1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
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.
7 #ifndef _UAPI__LINUX_BPF_H__
8 #define _UAPI__LINUX_BPF_H__
10 #include <linux/types.h>
11 #include <linux/bpf_common.h>
13 /* Extended instruction set based on top of classic BPF */
15 /* instruction classes */
16 #define BPF_ALU64 0x07 /* alu mode in double word width */
19 #define BPF_DW 0x18 /* double word */
20 #define BPF_XADD 0xc0 /* exclusive add */
23 #define BPF_MOV 0xb0 /* mov reg to reg */
24 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
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
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 */
39 /* Register numbers */
55 /* BPF has 10 general purpose 64-bit registers and stack frame. */
56 #define MAX_BPF_REG __MAX_BPF_REG
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 */
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 */
72 /* BPF syscall commands, see bpf(2) man-page for details. */
90 BPF_MAP_TYPE_PROG_ARRAY,
91 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
92 BPF_MAP_TYPE_PERCPU_HASH,
93 BPF_MAP_TYPE_PERCPU_ARRAY,
94 BPF_MAP_TYPE_STACK_TRACE,
95 BPF_MAP_TYPE_CGROUP_ARRAY,
96 BPF_MAP_TYPE_LRU_HASH,
97 BPF_MAP_TYPE_LRU_PERCPU_HASH,
98 BPF_MAP_TYPE_LPM_TRIE,
102 BPF_PROG_TYPE_UNSPEC,
103 BPF_PROG_TYPE_SOCKET_FILTER,
104 BPF_PROG_TYPE_KPROBE,
105 BPF_PROG_TYPE_SCHED_CLS,
106 BPF_PROG_TYPE_SCHED_ACT,
107 BPF_PROG_TYPE_TRACEPOINT,
109 BPF_PROG_TYPE_PERF_EVENT,
110 BPF_PROG_TYPE_CGROUP_SKB,
111 BPF_PROG_TYPE_CGROUP_SOCK,
112 BPF_PROG_TYPE_LWT_IN,
113 BPF_PROG_TYPE_LWT_OUT,
114 BPF_PROG_TYPE_LWT_XMIT,
117 enum bpf_attach_type {
118 BPF_CGROUP_INET_INGRESS,
119 BPF_CGROUP_INET_EGRESS,
120 BPF_CGROUP_INET_SOCK_CREATE,
121 __MAX_BPF_ATTACH_TYPE
124 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
126 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command
127 * to the given target_fd cgroup the descendent cgroup will be able to
128 * override effective bpf program that was inherited from this cgroup
130 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
132 #define BPF_PSEUDO_MAP_FD 1
134 /* flags for BPF_MAP_UPDATE_ELEM command */
135 #define BPF_ANY 0 /* create new element or update existing */
136 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
137 #define BPF_EXIST 2 /* update existing element */
139 #define BPF_F_NO_PREALLOC (1U << 0)
140 /* Instead of having one common LRU list in the
141 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
142 * which can scale and perform better.
143 * Note, the LRU nodes (including free nodes) cannot be moved
144 * across different LRU lists.
