xref: /linux-6.15/tools/include/uapi/linux/bpf.h (revision bbb03029)
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 #define BPF_JNE		0x50	/* jump != */
34 #define BPF_JSGT	0x60	/* SGT is signed '>', GT in x86 */
35 #define BPF_JSGE	0x70	/* SGE is signed '>=', GE in x86 */
36 #define BPF_CALL	0x80	/* function call */
37 #define BPF_EXIT	0x90	/* function return */
38 
39 /* Register numbers */
40 enum {
41 	BPF_REG_0 = 0,
42 	BPF_REG_1,
43 	BPF_REG_2,
44 	BPF_REG_3,
45 	BPF_REG_4,
46 	BPF_REG_5,
47 	BPF_REG_6,
48 	BPF_REG_7,
49 	BPF_REG_8,
50 	BPF_REG_9,
51 	BPF_REG_10,
52 	__MAX_BPF_REG,
53 };
54 
55 /* BPF has 10 general purpose 64-bit registers and stack frame. */
56 #define MAX_BPF_REG	__MAX_BPF_REG
57 
58 struct bpf_insn {
59 	__u8	code;		/* opcode */
60 	__u8	dst_reg:4;	/* dest register */
61 	__u8	src_reg:4;	/* source register */
62 	__s16	off;		/* signed offset */
63 	__s32	imm;		/* signed immediate constant */
64 };
65 
66 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
67 struct bpf_lpm_trie_key {
68 	__u32	prefixlen;	/* up to 32 for AF_INET, 128 for AF_INET6 */
69 	__u8	data[0];	/* Arbitrary size */
70 };
71 
72 /* BPF syscall commands, see bpf(2) man-page for details. */
73 enum bpf_cmd {
74 	BPF_MAP_CREATE,
75 	BPF_MAP_LOOKUP_ELEM,
76 	BPF_MAP_UPDATE_ELEM,
77 	BPF_MAP_DELETE_ELEM,
78 	BPF_MAP_GET_NEXT_KEY,
79 	BPF_PROG_LOAD,
80 	BPF_OBJ_PIN,
81 	BPF_OBJ_GET,
82 	BPF_PROG_ATTACH,
83 	BPF_PROG_DETACH,
84 	BPF_PROG_TEST_RUN,
85 	BPF_PROG_GET_NEXT_ID,
86 	BPF_MAP_GET_NEXT_ID,
87 	BPF_PROG_GET_FD_BY_ID,
88 	BPF_MAP_GET_FD_BY_ID,
89 	BPF_OBJ_GET_INFO_BY_FD,
90 };
91 
92 enum bpf_map_type {
93 	BPF_MAP_TYPE_UNSPEC,
94 	BPF_MAP_TYPE_HASH,
95 	BPF_MAP_TYPE_ARRAY,
96 	BPF_MAP_TYPE_PROG_ARRAY,
97 	BPF_MAP_TYPE_PERF_EVENT_ARRAY,
98 	BPF_MAP_TYPE_PERCPU_HASH,
99 	BPF_MAP_TYPE_PERCPU_ARRAY,
100 	BPF_MAP_TYPE_STACK_TRACE,
101 	BPF_MAP_TYPE_CGROUP_ARRAY,
102 	BPF_MAP_TYPE_LRU_HASH,
103 	BPF_MAP_TYPE_LRU_PERCPU_HASH,
104 	BPF_MAP_TYPE_LPM_TRIE,
105 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
106 	BPF_MAP_TYPE_HASH_OF_MAPS,
107 	BPF_MAP_TYPE_DEVMAP,
108 };
109 
110 enum bpf_prog_type {
111 	BPF_PROG_TYPE_UNSPEC,
112 	BPF_PROG_TYPE_SOCKET_FILTER,
113 	BPF_PROG_TYPE_KPROBE,
114 	BPF_PROG_TYPE_SCHED_CLS,
115 	BPF_PROG_TYPE_SCHED_ACT,
116 	BPF_PROG_TYPE_TRACEPOINT,
117 	BPF_PROG_TYPE_XDP,
118 	BPF_PROG_TYPE_PERF_EVENT,
119 	BPF_PROG_TYPE_CGROUP_SKB,
120 	BPF_PROG_TYPE_CGROUP_SOCK,
121 	BPF_PROG_TYPE_LWT_IN,
122 	BPF_PROG_TYPE_LWT_OUT,
123 	BPF_PROG_TYPE_LWT_XMIT,
124 	BPF_PROG_TYPE_SOCK_OPS,
125 };
126 
127 enum bpf_attach_type {
128 	BPF_CGROUP_INET_INGRESS,
129 	BPF_CGROUP_INET_EGRESS,
130 	BPF_CGROUP_INET_SOCK_CREATE,
131 	BPF_CGROUP_SOCK_OPS,
132 	__MAX_BPF_ATTACH_TYPE
133 };
134 
135 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
136 
137 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
138  * to the given target_fd cgroup the descendent cgroup will be able to
139  * override effective bpf program that was inherited from this cgroup
140  */
141 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
142 
143 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
144  * verifier will perform strict alignment checking as if the kernel
145  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
146  * and NET_IP_ALIGN defined to 2.
