xref: /linux-6.15/include/uapi/linux/bpf.h (revision 9d64fc08)
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_SK_SKB_STREAM_PARSER,
140 	BPF_SK_SKB_STREAM_VERDICT,
141 	__MAX_BPF_ATTACH_TYPE
142 };
143 
144 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
145 
146 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
147  * to the given target_fd cgroup the descendent cgroup will be able to
148  * override effective bpf program that was inherited from this cgroup
149  */
150 #define BPF_F_ALLOW_OVERRIDE	(1U << 0)
151 
152 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
153  * verifier will perform strict alignment checking as if the kernel
154  * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
155  * and NET_IP_ALIGN defined to 2.
156  */
157 #define BPF_F_STRICT_ALIGNMENT	(1U << 0)
158 
159 #define BPF_PSEUDO_MAP_FD	1
160 
161 /* flags for BPF_MAP_UPDATE_ELEM command */
162 #define BPF_ANY		0 /* create new element or update existing */
163 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
164 #define BPF_EXIST	2 /* update existing element */
165 
166 /* flags for BPF_MAP_CREATE command */
167 #define BPF_F_NO_PREALLOC	(1U << 0)
168 /* Instead of having one common LRU list in the
169  * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
170  * which can scale and perform better.
171  * Note, the LRU nodes (including free nodes) cannot be moved
172  * across different LRU lists.
173  */
174 #define BPF_F_NO_COMMON_LRU	(1U << 1)
175 /* Specify numa node during map creation */
176 #define BPF_F_NUMA_NODE		(1U << 2)
177 
178 union bpf_attr {
179 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
180 		__u32	map_type;	/* one of enum bpf_map_type */
181 		__u32	key_size;	/* size of key in bytes */
182 		__u32	value_size;	/* size of value in bytes */
183 		__u32	max_entries;	/* max number of entries in a map */
184 		__u32	map_flags;	/* BPF_MAP_CREATE related
185 					 * flags defined above.
186 					 */
187 		__u32	inner_map_fd;	/* fd pointing to the inner map */
188 		__u32	numa_node;	/* numa node (effective only if
189 					 * BPF_F_NUMA_NODE is set).
190 					 */
191 	};
192 
193 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
194 		__u32		map_fd;
195 		__aligned_u64	key;
196 		union {
197 			__aligned_u64 value;
198 			__aligned_u64 next_key;
199 		};
200 		__u64		flags;
201 	};
202 
203 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
204 		__u32		prog_type;	/* one of enum bpf_prog_type */
205 		__u32		insn_cnt;
206 		__aligned_u64	insns;
207 		__aligned_u64	license;
208 		__u32		log_level;	/* verbosity level of verifier */
209 		__u32		log_size;	/* size of user buffer */
210 		__aligned_u64	log_buf;	/* user supplied buffer */
211 		__u32		kern_version;	/* checked when prog_type=kprobe */
212 		__u32		prog_flags;
213 	};
214 
215 	struct { /* anonymous struct used by BPF_OBJ_* commands */
216 		__aligned_u64	pathname;
217 		__u32		bpf_fd;
218 	};
219 
220 	struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
221 		__u32		target_fd;	/* container object to attach to */
222 		__u32		attach_bpf_fd;	/* eBPF program to attach */
223 		__u32		attach_type;
224 		__u32		attach_flags;
225 	};
226 
227 	struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
228 		__u32		prog_fd;
229 		__u32		retval;
230 		__u32		data_size_in;
231 		__u32		data_size_out;
232 		__aligned_u64	data_in;
233 		__aligned_u64	data_out;
234 		__u32		repeat;
235 		__u32		duration;
236 	} test;
237 
238 	struct { /* anonymous struct used by BPF_*_GET_*_ID */
239 		union {
240 			__u32		start_id;
241 			__u32		prog_id;
242 			__u32		map_id;
243 		};
244 		__u32		next_id;
245 	};
246 
247 	struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
248 		__u32		bpf_fd;
249 		__u32		info_len;
250 		__aligned_u64	info;
251 	} info;
252 } __attribute__((aligned(8)));
253 
254 /* BPF helper function descriptions:
255  *
256  * void *bpf_map_lookup_elem(&map, &key)
257  *     Return: Map value or NULL
258  *
