xref: /linux-6.15/tools/include/uapi/linux/bpf.h (revision f26de110)
1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
2  *
3  * This program is free software; you can redistribute it and/or
4  * modify it under the terms of version 2 of the GNU General Public
5  * License as published by the Free Software Foundation.
6  */
7 #ifndef _UAPI__LINUX_BPF_H__
8 #define _UAPI__LINUX_BPF_H__
9 
10 #include <linux/types.h>
11 #include <linux/bpf_common.h>
12 
13 /* Extended instruction set based on top of classic BPF */
14 
15 /* instruction classes */
16 #define BPF_ALU64	0x07	/* alu mode in double word width */
17 
18 /* ld/ldx fields */
19 #define BPF_DW		0x18	/* double word */
20 #define BPF_XADD	0xc0	/* exclusive add */
21 
22 /* alu/jmp fields */
23 #define BPF_MOV		0xb0	/* mov reg to reg */
24 #define BPF_ARSH	0xc0	/* sign extending arithmetic shift right */
25 
26 /* change endianness of a register */
27 #define BPF_END		0xd0	/* flags for endianness conversion: */
28 #define BPF_TO_LE	0x00	/* convert to little-endian */
29 #define BPF_TO_BE	0x08	/* convert to big-endian */
30 #define BPF_FROM_LE	BPF_TO_LE
31 #define BPF_FROM_BE	BPF_TO_BE
32 
33 /* jmp encodings */
34 #define BPF_JNE		0x50	/* jump != */
35 #define BPF_JLT		0xa0	/* LT is unsigned, '<' */
36 #define BPF_JLE		0xb0	/* LE is unsigned, '<=' */
37 #define BPF_JSGT	0x60	/* SGT is signed '>', GT in x86 */
38 #define BPF_JSGE	0x70	/* SGE is signed '>=', GE in x86 */
39 #define BPF_JSLT	0xc0	/* SLT is signed, '<' */
40 #define BPF_JSLE	0xd0	/* SLE is signed, '<=' */
41 #define BPF_CALL	0x80	/* function call */
42 #define BPF_EXIT	0x90	/* function return */
43 
44 /* Register numbers */
45 enum {
46 	BPF_REG_0 = 0,
47 	BPF_REG_1,
48 	BPF_REG_2,
49 	BPF_REG_3,
50 	BPF_REG_4,
51 	BPF_REG_5,
52 	BPF_REG_6,
53 	BPF_REG_7,
54 	BPF_REG_8,
55 	BPF_REG_9,
56 	BPF_REG_10,
57 	__MAX_BPF_REG,
58 };
59 
60 /* BPF has 10 general purpose 64-bit registers and stack frame. */
61 #define MAX_BPF_REG	__MAX_BPF_REG
62 
63 struct bpf_insn {
64 	__u8	code;		/* opcode */
65 	__u8	dst_reg:4;	/* dest register */
66 	__u8	src_reg:4;	/* source register */
67 	__s16	off;		/* signed offset */
68 	__s32	imm;		/* signed immediate constant */
69 };
70 
71 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
72 struct bpf_lpm_trie_key {
73 	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
74 	__u8	data[0];	/* Arbitrary size */
75 };
76 
77 /* BPF syscall commands, see bpf(2) man-page for details. */
78 enum bpf_cmd {
79 	BPF_MAP_CREATE,
80 	BPF_MAP_LOOKUP_ELEM,
81 	BPF_MAP_UPDATE_ELEM,
82 	BPF_MAP_DELETE_ELEM,
83 	BPF_MAP_GET_NEXT_KEY,
84 	BPF_PROG_LOAD,