147  */
148 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
149 
150 #define BPF_PSEUDO_MAP_FD	1
151 
152 /* flags for BPF_MAP_UPDATE_ELEM command */
153 #define BPF_ANY		0 /* create new element or update existing */
154 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
155 #define BPF_EXIST	2 /* update existing element */
156 
157 #define BPF_F_NO_PREALLOC	(1U << 0)
158 /* Instead of having one common LRU list in the
159  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
160  * which can scale and perform better.
161  * Note, the LRU nodes (including free nodes) cannot be moved
162  * across different LRU lists.
163  */
164 #define BPF_F_NO_COMMON_LRU	(1U << 1)
165 
166 union bpf_attr {
167 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
168 		__u32	map_type;	/* one of enum bpf_map_type */
169 		__u32	key_size;	/* size of key in bytes */
170 		__u32	value_size;	/* size of value in bytes */
171 		__u32	max_entries;	/* max number of entries in a map */
172 		__u32	map_flags;	/* prealloc or not */
173 		__u32	inner_map_fd;	/* fd pointing to the inner map */
174 	};
175 
176 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
177 		__u32		map_fd;
178 		__aligned_u64	key;
179 		union {
180 			__aligned_u64 value;
181 			__aligned_u64 next_key;
182 		};
183 		__u64		flags;
184 	};
185 
186 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
187 		__u32		prog_type;	/* one of enum bpf_prog_type */
188 		__u32		insn_cnt;
189 		__aligned_u64	insns;
190 		__aligned_u64	license;
191 		__u32		log_level;	/* verbosity level of verifier */
192 		__u32		log_size;	/* size of user buffer */
193 		__aligned_u64	log_buf;	/* user supplied buffer */
194 		__u32		kern_version;	/* checked when prog_type=kprobe */
195 		__u32		prog_flags;
196 	};
197 
198 	struct { /* anonymous struct used by BPF_OBJ_* commands */
199 		__aligned_u64	pathname;
200 		__u32		bpf_fd;
201 	};
202 
203 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
204 		__u32		target_fd;	/* container object to attach to */
205 		__u32		attach_bpf_fd;	/* eBPF program to attach */
206 		__u32		attach_type;
207 		__u32		attach_flags;
208 	};
209 
210 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
211 		__u32		prog_fd;
212 		__u32		retval;
213 		__u32		data_size_in;
214 		__u32		data_size_out;
215 		__aligned_u64	data_in;
216 		__aligned_u64	data_out;
217 		__u32		repeat;
218 		__u32		duration;
219 	} test;
220 
221 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
222 		union {
223 			__u32		start_id;
224 			__u32		prog_id;
225 			__u32		map_id;
226 		};
227 		__u32		next_id;
228 	};
229 
230 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
231 		__u32		bpf_fd;
232 		__u32		info_len;
233 		__aligned_u64	info;
234 	} info;
235 } __attribute__((aligned(8)));
236 
237 /* BPF helper function descriptions:
238  *
239  * void *bpf_map_lookup_elem(&map, &key)
240  *     Return: Map value or NULL
241  *
242  * int bpf_map_update_elem(&map, &key, &value, flags)
243  *     Return: 0 on success or negative error
244  *
245  * int bpf_map_delete_elem(&map, &key)
246  *     Return: 0 on success or negative error
247  *
248  * int bpf_probe_read(void *dst, int size, void *src)
249  *     Return: 0 on success or negative error
250  *
251  * u64 bpf_ktime_get_ns(void)
252  *     Return: current ktime
253  *
254  * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
255  *     Return: length of buffer written or negative error
256  *
257  * u32 bpf_prandom_u32(void)
258  *     Return: random value
259  *
260  * u32 bpf_raw_smp_processor_id(void)
261  *     Return: SMP processor ID
262  *
263  * int bpf_skb_store_bytes(skb, offset, from, len, flags)
264  *     store bytes into packet
265  *     @skb: pointer to skb
266  *     @offset: offset within packet from skb->mac_header
