1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 #ifndef _UAPI__LINUX_BPF_H__
9 #define _UAPI__LINUX_BPF_H__
11 #include <linux/types.h>
12 #include <linux/bpf_common.h>
14 /* Extended instruction set based on top of classic BPF */
16 /* instruction classes */
17 #define BPF_ALU64 0x07 /* alu mode in double word width */
20 #define BPF_DW 0x18 /* double word (64-bit) */
21 #define BPF_XADD 0xc0 /* exclusive add */
24 #define BPF_MOV 0xb0 /* mov reg to reg */
25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
27 /* change endianness of a register */
28 #define BPF_END 0xd0 /* flags for endianness conversion: */
29 #define BPF_TO_LE 0x00 /* convert to little-endian */
30 #define BPF_TO_BE 0x08 /* convert to big-endian */
31 #define BPF_FROM_LE BPF_TO_LE
32 #define BPF_FROM_BE BPF_TO_BE
35 #define BPF_JNE 0x50 /* jump != */
36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
42 #define BPF_CALL 0x80 /* function call */
43 #define BPF_EXIT 0x90 /* function return */
45 /* Register numbers */
61 /* BPF has 10 general purpose 64-bit registers and stack frame. */
62 #define MAX_BPF_REG __MAX_BPF_REG
65 __u8 code; /* opcode */
66 __u8 dst_reg:4; /* dest register */
67 __u8 src_reg:4; /* source register */
68 __s16 off; /* signed offset */
69 __s32 imm; /* signed immediate constant */
72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
73 struct bpf_lpm_trie_key {
74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
75 __u8 data[0]; /* Arbitrary size */
78 /* BPF syscall commands, see bpf(2) man-page for details. */
93 BPF_PROG_GET_FD_BY_ID,
95 BPF_OBJ_GET_INFO_BY_FD,
97 BPF_RAW_TRACEPOINT_OPEN,
104 BPF_MAP_TYPE_PROG_ARRAY,
105 BPF_MAP_TYPE_PERF_EVENT_ARRAY,
106 BPF_MAP_TYPE_PERCPU_HASH,
107 BPF_MAP_TYPE_PERCPU_ARRAY,
108 BPF_MAP_TYPE_STACK_TRACE,
109 BPF_MAP_TYPE_CGROUP_ARRAY,
110 BPF_MAP_TYPE_LRU_HASH,
111 BPF_MAP_TYPE_LRU_PERCPU_HASH,
112 BPF_MAP_TYPE_LPM_TRIE,
113 BPF_MAP_TYPE_ARRAY_OF_MAPS,
114 BPF_MAP_TYPE_HASH_OF_MAPS,
116 BPF_MAP_TYPE_SOCKMAP,
121 BPF_PROG_TYPE_UNSPEC,
122 BPF_PROG_TYPE_SOCKET_FILTER,
123 BPF_PROG_TYPE_KPROBE,
124 BPF_PROG_TYPE_SCHED_CLS,
125 BPF_PROG_TYPE_SCHED_ACT,
126 BPF_PROG_TYPE_TRACEPOINT,
128 BPF_PROG_TYPE_PERF_EVENT,
129 BPF_PROG_TYPE_CGROUP_SKB,
130 BPF_PROG_TYPE_CGROUP_SOCK,
131 BPF_PROG_TYPE_LWT_IN,
132 BPF_PROG_TYPE_LWT_OUT,
133 BPF_PROG_TYPE_LWT_XMIT,
134 BPF_PROG_TYPE_SOCK_OPS,
135 BPF_PROG_TYPE_SK_SKB,
136 BPF_PROG_TYPE_CGROUP_DEVICE,
137 BPF_PROG_TYPE_SK_MSG,
138 BPF_PROG_TYPE_RAW_TRACEPOINT,
141 enum bpf_attach_type {
142 BPF_CGROUP_INET_INGRESS,
143 BPF_CGROUP_INET_EGRESS,
144 BPF_CGROUP_INET_SOCK_CREATE,
146 BPF_SK_SKB_STREAM_PARSER,
147 BPF_SK_SKB_STREAM_VERDICT,
150 __MAX_BPF_ATTACH_TYPE
153 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
155 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
157 * NONE(default): No further bpf programs allowed in the subtree.