85 	BPF_OBJ_PIN,
86 	BPF_OBJ_GET,
87 	BPF_PROG_ATTACH,
88 	BPF_PROG_DETACH,
89 	BPF_PROG_TEST_RUN,
90 	BPF_PROG_GET_NEXT_ID,
91 	BPF_MAP_GET_NEXT_ID,
92 	BPF_PROG_GET_FD_BY_ID,
93 	BPF_MAP_GET_FD_BY_ID,
94 	BPF_OBJ_GET_INFO_BY_FD,
95 };
96 
97 enum bpf_map_type {
98 	BPF_MAP_TYPE_UNSPEC,
99 	BPF_MAP_TYPE_HASH,
100 	BPF_MAP_TYPE_ARRAY,
101 	BPF_MAP_TYPE_PROG_ARRAY,
102 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
103 	BPF_MAP_TYPE_PERCPU_HASH,
104 	BPF_MAP_TYPE_PERCPU_ARRAY,
105 	BPF_MAP_TYPE_STACK_TRACE,
106 	BPF_MAP_TYPE_CGROUP_ARRAY,
107 	BPF_MAP_TYPE_LRU_HASH,
108 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
109 	BPF_MAP_TYPE_LPM_TRIE,
110 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
111 	BPF_MAP_TYPE_HASH_OF_MAPS,
112 	BPF_MAP_TYPE_DEVMAP,
113 };
114 
115 enum bpf_prog_type {
116 	BPF_PROG_TYPE_UNSPEC,
117 	BPF_PROG_TYPE_SOCKET_FILTER,
118 	BPF_PROG_TYPE_KPROBE,
119 	BPF_PROG_TYPE_SCHED_CLS,
120 	BPF_PROG_TYPE_SCHED_ACT,
121 	BPF_PROG_TYPE_TRACEPOINT,
122 	BPF_PROG_TYPE_XDP,
123 	BPF_PROG_TYPE_PERF_EVENT,
124 	BPF_PROG_TYPE_CGROUP_SKB,
125 	BPF_PROG_TYPE_CGROUP_SOCK,
126 	BPF_PROG_TYPE_LWT_IN,
127 	BPF_PROG_TYPE_LWT_OUT,
128 	BPF_PROG_TYPE_LWT_XMIT,
129 	BPF_PROG_TYPE_SOCK_OPS,
130 };
131 
132 enum bpf_attach_type {
133 	BPF_CGROUP_INET_INGRESS,
134 	BPF_CGROUP_INET_EGRESS,
135 	BPF_CGROUP_INET_SOCK_CREATE,
136 	BPF_CGROUP_SOCK_OPS,
137 	__MAX_BPF_ATTACH_TYPE
138 };
139 
140 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
141 
142 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
143  * to the given target_fd cgroup the descendent cgroup will be able to
144  * override effective bpf program that was inherited from this cgroup
145  */
146 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
147 
148 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
149  * verifier will perform strict alignment checking as if the kernel
150  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
151  * and NET_IP_ALIGN defined to 2.
152  */
153 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
154 
155 #define BPF_PSEUDO_MAP_FD	1
156 
157 /* flags for BPF_MAP_UPDATE_ELEM command */
158 #define BPF_ANY		0 /* create new element or update existing */
159 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
160 #define BPF_EXIST	2 /* update existing element */
161 
162 #define BPF_F_NO_PREALLOC	(1U << 0)
163 /* Instead of having one common LRU list in the
164  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
165  * which can scale and perform better.
166  * Note, the LRU nodes (including free nodes) cannot be moved
167  * across different LRU lists.