267  *     @from: pointer where to copy bytes from
268  *     @len: number of bytes to store into packet
269  *     @flags: bit 0 - if true, recompute skb->csum
270  *             other bits - reserved
271  *     Return: 0 on success or negative error
272  *
273  * int bpf_l3_csum_replace(skb, offset, from, to, flags)
274  *     recompute IP checksum
275  *     @skb: pointer to skb
276  *     @offset: offset within packet where IP checksum is located
277  *     @from: old value of header field
278  *     @to: new value of header field
279  *     @flags: bits 0-3 - size of header field
280  *             other bits - reserved
281  *     Return: 0 on success or negative error
282  *
283  * int bpf_l4_csum_replace(skb, offset, from, to, flags)
284  *     recompute TCP/UDP checksum
285  *     @skb: pointer to skb
286  *     @offset: offset within packet where TCP/UDP checksum is located
287  *     @from: old value of header field
288  *     @to: new value of header field
289  *     @flags: bits 0-3 - size of header field
290  *             bit 4 - is pseudo header
291  *             other bits - reserved
292  *     Return: 0 on success or negative error
293  *
294  * int bpf_tail_call(ctx, prog_array_map, index)
295  *     jump into another BPF program
296  *     @ctx: context pointer passed to next program
297  *     @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
298  *     @index: index inside array that selects specific program to run
299  *     Return: 0 on success or negative error
300  *
301  * int bpf_clone_redirect(skb, ifindex, flags)
302  *     redirect to another netdev
303  *     @skb: pointer to skb
304  *     @ifindex: ifindex of the net device
305  *     @flags: bit 0 - if set, redirect to ingress instead of egress
306  *             other bits - reserved
307  *     Return: 0 on success or negative error
308  *
309  * u64 bpf_get_current_pid_tgid(void)
310  *     Return: current->tgid << 32 | current->pid
311  *
312  * u64 bpf_get_current_uid_gid(void)
313  *     Return: current_gid << 32 | current_uid
314  *
315  * int bpf_get_current_comm(char *buf, int size_of_buf)
316  *     stores current->comm into buf
317  *     Return: 0 on success or negative error
318  *
319  * u32 bpf_get_cgroup_classid(skb)
320  *     retrieve a proc's classid
321  *     @skb: pointer to skb
322  *     Return: classid if != 0
323  *
324  * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
325  *     Return: 0 on success or negative error
326  *
327  * int bpf_skb_vlan_pop(skb)
328  *     Return: 0 on success or negative error
329  *
330  * int bpf_skb_get_tunnel_key(skb, key, size, flags)
331  * int bpf_skb_set_tunnel_key(skb, key, size, flags)
332  *     retrieve or populate tunnel metadata
333  *     @skb: pointer to skb
334  *     @key: pointer to 'struct bpf_tunnel_key'
335  *     @size: size of 'struct bpf_tunnel_key'
336  *     @flags: room for future extensions
337  *     Return: 0 on success or negative error
338  *
339  * u64 bpf_perf_event_read(map, flags)
340  *     read perf event counter value
341  *     @map: pointer to perf_event_array map
342  *     @flags: index of event in the map or bitmask flags
343  *     Return: value of perf event counter read or error code
344  *
345  * int bpf_redirect(ifindex, flags)
346  *     redirect to another netdev
347  *     @ifindex: ifindex of the net device
348  *     @flags: bit 0 - if set, redirect to ingress instead of egress
349  *             other bits - reserved
350  *     Return: TC_ACT_REDIRECT
351  *
352  * u32 bpf_get_route_realm(skb)
353  *     retrieve a dst's tclassid
354  *     @skb: pointer to skb
355  *     Return: realm if != 0
356  *
357  * int bpf_perf_event_output(ctx, map, flags, data, size)
358  *     output perf raw sample
359  *     @ctx: struct pt_regs*
360  *     @map: pointer to perf_event_array map
