xref: /linux-6.15/include/uapi/linux/bpf.h (revision 3fd87127)
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 	BPF_MAP_TYPE_SOCKMAP,
114 };
115 
116 enum bpf_prog_type {
117 	BPF_PROG_TYPE_UNSPEC,
118 	BPF_PROG_TYPE_SOCKET_FILTER,
119 	BPF_PROG_TYPE_KPROBE,
120 	BPF_PROG_TYPE_SCHED_CLS,
121 	BPF_PROG_TYPE_SCHED_ACT,
122 	BPF_PROG_TYPE_TRACEPOINT,
123 	BPF_PROG_TYPE_XDP,
124 	BPF_PROG_TYPE_PERF_EVENT,
125 	BPF_PROG_TYPE_CGROUP_SKB,
126 	BPF_PROG_TYPE_CGROUP_SOCK,
127 	BPF_PROG_TYPE_LWT_IN,
128 	BPF_PROG_TYPE_LWT_OUT,
129 	BPF_PROG_TYPE_LWT_XMIT,
130 	BPF_PROG_TYPE_SOCK_OPS,
131 	BPF_PROG_TYPE_SK_SKB,
132 };
133 
134 enum bpf_attach_type {
135 	BPF_CGROUP_INET_INGRESS,
136 	BPF_CGROUP_INET_EGRESS,
137 	BPF_CGROUP_INET_SOCK_CREATE,
138 	BPF_CGROUP_SOCK_OPS,
139 	BPF_CGROUP_SMAP_INGRESS,
140 	__MAX_BPF_ATTACH_TYPE
141 };
142 
143 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
144 
145 /* If BPF_SOCKMAP_STRPARSER is used sockmap will use strparser on receive */
146 #define BPF_SOCKMAP_STRPARSER	(1U << 0)
147 
148 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
149  * to the given target_fd cgroup the descendent cgroup will be able to
150  * override effective bpf program that was inherited from this cgroup
151  */
152 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
153 
154 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
155  * verifier will perform strict alignment checking as if the kernel
156  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
157  * and NET_IP_ALIGN defined to 2.
158  */
159 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
160 
161 #define BPF_PSEUDO_MAP_FD	1
162 
163 /* flags for BPF_MAP_UPDATE_ELEM command */
164 #define BPF_ANY		0 /* create new element or update existing */
165 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
166 #define BPF_EXIST	2 /* update existing element */
167 
168 /* flags for BPF_MAP_CREATE command */
169 #define BPF_F_NO_PREALLOC	(1U << 0)
170 /* Instead of having one common LRU list in the
171  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
172  * which can scale and perform better.
173  * Note, the LRU nodes (including free nodes) cannot be moved
174  * across different LRU lists.
175  */
176 #define BPF_F_NO_COMMON_LRU	(1U << 1)
177 /* Specify numa node during map creation */
178 #define BPF_F_NUMA_NODE		(1U << 2)
179 
180 union bpf_attr {
181 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
182 		__u32	map_type;	/* one of enum bpf_map_type */
183 		__u32	key_size;	/* size of key in bytes */
184 		__u32	value_size;	/* size of value in bytes */
185 		__u32	max_entries;	/* max number of entries in a map */
186 		__u32	map_flags;	/* BPF_MAP_CREATE related
187 					 * flags defined above.
188 					 */
189 		__u32	inner_map_fd;	/* fd pointing to the inner map */
190 		__u32	numa_node;	/* numa node (effective only if
191 					 * BPF_F_NUMA_NODE is set).