168  */
169 #define BPF_F_NO_COMMON_LRU	(1U << 1)
170 
171 union bpf_attr {
172 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
173 		__u32	map_type;	/* one of enum bpf_map_type */
174 		__u32	key_size;	/* size of key in bytes */
175 		__u32	value_size;	/* size of value in bytes */
176 		__u32	max_entries;	/* max number of entries in a map */
177 		__u32	map_flags;	/* prealloc or not */
178 		__u32	inner_map_fd;	/* fd pointing to the inner map */
179 	};
180 
181 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
182 		__u32		map_fd;
183 		__aligned_u64	key;
184 		union {
185 			__aligned_u64 value;
186 			__aligned_u64 next_key;
187 		};
188 		__u64		flags;
189 	};
190 
191 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
192 		__u32		prog_type;	/* one of enum bpf_prog_type */
193 		__u32		insn_cnt;
194 		__aligned_u64	insns;
195 		__aligned_u64	license;
196 		__u32		log_level;	/* verbosity level of verifier */
197 		__u32		log_size;	/* size of user buffer */
198 		__aligned_u64	log_buf;	/* user supplied buffer */
199 		__u32		kern_version;	/* checked when prog_type=kprobe */
200 		__u32		prog_flags;
201 	};
202 
203 	struct { /* anonymous struct used by BPF_OBJ_* commands */
204 		__aligned_u64	pathname;
205 		__u32		bpf_fd;
206 	};
207 
208 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
209 		__u32		target_fd;	/* container object to attach to */
210 		__u32		attach_bpf_fd;	/* eBPF program to attach */
211 		__u32		attach_type;
212 		__u32		attach_flags;
213 	};
214 
215 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
216 		__u32		prog_fd;
217 		__u32		retval;
218 		__u32		data_size_in;
219 		__u32		data_size_out;
220 		__aligned_u64	data_in;
221 		__aligned_u64	data_out;
222 		__u32		repeat;
223 		__u32		duration;
224 	} test;
225 
226 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
227 		union {
228 			__u32		start_id;
229 			__u32		prog_id;
230 			__u32		map_id;
231 		};
232 		__u32		next_id;
233 	};
234 
235 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
236 		__u32		bpf_fd;
237 		__u32		info_len;
238 		__aligned_u64	info;
239 	} info;
240 } __attribute__((aligned(8)));
241 
242 /* BPF helper function descriptions:
243  *
244  * void *bpf_map_lookup_elem(&map, &key)
245  *     Return: Map value or NULL
246  *
247  * int bpf_map_update_elem(&map, &key, &value, flags)
248  *     Return: 0 on success or negative error
249  *
250  * int bpf_map_delete_elem(&map, &key)
251  *     Return: 0 on success or negative error
252  *
253  * int bpf_probe_read(void *dst, int size, void *src)
254  *     Return: 0 on success or negative error
255  *
256  * u64 bpf_ktime_get_ns(void)
257  *     Return: current ktime
258  *
259  * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
260  *     Return: length of buffer written or negative error
261  *
262  * u32 bpf_prandom_u32(void)
263  *     Return: random value
264  *
265  * u32 bpf_raw_smp_processor_id(void)
266  *     Return: SMP processor ID
267  *
268  * int bpf_skb_store_bytes(skb, offset, from, len, flags)
269  *     store bytes into packet
270  *     @skb: pointer to skb
271  *     @offset: offset within packet from skb->mac_header
272  *     @from: pointer where to copy bytes from
273  *     @len: number of bytes to store into packet
274  *     @flags: bit 0 - if true, recompute skb->csum
275  *             other bits - reserved
276  *     Return: 0 on success or negative error
277  *
278  * int bpf_l3_csum_replace(skb, offset, from, to, flags)
279  *     recompute IP checksum
280  *     @skb: pointer to skb
281  *     @offset: offset within packet where IP checksum is located
282  *     @from: old value of header field
283  *     @to: new value of header field
284  *     @flags: bits 0-3 - size of header field
285  *             other bits - reserved
286  *     Return: 0 on success or negative error
287  *
288  * int bpf_l4_csum_replace(skb, offset, from, to, flags)
289  *     recompute TCP/UDP checksum
290  *     @skb: pointer to skb
291  *     @offset: offset within packet where TCP/UDP checksum is located
292  *     @from: old value of header field
293  *     @to: new value of header field
294  *     @flags: bits 0-3 - size of header field
295  *             bit 4 - is pseudo header
296  *             other bits - reserved
297  *     Return: 0 on success or negative error
298  *
299  * int bpf_tail_call(ctx, prog_array_map, index)
300  *     jump into another BPF program
301  *     @ctx: context pointer passed to next program