361  *     @flags: index of event in the map or bitmask flags
362  *     @data: data on stack to be output as raw data
363  *     @size: size of data
364  *     Return: 0 on success or negative error
365  *
366  * int bpf_get_stackid(ctx, map, flags)
367  *     walk user or kernel stack and return id
368  *     @ctx: struct pt_regs*
369  *     @map: pointer to stack_trace map
370  *     @flags: bits 0-7 - numer of stack frames to skip
371  *             bit 8 - collect user stack instead of kernel
372  *             bit 9 - compare stacks by hash only
373  *             bit 10 - if two different stacks hash into the same stackid
374  *                      discard old
375  *             other bits - reserved
376  *     Return: >= 0 stackid on success or negative error
377  *
378  * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
379  *     calculate csum diff
380  *     @from: raw from buffer
381  *     @from_size: length of from buffer
382  *     @to: raw to buffer
383  *     @to_size: length of to buffer
384  *     @seed: optional seed
385  *     Return: csum result or negative error code
386  *
387  * int bpf_skb_get_tunnel_opt(skb, opt, size)
388  *     retrieve tunnel options metadata
389  *     @skb: pointer to skb
390  *     @opt: pointer to raw tunnel option data
391  *     @size: size of @opt
392  *     Return: option size
393  *
394  * int bpf_skb_set_tunnel_opt(skb, opt, size)
395  *     populate tunnel options metadata
396  *     @skb: pointer to skb
397  *     @opt: pointer to raw tunnel option data
398  *     @size: size of @opt
399  *     Return: 0 on success or negative error
400  *
401  * int bpf_skb_change_proto(skb, proto, flags)
402  *     Change protocol of the skb. Currently supported is v4 -> v6,
403  *     v6 -> v4 transitions. The helper will also resize the skb. eBPF
404  *     program is expected to fill the new headers via skb_store_bytes
405  *     and lX_csum_replace.
406  *     @skb: pointer to skb
407  *     @proto: new skb->protocol type
408  *     @flags: reserved
409  *     Return: 0 on success or negative error
410  *
411  * int bpf_skb_change_type(skb, type)
412  *     Change packet type of skb.
413  *     @skb: pointer to skb
414  *     @type: new skb->pkt_type type
415  *     Return: 0 on success or negative error
416  *
417  * int bpf_skb_under_cgroup(skb, map, index)
418  *     Check cgroup2 membership of skb
419  *     @skb: pointer to skb
420  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
421  *     @index: index of the cgroup in the bpf_map
422  *     Return:
423  *       == 0 skb failed the cgroup2 descendant test
424  *       == 1 skb succeeded the cgroup2 descendant test
425  *        < 0 error
426  *
427  * u32 bpf_get_hash_recalc(skb)
428  *     Retrieve and possibly recalculate skb->hash.
429  *     @skb: pointer to skb
430  *     Return: hash
431  *
432  * u64 bpf_get_current_task(void)
433  *     Returns current task_struct
434  *     Return: current
435  *
436  * int bpf_probe_write_user(void *dst, void *src, int len)
437  *     safely attempt to write to a location
438  *     @dst: destination address in userspace
439  *     @src: source address on stack
440  *     @len: number of bytes to copy
441  *     Return: 0 on success or negative error
442  *
443  * int bpf_current_task_under_cgroup(map, index)
444  *     Check cgroup2 membership of current task
445  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
446  *     @index: index of the cgroup in the bpf_map
447  *     Return:
448  *       == 0 current failed the cgroup2 descendant test
449  *       == 1 current succeeded the cgroup2 descendant test
450  *        < 0 error
451  *
452  * int bpf_skb_change_tail(skb, len, flags)
453  *     The helper will resize the skb to the given new size, to be used f.e.
454  *     with control messages.
