT	=	(1ULL << 0),
6326 	BTF_F_NONAME	=	(1ULL << 1),
6327 	BTF_F_PTR_RAW	=	(1ULL << 2),
6328 	BTF_F_ZERO	=	(1ULL << 3),
6329 };
6330 
6331 #endif /* _UAPI__LINUX_BPF_H__ */
6332