146 #define BPF_F_NO_COMMON_LRU (1U << 1)
149 struct { /* anonymous struct used by BPF_MAP_CREATE command */
150 __u32 map_type; /* one of enum bpf_map_type */
151 __u32 key_size; /* size of key in bytes */
152 __u32 value_size; /* size of value in bytes */
153 __u32 max_entries; /* max number of entries in a map */
154 __u32 map_flags; /* prealloc or not */
157 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
162 __aligned_u64 next_key;
167 struct { /* anonymous struct used by BPF_PROG_LOAD command */
168 __u32 prog_type; /* one of enum bpf_prog_type */
171 __aligned_u64 license;
172 __u32 log_level; /* verbosity level of verifier */
173 __u32 log_size; /* size of user buffer */
174 __aligned_u64 log_buf; /* user supplied buffer */
175 __u32 kern_version; /* checked when prog_type=kprobe */
178 struct { /* anonymous struct used by BPF_OBJ_* commands */
179 __aligned_u64 pathname;
183 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
184 __u32 target_fd; /* container object to attach to */
185 __u32 attach_bpf_fd; /* eBPF program to attach */
189 } __attribute__((aligned(8)));
191 /* BPF helper function descriptions:
193 * void *bpf_map_lookup_elem(&map, &key)
194 * Return: Map value or NULL
196 * int bpf_map_update_elem(&map, &key, &value, flags)
197 * Return: 0 on success or negative error
199 * int bpf_map_delete_elem(&map, &key)
200 * Return: 0 on success or negative error
202 * int bpf_probe_read(void *dst, int size, void *src)
203 * Return: 0 on success or negative error
205 * u64 bpf_ktime_get_ns(void)
206 * Return: current ktime
208 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
209 * Return: length of buffer written or negative error
211 * u32 bpf_prandom_u32(void)
212 * Return: random value
214 * u32 bpf_raw_smp_processor_id(void)
215 * Return: SMP processor ID
217 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
218 * store bytes into packet
219 * @skb: pointer to skb
220 * @offset: offset within packet from skb->mac_header
221 * @from: pointer where to copy bytes from
222 * @len: number of bytes to store into packet
223 * @flags: bit 0 - if true, recompute skb->csum
224 * other bits - reserved
225 * Return: 0 on success or negative error
227 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
228 * recompute IP checksum
229 * @skb: pointer to skb
230 * @offset: offset within packet where IP checksum is located
231 * @from: old value of header field
232 * @to: new value of header field
233 * @flags: bits 0-3 - size of header field
234 * other bits - reserved
235 * Return: 0 on success or negative error
237 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
238 * recompute TCP/UDP checksum
239 * @skb: pointer to skb
240 * @offset: offset within packet where TCP/UDP checksum is located
241 * @from: old value of header field
242 * @to: new value of header field
243 * @flags: bits 0-3 - size of header field
244 * bit 4 - is pseudo header
245 * other bits - reserved
246 * Return: 0 on success or negative error
248 * int bpf_tail_call(ctx, prog_array_map, index)
249 * jump into another BPF program
250 * @ctx: context pointer passed to next program
251 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
252 * @index: index inside array that selects specific program to run
253 * Return: 0 on success or negative error
255 * int bpf_clone_redirect(skb, ifindex, flags)
256 * redirect to another netdev
257 * @skb: pointer to skb
258 * @ifindex: ifindex of the net device
259 * @flags: bit 0 - if set, redirect to ingress instead of egress
260 * other bits - reserved
261 * Return: 0 on success or negative error
263 * u64 bpf_get_current_pid_tgid(void)
264 * Return: current->tgid << 32 | current->pid
266 * u64 bpf_get_current_uid_gid(void)
267 * Return: current_gid << 32 | current_uid
269 * int bpf_get_current_comm(char *buf, int size_of_buf)
270 * stores current->comm into buf
271 * Return: 0 on success or negative error
273 * u32 bpf_get_cgroup_classid(skb)
274 * retrieve a proc's classid
275 * @skb: pointer to skb
276 * Return: classid if != 0
278 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
279 * Return: 0 on success or negative error
281 * int bpf_skb_vlan_pop(skb)
282 * Return: 0 on success or negative error
284 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
285 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
286 * retrieve or populate tunnel metadata
287 * @skb: pointer to skb
288 * @key: pointer to 'struct bpf_tunnel_key'
289 * @size: size of 'struct bpf_tunnel_key'
290 * @flags: room for future extensions
291 * Return: 0 on success or negative error
293 * u64 bpf_perf_event_read(&map, index)
294 * Return: Number events read or error code