455  *     @skb: pointer to skb
456  *     @len: new skb length
457  *     @flags: reserved
458  *     Return: 0 on success or negative error
459  *
460  * int bpf_skb_pull_data(skb, len)
461  *     The helper will pull in non-linear data in case the skb is non-linear
462  *     and not all of len are part of the linear section. Only needed for
463  *     read/write with direct packet access.
464  *     @skb: pointer to skb
465  *     @len: len to make read/writeable
466  *     Return: 0 on success or negative error
467  *
468  * s64 bpf_csum_update(skb, csum)
469  *     Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
470  *     @skb: pointer to skb
471  *     @csum: csum to add
472  *     Return: csum on success or negative error
473  *
474  * void bpf_set_hash_invalid(skb)
475  *     Invalidate current skb->hash.
476  *     @skb: pointer to skb
477  *
478  * int bpf_get_numa_node_id()
479  *     Return: Id of current NUMA node.
480  *
481  * int bpf_skb_change_head()
482  *     Grows headroom of skb and adjusts MAC header offset accordingly.
483  *     Will extends/reallocae as required automatically.
484  *     May change skb data pointer and will thus invalidate any check
485  *     performed for direct packet access.
486  *     @skb: pointer to skb
487  *     @len: length of header to be pushed in front
488  *     @flags: Flags (unused for now)
489  *     Return: 0 on success or negative error
490  *
491  * int bpf_xdp_adjust_head(xdp_md, delta)
492  *     Adjust the xdp_md.data by delta
493  *     @xdp_md: pointer to xdp_md
494  *     @delta: An positive/negative integer to be added to xdp_md.data
495  *     Return: 0 on success or negative on error
496  *
497  * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
498  *     Copy a NUL terminated string from unsafe address. In case the string
499  *     length is smaller than size, the target is not padded with further NUL
500  *     bytes. In case the string length is larger than size, just count-1
501  *     bytes are copied and the last byte is set to NUL.
502  *     @dst: destination address
503  *     @size: maximum number of bytes to copy, including the trailing NUL
504  *     @unsafe_ptr: unsafe address
505  *     Return:
506  *       > 0 length of the string including the trailing NUL on success
507  *       < 0 error
508  *
509  * u64 bpf_get_socket_cookie(skb)
510  *     Get the cookie for the socket stored inside sk_buff.
511  *     @skb: pointer to skb
512  *     Return: 8 Bytes non-decreasing number on success or 0 if the socket
513  *     field is missing inside sk_buff
514  *
515  * u32 bpf_get_socket_uid(skb)
516  *     Get the owner uid of the socket stored inside sk_buff.
517  *     @skb: pointer to skb
518  *     Return: uid of the socket owner on success or overflowuid if failed.
519  *
520  * u32 bpf_set_hash(skb, hash)
521  *     Set full skb->hash.
522  *     @skb: pointer to skb
523  *     @hash: hash to set
524  *
525  * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
526  *     Calls setsockopt. Not all opts are available, only those with
527  *     integer optvals plus TCP_CONGESTION.
528  *     Supported levels: SOL_SOCKET and IPROTO_TCP
529  *     @bpf_socket: pointer to bpf_socket
530  *     @level: SOL_SOCKET or IPROTO_TCP
531  *     @optname: option name
532  *     @optval: pointer to option value
533  *     @optlen: length of optval in byes
534  *     Return: 0 or negative error
535  *
536  * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
537  *     Grow or shrink room in sk_buff.