159 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
160 * the program in this cgroup yields to sub-cgroup program.
162 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
163 * that cgroup program gets run in addition to the program in this cgroup.
165 * Only one program is allowed to be attached to a cgroup with
166 * NONE or BPF_F_ALLOW_OVERRIDE flag.
167 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
168 * release old program and attach the new one. Attach flags has to match.
170 * Multiple programs are allowed to be attached to a cgroup with
171 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
172 * (those that were attached first, run first)
173 * The programs of sub-cgroup are executed first, then programs of
174 * this cgroup and then programs of parent cgroup.
175 * When children program makes decision (like picking TCP CA or sock bind)
176 * parent program has a chance to override it.
178 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
179 * A cgroup with NONE doesn't allow any programs in sub-cgroups.
181 * cgrp1 (MULTI progs A, B) ->
182 * cgrp2 (OVERRIDE prog C) ->
183 * cgrp3 (MULTI prog D) ->
184 * cgrp4 (OVERRIDE prog E) ->
185 * cgrp5 (NONE prog F)
186 * the event in cgrp5 triggers execution of F,D,A,B in that order.
187 * if prog F is detached, the execution is E,D,A,B
188 * if prog F and D are detached, the execution is E,A,B
189 * if prog F, E and D are detached, the execution is C,A,B
191 * All eligible programs are executed regardless of return code from
194 #define BPF_F_ALLOW_OVERRIDE (1U << 0)
195 #define BPF_F_ALLOW_MULTI (1U << 1)
197 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
198 * verifier will perform strict alignment checking as if the kernel
199 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
200 * and NET_IP_ALIGN defined to 2.
202 #define BPF_F_STRICT_ALIGNMENT (1U << 0)
204 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
205 #define BPF_PSEUDO_MAP_FD 1
207 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
208 * offset to another bpf function
210 #define BPF_PSEUDO_CALL 1
212 /* flags for BPF_MAP_UPDATE_ELEM command */
213 #define BPF_ANY 0 /* create new element or update existing */
214 #define BPF_NOEXIST 1 /* create new element if it didn't exist */
215 #define BPF_EXIST 2 /* update existing element */
217 /* flags for BPF_MAP_CREATE command */
218 #define BPF_F_NO_PREALLOC (1U << 0)
219 /* Instead of having one common LRU list in the
220 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
221 * which can scale and perform better.
222 * Note, the LRU nodes (including free nodes) cannot be moved
223 * across different LRU lists.
225 #define BPF_F_NO_COMMON_LRU (1U << 1)
226 /* Specify numa node during map creation */
227 #define BPF_F_NUMA_NODE (1U << 2)
229 /* flags for BPF_PROG_QUERY */
230 #define BPF_F_QUERY_EFFECTIVE (1U << 0)
232 #define BPF_OBJ_NAME_LEN 16U
234 /* Flags for accessing BPF object */
235 #define BPF_F_RDONLY (1U << 3)
236 #define BPF_F_WRONLY (1U << 4)
238 /* Flag for stack_map, store build_id+offset instead of pointer */
239 #define BPF_F_STACK_BUILD_ID (1U << 5)
241 enum bpf_stack_build_id_status {
242 /* user space need an empty entry to identify end of a trace */
243 BPF_STACK_BUILD_ID_EMPTY = 0,
244 /* with valid build_id and offset */
245 BPF_STACK_BUILD_ID_VALID = 1,
246 /* couldn't get build_id, fallback to ip */
247 BPF_STACK_BUILD_ID_IP = 2,
250 #define BPF_BUILD_ID_SIZE 20
251 struct bpf_stack_build_id {
253 unsigned char build_id[BPF_BUILD_ID_SIZE];
261 struct { /* anonymous struct used by BPF_MAP_CREATE command */
262 __u32 map_type; /* one of enum bpf_map_type */
263 __u32 key_size; /* size of key in bytes */
264 __u32 value_size; /* size of value in bytes */
265 __u32 max_entries; /* max number of entries in a map */
266 __u32 map_flags; /* BPF_MAP_CREATE related
267 * flags defined above.