259  * int bpf_map_update_elem(&map, &key, &value, flags)
260  *     Return: 0 on success or negative error
261  *
262  * int bpf_map_delete_elem(&map, &key)
263  *     Return: 0 on success or negative error
264  *
265  * int bpf_probe_read(void *dst, int size, void *src)
266  *     Return: 0 on success or negative error
267  *
268  * u64 bpf_ktime_get_ns(void)
269  *     Return: current ktime
270  *
271  * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
272  *     Return: length of buffer written or negative error
273  *
274  * u32 bpf_prandom_u32(void)
275  *     Return: random value
276  *
277  * u32 bpf_raw_smp_processor_id(void)
278  *     Return: SMP processor ID
279  *
280  * int bpf_skb_store_bytes(skb, offset, from, len, flags)
281  *     store bytes into packet
282  *     @skb: pointer to skb
283  *     @offset: offset within packet from skb->mac_header
284  *     @from: pointer where to copy bytes from
285  *     @len: number of bytes to store into packet
286  *     @flags: bit 0 - if true, recompute skb->csum
287  *             other bits - reserved
288  *     Return: 0 on success or negative error
289  *
290  * int bpf_l3_csum_replace(skb, offset, from, to, flags)
291  *     recompute IP checksum
292  *     @skb: pointer to skb
293  *     @offset: offset within packet where IP checksum is located
294  *     @from: old value of header field
295  *     @to: new value of header field
296  *     @flags: bits 0-3 - size of header field
297  *             other bits - reserved
298  *     Return: 0 on success or negative error
299  *
300  * int bpf_l4_csum_replace(skb, offset, from, to, flags)
301  *     recompute TCP/UDP checksum
302  *     @skb: pointer to skb
303  *     @offset: offset within packet where TCP/UDP checksum is located
304  *     @from: old value of header field
305  *     @to: new value of header field
306  *     @flags: bits 0-3 - size of header field
307  *             bit 4 - is pseudo header
308  *             other bits - reserved
309  *     Return: 0 on success or negative error
310  *
311  * int bpf_tail_call(ctx, prog_array_map, index)
312  *     jump into another BPF program
313  *     @ctx: context pointer passed to next program
314  *     @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
315  *     @index: 32-bit index inside array that selects specific program to run
316  *     Return: 0 on success or negative error
317  *
318  * int bpf_clone_redirect(skb, ifindex, flags)
319  *     redirect to another netdev
320  *     @skb: pointer to skb
321  *     @ifindex: ifindex of the net device
322  *     @flags: bit 0 - if set, redirect to ingress instead of egress
323  *             other bits - reserved
324  *     Return: 0 on success or negative error
325  *
326  * u64 bpf_get_current_pid_tgid(void)
327  *     Return: current->tgid << 32 | current->pid
328  *
329  * u64 bpf_get_current_uid_gid(void)
330  *     Return: current_gid << 32 | current_uid
331  *
332  * int bpf_get_current_comm(char *buf, int size_of_buf)
333  *     stores current->comm into buf
334  *     Return: 0 on success or negative error
335  *
336  * u32 bpf_get_cgroup_classid(skb)
337  *     retrieve a proc's classid
338  *     @skb: pointer to skb
339  *     Return: classid if != 0
340  *
341  * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
342  *     Return: 0 on success or negative error
343  *
344  * int bpf_skb_vlan_pop(skb)
345  *     Return: 0 on success or negative error
346  *
347  * int bpf_skb_get_tunnel_key(skb, key, size, flags)
348  * int bpf_skb_set_tunnel_key(skb, key, size, flags)
349  *     retrieve or populate tunnel metadata
350  *     @skb: pointer to skb
351  *     @key: pointer to 'struct bpf_tunnel_key'
352  *     @size: size of 'struct bpf_tunnel_key'
353  *     @flags: room for future extensions
354  *     Return: 0 on success or negative error
355  *
356  * u64 bpf_perf_event_read(map, flags)
357  *     read perf event counter value
358  *     @map: pointer to perf_event_array map