296 * int bpf_redirect(ifindex, flags)
297 * redirect to another netdev
298 * @ifindex: ifindex of the net device
299 * @flags: bit 0 - if set, redirect to ingress instead of egress
300 * other bits - reserved
301 * Return: TC_ACT_REDIRECT
303 * u32 bpf_get_route_realm(skb)
304 * retrieve a dst's tclassid
305 * @skb: pointer to skb
306 * Return: realm if != 0
308 * int bpf_perf_event_output(ctx, map, index, data, size)
309 * output perf raw sample
310 * @ctx: struct pt_regs*
311 * @map: pointer to perf_event_array map
312 * @index: index of event in the map
313 * @data: data on stack to be output as raw data
314 * @size: size of data
315 * Return: 0 on success or negative error
317 * int bpf_get_stackid(ctx, map, flags)
318 * walk user or kernel stack and return id
319 * @ctx: struct pt_regs*
320 * @map: pointer to stack_trace map
321 * @flags: bits 0-7 - numer of stack frames to skip
322 * bit 8 - collect user stack instead of kernel
323 * bit 9 - compare stacks by hash only
324 * bit 10 - if two different stacks hash into the same stackid
326 * other bits - reserved
327 * Return: >= 0 stackid on success or negative error
329 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
330 * calculate csum diff
331 * @from: raw from buffer
332 * @from_size: length of from buffer
334 * @to_size: length of to buffer
335 * @seed: optional seed
336 * Return: csum result or negative error code
338 * int bpf_skb_get_tunnel_opt(skb, opt, size)
339 * retrieve tunnel options metadata
340 * @skb: pointer to skb
341 * @opt: pointer to raw tunnel option data
342 * @size: size of @opt
343 * Return: option size
345 * int bpf_skb_set_tunnel_opt(skb, opt, size)
346 * populate tunnel options metadata
347 * @skb: pointer to skb
348 * @opt: pointer to raw tunnel option data
349 * @size: size of @opt
350 * Return: 0 on success or negative error
352 * int bpf_skb_change_proto(skb, proto, flags)
353 * Change protocol of the skb. Currently supported is v4 -> v6,
354 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
355 * program is expected to fill the new headers via skb_store_bytes
356 * and lX_csum_replace.
357 * @skb: pointer to skb
358 * @proto: new skb->protocol type
360 * Return: 0 on success or negative error
362 * int bpf_skb_change_type(skb, type)
363 * Change packet type of skb.
364 * @skb: pointer to skb
365 * @type: new skb->pkt_type type
366 * Return: 0 on success or negative error
368 * int bpf_skb_under_cgroup(skb, map, index)
369 * Check cgroup2 membership of skb
370 * @skb: pointer to skb
371 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
372 * @index: index of the cgroup in the bpf_map
374 * == 0 skb failed the cgroup2 descendant test
375 * == 1 skb succeeded the cgroup2 descendant test
378 * u32 bpf_get_hash_recalc(skb)
379 * Retrieve and possibly recalculate skb->hash.
380 * @skb: pointer to skb
383 * u64 bpf_get_current_task(void)
384 * Returns current task_struct
387 * int bpf_probe_write_user(void *dst, void *src, int len)
388 * safely attempt to write to a location
389 * @dst: destination address in userspace
390 * @src: source address on stack
391 * @len: number of bytes to copy
392 * Return: 0 on success or negative error
394 * int bpf_current_task_under_cgroup(map, index)
395 * Check cgroup2 membership of current task
396 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
397 * @index: index of the cgroup in the bpf_map
399 * == 0 current failed the cgroup2 descendant test
400 * == 1 current succeeded the cgroup2 descendant test
403 * int bpf_skb_change_tail(skb, len, flags)
404 * The helper will resize the skb to the given new size, to be used f.e.
405 * with control messages.
406 * @skb: pointer to skb
407 * @len: new skb length
409 * Return: 0 on success or negative error
411 * int bpf_skb_pull_data(skb, len)
412 * The helper will pull in non-linear data in case the skb is non-linear
413 * and not all of len are part of the linear section. Only needed for
414 * read/write with direct packet access.
415 * @skb: pointer to skb
416 * @len: len to make read/writeable
417 * Return: 0 on success or negative error
419 * s64 bpf_csum_update(skb, csum)
420 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
421 * @skb: pointer to skb
423 * Return: csum on success or negative error
425 * void bpf_set_hash_invalid(skb)
426 * Invalidate current skb->hash.
427 * @skb: pointer to skb
429 * int bpf_get_numa_node_id()
430 * Return: Id of current NUMA node.
432 * int bpf_skb_change_head()
433 * Grows headroom of skb and adjusts MAC header offset accordingly.
434 * Will extends/reallocae as required automatically.
435 * May change skb data pointer and will thus invalidate any check
436 * performed for direct packet access.