269 __u32 inner_map_fd; /* fd pointing to the inner map */
270 __u32 numa_node; /* numa node (effective only if
271 * BPF_F_NUMA_NODE is set).
273 char map_name[BPF_OBJ_NAME_LEN];
274 __u32 map_ifindex; /* ifindex of netdev to create on */
277 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
282 __aligned_u64 next_key;
287 struct { /* anonymous struct used by BPF_PROG_LOAD command */
288 __u32 prog_type; /* one of enum bpf_prog_type */
291 __aligned_u64 license;
292 __u32 log_level; /* verbosity level of verifier */
293 __u32 log_size; /* size of user buffer */
294 __aligned_u64 log_buf; /* user supplied buffer */
295 __u32 kern_version; /* checked when prog_type=kprobe */
297 char prog_name[BPF_OBJ_NAME_LEN];
298 __u32 prog_ifindex; /* ifindex of netdev to prep for */
299 /* For some prog types expected attach type must be known at
300 * load time to verify attach type specific parts of prog
301 * (context accesses, allowed helpers, etc).
303 __u32 expected_attach_type;
306 struct { /* anonymous struct used by BPF_OBJ_* commands */
307 __aligned_u64 pathname;
312 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
313 __u32 target_fd; /* container object to attach to */
314 __u32 attach_bpf_fd; /* eBPF program to attach */
319 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
324 __aligned_u64 data_in;
325 __aligned_u64 data_out;
330 struct { /* anonymous struct used by BPF_*_GET_*_ID */
340 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
346 struct { /* anonymous struct used by BPF_PROG_QUERY command */
347 __u32 target_fd; /* container object to query */
351 __aligned_u64 prog_ids;
359 } __attribute__((aligned(8)));
361 /* BPF helper function descriptions:
363 * void *bpf_map_lookup_elem(&map, &key)
364 * Return: Map value or NULL
366 * int bpf_map_update_elem(&map, &key, &value, flags)
367 * Return: 0 on success or negative error
369 * int bpf_map_delete_elem(&map, &key)
370 * Return: 0 on success or negative error
372 * int bpf_probe_read(void *dst, int size, void *src)
373 * Return: 0 on success or negative error
375 * u64 bpf_ktime_get_ns(void)
376 * Return: current ktime
378 * int bpf_trace_printk(const char *fmt, int fmt_size, ...)
379 * Return: length of buffer written or negative error
381 * u32 bpf_prandom_u32(void)
382 * Return: random value
384 * u32 bpf_raw_smp_processor_id(void)
385 * Return: SMP processor ID
387 * int bpf_skb_store_bytes(skb, offset, from, len, flags)
388 * store bytes into packet
389 * @skb: pointer to skb
390 * @offset: offset within packet from skb->mac_header
391 * @from: pointer where to copy bytes from
392 * @len: number of bytes to store into packet
393 * @flags: bit 0 - if true, recompute skb->csum
394 * other bits - reserved
395 * Return: 0 on success or negative error
397 * int bpf_l3_csum_replace(skb, offset, from, to, flags)
398 * recompute IP checksum
399 * @skb: pointer to skb
400 * @offset: offset within packet where IP checksum is located
401 * @from: old value of header field
402 * @to: new value of header field
403 * @flags: bits 0-3 - size of header field
404 * other bits - reserved
405 * Return: 0 on success or negative error
407 * int bpf_l4_csum_replace(skb, offset, from, to, flags)
408 * recompute TCP/UDP checksum
409 * @skb: pointer to skb
410 * @offset: offset within packet where TCP/UDP checksum is located
411 * @from: old value of header field
412 * @to: new value of header field
413 * @flags: bits 0-3 - size of header field
414 * bit 4 - is pseudo header
415 * other bits - reserved
416 * Return: 0 on success or negative error
418 * int bpf_tail_call(ctx, prog_array_map, index)
419 * jump into another BPF program
420 * @ctx: context pointer passed to next program
421 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
422 * @index: 32-bit index inside array that selects specific program to run
423 * Return: 0 on success or negative error
425 * int bpf_clone_redirect(skb, ifindex, flags)