359  *     @flags: index of event in the map or bitmask flags
360  *     Return: value of perf event counter read or error code
361  *
362  * int bpf_redirect(ifindex, flags)
363  *     redirect to another netdev
364  *     @ifindex: ifindex of the net device
365  *     @flags:
366  *	  cls_bpf:
367  *          bit 0 - if set, redirect to ingress instead of egress
368  *          other bits - reserved
369  *	  xdp_bpf:
370  *	    all bits - reserved
371  *     Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
372  *	       xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
373  * int bpf_redirect_map(map, key, flags)
374  *     redirect to endpoint in map
375  *     @map: pointer to dev map
376  *     @key: index in map to lookup
377  *     @flags: --
378  *     Return: XDP_REDIRECT on success or XDP_ABORT on error
379  *
380  * u32 bpf_get_route_realm(skb)
381  *     retrieve a dst's tclassid
382  *     @skb: pointer to skb
383  *     Return: realm if != 0
384  *
385  * int bpf_perf_event_output(ctx, map, flags, data, size)
386  *     output perf raw sample
387  *     @ctx: struct pt_regs*
388  *     @map: pointer to perf_event_array map
389  *     @flags: index of event in the map or bitmask flags
390  *     @data: data on stack to be output as raw data
391  *     @size: size of data
392  *     Return: 0 on success or negative error
393  *
394  * int bpf_get_stackid(ctx, map, flags)
395  *     walk user or kernel stack and return id
396  *     @ctx: struct pt_regs*
397  *     @map: pointer to stack_trace map
398  *     @flags: bits 0-7 - numer of stack frames to skip
399  *             bit 8 - collect user stack instead of kernel
400  *             bit 9 - compare stacks by hash only
401  *             bit 10 - if two different stacks hash into the same stackid
402  *                      discard old
403  *             other bits - reserved
404  *     Return: >= 0 stackid on success or negative error
405  *
406  * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
407  *     calculate csum diff
408  *     @from: raw from buffer
409  *     @from_size: length of from buffer
410  *     @to: raw to buffer
411  *     @to_size: length of to buffer
412  *     @seed: optional seed
413  *     Return: csum result or negative error code
414  *
415  * int bpf_skb_get_tunnel_opt(skb, opt, size)
416  *     retrieve tunnel options metadata
417  *     @skb: pointer to skb
418  *     @opt: pointer to raw tunnel option data
419  *     @size: size of @opt
420  *     Return: option size
421  *
422  * int bpf_skb_set_tunnel_opt(skb, opt, size)
423  *     populate tunnel options metadata
424  *     @skb: pointer to skb
425  *     @opt: pointer to raw tunnel option data
426  *     @size: size of @opt
427  *     Return: 0 on success or negative error
428  *
429  * int bpf_skb_change_proto(skb, proto, flags)
430  *     Change protocol of the skb. Currently supported is v4 -> v6,
431  *     v6 -> v4 transitions. The helper will also resize the skb. eBPF
432  *     program is expected to fill the new headers via skb_store_bytes
433  *     and lX_csum_replace.
434  *     @skb: pointer to skb
435  *     @proto: new skb->protocol type
436  *     @flags: reserved
437  *     Return: 0 on success or negative error
438  *
439  * int bpf_skb_change_type(skb, type)
440  *     Change packet type of skb.
441  *     @skb: pointer to skb
442  *     @type: new skb->pkt_type type
443  *     Return: 0 on success or negative error
444  *
445  * int bpf_skb_under_cgroup(skb, map, index)
446  *     Check cgroup2 membership of skb
447  *     @skb: pointer to skb
448  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
449  *     @index: index of the cgroup in the bpf_map
450  *     Return:
451  *       == 0 skb failed the cgroup2 descendant test
452  *       == 1 skb succeeded the cgroup2 descendant test
453  *        < 0 error
454  *
455  * u32 bpf_get_hash_recalc(skb)
456  *     Retrieve and possibly recalculate skb->hash.