302  *     @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
303  *     @index: index inside array that selects specific program to run
304  *     Return: 0 on success or negative error
305  *
306  * int bpf_clone_redirect(skb, ifindex, flags)
307  *     redirect to another netdev
308  *     @skb: pointer to skb
309  *     @ifindex: ifindex of the net device
310  *     @flags: bit 0 - if set, redirect to ingress instead of egress
311  *             other bits - reserved
312  *     Return: 0 on success or negative error
313  *
314  * u64 bpf_get_current_pid_tgid(void)
315  *     Return: current->tgid << 32 | current->pid
316  *
317  * u64 bpf_get_current_uid_gid(void)
318  *     Return: current_gid << 32 | current_uid
319  *
320  * int bpf_get_current_comm(char *buf, int size_of_buf)
321  *     stores current->comm into buf
322  *     Return: 0 on success or negative error
323  *
324  * u32 bpf_get_cgroup_classid(skb)
325  *     retrieve a proc's classid
326  *     @skb: pointer to skb
327  *     Return: classid if != 0
328  *
329  * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
330  *     Return: 0 on success or negative error
331  *
332  * int bpf_skb_vlan_pop(skb)
333  *     Return: 0 on success or negative error
334  *
335  * int bpf_skb_get_tunnel_key(skb, key, size, flags)
336  * int bpf_skb_set_tunnel_key(skb, key, size, flags)
337  *     retrieve or populate tunnel metadata
338  *     @skb: pointer to skb
339  *     @key: pointer to 'struct bpf_tunnel_key'
340  *     @size: size of 'struct bpf_tunnel_key'
341  *     @flags: room for future extensions
342  *     Return: 0 on success or negative error
343  *
344  * u64 bpf_perf_event_read(map, flags)
345  *     read perf event counter value
346  *     @map: pointer to perf_event_array map
347  *     @flags: index of event in the map or bitmask flags
348  *     Return: value of perf event counter read or error code
349  *
350  * int bpf_redirect(ifindex, flags)
351  *     redirect to another netdev
352  *     @ifindex: ifindex of the net device
353  *     @flags: bit 0 - if set, redirect to ingress instead of egress
354  *             other bits - reserved
355  *     Return: TC_ACT_REDIRECT
356  *
357  * u32 bpf_get_route_realm(skb)
358  *     retrieve a dst's tclassid
359  *     @skb: pointer to skb
360  *     Return: realm if != 0
361  *
362  * int bpf_perf_event_output(ctx, map, flags, data, size)
363  *     output perf raw sample
364  *     @ctx: struct pt_regs*
365  *     @map: pointer to perf_event_array map
366  *     @flags: index of event in the map or bitmask flags
367  *     @data: data on stack to be output as raw data
368  *     @size: size of data
369  *     Return: 0 on success or negative error
370  *
371  * int bpf_get_stackid(ctx, map, flags)
372  *     walk user or kernel stack and return id
373  *     @ctx: struct pt_regs*
374  *     @map: pointer to stack_trace map
375  *     @flags: bits 0-7 - numer of stack frames to skip
376  *             bit 8 - collect user stack instead of kernel
377  *             bit 9 - compare stacks by hash only
378  *             bit 10 - if two different stacks hash into the same stackid
379  *                      discard old
380  *             other bits - reserved
381  *     Return: >= 0 stackid on success or negative error
382  *
383  * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
384  *     calculate csum diff
385  *     @from: raw from buffer
386  *     @from_size: length of from buffer
387  *     @to: raw to buffer
388  *     @to_size: length of to buffer
389  *     @seed: optional seed
390  *     Return: csum result or negative error code
391  *
392  * int bpf_skb_get_tunnel_opt(skb, opt, size)
393  *     retrieve tunnel options metadata
394  *     @skb: pointer to skb
395  *     @opt: pointer to raw tunnel option data
396  *     @size: size of @opt
397  *     Return: option size
398  *
399  * int bpf_skb_set_tunnel_opt(skb, opt, size)
400  *     populate tunnel options metadata
401  *     @skb: pointer to skb
402  *     @opt: pointer to raw tunnel option data
403  *     @size: size of @opt
404  *     Return: 0 on success or negative error
405  *
406  * int bpf_skb_change_proto(skb, proto, flags)
407  *     Change protocol of the skb. Currently supported is v4 -> v6,
408  *     v6 -> v4 transitions. The helper will also resize the skb. eBPF
409  *     program is expected to fill the new headers via skb_store_bytes
410  *     and lX_csum_replace.