426 * redirect to another netdev
427 * @skb: pointer to skb
428 * @ifindex: ifindex of the net device
429 * @flags: bit 0 - if set, redirect to ingress instead of egress
430 * other bits - reserved
431 * Return: 0 on success or negative error
433 * u64 bpf_get_current_pid_tgid(void)
434 * Return: current->tgid << 32 | current->pid
436 * u64 bpf_get_current_uid_gid(void)
437 * Return: current_gid << 32 | current_uid
439 * int bpf_get_current_comm(char *buf, int size_of_buf)
440 * stores current->comm into buf
441 * Return: 0 on success or negative error
443 * u32 bpf_get_cgroup_classid(skb)
444 * retrieve a proc's classid
445 * @skb: pointer to skb
446 * Return: classid if != 0
448 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci)
449 * Return: 0 on success or negative error
451 * int bpf_skb_vlan_pop(skb)
452 * Return: 0 on success or negative error
454 * int bpf_skb_get_tunnel_key(skb, key, size, flags)
455 * int bpf_skb_set_tunnel_key(skb, key, size, flags)
456 * retrieve or populate tunnel metadata
457 * @skb: pointer to skb
458 * @key: pointer to 'struct bpf_tunnel_key'
459 * @size: size of 'struct bpf_tunnel_key'
460 * @flags: room for future extensions
461 * Return: 0 on success or negative error
463 * u64 bpf_perf_event_read(map, flags)
464 * read perf event counter value
465 * @map: pointer to perf_event_array map
466 * @flags: index of event in the map or bitmask flags
467 * Return: value of perf event counter read or error code
469 * int bpf_redirect(ifindex, flags)
470 * redirect to another netdev
471 * @ifindex: ifindex of the net device
474 * bit 0 - if set, redirect to ingress instead of egress
475 * other bits - reserved
477 * all bits - reserved
478 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error
479 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error
480 * int bpf_redirect_map(map, key, flags)
481 * redirect to endpoint in map
482 * @map: pointer to dev map
483 * @key: index in map to lookup
485 * Return: XDP_REDIRECT on success or XDP_ABORT on error
487 * u32 bpf_get_route_realm(skb)
488 * retrieve a dst's tclassid
489 * @skb: pointer to skb
490 * Return: realm if != 0
492 * int bpf_perf_event_output(ctx, map, flags, data, size)
493 * output perf raw sample
494 * @ctx: struct pt_regs*
495 * @map: pointer to perf_event_array map
496 * @flags: index of event in the map or bitmask flags
497 * @data: data on stack to be output as raw data
498 * @size: size of data
499 * Return: 0 on success or negative error
501 * int bpf_get_stackid(ctx, map, flags)
502 * walk user or kernel stack and return id
503 * @ctx: struct pt_regs*
504 * @map: pointer to stack_trace map
505 * @flags: bits 0-7 - numer of stack frames to skip
506 * bit 8 - collect user stack instead of kernel
507 * bit 9 - compare stacks by hash only
508 * bit 10 - if two different stacks hash into the same stackid
510 * other bits - reserved
511 * Return: >= 0 stackid on success or negative error
513 * s64 bpf_csum_diff(from, from_size, to, to_size, seed)
514 * calculate csum diff
515 * @from: raw from buffer
516 * @from_size: length of from buffer
518 * @to_size: length of to buffer
519 * @seed: optional seed
520 * Return: csum result or negative error code
522 * int bpf_skb_get_tunnel_opt(skb, opt, size)
523 * retrieve tunnel options metadata
524 * @skb: pointer to skb
525 * @opt: pointer to raw tunnel option data
526 * @size: size of @opt
527 * Return: option size
529 * int bpf_skb_set_tunnel_opt(skb, opt, size)
530 * populate tunnel options metadata
531 * @skb: pointer to skb
532 * @opt: pointer to raw tunnel option data
533 * @size: size of @opt
534 * Return: 0 on success or negative error
536 * int bpf_skb_change_proto(skb, proto, flags)
537 * Change protocol of the skb. Currently supported is v4 -> v6,
538 * v6 -> v4 transitions. The helper will also resize the skb. eBPF
539 * program is expected to fill the new headers via skb_store_bytes
540 * and lX_csum_replace.