192 					 */
193 	};
194 
195 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
196 		__u32		map_fd;
197 		__aligned_u64	key;
198 		union {
199 			__aligned_u64 value;
200 			__aligned_u64 next_key;
201 		};
202 		__u64		flags;
203 	};
204 
205 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
206 		__u32		prog_type;	/* one of enum bpf_prog_type */
207 		__u32		insn_cnt;
208 		__aligned_u64	insns;
209 		__aligned_u64	license;
210 		__u32		log_level;	/* verbosity level of verifier */
211 		__u32		log_size;	/* size of user buffer */
212 		__aligned_u64	log_buf;	/* user supplied buffer */
213 		__u32		kern_version;	/* checked when prog_type=kprobe */
214 		__u32		prog_flags;
215 	};
216 
217 	struct { /* anonymous struct used by BPF_OBJ_* commands */
218 		__aligned_u64	pathname;
219 		__u32		bpf_fd;
220 	};
221 
222 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
223 		__u32		target_fd;	/* container object to attach to */
224 		__u32		attach_bpf_fd;	/* eBPF program to attach */
225 		__u32		attach_type;
226 		__u32		attach_flags;
227 		__u32		attach_bpf_fd2;
228 	};
229 
230 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
231 		__u32		prog_fd;
232 		__u32		retval;
233 		__u32		data_size_in;
234 		__u32		data_size_out;
235 		__aligned_u64	data_in;
236 		__aligned_u64	data_out;
237 		__u32		repeat;
238 		__u32		duration;
239 	} test;
240 
241 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
242 		union {
243 			__u32		start_id;
244 			__u32		prog_id;
245 			__u32		map_id;
246 		};
247 		__u32		next_id;
248 	};
249 
250 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
251 		__u32		bpf_fd;
252 		__u32		info_len;
253 		__aligned_u64	info;
254 	} info;
255 } __attribute__((aligned(8)));
256 
257 /* BPF helper function descriptions:
258  *
259  * void *bpf_map_lookup_elem(&map, &key)
260  *     Return: Map value or NULL
261  *
262  * int bpf_map_update_elem(&map, &key, &value, flags)
263  *     Return: 0 on success or negative error
264  *
265  * int bpf_map_delete_elem(&map, &key)
266  *     Return: 0 on success or negative error
267  *
268  * int bpf_probe_read(void *dst, int size, void *src)
269  *     Return: 0 on success or negative error
270  *
271  * u64 bpf_ktime_get_ns(void)
272  *     Return: current ktime
273  *
274  * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
275  *     Return: length of buffer written or negative error
276  *
277  * u32 bpf_prandom_u32(void)
278  *     Return: random value
279  *
280  * u32 bpf_raw_smp_processor_id(void)
281  *     Return: SMP processor ID
282  *
283  * int bpf_skb_store_bytes(skb, offset, from, len, flags)
284  *     store bytes into packet
285  *     @skb: pointer to skb
286  *     @offset: offset within packet from skb->mac_header
287  *     @from: pointer where to copy bytes from
288  *     @len: number of bytes to store into packet
289  *     @flags: bit 0 - if true, recompute skb->csum
290  *             other bits - reserved
291  *     Return: 0 on success or negative error
292  *
293  * int bpf_l3_csum_replace(skb, offset, from, to, flags)
294  *     recompute IP checksum
295  *     @skb: pointer to skb
296  *     @offset: offset within packet where IP checksum is located
297  *     @from: old value of header field
298  *     @to: new value of header field
299  *     @flags: bits 0-3 - size of header field
300  *             other bits - reserved
301  *     Return: 0 on success or negative error
302  *
303  * int bpf_l4_csum_replace(skb, offset, from, to, flags)
304  *     recompute TCP/UDP checksum
305  *     @skb: pointer to skb
306  *     @offset: offset within packet where TCP/UDP checksum is located
307  *     @from: old value of header field
308  *     @to: new value of header field
309  *     @flags: bits 0-3 - size of header field
310  *             bit 4 - is pseudo header
311  *             other bits - reserved
312  *     Return: 0 on success or negative error
313  *
314  * int bpf_tail_call(ctx, prog_array_map, index)
315  *     jump into another BPF program
316  *     @ctx: context pointer passed to next program
317  *     @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
318  *     @index: index inside array that selects specific program to run
319  *     Return: 0 on success or negative error
320  *
321  * int bpf_clone_redirect(skb, ifindex, flags)
322  *     redirect to another netdev
323  *     @skb: pointer to skb
324  *     @ifindex: ifindex of the net device