411  *     @skb: pointer to skb
412  *     @proto: new skb->protocol type
413  *     @flags: reserved
414  *     Return: 0 on success or negative error
415  *
416  * int bpf_skb_change_type(skb, type)
417  *     Change packet type of skb.
418  *     @skb: pointer to skb
419  *     @type: new skb->pkt_type type
420  *     Return: 0 on success or negative error
421  *
422  * int bpf_skb_under_cgroup(skb, map, index)
423  *     Check cgroup2 membership of skb
424  *     @skb: pointer to skb
425  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
426  *     @index: index of the cgroup in the bpf_map
427  *     Return:
428  *       == 0 skb failed the cgroup2 descendant test
429  *       == 1 skb succeeded the cgroup2 descendant test
430  *        < 0 error
431  *
432  * u32 bpf_get_hash_recalc(skb)
433  *     Retrieve and possibly recalculate skb->hash.
434  *     @skb: pointer to skb
435  *     Return: hash
436  *
437  * u64 bpf_get_current_task(void)
438  *     Returns current task_struct
439  *     Return: current
440  *
441  * int bpf_probe_write_user(void *dst, void *src, int len)
442  *     safely attempt to write to a location
443  *     @dst: destination address in userspace
444  *     @src: source address on stack
445  *     @len: number of bytes to copy
446  *     Return: 0 on success or negative error
447  *
448  * int bpf_current_task_under_cgroup(map, index)
449  *     Check cgroup2 membership of current task
450  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
451  *     @index: index of the cgroup in the bpf_map
452  *     Return:
453  *       == 0 current failed the cgroup2 descendant test
454  *       == 1 current succeeded the cgroup2 descendant test
455  *        < 0 error
456  *
457  * int bpf_skb_change_tail(skb, len, flags)
458  *     The helper will resize the skb to the given new size, to be used f.e.
459  *     with control messages.
460  *     @skb: pointer to skb
461  *     @len: new skb length
462  *     @flags: reserved
463  *     Return: 0 on success or negative error
464  *
465  * int bpf_skb_pull_data(skb, len)
466  *     The helper will pull in non-linear data in case the skb is non-linear
467  *     and not all of len are part of the linear section. Only needed for
468  *     read/write with direct packet access.
469  *     @skb: pointer to skb
470  *     @len: len to make read/writeable
471  *     Return: 0 on success or negative error
472  *
473  * s64 bpf_csum_update(skb, csum)
474  *     Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
475  *     @skb: pointer to skb
476  *     @csum: csum to add
477  *     Return: csum on success or negative error
478  *
479  * void bpf_set_hash_invalid(skb)
480  *     Invalidate current skb->hash.
481  *     @skb: pointer to skb
482  *
483  * int bpf_get_numa_node_id()
484  *     Return: Id of current NUMA node.
485  *
486  * int bpf_skb_change_head()
487  *     Grows headroom of skb and adjusts MAC header offset accordingly.
488  *     Will extends/reallocae as required automatically.
489  *     May change skb data pointer and will thus invalidate any check
490  *     performed for direct packet access.