457  *     @skb: pointer to skb
458  *     Return: hash
459  *
460  * u64 bpf_get_current_task(void)
461  *     Returns current task_struct
462  *     Return: current
463  *
464  * int bpf_probe_write_user(void *dst, void *src, int len)
465  *     safely attempt to write to a location
466  *     @dst: destination address in userspace
467  *     @src: source address on stack
468  *     @len: number of bytes to copy
469  *     Return: 0 on success or negative error
470  *
471  * int bpf_current_task_under_cgroup(map, index)
472  *     Check cgroup2 membership of current task
473  *     @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
474  *     @index: index of the cgroup in the bpf_map
475  *     Return:
476  *       == 0 current failed the cgroup2 descendant test
477  *       == 1 current succeeded the cgroup2 descendant test
478  *        < 0 error
479  *
480  * int bpf_skb_change_tail(skb, len, flags)
481  *     The helper will resize the skb to the given new size, to be used f.e.
482  *     with control messages.
483  *     @skb: pointer to skb
484  *     @len: new skb length
485  *     @flags: reserved
486  *     Return: 0 on success or negative error
487  *
488  * int bpf_skb_pull_data(skb, len)
489  *     The helper will pull in non-linear data in case the skb is non-linear
490  *     and not all of len are part of the linear section. Only needed for
491  *     read/write with direct packet access.
492  *     @skb: pointer to skb
493  *     @len: len to make read/writeable
494  *     Return: 0 on success or negative error
495  *
496  * s64 bpf_csum_update(skb, csum)
497  *     Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
498  *     @skb: pointer to skb
499  *     @csum: csum to add
500  *     Return: csum on success or negative error
501  *
502  * void bpf_set_hash_invalid(skb)
503  *     Invalidate current skb->hash.
504  *     @skb: pointer to skb
505  *
506  * int bpf_get_numa_node_id()
507  *     Return: Id of current NUMA node.
508  *
509  * int bpf_skb_change_head()
510  *     Grows headroom of skb and adjusts MAC header offset accordingly.
511  *     Will extends/reallocae as required automatically.
512  *     May change skb data pointer and will thus invalidate any check
513  *     performed for direct packet access.
514  *     @skb: pointer to skb
515  *     @len: length of header to be pushed in front
516  *     @flags: Flags (unused for now)
517  *     Return: 0 on success or negative error
518  *
519  * int bpf_xdp_adjust_head(xdp_md, delta)
520  *     Adjust the xdp_md.data by delta
521  *     @xdp_md: pointer to xdp_md
522  *     @delta: An positive/negative integer to be added to xdp_md.data
523  *     Return: 0 on success or negative on error
524  *
525  * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
526  *     Copy a NUL terminated string from unsafe address. In case the string
527  *     length is smaller than size, the target is not padded with further NUL
528  *     bytes. In case the string length is larger than size, just count-1
529  *     bytes are copied and the last byte is set to NUL.
530  *     @dst: destination address
531  *     @size: maximum number of bytes to copy, including the trailing NUL
532  *     @unsafe_ptr: unsafe address
533  *     Return:
534  *       > 0 length of the string including the trailing NUL on success
535  *       < 0 error
536  *
537  * u64 bpf_get_socket_cookie(skb)
538  *     Get the cookie for the socket stored inside sk_buff.
539  *     @skb: pointer to skb
540  *     Return: 8 Bytes non-decreasing number on success or 0 if the socket
541  *     field is missing inside sk_buff
542  *
543  * u32 bpf_get_socket_uid(skb)
544  *     Get the owner uid of the socket stored inside sk_buff.
545  *     @skb: pointer to skb
546  *     Return: uid of the socket owner on success or overflowuid if failed.
547  *
548  * u32 bpf_set_hash(skb, hash)
549  *     Set full skb->hash.
550  *     @skb: pointer to skb
551  *     @hash: hash to set
552  *
553  * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
554  *     Calls setsockopt. Not all opts are available, only those with
555  *     integer optvals plus TCP_CONGESTION.