325  *     @flags: bit 0 - if set, redirect to ingress instead of egress
326  *             other bits - reserved
327  *     Return: 0 on success or negative error
328  *
329  * u64 bpf_get_current_pid_tgid(void)
330  *     Return: current->tgid << 32 | current->pid
331  *
332  * u64 bpf_get_current_uid_gid(void)
333  *     Return: current_gid << 32 | current_uid
334  *
335  * int bpf_get_current_comm(char *buf, int size_of_buf)
336  *     stores current->comm into buf
337  *     Return: 0 on success or negative error
338  *
339  * u32 bpf_get_cgroup_classid(skb)
340  *     retrieve a proc's classid
341  *     @skb: pointer to skb
342  *     Return: classid if != 0
343  *
344  * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
345  *     Return: 0 on success or negative error
346  *
347  * int bpf_skb_vlan_pop(skb)
348  *     Return: 0 on success or negative error
349  *
350  * int bpf_skb_get_tunnel_key(skb, key, size, flags)
351  * int bpf_skb_set_tunnel_key(skb, key, size, flags)
352  *     retrieve or populate tunnel metadata
353  *     @skb: pointer to skb
354  *     @key: pointer to 'struct bpf_tunnel_key'
355  *     @size: size of 'struct bpf_tunnel_key'
356  *     @flags: room for future extensions
357  *     Return: 0 on success or negative error
358  *
359  * u64 bpf_perf_event_read(map, flags)
360  *     read perf event counter value
361  *     @map: pointer to perf_event_array map
362  *     @flags: index of event in the map or bitmask flags
363  *     Return: value of perf event counter read or error code
364  *
365  * int bpf_redirect(ifindex, flags)
366  *     redirect to another netdev
367  *     @ifindex: ifindex of the net device
368  *     @flags:
369  *	  cls_bpf:
370  *          bit 0 - if set, redirect to ingress instead of egress
371  *          other bits - reserved
372  *	  xdp_bpf:
373  *	    all bits - reserved
374  *     Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
375  *	       xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
376  * int bpf_redirect_map(map, key, flags)
377  *     redirect to endpoint in map
378  *     @map: pointer to dev map
379  *     @key: index in map to lookup
380  *     @flags: --
381  *     Return: XDP_REDIRECT on success or XDP_ABORT on error
382  *
383  * u32 bpf_get_route_realm(skb)
384  *     retrieve a dst's tclassid
385  *     @skb: pointer to skb
386  *     Return: realm if != 0
387  *
388  * int bpf_perf_event_output(ctx, map, flags, data, size)
389  *     output perf raw sample
390  *     @ctx: struct pt_regs*
391  *     @map: pointer to perf_event_array map
392  *     @flags: index of event in the map or bitmask flags
393  *     @data: data on stack to be output as raw data
394  *     @size: size of data
395  *     Return: 0 on success or negative error
396  *
397  * int bpf_get_stackid(ctx, map, flags)
398  *     walk user or kernel stack and return id
399  *     @ctx: struct pt_regs*
400  *     @map: pointer to stack_trace map
401  *     @flags: bits 0-7 - numer of stack frames to skip
402  *             bit 8 - collect user stack instead of kernel
403  *             bit 9 - compare stacks by hash only
404  *             bit 10 - if two different stacks hash into the same stackid
405  *                      discard old
406  *             other bits - reserved
407  *     Return: >= 0 stackid on success or negative error
408  *
409  * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
410  *     calculate csum diff
411  *     @from: raw from buffer
412  *     @from_size: length of from buffer
413  *     @to: raw to buffer
414  *     @to_size: length of to buffer
415  *     @seed: optional seed
416  *     Return: csum result or negative error code
417  *
418  * int bpf_skb_get_tunnel_opt(skb, opt, size)
419  *     retrieve tunnel options metadata
420  *     @skb: pointer to skb
421  *     @opt: pointer to raw tunnel option data
422  *     @size: size of @opt
423  *     Return: option size
424  *
425  * int bpf_skb_set_tunnel_opt(skb, opt, size)
426  *     populate tunnel options metadata
427  *     @skb: pointer to skb
428  *     @opt: pointer to raw tunnel option data
429  *     @size: size of @opt
430  *     Return: 0 on success or negative error
431  *
432  * int bpf_skb_change_proto(skb, proto, flags)
433  *     Change protocol of the skb. Currently supported is v4 -> v6,
434  *     v6 -> v4 transitions. The helper will also resize the skb. eBPF
435  *     program is expected to fill the new headers via skb_store_bytes
436  *     and lX_csum_replace.