491  *     @skb: pointer to skb
492  *     @len: length of header to be pushed in front
493  *     @flags: Flags (unused for now)
494  *     Return: 0 on success or negative error
495  *
496  * int bpf_xdp_adjust_head(xdp_md, delta)
497  *     Adjust the xdp_md.data by delta
498  *     @xdp_md: pointer to xdp_md
499  *     @delta: An positive/negative integer to be added to xdp_md.data
500  *     Return: 0 on success or negative on error
501  *
502  * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
503  *     Copy a NUL terminated string from unsafe address. In case the string
504  *     length is smaller than size, the target is not padded with further NUL
505  *     bytes. In case the string length is larger than size, just count-1
506  *     bytes are copied and the last byte is set to NUL.
507  *     @dst: destination address
508  *     @size: maximum number of bytes to copy, including the trailing NUL
509  *     @unsafe_ptr: unsafe address
510  *     Return:
511  *       > 0 length of the string including the trailing NUL on success
512  *       < 0 error
513  *
514  * u64 bpf_get_socket_cookie(skb)
515  *     Get the cookie for the socket stored inside sk_buff.
516  *     @skb: pointer to skb
517  *     Return: 8 Bytes non-decreasing number on success or 0 if the socket
518  *     field is missing inside sk_buff
519  *
520  * u32 bpf_get_socket_uid(skb)
521  *     Get the owner uid of the socket stored inside sk_buff.
522  *     @skb: pointer to skb
523  *     Return: uid of the socket owner on success or overflowuid if failed.
524  *
525  * u32 bpf_set_hash(skb, hash)
526  *     Set full skb->hash.
527  *     @skb: pointer to skb
528  *     @hash: hash to set
529  *
530  * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
531  *     Calls setsockopt. Not all opts are available, only those with
532  *     integer optvals plus TCP_CONGESTION.
533  *     Supported levels: SOL_SOCKET and IPROTO_TCP
534  *     @bpf_socket: pointer to bpf_socket
535  *     @level: SOL_SOCKET or IPROTO_TCP
536  *     @optname: option name
537  *     @optval: pointer to option value
538  *     @optlen: length of optval in byes
539  *     Return: 0 or negative error
540  *
541  * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
542  *     Grow or shrink room in sk_buff.
543  *     @skb: pointer to skb
544  *     @len_diff: (signed) amount of room to grow/shrink
545  *     @mode: operation mode (enum bpf_adj_room_mode)
546  *     @flags: reserved for future use
547  *     Return: 0 on success or negative error code
548  */
549 #define __BPF_FUNC_MAPPER(FN)		\
550 	FN(unspec),			\
551 	FN(map_lookup_elem),		\
552 	FN(map_update_elem),		\
553 	FN(map_delete_elem),		\
554 	FN(probe_read),			\
555 	FN(ktime_get_ns),		\
556 	FN(trace_printk),		\
557 	FN(get_prandom_u32),		\
558 	FN(get_smp_processor_id),	\
559 	FN(skb_store_bytes),		\
560 	FN(l3_csum_replace),		\
561 	FN(l4_csum_replace),		\
562 	FN(tail_call),			\
563 	FN(clone_redirect),		\
564 	FN(get_current_pid_tgid),	\
565 	FN(get_current_uid_gid),	\
566 	FN(get_current_comm),		\
567 	FN(get_cgroup_classid),		\
568 	FN(skb_vlan_push),		\
569 	FN(skb_vlan_pop),		\
570 	FN(skb_get_tunnel_key),		\
571 	FN(skb_set_tunnel_key),		\
572 	FN(perf_event_read),		\
573 	FN(redirect),			\
574 	FN(get_route_realm),		\
575 	FN(perf_event_output),		\
576 	FN(skb_load_bytes),		\