556  *     Supported levels: SOL_SOCKET and IPROTO_TCP
557  *     @bpf_socket: pointer to bpf_socket
558  *     @level: SOL_SOCKET or IPROTO_TCP
559  *     @optname: option name
560  *     @optval: pointer to option value
561  *     @optlen: length of optval in byes
562  *     Return: 0 or negative error
563  *
564  * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
565  *     Grow or shrink room in sk_buff.
566  *     @skb: pointer to skb
567  *     @len_diff: (signed) amount of room to grow/shrink
568  *     @mode: operation mode (enum bpf_adj_room_mode)
569  *     @flags: reserved for future use
570  *     Return: 0 on success or negative error code
571  *
572  * int bpf_sk_redirect_map(map, key, flags)
573  *     Redirect skb to a sock in map using key as a lookup key for the
574  *     sock in map.
575  *     @map: pointer to sockmap
576  *     @key: key to lookup sock in map
577  *     @flags: reserved for future use
578  *     Return: SK_REDIRECT
579  *
580  * int bpf_sock_map_update(skops, map, key, flags)
581  *	@skops: pointer to bpf_sock_ops
582  *	@map: pointer to sockmap to update
583  *	@key: key to insert/update sock in map
584  *	@flags: same flags as map update elem
585  */
586 #define __BPF_FUNC_MAPPER(FN)		\
587 	FN(unspec),			\
588 	FN(map_lookup_elem),		\
589 	FN(map_update_elem),		\
590 	FN(map_delete_elem),		\
591 	FN(probe_read),			\
592 	FN(ktime_get_ns),		\
593 	FN(trace_printk),		\
594 	FN(get_prandom_u32),		\
595 	FN(get_smp_processor_id),	\
596 	FN(skb_store_bytes),		\
597 	FN(l3_csum_replace),		\
598 	FN(l4_csum_replace),		\
599 	FN(tail_call),			\
600 	FN(clone_redirect),		\
601 	FN(get_current_pid_tgid),	\
602 	FN(get_current_uid_gid),	\
603 	FN(get_current_comm),		\
604 	FN(get_cgroup_classid),		\
605 	FN(skb_vlan_push),		\
606 	FN(skb_vlan_pop),		\
607 	FN(skb_get_tunnel_key),		\
608 	FN(skb_set_tunnel_key),		\
609 	FN(perf_event_read),		\
610 	FN(redirect),			\
611 	FN(get_route_realm),		\
612 	FN(perf_event_output),		\
613 	FN(skb_load_bytes),		\
614 	FN(get_stackid),		\
615 	FN(csum_diff),			\
616 	FN(skb_get_tunnel_opt),		\
617 	FN(skb_set_tunnel_opt),		\
618 	FN(skb_change_proto),		\
619 	FN(skb_change_type),		\
620 	FN(skb_under_cgroup),		\
621 	FN(get_hash_recalc),		\
622 	FN(get_current_task),		\
623 	FN(probe_write_user),		\
624 	FN(current_task_under_cgroup),	\
625 	FN(skb_change_tail),		\
626 	FN(skb_pull_data),		\
627 	FN(csum_update),		\
628 	FN(set_hash_invalid),		\
629 	FN(get_numa_node_id),		\
630 	FN(skb_change_head),		\
631 	FN(xdp_adjust_head),		\
632 	FN(probe_read_str),		\
633 	FN(get_socket_cookie),		\
634 	FN(get_socket_uid),		\
635 	FN(set_hash),			\
636 	FN(setsockopt),			\
637 	FN(skb_adjust_room),		\
638 	FN(redirect_map),		\
639 	FN(sk_redirect_map),		\
640 	FN(sock_map_update),		\
641 
642 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
643  * function eBPF program intends to call
644  */
645 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
646 enum bpf_func_id {
647 	__BPF_FUNC_MAPPER(__BPF_ENUM_FN)
648 	__BPF_FUNC_MAX_ID,
649 };
650 #undef __BPF_ENUM_FN
651 
652 /* All flags used by eBPF helper functions, placed here. */
653 
654 /* BPF_FUNC_skb_store_bytes flags. */
655 #define BPF_F_RECOMPUTE_CSUM		(1ULL << 0)
656 #define BPF_F_INVALIDATE_HASH		(1ULL << 1)
657 
658 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
659  * First 4 bits are for passing the header field size.