437  *     @skb: pointer to skb
438  *     @proto: new skb->protocol type
439  *     @flags: reserved
440  *     Return: 0 on success or negative error
441  *
442  * int bpf_skb_change_type(skb, type)
443  *     Change packet type of skb.
444  *     @skb: pointer to skb
445  *     @type: new skb->pkt_type type
446  *     Return: 0 on success or negative error
447  *
448  * int bpf_skb_under_cgroup(skb, map, index)
449  *     Check cgroup2 membership of skb
450  *     @skb: pointer to skb
451  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
452  *     @index: index of the cgroup in the bpf_map
453  *     Return:
454  *       == 0 skb failed the cgroup2 descendant test
455  *       == 1 skb succeeded the cgroup2 descendant test
456  *        < 0 error
457  *
458  * u32 bpf_get_hash_recalc(skb)
459  *     Retrieve and possibly recalculate skb->hash.
460  *     @skb: pointer to skb
461  *     Return: hash
462  *
463  * u64 bpf_get_current_task(void)
464  *     Returns current task_struct
465  *     Return: current
466  *
467  * int bpf_probe_write_user(void *dst, void *src, int len)
468  *     safely attempt to write to a location
469  *     @dst: destination address in userspace
470  *     @src: source address on stack
471  *     @len: number of bytes to copy
472  *     Return: 0 on success or negative error
473  *
474  * int bpf_current_task_under_cgroup(map, index)
475  *     Check cgroup2 membership of current task
476  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
477  *     @index: index of the cgroup in the bpf_map
478  *     Return:
479  *       == 0 current failed the cgroup2 descendant test
480  *       == 1 current succeeded the cgroup2 descendant test
481  *        < 0 error
482  *
483  * int bpf_skb_change_tail(skb, len, flags)
484  *     The helper will resize the skb to the given new size, to be used f.e.
485  *     with control messages.
486  *     @skb: pointer to skb
487  *     @len: new skb length
488  *     @flags: reserved
489  *     Return: 0 on success or negative error
490  *
491  * int bpf_skb_pull_data(skb, len)
492  *     The helper will pull in non-linear data in case the skb is non-linear
493  *     and not all of len are part of the linear section. Only needed for
494  *     read/write with direct packet access.
495  *     @skb: pointer to skb
496  *     @len: len to make read/writeable
497  *     Return: 0 on success or negative error
498  *
499  * s64 bpf_csum_update(skb, csum)
500  *     Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
501  *     @skb: pointer to skb
502  *     @csum: csum to add
503  *     Return: csum on success or negative error
504  *
505  * void bpf_set_hash_invalid(skb)
506  *     Invalidate current skb->hash.
507  *     @skb: pointer to skb
508  *
509  * int bpf_get_numa_node_id()
510  *     Return: Id of current NUMA node.
511  *
512  * int bpf_skb_change_head()
513  *     Grows headroom of skb and adjusts MAC header offset accordingly.
514  *     Will extends/reallocae as required automatically.
515  *     May change skb data pointer and will thus invalidate any check
516  *     performed for direct packet access.
517  *     @skb: pointer to skb
518  *     @len: length of header to be pushed in front
519  *     @flags: Flags (unused for now)
520  *     Return: 0 on success or negative error
521  *
522  * int bpf_xdp_adjust_head(xdp_md, delta)
523  *     Adjust the xdp_md.data by delta
524  *     @xdp_md: pointer to xdp_md
525  *     @delta: An positive/negative integer to be added to xdp_md.data
526  *     Return: 0 on success or negative on error
527  *
528  * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
529  *     Copy a NUL terminated string from unsafe address. In case the string
530  *     length is smaller than size, the target is not padded with further NUL
531  *     bytes. In case the string length is larger than size, just count-1
532  *     bytes are copied and the last byte is set to NUL.
533  *     @dst: destination address
534  *     @size: maximum number of bytes to copy, including the trailing NUL
535  *     @unsafe_ptr: unsafe address
536  *     Return:
537  *       > 0 length of the string including the trailing NUL on success
538  *       < 0 error
539  *
540  * u64 bpf_get_socket_cookie(skb)
541  *     Get the cookie for the socket stored inside sk_buff.