538  *     @skb: pointer to skb
539  *     @len_diff: (signed) amount of room to grow/shrink
540  *     @mode: operation mode (enum bpf_adj_room_mode)
541  *     @flags: reserved for future use
542  *     Return: 0 on success or negative error code
543  */
544 #define __BPF_FUNC_MAPPER(FN)		\
545 	FN(unspec),			\
546 	FN(map_lookup_elem),		\
547 	FN(map_update_elem),		\
548 	FN(map_delete_elem),		\
549 	FN(probe_read),			\
550 	FN(ktime_get_ns),		\
551 	FN(trace_printk),		\
552 	FN(get_prandom_u32),		\
553 	FN(get_smp_processor_id),	\
554 	FN(skb_store_bytes),		\
555 	FN(l3_csum_replace),		\
556 	FN(l4_csum_replace),		\
557 	FN(tail_call),			\
558 	FN(clone_redirect),		\
559 	FN(get_current_pid_tgid),	\
560 	FN(get_current_uid_gid),	\
561 	FN(get_current_comm),		\
562 	FN(get_cgroup_classid),		\
563 	FN(skb_vlan_push),		\
564 	FN(skb_vlan_pop),		\
565 	FN(skb_get_tunnel_key),		\
566 	FN(skb_set_tunnel_key),		\
567 	FN(perf_event_read),		\
568 	FN(redirect),			\
569 	FN(get_route_realm),		\
570 	FN(perf_event_output),		\
571 	FN(skb_load_bytes),		\
572 	FN(get_stackid),		\
573 	FN(csum_diff),			\
574 	FN(skb_get_tunnel_opt),		\
575 	FN(skb_set_tunnel_opt),		\
576 	FN(skb_change_proto),		\
577 	FN(skb_change_type),		\
578 	FN(skb_under_cgroup),		\
579 	FN(get_hash_recalc),		\
580 	FN(get_current_task),		\
581 	FN(probe_write_user),		\
582 	FN(current_task_under_cgroup),	\
583 	FN(skb_change_tail),		\
584 	FN(skb_pull_data),		\
585 	FN(csum_update),		\
586 	FN(set_hash_invalid),		\
587 	FN(get_numa_node_id),		\
588 	FN(skb_change_head),		\
589 	FN(xdp_adjust_head),		\
590 	FN(probe_read_str),		\
591 	FN(get_socket_cookie),		\
592 	FN(get_socket_uid),		\
593 	FN(set_hash),			\
594 	FN(setsockopt),			\
595 	FN(skb_adjust_room),
596 
597 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
598  * function eBPF program intends to call
599  */
600 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
601 enum bpf_func_id {
602 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
603 	__BPF_FUNC_MAX_ID,
604 };
605 #undef __BPF_ENUM_FN
606 
607 /* All flags used by eBPF helper functions, placed here. */
608 
609 /* BPF_FUNC_skb_store_bytes flags. */
610 #define BPF_F_RECOMPUTE_CSUM		(1ULL << 0)
611 #define BPF_F_INVALIDATE_HASH		(1ULL << 1)
612 
613 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
614  * First 4 bits are for passing the header field size.
615  */
616 #define BPF_F_HDR_FIELD_MASK		0xfULL
617 
618 /* BPF_FUNC_l4_csum_replace flags. */
619 #define BPF_F_PSEUDO_HDR		(1ULL << 4)
620 #define BPF_F_MARK_MANGLED_0		(1ULL << 5)
621 #define BPF_F_MARK_ENFORCE		(1ULL << 6)
622 
623 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
624 #define BPF_F_INGRESS			(1ULL << 0)
625 
626 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
627 #define BPF_F_TUNINFO_IPV6		(1ULL << 0)
628 
629 /* BPF_FUNC_get_stackid flags. */
630 #define BPF_F_SKIP_FIELD_MASK		0xffULL
631 #define BPF_F_USER_STACK		(1ULL << 8)
632 #define BPF_F_FAST_STACK_CMP		(1ULL << 9)
633 #define BPF_F_REUSE_STACKID		(1ULL << 10)
634 
635 /* BPF_FUNC_skb_set_tunnel_key flags. */
636 #define BPF_F_ZERO_CSUM_TX		(1ULL << 1)
637 #define BPF_F_DONT_FRAGMENT		(1ULL << 2)
638 
639 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
640 #define BPF_F_INDEX_MASK		0xffffffffULL
641 #define BPF_F_CURRENT_CPU		BPF_F_INDEX_MASK
642 /* BPF_FUNC_perf_event_output for sk_buff input context. */
643 #define BPF_F_CTXLEN_MASK		(0xfffffULL << 32)
644 
645 /* Mode for BPF_FUNC_skb_adjust_room helper. */
646 enum bpf_adj_room_mode {
647 	BPF_ADJ_ROOM_NET_OPTS,
648 };
649 
650 /* user accessible mirror of in-kernel sk_buff.