541 * @skb: pointer to skb
542 * @proto: new skb->protocol type
544 * Return: 0 on success or negative error
546 * int bpf_skb_change_type(skb, type)
547 * Change packet type of skb.
548 * @skb: pointer to skb
549 * @type: new skb->pkt_type type
550 * Return: 0 on success or negative error
552 * int bpf_skb_under_cgroup(skb, map, index)
553 * Check cgroup2 membership of skb
554 * @skb: pointer to skb
555 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
556 * @index: index of the cgroup in the bpf_map
558 * == 0 skb failed the cgroup2 descendant test
559 * == 1 skb succeeded the cgroup2 descendant test
562 * u32 bpf_get_hash_recalc(skb)
563 * Retrieve and possibly recalculate skb->hash.
564 * @skb: pointer to skb
567 * u64 bpf_get_current_task(void)
568 * Returns current task_struct
571 * int bpf_probe_write_user(void *dst, void *src, int len)
572 * safely attempt to write to a location
573 * @dst: destination address in userspace
574 * @src: source address on stack
575 * @len: number of bytes to copy
576 * Return: 0 on success or negative error
578 * int bpf_current_task_under_cgroup(map, index)
579 * Check cgroup2 membership of current task
580 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type
581 * @index: index of the cgroup in the bpf_map
583 * == 0 current failed the cgroup2 descendant test
584 * == 1 current succeeded the cgroup2 descendant test
587 * int bpf_skb_change_tail(skb, len, flags)
588 * The helper will resize the skb to the given new size, to be used f.e.
589 * with control messages.
590 * @skb: pointer to skb
591 * @len: new skb length
593 * Return: 0 on success or negative error
595 * int bpf_skb_pull_data(skb, len)
596 * The helper will pull in non-linear data in case the skb is non-linear
597 * and not all of len are part of the linear section. Only needed for
598 * read/write with direct packet access.
599 * @skb: pointer to skb
600 * @len: len to make read/writeable
601 * Return: 0 on success or negative error
603 * s64 bpf_csum_update(skb, csum)
604 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE.
605 * @skb: pointer to skb
607 * Return: csum on success or negative error
609 * void bpf_set_hash_invalid(skb)
610 * Invalidate current skb->hash.
611 * @skb: pointer to skb
613 * int bpf_get_numa_node_id()
614 * Return: Id of current NUMA node.
616 * int bpf_skb_change_head()
617 * Grows headroom of skb and adjusts MAC header offset accordingly.
618 * Will extends/reallocae as required automatically.
619 * May change skb data pointer and will thus invalidate any check
620 * performed for direct packet access.
621 * @skb: pointer to skb
622 * @len: length of header to be pushed in front
623 * @flags: Flags (unused for now)
624 * Return: 0 on success or negative error
626 * int bpf_xdp_adjust_head(xdp_md, delta)
627 * Adjust the xdp_md.data by delta
628 * @xdp_md: pointer to xdp_md
629 * @delta: An positive/negative integer to be added to xdp_md.data
630 * Return: 0 on success or negative on error
632 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
633 * Copy a NUL terminated string from unsafe address. In case the string
634 * length is smaller than size, the target is not padded with further NUL
635 * bytes. In case the string length is larger than size, just count-1
636 * bytes are copied and the last byte is set to NUL.
637 * @dst: destination address
638 * @size: maximum number of bytes to copy, including the trailing NUL
639 * @unsafe_ptr: unsafe address
641 * > 0 length of the string including the trailing NUL on success
644 * u64 bpf_get_socket_cookie(skb)
645 * Get the cookie for the socket stored inside sk_buff.
646 * @skb: pointer to skb
647 * Return: 8 Bytes non-decreasing number on success or 0 if the socket
648 * field is missing inside sk_buff
650 * u32 bpf_get_socket_uid(skb)
651 * Get the owner uid of the socket stored inside sk_buff.
652 * @skb: pointer to skb
653 * Return: uid of the socket owner on success or overflowuid if failed.