542  *     @skb: pointer to skb
543  *     Return: 8 Bytes non-decreasing number on success or 0 if the socket
544  *     field is missing inside sk_buff
545  *
546  * u32 bpf_get_socket_uid(skb)
547  *     Get the owner uid of the socket stored inside sk_buff.
548  *     @skb: pointer to skb
549  *     Return: uid of the socket owner on success or overflowuid if failed.
550  *
551  * u32 bpf_set_hash(skb, hash)
552  *     Set full skb->hash.
553  *     @skb: pointer to skb
554  *     @hash: hash to set
555  *
556  * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
557  *     Calls setsockopt. Not all opts are available, only those with
558  *     integer optvals plus TCP_CONGESTION.
559  *     Supported levels: SOL_SOCKET and IPROTO_TCP
560  *     @bpf_socket: pointer to bpf_socket
561  *     @level: SOL_SOCKET or IPROTO_TCP
562  *     @optname: option name
563  *     @optval: pointer to option value
564  *     @optlen: length of optval in byes
565  *     Return: 0 or negative error
566  *
567  * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
568  *     Grow or shrink room in sk_buff.
569  *     @skb: pointer to skb
570  *     @len_diff: (signed) amount of room to grow/shrink
571  *     @mode: operation mode (enum bpf_adj_room_mode)
572  *     @flags: reserved for future use
573  *     Return: 0 on success or negative error code
574  *
575  * int bpf_sk_redirect_map(map, key, flags)
576  *     Redirect skb to a sock in map using key as a lookup key for the
577  *     sock in map.
578  *     @map: pointer to sockmap
579  *     @key: key to lookup sock in map
580  *     @flags: reserved for future use
581  *     Return: SK_REDIRECT
582  *
583  * int bpf_sock_map_update(skops, map, key, flags, map_flags)
584  *	@skops: pointer to bpf_sock_ops
585  *	@map: pointer to sockmap to update
586  *	@key: key to insert/update sock in map
587  *	@flags: same flags as map update elem
588  *	@map_flags: sock map specific flags
589  *	   bit 1: Enable strparser
590  *	   other bits: reserved
591  */
592 #define __BPF_FUNC_MAPPER(FN)		\
593 	FN(unspec),			\
594 	FN(map_lookup_elem),		\
595 	FN(map_update_elem),		\
596 	FN(map_delete_elem),		\
597 	FN(probe_read),			\
598 	FN(ktime_get_ns),		\
599 	FN(trace_printk),		\
600 	FN(get_prandom_u32),		\
601 	FN(get_smp_processor_id),	\
602 	FN(skb_store_bytes),		\
603 	FN(l3_csum_replace),		\
604 	FN(l4_csum_replace),		\
605 	FN(tail_call),			\
606 	FN(clone_redirect),		\
607 	FN(get_current_pid_tgid),	\
608 	FN(get_current_uid_gid),	\
609 	FN(get_current_comm),		\
610 	FN(get_cgroup_classid),		\
611 	FN(skb_vlan_push),		\
612 	FN(skb_vlan_pop),		\
613 	FN(skb_get_tunnel_key),		\
614 	FN(skb_set_tunnel_key),		\
615 	FN(perf_event_read),		\
616 	FN(redirect),			\
617 	FN(get_route_realm),		\
618 	FN(perf_event_output),		\
619 	FN(skb_load_bytes),		\
620 	FN(get_stackid),		\
621 	FN(csum_diff),			\
622 	FN(skb_get_tunnel_opt),		\
623 	FN(skb_set_tunnel_opt),		\
624 	FN(skb_change_proto),		\
625 	FN(skb_change_type),		\
626 	FN(skb_under_cgroup),		\
627 	FN(get_hash_recalc),		\
628 	FN(get_current_task),		\
629 	FN(probe_write_user),		\
630 	FN(current_task_under_cgroup),	\
631 	FN(skb_change_tail),		\
632 	FN(skb_pull_data),		\
633 	FN(csum_update),		\
634 	FN(set_hash_invalid),		\
635 	FN(get_numa_node_id),		\
636 	FN(skb_change_head),		\
637 	FN(xdp_adjust_head),		\
638 	FN(probe_read_str),		\
639 	FN(get_socket_cookie),		\
640 	FN(get_socket_uid),		\
641 	FN(set_hash),			\
642 	FN(setsockopt),			\
643 	FN(skb_adjust_room),		\
644 	FN(redirect_map),		\
645 	FN(sk_redirect_map),		\
646 	FN(sock_map_update),		\
647 
648 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
649  * function eBPF program intends to call
650  */
651 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
652 enum bpf_func_id {
653 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
654 	__BPF_FUNC_MAX_ID,
655 };
656 #undef __BPF_ENUM_FN
657 
658 /* All flags used by eBPF helper functions, placed here. */
659 
660 /* BPF_FUNC_skb_store_bytes flags. */
661 #define BPF_F_RECOMPUTE_CSUM		(1ULL << 0)
662 #define BPF_F_INVALIDATE_HASH		(1ULL << 1)
663 
664 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
665  * First 4 bits are for passing the header field size.