577 	FN(get_stackid),		\
578 	FN(csum_diff),			\
579 	FN(skb_get_tunnel_opt),		\
580 	FN(skb_set_tunnel_opt),		\
581 	FN(skb_change_proto),		\
582 	FN(skb_change_type),		\
583 	FN(skb_under_cgroup),		\
584 	FN(get_hash_recalc),		\
585 	FN(get_current_task),		\
586 	FN(probe_write_user),		\
587 	FN(current_task_under_cgroup),	\
588 	FN(skb_change_tail),		\
589 	FN(skb_pull_data),		\
590 	FN(csum_update),		\
591 	FN(set_hash_invalid),		\
592 	FN(get_numa_node_id),		\
593 	FN(skb_change_head),		\
594 	FN(xdp_adjust_head),		\
595 	FN(probe_read_str),		\
596 	FN(get_socket_cookie),		\
597 	FN(get_socket_uid),		\
598 	FN(set_hash),			\
599 	FN(setsockopt),			\
600 	FN(skb_adjust_room),		\
601 	FN(redirect_map),
602 
603 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
604  * function eBPF program intends to call
605  */
606 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
607 enum bpf_func_id {
608 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
609 	__BPF_FUNC_MAX_ID,
610 };
611 #undef __BPF_ENUM_FN
612 
613 /* All flags used by eBPF helper functions, placed here. */
614 
615 /* BPF_FUNC_skb_store_bytes flags. */
616 #define BPF_F_RECOMPUTE_CSUM		(1ULL << 0)
617 #define BPF_F_INVALIDATE_HASH		(1ULL << 1)
618 
619 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
620  * First 4 bits are for passing the header field size.
621  */
622 #define BPF_F_HDR_FIELD_MASK		0xfULL
623 
624 /* BPF_FUNC_l4_csum_replace flags. */
625 #define BPF_F_PSEUDO_HDR		(1ULL << 4)
626 #define BPF_F_MARK_MANGLED_0		(1ULL << 5)
627 #define BPF_F_MARK_ENFORCE		(1ULL << 6)
628 
629 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
630 #define BPF_F_INGRESS			(1ULL << 0)
631 
632 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
633 #define BPF_F_TUNINFO_IPV6		(1ULL << 0)
634 
635 /* BPF_FUNC_get_stackid flags. */
636 #define BPF_F_SKIP_FIELD_MASK		0xffULL
637 #define BPF_F_USER_STACK		(1ULL << 8)
638 #define BPF_F_FAST_STACK_CMP		(1ULL << 9)
639 #define BPF_F_REUSE_STACKID		(1ULL << 10)
640 
641 /* BPF_FUNC_skb_set_tunnel_key flags. */
642 #define BPF_F_ZERO_CSUM_TX		(1ULL << 1)
643 #define BPF_F_DONT_FRAGMENT		(1ULL << 2)
644 
645 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
646 #define BPF_F_INDEX_MASK		0xffffffffULL
647 #define BPF_F_CURRENT_CPU		BPF_F_INDEX_MASK
648 /* BPF_FUNC_perf_event_output for sk_buff input context. */
649 #define BPF_F_CTXLEN_MASK		(0xfffffULL << 32)
650 
651 /* Mode for BPF_FUNC_skb_adjust_room helper. */
652 enum bpf_adj_room_mode {
653 	BPF_ADJ_ROOM_NET_OPTS,
654 };
655 
656 /* user accessible mirror of in-kernel sk_buff.
657  * new fields can only be added to the end of this structure
658  */
659 struct __sk_buff {
660 	__u32 len;
661 	__u32 pkt_type;
662 	__u32 mark;
663 	__u32 queue_mapping;
664 	__u32 protocol;
665 	__u32 vlan_present;
666 	__u32 vlan_tci;
667 	__u32 vlan_proto;
668 	__u32 priority;
669 	__u32 ingress_ifindex;
670 	__u32 ifindex;
671 	__u32 tc_index;
672 	__u32 cb[5];
673 	__u32 hash;
674 	__u32 tc_classid;
675 	__u32 data;
676 	__u32 data_end;
677 	__u32 napi_id;
678 };
679 
680 struct bpf_tunnel_key {
681 	__u32 tunnel_id;
682 	union {
683 		__u32 remote_ipv4;
684 		__u32 remote_ipv6[4];
685 	};
686 	__u8 tunnel_tos;
687 	__u8 tunnel_ttl;
688 	__u16 tunnel_ext;
689 	__u32 tunnel_label;
690 };
691 
692 /* Generic BPF return codes which all BPF program types may support.