437 * @skb: pointer to skb
438 * @len: length of header to be pushed in front
439 * @flags: Flags (unused for now)
440 * Return: 0 on success or negative error
442 * int bpf_xdp_adjust_head(xdp_md, delta)
443 * Adjust the xdp_md.data by delta
444 * @xdp_md: pointer to xdp_md
445 * @delta: An positive/negative integer to be added to xdp_md.data
446 * Return: 0 on success or negative on error
448 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
449 * Copy a NUL terminated string from unsafe address. In case the string
450 * length is smaller than size, the target is not padded with further NUL
451 * bytes. In case the string length is larger than size, just count-1
452 * bytes are copied and the last byte is set to NUL.
453 * @dst: destination address
454 * @size: maximum number of bytes to copy, including the trailing NUL
455 * @unsafe_ptr: unsafe address
457 * > 0 length of the string including the trailing NUL on success
460 #define __BPF_FUNC_MAPPER(FN) \
462 FN(map_lookup_elem), \
463 FN(map_update_elem), \
464 FN(map_delete_elem), \
468 FN(get_prandom_u32), \
469 FN(get_smp_processor_id), \
470 FN(skb_store_bytes), \
471 FN(l3_csum_replace), \
472 FN(l4_csum_replace), \
474 FN(clone_redirect), \
475 FN(get_current_pid_tgid), \
476 FN(get_current_uid_gid), \
477 FN(get_current_comm), \
478 FN(get_cgroup_classid), \
481 FN(skb_get_tunnel_key), \
482 FN(skb_set_tunnel_key), \
483 FN(perf_event_read), \
485 FN(get_route_realm), \
486 FN(perf_event_output), \
487 FN(skb_load_bytes), \
490 FN(skb_get_tunnel_opt), \
491 FN(skb_set_tunnel_opt), \
492 FN(skb_change_proto), \
493 FN(skb_change_type), \
494 FN(skb_under_cgroup), \
495 FN(get_hash_recalc), \
496 FN(get_current_task), \
497 FN(probe_write_user), \
498 FN(current_task_under_cgroup), \
499 FN(skb_change_tail), \
502 FN(set_hash_invalid), \
503 FN(get_numa_node_id), \
504 FN(skb_change_head), \
505 FN(xdp_adjust_head), \
508 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
509 * function eBPF program intends to call
511 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
513 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
518 /* All flags used by eBPF helper functions, placed here. */
520 /* BPF_FUNC_skb_store_bytes flags. */
521 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
522 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
524 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
525 * First 4 bits are for passing the header field size.
527 #define BPF_F_HDR_FIELD_MASK 0xfULL
529 /* BPF_FUNC_l4_csum_replace flags. */
530 #define BPF_F_PSEUDO_HDR (1ULL << 4)
531 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
532 #define BPF_F_MARK_ENFORCE (1ULL << 6)
534 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
535 #define BPF_F_INGRESS (1ULL << 0)
537 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
538 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
540 /* BPF_FUNC_get_stackid flags. */
541 #define BPF_F_SKIP_FIELD_MASK 0xffULL
542 #define BPF_F_USER_STACK (1ULL << 8)
543 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
544 #define BPF_F_REUSE_STACKID (1ULL << 10)
546 /* BPF_FUNC_skb_set_tunnel_key flags. */
547 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
548 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
550 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */
551 #define BPF_F_INDEX_MASK 0xffffffffULL
552 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
553 /* BPF_FUNC_perf_event_output for sk_buff input context. */
554 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
556 /* user accessible mirror of in-kernel sk_buff.
557 * new fields can only be added to the end of this structure
569 __u32 ingress_ifindex;
579 struct bpf_tunnel_key {
583 __u32 remote_ipv6[4];
591 /* Generic BPF return codes which all BPF program types may support.
592 * The values are binary compatible with their TC_ACT_* counter-part to
593 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
596 * XDP is handled seprately, see XDP_*.
604 /* >127 are reserved for prog type specific return codes */
614 #define XDP_PACKET_HEADROOM 256
616 /* User return codes for XDP prog type.
617 * A valid XDP program must return one of these defined values. All other
618 * return codes are reserved for future use. Unknown return codes will result
628 /* user accessible metadata for XDP packet hook
629 * new fields must be added to the end of this structure
636 #endif /* _UAPI__LINUX_BPF_H__ */