666  */
667 #define BPF_F_HDR_FIELD_MASK		0xfULL
668 
669 /* BPF_FUNC_l4_csum_replace flags. */
670 #define BPF_F_PSEUDO_HDR		(1ULL << 4)
671 #define BPF_F_MARK_MANGLED_0		(1ULL << 5)
672 #define BPF_F_MARK_ENFORCE		(1ULL << 6)
673 
674 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
675 #define BPF_F_INGRESS			(1ULL << 0)
676 
677 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
678 #define BPF_F_TUNINFO_IPV6		(1ULL << 0)
679 
680 /* BPF_FUNC_get_stackid flags. */
681 #define BPF_F_SKIP_FIELD_MASK		0xffULL
682 #define BPF_F_USER_STACK		(1ULL << 8)
683 #define BPF_F_FAST_STACK_CMP		(1ULL << 9)
684 #define BPF_F_REUSE_STACKID		(1ULL << 10)
685 
686 /* BPF_FUNC_skb_set_tunnel_key flags. */
687 #define BPF_F_ZERO_CSUM_TX		(1ULL << 1)
688 #define BPF_F_DONT_FRAGMENT		(1ULL << 2)
689 
690 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
691 #define BPF_F_INDEX_MASK		0xffffffffULL
692 #define BPF_F_CURRENT_CPU		BPF_F_INDEX_MASK
693 /* BPF_FUNC_perf_event_output for sk_buff input context. */
694 #define BPF_F_CTXLEN_MASK		(0xfffffULL << 32)
695 
696 /* Mode for BPF_FUNC_skb_adjust_room helper. */
697 enum bpf_adj_room_mode {
698 	BPF_ADJ_ROOM_NET,
699 };
700 
701 /* user accessible mirror of in-kernel sk_buff.
702  * new fields can only be added to the end of this structure
703  */
704 struct __sk_buff {
705 	__u32 len;
706 	__u32 pkt_type;
707 	__u32 mark;
708 	__u32 queue_mapping;
709 	__u32 protocol;
710 	__u32 vlan_present;
711 	__u32 vlan_tci;
712 	__u32 vlan_proto;
713 	__u32 priority;
714 	__u32 ingress_ifindex;
715 	__u32 ifindex;
716 	__u32 tc_index;
717 	__u32 cb[5];
718 	__u32 hash;
719 	__u32 tc_classid;
720 	__u32 data;
721 	__u32 data_end;
722 	__u32 napi_id;
723 
724 	/* accessed by BPF_PROG_TYPE_sk_skb types */
725 	__u32 family;
726 	__u32 remote_ip4;	/* Stored in network byte order */
727 	__u32 local_ip4;	/* Stored in network byte order */
728 	__u32 remote_ip6[4];	/* Stored in network byte order */
729 	__u32 local_ip6[4];	/* Stored in network byte order */
730 	__u32 remote_port;	/* Stored in network byte order */
731 	__u32 local_port;	/* stored in host byte order */
732 };
733 
734 struct bpf_tunnel_key {
735 	__u32 tunnel_id;
736 	union {
737 		__u32 remote_ipv4;
738 		__u32 remote_ipv6[4];
739 	};
740 	__u8 tunnel_tos;
741 	__u8 tunnel_ttl;
742 	__u16 tunnel_ext;
743 	__u32 tunnel_label;
744 };
745 
746 /* Generic BPF return codes which all BPF program types may support.