655 * u32 bpf_set_hash(skb, hash)
656 * Set full skb->hash.
657 * @skb: pointer to skb
660 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen)
661 * Calls setsockopt. Not all opts are available, only those with
662 * integer optvals plus TCP_CONGESTION.
663 * Supported levels: SOL_SOCKET and IPPROTO_TCP
664 * @bpf_socket: pointer to bpf_socket
665 * @level: SOL_SOCKET or IPPROTO_TCP
666 * @optname: option name
667 * @optval: pointer to option value
668 * @optlen: length of optval in bytes
669 * Return: 0 or negative error
671 * int bpf_getsockopt(bpf_socket, level, optname, optval, optlen)
672 * Calls getsockopt. Not all opts are available.
673 * Supported levels: IPPROTO_TCP
674 * @bpf_socket: pointer to bpf_socket
675 * @level: IPPROTO_TCP
676 * @optname: option name
677 * @optval: pointer to option value
678 * @optlen: length of optval in bytes
679 * Return: 0 or negative error
681 * int bpf_sock_ops_cb_flags_set(bpf_sock_ops, flags)
682 * Set callback flags for sock_ops
683 * @bpf_sock_ops: pointer to bpf_sock_ops_kern struct
684 * @flags: flags value
685 * Return: 0 for no error
686 * -EINVAL if there is no full tcp socket
687 * bits in flags that are not supported by current kernel
689 * int bpf_skb_adjust_room(skb, len_diff, mode, flags)
690 * Grow or shrink room in sk_buff.
691 * @skb: pointer to skb
692 * @len_diff: (signed) amount of room to grow/shrink
693 * @mode: operation mode (enum bpf_adj_room_mode)
694 * @flags: reserved for future use
695 * Return: 0 on success or negative error code
697 * int bpf_sk_redirect_map(map, key, flags)
698 * Redirect skb to a sock in map using key as a lookup key for the
700 * @map: pointer to sockmap
701 * @key: key to lookup sock in map
702 * @flags: reserved for future use
705 * int bpf_sock_map_update(skops, map, key, flags)
706 * @skops: pointer to bpf_sock_ops
707 * @map: pointer to sockmap to update
708 * @key: key to insert/update sock in map
709 * @flags: same flags as map update elem
711 * int bpf_xdp_adjust_meta(xdp_md, delta)
712 * Adjust the xdp_md.data_meta by delta
713 * @xdp_md: pointer to xdp_md
714 * @delta: An positive/negative integer to be added to xdp_md.data_meta
715 * Return: 0 on success or negative on error
717 * int bpf_perf_event_read_value(map, flags, buf, buf_size)
718 * read perf event counter value and perf event enabled/running time
719 * @map: pointer to perf_event_array map
720 * @flags: index of event in the map or bitmask flags
722 * @buf_size: size of the buf
723 * Return: 0 on success or negative error code
725 * int bpf_perf_prog_read_value(ctx, buf, buf_size)
726 * read perf prog attached perf event counter and enabled/running time
727 * @ctx: pointer to ctx
729 * @buf_size: size of the buf
730 * Return : 0 on success or negative error code
732 * int bpf_override_return(pt_regs, rc)
733 * @pt_regs: pointer to struct pt_regs
734 * @rc: the return value to set
736 * int bpf_msg_redirect_map(map, key, flags)
737 * Redirect msg to a sock in map using key as a lookup key for the
739 * @map: pointer to sockmap
740 * @key: key to lookup sock in map
741 * @flags: reserved for future use
745 #define __BPF_FUNC_MAPPER(FN) \
747 FN(map_lookup_elem), \
748 FN(map_update_elem), \
749 FN(map_delete_elem), \
753 FN(get_prandom_u32), \
754 FN(get_smp_processor_id), \
755 FN(skb_store_bytes), \
756 FN(l3_csum_replace), \
757 FN(l4_csum_replace), \
759 FN(clone_redirect), \
760 FN(get_current_pid_tgid), \
761 FN(get_current_uid_gid), \
762 FN(get_current_comm), \
763 FN(get_cgroup_classid), \
766 FN(skb_get_tunnel_key), \
767 FN(skb_set_tunnel_key), \
768 FN(perf_event_read), \
770 FN(get_route_realm), \