660  */
661 #define BPF_F_HDR_FIELD_MASK		0xfULL
662 
663 /* BPF_FUNC_l4_csum_replace flags. */
664 #define BPF_F_PSEUDO_HDR		(1ULL << 4)
665 #define BPF_F_MARK_MANGLED_0		(1ULL << 5)
666 #define BPF_F_MARK_ENFORCE		(1ULL << 6)
667 
668 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
669 #define BPF_F_INGRESS			(1ULL << 0)
670 
671 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
672 #define BPF_F_TUNINFO_IPV6		(1ULL << 0)
673 
674 /* BPF_FUNC_get_stackid flags. */
675 #define BPF_F_SKIP_FIELD_MASK		0xffULL
676 #define BPF_F_USER_STACK		(1ULL << 8)
677 #define BPF_F_FAST_STACK_CMP		(1ULL << 9)
678 #define BPF_F_REUSE_STACKID		(1ULL << 10)
679 
680 /* BPF_FUNC_skb_set_tunnel_key flags. */
681 #define BPF_F_ZERO_CSUM_TX		(1ULL << 1)
682 #define BPF_F_DONT_FRAGMENT		(1ULL << 2)
683 
684 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
685 #define BPF_F_INDEX_MASK		0xffffffffULL
686 #define BPF_F_CURRENT_CPU		BPF_F_INDEX_MASK
687 /* BPF_FUNC_perf_event_output for sk_buff input context. */
688 #define BPF_F_CTXLEN_MASK		(0xfffffULL << 32)
689 
690 /* Mode for BPF_FUNC_skb_adjust_room helper. */
691 enum bpf_adj_room_mode {
692 	BPF_ADJ_ROOM_NET,
693 };
694 
695 /* user accessible mirror of in-kernel sk_buff.
696  * new fields can only be added to the end of this structure
697  */
698 struct __sk_buff {
699 	__u32 len;
700 	__u32 pkt_type;
701 	__u32 mark;
702 	__u32 queue_mapping;
703 	__u32 protocol;
704 	__u32 vlan_present;
705 	__u32 vlan_tci;
706 	__u32 vlan_proto;
707 	__u32 priority;
708 	__u32 ingress_ifindex;
709 	__u32 ifindex;
710 	__u32 tc_index;
711 	__u32 cb[5];
712 	__u32 hash;
713 	__u32 tc_classid;
714 	__u32 data;
715 	__u32 data_end;
716 	__u32 napi_id;
717 
718 	/* accessed by BPF_PROG_TYPE_sk_skb types */
719 	__u32 family;
720 	__u32 remote_ip4;	/* Stored in network byte order */
721 	__u32 local_ip4;	/* Stored in network byte order */
722 	__u32 remote_ip6[4];	/* Stored in network byte order */
723 	__u32 local_ip6[4];	/* Stored in network byte order */
724 	__u32 remote_port;	/* Stored in network byte order */
725 	__u32 local_port;	/* stored in host byte order */
726 };
727 
728 struct bpf_tunnel_key {
729 	__u32 tunnel_id;
730 	union {
731 		__u32 remote_ipv4;
732 		__u32 remote_ipv6[4];
733 	};
734 	__u8 tunnel_tos;
735 	__u8 tunnel_ttl;
736 	__u16 tunnel_ext;
737 	__u32 tunnel_label;
738 };
739 
740 /* Generic BPF return codes which all BPF program types may support.
741  * The values are binary compatible with their TC_ACT_* counter-part to
742  * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
743  * programs.
744  *
745  * XDP is handled seprately, see XDP_*.