693  * The values are binary compatible with their TC_ACT_* counter-part to
694  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
695  * programs.
696  *
697  * XDP is handled seprately, see XDP_*.
698  */
699 enum bpf_ret_code {
700 	BPF_OK = 0,
701 	/* 1 reserved */
702 	BPF_DROP = 2,
703 	/* 3-6 reserved */
704 	BPF_REDIRECT = 7,
705 	/* >127 are reserved for prog type specific return codes */
706 };
707 
708 struct bpf_sock {
709 	__u32 bound_dev_if;
710 	__u32 family;
711 	__u32 type;
712 	__u32 protocol;
713 };
714 
715 #define XDP_PACKET_HEADROOM 256
716 
717 /* User return codes for XDP prog type.
718  * A valid XDP program must return one of these defined values. All other
719  * return codes are reserved for future use. Unknown return codes will result
720  * in packet drop.
721  */
722 enum xdp_action {
723 	XDP_ABORTED = 0,
724 	XDP_DROP,
725 	XDP_PASS,
726 	XDP_TX,
727 };
728 
729 /* user accessible metadata for XDP packet hook
730  * new fields must be added to the end of this structure
731  */
732 struct xdp_md {
733 	__u32 data;
734 	__u32 data_end;
735 };
736 
737 #define BPF_TAG_SIZE	8
738 
739 struct bpf_prog_info {
740 	__u32 type;
741 	__u32 id;
742 	__u8  tag[BPF_TAG_SIZE];
743 	__u32 jited_prog_len;
744 	__u32 xlated_prog_len;
745 	__aligned_u64 jited_prog_insns;
746 	__aligned_u64 xlated_prog_insns;
747 } __attribute__((aligned(8)));
748 
749 struct bpf_map_info {
750 	__u32 type;
751 	__u32 id;
752 	__u32 key_size;
753 	__u32 value_size;
754 	__u32 max_entries;
755 	__u32 map_flags;
756 } __attribute__((aligned(8)));
757 
758 /* User bpf_sock_ops struct to access socket values and specify request ops
759  * and their replies.
760  * New fields can only be added at the end of this structure
761  */
762 struct bpf_sock_ops {
763 	__u32 op;
764 	union {
765 		__u32 reply;
766 		__u32 replylong[4];
767 	};
768 	__u32 family;
769 	__u32 remote_ip4;
770 	__u32 local_ip4;
771 	__u32 remote_ip6[4];
772 	__u32 local_ip6[4];
773 	__u32 remote_port;
774 	__u32 local_port;
775 };
776 
777 /* List of known BPF sock_ops operators.
778  * New entries can only be added at the end
779  */
780 enum {
781 	BPF_SOCK_OPS_VOID,
782 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
783 					 * -1 if default value should be used
784 					 */
785 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
786 					 * window (in packets) or -1 if default
787 					 * value should be used
788 					 */
789 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
790 					 * active connection is initialized
791 					 */
792 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
793 						 * active connection is
794 						 * established
795 						 */
796 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
797 						 * passive connection is
798 						 * established
799 						 */
800 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
801 					 * needs ECN
802 					 */
803 };
804 
805 #define TCP_BPF_IW		1001	/* Set TCP initial congestion window */
806 #define TCP_BPF_SNDCWND_CLAMP	1002	/* Set sndcwnd_clamp */
807 
808 #endif /* _UAPI__LINUX_BPF_H__ */
809