771 FN(perf_event_output), \
772 FN(skb_load_bytes), \
775 FN(skb_get_tunnel_opt), \
776 FN(skb_set_tunnel_opt), \
777 FN(skb_change_proto), \
778 FN(skb_change_type), \
779 FN(skb_under_cgroup), \
780 FN(get_hash_recalc), \
781 FN(get_current_task), \
782 FN(probe_write_user), \
783 FN(current_task_under_cgroup), \
784 FN(skb_change_tail), \
787 FN(set_hash_invalid), \
788 FN(get_numa_node_id), \
789 FN(skb_change_head), \
790 FN(xdp_adjust_head), \
791 FN(probe_read_str), \
792 FN(get_socket_cookie), \
793 FN(get_socket_uid), \
796 FN(skb_adjust_room), \
798 FN(sk_redirect_map), \
799 FN(sock_map_update), \
800 FN(xdp_adjust_meta), \
801 FN(perf_event_read_value), \
802 FN(perf_prog_read_value), \
804 FN(override_return), \
805 FN(sock_ops_cb_flags_set), \
806 FN(msg_redirect_map), \
807 FN(msg_apply_bytes), \
808 FN(msg_cork_bytes), \
811 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
812 * function eBPF program intends to call
814 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
816 __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
821 /* All flags used by eBPF helper functions, placed here. */
823 /* BPF_FUNC_skb_store_bytes flags. */
824 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
825 #define BPF_F_INVALIDATE_HASH (1ULL << 1)
827 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
828 * First 4 bits are for passing the header field size.
830 #define BPF_F_HDR_FIELD_MASK 0xfULL
832 /* BPF_FUNC_l4_csum_replace flags. */
833 #define BPF_F_PSEUDO_HDR (1ULL << 4)
834 #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
835 #define BPF_F_MARK_ENFORCE (1ULL << 6)
837 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
838 #define BPF_F_INGRESS (1ULL << 0)
840 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
841 #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
843 /* BPF_FUNC_get_stackid flags. */
844 #define BPF_F_SKIP_FIELD_MASK 0xffULL
845 #define BPF_F_USER_STACK (1ULL << 8)
846 #define BPF_F_FAST_STACK_CMP (1ULL << 9)
847 #define BPF_F_REUSE_STACKID (1ULL << 10)
849 /* BPF_FUNC_skb_set_tunnel_key flags. */
850 #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
851 #define BPF_F_DONT_FRAGMENT (1ULL << 2)
852 #define BPF_F_SEQ_NUMBER (1ULL << 3)
854 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
855 * BPF_FUNC_perf_event_read_value flags.
857 #define BPF_F_INDEX_MASK 0xffffffffULL
858 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
859 /* BPF_FUNC_perf_event_output for sk_buff input context. */
860 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
862 /* Mode for BPF_FUNC_skb_adjust_room helper. */
863 enum bpf_adj_room_mode {
867 /* user accessible mirror of in-kernel sk_buff.
868 * new fields can only be added to the end of this structure
880 __u32 ingress_ifindex;
890 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
892 __u32 remote_ip4; /* Stored in network byte order */
893 __u32 local_ip4; /* Stored in network byte order */
894 __u32 remote_ip6[4]; /* Stored in network byte order */
895 __u32 local_ip6[4]; /* Stored in network byte order */
896 __u32 remote_port; /* Stored in network byte order */
897 __u32 local_port; /* stored in host byte order */
903 struct bpf_tunnel_key {
907 __u32 remote_ipv6[4];
915 /* Generic BPF return codes which all BPF program types may support.
916 * The values are binary compatible with their TC_ACT_* counter-part to
917 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
920 * XDP is handled seprately, see XDP_*.
928 /* >127 are reserved for prog type specific return codes */
940 #define XDP_PACKET_HEADROOM 256
942 /* User return codes for XDP prog type.