747  * The values are binary compatible with their TC_ACT_* counter-part to
748  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
749  * programs.
750  *
751  * XDP is handled seprately, see XDP_*.
752  */
753 enum bpf_ret_code {
754 	BPF_OK = 0,
755 	/* 1 reserved */
756 	BPF_DROP = 2,
757 	/* 3-6 reserved */
758 	BPF_REDIRECT = 7,
759 	/* >127 are reserved for prog type specific return codes */
760 };
761 
762 struct bpf_sock {
763 	__u32 bound_dev_if;
764 	__u32 family;
765 	__u32 type;
766 	__u32 protocol;
767 };
768 
769 #define XDP_PACKET_HEADROOM 256
770 
771 /* User return codes for XDP prog type.
772  * A valid XDP program must return one of these defined values. All other
773  * return codes are reserved for future use. Unknown return codes will result
774  * in packet drop.
775  */
776 enum xdp_action {
777 	XDP_ABORTED = 0,
778 	XDP_DROP,
779 	XDP_PASS,
780 	XDP_TX,
781 	XDP_REDIRECT,
782 };
783 
784 /* user accessible metadata for XDP packet hook
785  * new fields must be added to the end of this structure
786  */
787 struct xdp_md {
788 	__u32 data;
789 	__u32 data_end;
790 };
791 
792 enum sk_action {
793 	SK_ABORTED = 0,
794 	SK_DROP,
795 	SK_REDIRECT,
796 };
797 
798 #define BPF_TAG_SIZE	8
799 
800 struct bpf_prog_info {
801 	__u32 type;
802 	__u32 id;
803 	__u8  tag[BPF_TAG_SIZE];
804 	__u32 jited_prog_len;
805 	__u32 xlated_prog_len;
806 	__aligned_u64 jited_prog_insns;
807 	__aligned_u64 xlated_prog_insns;
808 } __attribute__((aligned(8)));
809 
810 struct bpf_map_info {
811 	__u32 type;
812 	__u32 id;
813 	__u32 key_size;
814 	__u32 value_size;
815 	__u32 max_entries;
816 	__u32 map_flags;
817 } __attribute__((aligned(8)));
818 
819 /* User bpf_sock_ops struct to access socket values and specify request ops
820  * and their replies.
821  * Some of this fields are in network (bigendian) byte order and may need
822  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
823  * New fields can only be added at the end of this structure
824  */
825 struct bpf_sock_ops {
826 	__u32 op;
827 	union {
828 		__u32 reply;
829 		__u32 replylong[4];
830 	};
831 	__u32 family;
832 	__u32 remote_ip4;	/* Stored in network byte order */
833 	__u32 local_ip4;	/* Stored in network byte order */
834 	__u32 remote_ip6[4];	/* Stored in network byte order */
835 	__u32 local_ip6[4];	/* Stored in network byte order */
836 	__u32 remote_port;	/* Stored in network byte order */
837 	__u32 local_port;	/* stored in host byte order */
838 };
839 
840 /* List of known BPF sock_ops operators.
841  * New entries can only be added at the end
842  */
843 enum {
844 	BPF_SOCK_OPS_VOID,
845 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
846 					 * -1 if default value should be used
847 					 */
848 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
849 					 * window (in packets) or -1 if default
850 					 * value should be used
851 					 */
852 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
853 					 * active connection is initialized
854 					 */
855 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
856 						 * active connection is
857 						 * established
858 						 */
859 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
860 						 * passive connection is
861 						 * established
862 						 */
863 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
864 					 * needs ECN
865 					 */
866 };
867 
868 #define TCP_BPF_IW		1001	/* Set TCP initial congestion window */
869 #define TCP_BPF_SNDCWND_CLAMP	1002	/* Set sndcwnd_clamp */
870 
871 #endif /* _UAPI__LINUX_BPF_H__ */
872