651  * new fields can only be added to the end of this structure
652  */
653 struct __sk_buff {
654 	__u32 len;
655 	__u32 pkt_type;
656 	__u32 mark;
657 	__u32 queue_mapping;
658 	__u32 protocol;
659 	__u32 vlan_present;
660 	__u32 vlan_tci;
661 	__u32 vlan_proto;
662 	__u32 priority;
663 	__u32 ingress_ifindex;
664 	__u32 ifindex;
665 	__u32 tc_index;
666 	__u32 cb[5];
667 	__u32 hash;
668 	__u32 tc_classid;
669 	__u32 data;
670 	__u32 data_end;
671 	__u32 napi_id;
672 };
673 
674 struct bpf_tunnel_key {
675 	__u32 tunnel_id;
676 	union {
677 		__u32 remote_ipv4;
678 		__u32 remote_ipv6[4];
679 	};
680 	__u8 tunnel_tos;
681 	__u8 tunnel_ttl;
682 	__u16 tunnel_ext;
683 	__u32 tunnel_label;
684 };
685 
686 /* Generic BPF return codes which all BPF program types may support.
687  * The values are binary compatible with their TC_ACT_* counter-part to
688  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
689  * programs.
690  *
691  * XDP is handled seprately, see XDP_*.
692  */
693 enum bpf_ret_code {
694 	BPF_OK = 0,
695 	/* 1 reserved */
696 	BPF_DROP = 2,
697 	/* 3-6 reserved */
698 	BPF_REDIRECT = 7,
699 	/* >127 are reserved for prog type specific return codes */
700 };
701 
702 struct bpf_sock {
703 	__u32 bound_dev_if;
704 	__u32 family;
705 	__u32 type;
706 	__u32 protocol;
707 };
708 
709 #define XDP_PACKET_HEADROOM 256
710 
711 /* User return codes for XDP prog type.
712  * A valid XDP program must return one of these defined values. All other
713  * return codes are reserved for future use. Unknown return codes will result
714  * in packet drop.
715  */
716 enum xdp_action {
717 	XDP_ABORTED = 0,
718 	XDP_DROP,
719 	XDP_PASS,
720 	XDP_TX,
721 };
722 
723 /* user accessible metadata for XDP packet hook
724  * new fields must be added to the end of this structure
725  */
726 struct xdp_md {
727 	__u32 data;
728 	__u32 data_end;
729 };
730 
731 #define BPF_TAG_SIZE	8
732 
733 struct bpf_prog_info {
734 	__u32 type;
735 	__u32 id;
736 	__u8  tag[BPF_TAG_SIZE];
737 	__u32 jited_prog_len;
738 	__u32 xlated_prog_len;
739 	__aligned_u64 jited_prog_insns;
740 	__aligned_u64 xlated_prog_insns;
741 } __attribute__((aligned(8)));
742 
743 struct bpf_map_info {
744 	__u32 type;
745 	__u32 id;
746 	__u32 key_size;
747 	__u32 value_size;
748 	__u32 max_entries;
749 	__u32 map_flags;
750 } __attribute__((aligned(8)));
751 
752 /* User bpf_sock_ops struct to access socket values and specify request ops
753  * and their replies.
754  * New fields can only be added at the end of this structure
755  */
756 struct bpf_sock_ops {
757 	__u32 op;
758 	union {
759 		__u32 reply;
760 		__u32 replylong[4];
761 	};
762 	__u32 family;
763 	__u32 remote_ip4;
764 	__u32 local_ip4;
765 	__u32 remote_ip6[4];
766 	__u32 local_ip6[4];
767 	__u32 remote_port;
768 	__u32 local_port;
769 };
770 
771 /* List of known BPF sock_ops operators.
772  * New entries can only be added at the end
773  */
774 enum {
775 	BPF_SOCK_OPS_VOID,
776 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
777 					 * -1 if default value should be used
778 					 */
779 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
780 					 * window (in packets) or -1 if default
781 					 * value should be used
782 					 */
783 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
784 					 * active connection is initialized
785 					 */
786 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
787 						 * active connection is
788 						 * established
789 						 */
790 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
791 						 * passive connection is
792 						 * established
793 						 */
794 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
795 					 * needs ECN
796 					 */
797 };
798 
799 #define TCP_BPF_IW		1001	/* Set TCP initial congestion window */
800 #define TCP_BPF_SNDCWND_CLAMP	1002	/* Set sndcwnd_clamp */
801 
802 #endif /* _UAPI__LINUX_BPF_H__ */
803