943 * A valid XDP program must return one of these defined values. All other
944 * return codes are reserved for future use. Unknown return codes will
945 * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
955 /* user accessible metadata for XDP packet hook
956 * new fields must be added to the end of this structure
962 /* Below access go through struct xdp_rxq_info */
963 __u32 ingress_ifindex; /* rxq->dev->ifindex */
964 __u32 rx_queue_index; /* rxq->queue_index */
972 /* user accessible metadata for SK_MSG packet hook, new fields must
973 * be added to the end of this structure
980 #define BPF_TAG_SIZE 8
982 struct bpf_prog_info {
985 __u8 tag[BPF_TAG_SIZE];
986 __u32 jited_prog_len;
987 __u32 xlated_prog_len;
988 __aligned_u64 jited_prog_insns;
989 __aligned_u64 xlated_prog_insns;
990 __u64 load_time; /* ns since boottime */
991 __u32 created_by_uid;
993 __aligned_u64 map_ids;
994 char name[BPF_OBJ_NAME_LEN];
998 } __attribute__((aligned(8)));
1000 struct bpf_map_info {
1007 char name[BPF_OBJ_NAME_LEN];
1011 } __attribute__((aligned(8)));
1013 /* User bpf_sock_ops struct to access socket values and specify request ops
1014 * and their replies.
1015 * Some of this fields are in network (bigendian) byte order and may need
1016 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
1017 * New fields can only be added at the end of this structure
1019 struct bpf_sock_ops {
1022 __u32 args[4]; /* Optionally passed to bpf program */
1023 __u32 reply; /* Returned by bpf program */
1024 __u32 replylong[4]; /* Optionally returned by bpf prog */
1027 __u32 remote_ip4; /* Stored in network byte order */
1028 __u32 local_ip4; /* Stored in network byte order */
1029 __u32 remote_ip6[4]; /* Stored in network byte order */
1030 __u32 local_ip6[4]; /* Stored in network byte order */
1031 __u32 remote_port; /* Stored in network byte order */
1032 __u32 local_port; /* stored in host byte order */
1033 __u32 is_fullsock; /* Some TCP fields are only valid if
1034 * there is a full socket. If not, the
1035 * fields read as zero.
1038 __u32 srtt_us; /* Averaged RTT << 3 in usecs */
1039 __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
1048 __u32 rate_delivered;
1049 __u32 rate_interval_us;
1052 __u32 total_retrans;
1056 __u32 data_segs_out;
1060 __u64 bytes_received;
1064 /* Definitions for bpf_sock_ops_cb_flags */
1065 #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
1066 #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
1067 #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
1068 #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
1069 * supported cb flags
1072 /* List of known BPF sock_ops operators.
1073 * New entries can only be added at the end
1077 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
1078 * -1 if default value should be used
1080 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
1081 * window (in packets) or -1 if default
1082 * value should be used
1084 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
1085 * active connection is initialized
1087 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
1088 * active connection is
1091 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
1092 * passive connection is
1095 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
1098 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
1099 * based on the path and may be
1100 * dependent on the congestion control
1101 * algorithm. In general it indicates
1102 * a congestion threshold. RTTs above
1103 * this indicate congestion
1105 BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
1106 * Arg1: value of icsk_retransmits
1107 * Arg2: value of icsk_rto
1108 * Arg3: whether RTO has expired
1110 BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
1111 * Arg1: sequence number of 1st byte
1113 * Arg3: return value of
1114 * tcp_transmit_skb (0 => success)
1116 BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
1122 /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
1123 * changes between the TCP and BPF versions. Ideally this should never happen.
1124 * If it does, we need to add code to convert them before calling
1125 * the BPF sock_ops function.
1128 BPF_TCP_ESTABLISHED = 1,
1138 BPF_TCP_CLOSING, /* Now a valid state */
1139 BPF_TCP_NEW_SYN_RECV,
1141 BPF_TCP_MAX_STATES /* Leave at the end! */
1144 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
1145 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
1147 struct bpf_perf_event_value {
1153 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
1154 #define BPF_DEVCG_ACC_READ (1ULL << 1)
1155 #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
1157 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
1158 #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
1160 struct bpf_cgroup_dev_ctx {
1161 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
1167 struct bpf_raw_tracepoint_args {
1171 #endif /* _UAPI__LINUX_BPF_H__ */