746  */
747 enum bpf_ret_code {
748 	BPF_OK = 0,
749 	/* 1 reserved */
750 	BPF_DROP = 2,
751 	/* 3-6 reserved */
752 	BPF_REDIRECT = 7,
753 	/* >127 are reserved for prog type specific return codes */
754 };
755 
756 struct bpf_sock {
757 	__u32 bound_dev_if;
758 	__u32 family;
759 	__u32 type;
760 	__u32 protocol;
761 	__u32 mark;
762 	__u32 priority;
763 };
764 
765 #define XDP_PACKET_HEADROOM 256
766 
767 /* User return codes for XDP prog type.
768  * A valid XDP program must return one of these defined values. All other
769  * return codes are reserved for future use. Unknown return codes will
770  * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
771  */
772 enum xdp_action {
773 	XDP_ABORTED = 0,
774 	XDP_DROP,
775 	XDP_PASS,
776 	XDP_TX,
777 	XDP_REDIRECT,
778 };
779 
780 /* user accessible metadata for XDP packet hook
781  * new fields must be added to the end of this structure
782  */
783 struct xdp_md {
784 	__u32 data;
785 	__u32 data_end;
786 };
787 
788 enum sk_action {
789 	SK_ABORTED = 0,
790 	SK_DROP,
791 	SK_REDIRECT,
792 };
793 
794 #define BPF_TAG_SIZE	8
795 
796 struct bpf_prog_info {
797 	__u32 type;
798 	__u32 id;
799 	__u8  tag[BPF_TAG_SIZE];
800 	__u32 jited_prog_len;
801 	__u32 xlated_prog_len;
802 	__aligned_u64 jited_prog_insns;
803 	__aligned_u64 xlated_prog_insns;
804 } __attribute__((aligned(8)));
805 
806 struct bpf_map_info {
807 	__u32 type;
808 	__u32 id;
809 	__u32 key_size;
810 	__u32 value_size;
811 	__u32 max_entries;
812 	__u32 map_flags;
813 } __attribute__((aligned(8)));
814 
815 /* User bpf_sock_ops struct to access socket values and specify request ops
816  * and their replies.
817  * Some of this fields are in network (bigendian) byte order and may need
818  * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
819  * New fields can only be added at the end of this structure
820  */
821 struct bpf_sock_ops {
822 	__u32 op;
823 	union {
824 		__u32 reply;
825 		__u32 replylong[4];
826 	};
827 	__u32 family;
828 	__u32 remote_ip4;	/* Stored in network byte order */
829 	__u32 local_ip4;	/* Stored in network byte order */
830 	__u32 remote_ip6[4];	/* Stored in network byte order */
831 	__u32 local_ip6[4];	/* Stored in network byte order */
832 	__u32 remote_port;	/* Stored in network byte order */
833 	__u32 local_port;	/* stored in host byte order */
834 };
835 
836 /* List of known BPF sock_ops operators.
837  * New entries can only be added at the end
838  */
839 enum {
840 	BPF_SOCK_OPS_VOID,
841 	BPF_SOCK_OPS_TIMEOUT_INIT,	/* Should return SYN-RTO value to use or
842 					 * -1 if default value should be used
843 					 */
844 	BPF_SOCK_OPS_RWND_INIT,		/* Should return initial advertized
845 					 * window (in packets) or -1 if default
846 					 * value should be used
847 					 */
848 	BPF_SOCK_OPS_TCP_CONNECT_CB,	/* Calls BPF program right before an
849 					 * active connection is initialized
850 					 */
851 	BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB,	/* Calls BPF program when an
852 						 * active connection is
853 						 * established
854 						 */
855 	BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB,	/* Calls BPF program when a
856 						 * passive connection is
857 						 * established
858 						 */
859 	BPF_SOCK_OPS_NEEDS_ECN,		/* If connection's congestion control
860 					 * needs ECN
861 					 */
862 };
863 
864 #define TCP_BPF_IW		1001	/* Set TCP initial congestion window */
865 #define TCP_BPF_SNDCWND_CLAMP	1002	/* Set sndcwnd_clamp */
866 
867 #endif /* _UAPI__LINUX_BPF_H__ */
868